Reduced Impact Logging in the Tropical Rain Forest of Guyana Ecological, Economic and Silvicultural Consequences P. van der Hout Tropenbos - Guyana Series 6 Reduced impact logging in the tropical rain forest of Guyana The Tropenbos-Guyana Series publishes results of research projects carried out in the framework of the Tropenbos-Guyana Programme. The Tropenbos-Guyana Programme operates within the framework of the international programme of the Tropenbos Foundation. The multidisciplinary Tropenbos-Guyana Programme contributes to conservation and wise utilisation of forest resources in Guyana by conducting strategic and applied research and upgrading Guyanese capabilities in the field of forest-related sciences. P. van der Hout Reduced impact logging in the tropical rain forest of Guyana: ecological, economic and silvicultural consequences. Tropenbos-Guyana Series 6 Tropenbos-Guyana Programme, Georgetown, Guyana. ISBN 90-393-2185-X Keywords: tropical rain forest, reduced impact logging, logging damage, stumpage cost, silvicultural treatment, Guyana. © 1999 Tropenbos-Guyana Programme, Peter van der Hout All rights reserved. No part of this publication, apart from bibliographic data and brief quotation in critical reviews, may be reproduced, re-recorded or published in any form including photocopy, microfilm, electronic or electromagnetic record, without written permission. Printed by: Ponsen & Looijen bv, Wageningen Cover: Skidder winching a bole to the skid trail. Felling of a Peltogyne venosa (Purpleheart). Directional felling; driving a wedge in a Chlorocardium rodiei (Greenheart). Photos taken in Pibiri, Guyana by Peter van der Hout Reduced impact logging in the tropical rain forest of Guyana Ecological, economic and silvicultural consequences Schadebeperkende houtoogst in het tropisch regenwoud van Guyana Ecologische, economische en bosteeltkundige gevolgen (met een samenvatting in het Nederlands) Proefschrift ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de Rector Magnificus, Prof. Dr. H.O. Voorma, ingevolge het besluit van het College voor Promoties in het openbaar te verdedigen op donderdag 16 september 1999 des middags te 12.45 uur door Pieter van der Hout geboren op 30 maart 1957 te Rotterdam Promotoren: Prof. Dr. M.J.A. Werger Verbonden aan de Faculteit Biologie van de Universiteit Utrecht Prof. Dr. R.A.A. Oldeman Verbonden aan de Landbouwuniversiteit Wageningen The investigations reported in this thesis were carried out at the Mabura Hill fieldwork site of the TropenbosGuyana Programme, 12E Garnett Street, Campbellville, Georgetown, Guyana and at the Department of Plant Ecology, Utrecht University, PO Box 800.48, 3508 TB Utrecht, the Netherlands. Funding was provided by the Tropenbos Foundation, PO Box 232, 6700 AE Wageningen, the Netherlands, and Demerara Timbers Limited, Mabura Hill, Upper Demerara, Guyana. Ter nagedachtenis aan mijn vader In memory of my father For Forest Forest could keep secrets Forest could keep secrets Forest tune in everyday to watersound and birdsound Forest letting her hair down to the teeming creeping of her forest-ground But Forest don’t broadcast her business no Forest cover her business down from sky and fast-eye sun and when night come and darkness wrap her like a gown Forest is a bad dream woman Forest dreaming about mountain and when earth was young Forest dreaming of caress of gold Forest rootsing with mysterious eldorado and when howler monkey wake her up with howl Forest just stretch and stir to a new day of sound but coming back to secrets Forest could keep secrets Forest could keep secrets And we must keep Forest Grace Nichols PREFACE My first encounter with the tropical rain forest was in Suriname in 1981 during my M.Sc. studies at the Wageningen Agricultural University. Wyb Jonkers drove us along a neverending gravel road to the CELOS research plots at Kabo. Wyb showed me around in the plots and warned me of the greatest threat in these forests: the ‘patata losso’, better known among my colleagues as the ‘bête rouge’. To my amazement, there were fewer big trees than I had imagined and even those weren’t all that big. Still, the enormous complexity of this ‘green hell’ with so many different tree species was overwhelming. How could the functioning of such a forest be understood, not to mention managed? Reitze de Graaf and John Hendrison showed me some of the ropes. Their enthusiastic and knowledgeable guidance instilled an interest in me that finally resulted in this thesis. To these three people I wish to express my first and special gratitude. The Suriname experience had a lot more to it than just the forest. The people, the culture, the tropical heat gave a sensation that I’ll never forget. This magnetic force drew me back to the tropics, but this time not SouthAmerica but Africa. In Côte d’Ivoire, I had another look at natural rain forest management. In 1989, the tropical wave brought me ashore to Guyana to be precise to the Guyana Forestry Commission. There I was first introduced to ‘5 year old’ at the Lighthouse, ate labba and drank creek water (Guyanese would know what I mean). My comrades at the Forestry Commission always seemed to know how to get things done, and taught me a lot about Greenheart, logging and sawmilling. My special thanks goes to Howard Boyan, Sam Roberts, Royston Peters, David Ramchand, Ulrich Jacobs, Das, Campbell, Waterman, Archie, the girls from the typing pool, my Commissioners Udit, Cort and Black and all the other staff. They made my stay an enjoyable and memorable one. During this time, I came in contact with Hans ter Steege and Ben ter Welle, who had just initiated the TropenbosGuyana Programme. They thought that having a lumberjack in the programme was a good idea. Together we succeeded in obtaining the funds on the basis of which the Utrecht University employed me. So, Mabura Hill, here I come. Mabura Hill (sometimes Hell), the place of different tunes from different directions played at the same, loud volume, barking dogs, the four o’clock horn, driving maniacs and long lines in the ration store. But, in the non-pay-weekends it could be heaven and there were Hector’s, George’s and ‘Boots-Mouth’s’ disco to take some of the stress away. This study wouldn’t have been possible without the support of Demerara Timbers Ltd. They gave their kind co-operation by providing housing, carving out a large chunk from their concession for the experiment, paying my field-staff and providing machinery and operators. Not only for their professional support but also for entertainment, special thanks goes to Ivan Harris, Leroy Welcome, Wayne Marshall, Winston (‘Churchill’) Harper, Albert Alcie, Grace Peters, Karsten Borch (for making our quarters more comfortable and for dinner parties), Hans-Jörgen Sejer-Hansen, Finn Knudsen and Barney Quicke. Many people contributed significantly to the collection of data used in this thesis. The Pibiri crew was engaged in strenuous and arduous work. Measuring thousands of trees, sometimes twenty feet above the ground, wedging hundreds of trees, line cutting, painting, and lots and lots of walking. The crew varied in composition during the four years of whom I want to give special thanks to: Dane (‘scatterfoot’) Roberts, Julian (‘rasta’) Williams, Wrenny (‘short man’) Nathan, Ryan (‘conscious lyrics’) Roberts, Julian (‘king’) Pilay, Lindsey (‘broths’) Nathan, Montgomery (‘Monty’) Williams, Owen (‘inchie boy’) Spencer, ‘uncle’ Chris Williams, Eloma (‘fat girl’) Pollard and Norland (‘bounty killer’) Bovell. They stayed despite my mood swings, trees falling through the roof (twice), being banned from making wine from aubergine peels, spoilt rice, etc. and jumbies knocking on treespurs in the night. Thanks also to my students Bram Laumans, Dénise Bus, Bettine Robers, Mathilde Elsinga, Liesbeth Vanmechelen, Chris Brils, Edwin Laan and Sandra Williams. For applying their special skills to this experiment I would like to thank the three wise men: ‘uncle’ Sam Roberts, Aaron ‘Texas’ Braithewaite and Ivan ‘uncle – blue psychedelic shirt – Johno’ Johnson, who could recognise more trees than I could count. Furthermore, the fellers Wilfred ‘Chavez’ Jarvis and Hugh ‘rugged’ Abrams. Finally, for assisting with the data entry, Mark (‘dog’) Stephens. For their support and co-operation, I’d like to thank all of my colleagues in the Tropenbos-Guyana Programme. For looking after matters during periods of absence, I wish to specifically thank Simmoné Rose, Hans ter Steege and David Hammond. In Georgetown, there was always a welcome smile from Sybil Stewart, Joan Watson, ‘Harry’ Haripersaud, Mina Rambali or Celina Haywood, while Kissoon always managed to fix the ‘White Horse’ and other things. The TropenbosGuyana Programme rested on the strong shoulders of Dr. George Walcott, first Ben ter Welle and later Thijs Pons. I sincerely thank you for your support. Too much work and no play makes Jack a dull boy. Fun and relaxation were provided by Oscar Hector (‘the Protector’), Mark (‘dog’) Stephens, Jill & Stuart Cheney, Yves (‘bigpelastik’) Basset, Floyd (Mr. Yatzhee) Spencer, Abby Rose and Ivan & Faye Harris. In the Netherlands, guidance and encouragement from my supervisors kept me going. The many discussions I had with Prof. R.A.A. Oldeman and Prof. M.J.A. Werger were a continuous source of inspiration and insight. Thank you Professors, for your never-ceasing guidance and confidence. My thanks also goes to my colleagues at Vegetation Ecology; to Gerrit Heil, Heinjo During, Lourens Poorter, Pieter Zuidema, Marielos Peña, Zeng Bo, Liesbeth van der Hoeven, René Verburg and all the others. Discussions with Gart van Leersum, John Hendrison, Wyb Jonkers, Reitze de Graaf, Toon Rijkers and Marc Parren at the Department of Forestry in Wageningen were invaluable. I thank Joop Faber of the Biostatistics Department for showing me the ropes with respect to ‘dummy’ variables and logistic regression and Henk Koop and Rienk-Jan Bijlsma of the Dutch Institute for Forestry and Nature Research for their assistance with the use of SILVI-STAR. I thank Wanda Tammens and Jelle Maas of Tropenbos for assisting me with the final layout. This thesis never would have seen the light of day without the incredible patience and support of Marissa, who did much of the data entry and chased me back to my computer when I had sneaked out to watch football on TV. CONTENTS 1. GENERAL INTRODUCTION 1.1 1.2 1.3 1.4 1.5 Tropical rain forests and their use Tropical rain forest logging Logging in Guyana Objectives and outline of this thesis Guyana 1.5.1 Brief history, population and economy 1.5.2 Landscape and forest formations 1.5.3 Forest utilisation 1.6 The Tropenbos Programme 1.7 Study area 1.7.1 Location and landscape 1.7.2 Forest composition 1.7.3 Experimental lay-out 1 1 2 3 3 7 7 9 10 12 13 13 14 15 2. TROPICAL RAIN FOREST LOGGING AND SILVICULTURE: GUYANA 19 2.1 2.2 2.3 2.4 Forest management Tropical rain forest dynamics The history of tropical rain forest silviculture Logging, forest management and silviculture in Guyana 2.4.1 Forest utilisation before 1890 2.4.2 The period 1890-1954 2.4.3 The period 1954-1986 2.4.4 The period 1986-1998 2.4.5 The future 3. LOGGING METHODS 3.1 The emergence of Reduced Impact Logging 3.2 Conventional logging in Guyana 3.2.1 Felling 3.2.2 Skidding 3.3 Reduced impact logging 3.3.1 Pre-harvest inventory 3.3.2 Climber cutting 3.3.3 Tree selection criteria 3.3.4 Directional felling 3.3.5 Planned skidding 3.4 Equipment and personnel 4. LOGGING DAMAGE 4.1 Introduction 19 20 25 28 28 30 32 36 38 41 41 43 44 45 46 46 47 47 48 52 53 57 57 4.2 Material and methods 4.2.1 Study area and pre-logging measurements 4.2.2 Post-logging measurements 4.2.3 Volume measurement 4.2.4 Analysis 4.3 Composition and structure of the initial stand 4.4 Overview of the logging operations 4.4.1 Conventional logging operation 4.4.2 Reduced impact logging operation 4.4.3 Variation in the dimensions of felled trees 4.5 Loss of canopy cover 4.5.1 Accumulated loss of canopy cover 4.5.2 Size of canopy openings 4.5.3 Sizes of single and multiple treefall chablis 4.5.4 Single treefall chablis, tree diameter and climber cutting 4.6 Extent of disturbed ground area 4.6.1 Skid trails and traces of other skidder movements 4.6.2 Ground disturbance in felling gaps 4.7 Impact on the residual trees 4.7.1 Degree of damage and chances of survival 4.7.2 Impact of logging method and intensity 4.7.3 Impact of felling and skidding 4.7.4 Impact and size of damaged trees 4.7.5 Relation between tree damage and loss of canopy cover 4.7.6 Impact on the commercial stand 4.7.7 Impact on sustainable production of timber 4.8 Discussion 4.8.1 Comparison of results of logging damage studies 4.8.2 The effect of logging intensity 4.8.3 Comparison with natural gap mosaic 4.8.4 Implications for management and monitoring 5. LOGGING EFFICIENCY 5.1 Introduction 5.2 Methods 5.2.1 Cost elements 5.2.2 Time elements 5.2.3 Labour costing 5.2.4 Machine costing 5.2.5 Work study 5.2.1 Felling model 5.2.2 Skidding model 5.2.3 Skidding model scenarios 5.2.4 Limitations 5.3 Felling efficiency 59 59 60 63 63 64 67 67 68 69 74 74 79 83 86 88 88 92 93 94 95 101 106 113 114 119 123 123 130 131 134 139 139 140 141 142 143 143 146 147 150 154 154 155 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.3.1 Work study statistics 5.3.2 Operative work elements 5.3.3 Supportive work elements and delays 5.3.4 Felling time model Skidding efficiency 5.4.1 Work study statistics 5.4.2 Collecting 5.4.3 Travelling 5.4.4 Other operative work elements 5.4.5 Supportive work elements and delays 5.4.6 Skidding time model Cost of operation 5.5.1 Planning costs 5.5.2 Harvest preparation costs 5.5.3 Harvesting cost and output 5.5.4 Main transport 5.5.5 Support, logistics, supervision and overhead costs 5.5.6 Aggregate costs of logging 5.5.7 Effect of wood recovery and felling pattern The influence of logging intensity on the cost of an operation Comparison with other studies Required precision in directional felling Benefits of reduced impact logging Conclusion 155 157 164 167 169 169 172 176 184 185 185 187 187 189 192 194 195 196 198 199 202 206 207 211 6. SIMULATED SILVICULTURAL TREATMENT AND CROWN ILLUMINATION 213 6.1 Introduction 6.1.1 Silvicultural operations 6.1.2 Light conditions for potential crop trees 6.2 Methods 6.2.1 Selection of potential crop trees and their social position 6.2.2 The three-dimensional forest model SILVI-STAR 6.2.1 Assessment of social position 6.2.2 Evaluation of treatments 6.3 Results and discussion 6.3.1 Description of the tree populations 6.3.2 Selection and social position of favoured trees 6.3.3 Relationship between social position and received irradiation 6.3.4 Social position and tree characteristics 6.3.5 Crown size of favoured trees and impeders 6.3.6 Evaluation of different prescriptions for release 6.4 Silvicultural considerations 6.4.1 Why apply silvicultural treatment? 6.4.2 Selection of the ‘best’ method: the empirical way 213 213 217 218 218 219 221 224 225 225 228 231 233 238 241 248 248 248 6.4.3 Selection of the ‘best’ method: the deductive way 6.4.4 Conclusion 7. SUMMARY AND GENERAL DISCUSSION 7.1 7.2 7.3 7.4 7.5 7.6 7.7 DESIGN OF A REDUCED IMPACT LOGGING SYSTEM Canopy fragmentation Ground disturbance Damage to the residual stand The cost of introducing reduced impact logging The prospects of silvicultural intervention after first harvest Silvicultural considerations regarding Greenheart forests 251 252 255 255 259 262 263 265 270 274 SAMENVATTING 281 REFERENCES 293 APPENDIX A: LIST OF SPECIES 311 APPENDIX B: ESTIMATION OF TREE VOLUMES 317 APPENDIX C: COST CALCULATIONS 321 CURRICULUM VITAE 335 Chapter 1: General introduction 1. GENERAL INTRODUCTION ‘There is no doubt that most current logging techniques cause unnecessary damage to the forest; the fact is a blight on the forestry profession. Even without reducing volume cut, better logging techniques will have an immediate positive effect on the postharvest value of the forest and will increase the long term chances of sustainability’ (Sarre 1996) 1.1 TROPICAL RAIN FORESTS AND THEIR USE Tropical rain forests occur in hot, humid areas in a belt around the entire globe. Its northern and southern boundaries do not coincide exactly with any latitudinal limits: in some places they do not reach the Tropics of Cancer and Capricorn, in others they extend somewhat beyond them. Average annual temperatures are usually above 25°C, the mean monthly temperature never falling below 18°C. Average annual rainfall exceeds 1800 mm and no month has an average rainfall less than 60 mm (Jacobs 1988; Whitmore 1975, 1984, 1990; Richards 1952, 1996). Tropical rain forests are characterised by an immense diversity of plant and animal species, a complex structure and architecture of the vegetation and a mosaic of vegetation development phases (Hallé et al. 1978; Whitmore 1978, 1990). They are never homogeneous, neither in structure nor in architecture, even where there has been no felling or other artificial impacts. Gaps, varying in size and frequency, and here and there ‘climber tangles’ or thickets of immature trees interrupt stretches of high forest (Richards 1952, 1996; Oldeman 1978, 1990). Tropical rain forests have been used for centuries but ‘late twentieth century man, with his immense technological power, is out of harmony with nature and fast destroying the forests’ (Whitmore 1990). Millions of hectares of forest in South America, West and Central Africa and Southeast Asia are being converted to other forms of land-use, or degraded by the extraction of timber, fuelwood and a range of other products (e.g. Myers 1984; Van Beusekom et al. 1987; Werger 1992; Whitmore & Sayer 1992; Bruijnzeel & Critchley 1994). Estimates suggest that forest transformation in the tropics has increased to an alarming rate of some 15.5 million ha annually (FAO 1997), 11 million ha of which being converted each year to other uses, counterbalanced by 600,000 ha of new forest plantations. In addition, some 4.5 million ha are being degraded through selective logging (Marshall 1992), the vast majority without active management (Poore et al. 1989). Among scientists, politicians and the general public there is great concern about the future of the tropical rain forests. Issues contributing to this concern include: the alarming rate of tropical rain forest depletion; the often unsustainable uses to which the land is being converted; the violation of rights of indigenous forest-dwelling communities; the fast rate of species extinction and the potentially disastrous effects of deforestation and forest degradation on soil, water, 1 Reduced impact logging in the tropical rain forest of Guyana climate, genetic richness and the future supply of economic goods (Jacobs 1988; Poore et al. 1989). 1.2 TROPICAL RAIN FOREST LOGGING The most common form of timber extraction from tropical rain forest is selective logging. Such logging is selective because it involves the extraction of a limited number of marketable tree species, which generally comprise only a small proportion of the trees and species present (Baur 1964). In the last four decades, evolving technologies for wood processing and increasing global demand for wood products resulted in a substantial increase in the number of marketable species. Additionally, technological advances, such as the introduction of the chainsaw and of increasingly powerful road construction and log extraction vehicles, have greatly intensified logging operations (Dawkins & Philip 1998). As a result, the term selective logging has become misleading, because, although only a small percentage of the trees are generally felled for timber today still, a larger percentage is killed or damaged during logging operations (e.g. Johns 1988; Hendrison 1990; Jonsson & Lindgren 1990). The impact of selective logging includes the removal of harvested trees and associated damage to the residual stand. Removal of a significant proportion of the mature trees of marketable species reduces the potential seed sources for subsequent regeneration. Logging also leads to significant disturbance of the soil and catchment hydrology caused by the extraction by heavy machinery and the construction of roads (White 1994, Nussbaum et al. 1995; Pinard et al. 1996). Forest recovery following logging depends on the severity of the impact. Canopy openings and debris piles formed by felling are comparable to the natural chablis found in rain forests, and are often able to follow a similar successional pathway of canopy closure (Whitmore 1991). The areas traversed by log extraction machines, such as crawler tractors and wheeled skidders, are more severely damaged. All vegetation is crushed and most topsoil is removed (Gillman et al. 1985), and regeneration on such areas is extremely slow (Malmer & Grip 1990; Pinard et al 1996; Guariguata & Dupuy 1997). The severity of the impact of selective logging varies along with the intensity of harvesting and the logging method that is used. Reduced impact logging techniques aiming at diminished logging damage have been recently introduced in the tropics (Hendrison 1990; Bertault & Sist 1995, 1997; Johns et al. 1996; Pinard & Putz 1996; Webb 1997). Reduced impact logging techniques commonly consist of strict planning, control of logging operations and training of a well-motivated workforce (Dykstra & Heinrich 1992; Heinrich 1995; Dykstra 1996). Planning includes pre-harvest stock inventory and mapping, and topographic survey. Climber cutting, planning and demarcation of skid trails and directional felling also are important components. High logging intensities may have a disastrous effect on forest 2 Chapter 1: General introduction recovery (e.g. Chai & Udarbe 1977) and can seriously compromise the benefit of using reduced impact logging methods (Bertault & Sist 1997; Sist et al. 1998). 1.3 LOGGING IN GUYANA Commercial timber extraction has been practised in Guyana for centuries. Yet, despite this long history, the country retains one of the highest fractions of its land area covered by forest in the world, only rivalled by the neighbouring countries Suriname and French Guiana (Lanly 1982; Luning 1987). Until recently, the changes in the world timber trade and mechanisation of logging, that have changed the face of logging in other parts of the world, have not blotted this picture, and deforestation rates have remained consistently low (Burgess 1993). Recently however, international timber companies have started knocking on its door searching to offset a decline in harvestable volumes in Southeast Asia and West Africa (Repetto 1988; Colchester 1994, 1997). This development brought Guyana at a cross-roads where forest resources utilisation, conservation and preservation are concerned (Ter Steege 1998). Whilst acknowledging the value of its forests in respect to their importance in the conservation of biodiversity and in reducing the effects of carbon dioxide emissions, and ratifying many treaties in that respect, Guyana also indicated that the forest resources have to be beneficial to the country and its people, thereby stressing that this should conform to the principle of sustainable development (cf. Maini 1992). The habitual logging practice has not been much different from the one in other parts of the tropics. Most forest land is state-owned and concessions are given out to timber companies that log the forest, often in an unplanned manner, but always without much care for the remaining stand or future yields. For centuries logging focused on one species only, Chlorocardium rodiei, which provides a renowned timber for maritime purposes: Greenheart. Concern about the sustainability of Greenheart logging has been voiced repeatedly in the past without any tangible changes in practice (e.g. McTurk et al. 1882; Hohenkerk 1923; Swabey 1950; GNRA 1989). This concern has become immediate because of threatened and realised boycotts on the import of timber that is not extracted on a sustainable basis (cf. Putz & Viana 1996). A more extensively documented introduction concerning tropical rain forest ecology, logging and silviculture, in general and concerning Guyana in particular, which shaped the fundaments of the work reported in this thesis is provided in the following part of this chapter and chapters two and three. 1.4 OBJECTIVES AND OUTLINE OF THIS THESIS The main objective of the present study is to contribute to the formulation of a silvicultural concept under which timber can be extracted from Greenheart forest in 3 Reduced impact logging in the tropical rain forest of Guyana Guyana, and similar forests elsewhere, on a sustained yield basis. The specific objectives derived from this main objective are: I. to describe and understand the impact of the current logging practice on the forest and its consequences for forest recovery and future timber yields; II. to develop a reduced impact logging system that leaves the forest in a condition that favours a rapid recovery to a state that is silviculturally, ecologically and economically desirable; III. to analyse the effect of logging intensity on forest recovery and to determine at which intensity the benefit of using reduced impact logging techniques, if any, starts to be compromised; IV. to examine the costs and benefits associated with a change-over from habitual practice to reduced impact logging; V. to examine whether or not post-harvest silviculture is an option to increase productivity of logged forest. Following the present introduction, the text continues by providing the setting for this thesis, describing Guyana - its history, economy and forests - and the area where the present study was carried out; i.e. Pibiri, situated in the Demerara Timbers Ltd. concession in Central Guyana. Chapter 2 introduces the reader to tropical rain forest management and silviculture with special reference to Guyana. The chapter starts with an introduction of the basic concepts of sustainable forest management. This is followed by an overview of the main achievements in research into the broad field of natural forest dynamics, the latter dictating the silvicultural concept that can be applied. The development of an appropriate silvicultural concept and of a logging method that fits into that will have to take into account historical achievements and failures. General history of and developments in tropical rain forest logging, administration and silviculture are therefore reviewed. Finally, the history and development of logging and silviculture in Guyana is reviewed. Apart from a general description to put the study into perspective, major attention is given to the recurrent perception that continued supply of Greenheart is endangered by the forest administration system and logging methods of the different eras and that lessons can be learned from the answers to these problems. Guyana’s forest resources are also said to be extremely valuable and to be able to contribute strongly to the sustainable economic development of the country. Yet, Guyana’s government and business community are far from receiving even a small fraction of this potential revenue. Low timber density is named as one of the causes and indeed the average logging intensity does not seem to exceed two trees per hectare. Also repeated efforts to promote other species have failed until recently. Chapter 3 starts off describing the mechanisms behind the emergence of reduced impact logging techniques. This is followed by a description of the conventional logging system, which qualifies as a ‘close to best practice operation’; the reference against which the reduced impact logging system was tested. The reduced impact 4 Chapter 1: General introduction logging system, its rationale and components are described subsequently. Special attention is given to the fact that the system applied in the present study includes both ‘traditional’ reduced impact logging techniques and tree selection criteria. It has often been stated that Greenheart is being overharvested in proportion to its relative abundance among commercial species in Guyana. The first question that is tackled in Chapter 4 is: 1. What are the opportunities to shift the attention away from Chlorocardium rodiei to other commercial tree species in the Greenheart forest? Selective logging has a direct impact on the residual stand through changes in species composition and size class distribution. It also has an impact on the regeneration processes that will take place after logging through the resulting coarseness of the forest mosaic, the size of felling gaps, disruption of the propagule bank and soil compaction. The reduced impact logging system needed to be compared with conventional logging on account of both of these aspects. In order to describe and understand the differences in impact properly, a distinction is made between damage by felling and damage by skidding. It is also examined to what extent a difference between the two systems, if any, is compromised by logging intensity. The following questions were raised: 2. What is the extent of the total loss of canopy cover under the two logging systems and how is this influenced by logging intensity? 3. To what extent does the coarseness of grain of the forest mosaic differ between the two systems and how is this influenced by logging intensity? 4. What is the extent of the area traversed by the skidder under the two logging systems and how is this influenced by logging intensity? 5. What is the extent of the damage to the residual stand under the two logging systems in terms of total number of trees, total basal area and severity of the inflicted damage and how is this influenced by logging intensity? 6. If such differences exist, are these due to felling or skidding and are there differences in the size classes that are affected? 7. Is there any difference in damage between commercial and non-commercial species? It is also considered whether the experimental design has been appropriate to find answers to these questions. Obviously, the answers to these questions should enable the formulation of an optimum logging regime. Finally, it is considered important to find out whether the tree selection criteria and logging techniques had any specific effect on the results. This led to the following questions: 8. How do the initial forest composition, structure and architecture influence logging damage? 9. What are the specific effects of the tree selection criteria and the reduced impact logging techniques? 10. What is the optimal logging regime for these forests? 5 Reduced impact logging in the tropical rain forest of Guyana Apart from the importance of silvicultural and ecological aspects, it can be assumed that the country-wide establishment of a sustainable forest management system ultimately depends on its economic merits. The economic comparison of reduced impact logging with conventional logging is thus of especially great importance for the acceptance of the suggested harvest regime. In Chapter 5 the following questions were addressed: 11. What is the effect of the reduced impact logging system on the performance of the felling and the skidding operation? 12. If there are any differences, what is the effect on the cost of logging? 13. What are the specific effects of differences in wood recovery per extracted tree, the tree selection criteria and the reduced impact logging techniques? 14. What is the effect of logging intensity on the cost of logging? 15. How widely valid are the results obtained in the present study? 16. If no unequivocal results are obtained, which specific factors drive the cost of logging? 17. Is the implementation of the reduced impact logging system viable, and if not, is there room for improvement? Reduced impact logging may assist in reducing damage to ‘advance growth’ to a minimum, but generally sound harvesting alone is not sufficient to cause a uniform positive reaction from advance growth or natural regeneration of commercially desirable species. The forest therefore usually shows only marginal volume increment of commercially desirable species (e.g. Synnott & Kemp 1976; De Graaf 1986). Moreover, the remaining canopy is stripped of seed trees belonging to commercial species, in the end leading to progressive economic impoverishment of the forest. The desire to apply a silvicultural post-harvest treatment that ought to restore the balance in the species composition is obvious. Silvicultural treatments are usually directed at increasing the crown exposure of potential crop trees to light. In Chapter 6 it is examined whether or not silvicultural treatment can improve the social position of potential crop trees. A three-dimensional forest model was used to address the following questions: 18. What are the light conditions for potential crop trees in forest logged selectively under the reduced impact logging regime? 19. What is the relationship between the social position of potential crop trees and their crown illumination? 20. What type of treatment is most successful in improving the crown illumination of potential crop trees? 21. Can allometric relations, e.g. between crown and trunk diameter, be used in combination with the distance between trees to predict which neighbouring trees should be eliminated in order to improve the crown illumination of potential crop trees? Chapter 7 finally integrates the findings concerning the ecological and silvicultural impact of the two logging systems, the cost-benefit ratio of each of the two systems and the opportunity for post-harvest silviculture. The chapter concludes by outlining a 6 Chapter 1: General introduction proposed silvicultural concept for greenheart forests including recommendations for a logging regime and monitoring, and defining which elements are missing for the formulation of a complete silvicultural system. 1.5 GUYANA Guyana is situated on the northern coast of South America, between 0°45' and 8°38' north latitude and 56°32' and 61°22' west longitude. It is bounded on the north by the Atlantic Ocean, on the south and south-west by Brazil, on the east by Suriname and on the north-west by Venezuela. The country covers an area of 214,970 km2. It extends 800 km south from the Atlantic coast and about 480 km from east to west. The climate is tropical with average daily temperatures of 26°C. Rainfall is about 2700 mm/year and shows a bimodal pattern. Most rain falls from May to August with a precipitation between 300 and 400 mm/month. The second rainy season is in the period December-January. The ‘dry’ months still have a rainfall of more than 100 mm, with October being the driest month. For further details see Jetten (1994) and Van Kekem et al. (1996) 1.5.1 Brief history, population and economy ‘Guiana’ (= land of many waters) was the name given the area sighted by Columbus in 1498, comprising modern Guyana, Suriname, French Guiana and parts of Brazil and Venezuela, but it was not until a century later that the first Europeans settled in the area. The Dutch settled in Guyana in the early 17th Century. Initially welcomed by the indigenous Amerindians as trading partners, colonial government and exploitation of the Amerindians, and later of African slaves, followed. Interrupted briefly by the French and British, Dutch control ended when the British became the de facto rulers in 1796. In 1815, the colonies of Essequibo, Demerara, and Berbice were officially ceded to Great Britain at the Congress of Vienna. Following the abolition of slavery in 1834, indentured labourers were brought to Guyana to replace African slaves on the plantations, primarily from India but also from Portugal and China. The British stopped the practice in 1917. Many of the Afro-Guyanese former slaves moved to the towns and became the majority of the urban population, whereas the Indo-Guyanese remained predominantly rural. The small Amerindian population still lives in communal settlements in the interior. Guyana obtained independence from Great Britain in 1966 and became a republic in 1970. In 1998, Guyana counted 863,000 inhabitants with an annual growth rate of 2.3%. This population growth was countered by a high rate of emigration during the last two and a half decades, 1.5% per annum on average, which has resulted in a loss of persons with managerial and technical skills and has been a constraining factor to development (GNRA 1989). Ninety percent of the inhabitants live on the narrow coastal plain, where population density is more than 115 persons per km2, with about 180,000 living in Georgetown, the capital, and surrounding suburbs. Whereas most of the coastline is more or less continuously settled, the population in the interior is sparse, the population 7 Reduced impact logging in the tropical rain forest of Guyana density for Guyana as a whole being less than four persons per km2. Guyana's population comprises five main ethnic groups: 51% of East-Indian origin, 29% of African origin, 4% Amerindian, 2% of European and Chinese origin and 14% of mixed origin. Amerindians, the original inhabitants now numbering about 38,000, occupy 65 reserved areas, districts and villages, but also settlements not officially recognised as Amerindian lands. The settlements are mainly found in the south, south-west and north-west of the country and the sub-coastal parts of Berbice and Demerara. Both gold and diamond mining and logging have been on the increase and the Amerindians have been experiencing both beneficial and deleterious effects. There are no settlements within the limits of the timber concession where the study area is located. Two small settlements occur just outside the boundary. The people there have largely departed from the traditional livelihoods and are engaged in logging and mining. The Guyana economy has traditionally been dependent on a narrow range of products for its exports, employment and gross domestic product. Four products: bauxite, gold, sugar and rice, have been its principal foreign exchange earners. At independence, Guyana inherited an economy almost wholly dependent on exports of primary commodities. The industries producing these goods were largely owned by foreign, mainly British and American companies, whose influence over the colonial state had been paramount. Consequently, the colonial powers had invested little, except in infrastructure linked to these industries. After independence in 1966 Guyana pursued a policy of ‘co-operative socialism’, with increasing state intervention in the economy. The results were dismal. Real Gross Domestic Product grew only at 0.4 % a year over 1966-1988. The sugar and bauxite industries, the two pillars of the economy, collapsed after they were nationalised. After the oil price hike of 1973 and the collapse of world prices for sugar and bauxite, economic performance deteriorated further in the 1980s. Demand policies were expansionary, the real exchange rate appreciated, and the government relied increasingly on price controls and quantitative restrictions on trade. These policies resulted in a decline in Real Gross Domestic Product by 2.8 % a year over 1980-88, a dramatic increase in debt to about US$ 1.9 billion and a large parallel economy. With the economy in a precarious state, the government reversed course in 1988, and embarked on an economic recovery programme of broad macro-economic measures and structural reforms to realign relative prices, dismantle state controls, and establish a market-oriented economy. The economic recovery programme has improved Guyana's economic performance considerably. Economic activity began to recover in 1991 and Real Gross Domestic Product increased by 6 %. Price and exchange rate liberalisation and improved incentives for private sector development had a significant impact on increasing production in all sectors and began integrating the parallel economy into the official economy. Economic recovery continued over 1992-1997, with Real Gross Domestic Product growth averaging about 7 %, resulting in a Real Gross Domestic Product of US$ 634 million or US$ 734 per capita. At the same time, inflation declined significantly, from 83 % in 1991 to 4 % in 1997, and the nominal exchange rate, which initially had depreciated considerably, remained stable until 8 Chapter 1: General introduction 1998. Guyana experienced its first recession in seven years in 1998, but economic growth is expected to resume although at lower rates than in the recent past. The currency is presently at the level of US$ 1.00 = G$ 172 (February 1999). Guyana's external financial position, however, remains precarious because of the heavy debt service for the large public debt of US$ 1.5 billion. With the majority of the government revenues committed to debt servicing - debt to Gross Domestic Product ratio is 204% -, any contribution from its national budget toward maintaining the basic infrastructure, crucial to all economic sectors, is severely limited. Sources: Guyana news and information: http://www.guyana.org/ backgrou.htm; The World Bank: http://www.worldbank.org/html/extdr/offrep/lac/gy.htm, The Inter-American Development Bank: http://www.iadb.org/int/sta/brptnet/ guybrpt.htm and http://www.iadb.org/regions/re3/sep/gy-sep.htm; and Colchester (1997). 1.5.2 Landscape and forest formations Four main landscapes can be recognised in Guyana: 1. the Coastal Belt of low-lying alluvium, which has either been reclaimed for agriculture or which still carries a complex series of forest, swamp, and marsh communities; 2. the Lowland Region of undulating forest land, generally below 150 m in elevation, where the main forestry and mining undertakings are carried out; 3. the Pakaraima Montane Region, an elevated tableland, rising to an altitude of 3000 m; 4. the Southern Upland Region of undulating forest land and savannah, generally above 150 m in elevation. The Lowland Region is usually divided into the Pre-Cambrian Lowland Region and the Sandy Rolling Lands or Berbice formation. The intermediate savannahs, the dry evergreen forest, the seasonal forest and the rain forest occupy the latter. The Pre-Cambrian lowland supports mainly rain forest. Soils pertaining to the Berbice formation are old, infertile and highly acidic, thereby low in available salts. Soil types correspond closely to geomorphology. An often-observed gradient from watershed to valley bottom is white sands, light brown sands, sandy loam, silty clay. Soils pertaining to the Pre-Cambrian landscape are loams and clays and may contain ironstone gravel. The land is well watered and typically drained by a dendritic pattern of many larger and smaller creeks that flow into the large rivers that discharge into the Atlantic Ocean. Fanshawe (1952) recognised four main forest formations: 1. rain forests on well drained sites; brown sands, loams or red earths; 2. dry evergreen forest on sites where the soil moisture is inadequate to meet potential evapotranspiration; mainly the white sand plateaux; 3. marsh forest or seasonal swamp on temporarily inundated sites, where moisture conditions fluctuate from water-logged to dry; 9 Reduced impact logging in the tropical rain forest of Guyana 4. swamp forest, subjected to temporary inundation, where the soil rarely has a chance to dry out completely. The tropical rain forest of Guyana is exceptional in the sense that large parts are dominated by a few or one species (Davis & Richards 1933, 1934; Fanshawe 1952; Richards 1952, 1996). Dominance may result in clearly identifiable forest types. Forest types are locally often described by the dominant species, such as e.g., Greenheart forest (with dominance of Chlorocardium rodiei1 (Schomb.) Rohwer, Richter & v.d. Werff [syn. Ocotea rodiaei (Schomb.) Mez]), Mora forest (Mora excelsa Benth.), Morabukea forest (Mora gonggrijpii (Kleinh.) Sandw.), Wallaba forest (with Eperua falcata Aubl. - Soft Wallaba - and Eperua grandiflora (Aubl.) Benth. - Ituri Wallaba). Co-dominant species may differ, especially in Greenheart forests, which makes a typological classification in plant communities difficult (Ter Steege et al. 1993). 1.5.3 Forest utilisation Some 80% of Guyana’s land area is covered with forest; i.e. some 16.8 million hectares. The major part of these forests is still untouched. In 1997, the permanent forest estate (‘State Forests’) covered 8.9 million hectares, 73% of which had been allocated for timber harvesting according to three forms of tenure. These are the Timber Sales Agreement (TSA), the Woodcutting Lease (WCL) and the State Forest Permission (SFP). TSAs are issued for areas larger than 24,200 ha for periods of 10 to 25 years with a right to renewal. TSA holders are required to undertake forest inventories and submit management plans. WCLs are issued for areas between 8,100 ha and 24,200 ha for periods ranging from 3 to 10 years with an option for renewal. WCL holders are subjected to the same conditions as TSA holders. SFPs are issued for areas less than 8,100 ha for one year only and do not grant the concessionaire exclusive rights to the area. In 1987, approximately 2.4 million hectares had been given out through the three forms of tenure for harvesting. Dwindling timber resources in South East Asia and the change in Guyana’s economic policy attracted foreign logging companies and, by 1996, all types of permits combined covered approximately 6.5 million hectares. Forty-one per cent of this area is covered by 17 TSAs and 14% by nine WCLs, together covering 4.8 million hectares, mainly located in the central and north-western parts of the country. The approximately 480 SFPs, comprising 1.7 million hectares, are mainly located in the north-eastern part of the country (Fig. 1.1). The contribution of the forest industry to Gross Domestic Product rose from barely 1% to almost 5% during this decade (1987-1996). In 1981 the area of logged forest was estimated at 1.4 million ha (FAO, 1981). Log production was steady during 1981 to 1992 at an average annual output of 1 Names and authorities follow Mennega et al. 1988, except species that were renamed since, following Ek and Ter Steege (in prep.). No further reference to authorities or families is made in the text. Mostly scientific names are used. When first used the scientific name will be followed by the vernacular name in parentheses. In appendix A, a checklist of all species encountered in the plots is provided including authorities and families. 10 Chapter 1: General introduction 160,000 m3/yr, but has been rising since; i.e. from 220,000 m3/yr in 1993 to 520,000 m3/yr in 1997. It has been reported that forests are being logged at an intensity of 5 m3/ha or less. This indicates that since 1981 another 770,000 hectares would have been logged, resulting in a total estimated area of logged forest of 2.2 million ha. Figure 1.1 Map of Guyana showing the area occupied by large concessions (TSAs and WCLs) - shaded dark grey - and by small concessions (SFPs) - shaded light grey -. The location of the Mabura Hill township and the study area at Pibiri are also shown. Just as the national economy was narrowly based on a few primary commodities, the timber industry, traditionally, relied on a handful of species only. Greenheart used to be the mainstay of the forest industry, with over 40% of the annual log extraction of 200,000 m3, during 1955-90. In terms of export, the species made up as much as 70% of the 40,000 m3 of wood products that were exported. After the arrival of foreign-owned companies and the erection of a large plywood mill in the early 1990s, species suitable for plywood came in demand. Annual log extraction soared to 520,000 11 Reduced impact logging in the tropical rain forest of Guyana m3 in 1997, with Catostemma commune and C. fragrans (Baromalli) taking over Greenheart’s position as the most extracted species. Although the contribution of Greenheart declined to 15% of both extracted volume and exported wood products, the pressure on Greenheart was not released, the annually extracted volume remaining at the same level. At the start of the present study in 1992, Greenheart was thus the main timber producing species of Guyana, which explains why this study - and most other Tropenbos studies in Guyana - concentrate on Greenheart forests. Sources: Guyana Forestry Commission (Anonymous 1998); Colchester (1997); Sizer (1996). 1.6 THE TROPENBOS PROGRAMME The Tropenbos Foundation was established in 1988, to continue and expand the International Tropenbos Programme set up by the government of the Netherlands in 1986. The main objectives of the Foundation are to contribute to the conservation of tropical rain forests, promoting wise use, research and capacity building in a number of tropical countries (Tropenbos 1998). With project sites in Cameroon, Colombia, Côte d’Ivoire, Guyana and Indonesia, its contribution to global understanding of rain forests is increasingly well known. Numerous academic publications have been produced as a result of Tropenbos-funded research and a large number of post-graduate students from tropical countries have been supported by the Foundation. With forest management as an important theme, Tropenbos-funded research has contributed importantly to the understanding of forest ecology, water and nutrient cycling, and parameters for sustainability. In recent years, specific research themes include the use of non-timber forest products, biodiversity and criteria and indicators for sustainable forest management. The Tropenbos-Guyana programme, in which the Guyana Natural Resources Agency and the Universities of Guyana and Utrecht collaborate on forest research, commenced in 1989 with an intergovernmental agreement between the Netherlands and Guyana. For many years, this work has predominantly involved biological and ecological studies focusing on sustainable forest management for timber production in particular. Recently the work was extended with social, economic and anthropological research into the use of non-timber forest products. Presently its activities focus on: ƒ knowledge of natural resources: physical environment, biodiversity and timber characteristics; ƒ development of parameters for sustainable forest management; ƒ the significance of non-timber forest products for indigenous communities. The research is carried out at two sites - Mabura Hill and the North West District - in or near logging concessions. A comprehensive review of results of the first phase of 12 Chapter 1: General introduction the programme (1989-1993) and recommendation for forest management have been compiled by Ter Steege et al. (1995, 1996). Currently the second phase of the programme is nearing its completion. 1.7 STUDY AREA 1.7.1 Location and landscape The study was conducted in the West-Pibiri compartment of the Demerara Timbers Ltd. (DTL) concession in Upper Demerara. Logging activities in the DTL concession had started around 1985 and by the end of 1995 a total area of 68,600 hectares (Quicke 1996) had been selectively logged, mainly for Greenheart. Large parts of the logged area are covered by dry evergreen forest and have been worked less intensively than the mixed forest. An area covering approximately 1000 ha was set aside for experimentation and handed over to the Tropenbos Foundation in 1992 before logging had commenced there. The area is located about 40 km south of the Mabura Hill township and roughly 250 km south of Georgetown, between 5°05' and 5°10' north latitude and 58°25' and 58°35' west longitude. This location was selected because of the relative homogeneity in site characteristics and vegetation, as well as the fact that, at the time, the site represented the nearest, accessible area of unlogged Greenheart-bearing forest that could accommodate such an experiment. The study site at West-Pibiri spans an altitudinal range from 50 to 100 m above sea level. The slopes of divides and interfluves being gentle (< 20 %), the site is part of an undulating to rolling sedimentary plain. The Pibiri creek with its many tributaries drains the area and flows in easterly direction into the Demerara River. Most soils belong to the Berbice formation, consisting of more or less horizontally lying sediments that overlie the older Basement Rocks. Ridges of the Basement Rocks, covered by a sheet of lateritic ironstone, are surrounding the lower lying Pibiri plain on the northern and western side, thereby separating it from the white and brown sands in the northern part of the concession. The soils are very deep and well drained with topsoil textures varying between unbleached sands, loamy sands, sandy loams and sandy clay loams. Deeper subsoils are locally yellowish red clays, and may contain ironstone gravel (Van Kekem et al. 1996). The first major soil group in the area consists of Ferralsols (USDA Soil Taxonomy: Kandiudults) with a moderate (Guyana series: Kasarama sandy loams) to high percentage of clay (Guyana series: Ebini sandy clay loams - Kahn 1994, pers.comm.; Vanmechelen, 1994; van Kekem et al. 1996). The second major soil group is made up by Arenosols. These soils consist of excessively drained, bleached coarse sand. The reduced impact logging experiment was carried out on two brown sand ridges and adjacent slopes separated by a white sand ridge in the study area. Conventional logging was studied on two brown sand ridges and adjacent slopes in an area bordering the study area on the Southeast (Fig. 1.3). 13 Reduced impact logging in the tropical rain forest of Guyana 1.7.2 Forest composition The forests studied are evergreen, situated on brown sand and contained considerable quantities of Greenheart. Greenheart forest is found interspersed by other forest types; e.g., Wallaba forest on white sand, Mora forest along creeks, mixed forest lacking Greenheart and liana forest. The forests had not been logged before. Occurrence of charcoal in the soil indicates that the forest suffered disturbance in the not-so-distant past (see also Hammond & Ter Steege 1998). Residents of a nearby Amerindian community reported that forest fires had occurred in the area only 50 years ago. Charcoal was indeed found near the soil surface close to the Pibiri Creek. Further away from the creek where the study took place, however, charcoal was found more deeply in the soil and hence having originated much longer ago. Average canopy height is 30-40 m with emergents up to 50 m. Stem density of trees with a diameter of 20 cm and over is on average 180 stems per hectare (Vanmechelen 1994; Van der Hout 1996). Canopy dominants often occur in groups, so that the canopy presents an uneven appearance. Monodominance occurs frequently and groups are often composed of one or few species. Lianas are occasional to very common, epiphytes scarce to moderately frequent. Buttresses, except in the stands dominated by Mora gonggrijpii (Morabukea), are not common. About 10% of the tree species are strongly buttressed. Palms are few, mainly stemless and occur mainly in the undergrowth. Patches dominated by Greenheart (locally called ‘reefs’) vary from small groups of some four to six mature trees on some 500 m2 to large groups of some 20 to 30 mature trees on some 2500 m2. The patches usually have an elongated shape and are mainly found on the upper slopes along ridges. Species composition is variable, partially influenced by site factors such as soil texture, topography and water availability (see also Ter Steege et al. 1993), and partially by differences in successional phase in forest dynamics sensu Hallé et al. (1978) and Oldeman (1974a; 1983). Greenheart and Lecythis confertiflora (Wirimiri) frequently dominate the rather open canopy with Licania spp., Swartzia leiocalycina (Wamara) and Catostemma fragrans (Sand Baromalli) as prevalent co-dominant species. In some parts Mora gonggrijpii (Morabukea) is very abundant and on wetter soils Carapa guianensis (Crabwood) and Pentaclethra macroloba (Trysil) are common. Among the emergents, Peltogyne venosa (Purpleheart), Hymenaea courbaril (Locust) and Parinari campestris (Burada) are the most common. Prevalent understorey species are Oxandra asbeckii (Karishiri), Tapura guianensis (Waiaballi) and Paypayrola longifolia (Adebero). See Vanmechelen (1994), Van der Hout (1996) and Ek & Van der Hout (in prep.) for more detailed information on species composition. Ek (1997) estimated that the experimentally logged stands hosted up to 83 climber genera and 132 species (98 woody climbers, 22 vines, and 12 woody hemi-epiphytes) prior to logging. The total stocking of climbers was estimated at 2660 per ha (height > 2 m), but most of them were small; i.e. 83% being smaller than 2 cm dbh. The number of climbers larger than 10 cm was estimated at 21 per hectare. Most climbers (61%) climb on supports with a diameter < 5 cm, most of these being a climbers themselves. 14 Chapter 1: General introduction The estimates of the number of lianas > 2 cm were similar to estimates quoted by Fox (1968a) for lowland rain forest in Sabah, but lower than estimates by Pinard & Putz (1996) and Campbell & Newbery (1993). Only some parts of the forest can be classified according to Fanshawe's (1952) community classifications; i.e. the Lecythis confertiflora (Wirimiri) ‘lociation’, the Mora gonggrijpii (Morabukea) ‘faciation’ and the Vouacapoua macropetala (Sarebebeballi) ‘faciation’. Although most parts cannot be fitted in his classification exactly, the forest can best be described as belonging to the Eschweilera-Licania (Kakaralli-Kautaballi) association. Before 1992, most Tropenbos research had been carried out in or near the Tropenbos Ecological Reserve, which is located 25 km north of the Pibiri study area. There are a few but important differences between the Pibiri area and the reserve and surrounding area; i.e., ƒ the majority of the Pibiri soils belong to the so-called brown sands whereas white sands and reddish clays containing ironstone gravel constitute the majority in and around the reserve, and ƒ some species that are very common in and around the reserve such as Dicymbe altsonii Sandw. (Clump Wallaba) and Eperua rubiginosa Miq. (Water Wallaba) are not found in the Pibiri area (Ter Steege et al. 1993; Ek 1997). Following Fanshawe’s (1952) classification these differences could even be classified as major ones, because the Pibiri area would classify as his first major floristic group; the Eschweilera-Licania association, and the area around the reserve as the other major floristic group; the Eschweilera-Dicymbe association. In the wake of this study, more Tropenbos studies were initiated in the Pibiri area; a study of the effect of logging on botanical diversity (Ek 1997) and several experimental studies at the size of artificial gaps (Van Dam & Rose 1997; Rose 1997; Van Dam et al. 1999) 1.7.3 Experimental lay-out The Pibiri area had been subdivided by the logging company into 100 ha blocks, their borders demarcated by cut lines, in 1992. Seven blocks appeared to harbour Greenheart forest, two blocks were mainly covered by Wallaba forest and the remainder was classified as swamp or Mora forest. Within a block different forest types occurred usually coinciding with sites as defined by topography, soil type or drainage class. For reasons of accessibility, only four blocks were considered suitable for experimental logging. In those blocks trees belonging to 20 commercial species down to 40 cm dbh were inventoried and mapped in the latter part of 1992. This information was used to select 15 plots containing sufficient harvestable trees. The plots, measuring. 240 × 240 m2, and numbered from 1 to 15, were established between March and July 1993. The 49 sub-plots in each plot are numbered 11 to 17, 21 to 27, 31 to 37, etc., the first digit counting from west to east and the second from south to north (Fig. 1.2). Each plot contains a central assessment area that measured 15 Reduced impact logging in the tropical rain forest of Guyana 140 × 140 m2, which was subdivided into 20× 20 m2 recording units. Between July and November 1993 all trees ≥ 20 cm dbh were measured at 1.3 m above the ground, or 30 cm above the buttress, using a steel diameter tape. Nested subplots, 25 in each plot, were used for smaller trees: 100 m2 for stems from 5 to 20 cm dbh; and 25 m2 for saplings from 2 to 5 cm dbh. This set-up followed Alder & Synnott (1992). Plot lay-out Nested recording units 2m 140 m 5m 10 m 20 m 240 m Figure 1.2 Plot lay-out and subdivision of nested subplots in the reduced impact logging experiment. Trees in the two-ha plots were recorded with differential intensity according to size. See text for explanation In the area where conventional logging was to be monitored, stock maps produced by the timber company were used to select suitable areas. In all three units, measuring 200 × 600 m2 were selected and demarcated in May 1995 (Fig. 1.3). In the centre of each unit a 20 m wide, 500 m long strip was enumerated. All trees ≥ 20 cm dbh were measured in each strip. A nested five meters wide section of the strip was used for the smaller trees with stems from 5 to 20 cm dbh. In both conventional and experimental logging areas, each tree was labelled with an aluminium number tag for reliable identification after logging and to facilitate long-term monitoring. All trees were identified up to species or genus using a combination of characters: general form (e.g. buttressing, height), bark texture, slash, odour and exudates, and leaf type and shape (cf. Lindeman & Mennega 1963). Stem and crown damage was described and any other tree characteristics that might be mistaken for logging damage were noted. The reduced impact logging experiment serves two purposes. Besides a comparison between experimental and conventional logging, it was also set up to gain insight in the effect of logging intensity and post-harvest liberation thinning on growth and yield. Four treatments and a control are being tested in the experimentally logged area: 1. low intensity reduced impact logging, removing 4 trees per hectare; 16 Chapter 1: General introduction 2. 3. 4. 5. moderate intensity reduced impact logging, removing 8 trees per hectare, which yields approximately 25 m3/ha as recommended by De Graaf (1986); high intensity reduced impact logging, removing 16 trees per hectare, which is the average stock of currently merchantable stems down to 40 cm dbh in these forests; moderate intensity reduced impact logging, removing 8 trees per hectare, to be succeeded by a silvicultural post-harvest treatment; control; no intervention. Figure 1.3 Map showing the location of the experimentally logged plots (≈ 6 ha), with logging intensity, and conventionally logged blocks (≈ 12 ha) that were studied in the Pibiri area, Central Guyana. (Contours have a spacing of 50 ft.) Treatments were assigned to the plots at random (Fig. 1.3). The twelve experimental plots were logged in 1994 following the procedures described in Section 3.3. Climber cutting took place in advance of logging in November-December 1993. Skid trails and 17 Reduced impact logging in the tropical rain forest of Guyana road alignments were marked on the ground during January-February 1994. Felling operations started in March 1994 and were completed in July 1994. Skidding started in September 1994 and was completed by November 1994. The three conventionally logged blocks were logged in March and April 1996. After logging the blocks were surveyed in order to outline 2-hectare plots, measuring 100 × 200 m2, with logging intensities of 8 and 16 trees per hectare. Plots with a logging intensity of 4 trees per hectare could not to be found. 18 Chapter 2: Tropical rain forest logging and silviculture 2. TROPICAL RAIN FOREST LOGGING AND SILVICULTURE: GUYANA 2.1 FOREST MANAGEMENT It is commonly held that the term ‘management’ stands for a series of technical and administrative activities that pursue economic returns. In this sense, forest management would accrue direct and indirect economic benefits (Ciancio & Nocentini 1997). In recent years, forest management has taken on a wider and more general connotation. A similar change occurred in the connotation of ‘sustainable forest management’. Its context widened from the (age-old) silvicultural principle that aimed at sustainable timber production to a much broader context including ecological, economic and ethical considerations. Sustainable forest management is an assumption of virtually every forest manager and the term is used widely, but also rather loosely. To understand sustainability in reference to natural resources and management of complex ecosystems, the following questions must be asked (Maser, 1992): ‘what is to be sustained, at what level is ‘whatever’ to be sustained, how long is it to be sustained, and for whom?’ In some instances sustainable timber production is used to refer to continuity of supply of timber in general, implying that when one source (species or stand) is exhausted (depleted), another will be found. Such a conception of ‘sustainability’ is dangerous (Poore et al. 1989). Indeed, provisions for the continued output from harvested sites are not necessarily included and this can lead to the total destruction of the resource in the longer run. According to Poore et al. (1989), there are various levels of intensity at which forest management could be termed ‘management for sustainable timber production’: ‘wait and see’; ‘log and leave’; ‘minimum intervention’; ‘stand treatment’ and ‘enrichment planting’. The intensity thus varies between zero intervention and highly intensive management. The first two categories speak for themselves. Minimum intervention implies that marked trees are removed up to the limit defined by the allowable cut, leaving behind an adequate stocking for the next crop to be assumed to be available after a defined number of years. Stand treatment implies that logging is carried out in the same way as in the previous strategy, the latter being assumed to open the door to regeneration, but certainly not of commercially desirable species alone. Since logging has a two-sided effect and also impoverishes the stand, the growth and survival of the remaining stems or regeneration is supposed to be enhanced by certain silvicultural treatments, on historical grounds related to the thinning concept developed in timber plantations as long as two centuries ago in Europe. The level of intensity is, obviously, foremost determined by the forest itself. Only sites suited to generate new stands of high volume, quality, and value should be managed intensively (and expensively), and the remaining sites, which would constitute the major part of the area, should be placed under extensive (and cheap), minimum-intervention management (Kleine & Heuveldop 1993). Poore et al. (1989) stress that these levels of intensity can be termed management options if - and only if - a conscious decision has been made to manage 19 Reduced impact logging in the tropical rain forest of Guyana the forest at that particular level, if this decision is being conscientiously applied, and if the results are being monitored. One of the hottest items of discussion has been the estimation of the volume of timber, which can be harvested on a sustainable basis; the annual allowable cut. The traditional annual allowable cut calculations date back to the 1800s, when European foresters started to emphasise the notion of regulating yields to produce an even flow of timber. Over the years of forest history, many methods for determining the cut for both evenand uneven-aged forest stands were developed, for the latter especially in Central Europe. These methods have in common that they aim to attain a ‘regulated structure’ of the forest. The essential requirements of a fully regulated forest are that age and size classes be represented in such proportion and be consistently growing at such rates that an approximately equal periodic yield of products of desired sizes and quality may be obtained in perpetuity. Application of this concept is fraught with difficulty; one common problem is that few stands are composed of balanced mixtures of stand ages (Smith et al. 1996). Furthermore, the following complication arises (Poore et al. 1989): when a forest is logged for the first time it normally contains a higher standing volume of timber than at subsequent cuts. The growth rate of the remaining trees is likely to change after the first harvest. The annual allowable cut calculated for the primary forest is thus likely to be different from - probably higher than - the one that may be expected in the future. It is equally important to realise that even with the most suited silvicultural system future outputs of goods and services at desired levels cannot be assured. Complex ecosystems often do not behave as expected (Oldeman, 1991; Moir & Mowrer, 1995). This suggests a precautionary, conservative approach to forest management. It is therefore suggested that management of tropical rain forests should be oriented towards mimicking natural forest dynamics, indicating a retention of the natural mosaic of different stages of forest development, minimising logging damage, adapting yield regulation and diversifying forest utilisation (Kleine & Heuveldop 1993; Oldeman & Vooren 1980 ex Vooren 1987). 2.2 TROPICAL RAIN FOREST DYNAMICS The long-term consequences of selective logging in tropical rain forest are not very well known. Are the felled trees replaced by the same species in the long-term, or does logging permanently upset the balance of species? If the forest is to be utilised for sustainable timber production then apart from the impact of logging on the residual trees and advanced regeneration, it is of fundamental long-term concern to see how the future species composition is affected. Studies on forest regeneration processes after selective logging have been carried out mainly in Southeast Asia and Africa, most of them shortly after logging (e.g. Liew & Wong 1973 in Sabah; Van Gardingen et al. 1998 in Indonesia; Hawthorne 1993 in Ghana; and Van Loon 1997 in Cameroon) and a 20 Chapter 2: Tropical rain forest logging and silviculture few after a longer period (e.g. Meijer 1970 in Sabah; Oldeman & Van der Meer 1988 in Queensland; and Plumptre 1996 and Chapman & Chapman 1997 in Uganda). Only recently, some studies have been conducted in the neotropics (e.g. Silva et al. 1995; Zagt 1997; Webb 1998; Magnusson et al. 1999). The last two decades, however, saw an increasing number of studies focusing on regeneration in natural treefall gaps. This accumulated knowledge will be used here to elucidate the consequences of logging for forest regeneration. The forest mosaic Natural forests are shifting mosaics of forest patches in different stages of development, locked in a constant cycle of death and renewal by the process of tree falls (Hallé et al. 1978). Such tree-by-tree forest regeneration was described for the first time by Aubréville (1938) in the forests of Côte d’Ivoire. A tree fall is the first phase in a process called the forest growth or silvigenetic cycle (Watt 1947; Whitmore 1975, 1978; Oldeman 1974a, 1978; Hartshorn 1978, 1980; Brünig 1983) in which several development phases can be distinguished: i.e. gap, building and mature phases (Whitmore 1975) or innovation, aggradation, steady-state and degradation phases (Oldeman 1978, 1989; see also Mayer 1980; Bormann & Likens 1979; Torquebiau 1981, 1986a, 1986b). Gap formation is widely recognised as the driving force behind the establishment and growth of rain forest seedlings (Denslow 1980; Hartshorn 1980; Rollet 1983; Brokaw 1985a; Pickett & White 1985). Hartshorn (1978) suggests that some 75% of the tree species at La Selva, Costa Rica, are dependent on canopy openings for seed germination or for growth beyond the sapling size. Gap size definition Gap size is commonly measured as the area between the edges of the crowns of peripheral trees projected vertically down to ground level. At least five different gap definitions have been proposed (Runkle 1981, 1982; Brokaw 1982a, 1982b; Riéra 1982; Popma et al. 1988; Van der Meer & Bongers 1996a), of which Brokaw’s definition is the most widely used. He defined a gap as ‘a ‘hole’ in the forest extending through all levels down to an average height of 2 m above ground. The sides of forest openings are irregular in profile, but, for a workable definition, the sides of that space defined as a gap are vertical.’ However, as often as it has been used, just as often has it been criticised; on both practically and theoretically grounds (e.g. Oldeman 1983, 1989; Popma et al. 1988; Brown 1993; Whitmore et al. 1993; Whitmore 1996). First, in practice it is difficult to perform the vertical projection precisely. Secondly, geometric measures of gap size cannot allow for holes in the canopy other than the main gap (Van der Meer et al. 1994, Whitmore 1996), which become increasingly important sources of radiation as gaps get smaller (Mitchell & Whitmore 1993; Walter & Torquebiau 1997). Thirdly, the size of a gap at two meters high is often much smaller than when measured higher up in the canopy (Hubbell & Foster 1986a; Van der Meer & Bongers 1996b), and excludes ‘upper-level gaps’, that can totally illuminate the crowns of pole sized trees (Clark & Clark 1992). Finally, and most important, as indicated by several authors (e.g. Canham 1988; Popma et al. 1988; Koop 1989; Oldeman, 1989, 1990; Canham et al. 1990; Brown 1993; Whitmore et al. 21 Reduced impact logging in the tropical rain forest of Guyana 1993), the micro-climatic effects of a canopy gap and hence its effects on plant growth are not restricted to that area immediately beneath the hole in the canopy. These changes are not only dependent on the size of the hole in the canopy, but also on gap shape, orientation, and height of the surrounding forest in as much as these factors determine the daily duration of direct insolation (Mitchell & Whitmore 1993). A measure of gap size therefore needs to reflect the magnitude of the change in radiation rather than the strictly geometric perceptions of gap limits. Notwithstanding the fact that direct vertical illumination in gaps is important for those plants exposed to it, side light also has great importance for seedling growth as has been demonstrated by Oberbauer et al. (1988), Koop (1989) and Clark & Clark (1992). A more realistic estimate of the light environment may be the area in ‘expanded gaps’; i.e. the area of the surrounding undergrowth affected by, but not directly under, the canopy opening. Runkle (1981) defined a gap as ‘the ground area under a canopy opening extending to the bases of the canopy trees surrounding the opening.’ A gap measured according to Runkle’s definition indeed coincides more strongly with the area at the forest floor with an enhanced light level compared to the surrounding forest than according to Brokaw’s definition (Van der Meer 1995). Gap versus non-gap dichotomy The division of forests into gaps and closed canopy is not sharp (‘forests are not just Swiss cheese’ - Liebermann et al. 1989). It has been shown by geometric gap simulations (Canham 1988a/b, Canham et al. 1990; Ek 1997) and by field measurements (Popma et al. 1988; Brown 1993; Rich et al. 1993; Walter & Torquebiau 1997; Poorter 1998) that there is a continuum of change from closed forest to large gaps in terms of light environment. The central core of the gap receives highest total radiation and therefore exhibits greatest contrasts with the surrounding forest. Low solar altitude allows direct solar radiation to penetrate obliquely under the closed forest canopy surrounding a gap. Hence, there is a ‘penumbra’ of progressively less extreme light environments around a central core of the canopy gap. The gradient away from the central core of the gap also involves micro-climatic parameters that depend on incident radiation, such as amount of PAR, maximum air temperature, relative humidity, as well as light quality as measured by the ratio of red to far-red wavelengths (e.g. Brown 1993). Increasing gap size increases the extreme values of micro-climatic parameters and the size of the ‘penumbral’ influence of a gap. Large gaps have a wide area around their edges with micro-climates similar to those found in the centre of small gaps (Canham et al. 1990). Brown (1993) found the influence of a large gap (1000 m2) to extend 30 m beyond its perceived edge, while Raich (1989) found elevated PAR levels for at least 20 m into the forest. It is therefore suggested by many researchers that the interest should focus on light availability or canopy closure rather than in terms of characteristics and size of gaps, and, accordingly, alternative approaches have been proposed (e.g. Hubbell & Foster 1983, 1986a/b; Lieberman et al. 1989; Whitmore et al. 1993) However, by focusing on 22 Chapter 2: Tropical rain forest logging and silviculture light levels alone and studying responses to changes in the canopy this way, the link with the silvigenetic cycle is lost. In order to counterbalance this loss, multiple approaches are proposed. Those include a crown illumination index as well as a classification of the forest development phase (Koop 1989; Clark & Clark 1992; Zagt 1997). Classification of tree species Tropical tree species have been classified in various ways into guilds, strategists or functional groups. The underlying idea to propose those groups is that species are niche-segregated (e.g. by specialisation on a particular gap size or zone: Hartshorn 1978, 1980; Denslow 1980; Bongers & Popma 1988). Van Steenis (1958) was the first to classify tropical tree species, distinguishing between nomad and dryad species. In a landmark publication Swaine & Whitmore (1988) proposed that, on the basis of seed germination requirements, two ecological groups of tree species can be distinguished: the by now familiar paradigm of pioneer (or shade-intolerant) species, needing full light for germination, and non-pioneers (or shade-tolerant) that can germinate under a closed forest canopy. The dichotomy was rejected by Martínez-Ramos et al. (1989), Raich & Gong (1990) and Schupp et al. (1989) because it neglects influence of differential dispersal and survivorship. Although pioneers germinate best in large gaps, low levels of germination can occur even in closed forest (e.g. Florence 1981; Popma et al. 1988), but germination is soon followed by death (Kennedy & Swaine 1992; Vázquez-Yanes & Orozco-Segovia 1984). In 1989, Whitmore restated the paradigm; defining pioneers on their requirement for full light for both germination and establishment. A search for further subdivision of these two clear but somewhat broad groups continued with increased vivacity (e.g. Brokaw 1985a; Hubbell & Foster 1986b; Denslow 1987; Popma & Bongers 1988; Alexandre 1989; Welden et al. 1991; Oldeman & Van Dijk 1991; Clark & Clark 1992; Hawthorne 1993; Lieberman et al. 1995; Denslow 1996). Swaine & Whitmore’s two guilds can be recognised in most classifications, but, in fact, a wide range of seedling shade tolerances exists within the climax guild species (Whitmore 1989, 1996). Gap size and within-gap zone specialisation The view that there is a near continuum of adaptive strategies in tropical tree species led to the hypothesis that different species are preferentially adapted to gaps of particular size ranges (Denslow 1980; Pickett 1983; Augspurger 1984). Except for common visual observation by everybody who knows tropical rain forests, there is also formal evidence that pioneer species occur in greater density in large gaps (Kramer 1933; Richards 1952; Schulz 1960; Whitmore 1975; Hartshorn 1978; Barton 1984; Brokaw 1985b; Brokaw & Scheiner 1989). Reported threshold gap sizes for pioneers, however, vary greatly. Gap partitioning has been doubted by Kennedy & Swaine (1992) and Brown & Whitmore (1992). They found that rather than emergence of seedlings from the seed 23 Reduced impact logging in the tropical rain forest of Guyana bank, an increase in height growth and decline in mortality rate of pre-existing seedlings and saplings occurred after gap formation. Similar observations were made earlier by Schulz (1960), Rollet (1983) and Uhl et al. (1988) and later by Van der Meer et al. (1998). Whitmore (1996) attributed the abundance and greater densities of seed bank species in the studies of Brokaw (1982b, 1985b) and Denslow (1980) to the closeness of the sites to sources of pioneer seed. The French term chablis for a forest canopy gap reflects the considerable internal heterogeneity that exists inside a tree fall gap. Where a tree is uprooted the forest floor is disturbed and hollows and root floors are formed. The litter layer is disturbed and bare mineral soil is exposed. Differences in the level of destruction of pre-existing vegetation and concentrations of debris also lead to specific micro-environments. Oldeman (1978) and Orians (1982) postulated that these specific environments would favour the establishment of adapted species. This was demonstrated by Barton (1984), who showed that among a selected group of species, pioneers tend to be found in the middle of large gaps, and by Baur (1964) and Florence (1981) who observed middleto-edge transitions from pioneer to shade-tolerant individuals in large gaps. Brandani et al. (1988) found root, bole and crown zone to be more similar in seedling species composition than other zones of the same gap. Preference by pioneer species for root mounds was shown by Putz (1983) and Riéra (1985). A gap may thus provide as many sites for the regeneration of shade-tolerant seedlings as it does for light-demanding species, indicating that within-gap heterogeneity allows a wide range of species to be successful in any size of gap (Poorter 1998). Seedling release There is unmistakable evidence that in most gaps an increase in height growth and a decline in mortality of pre-existing seedlings and saplings play an important role in ‘gap-filling’, although sprouts of damaged trees (Putz & Brokaw 1989; Rijks et al. 1998) and lateral growth of surrounding trees should not be omitted (Uhl et al. 1988; Van der Meer & Bongers 1996b). Brown & Whitmore (1982) found that growth and survival of seedlings were most importantly affected by the seedling height at the time of gap formation, and not by the location in the gap or gap size. Similar results were obtained by Zagt (1997) in Guyana. Changes in height-rank through time may occur and greater height growth of light-demanding species may manifest itself in time, enabling them to overtake taller pre-existing shade-tolerant seedlings or saplings (Whitmore & Brown 1996; Zagt & Werger 1998). A seedling of a shade-tolerant species may then have to go through several cycles of release and stasis before it reaches the canopy (Canham 1989; Brown 1996; Zagt 1997). Application of natural gap dynamics in selective logging Knowledge about the various micro-habitat requirements is non-existent for most commercial species. Neither is much known about the growth rate of their seedlings 24 Chapter 2: Tropical rain forest logging and silviculture nor the ability to respond repeatedly to release from shade suppression until they attain maturity. Predicting the future composition of selectively logged forest is therefore not very easy. However, there are clear indications about the restrictions that should be imposed on a logging operation. There is unmistakable evidence that (undesirable) pioneer species will colonise gaps once they are overly large, whereas small gaps are often closed by lateral expansion of surrounding trees. Moreover, it has been shown that pre-existing seedlings play an essential role in the closure of canopy openings. It is therefore of paramount importance that pre-existing seedlings in man-made gaps should be left intact, implying that restrictions should be imposed on movements by heavy machinery in felling gaps. 2.3 THE HISTORY OF TROPICAL RAIN FOREST SILVICULTURE Commercial timber extraction in the humid tropics started in the seventeenth century. The first attempt to control timber extraction in the tropical moist forest dates back to 1770 when a minimum felling diameter was introduced for the extraction of Teak in Java. In 1777, instructions were added there to spread the felling and use the largest trees first. Also, many closures were ordered in worked-out forest and the first felling cycle was born in the tropics (Boomgaard 1988). The first efforts to bring tropical forests under scientific management were the ‘Improvement Fellings’ and ‘Selection Systems’ of Burma and India. These systems were mostly of European origin, based on innovations introduced by the German botanist Sir Dr. Dietrich Brandis who aimed to implement the Central European Selection System (G.: Plenterung or Plenterwald, Fr.: Jardinage par pieds d’arbres or Futaie jardinée - Dawkins 1958; Neil 1981). The basis of the concept is to change the initial ‘stand table’ (that is, the frequency in each diameter class of the untouched forest) into a ‘normal forest’ (balanced forest; see Oldeman 1990 p. 394-396), a forest in which the growing stock is comprised of trees of every age (and size) class represented in proportions that permit the harvesting in perpetuity of a constant volume (and number) of both mature trees and thinnings. The frequency of each diameter class has to be underwritten by a larger frequency in the class below to allow for mortality. The graphical form of the normal stand table takes a negative exponential distribution, which is known as the ‘de Liocourt curve’ or ‘reverse J-shape curve’. Of course, this concept implies stability in all aspects of the environment and the normal stand table is rarely, if ever, achieved. The system that Brandis worked out for the Teak forests of lower Burma is excellently described by Dawkins & Philip (1998). The system included a prescribed felling cycle and an allowable annual yield, with several stipulations; the most important - and most often forgotten when Brandis-type management spread in to the equatorial tropics - being to reserve seed bearers when nearby juveniles were lacking. The Brandis method inspired nearly all early efforts at tropical moist forest silviculture throughout the tropical world. The Brandis method has often been called a ‘selection’ system, but this term, transferred from European theory and practice, is widely misapplied and the method is better termed a diameter limit system (Dawkins 1958, 25 Reduced impact logging in the tropical rain forest of Guyana 1961; Neil 1981; Jonkers 1987). The theory of the selection system sensu stricto demands that trees are felled one by one from the whole forest area when they individually reach their maximum value, or when they begin to inhibit neighbours of potentially greater value (Leibundgut 1984; Bernasconi 1998). Successful application demands that regeneration should occur throughout the forest. The applicability of the system rests on a high proportion of shade bearers, which is linked to the region of origin, the Fagus - Picea - Abies mixed forests in Central Europe (but see Kuper 1994). The system also requires nearby markets for both large and small produce and a large number of skilled foresters. Clearly, the conditions faced by Brandis were quite different. He was mainly concerned with a single species in poly-specific forests. The system was not entirely suited to teak forest, because teak is a light-demanding pioneer, a temperament that has no counterpart in Central European Selection Silviculture. Moreover, markets and proper supervision were not as widely available as in Central Europe. When the method was adopted in the equatorial forests, more often than not with a multitude of target species with different temperaments and hence varying stand tables, the concept of a selection system sensu stricto appeared to be even less applicable. At the beginning of the twentieth century, attempts at assuring future timber supplies were extending to other parts of the world. This was the start of modern tropical rain forest silviculture in areas such as Malaysia, Africa, Australia and the Caribbean. The minimum diameter limit for felling appears to be the only aspect of the Brandis method that survived its author. Although the application of minimum diameter limits to felling has been considered a practical approach to selection, its application has been criticised widely (Neil 1981). Wadsworth (1987), for instance, argued that unless accompanied by silvicultural treatments to favour immature trees of valuable species, reliance on minimum diameter limits is counterproductive in the long run, because in forests where the valuable species are in the minority, the overstorey left is virtually without trees of species that would be valuable for future harvests or as sources of seeds. Economic impoverishment of the stand inevitably follows. It also has been repeatedly argued that, because tree selection is based on the choice of the logger rather than of the forester, the system incites negative selection (Champion 1936; Jonkers 1987). Another important argument against applying a diameter limit without distinction by species is that different species display different growth and regeneration characteristics, hence display different natural stand tables (Oldeman, pers.comm.), and differences in diameter at which the stems will begin to depreciate - that is e.g. the diameter at which the mean annual increment culminates - (but see also Vooren 1992 and Kuper 1994). The opposite of the selection system is the uniform or monocyclic system, in which the trees that constitute the forest are in even-aged blocks. In this case, the ‘normal forest’ is made up of blocks of each age class that are either of equal area or of equal productivity. In theory, the individual blocks differ from each other by 1 year of age and one is felled and regenerated every year, so as to obtain a continuous, uniform flow of timber. Artificial regeneration by clear-felling and replanting is the ancestral uniform system. The uniform system using natural regeneration refers to regenerating 26 Chapter 2: Tropical rain forest logging and silviculture with naturally present seedlings; e.g. the Malaysian Uniform System (e.g. Wyatt-Smith et al. 1963). The uniform system and the selection system stand at the two extremes of the range of silvicultural systems; the former oriented towards ‘stand-wise’ silviculture and the latter towards ‘tree-wise’ silviculture. Between these extremes many forms occur, such as group selection, shelterwood and the polycyclic systems. The chronology of the application of silvicultural systems in different parts of the world has been comprehensively summarised by Baur (1964), Neil (1981), Jonkers (1987), Schmidt (1987), Wadsworth (1997) and Dawkins & Philip (1998). Efforts in francophone Africa, French Guiana and Kalimantan after 1976 were not covered by these authors, except by Schmidt and Wadsworth, but recently several publications appeared on the trials initiated with the aid of CIRAD-forêt (Catinot 1997; Favrichon 1997; Dupuy et al. 1998; Nguyen-Thé et al. 1999) and on the Indonesian successes using mycorrhizal inoculation (Smits 1994; Yasman 1995). A brief history of the major social and economic developments that steered the direction of tropical silviculture and management is given below. The period 1900-1950 saw a rapid expansion of tropical silviculture and management, radiating from India, Burma and Indonesia throughout the colonial territories of the European powers. This expansion was mirrored by an expansion in trade and timber exports to Europe and North America. Apart from the geographical spread, there were few developments. In most countries, the example of Brandis was followed. A radical change occurred in the years that followed. The age of European colonial powers, such as Britain, France, Belgium and the Netherlands came to an end. This initiated a series of effects that had implications for forest management, among others a loss of continuity, experience and skills. Bilateral and multilateral aid projects offered a makeshift substitution for this loss, but these were of limited duration and were not aimed at support for continuous development of forest management practices. Moreover, the period 1950-1990 saw a rapid expansion of the world trade of wood products, a rise in the local demand for timber, the development of plywood and other panel products and last but not least the introduction of the chainsaw, the crawler tractor, articulated wheeled skidder and heavy duty truck-and-trailers for timber harvesting. Because of these developments, the intensity of logging rose tremendously with strong implications for the ambitions of silviculture and management. The 1980s saw the growth of the conservation movement, the acknowledgement of the effects of rain forest destruction on the global environment and an awakening desire to conserve the biodiversity of the rain forest. This shifted the attention - especially of aid projects from timber production to other forest values, also changing the goals of forest management. The development of most silvicultural systems thus was hardly ever based on silvicultural or biological factors, but rather on industrial and social changes. For instance, the development of uniform systems, such as the Malaysian Uniform System and the (classical) Tropical Shelterwood System, during and after the Second World War, was strongly related to the greater demand for a wider variety of species and the 27 Reduced impact logging in the tropical rain forest of Guyana impetus of mechanisation in extraction and milling practice, which favoured a single, heavy exploitation (Lowe 1978; Lee 1982a). The abandonment of these systems was not at all provoked by ecological factors but by increased land pressure for agricultural development in Peninsular Malaysia, or shortage of funds and staff and later increased population pressure after independence in Nigeria. In Sabah the (modified) Malaysian Uniform System was abandoned in 1977, owing to an increased intensity of logging that affected soils and regeneration so seriously that treatments were considered to be no longer appropriate (Lee 1982b). Polycyclic silvicultural systems, such as the ‘selective management system’ of Ghana (Asabere 1987) survived longer. Its demise was also caused by an increased pressure from the forest industry, looking for higher yields and shorter cutting cycles. The CELOS management system in Suriname was not lifted from an experimental to a practical level following a coup d’état in 1983 (De Graaf 1988). The often heard sweeping generalisation that natural regeneration systems have been abandoned because of lack of success thus in reality covers a multitude of problems, which is eloquently described by e.g. Fox (1976) and more recently by Clément (1997) 2.4 LOGGING, FOREST MANAGEMENT AND SILVICULTURE IN GUYANA Logging has been conducted in Guyana for over 200 years (Vieira 1980). During most of this time logging involved manual overland extraction within a narrow strip of land along the banks of Guyana’s many rivers. Mechanisation of logging first appeared in the 1920s with the introduction of motorised sleigh-winches, which were soon superseded by wheeled agricultural tractors. It was not until 1967, much later than in Africa and South East Asia, that the use of chainsaws for felling and articulated rubber tyred skidders for extraction were first introduced in Guyana (Vieira 1980). Uncontrolled, selective logging was carried out until 1871 when the first attempt was made to regularise cutting. A major development occurred in 1925 when a forest department was established. This resulted in the first systematic forest inventories and the first silvicultural trials. Numerous developments have occurred in forest management since 1871; i.e. several revisions of forest legislation, the establishment of a permanent forest estate and recently the emergence of a code of practice. The history of logging, management and silviculture is reviewed to show how the debate on the sustainability of the current logging practice, which is still mainly ‘high-grading’ and can be characterised as a ‘log-and-leave’ form of forest use, is nothing recent, but has been a recurrent cry in the Greenheart history of the country. 2.4.1 Forest utilisation before 1890 The oldest record of timber trade in Guyana dates from as early as 1624. In the early days of Dutch settlement in Guyana, one of the most important activities was the bartering of trade goods with the Amerindians, one of these goods being Piratinera guianensis (syn. Brosimum guyanense - Letterwood). The trade in Letterwood must 28 Chapter 2: Tropical rain forest logging and silviculture have flourished because in 1669 a single ship is reported to have transported 10,000 kg of the timber (Swabey 1950). By the middle of the 18th century, the granting of permissions to extract timber by the Dutch West India Company was established. These restrictions were not sparked by a notion that cutting should be controlled, but rather to prevent the dissipation of limited resources of slave labour by granting timber concessions on the ‘far’ rivers, which was presumed to weaken the defences of the settlement of the Demerara and Essequibo rivers (Swabey 1950). Letterwood dominated the timber trade until the latter part of the 18th century, when Chlorocardium rodiei (Greenheart) was first produced in commercial quantities. Most timber came from the old ‘Dutch’ estates, which, being private property, resulted in a lack of control of woodcutting and, moreover, no direct return to the colonial power, which was by then British (McTurk et al. 1882). This led to the first attempt to regularise woodcutting under the Crown Lands Ordinance of 1871, which provided that rent was payable on woodcutting tracts. Sawmilling also commenced about this time and, in 1880, there were two steam sawmills in Georgetown and one waterpower mill at Christianburg. At that time, logging was restricted to the lower rivers. Strenuous manual overland extraction was undertaken within a narrow strip of land about 1 km deep inland along the banks of the rivers. A system of rollers was adopted to reduce the resistance of logs being dragged over the ground. As the log advanced, the last roller, freed at the rear end, was taken forward and used again at the front end. Movements of logs in this way was carried out by workers pushing from the rear of the log by means of levers of suitably sized poles (locally named the ‘grail-stick’ method - Vieira 1980). During this period, for the first time, there was uneasiness at the destructive methods of the woodcutters; McTurk et al. (1882) remarked: ‘The necessity for the conservance of the forests has long been apparent to many who have had the opportunity of seeing the reckless manner in which the forest is being denuded of its most valuable trees. No restrictions as to size, time of cutting or kind to be cut are placed on the licensed wood-cutter.’ This and other comments culminated in the passing of the Crown Lands Ordinance of 1886. During this time, it was remarked for the first time (report of the Crown Surveyor 1889 - ex Swabey 1950) that: ‘the number of reefs of Greenheart within easy reach of water carriage are yearly becoming more and more scarce and further away from navigable watercourses. In no very remote time the stock of timber will be exhausted… I am convinced that some system of forest conservation should be established.’ The new Regulations of 1890 featured the following measures: the institution of minimum cutting limits, the obligation to retain economic species spaced throughout the forest, the payment of royalty, the institution of grant registers and removal permits and the marking out of working blocks. Otherwise during the 1880s, the coastal population in the interior began to increase due to gold mining and a growing trade in Balata - the coagulated latex obtained from tapping Manilkara bidentata (Bulletwood). During the later part of the 19th century and the earlier part of the 20th century, firewood and charcoal production, mainly for 29 Reduced impact logging in the tropical rain forest of Guyana domestic use and export to the Caribbean, and Balata bleeding outstripped timber production. 2.4.2 The period 1890-1954 An important step in the recognition of forest management problems was the establishment of a forestry branch at the Department of Lands and Mines in 1908 and the appointment in 1910 of five forest rangers and a forest officer. During the next 15 years the first forest surveys were undertaken by C.W. Anderson, the first forest officer and his successor L.S. Hohenkerk. Between 1908 and 1916 a total of 34 expeditions were made covering all of the easily accessible forest areas (Welch et al. 1975). Up until 1919 there were no leases, timber, firewood, charcoal and Balata being exploited under a licensing system. A revision of the Crown Lands Regulations made the issuing of woodcutting leases possible. These amended Regulations were primarily designed to safeguard revenue and there was little attempt to control woodcutting operations (Swabey 1950). With the increase in volume and complexity of duties regarding the forests and the forest industry - by this time the number of steam sawmills had increased to 20 - it became obvious that more staff was needed. Hohenkerk (1923, p. 5) commented again on the reckless method of logging -especially on the large number of poles that were cut for rollers - and was one of the first to contest the practice of high-grading (‘creaming’): ‘The cutting of Greenheart is purely selective, both as regards kind and size of tree felled. All charges for prospecting, licensing and logging are borne solely by this timber, with the result that the cost of marketing is excessive.…The trees are generally cut high on account of root-buttresses or holes in the trunk and great waste occurs through this and also through defective and careless logging methods. On grants already cut over, numerous trees have been left which, although of medium to small sizes, could profitably have been worked for the local market….’ In 1925 a Forest Department, independent of the Lands and Mines Department was established, with B.R. Wood as the first Conservator of Forests. During the next 25 years the Forest Department undertook forest inventories of the most important forest areas by strip enumeration surveys, identified and took samples of hundreds of tree species, prepared and distributed samples of timbers for testing, started investigations in the regeneration of Greenheart forests (see below) and established seasoning techniques for local timbers. By 1935, the distribution of the main commercial species and in particular Greenheart was well understood. During the period from 1919 to 1953, there was virtually no change in the methods of control of woodcutting operations. However, drastic changes had taken place in the timber industry. With the introduction of kerosene and other petroleum-based fuels, firewood and charcoal production dropped and was surpassed by timber production in the late 1930s (Hammond et al., in prep.). Conditions arising from the first world war induced an increase in the demand for local sawn timber and a consequent increase in sawmills, whereas the economic depression in the late 1920s and early 1930s brought in lumber at low prices from abroad and the local industry declined again. By 1939, 30 Chapter 2: Tropical rain forest logging and silviculture however, production had improved and there were 30 mills operating. The Second World War affected the supply of mechanical equipment on which the industry depended by that time. Timbers other than Greenheart - especially Mora excelsa (Mora) and Carapa guianensis (Crabwood) - were available adjacent to rivers and could be extracted with little or no mechanical equipment and an increase in the production of these species was seen. Sawmills to deal with this production sprang up everywhere, and by 1949 their number had increased to 79 (Swabey 1950). Before the war, Greenheart had a prominent position with a log output of 20,000 m3 per year, accounting for 77% of the total roundwood output in the 1920s, rising to 80% in 1939. During the war the production of timber other than Greenheart increased from 5,000 m3 in 1939 to 57,000 m3 in 1949, reducing to 39% the share of Greenheart in total roundwood output. The Golden Age of Greenheart was, however, still to come. Just after the war, a Developmental Committee was set up to consider the allocation of development aid provided by the British government. A forestry sub-committee was appointed with the task to 1) frame a forest policy, 2) estimate future forest production targets and 3) draw up specific development projects. C. Swabey, who succeeded B.R. Wood as Conservator, strongly urged that the forest policy should have two basic aims namely increasing production in order to fill local as well as export demands; and managing the forests on the basis of a sustained yield concept. A third major aim that was included in the final report of the forestry sub-committee was that the country should obtain a reasonable, direct return from forest exploitation by means of royalties (Welch et al. 1975). In 1948, the Governor presented a far-reaching statement on forest policy to the legislative council: the forests should be worked in such a way that they will be permanently productive and not be ‘mined’ as for a mineral. Swabey submitted draft amendments to the Forest Ordinance and the Regulations to go along with it. The main features were (Welch et al. 1975, p.50): ƒ ‘exploiters were required to submit logging programmes to work their leases in a sequence of contiguous blocks; ƒ all merchantable timber should be worked out from one block before exploitation extended into the next; ƒ seed bearers were to be retained for silvicultural purposes; ƒ worked out blocks were to be surrendered to the department for regeneration or improvement work to be undertaken.’ However, these proposed amendments and Regulations appeared to be ultra vires, because they conflicted with the 1903 Crowns Lands Ordinance. Practically all of the commercial forests occurred on Crown Lands, hence governed by the Crowns Lands Ordinance, and through that Ordinance by the Commissioner of Lands and Mines and the Forest Department had, apparently, from its inception little or no real control over the forests. It was therefore seen that in order to vest the kind of control Swabey wanted in the Forest Department, alterations to basic land laws were required instead of amendments. The thorough legal revision took several years and the new Forest Ordinance and Regulations became law in 1953. 31 Reduced impact logging in the tropical rain forest of Guyana The first silvicultural trials were established at Aruka in the North West district, where exotic Khaya ivorensis and Tectona grandis were planted in 1926, but these trials were soon afterwards abandoned. In 1931 planting experiments continued at the new headquarters of the Forestry Department at the Penal Settlement on the Mazaruni river. Seeds of exotic species, such as Swietenia macrophylla, S. mahagoni, Cedrela mexicana and Tectona grandis, as well as indigenous species, such as Hymenaea courbaril (Locust), Virola michelli (Hill Dalli), Peltogyne venosa (Purpleheart), Diplotropis purpurea (Tatabu) and Dipteryx odorata (Tonka Bean), were sown. Simultaneously, experiments concentrating on the natural regeneration of Chlorocardium rodiei (Greenheart) were started. The first operations consisted of underbrushing, removing all undergrowth competing with Greenheart seedlings and saplings, in lightly creamed forest near the Mazaruni Station and were probably inspired by the Malayan Regeneration Improvement Felling System (e.g. Wyatt-Smith et al. 1963). In 1935, T.A.W. Davis embarked on a series of experiments in order to gain insight into the conditions favourable to the regeneration and growth of Greenheart. Several treatments were compared, including different combinations and levels of thinning of undergrowth, ‘understorey’ and canopy in unexploited Greenheart forest (Clarke 1956). In 1936, casual examination of the first improvement operations indicated that the technique used was successful to a remarkable degree in stimulating regeneration of Greenheart (Welch et al. 1975). It was decided to concentrate silvicultural work on natural regeneration techniques and the plantation trials were stopped around 1939. Before the results from Davis’ experiments were available, routine forest improvement operations were launched. About 240 ha of logged forest near the Mazaruni Station were treated in 1937. Further improvement operations were carried out over selected areas of partially exploited forest on both banks of the Essequibo river. Between 1937 and 1943, a total of 8,750 ha of forest were treated in the Labakabra-Tiger creek and the Moraballi-Seba creek areas (Fanshawe 1944a, 1944b). The treatment consisted of climber cutting and manipulating the canopy over groups of seedlings, saplings or poles by poison girdling large undesirable trees. During 1945-46, some 1,290 ha were treated a second time. The considerable expenditure in time and funds, however, was not considered justifiable until the envisaged systematic block method of logging was implemented. The improvement work was suspended after 1946. 2.4.3 The period 1954-1986 The 1953 Forest Ordinance together with the approved Forest Policy committed the government to create another legal land category - namely the Crown Forests. The year 1954 marked the legal beginning of control of the Forest Department over 7,5 million ha of Crown Lands now declared to be Crown Forests. The year 1954 also marked the beginning of a new Ten Year Development Plan for forestry through grants from the Colonial Development and Welfare Scheme of the British government. Four schemes got off the ground; i.e. training of staff, the establishment of a Central Timber Manufacturing Plant, introduction of aerial photo-interpretation for forest inventory 32 Chapter 2: Tropical rain forest logging and silviculture and a silvicultural programme. The major objectives of the Central Timber Manufacturing Plant were to put new species on the market and so break the popular prejudice for Greenheart and Crabwood lumber, to promote seasoned lumber, and to serve as a central marketing agency for small producers. The silvicultural programme focused on the continuation of natural regeneration in exploited Greenheart forests and on planting experiments with Pinus caribaea on the white sands (see below). In 1954 also new lease agreements under the new Forest Ordinance were concluded for three major firms operating in the Bartica Triangle - British Guiana Timbers Ltd. (owned by the British Colonial Development Corporation), Willems Timber and Trading Co. Ltd. and Charlestown Sawmills Ltd. These lessees were subjected to the new Regulations compelling them to extract all merchantable timber to the satisfaction of the Forest Department and to work a sequence of contiguous blocks, which were to be handed back to the government (Fanshawe ex Clarke 1956). The intention was then to assist natural regeneration of Greenheart or, at least, to increase the stocking of Greenheart substantially by poison-girdling undesired trees and climber cutting, using funds from the Development Programme (Clarke 1956). Two years later, it was noted that the only forest management practised by the Forest Department was the issue of woodcutting leases and permits, and the enforcement of minimum felling diameter limits. On the larger leases a sort of block system was in operation, but there neither was a serious attempt by the Forest Department to check if these were worked according to the Regulations, nor were planned silvicultural operations carried out (Welch et al. 1975). There are two reasons why the planned systematic forest management system failed. The first is that the silvicultural treatment was considered too expensive and too slow to keep pace with the rate of exploitation (see below). The second, maybe more important reason, is that from 1960 the Forest Department assumed full responsibility for the collection and accounting of all revenue from Crown Forests which directed the energies of most of its professional and field staff towards the collection, recording and accounting of revenue, and the tabulation of production statistics. Meanwhile, the above-mentioned three firms were laying the foundations of the Golden Age of Greenheart. This was related to the higher level of capitalisation, vertically integrated operations, the higher level of mechanisation and road access to previously inaccessible areas. The impetus for greater mechanisation came from a British organisation - the Colonial Development Corporation - which invested a huge capital in a large integrated timber industry: British Guiana Timbers Ltd. The magnitude of this enterprise was such that its inception would be seen in later years to have marked the beginning of a new chapter in the history of logging in the country (Vieira 1980). By 1957, there were 104 woodcutting leases with 92 sawmills covering nearly 860,000 ha. The output of timber was 210,000 m3 per year (roundwood equivalent) of which 63% was Greenheart. About 50% of all timber was exported; some 90% of which was Greenheart. These levels of timber output and export were maintained up to 1974. 33 Reduced impact logging in the tropical rain forest of Guyana Early exploitation had taken the form of highly selective felling of choice stems of Greenheart, mainly geared towards the export of roundwood pilings and medium-sized squared timber. Since the new Forest Regulations were introduced and the mechanisation of the timber industry had increased - especially regarding the production of dressed lumber, creaming had become less selective. However, this happened in the sense that almost all the Greenheart was now removed from the felling area rather than the utilisation of other species being increased as intended with the new legislation (Gordon 1961). Clarke in 1956 and Gordon in 1961 stated (again) that supplies of old-growth Greenheart were limited and would not last for much more than 30 years at the rate of exploitation at that time, and that, as the more accessible areas would have been worked out, the cost of obtaining Greenheart was to start rising very soon. Just as at the end of the 19th century, 1920s and 1940s, this fear was allayed by the opening up of virgin forests, which was made possible by the development of new methods of extraction: from manual (‘grail-stick’) haulage to cattle haulage by the end of the 19th century, to steam (and later gasoline) sleigh winches in the 1920s. In the 1930s, the introduction of trucks and semi-trailers, extracting logs up to 20 kilometres had thrown open large areas of forests for logging. In the 1950s, access was further expanded through the introduction of heavy road construction machinery, crawler tractors for stumping and truck-and-trailers for road transportation and since 1967 through the introduction of the chainsaw and articulated, wheeled skidders. Swabey commented already in 1950 that opening virgin forest was not the real answer but that ‘with proper management, we should have been drawing these supplies (i.e. Greenheart pilings), to a considerable extent, from previously worked-over forest areas.’ Apparently his comments did not obtain much hearing, because in 1973 the State Forests (formerly named Crown Forests) were extended with another 1,3 million ha. In addition, in 1973 the involvement of the Colonial Development Corporation in forest exploitation in Guyana came to an end, because of continuing financial loss. In 1966, a Forest Industries Development Programme was launched with the assistance of FAO/UNDP with the aim to carry out a full scale appraisal of the forest resource potential of the country including forest inventories using aerial photography, and utilisation and marketing surveys. Significant achievements were recorded in the field of sawmilling and saw-doctoring, forest inventory, modernising logging equipment and wood preservation by the time the programme was concluded in 1970 (Welch et al. 1975). During the course of its operations, the project discovered and evaluated a large concentration of marketable wood species in the Upper Demerara region (Quicke et al. 1972). Meanwhile, the Government of the Republic of Guyana, that had attained independence in 1966, enunciated the policy that the ownership and control of the natural resources had to be firmly in Guyanese hands and that the government itself would take part directly in the exploitation of forest resources. With financial assistance from the World Bank and the European Community a state-owned company, called Demerara Woods Ltd, was established for that purpose in the before-mentioned Upper Demerara region in the early 1980s, involving a large concession area of 34 Chapter 2: Tropical rain forest logging and silviculture 250,000 ha, a sawmill said to be the largest in South America and a wood gasification plant. From the inception, neither of the latter two functioned properly. Apparently the failure of the Colonial Development Corporation’s venture to achieve integrated forestry development was not appreciated, maybe because the concession was considered to be the richest in Guyana, containing ‘the largest remaining stand of Greenheart’ (Quicke et al. 1972). In the following years adverse economic conditions started to take their toll from the logging industry. The annual roundwood output dropped below 200,000 m3 and shortage of staff and funding started to affect the Forest Department. In 1979, the Guyana Forestry Commission was established to replace the Forest Department. The intention was to finance the Commission using the revenues derived from the harvest of timber, fuelwood and other forest products, but circumstances, due in large part to the decline in the economy, resulted in revenues lower than the cost of its staff and its activities (GNRA 1989). Consequently, these activities were limited to the issue of woodcutting licenses, the occasional adjudication of boundary disputes, and the assessment and collection of revenues. Despite these constraints, the Forest Regulations were amended, creating a new type of harvesting right, the Timber Sales Agreement. Timber Sales Agreements were made available in 1985 and were issued to larger operators. The terms of these agreements required the holder, besides the stipulations of the 1953 Forest Act (formerly Forest Ordinance), to conduct forest inventories and to submit an operating plan for three years’ logging and road construction to the Commissioner for approval. After the new forest Regulations were passed in 1954, silvicultural work aiming at the natural regeneration of exploited Greenheart stands was resumed at Barabara at the right bank of the Mazaruni river (Clarke 1956). The silvicultural prescriptions differed from the earlier work. Treatments of increasing intensity were staged over several years (Prince 1971a, 1973): i.e. 1. removal of undergrowth surrounding Greenheart regeneration in the first year; 2. removal of undergrowth and ‘understorey’ where appropriate in the third year; 3. removal of undergrowth, ‘understorey’ and some canopy where appropriate in the fifth and sixth year; and 4. finally, removal of undergrowth and ‘understorey’ where competing and poison-girdling of all canopy trees other than greenheart in the tenth year. These prescriptions aimed at a gradual conversion to pure stands of Greenheart, and were probably inspired by the Tropical Shelterwood System (e.g. Lowe 1978; Kio et al. 1986). In 1957, improvement operations continued in the Moraballi-Seba creek area, although with a modification of the original aims of 1937. Preference of treatment was still given to Greenheart where it occurred, but instead of promoting regeneration of Greenheart solely, regeneration of all valuable species over as much of the forest as possible was promoted. Between 1957 and 1959, 888 ha were given a first treatment, of which 583 ha were given a second treatment to remove all remaining undesirable 35 Reduced impact logging in the tropical rain forest of Guyana species from the upper canopy. Two permanent sample plots measuring 80 x 80 m2 were laid down. One half of each plot - 40 x 80 m2 - was treated in the way described above and the other half was left untreated. Tallies were made of the species considered merchantable in eleven size classes. The plots were re-tallied in 1970. By means of simple analysis of recruitment and time of passage, Prince (1971b) made an estimate of rotation up to 50 cm dbh. For the fifteen desirable species selected for improvement, he found a rotation of 136 years in the treated and 218 years in the untreated forest. The rate of return on investment was estimated at 2%. Prince stressed that little confidence could be placed on these figures because all fifteen species were lumped due to the small plot size, and because of the nature of the measurements (tallying) which implied that no individual records of growth and mortality were available. In 1963, K.F.S. King laid down two increment plots, the first in exploited forest near the Mazaruni Station, the second in exploited forest at Barabara which was treated as described above. All Greenheart trees of 5 cm dbh and over were measured annually until 1969. Based on these records, Prince (1971a, 1973) estimated a rotation for Greenheart up to 50 cm of 150 years in untreated forest and of 74 years in treated forest. Given a felling limit of 34 cm (legal limit), he suggested a felling cycle of between 80 and 100 years in untreated forest and of 60 years in treated forest (see also Ter Steege 1990). The two studies by Prince showed that the treatments were successful in stimulating growth and survival rates of Greenheart. However, these treatments demanded huge investments in time and funds and Prince (1973) concluded that the marginal return on investment was stretched over such a long period and therefore so low that it was not worthwhile to continue improvement operations along these lines. Other silvicultural investigations during this period focused on Pinus caribaea trials on white sands and on increasing growth of Virola surinamensis (Dalli) in the coastal swamp forest. Interest in the latter species was instilled by a great demand for plywood production in Suriname at the time. These forest types are not the focus of this study and the interested reader is therefore referred to Vieira (1967) and Paul (1977) concerning the former trials and to John (1961) concerning the latter. 2.4.4 The period 1986-1998 In 1986, the Hoyte government reversed Guyana’s economic policy, adopting a policy of economic liberalisation, which radically changed the context in which the timber industry could develop. In 1989, a ‘National Forestry Action Plan’, jointly prepared by the Guyana Natural Resources Agency and the Canadian International Development Agency, was endorsed (GNRA 1989). The plan focused mainly on the shortage of foreign exchange in the local timber industry, the poor capacity of the Guyana Forestry Commission to regulate the industry and the insufficient utilisation of the forest resources. Indeed, the aspect of high-grading was put high on the agenda again. A dual policy to expand logging and to strengthen the Commission - it was estimated that 74 additional professional staff would be needed to effectively monitor both old and new 36 Chapter 2: Tropical rain forest logging and silviculture concessions - was advocated (GNRA 1989). As part of its programme of privatisation, the government sold the state-owned companies, which had been running at a loss, such as Guyana Timbers (the former British Guiana Timbers) and Demerara Woods Ltd., to foreign investors. A new era dawned from 1990 on. Then, in light of the projected decline of timber stocks in Asia and Africa, the global market started to look for South America to fill the gap (Colchester 1994, 1997; Sizer & Rice 1995; Sizer 1996). This led to an increasing number of originally Asian timber companies establishing themselves in Guyana and elsewhere in South America. Probably the most radical change for the timber industry in Guyana was the establishment of a large plywood mill in 1992. From 1994, the documented annual timber production soared from 130,000 m3⋅yr-1 to 420,000 m3⋅yr-1 in 1996, accompanied by a shift from sawn timber to plywood species: the share of plywood species in the total production rose from 3% in 1990 to 53% in 1996. The arrival of foreign-owned companies also led to a recovery of the production of Greenheart. Since its Golden Age during 1954-1975 when annual roundwood output reached highs of 130,000 m3⋅yr-1, its production had receded to about 50,000 m3⋅yr-1 by the end of the last decade, but, since 1993, the annual production has been on the rise again with an average of 68,000 m3⋅yr-1 over the years 1993 to 1996 (Guyana Forestry Commission records). The establishment of a large plywood mill thus shifted the attention towards peeler species, raising the number of readily merchantable species from a handful to about 25. The most important peeler species are Catostemma commune and C. fragrans (Baromalli), Trattinickia demerarae and T. rhoifolia (Ulu) and Alexa imperatricis and A. leiopetala (Haiariballi). During the present decade, the occurrence of several other events gave cause for optimism. The Government of Guyana committed itself firmly to the sustainable utilisation of the national forest resources as stated in the impressive National Development Strategy (Anonymous 1996). Guyana’s membership of ITTO, and being a signatory to the UNCED Rio Declaration bear witness of this commitment. Guyana also ratified the UNCED Convention on Biological Diversity, the Convention on International Trade in Endangered Species of wild Fauna and Flora (CITES) and the Amazon Co-operation Treaty (TCA). The government made impressive advances not only by signing and ratifying treaties, but also in other deeds. In 1989 the excision of 360,000 hectares of rain forest from the State Forest area was announced by the President to be donated to the Commonwealth for research into the conservation and the sustainable and equitable use of tropical rain forests (Kerr 1993) which finally resulted in the official establishment of the Iwokrama International Centre for Rainforest Conservation and Development in 1996. In addition, in 1989, an intergovernmental agreement was signed with the Netherlands, marking the start of the Tropenbos-Guyana Programme; a forest research programme aiming at sound forest management (see Section 1.6). In early 1995, the Government of Guyana declared a three-year moratorium on large-scale logging concessions (Woodcutting leases and Timber Sales Agreements) to allow time for the British 37 Reduced impact logging in the tropical rain forest of Guyana Department for International Development (formerly Overseas Development Administration) to strengthen the Guyana Forestry Commission. The agreement reached with the British agency called for reforms regarding policy, the royalty system and the Forestry Commission’s administrative, management and monitoring functions. During negotiation of the moratorium, it was agreed, however, that the Government of Guyana reserved the right to issue ‘Exploratory Forestry Leases’ as an instrument to keep would-be investors in the forestry sector interested. The first Exploratory Leases will be issued in 1999 (Chandarpal 1999). The Forestry Commission developed a draft Code of Practice for Forest Operations. This Code serves as a set of guidelines and requirements covering all aspects of logging. The draft has recently been tested to assess its commercial feasibility. After further modification, it is expected to become mandatory. Furthermore, a new Environmental Protection Act to establish the Environmental Protection Agency was passed in early 1996 (Sizer 1996) and a project was drafted with the World Bank to plan a programme for a National Protected Area System to be funded by the Global Environment Facility (e.g. Ter Steege 1998). 2.4.5 The future To date, Guyana’s forests have not been actively managed. The log-and-leave principle has been applied so far, made possible by the fact that logging only scratched the surface and because vast areas of untouched forest remained. In 1981, the area of logged forest in Guyana was estimated at 1.4 million ha (FAO 1981). Log production was steady during 1981 to 1992 at an average annual output of 160,000 m3⋅yr-1, but has been rising since; i.e. from 220,000 m3⋅yr-1 in 1993 to 520,000 m3⋅yr-1 in 1997. Assuming that forests were logged at an intensity of 5 m3⋅ha-1 - relative to the gross area, and not the so-called loggable area - the volume output suggests that since 1981 another 770,000 hectares have been logged. This brings the estimated, total logged area to 2.2 million ha. The State Forests where practically all commercial timber exploitation occurs cover 8.9 million hectares of which 6.5 million hectares have been allocated to logging companies (see Section 2.1.3). If extraction continues at the current rate and estimated intensity it will take 40 years before the allocated area will have been logged. A large part of the allocated area is, however, deemed to be unsuitable for timber extraction, because of restricted access due to steep slopes, or because of the occurrence of commercially unattractive forest types, such as dry shrub and dry low forest on white sands, xeric low forest on rocky laterite soils and various types of swamp forest (Fanshawe 1952; Ter Steege et al. 1996). These forest types also occur frequently within the areas that have been logged already, and their occurrence will not bring down the estimated logging intensity of 5 m3⋅ha-1 any further. However, the proportion of steeply dissected areas is larger in the untouched area and it is therefore almost certain that the accessible forest will be used up within less than 40 years. In the past, this has been a recurrent cry as described in the previous paragraphs, which was always answered by opening up virgin forests. The present size of the logging 38 Chapter 2: Tropical rain forest logging and silviculture operations together with annual coupes, which are much larger than before, rules out this escape route. Moreover, low logging intensities and increasingly further distances to markets (ports, sawmills) will inevitably lead to a situation where the cost of logging, especially of transport, will make logging uneconomical. The higher the transport cost the higher the pressure will be to resort to ‘high-grading’ practices. This implies lower logging intensities and the lower the logging intensity, the higher the rate at which allocated areas will be consumed. The fact that, over the years, repeated cries to install sustained yield forest management have remained unanswered in practice, suggests that managing logged forests in order to improve their productivity may have been - rightfully or wrongfully - deemed to be cost-prohibitive so far. However, a switch to active management of logged forests definitely will have to arise in the near future to keep the forest industry alive. An interesting conclusion can be drawn from the developments in the timber industry in Guyana during this century. Repeated calls to implement sustained yield forestry did not find any hearing. Neither did recurrent efforts to shift the attention away from Greenheart to other timbers. Efforts to promote other species started in the 1920s, were repeated in the 1950s, 1960s and 1980s, but sorted little effect. Changes in the market situation, on the other hand, were able to achieve this goal. Evidence is provided by the changes that took place during the Second World War, when supplies of mechanical equipment and imports of timber from abroad had dried up. Another case is the sudden predominance of Catostemma commune and C. fragrans (Baromalli) after Barama Company Ltd. had installed its plywood mill. Also, in the eastern part of the country, where Greenheart does not occur naturally, a thriving timber industry is running on species other than Greenheart - e.g. Goupia glabra, Aspidosperma album and Hymenolobium flavum. The market demand for timber is indeed a crucial issue. The properties of many species are unknown to the trade and to the consumer. Even when known, the available volume may be too small to be readily marketable, and exploiters are naturally reluctant to fell and extract timber unless a market is virtually assured (Wyatt-Smith, 1987). Buyers may even be reluctant to order species unless they are certain of continuity of supply for at least ten years (Asabere, 1987). The question is indeed, how common the other commercial species are; many of which have suitable wood properties (Gérard et al. 1996)? Still, there must be other reasons why promotion of other species has been unsuccessful for so long. Evident explanations are that there always have been sufficient supplies of Greenheart and that local builders (specifications for government buildings often require its use) have grown so accustomed to using it that other species are hardly accepted (e.g. Flemmich 1963). Another question might be: how reliable is the supply of Greenheart compared to other marketable species, particularly from the Greenheart forests? 39 Reduced impact logging in the tropical rain forest of Guyana 40 Chapter 3: Logging methods 3. LOGGING METHODS ‘…il doit s’agir pour le bucheron artisan moins de l’apparence de ce qu’il livre que de l’état de ce qu’il laisse; c’est la forêt sortant de ses mains qui est le tableau, le témoin de son art.’…… ‘Aussi nombreuse que les arbres à couper est la variété des cas qui se présentent. L’affaire n’est pas seulement de faire choir un arbre, ce qui est relativement facile et n’astreint que les muscles; l’affaire est de déterminer et d’orienter la chute de chaque arbre en tenant compte de sa forme, de son juste ou faux équilibre, de la charge inégale des branches, de l’empattement des racines qui créent autant de résistances différentes, de la qualité du bois, des tares éventuelles, du contact des arbres voisins, de la pente et des accidents du terrain, du vent qu’il fait, de la pluie qu’il vient de faire, de la surcharge en neige, du gel, du dégel, etc. Il faut savoir apprécier justement chacune de ces circonstances, en résumer en soi les effets probables, et savoir s’y prendre pour que la rupture de l’équilibre et la chute de l’arbre se déterminent dans la direction choisie, celle qui réunit le moins d’inconvénients pour la forêt qui reste.’…. ‘Il y a des moyens très variés pour obtenir ce résultat final; qu’on ne croie pas qu’il suffise de connaître le maniement de la scien et de la hache, de savoir battre le coin en cadence ou d’appliquer des leveiers et chaînes, ou bien le ‘diable’. Il s’agit de savoir où et comment les appliquer et à quel moment, après une bonne préparation. Il n’y faut ni grand déploiement de force, ni beaucoup de bruit…’ (Biolley 1930, architect of the ‘méthode du control’ in Swiss Selection Silviculture, on the art and necessity of directional felling in early twentieth century Europe). 3.1 THE EMERGENCE OF REDUCED IMPACT LOGGING Before the Second World War, logging was highly selective and mechanisation of harvesting and wood processing was nearly non-existent in most tropical countries. The close of the Second World War marked a time of radical change in forest utilisation resulting from an expansion of world trade of wood products, rises in local demand for sawn timber, advances in timber processing - the development of plywood, veneers and other panel products - and, last but not least, the widespread introduction of the chainsaw, the crawler tractor and heavy duty lorries for timber harvesting. Previously, felling large trees by hand with axe was slow, arduous and unpopular. The introduction of the chainsaw removed this last remaining barrier to the potential destruction of the forest. Fostered by these developments, the intensity of felling rose. It was realised soon afterwards that mechanised selective logging often damaged the residual stand heavily (e.g. Nicholson 1958; Fox 1968a; Chai & Udarbe 1977). However, improved harvesting techniques were only introduced in the early 1980s in Sabah (Mattson Mårn & Jonkers 1981) and Australia (Ward & Kanowksi 1985). In the 41 Reduced impact logging in the tropical rain forest of Guyana neotropics where logging intensities were lower, improved techniques were only introduced in the late 1980s (Suriname - Hendrison 1990). Strengthening of the conservation ideology and its popular success influenced policy makers (see Clément 1997) and, eventually led to two major events, influencing the scene of tropical timber trade and harvesting. In 1990, the International Tropical Timber Organisation (ITTO) targeted the year 2000 as a deadline for making the tropical timber trade sustainable. Then, in 1992, the United Nations Conference on Environment and Development, held in Rio de Janeiro, had as one of the most important outcomes the explicit recognition by policy-makers that forests should be sustained for future generations and that economic and social contributions of forestry should be improved. Fostered by these events, harvesting techniques, that aim at improving standards of utilisation and reducing environmental impacts, were being increasingly advocated. This led an increase in trials to investigate methods to reduce damage during selective logging: e.g. Bertault & Sist (1995) in East Kalimantan, Pinard & Putz (1996) and Cedergren et al. (1995) in Sabah, Johns et al. (1996) and Blate (1997) in Brazil, Webb (1997) in Costa Rica, Van Leersum (in prep.) in Cameroon and the present study in Guyana. The term ‘reduced impact logging’ (RIL) surfaced around the mid 1990s (Pinard et al. 1995), but the concept is also referred to as ‘low impact logging’ (Blate 1997; Holmes et al. 1999), ‘planned’ (as opposed to ‘unplanned’) logging (Johns et al. 1996; Barreto et al. 1998), ‘environmentally sound harvesting’ (Winkler 1997) and ‘damagecontrolled logging’ (Hendrison 1990). There is a need to clarify the substance covered by these terms, because we may be comparing apples and oranges. The adjective ‘reduced’ hints at a comparison with another logging method, which is obviously the current, local practice. The current practice may cover a broad range of methods, varying from ‘hit-and-miss, unplanned’ logging to ‘standard practice’ logging (Van der Hout & Van Leersum 1998). The place a current practice may take on this scale depends largely on the scale and level of capitalisation of the operation. Several elements are common to most RIL systems including the following (ITTO 1990): ƒ pre-harvest inventory and mapping; ƒ pre-harvest planning of roads and skid trails; ƒ pre-harvest climber cutting; ƒ directional felling; ƒ optimum recovery of utilisable timber; ƒ winching of logs to planned skid trails. Some of these elements may be already present in conventional practice; in some cases, they may be rudimentary, in other cases they may be close to ‘standard practice’. These elements are often depicted as being newly developed. However, most of them are common practice in logging in temperate zones (Conway 1982). Although directional felling and proper bucking techniques were already used by lumberjacks in the 17th century, the introduction of the chainsaw required proper sawing techniques to 42 Chapter 3: Logging methods ensure safety of the operator, to achieve high standards of utilisation and to maximise efficiency. Directional felling aims specifically at: 1. ensuring the safety of the felling crew; 2. preventing breakage of the log by felling the tree downhill, in depressions, over ridges or on a solid obstacle such as stumps, rocks or windfalls; 3. preventing damage to the standing commercial timber surrounding the tree to be felled; 4. avoiding excessive crushing of seedlings and saplings of commercially desirable species; and 5. producing a felling pattern geared for the subsequent operation - skidding. The lay-out of skid trails in certain patterns and the application of winching were also developed to increase efficiency and standards of utilisation. It is unknown whether such directives accompanied the introduction of the chainsaw and the frame-steered, wheeled skidder in tropical forestry. However, it is likely that points 3), 4) and 5) received less attention. The practices described by Conway (1982) apply mainly to the Pacific Coast of the United States, a situation where most of the standing timber belongs to commercial species and, consequently, often is a clear-cut. At the time of introduction of the chainsaw on the scene of selective tropical rain forest logging only a few trees were being harvested per hectare and most of the surrounding trees were regarded to be worthless. As far as the three other points are concerned, only rudimentary elements of what once was controlled felling are being practised nowadays, due to the poor educational system of the present shoddy on-the-job apprenticeship in most operations in tropical forests. 3.2 CONVENTIONAL LOGGING IN GUYANA Conventional logging systems are often described as unplanned, haphazard timber harvesting. As mentioned, there is no standard conventional practice at all, but the level of organisation and planning differs substantially between operators. The average logging operation in Guyana can be divided into five phases: 1. opening of a road and log landings by bull-dozer; 2. felling and bucking of trees; 3. skidding of logs to log landings; 4. landing operations, including trimming, measuring and loading of logs; 5. road transport to sawmill gate or riverside landing. Often the steps in this process do not follow this sequence but trees are scouted and successively felled before a road is constructed. Felling and extraction in such operations are inefficient due to a lack of information about terrain conditions and tree location. These features did not apply to the conventional operation studied here. Roads were planned and constructed well in advance of the harvest operation and a pre-harvest inventory was conducted. 43 Reduced impact logging in the tropical rain forest of Guyana During the time that I was monitoring their operation; i.e. 1996, the following pre-harvest activities were carried out. The concession was divided in logging compartments, that measured approximately 10,000 ha - the size of an annual coupe. Logging compartments were subdivided into square blocks measuring 100 ha (1 km2), that were marked on the ground by cut lines. Each block was surveyed by a team of ‘tree spotters’ and trees of commercial species down to a variable diameter limit were mapped along the principles described by Hendrison (1990). This pre-harvest planning phase, however, was considered insufficient. Tree location maps were not accurate, and skid trail alignments were neither marked nor planned. Chainsaw and skidder operators, like most of their colleagues in the tropics, had not received any formal training, but usually had risen through the ranks from assistant (‘choker-man’ or ‘oil-man’) to operator. Nevertheless, they were quite experienced and operated their equipment with great dexterity. During the operation, a logging clerk was present, whose role was restricted to handling the inventory map and taking care of the log flow registration. Loss of utilisable volume caused by overlooked left-behind felled trees was thereby minimised, but the felling and skidding crew still used their own judgement in how to work the area. Roads and landings were constructed before felling. The quality of road alignment and construction and of landing operations was regarded to be adequate and these activities were not considered to be in need of any changes. 3.2.1 Felling The felling process went as follows. The clerk entered the allotted tract of forest and searched the first harvestable stem with the aid of the map. By creating felling chablis at the entrance of the tract, debris was piled up close to trees that were to be felled later and potential skidder paths were often obstructed. The map that was used was not very accurate. Positions of trees were often incorrect and marked trees appeared to be defective on examination by clerk and feller, while some - commercially spoken impeccable trees were left out during the inventory. The tree was examined on defects after which the tree was felled in the direction of the gravity point of the crown, or any other direction convenient to the feller. Felling started by making a deep notch of one-third to one-half of the tree’s diameter at a height of approximately one metre above the ground. The back-cut was often cut at a slant, and thereby, ending too far below or above the notch. The operator had no notion of making a key (see Fig. 3.1) of unsawn wood to steer the tree. Because the tree was felled in the direction of the lean, the tree often started to topple before the back-cut was finished, thereby often splitting the lower part of the log. The feller was aware of the danger of splitting, and tried, without success, to prevent this by continuing to cut while the tree already had started to fall. He was unaware that by doing this he was loosing the bit of control that he had over the direction the tree would fall, thereby aggravating the hazards involved in felling. Often a misjudgement resulted in the cutting of another notch at the opposite side, in a hang-up, or in the guide bar becoming stuck and/or being pinched. Felling height was 44 Chapter 3: Logging methods usually taken above the buttress - sometimes the saw was lifted up to breast height, which wastes timber and which posture forms an occupational health hazard. According to the feller, the reason for sawing at this height was that Chlorocardium rodiei (Greenheart) often shows internal butt rot and that sand on the bark between the buttresses could not be removed easily. Both decay and sand would result in blunting of the chain. Once the tree was down, it would be topped. Then, the butt end was cross-cut; sometimes to remove any decayed bits, but mainly to obtain a clean face for painting a number on the tree. At the log-loading site, the logging supervisor determined the final bole length, allowing only minor defects and most logs were trimmed. While the tree was being felled, the clerk would search for the next tree. The result of this felling method was that the logs were scattered in all directions and in case of Chlorocardium rodiei groves; felled trees ending up in tangles of boles lying about criss-cross. 3.2.2 Skidding The skidder operator searched for logs guided by the openings in the canopy that were formed during felling. Alternatively, his assistant searched for logs while the skidder was making a trip to the landing. The assistant had no proper overview of the stump area and was not able to guide the operator along the shortest route Hence, the search was not carried out in a systematic manner and usually the skidder operator followed in the feller’s footsteps, mirroring the unsystematic search pattern of the feller in winding trails with sharp curves and loops. Logs were not only lying scattered in different directions, they were usually also lying flat on the ground. The large bull-hook with which the skidder was equipped could not pass underneath the bole. Hence, the operator tried to position it in such a way that the hook could pass by lifting the butt end with the dozer blade. As a matter of routine, the operator approached each log just behind the butt end and tried to shove it both in a lifted position and in a direction suitable for skidding. Often he did not succeed at once and the machine had to back off and re-approach the bole from a different angle. During these manoeuvres most of the residual saplings and seedlings around the stump and near the bole would be squashed. Sometimes the log, after all this manoeuvring, would still make an angle with the intended direction of travel. It would then sway before moving in the desired direction, resulting in more damage to the remaining vegetation and sometimes breakage of the log itself. Although the skidder was equipped with a winch, this tool was not used during the log collection phase, except when a log was located at the bottom of a steep slope. In general, the operator would only employ the winch if the condition of main trail had deteriorated to the extent that the wheels were starting to spin when climbing slopes. After positioning the first log, the skidder usually drove to the next log and repeated the positioning procedure. The skidder then drove to the rear end of the log to connect the winch line directly to the butt and dragged it to the first one, where it was unhooked. This procedure was repeated until sufficient logs were collected to compose a load. 45 Reduced impact logging in the tropical rain forest of Guyana The area near the stump of the first tree was thus used to bunch logs. The skidder made a small mound and pit with its blade. After positioning the first log with its butt sticking just over the mound, the skidder would collect up to four logs more and park them near the first log. Choker straps were not used and the main cable was wrapped around the ends of the logs. In order to do this the logs had to be repositioned. First, the skidder would approach the logs from the side and stack them. Next, it would approach the butts and shove them with its blade to get the butts lined up. This procedure was repeated each time that a bundle was composed. Because the skidder usually bunched the logs near the location of the first log of a load, the resulting skid trail was determined by the position of these first logs. When returning from the landing the same trail was used to approach the next set of logs. This explains how the winding trails come into being. 3.3 REDUCED IMPACT LOGGING A reduced impact logging system was designed for selective logging in Guyana. The system is inspired by the CELOS harvesting system (Hendrison 1990) and includes many elements typical of reduced impact logging systems described elsewhere (e.g. Pinard & Putz 1996; Johns et al. 1996; Van der Hout & Van Leersum 1998). It also includes tree selection criteria. Emphasis is placed on the planning of skid trails to reduce skidding damage. Trees are therefore not so much felled in a direction to cause as little damage as possible to the surrounding commercially desirable trees and regeneration, but rather to facilitate smooth execution of the skidding operation, thereby minimising skidding damage. The reduced impact logging system has six basic features: 1. Pre-harvest inventory 2. Climber cutting. 3. Tree selection 4. Skid trail planning 5. Directional felling. 6. Winching. 3.3.1 Pre-harvest inventory In both the conventional and the reduced impact logging system that was studied, pre-harvest inventories were conducted. The critical distinction between the inventories of the two systems lied in a difference in detail and accuracy. As mentioned in section 2.2.3, all trees with a dbh ≥ 20 cm were measured and mapped with 20×20 m2-recording units, in the central 2 hectares of the 6 ha plots of the reduced impact logging experiment. Harvestable trees in the 50 m zone surrounding these 2-ha plots were enumerated and mapped at a scale of 1:1,250. Tree locations were accurately determined by taking polar co-ordinates using a 50×50 m2 grid. 46 Chapter 3: Logging methods 3.3.2 Climber cutting Climbers are an important structural component of the forest canopy, often linking tree crowns. It is generally considered that the presence of climbers contributes a lot to logging damage (e.g. Fox 1968a, Liew 1973, Appanah & Putz 1984, Putz et al. 1984, but also see Jacobs 1976, 1988). By cutting climbers some time prior to exploitation, it may be expected that they will have died and the physical connections binding one crown to another will have weakened considerably. Falling trees may then be less liable to carry down with them their (smaller) neighbours. Similarly, fewer trees will be hung up when severed at the base. Studies on the effect of climber cutting on felling damage by Appanah & Putz (1984) and by Fox (1968a) showed a substantial reduction of felling damage after climber cutting. Putz (1984) found that trees infested by climbers carried down more trees in natural tree falls than climber-free trees. Climbers over 2 cm diameter, that were observed to be growing in or linked to crowns of trees to be harvested, were cut about six months prior to felling; i.e. by spot-wise instead of blanket treatment. 3.3.3 Tree selection criteria Tree selection has received little attention in the literature of Neotropical forestry. Generally, it is suggested that extraction rates are too low to be concerned about tree selection criteria (Lanly 1982; Estève 1983; Jonkers & Hendrison 1987). In Guyana's Greenheart forests, however, high yields can be obtained (Section 2.2.2). One characteristic of Chlorocardium rodiei is of special relevance: i.e. harvestable stems tend to occur in groves (locally named ‘reefs’). The only legal restriction in force is a minimum felling diameter and typically, all sound stems above the felling limit are taken in conventional operations, leading to large felling gaps. Previous Tropenbos research in Guyana (e.g. Brouwer 1996; Ter Steege et al. 1996) has indicated that such large gaps are silviculturally and ecologically undesirable. In order to discourage the formation of such large felling gaps, tree selection aimed at an even spacing of trees to be earmarked for felling. Besides a restriction of the extent of disruption of the forest mosaic, other selection criteria related to the population dynamics and the silviculturally allowable cut of the commercial species can be considered. Specifically, the maximum attainable diameter, reproductive behaviour, dispersal mechanism and the relation between bole diameter on the one hand and degree of maturity, volume increment and vitality on the other hand, could be considered. Little is known about the annual allowable cut per species or per stand. Neither is there sufficient information on the minimum density of seed trees that should be retained per species. Therefore, tree selection criteria were based on different diameter limits for different species according to their specific growth pattern, the abundance of a commercial species in a plot and the distribution of suppressed future crop trees. Ranked in order of priority the selection criteria were: 1. Trees to be felled must be distributed evenly over the area; 2. Trees must be selected according to the relative abundance of the various commercial species in a particular plot (6 ha); 3. A differential felling diameter limit must be applied. The felling limit is set at: 47 Reduced impact logging in the tropical rain forest of Guyana a. 4. 70 cm for species known to attain diameters up to 100 cm and not likely to suffer decay at that size; e.g. Peltogyne venosa (Purpleheart) and Hymenaea courbaril (Locust); b. 60 cm for species known to reach 80 cm and not likely to suffer decay at that size; e.g. Diplotropis purpurea (Tatabu), Aspidosperma album and A. cruentum (Shibadan); c. 45 cm; all other species - including Chlorocardium rodiei (exceeds 80 cm, but is usually defective from 70 cm on); In case these criteria leave a choice between trees, those trees must be selected that suppress immature crop trees. In contrast to conventional logging, where only perfectly sound stems are selected for harvest, trees are eligible for felling as long as they can produce at least one six metre length of sound, straight timber. 3.3.4 Directional felling Directional felling can be carried out for a number of reasons. With a view towards reduced impact logging five objectives of directional felling can be distinguished: 1. to ensure safety of the felling crew; 2. to prevent damage to the bole of the felled tree; 3. to avoid damage to potential and present crop trees; 4. to place logs in a position to facilitate extraction; 5. to fell trees into an existing natural or felling gap, in order to minimise damage to medium-sized and large residual trees. Felling a tree in a particular direction to ensure the safety of the felling crew and to prevent damage to the bole of the felled tree are of overriding importance and are not debated. Concerning the remaining three objectives, it is clear that they cannot all be achieved or optimised simultaneously. The formation of multiple treefall gaps is considered undesirable for the time being because such felling gaps tend to be too large (see above). Some authors stress the potential of directional felling in avoiding damage to potential crop trees (Pinard et al. 1995, Pinard & Putz 1996, Johns et al. 1996). Others stress that the damage resulting from excessive machine movements tends to be more severe (Mattsson-Mårn & Jonkers 1981, Malmer & Grip 1990, Gullison & Hardner 1993, Whitman et al. 1997) whereas reduction of that damage can be attained more easily than reduction of felling damage (Mattsson-Mårn & Jonkers 1981, Cedergren 1996). In the present study, it appeared that avoiding damage to potential crop trees was hard to achieve anyway, because of their high density, some 50 trees/ha, and, moreover because these trees were often clustered around present crop trees. Therefore, it was decided to give priority to the last out of the latter three objectives: i.e. place logs in a suitable position to facilitate extraction. If conditions posed by the safety to the feller and avoidance of damage to the harvested tree are fulfilled, trees are thus felled at obtuse angles with the skid trails (135° to 150°) for ease of skidding. This is the ‘herring-bone system’ (Conway 1982; 48 Chapter 3: Logging methods Hendrison 1990). Some flexibility is allowed. To avoid damage to potential crop trees, shifting the felling direction some 30° from the direction indicated by the ‘herring-bone’ system is tolerated. According to Hendrison (1990), such small deviations do not seriously affect the ease of skidding. If pre-designated felling direction either endangers the feller’s safety, or risks the felled tree to become lodged in a neighbouring tree or to be damaged, it should first be investigated whether the tree can be felled in the opposite direction, to be pulled out by the skidder with the top leading. Felling is executed in a pre-defined sequence, starting at the point furthest from the log market and proceeding towards it in a zigzag course. This way the tree to be felled next is generally free from the debris of trees that are felled earlier, which makes travelling between trees easier. Felling preparations include removal of the underbrush and lianas around the base of the stem, scraping of the stem near the height of felling, removing sand and mosses, cutting of a flight path for the felling crew and determining the felling direction. top cut felling (back) cut notch (undercut) bottom cut key wedge Figure 3.1 felling direction Classic method of directional felling. First, a notch, composed of a bottom and top cut, is sawn. The notch determines the front edge of the key and, therefore, the direction of the fall of a tree without lean and with a regular crown. Secondly, the felling cut is made in the back, opposite to notch and slightly above its level. Thirdly, a wedge is inserted behind the guide bar to keep the kerf open and to bring the tree down after the saw is removed. Classic directional felling techniques are described in Conway (1982), Hendrison (1990) and Klasson & Cedergren (1996). The classic directional felling technique includes a properly sawn notch and appropriate size and shape of the key. Wedges are inserted behind the saw's guide bar to keep the kerf open (Fig. 3.1). The notch should be deep enough to make the key sufficiently long to act as a strong hinge. When the depth of the notch is approximately 1/5 of the tree's diameter, the length of the key will be approximately 4/5 of the diameter, which normally is sufficient. Making the notch deeper will affect the leverage that can be obtained with wedging. The shorter the notch, the shorter will be the distance between the key and the point at which the 49 Reduced impact logging in the tropical rain forest of Guyana felling wedge is applied. This reduces the total leverage; thereby increasing the force needed to bring the tree down. For the same reason, the felling wedge should be applied directly in line with the centre of the notch. The direction of fall is determined by the front end of the key or, in other words, by the way the notch is cut. Consequently, if the notch is oriented incorrectly, this cannot be compensated for by leaving one end of the key thicker than the other end. If a correct key is left, the tree will fall perpendicular to the notch and will guide the tree in the desired direction throughout its fall - as if it were a hinge -, provided that the notch is sufficiently open, to prevent pinching. direction of fall key thicker opposite lean lean Figure 3.2 Felling of trees leaning to one side makes use of a ‘tapered hinge’. The felling cut is started on the side towards which the tree is leaning, to prevent the saw from jamming. A key is left that is thicker opposite to the lean to pull the tree in the desired direction. The shape of the key is determined by the angle between the lean and the desired direction. Only few trees tend to fall naturally in the desired direction. Natural inclinations can be categorised in three groups: trees leaning backwards, forwards or sideways. It is generally suggested that with the classic technique it is possible to fell a tree up to 45° from the direction of a not overly heavy lean, and that with special sawing techniques this can be extended to 90° (Cedergren 1996, Hendrison 1990). When the tree leans to the side the key is left extra thick opposite to the lean to ‘pull’ the tree in the desired direction (Fig. 3.2). The classic technique, however, is not geared to handle trees that are leaning backwards, because the tree tends to ‘sit back’ over the cut. This occurs especially with thin trees, because there is no opportunity to insert a lever behind the saw. Also a forward inclination causes a problem, because the tree may start to topple before the felling cut is finished, leaving a too thick key, which often results in splitting of the butt. In the present study, we used an innovative technique described in Brunberg et al. (1984). The essential improvement of the technique is that a small corner of the felling cut is left unsawn (Fig. 3.3). This prevents the tree form settling back on the guide bar when leaning backwards and from falling before the felling cut is finished when leaning forwards. The first steps do not differ from the classic approach. After the feller has determined the felling direction, he starts to cut the oblique top cut of the notch, standing with his shoulder to the tree, thus facing the felling direction. He checks if the bottom of the notch has become horizontal by peering into the cut. Next, he cuts the horizontal bottom cut of the notch. The saw operator checks the notch and 50 Chapter 3: Logging methods the direction in which it is pointing, by positioning himself with his back to the tree while holding the edges of the notch with his hands. Figure 3.3 1. The felling cut starts at the side of the tree instead of the back - a short distance behind the key; 2. By moving the nose of the guide bar gently (avoiding kick back) inward, a slot is cut parallel to the key. The slot is either cut right through or made as far as the bar will reach, depending on the size of the tree; 3. Sawing is continued towards the back of the tree - either leaving a corner as support for the tree or keeping the guide bar parallel to the key; 4. If the tree is too thick to make the first cut right through, a second slot is cut opposite to the first one; 5. The tree is standing firmly on the key and corner support and cannot fall. The saw is removed form the kerf and a felling wedge is inserted; 6. The remaining corner is cut. The felling method that was used for leaning trees in the reduced impact logging system in Pibiri, Central Guyana. The sequence is different for fine trees (to the left) and big trees (to the right). The direction in which the saw is moved in each step is indicated by arrows. (After Brunberg et al., 1984) Just as with the classic approach, the back or felling cut is made horizontally and slightly above the bottom level of the notch. The position from where the back cut is made differs from the classic method. It is started at right angles to the notch, at the side of the tree, shaping the key first and proceeding from the key towards the back of the tree instead of starting opposite to the notch and sawing towards it. 51 Reduced impact logging in the tropical rain forest of Guyana Trees with no lean will fall after the remaining corner has been cut. In case of backward lean, the wedge is driven in until the tree started falling. A lean to one side is approached by leaving a tapered hinge, just as in the classic method. When the tree has a strong forward lean no wedges are inserted but the tree is still cut from the key backward to avoid splitting. Large buttresses are in principle cut in advance, except for the one(s) opposite to the desired direction. These buttresses act as the corner support. A slot is cut in the spur and subsequently a wedge is inserted at right angles with it. Especially in case of large buttresses such as those of Peltogyne venosa (Purpleheart) or Mora gonggrijpii (Morabukea), this approach produces a lot of leverage, the distance between wedge and hinge sometimes exceeding one metre. 3.3.5 Planned skidding Skidder movements near the stump partially or completely destroy such vegetation as survived the felling. The area near the stump plays an important role in the regeneration process of species with shade-tolerant seedlings (cf. Zagt & Werger 1998). Many of Guyana's timber species belong to this category (Hammond & Brown 1995; Hammond et al. 1996). The skid trail system is designed to minimise skidding distance, skidding on steep slopes, skidding downhill, and stream or gully crossings. Skid trails basically follow a straight alignment with 80 m spacing, but adjustments are made to encompass topography. Prior to felling, a two-person planning team flags the main path for the skidder to follow with brightly coloured vinyl ribbon and indicates the direction in which each tree is to be felled on a map based on the actual skid trail alignment on the ground. Branch trails are flagged after felling. The skidder itself is used to open trails. The dozer blade of a skidder is principally a device for aligning logs and to remove small obstacles which may hamper its movements. A skidder normally can not push down any trees larger than 25 cm dbh. A Cat 528 skidder is, however, well capable of pushing its own trails. The low density and scattered distribution of trees larger than 25 cm dbh, combined with the narrow width over tyres of the skidder (Table 3.3) allow the machine to push rather straight trails. This method is referred to as non-trail skidding (FAO 1977) and is general practice in forest exploitation in Guyana. Non-trail skidding is juxtaposed with methods where skid trails are constructed by a crawler tractor (Winkler 1997; Dykstra 1996), or manually (Hendrison 1990). The skidder proceeds working trail by trail, starting with the logs furthest away from the trail. The crew chief locates the logs with the aid of the map and directs the skidder operator where to turn, where to reverse into the forest and where to halt. At the first entry, when the trail is being opened, the skidder penetrates the trail to the end, positioning logs that blocked the way alongside. At the end of the trail the skidder turns and backs up into the forest creating a branch trail if necessary. The crew chief halts the machine 20 to 30 m away from the stump depending on the condition of the undergrowth, terrain and position of the log. One choker man stays near the skidder and travels with it to and from the log market. The second choker man sets the choker straps while the skidder is on its trip to the market. Both choker men roll out the main 52 Chapter 3: Logging methods cable and hook the ferrule, at the end of the strap, to the choker bell on the cable. The log is then winched up to the skidder that subsequently skids it out the branch trail and parks it some 50 m down the main trail. The skidder turns and drives or reverses to the point where the crew chief directs it into the forest again. The same procedure is repeated until a full load is composed, 2 to 4 logs. These logs are then bundled and skidded to the log market. In order to facilitate winching and to improve the efficiency of the operation, the skidder is equipped with detached choker straps (see Section 3.4). The use of the dozer blade is minimised and scraping of the topsoil during skid trail construction is discouraged. Skid trail grades should not exceed 15%. Logs have to be approached in reverse. The winch is used to haul in logs up to 30 m from the bole zone (depending on distance from stump to trail, and ease of access), thereby reducing damage in the vicinity of stump and felling gap. 3.4 EQUIPMENT AND PERSONNEL In Guyana, the most popular chainsaw is the Stihl 070. This machine thanks its popularity to its extremely good record of low maintenance and few breakdowns, and to the fact that saws are often used to cut boards in small operations, whereby its weight is a decisive factor. The machine is extremely heavy, specially combined with the traditionally used 75 cm guide bar. A machine of this type was used with the conventional operation. In the reduced impact system the following equipment is used: a Stihl AV066 chainsaw with 62 cm bar, 30 cm long aluminium wedges, 2 cm thick steel plates for extra leverage to be inserted under the wedge, and a 3 kg sledge hammer with a handle made from the fluted bole of Aspidosperma excelsum (Yaruru). Table 3.3 shows that the power to weight ratio of the Stihl AV066 is much higher than of the outdated Stihl 070, and therefore an ergonomically preferable machine. Wheeled skidders are designed specifically for logging and are commonly used in Guyana instead of crawler tractors, which are widely used in South East Asia and Africa. Shuttle systems, where a crawler is used to ‘pre-skid’ or ‘stump’ the logs and a skidder for the longer haul to the landing, are used by some companies in Guyana. This system is imported from South East Asia, but the favourable terrain conditions in most parts of Guyana do not warrant its application there. In both conventional and reduced impact logging operations, the same extraction machine was used: a Caterpillar 528 cable-arch skidder equipped with winch. The only difference in equipment between the two operations is that the machine is equipped with a 35 m main line to which a Bardon choker hook (‘bell’) is attached in the reduced impact logging system, whereas it is was equipped with a ‘bull hook’ attached directly to the main line in the conventional operation. In the latter case, no choker straps were used, so the main line had to be wrapped around the end of the log. The use of choker straps with Bardon hooks and ferrules allows the setting of the strap even when the bole is situated flat on the ground, because only the ferrule has to be passed underneath the bole. The 53 Reduced impact logging in the tropical rain forest of Guyana conventional ‘bull hook’ cannot be passed underneath the bole, implying that the end of the log has to be lifted from the ground somehow. The change over from the bull hook to the choker bell is therefore a conditio sine qua non for implementing a winch system. In the experimental logging area, the felling was conducted by two chainsaw operators that had been trained by a felling instructor of the Agricultural University of Frederiksburg, Denmark, Mr. O. Madsen. Each operator was accompanied by two assistants, Tropenbos personnel, who were trained in directional felling as well. One assistant located the trees to be felled from a map and indicated the pre-determined felling direction, while the other would take care of the felling preparations and wedging. Skidding was carried out by one skidder operator, DTL personnel, and three assistants, Tropenbos personnel. One assistant located the trees to be extracted by means of a map and instructed the skidder operator where to turn, where to back up to a log and where to halt. The second assistant set the choker strap and locked it to a choker bell on the main line. The third assistant travelled with the skidder, and rolled out the main line together with the second one. A major change took place in the responsibility of the logging clerk. In the reduced impact system he functions as a crew leader; indicating the felling direction of the trees to be felled based on the designated skid trail alignment and the specific field situation, closely supervising the opening of the skid trail and guiding the skidder operator where he should position the machine to collect logs and where the logs should be concentrated. The latter decision is based on the tree location map and the specific field situation. The changes to the logging crews and the equipment used are summarised in Tables 3.1, 3.2 and 3.3 Table 3.1 Composition of the felling crew and equipment used in the conventional (CNV) and reduced impact (RIL) operations Felling crew CNV 1 chainsaw operator 1 assistant Felling crew RIL 1 chainsaw operator 2 assistants Equipment used CNV 1 Stihl 070 chainsaw 1 cutlass Equipment used RIL 1 Stihl AV 066 chainsaw 2 cutlasses 1 sledge hammer 3 felling wedges 54 Chapter 3: Logging methods Table 3.2 Composition of the skidding crew and equipment used in the conventional (CNV) and reduced impact (RIL) operations Skidding crew CNV 1 machine operator 1 assistant 1 logging clerk Skidding crew RIL 1 crew leader 1 machine operator 2 assistants Equipment used CNV 1 Caterpillar 528 cable-arch wheeled skidder 1 winch line with bull-hook 1 cutlass Equipment used RIL 1 Caterpillar 528 cable-arch wheeled skidder 1 winch line with choker bells 6 choker straps 1 cutlass Table 3.3 Specifications of the equipment used for felling and skidding in the study Chain saw model Stihl AV 066 Stihl 070 Skidder model Caterpillar 528 cable-arch wheel skidder Power (kW) Bar length (cm) Weight (kg) 62 76 7.3 13.0 Winch pull (t) Weight (t) 14.5 13.2 5.0 5.1 Power (kW) 130 55 Width over tyres (m) 2.9 Reduced impact logging in the tropical rain forest of Guyana 56 Chapter 4: Logging damage 4. LOGGING DAMAGE 4.1 INTRODUCTION Most logging operations in tropical rain forests are selective (Johns 1988; Hendrison 1990; Jonsson & Lindgren 1990). Selective logging concerns the commercially valuable trees from a forest area; i.e. only big, faultless trees of commercial species are harvested. Typically only a few trees are extracted per hectare, but logged forests are often left in a highly degraded state (e.g. Nicholson 1958, 1979; Fox 1968b, Liew & Wong 1973; Chai & Udarbe 1977; Lee 1982b; Johns 1988; Uhl & Vieira 1989; Hendrison 1990; Johns et al. 1996; Pinard & Putz 1996; Whitman et al. 1997). The damage done to the forest during initial log extraction co-determines the value of logged forests for future timber production, and the contribution of these forests to conservation of biodiversity (Gullison & Hardner 1993), and may be the crucial factor determining whether sustained-yield forestry is feasible within a regular planning time-span (Whitmore 1984). Damage levels vary greatly with the stocking density of commercial timber, the method of exploitation employed and current economic conditions. Damage to the residual stand ranges from 4.8% (Whitman et al. 1997) to 43.0% (Veríssimo et al. 1992) in neotropical forest, but is generally below 25% (Uhl & Vieira 1989; Hendrison 1990; Uhl et al. 1991; Gullison & Hardner 1993), while damage levels in South East Asia are generally around 50% of the stand (Nicholson 1958; Fox 1968a/b; Burgess 1971; Mattsson-Mårn & Jonkers 1981; Johns 1988). The differences in damage levels can be attributed mainly to the stocking density and market situations. Many neotropical studies focused on the exploitation of Mahogany (Gullison & Hardner 1993; Veríssimo et al. 1995; Whitman et al. 1997), that is characterised by the low density of this species. The South East Asian studies deal mainly with Dipterocarp forest, which is densely stocked with commercial species. Johns (1992) produced a theoretical minimum damage curve for different extraction levels. However, it is generally considered impossible to determine a relationship between harvest intensity and forest damage by comparing studies, due to differences in methodologies, stand characteristics, topography, forest types, extraction techniques and experience of workers (Nicholson 1958; Gullison & Hardner 1993). Logging in Guyana, although carried out with little care and respect for the remaining stand, has only scratched the surface and vast areas of untouched forest do remain. In most parts of Guyana, selective logging disturbs the landscape in a patchy fashion, because the most desirable timber species, Chlorocardium rodiei (Greenheart), occurs in monodominant patches (Davis & Richards 1933, 1934; Fanshawe 1954; Ter Steege et al. 1996; Zagt 1997 - see also Section 1.5.2). In areas of such forest with high densities of commercially viable stems logging damage is severe, while other areas may be untouched because of absence of harvestable trees. Extraction rates in Greenheart patches can be extremely high. Clarke (1956) counted 20 Greenheart stumps per hectare in the Bartica Triangle in forest that had been logged before 1950, while Zagt 57 Reduced impact logging in the tropical rain forest of Guyana (1997) found a similar density of stumps in the Mabura Hill area in forest that had been logged in the 1980s. These logging intensities correspond to a roundwood volume of approximately 60 m3/ha, a harvest rate that is usually only found in South East Asia. In Guyana, selective logging is removing commercially valuable trees with a minimum diameter of 40 to 60 cm, depending on the species, although according to the forest regulations even a threshold of 34 cm dbh is allowed. Felling of trees and extraction of logs are the two main causes of logging damage. Damage is inevitable as trees must fall somewhere, and must be extracted to a landing somehow. Removal of commercial trees is only a minor effect of selective logging. Incidental damage associated with the fall of the cut tree and the subsequent mechanised extraction of the bole, results in far greater losses of and damage to other (small) timber and non-timber trees. A logging operation disrupts the existing forest architecture, by impacts causing the formation of gaps, injury to crowns and trunks of remaining trees, and alteration of the frequency distribution of tree size classes. These changes can be recorded to compare the impact of different logging methods and intensities. The initial density and the size class frequency distribution of the trees that are liable to be affected by the operation will influence the extent of the associated losses and damage. Research in Suriname (Hendrison 1990), Brazil (Johns et al. 1996), Australia (Crome et al. 1992) and South East Asia (Mattson-Mårn & Jonkers 1981; Bertault & Sist 1995; Pinard & Putz 1996) has indicated that substantial reductions in forest damage can be achieved through improved felling and skidding practices and proper planning of operations. Most studies of logging damage are studies of damage caused by both felling and extraction. Separating the two causes a posteriori is almost impossible. It is at least not possible to determine consistently if a tree is lost during skidding or during felling (Crome et al. 1992). To better understand the potential of improved felling and skidding techniques per se in reducing logging damage, separate studies of each logging phase are required. Logging intensity by itself has not been widely studied (except Mattson-Mårn & Jonkers 1981; Jonkers 1987; Borhan et al. 1987; Schmitt 1989) and only one study is known to me where both reduced impact logging and logging intensity were examined simultaneously (Sist et al. 1998 - in Kalimantan). The objectives in this chapter are: ƒ to compare the damage in forest logged by conventional methods and in forest logged according to a reduced impact system; ƒ to examine the damage caused in each phase of the logging operations; ƒ to examine the relationship between logging intensity and damage; ƒ to examine the effect of the initial forest structure and architectural mosaic on logging damage. 58 Chapter 4: Logging damage Specifically, the change in the architectural forest mosaic and the composition of the tree population is considered by: ƒ mapping skid trails and canopy gaps to determine the amount of area affected by logging ƒ assessing damage to residual trees 4.2 MATERIAL AND METHODS A reduced impact logging system was designed for selective logging in Guyana. The system is inspired by the CELOS harvesting system (Hendrison 1990) and includes many elements typical of reduced impact logging systems described elsewhere (e.g. Pinard & Putz 1996; Johns et al. 1996; Van der Hout & Van Leersum 1998). In addition, measures were taken to discourage the formation of large felling gaps, by aiming at an even spacing of trees to be felled. Tree selection criteria furthermore considered differential felling limits - according to differences in specific growth patterns, the abundance of the various commercial species in a plot, and the distribution of suppressed future crop trees. Conventional logging is often described as unplanned, haphazard timber harvesting, but there are substantial differences between logging companies. The conventional operation that was monitored during the present study featured a relatively high level of organisation and planning. Felling and skidding operators, however, were left to their own judgement in the way that trees were felled and extracted. For details of both logging methods, see Sections 3.2 and 3.3. 4.2.1 Study area and pre-logging measurements The study was conducted in the West-Pibiri compartment of the Demerara Timbers Ltd. concession (Fig. 1.1). The total area that was reserved for experimentation covered about 1,000 ha, of which four blocks of 100 ha were judged to be suitable for the experiment (see Section 1.7.3). In these four blocks, trees belonging to 20 commercial species down to 40 cm dbh were enumerated and mapped. This information was used to select 15 plots, measuring 240 × 240 m2, that had to contain at least 16 harvestable trees per hectare. Twelve plots were to be submitted to reduced impact logging with logging intensities of 4, 8 and 16 trees/ha. The second intensity was duplicated with a view towards planned silvicultural post-harvest treatment. Logging intensities were assigned according to a randomised block design with three replicates (Fig. 1.3). Damage assessment plots were established in the 140 × 140 m2 core area of each plot. All trees ≥ 20 cm dbh were measured at 1.3 m above the ground, or 30 cm above the buttress, using a steel diameter tape between July and November 1993. Following Alder & Synnott (1992), nested subplots, 25 in each plot, were used for smaller trees: 100 m2, stems 5-20 cm dbh (Fig. 1.2). Three contiguous 100 ha working units immediately bordering the reduced impact logging area were earmarked to monitor conventional logging. According to the timber 59 Reduced impact logging in the tropical rain forest of Guyana company’s inventory records, the three blocks harboured 492 sawmill and 179 peeler logs, representing a potential harvest of just over two trees per hectare only. However, concentrations of harvestable stems could be distinguished on the stock maps. Three units, measuring approximately 200 × 600 m2, encompassing these concentrations were earmarked as the exact locations where conventional logging would be monitored (Fig. 1.3). Selection of trees to be felled was left up to the felling crew of the timber company. The logging intensity thus varied along with the density of stems that were regarded to be worthwhile of taking by the felling crew. In the centre of each unit a 20 m wide, 500 m long strip was demarcated, in which all trees ≥ 20 cm dbh were measured in May 1995. Trees between 5 cm and 20 cm dbh were recorded in a 5 m wide section of the strip. In both conventional and experimental logging areas, each tree was labelled with an aluminium number tag for reliable identification after logging and to facilitate long-term monitoring. All trees were identified up to species or genus. Stem and crown damage due to natural causes were described and any other tree characteristics that might be mistaken for logging damage were noted. Climber cutting took place in the twelve experimental plots in November-December 1993. Felling was conducted by two trained Guyana Forestry Commission / Tropenbos crews between March and July 1994, to be followed by skidding between during October-November 1994. A skidder plus crew were provided by the timber company. The operator was directed and assisted by trained Tropenbos staff (see Section 3.3.5). Damage assessments were carried out after felling and were repeated after skidding. Logs were collected from the landings by the timber company that retained the logs. Conventional logging took place in March and April 1996. Damage assessment commenced after completion of skidding. 4.2.2 Post-logging measurements In the reduced impact logging area, measurements were carried out before and after skidding commenced. Before logs were skidded, all permanent plots that underwent experimental felling were visited. At each felling gap within the perimeter of the assessment plot (140 × 140 m2), loss of canopy cover by felling and the damage due to felling were assessed. In the conventionally logged area, assessments could not be carried out before skidding. Stump positions were mapped with tape and compass after skidding. The spatial distribution of the stumps was used to delimit 2-ha plots (100 × 200 m2) with logging intensities of 8 and 16 trees/ha. No areas of this size could be found with a logging intensity of 4 trees/ha. In all, three replicates of both intensities were demarcated. In each gap within the perimeter of the plots, loss of canopy cover and damage to individual trees were assessed in the same manner as with reduced impact logging. 60 Chapter 4: Logging damage Felling damage The size of each canopy opening was estimated using a ‘centre-point’ system. Gap phenomena are familiar to many researchers, but definitions precise enough to be reproducible are not. Comparison of gap processes with natural disturbances has been limited by a lack of standardised methods for sampling and describing gaps (see Section 2.2). Several gap definitions have been used to measure gap sizes by natural disturbances, and within one definition methods of gap size estimation vary considerably; see reviews in Brokaw (1985a), Runkle (1992), Brils & Laan (1995), Van der Meer et al. (1994), Hammond (in prep.). In the experimentally logged area, assessments were made by Brils & Laan (1995) who followed a method used by Hammond & Brown (1992). The method follows Koop’s method (1989) to assess crown projections, but mirrors it to assess canopy openings. Gap area was estimated following the methods described in Brokaw (1982a/b), but used a modified definition. Instead that a gap was defined as ‘a canopy opening extending to all levels down to an average height of 2 m above the ground’, the size of the opening was determined at the local level of the dominant canopy layer (Brils & Laan 1995). This definition is similar, in practice, to the method suggested by Van der Meer & Bongers (1996a). The centre of the gap was determined by estimating the centre of the vertical projection of the hole in the canopy. The distance from the centre-point to the edge of intact canopy was subsequently measured in at least eight directions at 45° intervals. In case of large and/or irregularly shaped gaps, additional measurements are taken in the direction of the longest gap axis. The outline of a gap being defined by the edge of the hole created in the dominant canopy layer, implied that in some cases emergent trees were not considered to be part of this layer, while in case of a building phase (sensu Whitmore 1978) - late aggradation phase (sensu Oldeman 1990, p. 173) patch, pole sized trees were considered as forming the canopy. The co-ordinates of all vertices of the resulting polygon were obtained by subtraction of successive vectors. The area of the polygon was then calculated by a continuous product method (Gauss’ formula). If the (xiyi) are the co-ordinates of the vertices then the area is: A= 1 n ∑ xi ( yi −1 − yi +1 ) 2 i =1 Gaps that were crossing the perimeter of the assessment plots, either caused by trees having been felled outside that had fallen into the plot, or, conversely, felled inside and fallen to the outside, were only assessed for the segment that fell within the perimeter. The reasoning behind this was that the plot was used as the area of reference to determine canopy loss. Therefore, only the loss of canopy within the perimeter of a plot was considered to be of interest. It was realised that this was at the expense of the number of felling gaps on which an estimation of the average size of felling gaps could be based. This type of edge effects, however, can never be avoided completely when a delineated reference area is used. There will always be trees that fall over the border 61 Reduced impact logging in the tropical rain forest of Guyana of the reference area. Hence, gap size estimates will always be affected by such edge effects. Average gap sizes were estimated indirectly as shown below and it was therefore decided that this assessment method was adequate. In the conventionally logged areas, the size of canopy openings was estimated in the same way as with the experimental logging. In the conventionally logged plots less but larger gaps were formed. Therefore, in contrast to the assessments with experimental logging, we measured gaps that were crossing the border of the plot entirely, in order to obtain a sufficiently large number of observations to assess gap size distribution. In the experimentally logged plots, damage to individual trees ≥ 5 cm dbh was assessed in each gap. For each damaged stem, we noted the species - if it was a commercial species, and categorised the type of damage: cut (not for timber, but to enable felling of a neighbouring timber tree), crushed, uprooted, snapped, top broken off, split trunk, bark laceration (width and length), crown damage (percentage excised), or leaning or bending (deviation from vertical in both cases). Snapped-off and topped trees were distinguished from other severely damaged trees because it was expected that some of these would re-sprout. Skidding damage Directly after skidding, we mapped the affected ground area using a fibreglass tape and compass. We distinguished between planned main trails, unplanned branch trails and traces of manoeuvring in felling gaps and along trails. The width of a skid trail was the maximal extent of damaged vegetation according to visual assessment, and was recorded every tenth meter. The area occupied by trails was then calculated by summing the areas of the resulting trapezoid. Large disturbed patches, e.g. as a result of manoeuvring in the felling gap or creating a ‘ramp’ for bundling, were measured in the same manner as canopy openings; i.e. by selecting a point roughly in the centre of the area and measuring the distance and compass bearing to the edge of the disturbed vegetation in eight directions. Damage to individual trees ≥ 5 cm dbh was assessed on and along the skid trails, and in and along manoeuvring areas. Damage was categorised in the same way as with felling. Trees ≥ 20 cm dbh were measured and labelled before logging, so it could be easily verified whether or not a tree was missing during one of the two assessments, after which the situation could be corrected. This was not the case with trees < 20 cm dbh, so a few uprooted or crushed (and subsequently moved) trees may have been overlooked. Therefore, trees in the sub-samples were checked for signs of damage (or death) after logging in January 1995. In the latter case determining the cause of damage - felling or skidding - was based on signs of e.g. ground disturbance, felling debris or stump position. Tree survival was monitored annually until January 1997 and all dead trees were recorded. This is especially relevant for snapped-off trees that had re-sprouted the first 62 Chapter 4: Logging damage year, but died later, and for trees with bark lacerations, expected to be more susceptible to fungal attacks. In the conventionally logged area, some uprooted and crushed small trees may have been overlooked, because the first assessment could only take place after skidding. Therefore, it was decided to leave out the smallest trees and to take into account trees ≥ 10 cm dbh only. In each gap and along the trails, trees of this size class were checked for damage. 4.2.3 Volume measurement In the experiment, individual log volumes were determined at the stump after the boles were topped using Newton’s formula (e.g. Husch et al. 1982 – see Appendix B). In the experimentally logged plots, the volume extracted from each plot was thus determined per individual tree. There was no opportunity to measure the necessary bole dimensions during the conventional operation, but diameter at breast height could be taken before the feller started felling. Volume estimates were attained by using regression models. These models were constructed with the tree dimension data from the experimental logging. The models are provided in appendix B. 4.2.4 Analysis General The experimental design was incomplete – three logging intensities in reduced impact logging and two in conventional logging – and unbalanced – logging intensity of 8 trees/ha had six replicates whereas all other treatments had three replicates. This prevented the use of analysis of variance. In order to utilise as much of the gathered information as possible, linear regression analysis with dummy variables was used (Zar 1996, p.430): Y = a + b1 X 1 + b2 X 2 + b3 X 3 + b4 X 2 X 3 + ε , where Y is the response variable. If logging intensity is expressed in the categories 4, 8, and 16 trees/ha, it is split into two dummy variables: X1 = 1, X2 = 0 stands for intensity 4, X1 = 0, X2 = 1 stands for intensity 8 and X1 = 0, X2 = 0 stands for intensity 16. X3 is a dummy for logging method having the value 1 in case of conventionally logged plots and the value 0 otherwise. Because the combination X1 = 1, X2 = 0 does not occur when X3 = 1, only one interaction term is defined. The independent variables are entered into the model one by one, allowing the testing of a significance of each factor. In the cases that logging intensity is expressed as a continuous variable; i.e. basal area felled, dummy variables are only defined for the logging method. Regression analyses were performed with Statistica 4.5 (StatSoft 1994). Alternatively, type IV sum of squares general linear models are sometimes used for ease of presentation. These tests were performed with SPSS 7.5. This type of general linear model does not provide a unique result (Faber, pers.comm.; SPSS 1998). Therefore, the results of all presented Type IV sum of squares linear models were checked by performing linear regression analysis with dummy variables too. 63 Reduced impact logging in the tropical rain forest of Guyana In analyses where the dependent variable is categorised; e.g. gap size classes, loglinear analysis is used. Log-linear analysis was performed with Statistica 4.5. Graphs were made with the same application. Logistic regression was used to assess mortality, for which SPSS 7.5 was used. Damage to the residual stand Dead trees were classified as felled, uprooted, broken below reference point, or cut, the latter in order to reach a neighbouring timber tree that was to be felled or to allow it to fall in case of lodging. Trees snapped above reference point and inclined trees, which were either partly uprooted or bent, were classified as severely injured. Concerning injuries that affected the stem or the crown, as many as four stem damage and three crown damage classes were distinguished in the field (see above). In the analyses a comprehensive classification of six categories was used: ƒ felled tree; ƒ killed tree; ƒ severe injury: the whole crown has broken off (topped), the trunk has snapped above reference point, the trunk has split, or the tree has been partially uprooted or bent; ƒ moderate injury: at least half of the crown has broken off, the bark has been lacerated over at least 20 cm of the circumference of the stem or, if the tree is smaller than 20 cm dbh, over at least one-third of the circumference, or any injury on both stem and crown; ƒ minor injury: any stem or crown injury not qualifying as severe or moderate injury; ƒ tree undamaged. 4.3 COMPOSITION AND STRUCTURE OF THE INITIAL STAND Prior to logging, an inventory of all trees ≥20 cm dbh was conducted in each 2-hectare assessment plot of the experimentally logged areas and in three 1-hectare strip samples in the conventionally logged areas. The results, summarised in Table 4.1, showed an average of 22 harvestable stems per hectare prior to logging, of which 53% were Greenhearts. This strongly contrasts with the potential harvest of 2 trees/ha indicated by the survey of the timber company for the total area (see Section 4.2.1). This seeming contrast is largely caused by the spatial clustering of several commercial tree species. Our inventory, however, indicated that the average potential harvest for the total area must have been at least 6 trees/ha, half of which being Greenheart. The potential harvest indicated by the survey of the timber company may have been lower because of undue high-grading: i.e. only selecting prime species and of those only the choicest stems. The inventory revealed no differences in stand density (Table 4.1 - ANOVA, F3, 11 = 1.11, p = 0.39), or total basal area (ANOVA, F3, 11 = 2.24, p = 0.14) between the three experimentally logged blocks and the combined conventionally logged areas. The stocking with commercial stems - trees belonging to 33 prime species (see appendix A) with dbh≥40 cm and without serious defects - differed significantly (ANOVA, 64 Chapter 4: Logging damage F3, 11 = 13.66, p < 0.001). The conventionally logged area had a significantly higher stocking of commercial stems than experimental blocks 1 and 2 (Tukey multiple comparisons' tests, p < 0.005 and p < 0.001 respectively), while experimentally logged block 3 had a significantly higher stocking than block 2 (p < 0.05). The difference in stocking with harvestable Greenheart was even more pronounced (ANOVA, F3, 11 = 14.91, p < 0.001). Focusing on this species alone, the conventionally logged area was better stocked than any of the experimentally logged blocks. Nearest to the conventionally logged area came experimental blocks 1 and 3 (Tukey multiple comparisons test, p < 0.05), which were in turn markedly better stocked with Greenheart than block 2 (Tukey multiple comparisons test, p < 0.05). It should be noted that this comparison concerns commercially sound stems only. This does not have to reflect differences in species composition, because differences among plots in the proportions of defective stems per species can obscure similarity in species composition. Although the total number of stems was not found to differ between the areas, significant differences were found in diameter class frequency distributions (Pearson χ2 = 39.70, df = 6, p < 0.001). Chi-square analysis of diameter classes between separate plots showed significant heterogeneity. Examination of the frequency values reveals that the proportion of large trees (dbh ≥ 60 cm) is decidedly smaller in plots 4, 9, 10 and 14, whereas the proportion of small trees (dbh < 40) appears to be greater in plots 4 and 8, 9 and 10. Suspecting that the latter significant χ2 was due largely to these plots, plots 4, 8, 9, 10 and 14 were omitted and the remaining contingency table was considered. This manipulation indeed produced a non-significant χ2 (p = 0.14), confirming that the size class distribution of the remaining plots did not differ mutually. Because it is not proper to test a hypothesis after examination of the data, this analysis should only be considered as a guide. Investigation into differences in diameter distributions was also conducted by comparing median diameters between areas (pooled plots). These differed significantly (Kruskal-Wallis ANOVA by ranks, Hc (df = 3, n = 5251) = 52.05, p < 0.001). A similar analysis conducted for all plots separately revealed a significant difference as well (Kruskal-Wallis ANOVA by ranks, Hc (df = 14, n = 5251) = 97.71, p < 0.001). Subsequent non-parametric Tukey-type multiple comparisons (Zar 1996) showed that plots 13, 26 and 40 differed markedly from plots 4, 8, 9 and 10; plot 14 from plots 9 and 10; and plots 1, 2, 3, 7 and 15 from plot 10. In contrast to the χ2 test, this analysis ranked plot 14 among the ‘heavier-stocked’ plots. This is probably due to the fact that in this plot the proportion of middle sized trees is one of the largest of all plots. Plots 4 and 8, 9 and 10 thus have higher proportions of small trees than the other plots. The comparisons above revealed few differences between the experimentally and the conventionally logged areas prior to logging. Some of the plots that were to be submitted to experimental logging appeared to harbour a higher proportion of small trees. It was investigated whether or not these plots were inadvertently assigned the same logging intensity. 65 Stand structure and density of harvestable stems prior to logging in twelve plots subjected to reduced impact logging (RIL) with three logging intensities and in three strip samples subjected to conventional logging (CNV) - logging intensity depending on availability of harvestable stems - in Pibiri, central Guyana Method Block Plot Logging intensity (n/ha) Number of stems (n/ha) 10-20 RIL 1 1 2 66 2 3 3 CNV Diameter class (cm) 20-40 40-60 < 60 Total Median Maximum Basal area Harvestable Greenheart Harvestable diameter diameter stems dbh ≥ 40 cm Greenheart (cm) (cm) (m2/ha) (n/ha) (n/ha) (n/ha) 1 2 3 4 total 8 16 4 8 304 280 260 320 291 133 131 145 148 139 32 40 38 35 36 12 15 18 8 13 482 466 462 511 480 31.1 31.0 31.4 29.7 30.8 112.7 133.8 109.8 93.9 112.6 24.2 27.0 29.1 24.2 26.2 20 23 20 19 21 21 22 28 22 23 14 15 13 14 14 7 8 9 10 total 16 8 8 4 256 444 316 360 344 127 177 176 162 160 40 37 38 32 37 8 11 8 7 8 432 668 537 561 549 31.5 29.0 28.2 27.6 29.1 111.0 80.9 98.3 101.7 98.0 23.2 28.6 25.6 24.7 25.5 18 18 18 15 17 15 13 7 8 11 9 8 5 5 7 11 13 14 15 total 4 8 16 8 280 216 256 244 249 146 121 133 151 138 39 45 49 46 45 13 19 6 17 14 478 402 443 458 445 30.0 34.9 32.6 30.3 31.9 88.7 83.4 110.5 103.1 96.4 25.7 27.9 24.6 28.6 26.7 19 30 25 25 25 8 19 14 23 16 3 14 12 14 10 240 288 288 272 128 141 156 142 51 51 49 50 16 16 12 15 435 496 505 479 33.3 34.6 30.5 32.8 80.1 73.0 95.4 82.8 27.9 29.5 27.9 28.4 35 27 36 33 37 29 27 27 22 20 26 23 26 40 53 total Reduced impact logging in the tropical rain forest of Guyana Table 4.1 Chapter 4: Logging damage The inventory revealed no consistent differences in stand density prior to logging (ANOVA pooled error term, F2,7 = 1.75, p = 0.24), total basal area (ANOVA pooled error term, F2,7 = 0.10, p = 0.90), or stocking with commercial stems (ANOVA pooled error term, F2,7 = 2.64, p = 0.14) between the plots that were assigned different logging intensities. Even with respect to the number of commercial Greenheart stems no significant differences were found (ANOVA pooled error term, F2,7 = 3.49, p = 0.09). Nevertheless, the diameter class distribution was not comparable between plots with different planned logging intensity (partial association G = 12.80, df = 4, p < 0.05). However, the correlation between diameter class distribution prior to logging and the intended logging intensity was not the same in all of the blocks (three-way interaction G = 24.67, df = 8, p < 0.01). We thus cannot conclude that a different initial forest structure, e.g. a higher share of small or a higher share of large stems occurred consistently at one specific logging intensity. It was also examined whether or not the median diameter different along with the assigned logging intensity. The plots that were assigned a logging intensity of 16 trees per hectare had a median diameter of 31.5 cm (pooled plots), the ones with 8 trees per hectare a median diameter of 30.0 cm (pooled plots) and the ones with 4 trees per hectare a median diameter of 29.5 cm (pooled plots). The median diameters of the plots that were assigned the highest and lowest intensity differed significantly (Kruskal-Wallis ANOVA H (df = 2, n = 4631) = 7.98, p < 0.05; followed by Tukey type non-parametric multiple comparisons; q = 2.66, p < 0.05), but the magnitude of the difference is quite small. It is not known whether these differences in diameter class distribution and availability of harvestable stems prior to logging affected the logging damage reported below. One could imagine that a high proportion of smaller stems would result in higher numbers of incidental losses, while on the other hand in forest with lower share of large trees smaller gaps might be created. Therefore, it is important to realise that such differences existed and to keep them in mind when examining the results described below. 4.4 OVERVIEW OF THE LOGGING OPERATIONS 4.4.1 Conventional logging operation In the in total 32 ha submitted to conventional logging 279 trees were cut (8.7 trees/ha) corresponding to a total basal area of 72 m2 (2.2 m2/ha) and a total gross volume over bark of 888 m3 (27.8 m3/ha). The average dbh of the felled trees was 56.4 cm (sd = 9.4 cm) and the average volume over bark was 3.2 m3 (sd = 1.4 m3) per tree. The average felling intensity of 8.7 trees/ha conceals that, in some parts, the absolute intensity amounted to 25 trees/ha, while, in other parts, completely untouched hectares occurred. This meant that on a single hectare the harvested gross volume varied from nil to 78 m3/ha. Since this was a typical logging operation to yield saw logs, 96% of the extracted trees (91% of total gross volume) were Chlorocardium rodiei (Greenheart). Only three other species were cut: Peltogyne venosa subsp. densiflora (Purpleheart) with 6% of the 67 Reduced impact logging in the tropical rain forest of Guyana total volume, Hymenaea courbaril (Locust) with 2% and Diplotropis purpurea (Tatabu) with 1%. The fact that the operation focused for nearly 100% on gregariously growing Chlorocardium rodiei explains the high observed variation in logging intensity, even within tracts of only 10 to 12 hectares. 4.4.2 Reduced impact logging operation In the in total 69 ha submitted to experimental logging 612 trees were cut corresponding to a total basal area of 158 m2 and a total gross volume over bark of 2064 m3. The average dbh of the felled trees was 55.7 cm (sd = 13.6 cm), the average log length 17.1 m (sd = 3.7) and the average volume over bark 3.4 m3 (sd = 2.3 m3) per tree. One of the aims of the reduced impact logging concept was to reduce the size of felling gaps. In order to achieve this, trees to be felled were spread as evenly as possible over the area. This led indirectly to a relief of pressure on Chlorocardium rodiei. Twenty-six different species were involved, but Chlorocardium rodiei still was the main supplier with 53% of the stems. Another eight species made up 38% of the cut stems; i.e. Catostemma commune and C. fragrans (Swamp and Sand Baromalli) with 11%, Eperua falcata (Soft Wallaba) with 8%, Peltogyne venosa (Purpleheart) with 5%, Swartzia leiocalycina (Wamara), Goupia glabra (Kabukalli) and Mora gongrijpii (Morabukea), all with 4%, and Carapa guianensis (Crabwood) with 3%. In terms of volume Chlorocardium rodiei contributed 47%, Catostemma spp. 12%, Peltogyne venosa 11%, Eperua falcata 8%, Mora gongrijpii 4%, Goupia glabra and Swartzia leiocalycina both 3%, and Hymenaea courbaril and Carapa guianensis both 2%. Catostemma spp, Chlorocardium rodiei, Goupia glabra and Peltogyne venosa belong to the five species with the highest national production during 1990-1996 (GFC annual reports). Eighty-eight per cent of the trees that were cut belong to the 33 major species, 76% being sawnwood species and 12% peeler species. Eperua falcata, which made up 8% of the harvest, is normally used for special products such as poles, shingles and staves. It is usually not felled in the mixed forest, but in special operations in the Eperua spp. dominated forest on white sand. Mora gongrijpii is another species that is usually left standing. Its timber properties are fully comparable with Mora excelsa, but it is usually condemned because of the predominance of hollow stems (cf. De Milde 1970). The inventory confirmed this ill fame: of the 121 Mora gongrijpii trees larger than 40 cm dbh encountered in the inventory, 73 (60%) were rejected standing, while of the 22 presumably good, felled stems 8 were rejected over the total experimental area after felling. Chlorocardium rodiei, Eperua falcata and Swartzia leiocalycina also lived up to their bad reputation as far as decay is concerned. Chlorocardium rodiei had 34% of its stems rejected among 501 trees greater than 40 cm dbh on inspection prior to logging, Eperua falcata 51% of 118 trees and Swartzia leiocalycina 41% of 44 trees. After felling another 1% of the Chlorocardium rodiei trees, 4% of the Eperua falcata trees and 4% of the Swartzia leiocalycina trees were rejected in addition. In total 21 out of 612 stems were rejected after felling. These stems were nonetheless extracted from the plots, so as not to affect the experimental design. 68 Chapter 4: Logging damage 4.4.3 Variation in the dimensions of felled trees The size of felling gaps is largely determined by the size of the crown of the felled tree. Not only will the removal of a broad crown from the canopy leave a larger open space than the removal off a smaller crown, it is also likely to cause more damage to the surrounding (smaller) trees in the spot where it falls (cf. Mutoji-A-Kazadi 1977; Hendrison 1990; Brils & Laan 1995). Emergent trees - such as Hymenaea courbaril or Peltogyne venosa - are not only taller than the average canopy height, they also have a larger, wider-spreading crown, as dictated by their sympodial, largely plagiotropic growth and reiteration patterns. The felling of an emergent tree will result in a typical chablis. Its crown may not leave an empty space in the canopy after removal because it overtops other canopy trees, but its fall may crush or uproot many surrounding trees. Measuring tree crown sizes is a difficult and time-consuming task, even when considering the vertical projection of the crown on the forest floor alone. There is, however, a certain amount of correlation between stem diameter and crown diameter (e.g. Husch et al. 1982; Jonkers 1987). This relationship varies from species to species according to its architectural diagram (Edelin 1991). In section 6.3.5, relationships between crown diameter and bole diameter are described for three abundant commercial species; i.e. Chlorocardium rodiei, Mora gonggrijpii and Catostemma fragrans, by fitting regression models. For all three species a significant correlation between the two parameters is shown (0.60 < R2 < 0.65 - Fig. 6.18) and also clear differences between the species (Table 6.5). Having established that there is a relationship between bole diameter and crown diameter, and assuming that there is a relation between crown diameter and felling ‘gap size’ (cf. Brils & Laan 1995), it is clear that differences in the diameters of the felled trees should be taken into consideration. Besides crown size, tree height is assumed to play an important role. A tree shows allometric relations that usually change during its life. These variable ratios between dimensions such as total height, crown point height and diameter (at breast height) indicate phases of the development of a tree, or other properties of a tree at one moment of its life (Oldeman 1990 p. 98-103). The ratio between a tree’s total height and its diameter is related to the architecture of a tree. Oldeman (1974a) found in French Guiana that, as long as a tree is developing conform its specific architectural model, a ratio of h = 100 · d is found, but when a tree crown expands by abundant reiteration within the crown the ratio becomes h < 100 · d. Under tropical rain forest conditions, measuring tree heights is a difficult and time-consuming activity. Diameters, however, are easy to measure and were known for all felled trees. In order to assess whether these could be used to estimate total height, a regression model was fitted for Chlorocardium rodiei using logarithmic transformation. The data set prepared for chapter 6 was used. ht = 41.78 + 24.75 · 10log dbh (R2 = 0.54, F1,112 = 126.4, p < 0.001) where: ht = total height (m) dbh = diameter at breast height (m) 69 Reduced impact logging in the tropical rain forest of Guyana Length of the branch free bole, or crown point height, can also be used as a predictor of total tree height (Ogawa et al. 1965; Oldeman 1990, p. 103). The crown point height usually coincides with the height to the first large branch, which is also the traditional cut-off point for the gross bole length. Bole length information was available for all trees felled in the reduced impact logging operation. In order to assess whether these could be used to estimate total height, a regression model was fitted for crown point height and total tree height for Chlorocardium rodiei, using the data set prepared for chapter 6. The fitted relationship closely resembles Oldeman’s (1974a) empirical ratio of hf = 0.5 ⋅ ht: hf = 2.03 + 0.53 · ht (R2 = 0.64, F1,114 = 79.41, p < 0.001) where: ht = total height (m) hf = crown point height (m) Both stem diameter and crown point height are thus good indicators of total tree height and both tree dimensions were recorded during the reduced impact logging operation. In order to establish whether or not tree diameter and crown point height predicted total height independently, they were entered stepwise into one regression model. The result indicated that entering of both independent variables was significant (partial correlation (10log dbh) = 0.68; t112 = 9.70, p < 0.001, and partial correlation (hf) = 0.54; t112 = 6.71, p = 0.001). Both parameters can thus be used as a co-variate when comparing felling gap sizes. However, there were no bole length data for the conventional areas so that volumes had to be estimated by using the relationship between dbh and volume in the experiment (see Appendix B). Since in the latter case bole length itself is not known, it is suggested to use trunk volume as a predictor variable instead. Trunk volume by definition is an expression of bole length and basal area; the latter defined by the squared diameter. Means per plot and standard deviations of dbh, bole length and tree volume are shown in Table 4.2. The total expected gap ‘area’ per plot is presumed to be influenced by the mean diameter or volume of the felled trees. It is also presumed to be dependent on whether mainly trees of which the diameter was close to the average or some very large trees and a lot of small trees were felled (cf. Brils & Laan 1995). A difference in the mean felling diameter shows a difference in the central tendency of a distribution (measures of central tendency - or location statistics - are e.g. the arithmetic mean, the median or the mode). However, such a statistic does not describe the shape of the frequency distribution. This distribution may be long or narrow; it may be humped or it may be asymmetrical. Quantitative measures of the spread of frequency distributions are e.g. the range or the standard deviation. A measure for asymmetry is skewness; i.e. one tail of a distribution is drawn out more than the other. 70 Table 4.2 Method RIL Logging intensity expressed as number of trees, basal area and gross volume over bark per hectare and bole dimensions in twelve 2-ha plots subjected to reduced impact logging (RIL) with three logging intensities and six 2-ha plots subjected to conventional logging (CNV) with two logging intensities in Pibiri, central Guyana Block Logging Basal area intensity felled n/ha (m2/ha) Diameter (cm) Bole length (m) Gross volume ob. (m3) mean sd skewness mean sd per ha per tree mean sd Greenheart % of harvested volume 1 2 3 4 8 16 4 8 2.0 3.4 1.1 1.9 54.6 50.2 59.6 53.4 8.3 9.9 5.6 10.8 0.5 1.9 –1.9 1.3 16.6 14.5 15.2 16.5 4.0 3.9 2.6 4.9 25.6 39.7 14.3 26.3 3.1 2.4 3.5 3.2 1.7 1.2 0.8 2.3 62% 59% 84% 79% 2 7 8 9 10 16 8 8 4 4.0 1.8 1.8 1.2 53.4 51.1 51.9 59.4 16.0 12.2 8.9 20.4 2.5 1.2 1.0 1.3 16.7 15.2 17.6 16.6 3.5 3.1 4.5 3.2 53.8 20.8 23.0 16.3 3.3 2.6 2.8 4.0 2.8 1.4 1.0 3.7 38% 48% 23% 26% 3 11 13 14 15 4 8 16 8 1.3 2.7 3.6 1.9 62.4 64.7 51.8 54.3 13.1 9.3 12.3 9.2 0.3 -0.2 3.3 0.6 20.0 19.7 17.2 16.8 2.7 2.8 3.3 3.3 18.2 38.6 50.3 24.5 4.5 4.7 3.1 3.0 1.7 1.4 2.3 0.9 20% 40% 39% 61% 26N 26S 40W 40M 40E 53 16 16 16 8 8 8 3.7 4.6 3.9 1.6 2.2 2.1 54.1 60.2 55.3 50.8 57.5 56.7 7.2 7.2 8.6 5.1 13.5 13.3 0.8 0.6 0.5 1.2 2.5 2.1 45.6 56.5 48.1 19.9 27.9 27.4 2.9 3.5 3.0 2.5 3.5 3.4 0.8 0.9 1.0 0.5 2.5 2.3 100% 100% 100% 100% 78% 68% na na na na na na na na na na na na Chapter 4: Logging damage 1 71 CNV Plot Reduced impact logging in the tropical rain forest of Guyana The median diameter of the felled trees per plot differed significantly between plots (Kruskal-Wallis ANOVA by ranks for unequal n; Hc (df = 17, n = 360) = 506.2, p < 0.001). The median felling diameter for experimentally logged plot 2 was significantly smaller than the ones for experimentally logged plots 3, 11 and 13, and conventionally logged plot 26S (non-parametric Tukey-type multiple comparisons for unequal n; q-values between 3.65 and 6.27). Furthermore, the median felling diameters for plots 7, 8, 9, 14 and 40M were significantly smaller than the ones for plots 13 and 26S (q-values between 4.07 and 5.36), and for plots 4 and 26N they were smaller than for plot 13 (q = 3.57 and q = 3.66 respectively). The contrast between the median felling diameters of plots 3 and 11, on the one hand, and plots 2, 7, and 14, on the other, is interesting, because this is associated with logging intensity. Increasing the intensity apparently led to the selection of smaller trees after relatively large trees had been depleted. This may confound the relationship between logging intensity expressed as numbers of trees felled per ha and felling impact. It also shows that the initial diameter frequency distribution of harvestable trees may have co-determined the felling impact. In order to investigate whether the mean felling diameter was associated with either logging intensity or logging method a regression model with three dummy variables was fitted: Y = a + b1 ⋅ X1 + b2 ⋅ X2 + b3 ⋅ X3 + b4 ⋅ X2 ⋅ X3; where Y is the mean felling diameter in a plot. Logging intensity is split into two dummy variables X1 and X2: the combination X1 = 1, X2 = 0 stands for intensity 4, X1 = 0, X2 = 1 stands for intensity 8 and X1 = 0, X2 = 0 stands for intensity 16. X3 is a dummy for logging method having the value 1 in case of conventionally logged plots and the value 0 otherwise. The independent variables were entered into the model one by one. The final model only included variable X1 (t16 = –2.49, p < 0.05). None of the other independent variables was significant. This means that the experimental plots with a logging intensity of 4 trees/ha have a significantly larger mean felling diameter than the other treatment combinations, but this is the only marked difference. Figure 4.1 suggests that the effect of intensity is qualified by the effect of the logging method, but this could not be proved (interaction term: t14 = –1.18, p = 0.25). Felling diameters are not distributed normally because by definition there is a set lower limit, while the upper limit is determined by what is both available and selected in the stand. Usually the felling diameter distribution is skewed to the right. It was indicated above that the average felling diameter decreased with increasing logging intensity. It is obvious that this is mainly caused by an increased asymmetry of the felling diameter distribution. If one fells more trees, one has to resort to smaller diameter trees because the larger diameter class becomes depleted. To test whether distribution asymmetry is related to logging intensity or to logging method the skewness of the distribution per plot is used. Skewness is a measure of the extent to which the distribution of the respective variable is skewed to the left (negative value) or right (positive value), relative to the standard normal distribution (for which the skewness is 0). A model as the one above with dummy variables was used. In this case not only the experimental plots with a logging intensity of 4 trees/ha were significantly different 72 Chapter 4: Logging damage (t13 = 3.94, p < 0.01), but also the interaction between intensity and method was significant (t13 = 3.51, p < 0.01). Increasing the intensity in reduced impact logging increased the skewness to the right, the contrary being true in conventional logging. 70 Conventional Reduced impact Felling diameter (cm dbh) 65 60 55 50 45 4 8 16 Logging intensity (trees/ha) Figure 4.1 Mean felling diameter per plot in reduced impact logging and conventional logging at different logging intensities at Pibiri, Central Guyana. Boxes indicate the standard error, whiskers the range. The mean felling diameter decreases with increasing intensity in reduced impact logging, whereas the opposite is found in conventional logging. This is caused by a difference in tree selection procedures and inherent non-homogeneity of rain forest samples. This is related to a sampling artefact due to the ‘Van Rompaey effect’ (Van Rompaey 1993) caused by the inherent non-homogeneity of rain forest samples. In case of experimental logging, plots were randomly assigned a logging intensity, and the distribution of the felling diameters depended on what was available in the plot. Now, the plots are heterogeneous by nature. Hence in cases where the stand was relatively poor, tree selection could not but include smaller diameters. In the case of conventional logging, however, by the nature of the tree selection procedure - the felling crew decided which trees were felled -, selection of the plots could only take place after logging. There the logging intensity was related to what was available in the stand and it was not necessary to include smaller diameters. In fact, in conventional logging a tacit stand selection preceded tree selection, thereby indirectly determining plot selection. Similar regression models were fitted for the mean volume per plot. This allows calculating the influence of tree height in conjunction with the influence of diameter. The results for mean tree volume per plot were similar to above: only plots with logging intensity of 4 trees/ha had higher mean tree volumes (t16 = –2.42, p < 0.05), while no differences were found for the dispersion around the means. As far as asymmetry is concerned, stronger differences were found. Asymmetry in the distribution of tree volumes differed between all logging intensities (X1: t13 = 3.21, p < 0.01; X2: t13 = –2.26, p < 0.05), when corrected for interaction between logging intensity and method (X2X3: t13 = 3.38, p < 0.01). 73 Reduced impact logging in the tropical rain forest of Guyana In summary, it appears to be necessary to include a co-variate related to size in models testing the effect of the different factors, not only because the mean sizes of the trees felled in the experimental plots with logging intensity of 4 trees/ha were significantly different, but also because plot 13 appeared to be an outlier (standard residual < 2.5 s, when comparing mean dbh). There were small differences in the results obtained when comparing mean values and those obtained comparing symmetry of the distributions. Also there were small differences in results obtained when comparing tree volumes and when comparing diameters. 4.5 LOSS OF CANOPY COVER 4.5.1 Accumulated loss of canopy cover Loss of canopy cover was strongly associated with logging intensity, but not with logging method and differences between the three intensities were significant (observed values are shown in Table 4.3, fitted regression model in Table 4.4). Because the fitted model showed that there was no effect from the logging method, a simpler regression model with logging intensity as a continuous, independent variable can be used to demonstrate the linear relationship between canopy loss and logging intensity: p′= 13.42 + 1.13 ⋅ (logging intensity); r2 = 0.79, p < 0.001 ( p ′ = arcsin (canopy loss) , p′ in degrees - Fig. 4.2). Logging intensity explained 79 % of the variation in canopy loss. Other possible sources of variation such as the mean diameter or the mean volume of felled trees per plot appeared to be non-significant when entered into the model of Table 4.4. Diameter: p = 0.83, R2-change = 0.001; Volume: p = 0.37, R2-change = 0.01. This can be partly explained by the complex relationship between logging intensity, logging method and felling diameter, but it also results from the few degrees of freedom and the fact that mean diameter and volume of felled trees per plot were used. Table 4.3 Mean loss of canopy cover as percentage of total area and per felled tree, gap(-segment) size and number of gap(-segment)s and standard deviations (sd) for conventional and reduced impact logging at different logging intensities. Gap(-segment) size is a measure of fragmentation. Gap(-segment) size considers gaps that are lying over the perimeter of a plot only for the part within. N = 3 in all cases except for reduced impact logging with intensity of 8 trees per ha: n = 6. Type of logging Conventional 8/ha 16/ha Reduced Impact 4/ha 8/ha 16/ha Loss of canopy cover (%) Mean gap Loss of canopy cover per extracted (segment) size (m2) 2 tree (m ) Number of gaps (segments) per hectare 15.8 (1.4) 24.5 (1.6) 198 (18) 153 (10) 264 (23) 439 (76) 6.0 (0.0) 5.7 (0.8) 8.5 (2.8) 15.7 (3.6) 30.1 (7.2) 208 (68) 192 (44) 184 (44) 185 (59) 209 (65) 333 (71) 4.6 (0.5) 7.7 (1.1) 9.0 (0.3) 74 Chapter 4: Logging damage A better way of including the effect of the variation in size of the felled trees is to examine the relationship between the felled basal area and the extent of canopy loss directly. This revealed a significant effect of logging method (Table 4.5, Fig. 4.3). At a moderate logging intensity, canopy loss did not differ between conventional and reduced impact logging, but when the intensity is increased reduced impact logging resulted in greater loss of canopy cover. This may have been caused by the fact that in experimental logging emergent trees were felled at the high intensity while this was not the case with conventional logging. However, adding the felling of emergents as a dummy variable proved to be redundant (t13 = 0.70, p = 0.49). Table 4.4 Fitted coefficients for the regression model: Y = a + b1X1 + b2X2 + b3X3 + b4X2 X3 + ε; where Y = percentage canopy loss (arcsin฀p - transformed), X1 and X2 are dummies for logging intensity and X3 is a dummy for logging method. X1 = 1, X2 = 0 if a plot had a logging intensity of 4 trees per hectare, X1 = 0, X2 = 1 in case of intensity 8 and X1 = 0, X2 = 0 in case of intensity 16. X3 has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. R² = 0.83, F4,13 = 15.82, p < 0.001. Standardised regression coefficients (β), estimated coefficients (b), Student’s t-statistic and probability level. Parameter Intercept (a) Logging method (b3) Intensity = 4 (b1) Intensity = 8 (b2) Interaction (b4) Figure 4.2 β -0.29 -1.06 -0.87 0.25 b 33.19 -3.54 -16.40 -9.99 3.79 t13 20.56 -1.55 -7.18 -5.05 1.25 p-level 0.000 0.145 0.000 0.000 0.232 The relation between logging intensity and loss of canopy cover for conventional and reduced impact logging in Pibiri, Central Guyana. Canopy cover percentages were angularly transformed. There is no significant difference between conventional and reduced impact logging. See text for regression equations. As shown above, variation in forest structure and composition had been observed before logging. What has been the ‘Van Rompaey effect’ - the non-homogeneity of rain forest samples (Van Rompaey 1993) - of this variation on canopy loss, to be precise: 75 Reduced impact logging in the tropical rain forest of Guyana the non-homogeneity of the architectural forest mosaic prior to logging? This could only be examined for the experimentally logged area. Table 4.5 Fitted coefficients for the regression model Y = a + b ⋅ X, where a = a1 +a2 ⋅ [M], and b = b1 +b2 ⋅ [M]. Y = percentage canopy loss (arcsin฀p - transformed), X = basal area felled, [M] is a dummy for logging method, and ai and bi are fitted coefficients. [M] has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. *** denotes that the probability of b = 0 is smaller than 0.001. Slopes and intercepts differed significantly between methods (t14 = –2.79, p < 0.05; and t14 = 2.18, p < 0.05). Logging method Conventional Reduced impact Figure 4.3 Fitted coefficients a b 18.04 *** 2.80 *** 9.88 *** 6.38 *** n R2 6 12 0.92 0.82 The relation between basal area felled and loss of canopy cover for conventional and reduced impact logging in Pibiri, Central Guyana. Canopy cover loss percentages were angularly transformed. Loss of canopy cover decreases relatively with increasing intensity for conventional logging. See Table 4.5 for regression equations. Analysis of covariance was performed in order to adjust for differences in total standing basal area before felling and difference in the forest mosaic, particularly the presence of eco-units in state of aggradation (sensu Oldeman 1990, terminology by Bormann & Likens 1979) before felling. My aim in conducting analysis of covariance is not to increase precision, because we know already that the effect of logging intensity is highly significant and more precision would be meaningless to this understanding. Rather, my aim is to extract the noise caused by co-variates causing interference. Analysis of covariance requires the slopes of the regression lines being fitted in each group to be parallel. This condition was fulfilled for the basal area (F2,6 = 0.046, p = 0.95) and for the proportion of building phase patches (F2,6 = 0.16, p = 0.86). 76 Chapter 4: Logging damage The effect of total basal area per plot prior to logging was non-significant (F1,8 = 2.19, p = 0.18), but the presence of natural gaps had a significant effect. This can be explained as follows. In reduced impact logging it was tried to avoid that the felled tree would lodge in a neighbouring canopy tree. This reduced the likelihood that neighbouring canopy trees were uprooted. During its fall, however, it would crush or uproot smaller, subcanopy trees. Hence, if a tree were directed towards an aggrading patch, the regrowth there would be crushed. Because our gap definition used the dominant tree layer to determine its edge, it is likely that if a tree fell in a patch of regrowth this patch would be included in the felling gap. On the contrary, if the tree fell under neighbouring canopy trees, the obliterated smaller vegetation would not be included in the felling gap. Analysis of covariance showed that the effect of logging intensity becomes more pronounced after adjustment for the proportion of aggrading patches of the total forest matrix (F-value for adjusted means F2,8 = 31.45 compared to F-value among groups F2,9 = 19.15). Calculated adjusted means were 8.1%, 15.2%, 30.8% for the respective intensities (see Table 4.3 for raw means). These parametric values show that the rise in the loss of canopy cover was even steeper after removal of the effect of the disturbing variable than before; i.e. if there would not have been differences in the proportion of aggrading patches, canopy loss is even more dramatic in high intensity reduced impact logging. Unfortunately, there was no information on the proportion of aggrading patches in the conventionally logged area. Table 4.6 Fitted coefficients for the regression model Y = a + b ⋅ X1 + c ⋅ X2, where a = a1 + a2 ⋅ [M], and b = b1 + b2 ⋅ [M]. Y = percentage canopy loss (arcsin฀p - transformed), X1 = basal area felled, X2 = proportion aggrading patches before logging (arcsin฀p - transformed), [M] is a dummy for logging method, and ai, bi and c are fitted coefficients. [M] has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. Stars denote the probability of coefficients being 0: * p < 0.05, ** p < 0.01, *** p < 0.001. Slopes differed significantly between methods (t13 = –2.385, p < 0.05), but intercepts did not (t14 = 1.917, p = 0.077). Logging method Conventional Reduced impact a 3.98 *** –2.10 ns Fitted coefficients b 0.78 ** 1.41 *** c 0.51 na 0.51 * n R2 6 12 0.93 0.89 According to the pre-harvest inventory, the forest in the conventionally logged area had a higher basal area than most of the experimentally logged plots (Table 4.1). There is some indication that standing basal area is negatively correlated with the proportion of building phase patches (r = –0.56, t10 = –2.134, p = 0.058). Field observations confirm this. Substituting the missing values with means, is therefore an acceptable approximation. The overall regression equation then becomes: p’ = 1.07 + 1.19 ⋅ (logging intensity) + 0.45 ⋅ (percentage natural gaps, angularly transformed); R2 = 0.85, p < 0.001). After these adjustments, it appears that reduced impact logging also leads to greater canopy loss without checking for basal area felled (Table 4.6). The effect of the disturbing variable is visualised in Figure 4.4. The slope of the regression 77 Reduced impact logging in the tropical rain forest of Guyana line for the plots with a more than average percentage of aggrading patches before logging does not differ from the one for the plots with a less than average percentage. This implies that with my definition of gaps the effect of logging intensity is not modified by the presence of natural gaps. Figure 4.4 The relation between basal area felled and loss of canopy cover for conventional and reduced impact logging, qualified by the proportion of building phase patches before logging. Canopy cover percentages are angularly transformed. Two regression lines have been fitted for the experimentally logged plots: one for those with a more than average proportion of aggrading patches (p′ = 10.33* + 7.12** ⋅ (basal area felled), n = 6, R2 = 0.91), and one for those with less than average (p′ = 7.42ns + 6.60** ⋅ (basal area felled), n = 6, R2 = 0.91). Intercepts differ significantly (t8 = 2.84, p < 0.05), but slopes do not (t8 = 0.22, p = 0.83). Regression line for conventional is the same as in Figure 4.3. With increasing logging intensity, gaps formed in the experiment overlapped less often than gaps formed with conventional logging. In the conventional operation the average canopy opening per felled tree decreased from 198 m2 (sd = 18 m2) to 153 m2 (sd = 10 m2) when the intensity was increased from 8 to 16 trees/ha, whereas this reduction was minimal, from 208 m2 (sd = 68 m2) to 192 m2 (sd = 44 m2) to 185 m2 (sd = 44 m2), when increasing the intensity from 4 to 8 to 16 trees/ha in the reduced logging operation. Since this cannot be ascribed to any disturbing variables, the cause must lie either in the distribution of the cut trees over the area or in the felling directions. According to the tree selection method, stumps are indeed unclustered in the experiment, against clustered in conventional logging. This can be demonstrated by examining the average distance to the nearest neighbouring stump. Both the method and the intensity of logging had a marked effect on this distance (Table 4.7). In the conventionally logged area, the average distance to the nearest stump was about 14 m regardless of logging intensity. This indicates that trees were consistently felled in clusters. In the experiment, the average distance decreases linearly with increasing logging intensity due to the fact that stumps were more evenly spaced. The average nearest neighbour distances would converge at a logging intensity of 20 trees/ha. If 78 Chapter 4: Logging damage clustering were the decisive cause of overlapping gaps, one would expect to find the effect of logging method to decrease with increasing intensity. In reality, the contrary was found. This leaves one explanation only, which is that trees were purposely felled in the same gaps in the conventionally logged area, whereas this was not the case in the experiment. This is consistent with felling in a herring-bone pattern and with field observations. Table 4.7 Fitted coefficients for the regression model Y = a + b ⋅ X, where, a = a1 + a2 ⋅ [M] and b = b1 + b2 ⋅ [M]. Y = average distance to nearest neighbouring stump, X = logging intensity (trees/ha), [M] is a dummy for logging method, and ai and bi are fitted coefficients. [M] has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. Stars denote the probability of coefficients being 0: ns p < 0.05, ** p < 0.01, *** p < 0.001. Intercepts and slopes differ significantly (t14 = –5.35, p < 0.001; t14 = 3.16, p < 0.01 respectively). Logging method Conventional Reduced impact Fitted coefficients a b 13.99 ** -0.19 ns 40.92 *** -1.52 *** n 6 12 R2 0.18 0.78 4.5.2 Size of canopy openings In the section before, it was shown that loss of canopy cover indeed increases with increasing logging intensity. However, as a counterintuitive phenomenon, it was also shown that reduced impact logging resulted in a greater loss when the intensity was increased above a level of 8 trees per hectare. In Section 2.2, the ecological implications of large gaps were demonstrated to be quite different from the ones of small gaps. Besides the size of the gaps, the shape and exposure of the gap play an important role, because they determine the properties of the border zones (e.g. Oldeman 1990; Rossignol et al. 1998). In section 4.2.2, it was explained that the choice of comparing canopy loss over a geometrically delimited area had advantages but also disadvantages. One of the disadvantages is that the effect of the type of logging operation on gap size can not consider all canopy openings. This is so, because there will also be gaps that cross the boundary line. Including these gaps and measuring them entirely will indeed produce an unbiased mean gap size, but the link to the reference area is lost, making it impossible to relate the calculated mean to a certain logging intensity. Excluding these gaps will lead to a smaller number of observations, thereby increasing the standard error of the mean. Moreover, it will give a biased estimate of the average gap size, because the likelihood for small gaps to be fully enclosed within the plot boundary is greater than for large gaps. However, the size distribution of fully incorporated gaps together with the segments of the gaps that were overlying the perimeter of a plot characterise the coarseness of the canopy texture or mosaic (Heil 1998, pers.comm.). Therefore, I included both the gaps that were fully incorporated and of the gaps that crossed the perimeter, only the segments that were lying within the plot perimeter to calculate an average gap(-segment) size. The thus calculated mean ‘felling gap size’ seemed larger in conventional logging (351 m2, sd = 108 m2, n = 6) than in reduced 79 Reduced impact logging in the tropical rain forest of Guyana impact logging (234 m2, sd = 85 m2, n = 12), but not significantly so (ANOVA on logarithmically transformed plot means, Type IV sum of squares: F1,13 = 3.75, p = 0.075). Figure 4.5 The relation between felled basal area per hectare and mean ‘gap’ size for conventional and reduced impact logging. Mean gap size is based on gaps fully incorporated in the plots and segments within a plot’s perimeter of those gaps that crossed a plot’s boundary. Regression model: Y = a + b ⋅ X + ε, where, a = a1 + a2 ⋅ [M] and b = b1 + b2 ⋅ [M]. Y = mean ‘gap’ size (m2), X = basal area felled (m2/ha), [M] is a dummy for logging method, and ai and bi are fitted coefficients. [M] has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. The coefficients a2 and b2 did not differ significantly from zero (t15 = 1.94, p = 0.07 and t14 = 0.45, p = 0.66 respectively), resulting in the following model: Y = 73.34 + 79.97 ⋅ X; R2 = 0.67, F1,16 = 32.24, p < 0.001. The effect of logging intensity was significant (ANOVA on logarithmically transformed plot means, Type IV sum of squares: F2,13 = 7.21, p < 0.01). The mean ‘felling gap size’ with a logging intensity of 4 trees per hectare (185 m2, sd = 59 m2, n = 3) did not differ significantly from the one with an intensity of 8 trees per hectare (227 m2, sd = 60 m2, n = 9), but both had significantly (Tukey’s HSD, p < 0.01) smaller mean ‘gap’ sizes than 16 trees/ha (385 m2, sd = 88 m2, n = 6). Possible other sources of variation had no effect on the calculated mean ‘gap’ size: ƒ the mean diameter of the felled trees per plot (F1,12 = 0.65, p = 0.44); ƒ the mean volume of the felled trees per plot (F1,12 = 1.18, p = 0.30); ƒ the presence of natural gaps prior to logging (F1,8 = 2.146, p = 0.18), which is surprising because it did influence the accumulated loss of canopy cover; ƒ total basal area per plot prior to logging (F1,8 = 2.485, p = 0.15). In Section 4.5.1, it was shown, that using the felled basal area directly as an independent variable was more successful in correcting for the effect of the variation in size of the felled trees than using the mean diameter or volume per plot as a co-variate. The same approach was also used to determine whether or not gap sizes differed 80 Chapter 4: Logging damage between the two logging methods. However, in contrast to the accumulated loss of canopy, no such difference was found for gap sizes (Fig. 4.5). The fact that mean ‘gap’ sizes did not differ between the logging methods does not exclude that the coarseness of the architectural forest mosaic differs between them. Just as with the diameter distribution of the felled trees, the size class distribution of the gaps may differ in terms of dispersion or symmetry. Gap sizes were grouped into four size classes: smaller than 100 m2, from 100 to 250 m2, from 250 to 500 m2, and 500 m2 and larger. The size class frequency distributions were compared using G-test for independence. The gap size distributions differed between logging methods (partial association G = 8.24, df = 3, p < 0.05), and between logging intensities (partial association G = 9.45, df = 3, p < 0.05), but there was no interaction between these two factors (three-factor interaction G = 1.09, df = 3, p = 0.77). In order to find out in which size classes the frequencies differed, the test for independence was partitioned into three orthogonal comparisons (Sokal & Rohlf 1995, p. 755-757). First, the four size classes were condensed to two classes: < 250 m2 versus ≥ 250 m2). This condensed size class distribution differed according to logging intensity (partial association G = 5.04, df = 1, p < 0.05) as well as logging method (partial association ∆G = 4.02, df = 1, p < 0.05). Subsequently, the frequencies in the two smaller size classes that were condensed in the class < 250 m2 were examined separately. There were no patterns in these two classes associated with either logging intensity of method (G = 3.46, df = 4, p = 0.48). Finally, the frequencies in the two larger size classes were examined separately. Comparing the size classes 250 to 500 and ≥ 500 m2 did reveal an association with logging intensity (partial association G = 4.68, df = 1, p < 0.05), but not with logging method (partial association G = 1.80, df = 1, p = 0.18). This means that the effect of logging intensity is mainly reflected in a difference in the proportional frequency of gaps in the size classes ≥ 500 m2, whereas the effect of the logging method is mainly reflected in a difference between the frequencies in the size classes <250 m2 and >250 m2 (Fig. 4.6). The fact that there was no interaction between logging method and intensity implies that the differences between logging methods were consistent regardless of logging intensity and vice versa. Figure 4.6 suggests that the difference between the logging methods is mainly reflected in a difference in frequency in the 100 to 250 m2 and 250 to 500 m2 size classes. A test of independence comparing these two classes only indeed showed a significant difference between logging methods (partial association G = 7.88, df = 1, p < 0.01). This last comparison, however, is not orthogonal to the former three, so that the test is not independent and the test may have been too sensitive. The logging intensity of 4 trees/ha could not be included in the analysis of the above three-way contingency table because it did not occur in conventional logging. In order to find out whether the gap size distribution at a logging intensity of 4 trees/ha differed from the other intensities, a test of independence was carried out for the experimentally logged plots only. As expected, the association between logging intensity and gap size distribution was significant (G = 15.77, df = 6, p < 0.05). In order to detect differences between the levels of intensity, the test for conditional independence was decomposed 81 Reduced impact logging in the tropical rain forest of Guyana % of gaps % of gaps into two comparisons. First, the intensities 4 and 8 trees/ha were pooled and compared with 16 trees/ha. This had no influence on the conditional independence (G = 13.13, df = 3, p < 0.01). Next, the intensities 4 and 8 trees/ha were compared, a comparison which produced no significant difference in gap size class distribution between these two intensities (G = 3.67, df = 3, p = 0.30). The gap size frequency distribution after logging with 4 trees/ha does thus not differ from the distribution with 8 trees/ha, while a logging intensity of 16 trees per hectare led to a different distribution; i.e. more large gaps. gap size class (m2) Figure 4.6 The percentage of gaps(-segments) in each gap size class in conventionally logged forest (open bars) and in experimentally logged plots (hatched bars) at different logging intensities in Pibiri, Central Guyana. Data are pooled for three replicates of 2-ha (except reduced impact logging with eight trees/ha: six replicates). The percentages at the top of each bar represent the proportion of the total accumulated gap area. Note: gaps that were crossing the boundary of the reference area were measured up to the boundary only. Recapitulating, few large gaps ( < 500 m2) occurred at both 4 and 8 trees/ha; i.e. 4% and 9% (no significant difference) of the gap frequency respectively. At an intensity of 16 trees/ha this frequency steeply and significantly increased to 25% of the gaps. In terms of proportions of total gap area, however, these frequencies account for 18%, 32% and 63%. This indicates a linear relationship with the number of trees felled per hectare. The effect of the reduced impact logging method is shown by a proportional shift of gaps from the size class 250 to 500 m2 to the size class 100 to 250 m2. Ignoring the intensity of 4 trees/ha and pooling the remaining intensities, reduced impact logging decreased the proportion of gaps in the size class 250 to 500 m2 from 29% to 16% of total gap frequency, whereas the proportion of gaps in the size class 100 to 250 m2 82 Chapter 4: Logging damage increased from 24% to 42% of total gap frequency. As to the relative total gap area, the proportion of gaps in the size class 250 to 500 m2 reduced from 34% to 21% of the total gap area and the proportion of gaps in the size class 100 to 250 m2 increased from 12% to 20%. The occurrence of large gaps (≥ 500 m2) was slightly reduced, although not significantly so; from 17% of the total gap frequency, referring to 50% of total gap area, in case of conventional logging, it descended to 15% of the total gap frequency, referring to 49% of total gap area, in case of reduced impact logging. It should be remembered that the above ‘gap’ size distributions not always refer to entire gaps, but in some case to segments of gaps (in case a gap crosses the plot boundary). 4.5.3 Sizes of single and multiple treefall chablis The mean ‘gap’ sizes reported above do not always refer to entire gaps and therefore underestimate the actual mean gap sizes. In the in all 69 ha experimentally logged plots 171 gaps were found, 103 of which were fully included by the plot boundaries. In the 32 hectares conventionally logged area 70 gaps were found. These gaps were measured in two ways: the part of the gap that fell inside the plot boundary was measured and additionally, the total gap area, thus including the part lying outside the plot boundary. For each gap, it is known how many trees had contributed to the formation of a particular gap. Trees sometimes formed more than one opening in the canopy; i.e. the place where the crown of the tree was removed from the canopy and the place where the crown of the tree landed - the epicentre of the treefall (Oldeman 1978). At the epicentre, an opening may be formed because trees standing there are uprooted or broken. In case of multiple tree fall gaps epicentres may coincide or the crown of one tree may land near where the stump of another tree is located (cf. Mutoji-A-Kazadi 1977; Oldeman 1978; Florence 1981). These openings are usually separated by only a few remnant trees. Because the effect of a canopy opening extends beyond its projected outline (see Section 2.2) combining these neighbouring openings is a more useful measure to compare the long-term effects of the impact of logging than comparing sizes of separate canopy openings. The total area affected by a tree fall is called a chablis. This is a traditional French term that includes all changes to the architectural forest mosaic initiated by the fall of one or more trees (Oldeman 1990, p. 160): ‘the uprooting of a tree, the uprooted tree, the inaccessible heap of broken and surviving vegetation and branches, and the opening (gap) in the forest canopy’. From now on, I will use the chablis concept - although not in its full original sense - in order to relate the felling of a particular tree or set of trees felled in one spot to the (combined) canopy opening(s) formed by that event. The 171 gaps in the experimentally logged area belonged to 144 chablis, 79 of which were measured entirely. The 70 gaps in the conventionally logged area pertained to 49 chablis. The proportion of chablis formed by a single, two, three or more trees was associated with the logging method (G-test for conditional independence, plots pooled per treatment, intensity 4 trees/ha excluded; G = 21.97, df = 3, p < 0.001). The number of single, dual and triple or more treefalls also differed along with the logging intensity (G = 26.62, df = 3, p < 0.001), while there was no interaction between the logging 83 Reduced impact logging in the tropical rain forest of Guyana % of all chablis % of all chablis method and intensity (G-test for three-way interaction: G = 2.35, df = 3, p = 0.50). Reduced impact logging resulted in a higher percentage of single treefall chablis and lower percentages of multiple treefall chablis (Fig. 4.7). At the lowest logging intensity, which only occurs with reduced impact logging, chablis with more than two felled trees did not occur. At an intensity of 8 trees/ha, less than 2% of the chablis contained more than 3 felled trees in reduced impact logging, whereas 15% of the chablis fell into this class in conventional logging. At the highest intensity, 27% of the chablis fell into this class with reduced impact logging, and as much as 54% with conventional logging. Number of felled trees in one chablis Figure 4.7 The percentage of chablis caused by felling one, two, three or more trees in a conventional logging operation (open bars) and in a reduced impact logging operation (hatched bars) at different logging intensities in Pibiri, Central Guyana. Whiskers denote the standard deviation. In the conventionally logged areas, as many as 5 felled trees were felled were found in one chablis at an intensity of 8 trees/ha and as many as 17 trees at an intensity of 16 trees/ha. With reduced impact logging, one chablis contained as many as 4 felled trees at an intensity of 8 trees/ha and as many as 10 at an intensity of 16 trees/ha. The mean area of all chablis classes considered together was significantly larger (ANOVA on logarithmically transformed data, data pooled per treatment, type IV sum of squares, F1,123 = 9.99, p < 0.01) in the conventional operation (582 m2, sd = 529 m2, n = 49) than in the experimental operation (312 m2, sd = 362 m2, n = 79). There was no difference between intensities (ANOVA, F2,123 = 1.60, p = 0.21). However, the estimate for reduced impact logging is seriously biased, because, only those chablis were considered that were lying completely inside the boundary of a plot. As mentioned earlier, the likelihood for small chablis to be fully included is larger than for 84 Chapter 4: Logging damage large chablis. The appropriate way to compare the treatments is to investigate the area within each chablis class. The mean chablis size for a single treefall chablis; 181 m2 (sd = 140, n = 71 - Fig. 4.8), was not influenced by the logging method (Mann-Whitney U15,56 = 409, p = 0.87). The mean size for chablis with two treefalls seemed lower for reduced impact logging - 355 m2 (sd = 130, n = 13), than for conventional logging - 444 m2 (sd = 169, n = 12), but this difference was not significant (Mann-Whitney U12,13 = 52, p = 0.16). For chablis with more than two tree falls per chablis, the areas seemed larger with experimental logging; three treefalls: 637 m2 (sd = 208 m2, n = 6), more than three treefalls: 1504 m2 (sd = 642, n = 4), than with conventional logging; three treefalls: 548 m2 (sd = 191 m2, n = 6), more than three treefalls: 1076 m2 (sd = 642 m2, n = 16). These differences, however, were not statistically significant either (Mann-Whitney U6,6 = 12, p = 0.34, and Mann-Whitney U4,16 = 18, p = 0.19, respectively). In summary, it is the number of trees that together formed one chablis that determines the size of that chablis, whereas the way the trees are felled hardly influences the size, at least not significantly so. Figure 4.8 Mean area of chablis caused by felling one, two, three or more trees in a conventional logging operation (open bars) and in a reduced impact logging operation (hatched bars), in Pibiri, Central Guyana. Whiskers denote standard error of the mean. The mean chablis area per chablis class did not differ significantly between logging methods for any of the four classes. Multiple tree fall chablis occurred, however, more frequently in conventional logging (Fig. 4.7). Let us now regard the chablis classes (number of trees in one chablis) as being part of a stratified population, the composition of which was established after a first-phase sample (post-stratification), while corresponding chablis sizes were recorded in a smaller, second-phase sample. Unbiased estimates of the chablis sizes then can be obtained with a sample procedure referred to as double sampling with stratification (De Vries 1986). The estimated chablis areas that were obtained this way are markedly smaller with reduced impact logging than with conventional logging both at a logging intensity of 8 and of 16 trees/ha (Table 4.8). In reduced impact logging the chablis size 85 Reduced impact logging in the tropical rain forest of Guyana remained below 300 m2 as long as the logging intensity did not exceed eight trees per hectare. Ek & Van der Hout (in prep.) concluded that 300 m2 is the critical gap size above which size light-demanding pioneer species are likely to invade. At an intensity of 16 trees/ha, the mean chablis size doubles in reduced impact logging. In the conventional operation, the mean chablis size is about 60% larger than in reduced impact logging at an intensity of 8 trees/ha and about 40% at an intensity of 16 trees/ha. Table 4.8 Estimated mean chablis areas and standard deviations (sd) and estimated individual canopy opening areas for conventional and reduced impact logging at different logging intensities. Chablis areas were estimated by double sampling with post-stratification. The first phasesample (n1), including all encountered chablis, was stratified according to chablis class (defined by the number of felled that formed one chablis). In the second-phase sample (n2), both the chablis class and the chablis area were recorded. The estimated mean canopy opening was determined by multiplying with the average number of canopy openings per chablis for each category. Type of logging Conventional 8/ha 16/ha Reduced impact 4/ha 8/ha 16/ha Chablis area (m2) mean sd n1 425 775 56 145 27 22 249 263 566 53 27 154 22 81 41 n2 Canopy opening area (m2) mean 319 502 16 46 17 190 216 370 Chablis were often composed of multiple canopy openings. The estimated mean canopy opening size showed a similar trend as chablis size (Table 4.8). Comparison of these estimates with the mean of the canopy openings and segments of openings within the plot boundary (Table 4.3), shows that the mean canopy opening estimated by using double sampling is only slightly larger. 4.5.4 Single treefall chablis, tree diameter and climber cutting Multiple tree fall gaps often overlap. Such gaps overlapped to a larger extent in conventional logging, because the felled trees were often clumped. Multiple tree fall gaps can thus not be used to investigate the effect of climber cutting and directional felling on the loss of canopy cover. In section 4.5.3, it was shown that the size of a single treefall chablis was not influenced by the logging method. As demonstrated in section 4.5.1, the size of the felled trees had an influence on the total loss of canopy cover. The absence of a difference in single tree fall gap size between the logging methods may well be related to a difference in tree size. As expected, the size of a single treefall chablis was indeed strongly related to the diameter of the felled tree, but not to its species (Table 4.9, Fig. 4.9). Whether or not a tree was emergent, at first glance, seemed to affect this relation, but this can not be substantiated due to the small number of observations (Table 4.9). 86 Chapter 4: Logging damage Table 4.9 Single treefall chablis: relation between tree diameter, species and chablis area. Fitted coefficients for the regression model Y = a + b ⋅ X, where a = a1 + a2 ⋅ [G] + a3 ⋅ [E], and b = b1 + b2 ⋅ [G] + b3 ⋅ [E]. Y = chablis area (m2), X = dbh of felled tree (cm), [G] is a dummy for Greenheart, [E] is a dummy for emergent species, and ai and bi are fitted coefficients. [G] and [E] have the value 1 if an individual was a Greenheart or an emergent species, respectively, and 0 if it did not. Stars denote the probability of coefficients being zero: ns p < 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001. The results show that there was a significant response to tree diameter (t65 = 3.01, p < 0.01; model R2 = 0.55). Emergent species had a significantly lower intercept (t65 = - 2.61, p < 0.05) and significantly steeper slope (t65 = 2.23, p < 0.05). The difference between the regression lines was due to one outlying case (Fig. 4.9). Removal of this outlier eliminated the statistical differences between species. Species (group) Greenheart Emergent species Other species Figure 4.9 Fitted coefficients a b -228.5 ** 7.59 *** -968.1 ns 18.20 * 8.46 ** -245.1 ns n R2 41 7 23 0.40 0.61 0.33 The relation between chablis area and diameter of felled trees in single treefall chablis for three species groups. Chablis in conventional and reduced operations were pooled. See table 4.9 for regression equations. Entering logging method into the model of table 4.9 did not reveal any effect of climber cutting or directional felling. One may argue that the period of four to seven months between climber cutting and felling was too short. Indeed, field observations during felling confirmed that in some cases the climber stems had not weakened sufficiently. The moderate amount of variation explained by the diameter of the felled trees indicates that there are other sources of variation; for instance, the effect of the presence of natural gaps, which certainly exists but has not been confirmed yet (cf. Fig. 4.4). Also, there may have been differences between the areas in the density of lianas (unconfirmed). Otherwise, only 15 chablis were available for this analysis in the conventional operation, compared to 56 observations in the reduced logging operation. Effects of climber cutting are not as self-evident as they may seem at first sight. 87 Reduced impact logging in the tropical rain forest of Guyana Specific studies will be necessary, taking into account the specific biological features of lianas (cf. Oldeman 1990:119-124). 4.6 EXTENT OF DISTURBED GROUND AREA 4.6.1 Skid trails and traces of other skidder movements Reduced impact logging indeed substantially diminished the amount of ground disturbance caused by machine movements, as planned. At a logging intensity of 8 trees/ha, the skidder in the conventional operation affected an area almost twice that found in the reduced impact operation and at an intensity of 16 trees/ha it was more than twice (Table 4.10; Fig. 4.10). The logging intensity also influenced the affected area, but not as clearly. Table 4.10 Mean ground area (m2) affected per extracted tree and as percentage of total area during actual skidding and other movements in conventional and reduced impact logging operations at different logging intensities, and standard deviation (sd) in Pibiri, Central Guyana. Results are based on three replicates of 2-ha, except reduced impact logging with 8 trees/ha: six replicates. Other skidder movements include manoeuvring in the bole zone to position logs for hooking and manoeuvring along trails to position logs for bundling. Type of logging Conventional 8/ha 16/ha Reduced impact 4/ha 8/ha 16/ha Ground area affected Per tree extracted (m2) Percentage of total area (%) Trails Other Total Trails Other Total movements movements 103 (13) 62 (9) 59 (27) 67 (13) 161 (16) 129 (13) 8.2 (1.1) 9.9 (1.4) 4.7 (2.2) 10.8 (2.1) 12.9 (1.2) 20.7 (2.1) 121 (11) 93 (34) 51 (21) 3 (2) 5 (5) 3 (1) 124 (10) 98 (37) 54 (21) 4.9 (0.5) 7.6 (2.7) 8.3 (3.4) 0.1 (0.1) 0.4 (0.4) 0.5 (0.1) 5.1 (0.4) 8.0 (3.0) 8.1 (3.4) The differences due to the logging method can be mainly attributed to an almost complete reduction of damage incurred while positioning logs for hooking. This is in part because of the use of the winch. The winch was employed in 77% of the observed skidder hook-ups (n = 179) and allowed the skidder to remain, on average, 11.1 m from the hook-up point (sd = 8.5 m). The maximum observed winch distance amounted to 38.0 m. Omitting the cases where the logs were located along the trail (9% of the observed hook-ups), increases the employment of the winch to 84% of the cases with an average winch distance of 12.2 m (sd = 8.2 m, n = 163). Winching allowed the machine to stay on the designated trail in 30% of the observed cases. The feasibility of winching depends strongly upon the felling pattern. Another reason for the elimination of damage incurred during hooking was that shifting and lifting of the bole with the machine blade - needed for hooking in the conventional operation - was made redundant by the use of special equipment and the favourable position of the logs due to the improved felling pattern (see Section 3.3.5). In the experimental operation, skidding damage was thus not linked to the hooking of logs at stump, but rather to 88 Chapter 4: Logging damage manoeuvres along the trails, e.g. for bundling logs. In the conventional operation, the skidder constructed its trail without any overview over the stump area; largely following in the feller’s footsteps (see Section 3.2.2). This resulted in winding trails and sometimes unnecessary loops were formed (Fig. 4.11). Figure 4.10 Percentage of ground area affected in conventional (CNV) and reduced impact logging (RIL) operations at different logging intensities in Pibiri, Central Guyana. Affected ground area is subdivided into skid trails and other skidder movements. Other movements include manoeuvring in the bole zone to position logs for hooking and manoeuvring along trails to position logs for bundling. In the reduced impact operation a distinction can be made between planned (designated) and unplanned (branch) trails. Total affected area is significantly higher in conventional logging (ANOVA on arcsin฀p - transformed data, Type IV sum of squares: F1,13 = 29.95, p < 0.001). Logging intensity also significantly influenced the affected area (F2,13 = 3.85, p < 0.05). In the reduced impact operation, trails followed a planned, demarcated alignment and, in addition, some branch trails were made. Together, these trails produced a herringbone pattern (Fig. 4.12). Skidtrails in the conventional operation were significantly wider than in the experimental operation (Student’s t-test, t16 = −5.12, p < 0.001), with a width of 4.0 m (sd = 0.3, n = 6) as opposed to 3.5 m (sd = 0.1 m, n = 12). The designated trails in the reduced impact operation were significantly wider than the additional branch trails (Student’s t-test, t22 = 3.08, p < 0.01), with a width of 3.6 m (sd = 0.2 m) as opposed to 3.4 m (sd = 0.2 m). The area affected by trails alone, however, did not differ significantly between the two logging methods (Fig. 4.10). In the reduced impact operation, designated trails ran parallel at a spacing of 80 m on flat to undulating terrain. In case of more dissected terrain (slopes exceeding 10%) the alignments were adjusted to the topography and followed the water divides. The total area affected by trails appeared to be significantly smaller when the trails could run parallel to each other (Student’s t-test, t10 = −2.836, p = 0.018). 89 Reduced impact logging in the tropical rain forest of Guyana CNV - 16/ha CNV - 8/ha Figure 4.11 Conventional logging. Canopy opening and ground area damage due to conventional logging at two logging intensities on flat to undulating terrain at Pibiri, Central Guyana. Canopy openings are hatched with forward slashes, ground disturbance with backward slashes. Stumps are indicated by circles and the transport direction by an arrow. Note the skid trails extending from stump to stump and the frequent overlap of canopy openings and ground level disturbance. Both are associated with the skidder positioning each log with its blade, in order to hook it, which in turn is related to the criss-cross felling pattern. Separating plots with parallel trails from the ones with trails that were adjusted to topography, revealed that reduced impact logging in terrain that allowed parallel trails, led to a 40% decrease in area occupied by trails compared to conventional logging. In case the terrain dictated an adjustment of the alignment the area occupied by trails was increased by 6% (Table 4.11, Fig. 4.13). Different trail patterns influenced the total affected ground area as follows. In case parallel trails could be realised, reduced impact logging reduced the total affected ground area from 13% to 5% at 8 trees/ha and from 21% to 7% at 16 trees/ha; a reduction with 64% on average. In case the trail pattern was dictated by steep slopes the total affected ground area was reduced to 10% and 13% respectively; a reduction with 40% on average. The terrain in the conventionally logged area was flat to gently undulating, so only the first comparison - a reduction of 64% on average - is relevant. 90 Chapter 4: Logging damage RIL - 8/ha RIL - 16/ha Figure 4.12 Reduced impact logging. Canopy opening and ground area damage due to reduced impact logging at two logging intensities on flat to gently undulating terrain at Pibiri, Central Guyana. Black streaks indicate stretches over which was winched. If a stump is not ‘connected’ to either a winch track or a skid trail, the bole was removed by hooking the top instead of the butt end. Note the branch trails and winch tracks making an angle of more or less 135° with the transport direction resulting in a ‘herring-bone’ skidding pattern. This pattern was made possible by a similar felling pattern. Note also the even spacing of stumps in comparison with the clustered distribution in Fig. 4.11. Table 4.11 Effect of skid trail pattern and logging method on the area occupied by trails. Fitted coefficients for the regression model Y = a + b ⋅ X + ε, where a = a1 + a2⋅[M] + a3⋅[T], and b = b1 + b2⋅[M] + b3⋅[T], where. Y = area occupied by skid trails as percentage of total area (arcsin฀p transformed), X = logging intensity (trees/ha), [M] and [T] are dummies for logging method and designated trail pattern respectively, and ai and bi are fitted coefficients. [M] has the value 1 for conventional and the value 0 for reduced impact logging. [T] has the value 1 if the trail alignment was adjusted to terrain circumstances and the value 0 for parallel trails. Different letters behind a coefficient refer to comparisons between the three trail patterns. Symbols denote the probability of coefficients being zero: ns p < 0.05,* p < 0.05, *** p < 0.001. The coefficients b2 and b3 were non-significant, which resulted in the following final model: Y = 11.41*** + 3.59*** ⋅ [M] + 4.58*** ⋅ [T] + 0.21** ⋅ X + ε (R2 = 0.80). Reduced impact logging with parallel trails had a significantly smaller area occupied by skid trails, than reduced impact logging with adjusted trails or conventional logging. The area occupied by trails did not differ between the latter two. Logging method Conventional Reduced impact, parallel trails Reduced impact, trails adjusted to topography a 14.92 *** 11.85 *** 14.43 * fitted coefficients b a 0.21 b 0.16 a 0.37 91 ns ns ns x x x n R2 6 7 5 0.41 0.39 0.40 Reduced impact logging in the tropical rain forest of Guyana Figure 4.13 The relation between logging intensity and the area affected by skid trails with three different skid trail patterns. The percentage of the total area affected by skid trails angularly transformed. The conventionally logged plots were located on flat to undulating terrain. Reduced impact logging led to a reduction of trail area by 40% on average in case the terrain allowed parallel trails. In case of adjusted trails, there was no significant difference with conventional logging. See Table 4.11 for regression equations. Traversing of felling gaps was restricted to gaps along skid trails in the reduced impact operation, implying that traces of skidding occupied only 5% to 8% of the total felling gap area at logging intensities of 8 to 16 trees per hectare, respectively. In the conventional operation this amounted to as much as 30% to 36% of the total felling gap area. Logging intensity did influence the total area occupied by trails (ANOVA on arcsin฀p transformed data, Type III sum of squares, F2,15 = 3.84, p = 0.045), but only the difference between 4 and 16 trees/ha was significant (Tukey’s HSD test). 4.6.2 Ground disturbance in felling gaps Ground disturbance in felling gaps occurred significantly more often in conventional logging than in reduced impact logging; an average 6.7% of the total area as opposed to an average 1.1%. It is evident that the occurrence of ground disturbance in felling gaps increased with increasing logging intensity, because total gap area and affected ground area increased (Fig. 4.14). In case of reduced impact logging the total area of ground disturbance in felling gaps was significantly correlated with the area affected by trails (Pearson r = 0.68, , n = 12, p < 0.05) and not with manoeuvring (Pearson r = 0.55, p = 0.06). In case of conventional logging ground disturbance in felling gaps was correlated with the total area affected by manoeuvring (Pearson r = 0.99, n = 6, p < 0.0003), and not with trails (Pearson r = 0.34, p = 0.52). This indicates that in reduced impact logging ground disturbance in felling gaps predominantly occurred where trees were felled close to the 92 Chapter 4: Logging damage trail network, while in case of conventional logging it was predominantly linked to the manoeuvring of logs into a position favourable for hooking. Figure 4.14 The relation between the area of ground disturbance in logging gaps and logging intensity. The following regression model was fitted: Y = 1.26ns + 6.57** ⋅ [M] + 0.46***⋅ X + 0.11ns ⋅ [M] ⋅ X + ε. Y = disturbed ground area in felling gaps as percentage of total area (arcsin฀p - transformed), X = logging intensity (trees/ha), and [M] is a dummy for logging method. It has the value 1 if a plot was logged conventionally and the value 0 if a plot was subjected to reduced impact logging. The intercept of the regression line for conventional logging; Y = 7.83 + 0.58 ⋅ X + ε, r2 = 0.77, is significantly higher (t14 = 3.11, p < 0.01) than the intercept of the regression line for reduced impact logging; Y = 1.26 + 0.46 ⋅ X + ε, r2 = 0.72. The slopes do not differ (t14 = 0.65, p = 0.52). Also the mean size of the exposed patches of ground disturbance differed significantly between the logging methods (ANOVA F1,16 = 145.65, p < 0.001) and amounted to 56.5 m2 (sd = 10.3 m2) in the conventional operation and 10.8 m2 (sd = 5.9 m2) in the experimental operation. 4.7 IMPACT ON THE RESIDUAL TREES Selective logging not only removes a proportion of the tree population, it also damages a proportion of the residual stand, because some of the trees that were not felled are killed or injured. Trees were classified in six categories: felled, killed, severely injured, with moderate injury, with minor injury or unaffected (for definitions, see Section 4.2.4). Severely injured trees survived the logging operation, but the probability that they will fully recover is small. If they will happen to recover, the probability that their trunk will be crooked is very large and they should thus be discounted as a potential source of timber in the future. The probability of full recovery for trees with a moderate or a minor injury is much larger. Moderate injury, however, is more likely to lead to decay than minor injury and probably also will affect a tree’s growth rate more strongly. 93 Reduced impact logging in the tropical rain forest of Guyana 4.7.1 Degree of damage and chances of survival Trees in the reduced impact experiment were monitored during two and a half years after logging. The odds on survival were computed for the three degrees of damage minor, moderate and severe injury - and compared with the survival rate of undamaged trees. Table 4.12 shows that the odds on survival for trees with minor damage do not differ from the ones for undamaged trees. The odds that a tree dies within two and a half years after logging are two times higher for moderately damaged trees than for undamaged trees and as much as eleven times higher for severely damaged trees. Not all severe damage turns out to be fatal in the end. The question is whether this holds for all types of severe damage. The ‘severe’ damage category included trees of which the crown had broken off entirely (topped), trees of which the trunk had split or had broken below the crown point (snapped), and trees that were partly uprooted or bent. The odds on survival for trees belonging to these four classes were compared separately with the survival rate of undamaged trees. It appeared that two and a half years after logging 10% of the trees of which the trunk had split, had died. The odds that a tree with a split trunk died, did not differ significantly from the odds for undamaged trees (χ2 = 0.716, df = 1,, p = 0.40). This is not surprising, because a split in the stem barely hinders a tree’s sapstream activity, while many species have effective mechanisms to block insect and fungal attacks (Shigo 1989). The mortality rates were significantly higher in the other cases: 17% of the partly uprooted trees died (odds ratio: 3.75, χ2 = 19.09, df = 1 , p < 0.01), 34% of the trees of which the crown had broken off (odds ratio: 9.37, χ2 = 123.8, df = 1, p < 0.01), and 65% of the trees of which the trunk had snapped (odds ratio: 33.06, χ2 = 281.4, df = 1, p < 0.01). Trees of which the trunk had snapped, will thus most likely die in the end. Table 4.12 Damage category Undamaged Minor Moderate Severe Odds on survival until two and a half years after logging for four damage categories in the reduced impact logging experiment. Damage was assessed in 1994 and mortality until early 1997. Significance testing and confidence limits were obtained by logistic regression. For logistic regression, the categories were recoded to three binary design variables. Each design variable assigns 1 to one of the three damaged categories and zero to all others. The undamaged category is considered the referent group with all design variables coding zero. Number of trees Dead 212 27 24 134 Alive 3788 438 203 216 Mortality 5% 6% 11% 38% Odds ratio 1.00 1.10 2.11 11.09 95% Confidence χ2 , df = 1 limits for odds ratio Lower Upper 0.73 1.35 8.58 1.66 3.30 14.32 0.21 10.84 338.94 p 0.646 0.001 0.000 A period of two and a half years is in most cases too short to determine the eventual fatality of a severe injury, the involved time-span most likely varies depending on the type of injury. Final collapse following crown breakage or snapping, may take between two to six years as suggested by Putz & Brokaw (1989), but may also take as much as twenty years as indicated after cyclone damage in Queensland (Oldeman & Van de 94 Chapter 4: Logging damage Meer 1988). The time-span is probably longer in case of partly uprooted trees and trees with split trunks, because their demise mainly depends on external factors - such as wind or fungi - , whereas in case of topped and snapped trees it depends mainly on whether these trees resprout or not and whether the sprouts themselves survive or not. Field inspections showed that sprouts of many trees that had re-sprouted just after logging, had died after two and half years. Field observations led to the suspicion that tree size plays a role in the odds on survival. It was also intuitively expected that smaller trees would resprout more vigorously than large trees. Therefore, multiple logistic regression was performed with both the damage categories and tree diameter as independent variables. Summarised data with the proportion of dead trees with varying levels of damage in three diameter classes are presented in Table 4.13. This table shows that the mortality rate increased with increasing diameter (effect of tree size controlled for variation between damage categories: χ2 = 33.02, df = 1, p < 0.01), but it appeared that this increase in the mortality rate did not differ between the different damage categories (multiple logistic regression, interaction: χ2 = 2.96, df = 3, p = 0.40). Table 4.13 Dbh class (cm) 10-20 20-40 ≥40 Periodic mortality rates within two and a half years after logging for four damage categories in a reduced impact logging experiment. Numbers of monitored trees (N) and mortality rates in three diameter classes (cm dbh). Undamaged N dead 637 2% 2588 5% 775 8% Minor injury N dead 55 7% 291 4% 119 10% Moderate injury N dead 29 0% 151 14% 47 13% Severe injury N dead 68 25% 253 38% 29 48% 4.7.2 Impact of logging method and intensity In contrast to the accumulated canopy loss with the two logging methods (Table 4.3, Fig. 4.2), the total number of affected stems per hectare (dbh ≥ 10 cm) was significantly smaller with reduced impact logging than with conventional logging (Tables 4.14 and 4.15). As expected, the number of affected trees increased when more trees had been felled. There was no interaction between logging method and logging intensity regarding the total number of affected trees. Although the number of affected trees per ha was augmented by increasing the logging intensity, this was not in proportion to the number of trees extracted. In the conventional operation, the number of trees affected fell from 16.4 to 10.2 per extracted tree when the intensity was increased from 8 to 16 trees/ha. In the reduced impact operation the number of affected trees fell from 16.4 per extracted tree at 4 trees/ha to 12.8 at 8 trees/ha and 8.9 at 16 trees/ha. Higher extraction levels therefore appear to cause less damage per extracted tree. This result was intuitively expected. When felling intensity is low, almost every felled tree makes its own gap, and a certain basic skidtrail network is needed to extract even a few logs. Increasing the exploitation level increases the probability that crowns of felled trees fall into existing gaps, 95 Reduced impact logging in the tropical rain forest of Guyana causing little further damage. Besides the basic skidtrail network, only additional branch trails have to be opened up, most of which are short, and thus increasing the total area under trails only slightly (see Section 4.6.1). Table 4.14 Impact of logging on residual stems with dbh ≥ 10 cm. Number of stems killed, severely damaged or less severely damaged with standard deviations (sd). The initial stem number in the experimentally logged plots is based on complete counts of stems with a dbh ≥ 20 and on a 12.5 % sample for the trees with a dbh from 10 to 20 cm. For the conventional logged plots, the real initial stem number is unknown, but based on three 1 ha sample plots Type of logging Conventional 8/ha 16/ha Reduced impact 4/ha 8/ha 16 /ha Table 4.15 Number of stems per ha Severe Moderate injury injury Before logging Killed 499 (5) 455 (35) 42 (2) 57 (3) 34 (7) 44 (3) 500 (53) 510 (91) 447 (18) 28 (10) 33 (6) 59 (9) 14 (3) 23 (3) 38 (5) Minor injury Total damaged 24 (1) 30 (5) 32 (4) 33 (4) 131 (3) 163 (13) 7 (0) 14 (4) 15 (3) 19 (4) 35 (10) 33 (5) 67 (10) 104 (15) 145 (6) Results from five ANOVAs testing the effects of logging method and logging intensity (n/ha) on the number of trees (n/ha) in each damage category. Sums of Squares are Type IV; degrees of freedom are 13, 1, 2 and 1 for Error, Method, Intensity and Interaction respectively. In case of conventional logging, significantly more trees are found in the moderately and severely damaged tree categories. Damage category Killed Severe injury Moderate injury Minor injury Total damaged Error Effect of Method Sum of Sum of F Squares Squares 531.0 38.0 0.93 ns 244.7 251.1 13.34 ** 134.2 528.9 51.23 *** 640.0 9.1 0.19 ns 1793.5 1762.6 12.78 ** Effect of Intensity Sum of F Squares 2083.7 25.51 *** 1030.9 27.39 *** 122.6 5.94 * 357.8 3.63 ns 10224.9 37.06 *** Interaction Sum of F Squares 91.7 2.25 21.1 1.12 12.9 1.25 7.5 0.15 61.6 0.45 ns ns ns ns ns Tests were also carried out concerning the number of trees in the different damage categories separately. In the conventional operation, more trees were found with severe and moderate injuries than in the experimental operation, whereas the number of killed trees and trees with minor damage were similar in both operations (Tables 4.14 and 4.15). Increasing the logging intensity, increased the number of killed, severely and moderately injured trees, but surprisingly did not affect the number of trees with minor injury. The results above only give the number of affected trees but not their size. Basal area is correlated with tree size and is therefore a better indicator of how much timber volume or biomass was lost. Basal areas (m2/ha) were calculated by damage category and treatment (Table 4.16). In Section 4.5.1, it was shown that an effect of the logging method on the loss of canopy cover was only revealed, if logging intensity was expressed as the basal area felled. Therefore, stand damage was also described as a function of the basal area felled - thus 96 Chapter 4: Logging damage compensating for the size of the felled trees. Linear regression analysis with logging method as a dummy variable showed, that the total affected basal area was strongly associated with logging intensity, but was independent of the logging method (Table 4.17). Table 4.16 Impact of logging on residual stems with dbh ≥ 10 cm. Basal area per ha killed, severely damaged or less severely damaged. The initial basal area in the experimentally logged plots is based on complete counts of stems with a dbh ≥ 20 and on a 12.5 % sample for the trees with a dbh from 10 to 20 cm. For the conventionally logged plots, the initial basal area is unknown, but based on three 1-ha sample plots. Type of logging Before logging Conventional 8/ha 16/ha Reduced impact 4/ha 8/ha 16/ha Table 4.17 Felled Basal area (m2/ha) Killed Severe Moderate injury injury Total damaged 29.0 28.4 2.0 4.1 1.0 1.2 1.3 1.6 1.8 2.4 2.1 3.2 6.3 8.3 26.5 26.5 25.0 1.2 2.0 3.7 0.7 0.9 1.5 0.6 1.0 1.7 0.4 0.9 1.2 1.9 2.7 2.6 3.6 5.5 7.0 Impact of logging method and intensity distributed over four damage categories. Fitted coefficients for the regression models Yi = ai + bi ⋅ X + ε, where ai = ai1 + ai2 ⋅ [M] and bi = bi1 + bi2 ⋅ [M]. Yi = basal area (m2/ha) in damage category i , X = basal area felled (m2/ha), [M] is a dummy for logging method, and aij and bij are fitted coefficients. [M] has the value of 1 if a plot was logged conventionally and the value of 0 if a plot was subjected to reduced impact logging. Symbols denote the probability of coefficients being zero: ns p < 0.05,* p < 0.05, ** p < 0.01, *** p < 0.001. The letter d behind a coefficient denotes a significant difference between logging methods for that coefficient, while the letter x indicates non-significance. Intercepts for the moderately and severely injured tree categories are significantly higher in the conventional operation, while slopes are significantly steeper for the killed and severely injured tree categories in the reduced impact operation. Damage category Killed Severe injury Moderate injury Minor injury Total incidental losses and injuries Unaffected Minor injury Logging Method a 0.97 0.99 1.33 1.33 4.44 Conventional b * 0.10 ns * 0.15 ns * 0.25 ns * 0.44 * *** 0.94 ** 25.14 *** –2.23 ** R2 0.35 0.47 0.47 0.80 0.91 0.92 a 0.22 ns 0.08 ns 0.32 ns 1.75 ns 2.37 * Reduced impact b x 0.34 *** d 0.44 *** d 0.26 * x 0.31 ns x 1.35 ** 25.29 *** x d d x x x R2 0.77 0.89 0.38 0.08 0.51 –3.04 *** x 0.85 The total basal area of injured and incidentally lost trees thus does not appear to differ between the two methods, whereas the total number of damaged stems did. This suggests that the average size of the damaged trees was not entirely the same with both methods (see Section 4.7.4). Although the difference between the two methods is not significant, the amount of damage caused by reduced impact logging may seem to be somewhat lower than the amount of damage caused by conventional logging. According 97 Reduced impact logging in the tropical rain forest of Guyana to the linear regression model (Table 4.17), the basal area damaged per m2 extracted fell from 3.2 at a basal area felled of 2 m2/ha to 2.1 m2 at 4 m2/ha in the conventional operation. In the experimental operation, the affected basal area fell from 3.7 per extracted m2 at 1 m2/ha to 2.5 at 2 m2/ha and 1.9 at 4 m2/ha. Higher yields thus also appear to cause less damage per m2 felled. The relative reduction in the damaged basal area coincides with the relative reduction in the damaged number of trees. This suggests that the size of the damaged trees stayed more or less the same with an increase of the extraction level (see Section 4.7.4). Separate tests of the basal areas in each damage category showed, however, that conventional logging resulted in larger amounts severely and moderately damaged basal area. Furthermore, a significant correlation was found between extraction level and incidence for all damage categories, except minor damage. Finally, an interaction between logging method and extraction level was revealed concerning the amount of basal area killed and the amount of severely injured basal area (Table 4.17). According to the regression model, only a small amount of basal area was incidentally lost due to selective logging (Table 4.17): i.e. 0.7 m2/ha at a logging intensity of 1 m2/ha rising to 1.3 m2/ha at a logging intensity of 4 m2/ha (estimated to be 2.5 to 5.1% of the original stand). The difference between logging methods in respect to incidental losses fell just short of being significant (p = 0.051), but interaction between logging method and intensity was significant (p < 0.05). A significant difference between logging methods did occur with respect to the severely damaged tree category. Conventional logging led to a seriously damaged basal area of 1.5 m2/ha (an estimated 5.1% of the original stand), with no significant relation to logging intensity. Reduced impact logging seriously damaged a basal area of 0.5 m2/ha (2.1% of the original stand) after a light harvest, which damage increased significantly after increasing the intensity fourfold: 1.8 m2/ha (7.6% of the original stand). The incidence of moderate damage was also significantly higher with conventional logging than with reduced impact logging. In case of conventional logging the incidence of moderate damage amounted to 2.1 m2/ha (an estimated 7.3% of the original stand) independent of logging intensity. In case of reduced impact logging, on the other hand, it rose from 0.6 m2/ha (2.0% of the original stand) to 1.4 m2/ha (5.5% of original basal area) with an increase in logging intensity from 1 to 4 m2/ha. Minor damage occurred more often than any of the other damage categories. In case of conventional logging the incidence of minor damage rose from 1.8 to 3.1 m2/ha (an estimated 6.1% to 10.8% of the original stand) with a fourfold increase in logging intensity. In case of reduced impact logging, it amounted to 2.4 m2/ha (9.4% of the original stand), independent of logging intensity. It is also important to estimate how much of the original stand survived logging without damage. In the reduced impact operation the numbers and basal areas that survived without damage can be accurately estimated, because all trees ≥ 20 cm dbh were enumerated and trees with a dbh between 10 and 20 cm were measured in 25 systematically arranged sample plots of 100 m2 prior to logging. In the conventional operation, as mentioned before, it was not possible to determine the exploitation levels in the allocated tracts of forest beforehand, and, therefore, no pre-harvest enumeration 98 Chapter 4: Logging damage could be carried out. Three 1-ha plots were established prior to logging, one in each of the three tracts to be conventionally logged. The pre-harvest number of stems and basal area of a 1-ha plot in a tract were extrapolated to assess the original stands in the 2-ha plots in that particular tract. This way it was estimated that the pre-harvest basal areas were higher in the conventionally logged area than in the experimentally logged area (Table 4.16). The unaffected basal area in each plot was estimated by deducting all felled, incidentally lost and injured basal area from these estimated pre-harvest basal area in that particular plot. Linear regression analysis performed on the estimated unaffected basal area as a function of the basal area felled (Table 4.17) failed to show any difference between the two methods. The original basal area in the conventionally logged plots was estimated to be 9% higher on average. This may explain the outcome, even if the proportion of undamaged basal area were higher with reduced impact logging. The main difference between conventional logging and reduced impact logging, in the case studied, is that reduced impact logging caused less moderate damage in terms of number of stems and basal area. It also resulted in less severely damaged trees and basal area, the latter, however, at low and modest logging intensities only. The number of killed trees did not differ significantly between the two methods, but there are indications that reduced impact logging killed less basal area than conventional logging at a low extraction level and more basal area at a high extraction level. Another important point is the absence of a relation with logging intensity in certain categories. With neither method, the incidence of minor injuries appeared to have a significant relation with logging intensity in terms of number of stems. This suggests that in the reduced impact operation minor damages are related to skidding, because the area affected by skidding was only weakly associated with logging intensity (Table 4.11). In the conventional operation, neither the amount of killed, nor the amounts of severely or moderately damaged basal area were significantly augmented by felling more trees (Table 4.17). It was mentioned above that an increase of the extraction level causes less damage per m2 felled in the case studied here. This applied to the experimental logging operation, but it especially manifested itself in the conventional operation that was studied. This result is most certainly associated with the felling of clusters of gregarious Chlorocardium rodiei. In felling clusters of similar sized Chlorocardium rodiei, it is not unlikely that some of the trees neighbouring a tree earmarked for felling, are going to be or have already been felled. Besides that, a tree can only be killed once. These two aspects reduced the amount of neighbouring trees that potentially could be killed or damaged. Almost 40% of the trees with severe injury had died during the first three years after logging (see Section 4.7.1). The surviving ones will presumably either die or develop malformations. Most trees with less severe injury should not be discounted as a future source of timber. However, trees with moderate - especially stem - injury are more susceptible to decay and thus are less likely to recover fully than trees with minor injury. Thus, instead of comparing the damage classes separately, it is more informative 99 Reduced impact logging in the tropical rain forest of Guyana to compare the amount of basal area that can be surely (felled or at least severely damaged trees), or probably (felled or at least moderate injury) discarded as future timber suppliers. Fig. 4.15 is a graphical representation of cumulative damage categories. According to the regression model (Table 4.18), the removed or at least severely damaged basal area rose from 3.0 to 6.8 m2/ha (or from an estimated 10.3% to 23.1% of the original stand) in case of the conventional operation studied here when the logging intensity was increased from 1 to 4 m2/ha, while it rose from 2.1 to 7.4 m2/ha (or from 7.4% to 29.8% of the original stand) in case of the reduced impact operation. Table 4.18 Impact of logging method and intensity, using the basal area (m2/ha) felled or surviving logging with specified minima of injury as a function of basal area felled (m2/ha). Yi = ai + bi ⋅ X + ε, where ai = ai1 + ai2 ⋅ [M] and bi = bi1 + bi2 ⋅ [M]. Yi = basal area (m2/ha) in damage category i , X = basal area felled (m2/ha), [M] is a dummy for logging method, and aij and bij are fitted coefficients. [M] has the value of 1 for conventional and the value of 0 for reduced impact logging. Symbols denote the probability of coefficients being zero: ns p < 0.05,* p < 0.05, ** p < 0.01, *** p < 0.001. The letter d behind a coefficient denotes a significant difference between logging methods for that coefficient, while the letter x indicates non-significance. Damage category (cumulative) a 0.78 * 1.78 ** Conventional b 1.10 *** 1.25 *** R2 0.99 0.98 a 0.22 0.30 ns ns Reduced impact b x 1.34 *** d d 1.78 *** d R2 0.98 0.98 3.11 ** 1.50 ** 0.95 0.62 * d 2.04 *** d 0.97 4.44 *** 1.94 *** 0.98 2.37 * x 2.35 *** x 0.76 Basal area (m2/ha) Felled or killed Felled or at least severe injury Felled or at least moderate injury Felled or at least minor injury Logging method Basal area felled (m2/ha) Basal area felled (m2/ha) Figure 4.15 Impact of logging method and intensity. The basal area (m /ha; dbh ≥ 10 cm) felled or surviving logging with specified minima of injury in relation to the felled basal area (m2/ha). Coefficients for regression lines are given in Table 4.18. See text for further explanation 2 100 Chapter 4: Logging damage Including the moderate tree damage category resulted in a rise from 4.6 to 9.1 m2/ha (or from an estimated 15.7% to 32.2% of the original stand) in case of the conventional operation studied here with an increase in the logging intensity from 1 to 4 m2/ha. In case of the reduced impact operation, it rose from 2.7 to 8.8 m2/ha (or from 9.4% to 35.2% of the original stand). Assuming that it is realistic that only the most vigorous of these trees will fully recover, it can be concluded that reduced impact logging ‘saves’ 40% of the basal area, which was lost otherwise, at a light harvest. However, this gain reduces rapidly with increasing logging intensity and approaches zero at a high logging intensity. 4.7.3 Impact of felling and skidding Felling inflicted damage to more trees than skidding. All plots pooled, 15% of the initial tree population was damaged by felling, 9% by skidding and only 0.5% first by felling and subsequently by skidding. There were distinct differences between logging methods and intensities in the proportions affected by felling alone, by skidding alone, or damaged by both felling and skidding (Fig. 4.16). Figure 4.16 Mean numbers of trees per hectare (and standard deviations) affected by felling (hatched bars), skidding (open bars) or both (black bars), for reduced impact (RIL) and conventional logging (CNV) at two logging intensities in Pibiri, Central Guyana. More trees are damaged by felling than by skidding and only few trees suffered injury twice. In reduced impact logging, felling damage rose more steeply with an increase of logging intensity, whereas skidding damage remained the same. In the reduced impact operation a higher proportion of tree damage was due to felling (as opposed to skidding) than in the conventional operation. An increase in logging intensity raised the proportion of damage due to felling as opposed to skidding. However, the relative increase in felling damage (as opposed to skidding damage) with increasing logging intensity was stronger in case of reduced impact logging than in case of conventional logging. 101 Reduced impact logging in the tropical rain forest of Guyana Log-linear analysis proved that the logging method, logging intensity and logging phase; i.e. felling or skidding, were indeed not mutually independent (G = 10.10, df = 2, p < 0.01). In the previous section, it was shown that the number of casualties per ha was higher for conventional logging than for reduced impact logging, without distinguishing between damage due to felling or skidding. Analysis of variance (Type IV sum of squares) was carried out separately for trees damaged by felling and trees damaged by skidding (counting multiple injury twice). This showed that the difference in the numbers of trees damaged by felling (n/ha) between logging methods was not significant (F1,13 = 0.24, p = 0.64), whereas, not surprisingly, the effect of logging intensity was highly significant (F2,13 = 28.79, p < 0.001). With respect to trees damaged by skidding, however, the difference between the logging methods was clearly significant (F1,13 = 15.67, p < 0.01), whereas the effect of logging intensity was not significant (F2,13 = 1.88, p = 0.19). The latter result is opposed to the effect of logging intensity on the area affected by skidding (section 4.6.1), which was significantly related to the logging intensity. This can be partly explained by the fact that a skidder operator in general avoids large trees, as running into these would be time consuming. Another explanation could be that small trees were often dragged along, partially buried or even completely crushed by the machine. Therefore, it is not unlikely that some of them - small trees were only marked in a 12.5% sample prior to logging and only in the experimental area - may have been overlooked. If some of these crushed and shoved trees were indeed overlooked, chances are that this occurred especially in the conventionally logged area. There, the machine used its blade in the stump area, shoving around more trees at the higher than the lower logging intensity (more felled trees included in one chablis). If the latter problem indeed affected our tree damage estimates, this would suggest that skidding damage to the residual stand in the conventional operation was even heavier than reported above. The number of stems damaged by skidding appeared not to be influenced significantly by the logging intensity. This implies that the number of trees damaged per tree extracted was reduced strongly by increasing the logging intensity: i.e. an increase from 8 to 16 trees felled per ha led to a reduction from 5.3 to 2.6 trees damaged per tree extracted in case of reduced impact and from 7.5 to 4.4 in case of conventional logging. The number of affected trees per ha by felling was augmented by increasing the logging intensity, but, as before, not in proportion to the number of trees extracted. The number of trees affected by felling dropped from 8.4 to 6.4 per extracted tree when the intensity was increased from 8 to 16 trees/ha, independent of the logging method. Linear regression analysis was used to describe tree damage (m2/ha) as a function of basal area felled (m2/ha) with logging method and logging phase (felling versus skidding) as dummy variables (Table 4.19). If the parameter of the exploitation level is basal area felled - thus compensating for the size of the felled trees - , it is again found that felling causes significantly more damage than skidding (t29 = 5.33, p < 0.001), and that conventional logging affects more basal area than reduced impact logging 102 Chapter 4: Logging damage (t29 = 2.08, p < 0.05). The relation between damage due to felling and the basal area felled is highly significant (t15 = 6.98, p < 0.001), whereas the relation between skidding damage and the basal area felled is non-significant (t15 = 1.24, p = 0.23). Felling in the conventional operation caused significantly more damage than reduced impact felling (t15 = 2.16, p < 0.05), but the difference between logging methods in the damage due to skidding appears to be non-significant (t15 = 1.39, p = 0.18). Table 4.19 Impact of felling and skidding in terms of damaged basal area (m2/ha) by logging method and intensity. Fitted coefficients for the regression models Y = a + b ⋅ X + ε, where a = a1 + a2 ⋅ [M] + a3 ⋅ [P] + a4 ⋅ [M] ⋅ [P] and b = b1 + b2 ⋅ [M] + b3 ⋅ [P] + b4 ⋅ [M] ⋅ [P]; Y = basal area damaged (m2/ha); X = basal area felled (m2/ha); [M] is a dummy for logging method, [P] is a dummy for logging phase, and ai and bi are fitted coefficients. [M] has the value of 1 for conventional and the value of 0 for reduced impact logging. [P] has the value of 1 if damage was incurred during felling and the value of 0 if incurred during skidding. Trees that were damaged by both felling and skidding, were counted twice; once as a tree damaged by felling and once as a tree damaged by skidding. Symbols denote the probability of coefficients being zero: ns p < 0.05,* p < 0.05, ** p < 0.01, *** p < 0.001. See text for further explanation. Logging phase Felling Skidding Logging method Conventional a b 2.16 ** 0.95 *** 2.47 ** 0.07 ns R2 0.95 0.06 a 1.29 * 1.08 ns Reduced impact b 1.02 *** 0.36 ns R2 0.68 0.13 In terms of the damaged basal area, felling damage differed between methods, whereas skidding damage did not. According to the regression model (Table 4.19, ig. 4.17), an increase in the extraction level from 2 to 4 m2/ha felled per ha led to a reduction of skidding damage from 1.0 to 0.7 m2 per m2 extracted, independent of the logging method. The basal area affected by felling fell from 2.0 to 1.5 m.2 per m2 extracted in case of conventional logging and from 1.7 to 1.3 m2 per m2 extracted in case of reduced impact logging when the extraction level is increased from 2 to 4 m2/ha. Comparing the results of figures 4.16 and 4.17, it would appear that the outcome of the test concerning the affected basal area conflicts with the one concerning the number of affected trees, but this need not be so. It rather implies, that indeed more trees were damaged during skidding in the conventional operation, but that these trees were smaller on average than in the reduced impact operation. The outcome concerning felling is diametrically opposed to this; similar amounts of trees were damaged by felling in both operations, but these trees were smaller on average in the reduced impact operation. In the next section, the difference in size of damaged trees between conventional and reduced impact logging and the relation with either felling or skidding will be dealt with. In the previous section, it was shown that conventional logging damaged more trees than reduced impact logging. This difference was mainly caused by a difference in the number of severely and moderately damaged trees. Furthermore, it was shown that an increase of the exploitation level led to increased numbers of killed, severely and moderately damaged trees. Above it was shown that the number of casualties due to 103 Reduced impact logging in the tropical rain forest of Guyana felling did not differ between logging methods, whereas the number of trees damaged by skidding was higher in case of conventional logging. The number of trees that had severe, moderate or light felling damage, all seemed to be smaller in the reduced impact operation, but only the differences in the number of trees with moderate and minor injury were significant (F1,13 = 13.07, p = 0.003 and F1,13 = 8.23, p = 0.013 respectively). The extraction level appeared to influence only the number of trees that were killed or severely damaged by felling (F1,13 = 17.44, p < 0.001 and F1,13 = 20.00, p < 0.001 respectively). Figure 4.17 Mean basal area (m2/ha - and standard deviations) affected by felling (hatched bars), skidding (open bars) or both (black bars), for reduced impact (RIL) and conventional logging (CNV) at two logging intensities in Pibiri, Central Guyana. Felling in the conventional operation damaged significantly more basal area than reduced impact felling, but there is no difference in basal area damaged by skidding. Is there any difference in severity between damage due to felling and damage due to skidding, and, if so, is this independent of the logging method and logging intensity? Log-linear analysis applied to the proportion of trees in each damage category showed that the distribution among the different damage categories was independent of logging intensity (G = 5.09, df = 3, p = 0.17). The severity of damage was, however, associated simultaneously with both the logging method and the logging phase (felling or skidding) during which the damage was inflicted (G = 171.58, df = 3, p < 0.001, Fig. 4.18). Indeed, the severity of the damages differed between damage due to felling and damage due to skidding, but these differences were not the same with both logging methods. This result indicates the need for a more detailed analysis. Data presented in Table 4.20 indicate that reduced impact logging reduced the number of trees that were destroyed due to skidding significantly (F1,13 = 128.03, p < 0.001). The numbers of trees with severe and moderate damage decreased as well, whereas the 104 Chapter 4: Logging damage % of damaged basal area per phase % of damaged trees per phase number of trees with minor damage increased. However, these differences were not significant. In case of damages due to felling, on the other hand, reduced impact logging destroyed significantly more trees (F1,13 = 22.84, p < 0.001), with the difference between the logging methods becoming more pronounced with increasing logging intensity (F1,13 = 5.52, p < 0.05). Figure 4.18 Tree damage in Conventional (CNV) and Reduced impact logging (RIL). Trees per damage category as percentage of all damaged trees per logging phase (top) and basal area per damage category as percentage of totally damaged basal area per logging phase (bottom) in Pibiri, Central Guyana. Mean percentages for felling (hatched bars), skidding (black bars) and both phases combined (open bars) and standard deviations (whiskers). Observations were pooled among logging methods and intensities; 2474 damaged trees in twelve 1.96-ha plots subjected to reduced impact logging and 1768 damaged trees in six 2-ha conventionally logged plots. Skidding in the reduced impact operation thus appeared to have a lower impact than in the conventional operation with respect to the number of casualties (24 trees/ha less on the average). This difference lies mainly in a reduction in the number of destroyed trees (20 trees/ha less on the average). Apparently, this reduction in the number of trees destroyed by skidding was nullified by an increased number of trees destroyed by felling, with the understanding that this effect increased more or less exponentially at a high logging intensity (8 trees/ha more at 8 trees felled per ha and 25 trees/ha more at 16 trees felled per ha). The number of trees with less serious injury were smaller on average in the reduced impact operation. Consequently, the total number of casualties due to felling did not differ between the two methods. The smaller numbers in the reduced impact operation of trees with severe and moderate injuries that were observed without subdivision (Table 4.14), can thus be attributed to reductions achieved during 105 Reduced impact logging in the tropical rain forest of Guyana both felling and skidding. The distribution among damage categories in terms of basal area for the different logging methods and logging phases indicates that trees that suffered from serious and fatal injuries were smaller on average than the ones that suffered from less serious injuries (Fig. 4.18). Table 4.20 Impact of logging on residual stems with dbh ≥ 10 cm. Number of stems killed, severely, moderately and little damaged due to felling and skidding in a conventional (CNV) and a reduced impact operation (RIL) at different intensities. Type of logging Number of stems in each damage category (n/ha) Felling killedabc Skidding Multiple Total injuries severe moderate minor killeda severe moderate minor injuryb injurya injurya injury injury injury CNV 8/ha 16/ha 14.0 22.3 27.0 33.8 12.8 17.8 17.2 18.7 27.8 34.7 6.0 8.0 8.0 7.5 14.3 14.0 4.2 6.5 131.3 163.3 4/ha 8/ha 16/ha 20.6 22.4 47.1 10.5 19.0 33.0 4.3 8.7 9.2 8.5 12.0 14.8 7.1 10.8 11.6 3.4 3.7 4.4 2.4 4.3 5.3 10.2 22.6 18.0 0.0 0.9 1.5 67.0 104.4 144.9 RIL a difference between methods significant difference between intensities significant c interaction significant b 4.7.4 Impact and size of damaged trees In the previous section, seemingly conflicting results were obtained depending on whether damage was expressed in basal area or in number of stems. This suggested that trees damaged by felling were, on average, smaller in the reduced impact operation, whereas trees damaged by skidding were, on average, smaller in the conventional operation. The mean diameter of the damaged trees could be examined for each logging method and logging phase (felling or skidding). However, the conventionally logged stand featured a larger mean diameter prior to logging than most experimentally logged stands (Section 4.3). Interpretation of any difference in mean diameter of damaged trees in a particular category will hence be difficult. It may be just as well be ascribed to the logging method as to differences in initial stand structure. Therefore, it was examined whether or not damages in each diameter class were proportionate to their relative frequency in the tree populations prior to logging, and whether or not these relative frequencies differed for logging method, intensity and logging phase. A proportion of the trees ≥40 cm dbh had been felled. The data set for these was separated from the original population when making this comparison, because these were no longer potential casualties. Data presented in Table 4.21 indicate that, overall, trees between 20 and 40 cm dbh were proportionately more frequently damaged than the other size classes (partial association: G = 60.11, df = 2, p < 0.001), but there was significant variation with respect to logging phase, method or intensity. Felling damaged medium sized trees (20 to 40 cm dbh) more frequently than 106 Chapter 4: Logging damage other size classes, whereas skidding damaged small trees (10 to 20 cm dbh) more frequently (partial association: G = 130.79, df = 2, p < 0.001). Compared to conventional logging, reduced impact logging damaged a smaller proportion of the trees between 20 to 40 cm dbh and a larger proportion of the trees larger than 40 cm dbh (partial association: G = 16.22, df = 2, p < 0.001). An increase in logging intensity from 8 to 16 trees/ha resulted in an increase in the proportion that was damaged in the 10 to 20 cm class and a decrease in - mainly - the ‘40 cm dbh and larger’ class (partial association G = 9.60, df = 2, p < 0.01). Table 4.21 Logging damage to residual stand; proportions of residual number of stems (n/ha) by diameter class (cm) with different logging methods and intensities. The residual stand is defined as the original stand minus the felled trees. The frequencies of damaged stems after felling and after skidding are based on a 100% inventory. The number of stems prior to logging in the reduced impact plots is based on complete counts of stems with a dbh ≥ 20 cm and on a 12.5% sample for the trees with a dbh from 10 to 20 cm. Number of stems prior to logging in case of conventional logging is based on three 1 ha sample plots for trees ≥ 20 cm dbh and 25% sub-sample for trees 10-20 cm dbh. The total proportion of damaged trees can be lower than the summation of the proportions damaged by skidding and felling, because trees damaged fist by felling and subsequently by skidding were counted two times: as trees damaged by felling and as trees damaged by skidding. Type of logging Conventional 8/ha Dbh class (cm) n/ha 10-20 20-40 ≥40 288 146 57 491 12.2 22.5 12.6 15.3 13.0 12.7 7.6 12.3 25.1 33.0 19.0 26.7 10-20 20-40 ≥40 256 132 51 439 18.9 31.2 18.6 22.6 18.8 13.7 8.5 16.1 36.8 42.4 25.2 37.2 10-20 20-40 ≥40 300 151 36 500 8.2 10.0 9.1 8.8 5.2 3.9 3.8 4.7 13.4 13.9 12.9 13.5 10-20 20-40 ≥40 307 151 43 502 10.0 17.7 12.9 12.5 9.0 7.1 9.2 8.4 18.9 24.5 21.6 20.8 10-20 20-40 ≥40 264 130 37 431 22.5 29.5 21.8 24.5 10.6 6.9 10.1 9.5 32.9 35.7 31.9 33.6 Total 16/ha Total Reduced impact 4/ha Total 8/ha Total 16/ha Total Percentage damaged in each size class by: felling skidding total The observation that skidding generally damages a larger proportion of the small trees than felling is no surprise, because, as mentioned earlier, a skidder operator in general avoids large trees, while the trees that are in its way can not escape from - usually fatal 107 Reduced impact logging in the tropical rain forest of Guyana - injury. The fact that during felling, and, actually, during the entire operation, trees between 20 and 40 cm dbh were relatively more frequently damaged, is not so obvious. A plausible explanation may be that a small tree stands a bigger chance of escaping from being hit by a felled tree that comes crashing down. Not only is a small tree shorter, also its crown is much smaller than the one of a tree between 20 and 40 cm dbh (the size of the crown of a tree having a curvilinear relationship with its stem diameter - cf. Section 6.3.5). The former quality may allow it to escape the felled tree’s sweeping branches when it is positioned close to the stump. The latter quality allows it to escape the felled tree’s branches and trunk when it is positioned near the locality where the trunk will land (bole zone). Trees between 20 and 40 cm dbh are usually starting to reach the canopy and have started to expand their crowns (see also Section 6.3.5), which makes them an easier target. They are also more likely to be linked to the felled tree by lianas. There are two plausible reasons why larger trees (dbh ≥ 40 cm) got off lightly are. First, a feller would always try to avoid felling the tree in the direction of a large tree, because of the risk of a hang-up. Secondly, the falling tree would probably just graze the crown of a large tree underneath as it passes, because large trees are relatively more dispersed and usually of similar height as the felled tree. The outcome that, compared to conventional logging, reduced impact logging damaged trees between 20 and 40 cm dbh less frequently, large trees more frequently, and a similar proportion of the trees between 10 and 20 cm, could be partly explained by differences in the characteristics of the felling methods. Large trees are less frequently damaged by felling in the conventional operation, because trees are felled in groups in this operation. When the trees to be felled are dispersed, as in the reduced impact operation, neighbouring large trees are ready victims. By felling in groups, however, the number of neighbouring large trees is reduced and, thereby, the number of potential victims. Knowing, however, that more trees were damaged by skidding in the conventional operation and knowing now that skidding affects trees with 10 cm ≤ dbh < 20 cm proportionately most frequently, would not suggest that conventional and reduced impact logging damage equal proportions of this size class. The latter can be explained as follows. Small trees were probably damaged relatively more often during felling in the reduced impact operation, because of directional felling; i.e. in order to avoid damage to medium sized and large trees, trees were often directed at a patch of small trees. Conversely, during skidding small trees were more often damaged in the conventional operation, because of the movements in the bole zone and around the stump of the felled tree in this operation. During these manoeuvres small and medium sized trees that survived felling were knocked down or grazed. In the end, the higher proportion of small trees damaged during skidding then cancels out against the smaller proportion damaged during felling. The data in Table 4.21 and field observations confirm this explanation, although it is neither substantiated nor contradicted by finding significant interactions in the log-linear analysis. The observation that, by increasing the logging intensity, relatively more trees smaller than 20 cm dbh were damaged in comparison to the larger trees, is partly related to an increase in overlap of felling gaps. It is related, for another part, to the fact that an 108 Chapter 4: Logging damage increase in logging intensity rather led to an increase of the ground area affected by manoeuvring, instead of an increase in the total length of the skid trails. As mentioned above, these movements affect small trees especially. The degree of injury depended on the size of the tree. Data presented in Table 4.22 indicate that small trees are more vulnerable to destruction than larger ones, while less serious injury - especially minor injury - occurred more often among larger trees. For convenience of comparison, the four damage categories were reduced to two categories; ‘irreversible’ and ‘reversible’ damage. As mentioned in Section 4.7.2, severe damage can be regarded as delayed fatal damage, whereas trees with minor and moderate damage will most likely recover. The terminology is actually not appropriate because the final recovery cannot be established until several years from now (Section 4.7.1). Seventy-five percent of all damaged small trees were judged to be damaged irreversibly, but there was distinct variation depending simultaneously on whether the damage was incurred during skidding or felling, and on the logging method. It depended on the logging method whether felling or skidding was the most cause of damage (G = 115.29, df = 1, p < 0.001). Small trees damaged by felling were more frequently irreversibly damaged in the reduced impact operation than in the conventional operation (G = 31.81, df = 1, p < 0.001), whereas the reverse was found for trees damaged by skidding (G = 95.03, df = 1, p < 0.001). In sum, a larger proportion of the trees in the size class of 10 to 20 cm that were damaged - by felling, by skidding or with multiple injury - will most likely recover with reduced impact logging than with conventional logging (27% against 23%; G = 5.40, df = 1, p < 0.020). Of all damaged medium sized trees, 54% will probably recover, but also in this class there was distinct variation. Here, skidding consistently led to lighter injuries than felling, but otherwise the same pattern was seen as with the small trees (G = 30.40, df = 1, p < 0.001): i.e. trees damaged by felling were more often found to be irreversibly damaged in the reduced impact operation than in the conventional operation (G = 8.25, df = 1, p < 0.01), whereas the reverse was found for trees damaged by skidding (G = 22.36, df = 1, p < 0.001). Ultimately, equal proportions of damaged medium sized trees - by felling, by skidding or with multiple injury - will most likely recover with both logging methods (G = 2.80, df = 1, p = 0.09). Only 21% of all damaged large trees were irreversibly damaged. There were no differences between the logging methods in this class (G = 1.13, df = 1, p = 0.29), but skidding had a lighter impact than felling (G = 56.59, df = 1, p < 0.001). Including the trees that were damaged first by felling and subsequently by skidding had no influence on the difference between the logging methods in the balance between irreversibly and reversibly damaged trees. It is not surprising that small trees stand a much bigger chance on a fatal encounter than larger trees, for a falling tree gains momentum on its course through the forest which reaches highest value as it slams into the forest floor. 109 Type of logging Severity of logging damage to residual stand per diameter class by logging method and logging phase. Encountered number of stems damaged by felling, by skidding and damaged twice, and respective proportions in each damage category, in twelve 1.96 ha plots subjected to reduced impact logging (RIL), and six 2 ha plots subjected to conventional logging (CNV. Observations are pooled among logging intensities. Dbh class (cm) Felling Skidding Multiple injuries Total Damaged Percentage in each damage category Damaged Percentage in each damage category number of number of trees trees killed severe moderate minor killed severe moderate minor injury injury injury injury injury injury CNV 110 10-20 20-40 ≥40 Total 486 406 90 982 32 13 10 22 41 37 17 37 14 23 28 19 13 27 46 22 499 181 42 722 69 17 0 52 12 13 2 12 7 26 31 13 12 44 67 24 15 39 10 64 1000 626 142 1768 10-20 20-40 ≥40 Total 847 619 134 1600 60 22 13 41 26 37 22 30 6 17 19 11 9 23 46 17 574 205 76 855 40 2 0 28 13 8 0 11 7 18 24 11 40 71 76 51 7 9 3 19 1428 833 213 2474 RIL Reduced impact logging in the tropical rain forest of Guyana Table 4.22 Chapter 4: Logging damage The tree's impulse thus is small when its crown meets with large trees and becomes increasingly stronger while it hits decreasingly smaller trees. Occasionally a large tree may be uprooted but damage is most often restricted to excision of larger or smaller parts of the crown. Conversely, small trees are subdued to the maximum force and suffering at least serious injury. It must be noted, however, that the subjective division in reversible and irreversible damage ignores that small trees have a greater potential to recover from heavy impact than large trees. Often the young flexibility of a small tree’s stem allows it to redress itself and fully recover, or, if it is not so, by way of reiteration. Skidding typically avoids large trees and injuries to them are usually restricted to scraped-off pieces of bark. The small trees on the skidder's way, however, are nearly all squashed and few survive. The breaking down of the damages by size class and damaging agent showed, that pooling of these classes obscured important differences between the logging methods. In fact, the pooled data were not homogeneous. This often led to conflicting results instead of providing a more powerful analysis, particularly if it was - wrongfully assumed that a larger number of observations make analysis stronger. The differences between the two logging methods in absolute measures are summarised in Table 4.23. It is shown, that reduced impact felling did kill larger numbers of small trees than conventional logging, but the numbers of small trees with severe or moderate injury were smaller, resulting in equal numbers of damaged small trees. Trees larger than 20 cm dbh were affected by felling to a similar degree with both methods, except that more medium sized trees were found with minor damage in the conventional operation. The latter could not be retrieved from the total number of damages in this size class. In summary, reduced impact skidding destroyed fewer small and medium sized trees. In case of the small trees, the difference in the total number of damaged trees was slightly toned down by an increase in the number of trees with minor injury. In case of the medium sized trees, the difference was amplified by a drop in the number of trees with moderate injury. Multiple severe injury was far less common with reduced impact logging among trees below 40 cm dbh, whereas multiple moderate injury was less common among trees above 20 cm dbh. The difference in absolute numbers of destroyed and injured trees is not necessarily the result of a difference in logging method, but also related to differences in size class frequency distribution prior to logging. Comparing the damages in each size class in proportion to their relative frequencies before logging, showed that the differences shown in Table 4.23 remained for trees smaller than 40 cm dbh, but not for larger trees. It appeared that proportionately more large trees were damaged in the reduced impact operation, the cause of which is the augmentation in the proportion of trees damaged severely by felling and in the proportion damaged slightly by skidding. Furthermore, an interaction was observed between the logging method and logging intensity as regards the number of small trees that was destroyed by felling. Increasing the logging intensity led to a disproportionate increase of this number, which is can only be related 111 Reduced impact logging in the tropical rain forest of Guyana to group-wise felling in conventional logging, and which becomes more apparent at higher intensities. Table 4.23 Overview of the difference between logging methods in damage to the residual stand. The difference is measured as the average number of trees damaged with reduced impact logging in proportion to the damage with conventional logging. Differences were tested per size class / damage category by ANOVA (Type IV Sum of Squares): ± indicates no significant difference; +, + + and + + + indicate a significant increase in number in that class / category with reduced impact logging; and −, − − and − − − indicate a significant decrease. The number of + or - signs indicate an increase in magnitude of those differences. Empty cells indicate that this class / category did not occur in any of the logging methods. Damage caused during Dbh class (cm) Damage category killed Felling 10-20 20-40 ≥40 Total Skidding 10-20 20-40 ≥40 Total Multiple injuries ++ ± ± +++ −−− − severe injury − ± ± ± moderate injury − ± ± − − ± − ± ± − − − − −− ± ± ± ± −−− 10-20 − ± − 20-40 ≥40 Total Overall Total 10-20 20-40 ≥40 ± ± ± ± ± − ± ± − − − minor injury ± − ± − + ± ± ± ± ± ± ± total damages ± ± ± ± −− − ± −−− ± − − − −− −− ± −−− In conclusion, reduced impact logging is definitely successful in reducing skidding damage to small and medium sized trees, but at the expense of a higher impact of felling damage to small trees. In absolute numbers, felling damage to medium-sized and large trees remains unchanged. In proportion to their frequency before logging, however, felling damage to large trees is more frequent in reduced impact logging. Knowing that conventional logging is characterised by multiple treefall gaps, one would expect more extensive felling damage with reduced impact logging to medium-sized trees as well. The outcome that felling damage to medium-sized trees remains at the same level is confirmed by our field observations. In the field, it was observed that directional felling was not only successful in facilitating a reduction of skidding damage, but also in avoiding damage to medium sized trees, by redirecting the falling tree in the direction of smaller trees. 112 Chapter 4: Logging damage Eighty-five percent of the residual stand will probably survive at an intensity of 8 trees/ha with conventional logging, which reduces to 77% at 16 trees/ha. With reduced impact logging, 89% will probably survive at 8 trees/ha and 78% at 16 trees/ha. Survival rates are highest among large trees (92% to 96%) and lowest among small trees (71% to 87%). Increasing the logging intensity thus reduces the damage per extracted tree. Felling damage is reduced more strongly in conventional logging, while skidding damage is reduced more strongly in reduced impact logging. 4.7.5 Relation between tree damage and loss of canopy cover In Section 4.7.2, a discrepancy was noted between the accumulated canopy loss - no difference between logging methods (Fig, 4.2) - and the number of stems affected by one or the other logging method - conventional logging affects more trees than reduced impact logging. This dissimilarity is largely explained by breaking tree damage down by size class and cause of damage (felling or skidding). Figure 4.19 The basal area of felled trees, together with trees with dbh ≥ 30 cm that were destroyed or severely damaged by felling, in relation to the basal area felled. Crowns of felled, destroyed and severely damaged trees are assumed to have been removed from the canopy. For conventional logging the following regression line was fitted Y = 0.59ns + 1.07*** ⋅ X (R2 = 0.97) and for reduced impact logging Y = –0.08ns + 1.39*** ⋅ X (R2 = 0.97). Linear regression analysis with logging method as a dummy variable showed that the slopes differed significantly between methods (t14 =–2.61, p < 0.05), but intercepts did not (t14 = 1.89, p = 0.08). Besides the felled trees themselves, mainly the killed; i.e. uprooted and crushed trees, and severely damaged; i.e. topped and snapped trees, constitute the outline of a felling gap. Furthermore, it can be assumed that the canopy is mainly upheld by trees with dbh < 40 cm dbh. At an extraction level of 8 trees/ha on average 2.0 trees/ha with dbh < 40 cm dbh were irreversibly damaged by felling in conventional logging and 1.9 tree in reduced impact logging. At 16 trees/ha on average 2.0 trees/ha had the same fate in the conventional operation and on average 3.1 trees/ha in the reduced impact operation. Including the trees with 20 cm ≤ dbh < 40 cm, does not change the trend 113 Reduced impact logging in the tropical rain forest of Guyana observed: at an extraction level of 8 trees/ha on average 18 trees/ha with dbh < 20 cm were irreversibly damaged in the conventional operation as opposed to 17 trees/ha in the reduced impact operation. At an intensity of 16 trees/ha, this increased to 22 trees/ha in the conventional operation and 28 trees/ha in the reduced impact operation. Just as in accumulated canopy loss, everything points at an interaction between the logging method and the logging intensity. In neither case, however, this interaction proved to be significant. Nevertheless, compensating for the size of the felled trees, by expressing the extraction level as the basal area felled (m2/ha) showed a significant interaction between logging intensity and logging method: i.e. more canopy was lost with reduced impact logging at a high logging intensity. Therefore, a similar approach was used to compare the logging methods with respect to the trees that were assumed to have been removed from the canopy. The relation between the total basal area of the trees that were taken from the canopy and the total basal area felled were examined for the two logging methods. This comparison also revealed an interaction between the logging method and the logging intensity (Fig. 4.19), but certainly not as strong as when comparing canopy loss. On the one hand, this confirms the effect of group-wise felling. On the other hand, this once more shows that the estimation of canopy loss according to the method used in this study is not free of bias, mainly as a result of uncertainties in the application of the definition of a gap. 4.7.6 Impact on the commercial stand The consequences of selective logging are undeniably more serious for the commercial stand than for the non-commercial trees, because a proportion of the adult population is intentionally removed by felling. Acknowledging this, two questions arise. First, were commercial trees indeed privileged above non-commercial trees by reduced impact logging, as was intended? Secondly, how many potential timber trees survived selective logging without consequential damage? Pre-harvest inventory showed a considerable variation in the proportion of commercial trees (see Section 4.3). In the smallest size class - which was sampled at a lower rate the proportion of commercial stems varied between 16% and 35% in the 12 plots that were subjected to reduced impact logging. In the three plots that were established before logging in the area that was to be logged conventionally the proportion of commercial stems in this size class varied between 7% and 29%. Among medium-sized trees the proportion ranged from 31% to 53%, and among large trees from 50% to 86%. Because of this large variation, it was decided not to extrapolate data from samples, but to compare actual data only. This implies that logging intensity cannot be included among the design variables in case of conventional logging. The logging intensity in the three 1 ha samples varied between 13 and 20 trees/ha. For convenience of comparison, distinction was made between three diameter classes; i.e. 10 cm ≤ dbh < 20 cm (small); 20 cm ≤ dbh < 40 cm (medium-sized) and dbh ≥ 40 cm (large). The replications of each logging intensity in the reduced impact operation were tested for heterogeneity (Chi-square test) in the proportions of commercial trees damaged and 114 Chapter 4: Logging damage of non-commercial trees for each diameter class separately. In all cases, the replications appeared to be homogeneous in the proportion of commercial trees damaged by felling and in the proportion damaged by skidding. In this respect, the three samples in the conventionally logged area were also homogeneous for the size classes 10 to 20 and 20 to 40 cm, but not for the large trees. Log-linear analysis of the proportion of commercial species among damaged trees with reduced impact logging showed, that these percentages did not differ among small trees (G = 0.03, df = 1, p = 0.86), medium sized trees (G = 1.11, df = 1, p = 0.29), or large trees (G = 2.61, df = 1, p = 0.11), regardless whether the damage was caused by felling or by skidding and regardless of logging intensity. A chi-square test for heterogeneity showed that pooling of the three samples of conventional logging was allowed for two size classes; i.e. the small trees and the medium sized trees. In these two size classes, the percentage damaged trees did not differ between commercial and non-commercial species; i.e. small trees; G = 1.39, df = 1, p = 0.24, and medium-sized trees; G = 0.391, df = 1, p = 0.53, regardless whether the damage was caused by felling or by skidding. The conventionally logged plots were too heterogeneous to be pooled with regard to large trees. In one of these plots the proportion damaged trees of commercial species was smaller than for the non-commercial species (G = 6.13, df = 1, p < 0.05). Inspection of the data showed that this was caused by disproportionate skidding damage to trees of non-commercial species. It was intended to minimise damage to potential crop trees during felling in the reduced impact operation. To this extent, we regularly deviated from the felling direction that was indicated to achieve a herring-bone felling pattern. Allowance was made to deviate up to 30° from the indicated direction, since such small deviations were not regarded to hamper the extraction process (Section 3.3.4). Out of 387 observed felling events, the felling direction was adjusted for this purpose in 20% of the cases. In 50% of the latter cases the tree landed within an angle of 30° with the planned direction, and in the other 50% at a greater (undesirable) angle. Since damage assessments were carried out separately, no data were to confirm whether or not a deviation of 30° was sufficient to prevent damage to the commercial tree. Data presented in Table 4.24 suggest that in case of reduced impact logging the percentage damaged commercial trees is slightly lower than the percentage damaged non-commercial trees in the size class 20 to 40 cm dbh, and slightly higher in the size class 40 cm dbh and larger. In comparison with conventional logging, these differences are a bit more pronounced. However, the test results indicate that none of these differences is significant. It cannot be excluded that this is caused by the unbalanced sample procedure. The sample plots in the conventionally logged area were not only smaller in number, but also smaller in size. Another reason that there was no significant difference could be that the plots subjected to reduced impact logging were rather densely stocked with potential crop trees. On average, the residual stand contained an estimated 76 trees of commercial species per hectare between 10 and 20 cm dbh, and 60 trees of commercial species per hectare between 20 and 40 cm. The residual number 115 Reduced impact logging in the tropical rain forest of Guyana of trees of commercial species per hectare of 40 cm and above obviously varied with logging intensity; i.e. between 15 and 27 per hectare. If these trees were regularly distributed, thereby surrounding the tree to be felled in all directions, it could well have been impossible to significantly reduce the number of damaged potential crop trees. Table 4.24 The proportion of the residual stand belonging to commercial species (Com.) that was damaged by felling and by skidding and the proportion of trees belonging to non-commercial species (Non-com.) that was damaged with two logging methods and different logging intensities. The proportions were derived by pooling of replications in case of reduced impact logging. In case of conventional logging only the three 1 ha plots were used in this analysis. The proportion of commercial trees that was damaged did not differ from the proportion among non-commercial trees. The proportions remained similar whether damage was caused by felling or by skidding, whether the logging intensity was low or high, or whether logging was carried by the one or the other method. See text for further explanation. Type of logging Dbh class (cm) Com. Reduced impact 4/ha Percentage of residual stand damaged Felling Skidding Non-com. Com. Non-com. 10-20 20-40 ≥40 8.1 10.1 10.8 10.4 10.0 6.7 11.3 4.0 2.5 6.7 3.9 5.7 8/ha 10-20 20-40 ≥40 15.1 17.5 13.9 15.8 17.8 10.9 8.5 7.0 10.1 12.4 7.1 7.3 16/ha 10-20 20-40 ≥40 28.3 27.0 28.4 24.8 31.1 17.2 16.1 4.6 4.4 13.5 6.8 14.3 pooled intensities 10-20 20-40 ≥40 16.3 17.8 15.2 16.4 18.7 11.8 11.8 5.9 7.8 11.6 6.5 8.3 Conventional 10-20 20-40 ≥40 15.0 25.9 21.3 20.7 24.3 27.3 17.5 18.4 11.7 24.4 16.5 25.5 In order to find out if this line of reasoning cuts any ice, the spatial distribution of trees of commercial species was investigated further. Spatial patterns can be broadly classified as ‘random’, ‘regular’ or ‘aggregate’ (Diggle 1983; Fröhlich & Quednau 1995), without implying any specific causal mechanism. In order to establish whether the trees belonging to commercial species assumed a particular pattern, a simple quadrat count procedure was followed. Under complete spatial randomness, the quadrat counts would follow a Poisson distribution. Each plot was partitioned into 49 quadrats of 400 m2 and the counts of the number of trees with a diameter between 20 and 40 cm and of trees with a diameter of 40 cm and above prior to logging were used to test randomness of the spatial distribution of these two size classes. Pooling of plots was allowed with respect to the size class 40 cm dbh 116 Chapter 4: Logging damage and above (Heterogeneity χ2 = 18.32, df = 19, p = 0.50), but not for medium-sized trees (Heterogeneity χ2 = 47.19, df = 27, p < 0.01). The spatial distribution of the large commercial trees proved to be random (coefficient of dispersion: 1.12; χ2 = 5.490, df = 4, p = 0.24). With respect to medium-sized trees, most plots showed a random distribution. Quadrat counts in two out of the twelve plots did not fit a Poisson distribution; i.e. commercial trees in this size class were regularly distributed in plot 4 (coefficient of dispersion: 0.85; χ2 = 10.94, df = 3, p < 0.05) and showed aggregation in plot 15 (coefficient of dispersion: 2.24; χ2 = 9.29, df = 4, p < 0.05). The two diameter classes were combined and plots were subsequently tested for heterogeneity, which was non-significant (Heterogeneity χ2 = 39.20, df = 36, p = 0.33). Subsequent quadrat counts showed that commercial trees with dbh ≥20 cm showed a clumped spatial distribution (coefficient of dispersion: 1.20; χ2 = 17.79, df = 8, p < 0.05). This result is of great importance because it indicates that felled trees were surrounded by other commercial trees. Since most cut trees belonged to Chlorocardium rodiei, the spatial distribution of trees of this species was examined separately. Medium-sized Chlorocardium rodiei trees occurred in a distinct aggregated pattern (coefficient of dispersion: 1.77; χ2 = 25.22, df = 3, p < 0.001) and so did large Greenhearts (coefficient of dispersion: 1.57; χ2 = 30.20, df = 3, p < 0.001). Joining of the two size classes also resulted in a distinct aggregated pattern (coefficient of dispersion: 1.54; χ2 = 75.15, df = 4, p < 0.001). This indicates that the clusters of medium-sized trees were not clearly separated from the clusters of large trees. These results confirm the field observation that Greenhearts selected for felling were usually surrounded on all sides by conspecific trees (see Fig. 4.20). It was thus extremely difficult to find a felling direction that would not affect a tree belonging to a commercial species. Directional felling did not result in smaller numbers of affected commercial trees, but it may have had an effect on the severity of the damage. Table 4.25 shows that the damage to commercial trees was less severe than to non-commercial trees; there was a trend, but not significantly so, in the size classes 10 to 20 cm dbh and 20 to 40 cm dbh, and significantly so in the size class 40 cm dbh and above. In the conventionally logged area, on the other hand, trees of commercial species between 10 and 20 cm dbh were more seriously damaged than trees of non-commercial species, whereas there was a trend to small, non-significant differences in the two other size classes. Felling damage was more severe in the reduced impact operation which is probably related to a different felling pattern (see the previous section). In order to establish whether directional felling minimised severe damage to commercial trees, the proportion of irreversibly damaged commercial trees in relation to the proportion of irreversibly damaged non-commercial trees in conventional logging was compared with the same relation in reduced impact logging. 117 Reduced impact logging in the tropical rain forest of Guyana 0 20 40 60 80 0 20 40 60 80 Figure 4.20 Spatial distribution of trees of commercial species in two 2-ha plots that were subjected to reduced impact logging. Open dots portray Chlorocardium rodiei and black dots portray other commercial species. The size of the symbol is proportionate to the diameter of the tree. Note that lesser used species are not included. Table 4.25 suggests that, in relative terms, damage to commercial trees in the 10 to 20 cm dbh size class was less severe with reduced impact logging than with conventional 118 Chapter 4: Logging damage logging, but this was not significantly so (G = 3.60, df = 1, p = 0.058). There definitely were no differences between the logging methods concerning the 20 to 40 cm dbh class (G = 0.16, df = 1, p = 0.69) or the 40 cm and above dbh class (G = 2.09, df = 1, p = 0.15). Table 4.25 Proportions of irreversibly damaged (killed, topped, snapped, partly uprooted) trees among trees damaged by felling in three diameter classes, subdivided by commercial and non-commercial species. n is the number of encountered damaged trees per category, G is the statistic of the log likelihood ratio test on difference between species groups, and p the corresponding probability with one degree of freedom. Logging method Dbh class (cm) Reduced impact 10-20 20-40 ≥40 Conventional 10-20 20-40 ≥40 Proportion irreversibly damaged Commercial species n 190 85% 245 58% 89 29% 86 120 65 83% 47% 28% Non-com. species n 657 86% 374 60% 45 47% G 0.12 0.23 3.93 p 0.72 0.63 0.048* 400 286 25 4.70 0.75 0.09 0.03* 0.39 0.77 71% 51% 24% The effect of directional felling, though noted in the field, thus was statistically untraceable. This is most likely linked to the aggregated distribution of commercial trees, which makes it almost impossible to choose a felling direction that simultaneously fulfils the preconditions set to facilitate extraction, does not risk lodging into a neighbouring large tree and is devoid of potential crop trees. 4.7.7 Impact on sustainable production of timber Having established that, with regard to trees with a dbh of less than 40 cm, commercial species and non-commercial species were damaged to the same degree, it appears that the major difference in impact on commercial versus non-commercial species concerns the trees of 40 cm dbh and above. A portion of variable size, depending on the logging intensity, of the commercial species was removed by logging, which resulted in a reduction of their share in the residual population. What are the consequences of this impact for future harvests? Of course, marketability of species changes in time as clearly demonstrated in Sections 1.5.3 and 2.4.4 (see also Section 4.1). Building a silvicultural system, whereby lesser know species are included besides the presently commercial species, implies a general continuity of timber supply. In turn, this implies that such a system allows depletion of one species because it assumes its replacement by another. Such a conception of ‘sustainability’ is against silvicultural principle and therefore, in order to obtain a clear picture of the sustainability of selective logging, the system focuses on those species which are presently commercial (see Appendix A). Not all trees of those species qualify as future crop trees. Defective stems are a widespread phenomenon in rain forest exploitation. Stems can be rejected for several reasons; visible and 119 Reduced impact logging in the tropical rain forest of Guyana internal decay, deformities and wood splits or shakes being the most common ones. Trees of which the stems had serious defects neither qualified as growing stock. Because pre-harvest data for the conventional logged area were only available for three small plots and because no difference was found between the logging methods in the proportion of damaged trees of commercial species, further details are limited to the experimental plots only. In the reduced impact plots, 30% of the trees belonging to commercial species had serious defects. Occurrence of defects, especially decay, varied between size classes. Large trees (dbh ≥ 40 cm) were more often rejected (48% of the trees in this size class, sd = 10%, n = 12) than medium-sized (20 cm ≤ dbh < 40 cm 24% rejected, sd = 7%, n = 12) or small trees (10 cm ≤ dbh < 20 cm - 27% rejected, sd = 11%, n = 12). Prior to logging, trees of commercial species without defects had a 25% share of the total population with dbh ≥ 10 cm (sd = 5%, n = 12), which corresponded to a 33% share of the total basal area (sd = 4%, n = 12). The share of commercial species increased with increasing diameter: nineteen percent (sd = 6%) of the trees with 10 cm ≤ dbh < 20 cm belonged to commercial species; 32% (sd = 6%) of the trees with 20 cm ≤ dbh < 40 cm; and 42% (sd = 4%) of the trees with dbh ≥ 40 cm. Selective logging removed a portion of trees of commercial species with dbh ≥ 40 cm, which varied along with logging intensity. Another portion was destroyed or seriously damaged. Table 4.25 shows that commercial trees were damaged to the same degree as non-commercial trees with regard to the trees smaller than 40 cm dbh. Selective logging thus resulted in the reduction of the share of trees of commercial species with dbh ≥ 40 cm, no or reversible damage and without defect - from 38% to 32% at a logging intensity of 4 trees per hectare, from 43% to 33% at an intensity of 8 trees per hectare and from 45% to 21% at 16 trees per hectare. Focusing on Chlorocardium rodiei (Greenheart) - a species deserving particular attention, because it is both a major timber species and one of the most common commercial species in these forests - the following was observed. The proportion of defective Greenhearts amounted to 25% (sd = 9%). This proportion is smaller than the one observed above. This is mainly related to the small proportion of defects among the Greenhearts with 20 cm ≤ dbh < 40 cm; i.e. 16% (sd = 9%). Prior to logging, the percentage of Greenhearts without defect amounted to 8% (sd = 3%) of the total population (with dbh ≥ 10 cm), corresponding to a 14% (sd = 5%) share of the total basal area. Including the defective Greenhearts shows how dominant the species was in the upper storey; i.e. 32% of all trees with dbh ≥ 40 cm, while being less numerous in the lower storeys; i.e. 12% of all trees with 20 cm ≤ dbh < 40 cm and 6% of all trees with 10 cm ≤ dbh < 20 cm. After logging with an intensity of 4 trees/ha, the proportion of Greenhearts with dbh ≥ 40 cm without defect and without ‘irreversible’ damage was reduced from 14% of the initial to 10% of the residual population. After a logging intensity of 8 trees/ha its proportion was reduced from 23% to 17%, and at an intensity of 16 trees/ha from 23% to 10% 120 Chapter 4: Logging damage From a silvicultural point of view, it is important to establish how much of the original stand survived without irreversible damage, because those trees form the population from which the next crop will have to be recruited. If the remaining commercial portion of the stand is too small, the extraction level has been too high to expect a next harvest within a reasonable cutting cycle length. In the past, silvicultural ‘thinning’ treatments have been advocated to boost the share and growth rate of commercial trees (e.g. Dawkins 1958; Synnott & Kemp 1976; De Graaf 1986 see also Sections 2.3 and 6.1.1). In recent years, it has been realised that logging must be considered as a silvicultural tool, implying that logging intensities and methods should be designed accordingly (e.g. Lee 1982b; De Graaf 1986; Hendrison 1990; Dawkins & Philip 1998). From this point of view, it is thus more important to know how much has survived logging than how much has been lost. Linear regression analysis was used whereby the proportion of the original stand with less than a specified maximum damage was described as a function of logging intensity. Three equations were fitted for three levels of damage (Table 4.26). Differences between the regression lines were tested by using one regression model with two dummy variables X1 and X2 - for the three injury levels. The injury level ‘not felled’ (all residual and destroyed trees) is represented by X1 = 1 and X2 = 0, the injury level ‘no or reversible damage’ (all residual trees except those with severe damage) by X1 = 0 and X2 = 1, and the injury level ‘no damage’ by X1 = 0 and X2 = 0. The analysis showed that all intercepts of the regression lines in Table 4.26 differ significantly (p < 0.01) from each other and that the slope of the line for the category ‘not felled’ differed from the other two lines (p < 0.05), but the slopes of the lines for the categories ‘no or reversible damage’ and ‘no damage’ did not. Table 4.26 Impact of logging intensity on the commercial stand and on Chlorocardium rodiei , using the percentage of the original basal area with a specified maximum level of injury as a function of basal area felled (m2/ha). Fitted coefficients for the regression models Yi = ai + bi ⋅ X + ε, where Yi = percentage of the original basal area with level of injury i , X = basal area felled (m2/ha), and ai and bi are fitted coefficients. The probability of coefficients being zero is very small for all regression lines: p < 0.001. See text for differences between intercepts and slopes and further explanation. Level of injury Not felled No or reversible damage No damage Fitted coefficients Commercial species pooled a b R2 a 98.96 -11.04 0.91 93.25 97.72 -13.07 0.94 91.06 89.65 -14.45 0.97 83.62 Greenheart b -10.57 -10.75 -12.75 R2 0.60 0.67 0.80 The impact of the extraction of commercial trees on the percentage of the original basal area is considerable. According to the regression model, 12% is removed at a logging intensity of 1 m2/ha, which increases to 45% at a logging intensity of 4 m2/ha. The percentage injured concurs with the percentage removed after a light harvest (13%), and increases disproportionately to 23% after a fourfold increase of the logging intensity. Most of this damage is reversible. According to the regression model, the incidence of irreversible damage; i.e. broken, snapped, topped and (partly) uprooted trees, amounts to only 3% of the original basal area at a logging intensity of 1 m2/ha and rises to 9% at a logging intensity of 4 m2/ha. The incidence of mild damage; partial 121 Reduced impact logging in the tropical rain forest of Guyana excision of the crown and bark lacerations, does not increase significantly when augmenting the logging intensity. Assuming that most trees will recuperate from mild damage, the population from which the next cut will have to be recruited will be 85% of the original stand after a light harvest of 1 m2/ha, but will be reduced to only 45% after a heavy harvest of 4 m2/ha. Table 4.27 Logging intensity Impact of logging on residual commercial stems with dbh ≥ 10 cm. Percentage of original number of stems felled, irreversibly damaged (destroyed, snapped, topped, split, heavy lean), reversibly damaged (mild crown and/or mild stem damage) and undamaged in a reduced impact operation with varying logging intensity. The frequencies of damaged stems after logging were recorded in a 100% inventory of stems with a dbh ≥ 20 cm, sample size per exploitation level is 6 ha, except for intensity 8 trees/ha where the sample size is 12 ha. The frequencies of damaged stems with 10 cm ≤ dbh < 20 cm are extrapolated from a 12.5% sample, total sample size per exploitation level is 0.75 ha, except for intensity 8 trees/ha where the sample size is 1.5 ha. Dbh class (cm) n/ha Percentage in each damage category Felled 4/ha Total Unaffected 64 47 18 129 0.0 0.0 22.2 3.2 10-20 20-40 ≥40 52 47 22 121 0.0 0.2 36.3 6.7 16.7 10.2 2.7 11.6 3.8 14.6 12.2 9.5 79.5 75.0 48.9 72.1 10-20 20-40 ≥40 47 42 24 112 0.0 0.8 67.6 14.5 34.3 17.8 1.4 21.2 14.3 15.0 11.5 14.0 51.4 66.4 19.4 50.3 Total 16/ha Reversible damage 4.2 6.5 9.3 5.7 10-20 20-40 ≥40 Total 8/ha Irreversible damage 16.7 4.3 2.8 10.2 79.2 89.1 65.7 80.9 A similar regression model was constructed for Chlorocardium rodiei alone. According to that model, the percentage of the original Greenheart basal area extracted was greater than for other commercial species combined. The regression lines for the percentage ‘not felled’, ‘no or reversible damage’ and ‘no damage’ had different intercepts (p < 0.001), but slopes did not differ. This implies that there is no significant increase in incidental damage with increasing logging intensity. According to the model, 17% is removed at a logging intensity of 1 m2/ha, which increases to 49% at a logging intensity of 4 m2/ha. The percentage of basal area that survived logging with no or only mild damage amounts to 80% after a light harvest of 1 m2/ha which is reduced to 48% after a fourfold increase in logging intensity. Only a very small proportion of the original Greenheart stand is incidentally destroyed or seriously injured (between 2% and 3%, depending on the logging intensity), suggesting that, contrary to the results from the log-linear analysis, Greenheart has been spared to some degree by directional felling. 122 Chapter 4: Logging damage Table 4.27 shows that the incidence of damage is not proportional to the number of trees felled. Harvesting 4 trees/ha was found to result in the destruction or serious damage of 10% of the commercial trees with dbh ≥10 and to cause reversible injury to another 6%, compared with 21% irreversibly damaged and 14% reversibly damaged at a harvest of 16 trees/ha. In other words, quadrupling the logging intensity led to doubling of the damage. However, only 31% of the original number of trees with dbh ≥40 cm will probably recover from the logging operation at an intensity of 16 trees/ha, whereas this is still 61% at an intensity of 8 trees/ha. At 16 trees/ha, only an estimated 7 trees/ha will be left after logging. Assuming that the trees in the 20 to 40 cm dbh class are evenly distributed throughout the class - which, obviously, is a simplified representation of reality - and that each tree will grow at an average rate of 3 mm/yr., stand table projection suggests that it takes 10 trees 20 years to move into the next diameter class. With a cutting cycle of 25 years and neglecting mortality, a next harvest at this intensity may thus be possible, but for the following cutting cycle all trees will have to be recruited from the 20 to 40 cm size class, which will take much longer than 25 years. With the same assumptions, stand table projection with an intensity of 8 trees per hectare will result in 27 trees per ha in the 40 cm and above size class after 25 years, which suggests that this exploitation level may be sustainable. Clearly, at this stage it is not possible to demonstrate conclusively that a logging intensity of 8 trees/ha is a sustained yield of timber and maintains the forest ecosystem in a condition that will guarantee its ‘environmental sustainability’. A final conclusion can only be drawn at a much later stage (Poore et al. 1989, p. 191):only when a forest, managed under a specific silvicultural system, is in, at least, its third rotation, and still ‘retains the full forest structure, is fully stocked with commercial species which are growing well, and possesses adequate regeneration and an intact soil and ground flora.’ I will return to the matter of the desirable extraction level in chapter 6. 4.8 DISCUSSION 4.8.1 Comparison of results of logging damage studies Implementation of reduced impact logging techniques has been reported to result in spectacular reductions in damage to the remaining trees. Key elements in reducing damage are cited to be proper planning of the logging operation, directional felling, climber cutting and winching. Bertault & Sist (1995), Pinard & Putz (1996 ), Johns et al. (1996) and Hendrison (1990) reported a reduction of the proportion of the remaining trees that were damaged by around 20% or more. Furthermore, strong reductions in canopy removal and skid trail coverage have been reported. In this study, reduced impact logging - featuring the key elements mentioned above - resulted in a reduction in damage to the residual stand by 6% at an extraction level of 8 trees/ha and by 3% at a level of 16 trees/ha. These reductions are clearly lower than those found elsewhere (Fig. 4.21). The achievements are even more divergent when comparing canopy loss (Fig. 4.22). All studies report a reduction of canopy loss, but an increase in canopy loss was observed at 123 Reduced impact logging in the tropical rain forest of Guyana a logging intensity of 16 trees/ha in the present study. In fact, only the reduction of the skid trail coverage is in agreement with the results of other comparative studies (Fig. 4.23): from 13% to 8% of the total area at a logging intensity of 8 trees/ha and from 21% to 8% at an intensity of 16 trees/ha. Figure 4.21 Comparison between logging operations of total residual stem damage estimates for trees with dbh ≥ 10 cm. Damage is defined as the sum of destroyed and injured residual stems. The relationship between the percentage of residual stem damage and logging intensity is highly significant, but explains only a moderate amount of the total variation: (% canopy loss) = 13.16 + 1.96 ⋅ (number of trees removed ha–1); r2 = 0.49. Studies used in this comparison were conducted in Pará State, Brazil (Bz1: Uhl & Vieira 1989; Bz2: Veríssimo et al. 1992; Bz3: Veríssimo et al. 1995; Bz4: Johns et al. 1996), Costa Rica (CR: Webb 1997), Gabon (Gb: White 1994), French Guiana (FG: Schmitt & Bariteau 1990), East Kalimantan, Indonesia (EK1: Abdulhadi et al. 1981; EK2: Bertault & Sist 1995), West Malaysia (WM: Johns 1988), Sabah, Malaysia (Sb1; Nicholson 1979; Sb2: Pinard and Putz 1996), Sarawak, Malaysia (Sw: Mattson Mårn & Jonkers 1981), Suriname (Su: Jonkers 1987), and the present study (Gy). Multiple regression analysis with all sites included and logging method as a dummy variable increased the explained variation: R2 = 0.55, but failed to show a significant difference between logging methods (difference between intercepts: t23 =–1.76, p = 0.09). However, if the analysis is restricted to the five studies that compare conventional and reduced impact logging (Bz4, EK2, Sb2, Sw, Gy) a significant difference is revealed between logging methods (intercepts only - t11 =–2.68, p < 0.05). Still, only a moderate proportion of the total variation is explained; R2 = 0.64. It is difficult to make comparisons between sites, because few studies include all logging damage variables, and because of a perplexing array of methods of assessing logging damage. Logging intensities - once they are mentioned which is sometimes not the case - differ considerably. In general, yields from South East Asian logging operations are considerably higher than their neotropical (and African) counterparts. The number of boles removed per hectare lies generally below 10 in the neotropics, whereas it usually is above this level in South East Asia. However, also within a region considerable variation occurs. In order to adjust for these differences in logging intensity a regression approach was used in Figs. 4.21, 4.22 and 4.23. 124 Chapter 4: Logging damage The level of residual damage (Fig. 4.21) is significantly correlated with the number of boles extracted, but this appears to explain only a modest portion of the total variation (r2 = 0.49). A part of the remaining variation is explained by the difference in impact between reduced impact and conventional logging. If the regression analysis is restricted to the five comparative studies, it appears that the relative importance of the logging intensity and logging method do not differ much: the standardised partial regression coefficient β = 0.51, for logging intensity and β = 0.50, for logging method. The amount of explained variation remains modest (R2 = 0.64). There are several other explanations why variation is high. Harvesting forests of different statures can not but result in different damage levels at similar extraction intensities, one would expect. This was indeed demonstrated by e.g. Hendrison (1990), but could not be confirmed by the present study (see also Whitman 1997). Similarly, minimum felling diameter limits can frustrate a proper comparison. Unfortunately, few studies yield information on the average diameters of the cut trees. Another source of variation may result from sampling methods (cf. Van Rompaey 1993; Rossignol et al. 1998, p. 151). Except in the studies by Schmitt & Bariteau (1990) and Bertault & Sist (1995), small trees were sampled at lower rates than large trees. Furthermore, results were often based on small samples. This study suggests that this may be an important source of bias, because of the strong variation in density of - especially - small trees and most casualties are falling in that size class. Similarly, not all studies were so-called ‘before-and-after’ studies. If trees are not numbered and marked before logging, it is not possible to determine exactly how many trees were smashed, because such trees may be hidden under the crowns of fallen trees, or totally obliterated by the skidder. Furthermore, it is not possible to determine consistently whether for instance crown damage had not existed before logging commenced. Careful examination of Fig. 4.21 reveals that, in three out of the four studies that compared conventional and reduced impact logging, - intensity-adjusted - residual damage in the reduced impact type of logging higher than in the conventional type in Guyana. So, the spectacular damage reductions should be seen against a high damage benchmark. In case of a low damage benchmark as in Guyana, it apparently more difficult to achieve massive damage reductions. This comparison could be conceived to indicate that conventional logging results in extraordinarily low residual damage in Guyana. However, damage levels in Guyana are coherent with findings in two studies in the Brazilian Amazon (Uhl & Vieira 1989; Veríssimo et al. 1995), Costa Rica (Webb 1997), French Guiana (Schmitt & Bariteau 1990) and Suriname (Jonkers 1987), indicating that such an interpretation would not be just. In the latter perspective, the improvements achieved by engaging the key elements mentioned above appear to be modest in this study. The potential of directional felling alone in reducing logging damage is questionable (Nicholson 1965; Fox 1968a) and the usual perception is that damage reduction is rather related to climber cutting than to directional felling per se (e.g. Johns et al. 1996). Perhaps, the most important reasons for the marginal success can be found in 125 Reduced impact logging in the tropical rain forest of Guyana the numerically untraceable effect of liana cutting and the influence of the felling pattern (‘clustered tangles’ versus ‘herring-bone’). Cutting of lianas is one of the most frequently mentioned tools to reduce felling damage (Fox 1968; cf. Jacobs 1976, 1988; Appanah & Putz 1984; Pinard et al. 1995; Cedergren 1996). Figure 4.22 Comparison of canopy loss estimates as a function of logging intensity (n/ha) for conventional and reduced impact logging operations in several sites. The relationship between the loss of canopy cover and logging intensity is significant, but explains a meagre proportion of the total variation; (% canopy loss) = 12.79 + 1.13 ⋅ (number of trees removed ha–1); r2 = 0.22, F1,17 = 4.90, p < 0.05. Studies used in this comparison were conducted in North Queensland, Australia (Au: Crome et al. 1992), Pará State, Brazil (Bz1: Uhl & Vieira 1989; Bz2: Veríssimo et al. 1992; Bz3: Veríssimo et al. 1995; Bz4: Johns et al. 1996), Costa Rica (CR: Webb 1997), Gabon (Gb: White 1994), French Guiana (FG: Schmitt 1989), Suriname (Su: Jonkers 1987), Sarawak (Sw: Mattson Mårn & Jonkers 1981), and this study (Gy). Multiple regression analysis with all sites included and logging method as a dummy variable increased the explained variation: R2 = 0.36, but failed to show a significant difference between logging methods (difference between intercepts: t11 = –1.89, p = 0.08). If the analysis is restricted to the three studies that compare conventional and reduced impact logging (Bz4, Sw, Gy) no significant difference is revealed either (difference between intercepts: t11 = –1.53, p = 0.18). In this study liana cutting did not seem to have any effect: single tree fall gaps did not differ in area, gaps containing two trees were slightly smaller with reduced impact logging, whereas multiple tree gaps were larger (see also Parren & Bongers 1999; Liew 1973). It is usually recommended that lianas be cut about one year prior to logging (Cedergren 1996; Pinard et al. 1995). In this study, lianas were cut only six months prior to felling, which is probably too short a period. Parren & Bongers (1999) found that only 50% of lianas had died six months after being cut in Cameroon. Even when dead, vine stems connecting crowns may not have weakened sufficiently to prevent uprooting or snapping of neighbouring trees, and it may be necessary to cut lianas even two years prior to felling. On the other hand, it is known that many species of liana have the ability to resprout vigorously and vine sprouts may reach the canopy within a two year period (Oldeman 1990; Coudurier 1992). One may thus argue that the period of six months was too short a period. Indeed, field observations during felling confirmed that in some 126 Chapter 4: Logging damage cases the liana stems had not weakened sufficiently. Another explanation why there was no traceable effect of liana cutting, could be a low density of lianas in the conventionally logged area. In summary, there is probably an element of truth in the view that a lack of success of liana cutting explains why no reduction in canopy loss was achieved. In general, some reservation is warranted before advocating liana cutting - especially if its beneficial effect is not instantly clear as in this case - because of its implications on wildlife, particularly frugivores and hence the stability of the ecosystem. Figure 4.23 The relationship between logging intensity (n/ha) and the area affected by skidding as percentage of the total area for conventional and reduced impact logging operations in several sites. There is no significant linear relationship (r2 = 0.05, F1,24 = 1.24, p = 0.28), but the intercept differs significantly from zero; i.e. 9.6% (t24 = 3.36, p < 0.01). Studies used in this comparison were conducted in North Queensland, Australia (Au: Crome et al. 1992), Pará State, Brazil (Bz2: Veríssimo et al. 1992; Bz3: Veríssimo et al. 1995; Bz4: Johns et al. 1996), Costa Rica (CR: Webb 1997), Gabon (Gb: White 1994), French Guiana (FG: Schmitt 1989), East Kalimantan, Indonesia (EK1: Abdulhadi et al. 1981; EK2: Bertault & Sist 1995), Sabah, Malaysia (Sb1; Nicholson 1979; Sb2: Pinard & Putz 1996), Suriname (Su1: Jonkers 1987; Su2: Hendrison 1990), Sarawak (Sw: Mattson Mårn & Jonkers 1981), and this study (Gy). Multiple regression analysis with logging method as a dummy variable revealed a significant difference between methods (difference between intercepts: t23 = –2.56, p < 0.05) when all sites are included, but the explained variation increased only slightly R2 = 0.26. If the analysis is restricted to the studies that compare conventional and reduced impact logging (Bz4, EK2, Sb2, Sw, Gy) differences between methods remain unchanged (difference between intercepts: t10 = – 2.83, p < 0.05), but the explained variation becomes much higher (R2 = 0.46). Leaving unresolved for the time being the question whether liana cutting did meet the objectives or not, there are clear indications that the difference in felling patterns obscured a clear view on the effect of the introduction of directional felling, climber cutting, etc. In conventional Greenheart logging operations, trees are often felled in clusters and often resulting in felled trees ending up piled on top of one another. In the reduced impact logging operation, it was avoided as much as possible to fell trees in clusters and when this could not be avoided trees were directed in such a way as to 127 Reduced impact logging in the tropical rain forest of Guyana make separate gaps. Both measures were taken to avoid large composite gaps, as is often recommended (e.g. Johns et al. 1996; Webb 1997). Indeed fewer composite gaps occurred with reduced impact logging. In the reduced impact operation, 23% and 49% of the felling chablis were composite chablis at intensities of 8 trees/ha and 16 trees/ha respectively, while this amounted to 64% and 76% respectively in the conventional operation. Skid trail Skid trail Figure 4.24 Hypothetical felling patterns in a conventional (left) and reduced impact logging operation (right) at similar high logging intensity. Note that the chance that very large gaps are formed is higher in case of linked gaps with reduced impact logging than with conglomerate gaps formed by piling up of felled trees with conventional logging. The difference is created by the degree of clustering of target trees and by difference in felling directions The size of the composite gaps (with three or more trees) was larger with reduced impact logging (Fig. 4.8). This has two causes, to wit, the spatial distribution of the felled trees and the felling pattern. Greenheart has an aggregated spatial distribution (Section 4.7.6, Fig. 4.20). Conventional logging usually removes all sound Greenheart trees above a minimum diameter within a cluster. In the conventional operation that was monitored in this study, the stocking density varied and with it the number of trees in one cluster and the local logging intensity. It is therefore no surprise that composite tree fall gaps occurred more often at higher logging intensity. In the reduced impact operation, it was avoided as much as possible to fell neighbouring trees. The felling patterns differed as well. In the conventional operation, the feller tried to reduce the chance that the felled tree was caught in or damaged by surrounding trees - as far as his limited technique and tools afforded him to have any control over the felling direction This often resulted in the boles of felled trees being piled up in tangles. In the reduced impact operation, where felling technique and tools allowed the feller to have more control of the direction of the lie, the choice of the felling direction was primarily dictated by the desire to produce a herring-bone pattern, which would increase skidding efficiency and reduce skidding damage. If the distance between the target trees were sufficiently large, this would usually lead to the formation of separate gaps. However, increasing the 128 Chapter 4: Logging damage intensity increased the number of interlinked gaps. At similar extraction intensity, these ‘concatenated’ gaps would be larger than in case of the tangles produced in the conventional operation, because there the area of overlap is larger (Fig. 4.24). Although the rate of canopy loss was proved not to be reduced by implementing reduced impact logging techniques in the present study, damage to the remaining trees was reduced at both logging intensities. This reduction should be mainly ascribed to the strong decrease in skidding damage that was achieved (Fig. 4.16, Table 4.20). The proportion of residual trees damaged by skidding was reduced from 12% to 8% at a logging intensity of 8 trees/ha and from 16% to 9% at a logging intensity of 16 trees/ha. For the number of trees damaged by felling a reduction was found from 15% to 12% at a logging intensity of 8 trees/ha, whereas an increase from 22% to 24% was found at a logging intensity of 16 trees/ha. Felling and skidding damage taken together, residual damage was reduced from 26% to 20% and from 36% to 32% at logging intensities of 8 trees/ha and 16 trees/ha, respectively, by implementing improved logging methods. How can these results be reconciled with the fact that canopy loss was not reduced, respectively increased strongly? In the values listed, all damage categories are included. However, in order to link canopy loss to tree damage, the focus should be on severely damaged - uprooted, broken, snapped, topped and heavily leaning - trees, because it is mainly the removal of the crowns of these trees that constituted the gaps we measured. It was found that the number of trees irreversibly damaged by felling did not differ at an intensity of 8 trees/ha - remaining at a level of 8% - but increased strongly at an intensity of 16 trees/ha - an increase from 12% to 18% by applying reduced impact logging. These trends are more in agreement with the trends in the rates of canopy loss found. The number of trees irreversibly damaged by skidding was much smaller in the reduced impact operation: a reduction 7% to 3% and from 9% to 4%, respectively, at logging intensities of 8 trees/ha and 16 trees/ha respectively. Felling and skidding damage taken together, this meant a reduction of irreversible residual damage from 15% to 11% and from 22% to 21% respectively. What we see here, is a trade-off between trying to reduce skidding damage and trying to reduce felling damage. At a logging intensity of 8 trees/ha, where most gaps are separate gaps in reduced impact logging, neither total canopy loss nor severe residual felling damage was reduced, whereas skidding damage was reduced. The conventionally logged forest featured mainly composite tree fall gaps, irrespective of logging intensity. It is warranted to assume that composite gaps result in less residual damage than separate gaps, because there will always be some overlap in area, and hence in potential casualties. At high logging intensity this effect becomes more outspoken, because the numbers of neighbouring trees felled in one gap in the conventional operation became larger, and with them the overlap in areas affected by each separate tree fall. Strict adherence to the herring-bone felling pattern thus appears to have certain drawbacks, which became more outspoken at higher logging intensities. In the conventional operation, the spatial distribution of the felled trees became more 129 Reduced impact logging in the tropical rain forest of Guyana clumped with increasing intensity, whereas gaps started to interlink in the reduced impact operation. Indeed, the values for canopy loss are high when compared to the proportion of the residual stand that was severely damage. This discrepancy is partly caused by the removal of the target trees from the canopy. Adding the felled trees does not change things much, however, because only 2% to 4% of the original stand were harvested. The proportion of the original basal area that was felled, destroyed or severely injured (Table 4.14) shows a much stronger resemblance to the rate of canopy loss; i.e. 15% in both conventional and reduced impact logging operation at a logging intensity of 8 trees/ha, and 24% and 28% respectively, at a logging intensity of 16 trees/ha. Some of this basal area was destroyed or damaged during skidding and an assessment of logging damage based on the extent of the canopy loss therefore still seems likely to result in overestimation of the basal area removed or lost in the operation, and a serious overestimation of the number of trees removed or lost (cf. Jonkers 1987, p.92). As a matter of fact, measuring gaps as a two-dimensional projection of ‘hole’ in the canopy which is assessed from the ground with little precision, and delineated as if it were cylindrical chimney, although undoubtedly being useful in estimating abiotic conditions (Brown 1993; Whitmore et al. 1993), holds a poor bearing to the change that has taken place in the three-dimensional forest stand. First, by using such a method, changes in the canopy profile (cf. Crome et al. 1992) are neglected. The change of the canopy profile is not a simple loss of upper parts of the canopy, but heavy reductions in LAI take place in middle canopy regions as well. Secondly, it has been shown repeatedly that the interpretation of what constitutes the edge of a gap is not free of bias (Van der Meer et al. 1994; Brils & Laan 1995; Mekkink & Nijmeijer 1998; Hammond, in prep.). The latter is obviously associated with the fact that there inherently is no strict boundary between gap and forest, but rather a ‘fuzzy’ transition zone (Oldeman 1978, 1983, 1992; Florence 1981; Vester 1997; Rossignol et al. 1998). Thirdly, this study showed that the presence of natural chablis interferes with the assessment of the outline of a gap. Low regrowth can constitute a ‘canopy’ in a natural chablis. A tree directed into such a patch of low vegetation will inevitably obliterate much of this ‘young canopy’. This obliterated vegetation will be included in the felling gap as it is defined here. Obliteration of low vegetation under a mature forest canopy will not contribute to the felling gap and hence produce a smaller gap. Other methods, taking account of the three-dimensional nature of a forest stand and the spatial variability in damage that is so typical of logging operations in rain forest, are available (e.g. Lieberman et al. 1989; Crome et al. 1992; Koop & Sterck 1994; Van der Meer 1997), but unfortunately more time consuming than the two-dimensional projections methods. 4.8.2 The effect of logging intensity Despite the fact that the studies, referred to above, reported on different types of forest, felling diameters and logging techniques and used different methods to assess damage, 130 Chapter 4: Logging damage it is evident from all cases that damage to the residual stand increases as the density of the harvested trees increases. Introduction of reduced impact logging techniques appeared to have a similar weight in other studies, but less clearly so in the present study. From this point of view, it is suggested that implementation of reduced impact harvesting techniques should be paired with a maximum logging intensity, so that the retention of a minimum level of forest structure and sufficient post-logging stocking densities of commercial species are guaranteed. This of course is the basic thought behind all silvicultural systems (cf. Leibundgut 1970; Mayer 1980; Kuper 1994). However, there are trade-offs associated with low logging intensities. The amount of canopy lost per extracted tree and the number of damaged remaining trees per extracted tree increases with decreasing logging intensity. In the reduced impact logging operation, the number of damaged trees per extracted tree fell from 17 at a logging intensity of 4 trees/ha, to 13 at a logging intensity of 8 trees/ha and 9 at a logging intensity of 16 trees/ha. Much of this reduction results from a reduction in skid trail area. The damage area per tree due to skidding declines as each additional harvested tree requires less and less new skid road (Table 4.10). Canopy loss per extracted tree was only marginally reduced, which is due to the effort to space trees to be felled evenly and to produce a herring-bone felling pattern, whereby overlap between gaps is minimised. In conventional logging, a reduction in canopy loss per extracted tree was observed, because gaps overlapped increasingly (Table 4.3). Trends found in this study correspond with those from other tropical logging operations (Figs. 4.21, 4.22 and 4.23). The observed trend has prompted several authors to suggest that if a certain amount of timber is to be harvested, it is better to harvest a smaller area more intensively (Jonkers 1987; Gullison & Hardner; 1993; Whitman et al. 1997). There is a snake in the grass, however, because the harvest intensity will also strongly influence regeneration and growing stock, and hence is apt to be silviculturally unsustainable. Not only will a high harvest intensity remove most of the largest and often most fecund commercial trees, which will greatly reduce the seed production potential of the residual stand, also the forest gap mosaic will change dramatically. The challenge is thus to find an optimum balance between acceptable damage per extracted tree and minimum disruption of the forest mosaic. 4.8.3 Comparison with natural gap mosaic Given the assumption that natural forest management should relate to natural gap dynamics in order to be sustainable, it is important to compare ‘logging dynamics’ to natural forest dynamics. In this study, felling gaps formed by reduced impact logging, covered 8.5%, 15.7% and 30.1% of the area, respectively, at logging intensities of 4, 8 and 16 trees/ha. In the conventionally logged area, the total gap area was 15.8% and 24.5%, respectively, at logging intensities of 8 and 16 trees/ha. Undisturbed neotropical forests usually have less than 10% of their area covered by gaps at one time: Mabura Hill, Guyana (Hammond & Brown 1992): 3.1%; Curua Una, Brazil (Rollet 1983): 4.2%; Nouragues, French Guiana (Van der Meer & Bongers 1996b): 4.3%; San Carlos, 131 Reduced impact logging in the tropical rain forest of Guyana Venezuela (Uhl & Murphy 1981): 4.8%; Pibiri, Guyana (Van Dam & Rose 1997): 5.6%; Saül, French Guiana (Mutoji-A-Kazadi 1977): 7.0%; Paracou, French Guiana (Durrieu de Madron 1994): 8.8%. Realising that natural gaps were not included in this study, it can be assumed that for the total area in a gap condition, another 3% at least should be added to the values above in case of conventional logging, while a smaller percentage should be added in reduced impact logging because trees were purposely felled into existing openings. It can thus be concluded that both conventional and reduced impact logging left the forest in a more open, fragmented state, varying with logging intensity, than before logging. Logging gaps were larger on average than the average natural gap in the neotropics. The average natural gap size in most neotropical studies lies below 200 m2: Nouragues, French Guiana (Van der Meer & Bongers 1996b): 120 m2; Costa Rica (Whitman et al. 1997): 130 m2; Paragominas, Brazil (Johns et al. 1996): 200 m2; but Curua Una, Brazil (Rollet 1983): 260 m2. In Guyana, the average size of natural gaps tends to be smaller than these values: 100 m2 (Hammond & Brown 1992); 90 m2 (Van Dam & Rose 1997). Values are rounded to the nearest 10 m2 in view of indistinct gap limits. The average gap size for reduced impact logging with an intensity of 8 trees/ha was estimated at 220 m2 and is thus about twice the size of natural gaps in Guyana. The canopy openings formed by reduced impact logging were smaller than by conventional logging - the size in conventionally logged forest at the same intensity was estimated at 320 m2 (Table 4.8). It should be mentioned that there is some variation in the methods used in the different studies which undoubtedly contributes to the differences in gap sizes as discussed at length by Van der Meer et al. (1994), Van der Meer & Bongers (1996b) and Hammond. (in prep.), but studies in Guyana used congruent methods (Hammond, pers.comm.). Still, one should bear in mind that variation in gap delimitation due to inherently indistinct gap limits (e.g. Vester 1997), hence differences in perception of the boundaries of a gap, invariably plays a role in comparisons between studies. Felling often formed multiple clustered gaps. Because there is no clear border between gap and forest, micro-climatic factors change gradually in both gap and understorey (Oldeman 1989, 1990; Canham et al. 1990; Brown 1993; Lieberman et al. 1995; Vester 1997). Hence, the micro-climatic environment in the forest remnants within such a gap cluster is rather different from that under a closed forest canopy. Integrating the clustered gaps to the chablis that they belong to and regarding the area of the entire chablis will therefore give a more realistic indication of the change in micro-climatic environment. At the same logging intensity of 8 trees/ha the average chablis area amounted to 260 m2 in the reduced impact operation, but was as much as 430 m2 in the conventional operation. Interestingly, a logging intensity of 4 trees/ha in the reduced impact operation did not lead to a decline of the average area per chablis: 250 m2. Increasing the logging intensity to 16 trees/ha raised the average area per chablis to 570 m2 and 780 m2 in the reduced impact and conventional logging operation respectively (Table 4.8). 132 Chapter 4: Logging damage The average gap size after reduced impact logging was thus at least twice the average size of natural gaps, when the logging intensity is kept at 8 trees/ha or less. The forest mosaic after conventional logging or high-intensity reduced impact logging was characterised by much larger gaps. The density of felling gaps was just below the felling intensity at 8 trees/ha and just above this value at 4 trees/ha (Table 4.3), which is not far from the density of natural tree fall gaps. Hammond & Brown (1992) found 3.2 gaps/ha in 15 ha of untouched forest 25 km north of the experimental area and Van Dam & Rose (1997) found 6.3 gaps/ha in 89 ha of untouched forest in an area directly bordering the experiment. Again, it should be realised that felling gaps were formed in addition to - sometimes partially overlapping - pre-existing natural gaps. Although natural gaps are small on average, large gaps (over 500 m2) do occur in undisturbed forest (Rollet 1983; Uhl et al. 1988; Van der Meer 1995), but are infrequent (Jans et al. 1993; Denslow & Hartshorn 1994). Both studies in Guyana found that only 2% of the gaps were over 500 m2, the largest encountered gaps measuring 560 m2 and 940 m2 respectively. The majority of the gaps (91%) measured less than 250 m2 in both cases. In this study, the proportion of total gap area consisting of large gaps was much larger and increased substantially with increasing logging intensity; at a logging intensity of 4 trees/ha, 1 in 25 gaps was a large gap, at a logging intensity of 8 trees/ha, 1 in 10 gaps, and at a logging intensity of 16 trees/ha 1 in 4. Large gaps occurred with similar frequently in the reduced impact as in the conventional operation. The main difference in gap mosaic between the two logging methods was in the occurrence of medium-sized gaps (between 250 and 500 m2) which were less frequent in the reduced impact operation; 73% to 63% of felling gaps were smaller than 250 m2 in the reduced impact operation and 64% to 42% in the conventional operation, with logging intensity varying from 8 to 16 trees/ha. The gap mosaic of the forest thus changes with low- and medium-intensity, reduced impact logging, but far less than with medium-intensity, conventional logging. Conventional logging had a similar area in gap state as reduced impact logging at a logging intensity of 8 trees/ha, but this was concentrated in larger gaps. If the logging intensity is raised to 16 trees/ha the forest mosaic changes dramatically in both cases. An increase in large gap frequency affects future species composition. There is a large body of evidence that large gaps usually become dominated by soft-wooded secondary species for longer time periods than small gaps because of micro-climatic conditions and potentially decreased seed dispersal into gaps (Kramer 1933; Richards 1952; Schulz 1960; Whitmore 1975; Hartshorn 1978; Brokaw 1985b; Rossignol et al. 1998, table 5.1). In addition, establishment of lianas is promoted by large gap size (Putz 1984; Uhl & Vieira 1989). Vigorous liana growth may stifle regeneration of trees and, hence, in stead of returning to its original tall forest state, the large gaps may develop into low, impenetrable liana tangles forming a ‘liana blanket’ (cf. Kahn 1982; Oldeman 1990). Results of a study by Ek et al. (in prep.) confirm this hypothesis. Consequently, logging gaps with potentially low commercial value can comprise a substantial part of forest that was subjected to conventional logging or high-intensity reduced impact logging. The situation is strongly exacerbated by soil disturbance inside felling gaps (Ek & 133 Reduced impact logging in the tropical rain forest of Guyana Van der Hout, in prep.). Dormant seeds of pioneer species benefit strongly from soil disturbance (Putz 1983; Riéra 1985; De Foresta & Prévost 1986), while most shade-tolerant seedlings have been destroyed by the skidder. Shade-tolerant seedlings that survived will soon be overgrown by the fast-growing pioneers in large gaps (Zagt 1997). Designing a silvicultural system of mosaic management with prevention of the formation of large logging gaps and preventing soil disturbance in gaps thus should be a primary focus of sound forest management. The question in what case the first condition is fulfilled is hard to answer; because reported threshold gap sizes for pioneers vary greatly. Moreover, results from experimental tree-fall gaps have demonstrated that gap size sometimes has less influence than expected on floristic composition. In those studies, the most important change appears to be an increase in height growth and decline in mortality rate of pre-existing seedlings and saplings rather than emergence of seedlings from the seed bank (Popma et al. 1988; Uhl et al. 1988; Brown & Whitmore 1992; Kennedy & Swaine 1992; Whitmore & Brown 1996; Van der Meer et al. 1998). Based on nutrient leaching rates, Brouwer (1996) suggested a threshold gap size of 500 m2. He also noted that the strongest pulse of solute leaching took place in those parts of a gap where soil disturbance had occurred. Ek & Van der Hout (in prep.) studied species composition in logging gaps of several ages, whereby gaps were divided into small or large gaps, with 300 m2 as the dividing line. The species composition in logging gaps smaller than 300 m2 appeared to be quite similar to that of closed forest, whereas larger gaps differed in composition. The question remains, however, whether results would have differed if a different size limit for the distinction between large and small gaps were chosen (cf. Rossignol et al. 1998). It can thus not be said with certainty whether the threshold size should be lowered to 300 m2. This is quite unfortunate, because the main distinction between conventional and reduced impact logging appears to be in the frequency of medium sized gaps (between 250 and 500 m2). Until further information on the ‘allowable’ gap size in the forests studied here becomes available from ongoing research (Van Dam et al. 1999), it is wise to use a conservative threshold size. 4.8.4 Implications for management and monitoring In this thesis the effects of conventional and reduced impact selective logging operations at variable logging intensity in Guyana is described. Plots were placed in areas representative for Chlorocardium rodiei dominated mixed forest on brown sand. Descriptions of this forest type can be found in Davis & Richards (1933, 1934), Fanshawe (1952) and Ek & Van der Hout (in prep.). Chlorocardium rodiei (Greenheart) occurs accompanied by different co-dominant species, which prompted Fanshawe (1952) to distinguish several ‘faciations’ each with its own particular dominants. Those forest types all have in common that clusters of mature Greenheart trees occur. Although such clustering occurs in rain forests elsewhere (Richards 1952; Connell & Lowman 1989; Hart et al. 1989), the fact that logging has been leaning 134 Chapter 4: Logging damage heavily on this species, makes that the concept of selective logging takes on an exceptional meaning in these forests. There are also areas within the mixed forest where Greenheart is sparse. These areas were purposely avoided and our results should be viewed as the effect of logging on timber-producing areas rich in Greenheart. Five general conclusions stem from this research. First, the level of felling damage incurred during reduced impact logging did not appear to deviate substantially from conventional logging operations in Greenheart forest when measured in terms of relative canopy loss or residual stand damage. Yet, it is clear that skidding damage was mitigated significantly; i.e. largely sparing the propagule bank and minimising the area of compacted soil, by introducing elements such as proper planning of the logging operation, directional felling and winching. The reason why - in contrast to other comparative studies - no significant reductions in damage were found is explained by the fact that there are other differences between the two logging systems studied besides the implementation of special techniques and tools: i.e. differences in felling pattern and in the spatial distribution of the felled trees. Compared with the piling up of boles in the conventional operation, the consistent application of a herring-bone system and spreading of the target trees appeared to have drawbacks, especially at higher logging intensities. In the former case tree falls frequently overlapped, while in the latter case almost every felled tree formed its own gap. The herring-bone pattern was inspired by the objective to reduce skidding damage and improve skidding efficiency. The direction in which a tree could be directed was kept within a narrow range. Field observations during skidding gave the impression that more flexibility can be betrayed in this respect (see Section 5.8). Since it was shown that the size of a dual tree fall chablis was below 400 m2 on average - which could be regarded as an ‘allowable’ gap size - it is recommended to apply the herring-bone system less strictly and aim for overlapping tree falls when applicable. On similar grounds, the restrictions on felling neighbouring trees can be alleviated. Discouragement of the felling of neighbouring trees resulted in the selection of less-valued species in this study, which will make it difficult to reconcile the reduced impact logging system with the interests of the logging industry. There are also strong indications that liana cutting was unsuccessful due to a non-optimal short period between its application and the moment of logging (six months). Further research in which lianas are cut one year in advance is recommended to find out if logging damage can be reduced more substantially. The clustered, criss-cross felling pattern observed in the conventional Greenheart logging operation might have had its advantages in respect to the rate of canopy removal. The felling gaps produced here are disadvantageous to sustainable timber production. Even at a moderate logging intensity of 8 trees/ha, the average gap size was 50% larger than with reduced impact logging and exceeded the size recommended by Ek & Van der Hout (in prep.). Furthermore, given similar gap size, the micro-climatic environment is likely to be more adverse in conventional felling gaps because of their 135 Reduced impact logging in the tropical rain forest of Guyana more regular shape and the lower density of residual trees and the disruption of the propagule bank due to skidder movements within the gap area. It is therefore recommended to discontinue the current logging practice. Secondly, the level of damage per extracted tree decreases when the logging intensity is increased. This corresponds to findings elsewhere. However, there are limits to the allowable change in the forest mosaic architecture. The likelihood that the goal of sustainability can be achieved will do well by a conservative removal of trees. Besides gap replacement by residual juvenile commercial trees in felling gaps, the arrival of propagules of commercial species to the gap, is another prerequisite to ensure future timber yields. Since a large proportion of the commercial species in these forests are large-seeded, depending on gravity or mammals for dispersal of their seeds (Hammond & Brown 1995) - Zagt (1997) found that very few Greenheart seeds were dispersed further than 14 m from the parent tree - , regeneration will benefit from the presence of a reproductive mature tree adjacent to the gap (cf. Webb 1998). This also poses further restrictions on the allowable gap size and the spatial distribution of seed trees. The average felling gap size remained below the threshold gap size in the reduced impact operation at a logging intensity of 8 trees/ha and was not further decreased by reducing the logging intensity further. On the other hand, the area affected per extracted tree by skidding was reduced by 50% by doubling the logging intensity. Balancing these contrasting arguments, leads to the conclusion that the logging intensity should not exceed 10 tree/ha when applying reduced impact logging, but neither should it be lowered too much below this level. Thirdly, single tree falls formed a total canopy opening (= ‘chablis’-area) of 180 m2 on average, dual tree falls 360 m2 and composite gaps formed by three tree falls 640 m2. The distribution of the target trees and felling directions should thus be such that no more than two neighbouring trees are to be felled into one gap. It is important to focus on the mode of gap formation (i.e. single vs. multiple tree fall) instead of on the desirable precise gap area, because this is a more realistic and tangible objective for field workers when conceiving logging guidelines (Webb 1998). Personal experience during this study and during the establishment of an ongoing Tropenbos gap size experiment, showed that it is extremely difficult to obtain a gap of a particular size (see also Van Dam & Rose 1997; Mekkink & Nijmeijer 1998). It is difficult to assess the projected area of the gap formed by the removal of the crown of the target tree itself beforehand and even more difficult to predict whether or not the target tree will bring down any surrounding canopy trees, which in turn may bring down their neighbours (‘domino-effect’). For the same reason, it appears de facto to be impossible to preclude the formation of large (over 500 m2) canopy openings entirely. In such gaps, natural regeneration of timber species could be assisted by silvicultural means. Gap enrichment could be considered a silvicultural option (Webb 1998) or liberation of seedlings (Schulz 1960; Zagt 1997). However, it has been long shown that seedling or sapling liberation treatments need to be repeated frequently and thereby are cost-prohibitive (cf. Prince 1973; Boerboom 1964; De Graaf 1986). 136 Chapter 4: Logging damage Fourthly, controlled skidding and the application of winching appeared to be successful in reducing the area affected by skidding. Moreover, skidder movements around the stump were almost completely avoided. This is of utmost importance because, not only is the soil structure adversely affected (Hendrison 1990), which changes the nutrient, hydrological and biological characteristics of the soil (Malmer & Grip 1990; Brouwer 1996; Smits 1994; Yasman 1995), and hence the composition of the regeneration (Hawthorne 1993; Guariguata & Dupuy 1997), also damage to residual seedlings and saplings is mitigated. Because seedlings and saplings of many timber species are mainly contracted around the parent tree (stump), an important source of future timber production is salvaged compared to conventional practice. Fifthly, the use of Brokaw’s gap definition (1982) - a ‘hole’ in the forest canopy extending through all levels down to an average height of 2 m above ground - or for the same matter Van der Meer & Bongers (1996a) modified definition in which the gap size is measured at 20 m height, is an unsatisfactory parameter of the changes taking place in forest architecture due to logging. As indicated by Orians (1982), Popma et al. (1988) and Brown (1993) the ecologically altered area at the forest floor is often larger than the size of the gap at the forest canopy level. Van der Meer et al. (1994) concluded that Runkle’s expanded gap size (1981, 1992) gives a better estimate of the area on the forest floor where irradiation levels have changed. Still, both definitions neglect the fact that important changes take place between the forest floor and the top of the forest canopy. When a tree falls - by felling or by natural causes - its crown is removed from the canopy. Crowns of surrounding canopy trees will be hit first, but as the momentum is still small, most of these are not seriously damaged, although, occasionally, a large tree is uprooted or snaps. During its fall, the tree accelerates and smaller trees are hit harder (Jonkers 1987; the present study). The extent of damage among subcanopy trees is thus much larger than among their fellows in the canopy. Oldeman (1974a, 1978) used the medieval French word ‘chablis’ to denote the impact of a tree fall on the forest; i.e. including the fallen tree itself and all of the resulting destruction. The projected area of a chablis is much larger than the associated canopy opening. Oldeman (1974a) estimated the average size of a chablis in French Guiana at 600 m2, whereas Van der Meer et al. (1994) found an average gap size of 54 m2 at Les Nouragues, French Guiana after strict application of Brokaw’s definition and of 338 m2 according to Runkle’s (1981) expanded gap size definition. Since the projected area of a chablis reflects affected vegetation at any height, situated under or above intact vegetation, it is also of limited use in estimating the ecological changes that have taken place (Van der Meer et al. 1994). Field observations during logging in the present study showed that less pre-existing, subcanopy vegetation escaped damage in the conventional operation than in the reduced impact operation. However, this is not reflected in the canopy removal rates estimated using Van der Meer & Bongers’ method (1996a). Comparisons based on damage to the 137 Reduced impact logging in the tropical rain forest of Guyana remaining trees do not offer an alternative, because skidding damage varied considerably between the operations. Another major problem that was encountered was the fact that the local forest structure influences the size of a canopy opening when using a two-dimensional projection. As mentioned above, if a tree would be directed into a natural gap or a building (sensu Whitmore 1975) or late aggradation phase (Bormann & Likens 1979), the gap size would incorporate the pre-existing natural gap. From a timber grower’s point of view, aiming a tree in such a patch would be recommendable, because less basal area - stored in easy to replace small trees - will be affected. Using any canopy opening projection method a more open forest would suffer more canopy loss than a closed forest, using a chablis projection method (Riéra 1982) the canopy removal rates would not differ assuming that equal numbers of small trees are affected in a natural gap as in the understorey of the closed forest. To overcome the limitations of two-dimensional projections another approach is needed. The three-dimensional forest model SILVI-STAR (Koop 1989; Koop & Sterck 1994) attempts to bridge this gap. It models light penetration by calculation of interception of direct and diffuse light falling through a three-dimensional tree model. This model offers a good alternative, especially because it takes the spatial positions of individual tree crowns and unoccupied spaces into account, but is very time consuming (Koop & Sterck 1994). There are also simpler alternatives that are quicker to apply as described in (Lieberman et al. 1989; Crome et al. 1992 and Van der Meer 1997). Because these methods are being used infrequently and for reasons of simplicity, one might still opt for a two-dimensional projection. In that case, it is recommended to apply Runkle’s because 1) it is the only definition that allows a gap to be determined precisely, while with the other definitions it is hard to determine the exact vertical projection of the determined gap edge; 2) the objectivity in determining where the edge of the gap lies is much higher; 3) it gives a better estimate of the area of the forest floor with increased light availability and vegetation damage. By applying Runkle’s definition a weight factor is attached to the vertical projection of the canopy opening, which signifies that the affected area increases disproportionately with an increase in vertically projected gap size. 138 Chapter 5: Logging efficiency 5. LOGGING EFFICIENCY 5.1 INTRODUCTION One of the major objectives of the present study was to assess whether improving harvesting methods can reduce logging damage. In Chapter 3, methods to reduce felling and skidding damage are discussed. In Chapter 4, it was concluded that the current utilisation practice of Guyana’s Chlorocardium rodiei forest is not sustainable and that changes are required in the logging method and local logging intensity. Apart from the importance of silvicultural and ecological aspects, it can be assumed that the country-wide establishment of a sustainable forest management system ultimately depends on its economic merits. An economic comparison of reduced impact logging with conventional logging thus is of great importance for the acceptance of the suggested harvest regime. The harvesting of timber is usually subdivided into three main phases: cutting, off-road transport and long-distance transport. Within each of these phases there are a number of sub-operations such as: shaping of the felled tree into logs (‘bucking’); unloading and loading at a roadside assembly point (the ‘landing’ or ‘market’); and final unloading at the delivery point - being either a sawmill or a point from where the logs are despatched for further transport for instance to an export port. Any logging operation is based on some survey of the wealth of the forest. This may vary from a ‘tree spotter’, who cruises the forest in an unstructured manner to identify areas worthwhile of logging, to professional pre-harvest planning, including block layout, systematic pre-harvest inventory and mapping of trees to be harvested. Several studies have shown that additional cost incurred during planning and directional felling can be offset by an increased efficiency of the actual extraction operation - off-road transport - (e.g. Mattson Mårn & Jonkers 1981; Hendrison 1990; Barreto et al. 1998; Holmes et al. 1999). Assessment of the efficiency of a logging operation often concerns the first two logging phases only and occasionally the pre-harvest planning phase. The aggregate logging cost, however, also includes road transport, landing operations and unloading at the delivery point and efficiency assessment actually should include the cost of these phases as well. When assessing the economic merits of a ‘reduced (ecological) impact’ logging system, its efficiency must be compared with another logging method, which is obviously the current, local practice. However, the current practice may cover a wide range of methods, varying from ‘hit-and-miss’ logging to ‘standard practice’ logging (Van der Hout & Van Leersum 1998, see also Chapter 3). Studies by Flaming (1995) and Landell-Mills (1997) also showed a wide variation in this respect for Guyana. The place a current practice may take on this scale depends largely on the scale and level of capitalisation of the operation. These two factors weigh heavily on the cost incurred and the output achieved in each logging phase. Hendrison (1990) showed that the difference in performance and cost between ‘conventional’ and ‘damage-controlled logging’ in 139 Reduced impact logging in the tropical rain forest of Guyana Suriname could be partly attributed to differences in such organisational aspects and standards of equipment. It is important, however, to evaluate the financial consequences of introducing reduced impact logging on its own, technological merits. This implies that obscuring factors such as differences in terrain, stand characteristics, quality of equipment and organisation should be ruled out as much as possible. In this chapter, an attempt is made to compute the financial results of reduced impact logging as compared to those of conventional logging in Chlorocardium rodiei forest Guyana. The effect of logging intensity on logging cost is also calculated. The focus of the analysis lies on the cost of logging, which is often used as an argument against a change of traditional logging methods by logging companies. This will be accomplished by conducting a comparative financial analysis of a reduced impact operation relative to a conventional operation in the Demerara Timbers Ltd. (DTL) concession in Central Guyana. Although operation characteristics are variable across sites and across firms in Guyana, the Demerara Timbers operation is considered representative for several operations in Guyana. The present study does not deal with the effects on the respective profits for the timber industry. 5.2 METHODS Logging costs and performance are affected by many different factors. The quantity and quality of available labour, the type and size of the trees to be harvested, topography, carrying capacity of soils, accessibility of the area to be logged and the extraction (skidding) distances all influence the cost and performance of a particular logging system (Sundberg & Silversides 1988). In case of selective logging in tropical rain forest with its variety in species of trees to be harvested and a relatively high variety in terrain characteristics and extraction distances due to the small volume harvested per hectare, the influence of these factors is even more important. One can try to filter out such noise when comparing different systems, but, obviously, some aspects will always remain inseparable. Within the scope of the present study, it was only possible to study two logging phases in detail: felling and off-road transport. It should be noted that the study of the efficiency of reduced impact logging was carried out within the experimental lay-out as described in Chapter 1. This posed some limitations on the lay-out of the extraction network; i.e. skid trail planning was co-determined by the configuration of the experimental plots. Based on these considerations it was decided to construct models to estimate inputs and outputs. These models are based on detailed time and work studies as far as felling and skidding are concerned, and on standard rules of thumb (Caterpillar 1986; FAO 1977, 1978) for the remaining phases of the logging process, using data provided by the timber company. By costing each activity on a m3 - basis, the influence of differences in site quality on the cost-benefit analysis can be further reduced. Similarly, 140 Chapter 5: Logging efficiency modelling skidding costs for a certain fixed distance and load size can eliminate the effect of differences in extraction distance and load size. This chapter is composed as follows. First, the output in m3 per hour of felling and skidding is determined for both logging systems. Variation in the factors mentioned above are ‘filtered out’ by using regression models. Secondly, the cost of felling and skidding ($/m3) is determined by determining the hourly cost ($/hr) of each activity and combining this with its output (m3/hr). The cost and output of the subsequent phases - landing operations and long-distance transport - are determined by using company data and standard estimation procedures (Caterpillar 1986; FAO 1977, 1978). Thirdly, the effect of logging intensity is determined by using the time study data of the reduced impact operation for felling and skidding and standard estimation procedures for the subsequent activities. There are a number of textbooks available describing various methods of determining the cost of operating equipment (Samset 1972; Caterpillar 1986; FAO 1977, 1978; Sundberg & Silversides 1988, Silversides & Sundberg 1989). These methods differ from one another to various degrees and none is perfect. Some differences in the manner of treating depreciation, operator wages, fringe benefits, etc. are due to management policies or accounting practices. Other differences are due to the time unit over which the cost is distributed: there are several kinds of ‘hours’ and care must be taken to specify whether the term refers to shift hours, engine meter hours, operative machine or effective crew hours. The methodology used in the present study is mainly based on FAO (1978), Sundberg & Silversides (1988) and Silversides & Sundberg (1989). Detailed time data are available for the felling and skidding phases only. Cost and output of these phases are based on the average ‘effective crew time’. For the subsequent phases, simplified rules of thumb for estimating costs and output are adopted. 5.2.1 Cost elements Restricting the assessment of the operations until unloading at the delivery point, the costing comprises: ƒ Stump operation ƒ Off-road transport ƒ Landing operation ƒ Loading ƒ Road transport To which are added the cost of: ƒ Transport routes ƒ Overheads All operations require a certain input of manpower plus a certain input in machine power, which are both quantified in time units; i.e. m3/hr. The cost of each activity is 141 Reduced impact logging in the tropical rain forest of Guyana expressed as the cost per volume unit and calculated through a costing of the two time inputs and relating the combined time costs to the volume output in that time. The cost element related to transport routes comprises a network of truck roads of different standards. The construction cost of these transport routes is derived by calculating the cost per km as related to the assessed road lengths to be distributed over the total volume yield. Road maintenance costs usually apply only to the main road on which the trucks are travelling year after year. For the sake of simplicity, it was assumed that the feeder road density is 1 km per 1 km2 of logging area. Furthermore, it was assumed that 1 km of main road services 5 km2 of logging area. For a logging block of 1 km2, construction cost of 1 km of feeder road and 200 m of main road is taken into consideration, whereas maintenance cost covers the maintenance of the 200 m of main road. The cost of overheads include workshop expenses, office costs, management costs, housing and utility costs. Royalties and concession fees that are paid per volume and respectively per area are included as well. 5.2.2 Time elements The cost assessment is fundamentally based on an assessment of two time inputs, i.e. labour- and machine-time. Labour time Forest workers are paid for a number of fixed daily hours, or are paid based on yield. In the former case the paid working hours include time for transport to and from the work site (‘commuting time’) as well as time for a meal-break. These time elements are referred to as ‘down time’. Down time is deducted from the total time the worker is on duty (‘camp-to-camp time’), the outcome being ‘effective man-time’. Effective man-time is defined as the time the labourer is actively engaged in the work to which he is assigned. Work delays are in principle taken into account as reducing mean performance rather than quantifying their adverse effect in time units. In the present study, a distinction was made between organisational, unavoidable and avoidable delays. Organisational delays are caused by any incident which prevented the crew from utilising their time on operational work, for instance waiting for logs to be ready for extraction or waiting for scaling of logs at the landing. Unavoidable delays are caused for instance by waiting for heavy rain to stop or waiting for hazardous conditions to be cleared (wasp nests, snakes). Because these delays are either dependent on organisational quality of operations or truly unavoidable, they were assessed separately. Avoidable delays (work delays), such as time spent on clearing hang-ups, time to free the saw when stuck in the cut or to reset a choker to free a log that is stuck behind a stump or other obstacle, are related to the logging method itself and accounted for as reduced efficiency. Occasional rests due to work fatigue are regarded in the same way. Machine time The time a machine is available for work is referred to as the ‘in-shift time’ and equals the effective crew time. The ‘effective machine time’ (or ‘operative machine time’) is 142 Chapter 5: Logging efficiency the time the machine is actively working. In case of a skidder, it includes light and loaded travel as well as the time for collecting or discharging a load. In case of felling it includes only the actual sawing time (felling and bucking) and is only a small proportion of the effective crew time (see below). The time for routine servicing, refuelling, sharpening of chains, etc. is excluded but work delays are taken into account in the same way as lost crew time. Because the operative machine time equals the effective crew time minus the machine down time, it can be quantified as a certain percentage of the effective crew time. The following simple cost formula applies: cost per m 3 = where: tc tm cc cm v tc ⋅ cc + t m ⋅ cm v = the effective crew time, measured in effective crew hours, which are required to process the volume v, measured in m3. = the operative machine time, measured in operative machine hours, which are required to process the volume v, measured in m3. = the crew cost per effective crew hour = the machine cost per operative machine hour = the volume in m3 processed in the operational time which contains the two time elements: tc and tm 5.2.3 Labour costing The basic wage for an individual labourer is a fixed rate per day in the reduced impact logging system. In the conventional operation, a piece-rate payment system was in place. The rate paid varied between US$ 1.70 and US$ 2.00 per m3, which is divided among the entire logging crew. In addition to basic wages, other direct payments apply such as paid leave, food allowances, social service contributions and other financial incentives. Other services rendered such as free housing, free transport, medical care are included in the total cost of labour. Information on the number of days worked per year was derived from Landell-Mills (1997).The assessment of the essential time cost, the crew cost per effective crew hour is based on: ƒ the composition of each crew category; ƒ total cost per crew member per year; ƒ total number of crew hours per year. 5.2.4 Machine costing Machine costs are calculated according to standard costing methods (Caterpillar 1986; Sundberg & Silversides 1988). For all machines except trucks the cost unit is the cost per operative machine hour and comprises both fixed and variable costs. For trucks a different approach is warranted. Trucks are often idle during prolonged periods while being loaded and unloaded. In case of trucks, a distinction is made between ‘round-trip time’ and ‘standing time’. The cost per work hour will then be the sum of fixed and variable costs and for idle hours, the fixed hourly cost only. Items like saw chains of a power saw or tyres of trucks and skidders are customarily subject to separate cost estimates and not included in the aggregate unspecified repair and maintenance cost. 143 Reduced impact logging in the tropical rain forest of Guyana Fixed costs 1. Depreciation Depreciation is the diminution of value of equipment through use. To calculate depreciation, the acquisition cost of the equipment and the machine’s life expectancy is needed. Although, in Guyana, second hand equipment is frequently used, the present study uses acquisition cost and life expectancy of new machines, because of the complications involved in calculating depreciation for second hand machines (Landell-Mills 1997). Information from the timber company is used when available. Otherwise, the Caterpillar performance handbook (1986) is used. depreciation = delivered cost × 0.9 machine life in hours 2. Interest Interest charged on a machine captures the cost of borrowing or the opportunity cost of tying up funds in capital. The interest charge is based on the average annual investment in capital. It is expressed as a cost per unit of time by dividing the annual interest by the number of time units in the year, thus: interest per hour = where: Iaa i hrom I aa ⋅ i hrom = average annual investment = rate of interest = operative machine hours per year The average annual investment can be calculated as follows (disregarding salvage value of machine): I aa = where: c × ( yr + 1) 2 ⋅ yr c yr = = delivered cost of machine estimated useful life of machine in years The interest rate is based on an average of the yield on a one-year Treasury Bill and the prime Lending Rate for 1996: 16% (see Landell-Mills 1997). 3. Insurance Insurance is normally designed to cover public liability and property damage and loss of machine because of fire, theft or other hazard. The annual insurance is taken as a percentage of the average annual investment and converted to a rate per machine hour. The present study uses a percentage rate of 3%. The following formula applies: 144 Chapter 5: Logging efficiency insurance cost per hour = where: Iaa hrom I aa × 0.03 hrom = average annual investment = operative machine hours per year Variable costs 4. Fuel The fuel consumption rate of a piece of equipment depends on the engine size, the condition of the equipment, the operator’s handling habit, environmental conditions, the basic design of the equipment and the work load factor. An engine continuously producing full rated engine power is operating at a load factor of 1.0. A machine seldom operates at this level for extended periods. Periods spent at idle, travel in reverse, skidders travelling without load, close manoeuvring at part throttle and operating downhill are examples of conditions that reduce load factor. The work load factor appears to be rather constant for a particular operation (Sundberg & Silversides 1988). For practical purposes, the often-assumed value of 60% is used here. The engine power is stated by the manufacturer as the maximum power rating at the highest RPM for which the engine is designed. For practical purposes, a fuel consumption rate of a diesel engine of 0.18 kg/kWh is used. Gasoline engines have a fuel consumption rate in the order of 0.25 kg/kWh (Sundberg & Silversides 1988). fuel cost per hrom = gkW ⋅ X ⋅ c f where: hrom gkW X cf = = = = operative machine hour gross engine power Load factor × fuel consumption per kWh fuel cost per litre Alternatively, standard hourly fuel consumption tables supplied by the manufacturer can be applied (e.g. Caterpillar 1986). 5. Lubricants Lubricants include engine oil, transmission oil, final drive oil, hydraulic oil, grease and filters. The consumption rate varies with the type of equipment, environmental conditions (temperature), the design of the equipment, and the level of maintenance. The cost of oil and grease, including hydraulic oils, hence varies with engine power and the capacity of the hydraulic system. Lacking actual data, lubricant cost per hour can be estimated (cf. Sundberg & Silversides 1988): cost of oil and grease per hrom = 0.20 ⋅ gkW ⋅ co 100 145 Reduced impact logging in the tropical rain forest of Guyana where: hrom = operative machine hour gkW = gross engine power = cost of engine oil per litre co 6. Servicing and repair These activities include everything from simple maintenance to the periodic overhaul of engine, transmission, clutch, brakes and other major equipment components. These costs are influenced by the severity of working conditions, the operator’s use or abuse of equipment and maintenance policy. Servicing and repair cost of a machine increase with age. Still, it is normally averaged over machine life and expresses on a ‘straight line’ basis in the same manner as depreciation. Although this will overstate costs earlier in the machine’s life and understate costs towards the end of its life, this method is used here because the degree of precision of the present study does not need the precision of a power function for servicing and repairs. As a rule of thumb lifetime servicing and repair cost can be estimated as a percentage the delivered cost of the original machine. The hourly cost is derived by dividing by the life in operative machine hours, and thus will be estimated as a percentage of hourly depreciation cost. Miyata (1980) quotes percentages that vary according to the type of equipment. Other authors (FAO 1977, 1978; Sundberg & Silversides 1988) set the hourly repair and maintenance costs at a rate equal to the hourly depreciation cost. The latter approach is followed here because the present study does not need the degree of precision indicated by Miyata (1980). 7. Tyres Tyres wear off according to the mileage travelled and it has therefore become customary to assess this cost separately. As a rule of thumb, tyre costing is based on travelling hours. Tyre life for a vehicle is estimated as one quarter of the machine life (FAO 1978). This implies that three replacement sets will be needed during the lifetime of the vehicle. Recapping and repairs per tyre amount to 50% of the original cost (Sundberg & Silversides 1988), implying that the total cost for each set of tyres is 150% of the purchase price. The tyre cost per travelling hour is found as follows: ct = where: ct pt lm pt × 1.5 × 3 lm = tyre cost per operative machine hour = price of set of replacement tyres = Machine life in hours 5.2.5 Work study Work study is a systematic analysis for measuring and recording procedures, inputs and outputs in a work process. The field of work study comprises two types of studies: method study and time study. The field of method study concerns mapping or recording of the various elements in a work process. It is essentially concerned with improving organisation, working techniques, working conditions, etc. Time study 146 Chapter 5: Logging efficiency measures the time it takes to performing a particular job. It is usually preceded by a method study. Systematic work study was introduced in the beginning of this century by Taylor (1911 ex Apel 1993) and related to the era of industrialisation. The first systematic work studies in forestry followed soon after that (Apel 1993). An overview of the numerous aspects, special applications and terminology can be found in Brils (1995) and Wittering (1973). The complex environment in forestry poses constraints on performing time studies and this is even more so in case of selective logging in tropical rain forest (Brils 1995). The rain forest is a diverse working environment and the trees to be logged vary in appearance (buttresses) and physical properties (high- versus low-density wood, presence of silica, etc.). Workers and machines are under strain of heat and humid air. This comes on top of the fact that the energetic demands of many forestry jobs already are among the highest of all professions (between 5000 and 6000 kcal/day). The variation in stand characteristics, terrain, distances between trees implies that it is impossible to assume ceteris paribus conditions as the feller or skidder operator meets a steadily changing and challenging environment, demanding a work approach that is never entirely the same (Samset 1990 ex Brils 1995; Brils 1995). Both felling and skidding comprise a chain of activities. Some of these activities follow a regular sequence repeatedly applied to every work object - in felling: a tree, in skidding: a load. The sequence of elements that are required to complete a work object is referred to as the work cycle. Other work elements occur occasionally, such as delays, refuelling or sharpening of the chain of the chainsaw, or only once a day, such as a meal-break and the round-trip to the work site. In felling the following repetitive work elements form an elementary cycle: walking, felling preparations, cutting and bucking. Some work elements occur regularly as well, but span more than one elementary cycle: e.g. sharpening and refuelling. Other work elements that occur irregularly are work delays such as freeing of hang-ups or stuck saw blades, unavoidable delays, such as rain or snakes. In skidding the cyclic elements are return from landing to site, collect load, bunch load, skid load to landing, discharge. The load collection element can contain repeated elements. If a load consists of more than one log, the following elements are repeated for each log in case of reduced impact logging: move to next log, turn, pull out winch lime, choke, winch, pre-skid. In case of conventional logging, this would be in principle: move to next log, position log with blade, turn, hook and pre-skid. However, not every element is always repeated in conventional logging. It can happen that two logs are positioned at one time, or that two logs are pre-skidded together. For this reason, the comparison will have to be restricted to comparing the collection of load instead of collecting individual logs. Work elements that span more than one elementary cycle are refuelling and servicing. Irregular work delays are due to mechanical malfunctioning, waiting for felling, assisting the felling crew in freeing a lodged tree, etc. 147 Reduced impact logging in the tropical rain forest of Guyana 5.2.1 Felling model As mentioned before no ceteris paribus conditions exist when comparing two logging methods. To overcome this limitation the data were analysed by multiple regression with logging method as a dummy variable to develop extrapolating prediction equations for the different work elements. This approach also allows statistical testing of any difference between the logging methods. Because logging methods influenced certain time elements more than others did, regression analyses were conducted for the separate time elements instead of the aggregate work cycle. Time elements related to diameter Felling preparations, cutting, bucking and work delays are strongly related to the diameter (dbh) of the tree to be felled. Purpleheart trees were not included in determining the model because extremely deviant values were recorded for this species, which is due to its typically very large buttresses (see Brils 1995). Cutting and bucking are intrinsically related to the surface that has to be cut, i.e. to the basal area rather than the diameter of a tree. Therefore, polynomial regression models of the following form were fitted for these time elements: t = a1 + a 2 ⋅ [M] + b1 ⋅ d bh2 + b2 ⋅ [M] ⋅ d bh2 + ε For felling preparations and work delays the following linear version is used as a regression model: t = a1 + a 2 ⋅ [M ] + b1 ⋅ d bh + b2 ⋅ [M ]⋅ d bh + ε where: t dbh [M] ai and bi = time consumption (min) = diameter at breast height (m) = dummy variable; equalling 1 in case of reduced impact logging and equalling 0 in case of conventional logging = regression coefficients The predictor variables were entered one by one into the model, starting with dbh or (dbh)2, subsequently [M] and finally [M]⋅dbh or [M]⋅(dbh)2. In each step it was tested whether or not the respective coefficient (respectively b1, a2 and b2) was significantly different from zero. The resulting regression equations were applied to all harvested trees producing a distribution of predicted time element values for each logging method. Averages were calculated for each time element for each logging method, which were then put in the aggregate model. Sawing time related time elements Certain time elements are related to other time elements. This is the case for sharpening the chain of the saw and the frequency of refuelling. The activities take place once in several work cycles and can be expressed as a proportion of the total sawing time (felling and bucking combined) per day. These two time elements were compared using the following linear regression model: 148 t = a1 + a 2 ⋅ [M ] + b1 ⋅ s t + b2 ⋅ [M ]⋅ s t + ε where: t st [M] ai and bi Chapter 5: Logging efficiency = time consumption (min) = sawing time per day (min) = dummy variable; equalling 1 in case of reduced impact logging and equalling 0 in case of conventional logging = regression coefficients The total time spent on refuelling can be estimated by using the estimate for the frequency of refuelling multiplied by the average time that it takes to refuel the saw (2-stroke gasoline and chain oil). As mentioned below the methods used different chainsaws, each with a different tank volume. The difference in the average time of filling the tank was tested with Student’s t-test. The total time spent on servicing the machine (excluding minor repairs) was then estimated as follows: (servicing time) = (sharpening time) + (refuelling frequency) ⋅ (average refuelling time) Time elements unrelated to diameter or sawing time The following time elements were not related to the diameter of the tree or to other time elements. The time spent moving from one tree to the next (‘walking time’), spent on minor repairs and on freeing lodged trees. Difference between the logging methods respective to walking and repair times were tested with the non-parametric Mann-Whitney U-test. Trees that lodged in a neighbouring tree after being severed at the base (‘hang-ups’) were treated differently. The reason for this is that in the conventional operation the skidder was present at the time of felling, whereas this was not the case in the reduced impact operation. Consequently, the skidder pulled down hang-ups in the conventional operation. In the reduced impact operation, much time was spent on trying to force the tree down by wedging. Neither the close presence of the skidder in the conventional felling operation nor its absence in the reduced impact felling operation represent a ‘normal’ situation. Furthermore, this difference is judged an organisational aspect and it was decided to eliminate it. The time to bring down a hang-up was therefore by conjecture set at a time of 5 minutes for both operations. This time element is then estimated by multiplying the average number of hang-ups per day with this number. Crew down time and unavoidable delays The crew down time - time for commuting and meal-break - differed between the two operations. The difference in commuting time was related to a difference in proximity of the blocks to the camp and the closer proximity to passable roads of the reduced impact logging plots. Unavoidable delays also differed between the two operations. This was mainly related to the fact that logs were measured at the stump in the reduced impact logging operation in order to produce volume equations. The timber company routinely measured at the landing in both operations. Obviously, these delays bear no relation to the applied logging methods. The latter kind of delay is referred to as organisational delay. Applying the unavoidable delay time as measured in the 149 Reduced impact logging in the tropical rain forest of Guyana conventional operation also to the reduced impact operation eliminated this difference. Meal-breaks and commuting time were in both cases set at the average time for these elements. Waste Several publications address the aspect of increased volume recovery (utilisation) due to improved felling methods. Winkler (1997) reports a reduction of timber losses due to improper felling and bucking techniques from 8.5% to 3.9%. Hendrison (1990) reports a reduction of felling waste from about 7% to about 2%. Wasting of usable volume can take several forms, the most important being unnecessarily high stumps, inaccurate topping, splitting of the butt end on felling or the top end on bucking. Possibilities to reduce stump height or to increase log length by topping the tree further up the crown were found to be limited because of inherent architectural tree properties (forking). The reason why stump height can be lowered only slightly is related to buttressing of most of the trees. The wood properties between the buttresses and also at or beyond the first fork differ markedly from the wood properties in the bole itself. Between the buttresses and at the base of a branch or fork, tension and pressure wood is formed, which reduces its suitability for sawn timber or plywood. Still, the height of the stumps was often unnecessarily high in the conventional operation. Moreover, most logs produced in the conventional logging were split at the butt end due to improper felling techniques (Chapter 3). Mainly the absence of a hinge to direct the tree and the felling of the tree in the direction of its natural lean provoked this. The reduced wood recovery because of these two features amounted to a wasted butt end piece of one meter in length on average. The reduced wood recovery was expressed in a reduced volume by using volume equations instead of the actual individual tree volumes. By attaching a fixed volume to each tree with a certain diameter of a certain species, waste estimates are not confounded by individual differences in bole length. The latter is rather related to the site index or to the local forest architecture, which difference was intentionally eliminated (see the start of this section). Volume equations were constructed, based on the measurements carried out in the reduced impact logging operation (see Appendix A). Waste is deducted from the volume obtained from the equation. It is defined as: vob = π where 4 ⋅ dbh2 vob dbh = waste volume over bark in m3 = diameter at breast height in m 5.2.2 Skidding model While felling costs are mainly determined by the size and type of trees that are to be felled and relatively uniform for different types of rain forest and terrain, skidding costs are affected substantially by terrain features and by transport distances (Conway 1982; Hendrison 1990; Sundberg & Silversides 1988; Silversides & Sundberg 1989). In the present study, the average skidding distance was much greater in the 150 Chapter 5: Logging efficiency conventional operation as shown later. In addition, the terrain differed substantially. The conventional operation was performed on remarkably flat terrain, situated mainly on sandy soils, whereas the reduced impact operation was conducted on more dissected terrain, mainly on sandy clay soils (Khan 1994 pers.comm.). In skidding it is not so much the diameter of the trees that plays a decisive role, as the composition of the load in respect to weight and number of logs, that determines the time a particular work element will consume. Corrections that have to be made are therefore far-reaching. Collecting the load The time needed to collect a load is determined by the size of the load - in respect to the number of logs and its total volume, the distance between the logs and terrain characteristics. Obviously, given that all harvested trees are in a few diameter classes only, the volume of a load is strongly correlated with the number of logs that make up the load. The multiple regression analysis is carried out stepwise with volume and number of logs as putative variables. The volume of the load is added to the model with the number of logs already in it and vice versa. Each time the significance is tested of the additional variable. This type of model is used for four work elements: approaching, hooking, manoeuvring and bunching. These work elements have different features in the two logging methods. Approaching a new set of logs to be transported is carried out in different fashions. As described in Chapter 3, in the conventional operation the skidder extends the trail after returning from the log landing. In the reduced impact operation, on the other hand, main trails are established before the actual extraction commences. The skidder winches logs to this main trail whenever possible, but occasional branch trail construction cannot be avoided. This implies that, in the latter situation, the time involved in constructing the main trails has to be evenly distributed over those logs that are to be extracted along that trail. A separate analysis was carried out to determine this additional element. In the conventional operation, hooking the logs involves the positioning of the log by moving it with the blade - often including trying to raise the butt off the ground -, moving to the end of the log and hooking it. In the reduced impact operation, hooking involved pulling out of the winch line, choking the log (chokers are set by an assistant while the skidder makes the trip to the market) and hauling the log with the winch. Occasionally, the log has to be positioned - especially in case of large logs. This is regarded as a work delay. Manoeuvring involves the moving of the logs to the place where the load will be bunched. In the conventional operation, sometimes small bunches are made, which are later combined. All activities involved in this ‘pre-bunching’, such as unhooking, ‘ramping’ (positioning the logs for bunching) and bunching, are included in manoeuvring. Bunching involves unhooking of the logs, ‘ramping’ and the hooking of the final load. The multiple regression model includes the following putative predictor variables: 151 Reduced impact logging in the tropical rain forest of Guyana t = a1 + a 2 ⋅ [M ] + b1 ⋅ V + b2 ⋅ [M ]⋅ V + c1 ⋅ L + c 2 ⋅ [M ]⋅ L where: t [M] V L aI , bI , ci = time consumed (min) = dummy variable; equalling 1 in case of reduced impact logging and equalling 0 in case of conventional logging = volume of load (m3) = number of logs in load (1 to 5) = regression coefficients Travel to and from landings The time consumed for light and loaded travel was measured during the extraction of one and part of a second conventionally logged block and during the extraction of several plots in the reduced impact logging operation. In that conventionally logged unit the skidder was not guided along any cut lines and, according to our field observations elsewhere in this and other concessions, this is a representative situation. As mentioned above the terrain conditions differed for the two areas. In order to approach ceteris paribus conditions a selection of the reduced impact logging measurements is used, whereby extraction routes that involve slopes larger than 10% were excluded. Consumption of time is compared by using the following regression models: t lo = a1 + a 2 ⋅ [M ] + b1 ⋅ V + b2 ⋅ [M ]⋅ V + c1 ⋅ D + c 2 ⋅ [M ]⋅ D t li = a1 + a 2 ⋅ [M ] + c1 ⋅ D + c 2 ⋅ [M ]⋅ D where: tlo tli [M] V D = time consumed transporting load to landing (min) = time consumed returning light from landing (min) = dummy variable; equalling 1 in case of reduced impact logging and equalling 0 in case of conventional operation = volume of load (m3) = actual travelled distance (m) Average skidding distance and average load A model is constructed whereby an average load of 9.7 m3 is assumed. Because of a higher recovery of gross tree volume in the reduced impact operation, logs are assumed to contain a higher volume. Accordingly, an average load in reduced impact logging is assumed to contain 3.1 logs, whereas it is assumed to contain 3.4 logs in conventional logging. Furthermore, an average skidding distance is assumed based on the road density and distance between log landings in the concession. Logs are not piled up uniformly along roads, but transport from an area to a road terminates at certain landings. A simple method for determining the average skidding distance is to simplify the stumpage area to a rectangle and to assume that the landing will be situated in the middle of one of the sides of the rectangle. The average skidding distance is then given by the following equation (Peters 1978): 152 Chapter 5: Logging efficiency ⎧ ⎡ ⎛⎜ S + ( S 2 + L2 ) ⎞⎟ ⎤ ⎡ ⎛⎜ L + ( S 2 + L2 ) ⎞⎟ ⎤ ⎫ 2 ⎢⎝ 1 ⎪⎪ L2 ⎢ ⎝ ⎠⎥ S ⎠ ⎥ ⎪⎪ 2 2 ln ⎢ ASD = ⎨ ( S + L ) + ln ⎢ ⎥+ ⎥⎬ 6⎪ 2S ⎢ L S ⎥ 2L ⎢ ⎥⎪ ⎪⎩ ⎣ ⎦ ⎣ ⎦ ⎪⎭ where: ASD = average skidding distance (m) L = Landing spacing (m) S = Road spacing (m) Winding coefficient The ASD is a straight-line (as the crown flies) distance. The actual trail obviously does not follow a straight line. For instance, there are many obstacles (trees, chablis, logging debris), that cause the skid trails to be winding. In addition, poor organisation of a skid trail network takes its toll of skidding efficiency. Other deviations for which we have to correct are the fact that trails meet in junctions and terminate as dead ends (Von Segebaden 1964, 1969). Together these deviations determine a mathematical parameter the so-called ‘winding coefficient’. The winding coefficient is based on empirical measurements of straight-line distances and travelled distances. These measurements were carried out within the areas that were actually logged and within these areas only where skid trails were mapped. For reduced impact logging this means that the measurements are confined to the assessment plots and for conventional logging the demarcated 12-ha units (see Chapter 2). These empirical measurements render a fair picture of the degree to which the trails deviate from a straight line, but neglect the fact that the trails terminate at one landing. The winding coefficients of several extraction patterns are worked out and subsequently multiplied by the result of the empirical measurements. Other variable work elements Two more work elements may be influenced by the logging method. These are (avoidable) work delays and discharging of the load at the landing. Work delays usually occur during the gathering of the load. Differences between the logging methods were tested with the non-parametric Mann-Whitney U-test. The effect of load size and logging method on the time of discharging a load is tested with multiple regression analysis. Servicing and repairs Servicing of the skidder usually takes place at the start and at the end of the day. It includes refuelling and regular checks of oil, bearings, joints, etc. Minor repairs occur when one of the parts of the machine malfunctioned and it can be fixed by the skidder operator at the spot. The machines used in the operations were of the same model, fitted with the same gear and of similar age. However, the state of maintenance differed between the machines. Obviously, the effect of such incidental difference on the efficiency should be minimised in the generic computations. As mentioned under machine costing, the cost of servicing and repairs of a machine is directly related to the operative machine time. In order to eliminate the effect of the state of maintenance on 153 Reduced impact logging in the tropical rain forest of Guyana time spent on servicing and repairs, the time consumed by this work element is set at a fixed proportion; i.e. the ratio of the time consumed by servicing and minor repairs to the daily operative machine time in the conventional operation. Crew down time and unavoidable delays For the crew down time in the skidding operation, the same approach is used as with the felling operation. Unavoidable delays are mainly of organisational or technical nature and differed between the two operations. Many of these delays are not considered to be representative of a normal skidding operation: ƒ waiting for a tree to be felled in the conventional operation; ƒ decking of logs at the market because of reduced storage capacity at the landing occurring mainly in the reduced impact operation; and ƒ frequent machine breakdowns in the reduced impact operation. Such delays are excluded from the model. The remaining unavoidable delays that occurred in the conventional operation - such as freeing hang-ups, changing tyres in case of punctures, etc. are considered to be standard. The amount of time spent per working day on assisting the felling crew in freeing hang-ups was applied to both methods. Other unavoidable delays, such as malfunctioning of the skidder, were considered to represent a fixed proportion of the operative machine time. The proportion observed in the conventional operation was applied to both methods. 5.2.3 Skidding model scenarios Based on the procedures described above a basic skidding scenario is constructed for the two logging methods. The point of departure in both cases is an average (straight line) skidding distance based on the average road density in the concession, one winding coefficient to obtain an estimate of the travelled distance for both methods, and a load consisting of three logs of average diameter. An alternative scenario is based on the assumption that winding coefficients differ between the two methods. For the conventional operation, a third alternative is formulated. This is inspired by the fact that stumps were distributed regularly in the reduced impact logging operation, as opposed to the conventional logging operation where stumps, hence logs, had a clumped distribution. It is expected that this influenced the time spent on collecting logs. First, it is examined whether the time involved in the activities that make up the collection phase, is proportionate to the number of logs that are collected. Successively, it is examined whether there are any differences in this respect between the logging methods. 5.2.4 Limitations Considerations are confined to the stage of development when an operation has reached a ‘normal’ performance level. This means that no specific considerations have been made to the adverse impact from initial periods of reduced efficiency caused by employment of inexperienced staff, training of staff, etc. This does not imply that the level of training of staff was sufficient, especially as regards directional felling techniques. Costs depend on several factors including forest stand conditions, weather 154 Chapter 5: Logging efficiency and skill and dexterity of labour. Prices for logs vary by species, grade and market conditions at the time of sale. The results of the present study apply to forest stand conditions found in Central Guyana, to the harvesting of Greenheart forest and to the 1998 price situation. The estimates in the present study are vulnerable to error due to the assumptions made. Potential errors were touched upon in the methodology section. The validity of the result in absolute terms of cost thus must be expected to be limited, and the scope for improvements is evident. At this stage, the results of the study should therefore be considered to give no more than an indication of the difference in logging costs and benefits between the two methods of harvesting in Guyana. 5.3 FELLING EFFICIENCY The efficiency of a felling method can be determined by looking at the output obtained with certain input. The input can be reliably determined by measuring the time consumed by the work elements that make up the felling work cycle. This cycle is the total time involved in the first phase of harvesting and includes moving from one tree to the next, felling preparations, actual felling and bucking. The output is measured as the volume in cubic metres of wood that is processed. 5.3.1 Work study statistics The time study of the reduced impact operation was carried out between March and June of 1994 (Brils 1995). Only days that have been monitored during the entire day are included in this analysis, resulting in 21 days. Distributed over six plots of the logging experiment (Section 1.7.3) the felling of 147 trees was monitored. The 147 trees belonged to 15 different species, Chlorocardium rodiei and Catostemma fragrans being the most common species with 57% and 15%. A conventional operation was monitored during 8 days between March and April 1996. The felling comprised 130 trees, belonging to four species, 94% of which was Chlorocardium rodiei. The average time spent during a day’s work - the ‘camp-to-camp time’ - is subdivided in one out of three classes: i.e. ‘crew down time’, ‘effective crew time’ or ‘organisational delay’ (FAO 1978). Their distribution together with data concerning the trees that were involved in the work samples is listed in Table 5.1. Felling of trees is extremely heavy work, often so heavy that the physical work capacity of the operator puts a practical limit on performance. The weight of the chainsaw combined with the noise and vibration it produces puts a severe strain on the operator’s physique. In addition, working in a hot and humid environment, such as the tropical rain forest, causes heat stress to the body. High humidity combined with high temperatures is detrimental to the operator’s work capacity, because heat elimination at the body surface depends mainly on transpiration, which is restricted by the high humidity. Frequent rest breaks are thus needed during a working day. Nevertheless, the fatigue build up during the course of the day usually does not allow the operator to work for 155 Reduced impact logging in the tropical rain forest of Guyana more than 5 to 6 hours. The effective working hours amounted to 4.5 hours in the conventional operation and to 3.7 hours in reduced impact operation. The low score of around 4.5 effective working hours in the conventional operation is also related to a target set at 15 trees to be felled per day. The difference between conventional and reduced impact logging is probably related to fatigue being mainly dependent on the actual sawing time (Table 5.2). The actual sawing time (felling and bucking) amounted, on average, to 1.3 hours in the conventional and 1.6 hours in the reduced impact operation. Table 5.1 Felling efficiency: daily effective crew time and production statistics based on the raw recorded data of the time studies of the two logging methods Subject of observation Effective crew time (min) Operative machine time (min) Crew down time (min) Organisational delays (min) Camp-to-camp time Conventional 269 79 104 7 380 Trees felled/day Lodged trees/day Volume/day (m3) dbh (cm) Volume/tree (m3) Logging intensity (n/ha) 16.3 3.0 46.4 55 2.9 8.3 Logging method Reduced Impact 221 94 79 58 358 7.0 1.3 21.2 53 3.0 9.7 The difference in crew down time is related to a difference in accessibility of the areas. The area where the reduced impact logging took place was close to passable forest roads, whereas two of the three conventionally logged units were located at two kilometres walking distance. The large difference in organisational delays was already highlighted in the methodology. In the conventional operation, organisational delays were restricted to work organisation; e.g. waiting for transportation. In the reduced impact operation logs were measured at the stump and the felling crew often had to wait to make correction cuts, or had to assist in clearing debris. Moreover, the state of maintenance of the saw used during the first part of the reduced impact operation left a lot to be desired. Break-downs occurred frequently due to worn out parts. Excluding time for commuting, meal-breaks and organisational delays, the time required per felled tree was, on average, 17 minutes in the conventional operation and 32 minutes in the reduced impact logging system. The difference between the methods in volume produced per hour is somewhat smaller: 10.3 m3/hr as opposed to 5.8 m3/hr, which is related to the higher rate of wood recovery in the reduced impact operation. As mentioned in the methods section, besides organisational delays, several other interfering factors influenced the results, for which further corrections will be made in the following sections. In order to find and eliminate noise factors and to pinpoint the time elements leading to the difference in performance stated above, the times spent on 156 Chapter 5: Logging efficiency the individual work elements are compared. The uncorrected average time of work elements involved in the felling of a single tree for both systems is shown in Table 5.2. Table 5.2 Felling efficiency: actual operative and supportive time elements required to accomplish felling of a single tree in conventional and reduced impact operations Work element Location of tree Felling preparations Felling Bucking Operative work time Machine servicing Minor repairs Work delays Unavoidable delays Freeing hang-ups Supportive work time Felling cycle Mean time taken in felling method (min) Conventional Reduced Impact 0.6 0.9 2.2 4.8 2.8 10.7 2.0 2.7 7.7 19.1 3.1 0.4 4.8 0.1 0.4 8.8 3.5 0.2 4.4 0.5 3.8 12.4 16.6 31.5 The comparison of the time distribution of work elements that occurred in felling of a single tree in both systems reveals that both operative and supportive times were higher in the reduced impact operation. Among the operative work elements, especially the enormous increase in felling and preparation time catches the eye. The observed differences may be related to the inexperience with directional felling techniques and to differences in the size distribution of the felled trees. In case of the supportive work elements it is clear that certain differences are unrelated to the working method per se: unavoidable delays and time spent on freeing lodged trees. The improper felling and bucking techniques in the conventional operation more frequently led to the saw becoming stuck in the cut. This not only demanded time to cut it free with a cutlass (increased work delay), it also involved time to fix the saw (increased repair time). One saw was demolished in the conventional operation when a tree kicked back and landed on top of it. The influence of improper felling techniques on the longevity of equipment, although real without any doubt, was not taken into consideration in the present study. 5.3.2 Operative work elements Locating trees In both operations, trees were located with the aid of a tree location map. The accuracy and scale of the maps differed considerably but this did not seem to affect the time to locate the next tree too much. In the conventional operation felling started with the tree that was closest to the unit’s point of entry. The search for the next tree usually began while the operator was still busy felling the tree, but not systematically. Searching the area while the feller is still sawing is not free of danger, especially because the feller had little control over the direction in which the tree would fall. Sometimes the logging 157 Reduced impact logging in the tropical rain forest of Guyana clerk realised he had left out a tree and the crew had to return on their footsteps to relocate it. Since the skidder operator more or less followed the felling crew in their footsteps, the unsystematic search pattern was reflected in the skidding pattern. Felling trees in the direction of other commercial trees also hampered locating trees. In the reduced impact operation searching started from the back of the unit and proceeded systematically to the point of entry. If the skid trail was designated to run from east to west, searching proceeded from north to south and back. For reason of job safety, the system dictates the search to begin after the tree was felled and bucked. Figure 5.1 The time required to move to the next tree in relation to logging intensity in a conventional and a reduced impact operation in Pibiri, Central Guyana. Felling was monitored in plots with three different logging intensities in the reduced impact operation and in two units with different logging intensity in the conventional operation. For the observations in the conventional operation alone the following equation was found: (time to locate tree) = 1.60* – 0.11ns ⋅ (intensity); R2 = 0.02. For reduced impact operation alone: (time to locate tree) = 1.64*** – 0.05*** ⋅ (intensity); R2 = 0.09. Intercepts of these two regression lines differ significantly (t245 = 8.29, p < 0.001), but slopes do not (t244 = 0.78, p = 0.44). The best fitting model is (R2 = 0.10; in graph); conventional operation: (time to locate tree) = 1.13*** – 0.05*** ⋅ (intensity) and reduced impact operation: (time to locate tree) = 1.67*** – 0.05*** ⋅ (intensity); Symbols behind coefficients denoted their probability of not equalling zero: * p < 0.05, *** p < 0.001 and ns p > 0.05. The average recorded tree location time in the conventional operation was 0.6 minutes compared to 0.9 minutes in the reduced impact operation. Since the time to locate trees is with certainty related to the logging intensity, multiple regression analysis was performed to establish the effect of this co-variate. It appeared that consistently 0.5 minutes more time was needed in the reduced impact operation, increasing by 0.05 minutes with every extra tree harvested per hectare (Fig. 5.1). Walking to the first tree on a particular day is included in commuting time. Applying the regression model to a logging intensity of 10 trees per hectare including the remaining trees only yields an average time for locating a tree of 0.6 minutes for the conventional and 1.1 minute for the reduced impact operations. 158 Chapter 5: Logging efficiency Surprisingly, it took less time to locate trees in the conventional operation. The results are in contrast with what has been reported elsewhere in two ways. First, introduction of planning is suggested to generally reduce the time needed to locate trees. Secondly, the time spent on locating trees was lower in the present study than reported elsewhere. Hendrison (1990) found 4.1 minutes for a conventional and 1.2 minutes for his ‘damage-controlled’ operation in Suriname, both operations with a logging intensity of about 10 trees per hectare. In two conventional operations in French Guiana, searching of trees took about 3.3 minutes on average (Castanet 1989; logging intensity unknown). In the state of Amazonas, Brazil, more time was required to locate trees in a conventional and reduced impact operation too. In contrast to Hendrison’s study, more time was spent on locating a tree in the Brazilian reduced impact operation; i.e. 2.1 as opposed to 1.3 minutes in the Brazilian conventional operation (Winkler 1997). The latter is explained by the much lower logging intensity in the reduced impact operation; i.e. 6 trees/ha as opposed to 16 trees/a in the conventional operation. Taking the differences in logging intensity into account, locating trees to be felled does not appear to differ much between planned operations. In the conventional Greenheart operation that was studied, the time spent on locating trees was remarkably short. The difference with the studies in Brazil and Suriname is partly explained by the fact that a tree location map was used, and partly by the fact that trees to be harvested had a clumped distribution in the conventional operation in Guyana. Probably for the same reason, trees were located quicker than in the reduced impact operation in Guyana, despite an unsystematic search procedure. The fact that the search of the next tree started before the felling was completed in the conventional operation probably contributed to this difference too. Felling preparation Felling preparations involved cutting of undergrowth and lianas surrounding the base of the tree, removal of loose or sandy bark at the height of the felling cut and a quick judgement of the natural lean of the tree to determine the felling direction in the conventional operation. Flight paths were incidentally cut, but only in case of dense undergrowth. In the reduced impact operation removal of undergrowth and obstacles around the tree were carried out more painstakingly. Determination of the felling direction involved checking the tree’s position in relation to the designated skid trail and a thorough inspection of the surroundings of the tree. Flight paths were always cut. The average recorded preparation time in the conventional operation was 2.2 minutes compared to 4.8 minutes in the reduced impact operation. Since the size of the tree influences the time spent on cleaning the trunk at the height of the felling cut, multiple regression analysis was performed to determine the influence of the tree’s diameter. It appeared that, for small trees, the preparation time was longer in the reduced impact operation: i.e. 4.7 minutes in the reduced impact against 1.7 minutes in the conventional operation (trees with dbh = 47 cm). This difference appeared to diminish with increasing diameter; i.e. 3.7 minutes in the conventional as opposed to 4.7 minutes in the reduced impact operation when a large tree (dbh = 87 cm) is felled (Fig. 159 Reduced impact logging in the tropical rain forest of Guyana 5.2). Applying the regression model to the diameter distribution in the conventional operation, a preparation time of 2.1 minutes is found for the conventional operation and 4.7 minutes for reduced impact logging. Figure 5.2 Felling preparation time in relation to the diameter of the tree to be felled in a conventional and a reduced impact operation in Pibiri, Central Guyana. Felling preparation time is remarkably constant in the reduced impact operation, whereas it increases with diameter in the conventional operation. Reduced impact logging operation: (preparation time) = 4.69*** + 0.01ns ⋅ dbh; R2 = 0.00. For conventional: (preparation time) = –0.57ns + 4.86*** ⋅ dbh; R2 = 0.10. Intercepts and slopes of the two regression lines differ significantly (t264 = 4.68, p < 0.001 and t264 = –2.35, p < 0.05 respectively). Symbols behind coefficients denoted their probability of not equalling zero: *** p < 0.001 and ns p > 0.05. The amount of variation explained by including the diameter of the tree is minimal (R2 = 0.10). The felling preparation time varied between zero and eight minutes in the conventional and between zero and twelve minutes in the reduced impact operation. The observations at the lower end are probably related to the fact that preparations on a tree sometimes commenced before the operator had completed the cross-cutting of the previous bole. The fact that the preparation time was remarkably constant with increasing tree diameter in the reduced impact operation is explained by the fact that the operator determined the felling direction, while his assistant removed loose bark and established a flight path. Determining the felling direction was the decisive factor here, whereas this was the cleaning of the stem in the conventional operation. The difference in preparation time reflects the crew’s position on felling work. The main objective of the crew in the conventional operation was to achieve their target of 15 trees without paying too much attention to safety or timber and stand damage. Although they were aware of the dangers involved in felling, they were unaware that their way of felling aggravated this. Preparation times in the present study are similar to the findings of Hendrison (1990) in Suriname, 2.0 minutes and 4.3 minutes for conventional and damage-controlled felling respectively, and of Winkler (1997) in Amazonas state, 1.7 minutes and 3.5 minutes respectively. 160 Chapter 5: Logging efficiency Felling In the conventional operation, the operator tried to bring the tree down as quickly and easily as possible, without paying too much attention to potential timber or stand damage (see Chapter 3). For instance, trees were regularly felled cross-wise, whereby the tree fell on a previously felled bole, which substantially increases the chance of breaking of the log. In the reduced impact operation, a felling technique was used that maximised the control over the felling direction of the tree, so reducing the hazards inherent to tree felling as regards safety and damage prevention (Brunberg et al. 1984, Conway 1982). To this effect, felling aids were introduced; i.e. aluminium felling wedges and a sledgehammer, and the crew was expanded with one assistant to carry out wedging. (For details on the felling methods, see Sections 3.2.1 and 3.3.4). A one-month period of training and supervision preceded the actually monitored work. Training not only focused on performing the required felling technique, but also aimed to increase the felling crew’s awareness in regard to safety, efficiency, and stand and timber damage prevention. The emphasis of felling work so shifted from work targets to work quality. The operators were particularly impressed by the increase of job safety by employing the technique1. The mean actual felling time in the conventional operation was 2.8 minutes as opposed to 10.7 minutes in the reduced impact operation. The felling time is determined by felling technique, wood density and other properties (silica, latex) and surface to be cut; i.e. the basal area of the tree. The average tree diameter was smaller in the reduced impact operation. In the reduced impact operation, however, a higher number of large trees (dbh > 70 cm) and a higher number of small trees (dbh < 50 cm) were involved, providing for a significant difference in the distribution of the diameters (Kolmogorov-Smirnov two-sample test D = 0.180, p < 0.05). In order to correct for this difference, multiple regression analysis was performed (Fig. 5.3). The equation was applied to the diameter distribution found in the conventional operation. The average felling time then becomes 2.6 minutes for conventional and 10.1 minutes for reduced impact felling. The time required to fell a tree was longer than found by Hendrison in Suriname: 1.5 minutes for conventional felling and 2.7 for damage-controlled felling. In the State of Amazonas, Brazil, conventional felling was found to require 6.3 minutes, whereas environmentally sound felling required 8.7 minutes (Winkler 1997). Castanet (1989) found times varying between 10 and 15 minutes for the entire felling operation (preparation, felling and bucking) in French Guiana, which is comparable to what was found in Brazil. As regards conventional felling, the differences between the studies are probably determined by differences in tree diameter. Reduced impact felling, however, took relatively much more time in the present study than what was found elsewhere. This is probably due to differences in the felling technique. The felling technique used in the studies in Brazil and Suriname was a conventional directional 1 One of the operators, Mr. W. Jarvis, demonstrated this to other fellers by sitting down under a tree of which only the piece of holding wood (the ‘heel’) was left uncut during heavy gushes of wind. When a traditional felling cut is partly finished, a feller usually takes to his heels in such weather conditions. 161 Reduced impact logging in the tropical rain forest of Guyana felling technique as described in e.g. Conway (1982) and Klasson & Cedergren (1996). This technique involves the use of wedges, but does not involve cutting backward from the hinge to the back of the tree. Figure 5.3 Felling time in relation to the diameter of the felled tree in conventional and reduced impact operations in a conventional and a reduced impact operation in Pibiri, Central Guyana. The time required to fell a tree is substantially increased by employing the directional felling method. The difference increases with increasing diameter: (felling time) = 3.17** + 21.98*** ⋅ dbh2 for reduced impact logging (R2 = 0.19). For conventional logging: (felling time) = 0.51* + 6.50*** ⋅ dbh2; R2 = 0.40. Note the square function indicating the relation between time spent and surface [π⋅(d/2)2] cut. The second coefficient of the two regression curves differ significantly (t265 = 3.62, p < 0.001), indicating an interaction between the influence of the felling method and squared tree diameter. The complete model is: (felling time) = 0.51ns + 2.65ns ⋅ [M] + 6.50* ⋅ dbh2 + 15.47*** ⋅ [M] ⋅ dbh2; and has an R2 = 0.50. [M] is a dummy variable that equals 1 for reduced impact and 0 for conventional logging. Symbols behind coefficients denoted their probability of not equalling zero: ns p > 0.05, * p < 0.05, ** p < 0.01 and *** p < 0.001. Bucking Bucking (‘junking’) includes the topping of a felled tree and corrective cross-cutting at either butt or top end or both, in case of internal, hidden decay or in case of felling damage such as splits or shakes. Proper bucking requires considerable skill and dexterity. The top of the bole is often under stress and may split if cross-cutting is not carried out correctly. On flat terrain, the top of the bole is raised above the ground, because it is still attached to the crown and compression forces act upon the upper side. If due to terrain micro-topography or obstacles, the top is not raised, compression forces will act upon the lower side. Proper cross-cutting techniques are described in e.g. Conway (1982). Besides the hazard of damage to the log, a lack of understanding of stress and lack of experience how to deal with it, may lead to the saw bar becoming stuck in the cut. Bucking is not free of safety risks, because of the real chances that the bole may ‘kick back’ when severed from the crown. 162 Chapter 5: Logging efficiency In the conventional operation, misjudgements were often made when assessing the side of the log to be cut, as being under tension or compression forces. The decision as to the cross-cut point on the log was determined by accessibility and ease, and not on maximum length of the bole. Correction cuts were seldom made. This is partly because butt decay often does not extend far up the trunk in Greenheart (De Milde 1970). By felling trees, in the size classes that were felled, at a height of about one metre above the ground, internal decay is ever encountered anymore. If correction cuts were needed, they were often postponed and carried out at the log landing after extraction. In the reduced impact operation, logs were being topped at such a point as to maximise bole length. This could require removal of undergrowth, felling debris or branches before actual topping. Correction cuts were frequently carried out. Partly, this is related to the lower stump height of about 50 cm above ground level, exposing hidden decay. In addition, more species were involved in this operation. In Chlorocardium rodiei, the operator can get a fair impression of the soundness of the bole by tapping on it, because of its high wood density and thin, hard bark. In other species, with a thicker and/or softer bark and/or lower wood density, this procedure did not prove to be reliable, and defects often remained hidden until revealed by felling. Obviously, these aspects have a negative influence on bucking time. The use of felling wedges and proper techniques only partly compensated for this loss of time. Moreover, in case of pinching of the saw bar, the time spent to free the guidebar was no longer recorded as ‘bucking’, but as ‘work delay’. The average time required to buck a log amounted to 2.0 minutes for conventional and 2.7 minutes for reduced impact felling. Similar to the time required to fell a tree, the time to cross-cut a log is related to the technique, the wood characteristics, and the surface to be cut. Multiple regression analysis was performed for bucking time as well (Fig. 5.4). The equation was applied to the diameter distribution found in the conventional operation. The average bucking time then becomes 1.9 minutes for conventional and 2.8 minutes for reduced impact felling. The difference in time required to buck a log is explained by the difference in approach to this activity. Another plausible explanation is the difference in saw bar length. The saw bar used in the conventional operation had a length of 76 cm. Hence trees up to 70 cm could be crosscut in one single cut. In the reduced impact operation, the guidebar measured 14 cm less, so only trees up to 55 cm could be cut. One actually would expect a twist in the curves shown in figures 5.3 and 5.4 around 70 cm dbh for conventional and around 55 cm for reduced impact logging. The reason this was not observed is the small number of observations on large trees in conventional and the enormous variation in time recordings in reduced impact logging. In Hendrison’s (1990) study in Suriname, less time was required for bucking than in the present study. He found bucking to take 1.8 minutes in conventional and 2.1 minutes in damage-controlled felling. Winkler (1997), on the other hand, reported longer times; i.e. 4.9 and 5.1 minutes respectively. These differences are presumably caused by differences in tree diameter. 163 Reduced impact logging in the tropical rain forest of Guyana Figure 5.4 Time required to buck a log in relation to tree diameter in conventional and reduced impact operations in a conventional and a reduced impact operation in Pibiri, Central Guyana. Time required significantly depends on tree size and consistently is about a minute more in the reduced impact operation than in the conventional operation. For the reduced impact operation the following equation was found: (bucking time) = 0.47ns + 7.34*** ⋅ dbh2; R2 = 0.28. For conventional logging: (bucking time) = 0.07ns + 5.81*** ⋅ dbh2; R2 = 0.26. Intercepts of these two regression lines differed significantly (t246 = 5.21, p < 0.001), but the coefficients of the second (square) term did not (t245 = 1.09, p = 0.28). The best fitting model is (R2 = 0.31; in graph); conventional operation: (bucking time) = –0.19 + 6.66 ⋅ (dbh)2 and reduced impact operation: (bucking time) = 0.66 + 6.66 ⋅ (dbh)2; Symbols behind coefficients denoted their probability of not equalling zero: *** p < 0.001 and ns p > 0.05. 5.3.3 Supportive work elements and delays Servicing and repairs The servicing of a chain saw in the field consists of refilling of the fuel and chain oil tanks, sharpening of the chain, cleaning the saw (removing sawdust) and adjustment of the chain’s tension. Occasionally, a chain is exchanged for a spare one. Fuel consumption depends for one on the power rating and engineering of the machine in question, and on how the machine is handled by its operator. On the other hand, the condition of the chain (sharp or blunt), the surface to be cut, and wood and bark properties (density, presence of silica or latex) influence the time needed to cut a tree and hence the daily fuel consumption. Therefore fuel consumption, hence time required to refuel, relates to the time involved in sawing. The time to sharpen the chain also depends on the time used for sawing. It is also influenced by how dull the chain is, which is related to the nature of the cut. Brils (1995) found that frequent sharpening of not too dull a chain is more efficient than waiting for the chain to become completely dulled. The average time spent per day on refuelling amounted to 12.4 minutes in the conventional operation and 5.0 minutes in the reduced impact operation. The average time consumed by sharpening per day was 37.8 minutes against 20.7 minutes. The 164 Chapter 5: Logging efficiency average time spent on cleaning the saw did not differ: about 1 minute. As mentioned in Section 5.3.1 the actual time the saw was running was longer in the reduced impact operation. Machine maintenance therefore appears relatively less time-consuming in that case. This is caused by a difference in engine capacity and fuel efficiency between the two chainsaws, but also by a different attitude of the operators. The operator in the conventional operation worked his machine at full throttle most of the time, while the other operator worked often at part throttle. Sharpening taking more time in the conventional operation is partly explained by its long postponement, which is inefficient (Brils 1995). The regular jamming of the saw affected the condition of the chain too. The total servicing time amounted to 50.9 minutes in the conventional and 26.2 minutes in the reduced impact operation. This is considerably more than Hendrison (1990) found in Suriname and reversed; i.e. 18 minutes in damagecontrolled against 34 minutes in conventional logging. Winkler (1997) reported the average time to refuel and maintenance time for the felling of a single tree; both refuelling and maintenance were 0.3 minutes for the ‘environmentally sound’ system and 0.6 minutes for the conventional system. In the present study, refuelling took 0.8 minutes and maintenance 2.4 minutes per single felled tree for the conventional felling, and 0.7 minutes and 3.0 minutes for reduced impact felling. A model was constructed describing the time consumed by servicing. Regression models were constructed with the daily sawing time as predictor for sharpening time and frequency of tank refilling. The estimated frequency of refuelling is then multiplied by the average time used for refilling. Sharpening time was shown to be independent of sawing time in the conventional operation, and to increase with sawing time in the reduced impact operation. Sharpening time is predicted as (sharpening time) = 93.42* – 0.70ns⋅(sawing time); R2 = 43, in the conventional and as (sharpening time) = 4.83ns + 0.17* ⋅ (sawing time); R2 = 0.21 in the reduced impact operation. Multiple regression analysis showed a significant interaction between sawing time and felling method. The best fitting model involving the felling method and sawing time was (sharpening time) = 93.42 – 88.95 ⋅ [M] – 0.70 ⋅ (sawing time) + 0.87 ⋅ [M] ⋅ (sawing time); R2 = 0.51. [M] is a dummy variable that is coded 0 in case of conventional and 1 in case of reduced impact operation. If the interaction term is ignored, it appears that the ratio sharpening time to sawing time is significantly higher in the conventional operation (t26 = –3.67, p < 0.01). The corrected sharpening time becomes 26.1 minutes in the reduced impact operation and remains unchanged in the conventional operation: 37.8 minutes. It took significantly more time to refill the fuel and chain oil tanks in the conventional operation (t112 = 2.89, p < 0.01); i.e. 1.9 minutes (sd = 0.8) as opposed to 1.3 minutes (sd = 1.1). The latter is a reflection of the larger tank volume 1.2-l as opposed to 0.8-l. The best fitting model for the frequency of refilling indicates that the frequency was lower for the reduced impact operation. This is a reflection of the lower fuel consumption rate of the modern Stihl AV066 compared to the outdated Stihl 070 and a difference in work attitude: the operator in the conventional operation working at full throttle all the time, whereas the other operator worked often at half throttle. The 165 Reduced impact logging in the tropical rain forest of Guyana frequency of refilling related to the sawing time in the reduced impact operation as: (frequency) = 1.13 + 0.03 ⋅ (sawing time), and in the conventional operation as: (frequency) = 3.19 + 0.03 ⋅ (sawing time). The best fitting model had a coefficient of determination of R2 = 0.51. Crew down time and delays Short delays that occur during the felling cycle, that are not caused by differences in organisation (see Sections 5.2.2 and 5.3.1), and thus form part of the effective crew time, are referred to as work delays. Short delays that have natural causes, such as heavy rain, snakes and wasp or ant nests, are referred to as unavoidable delays. Work delays fall into three groups: short, intermediate rest periods, minor repairs and accidental delays. Short rest periods occur between the different phases of the felling cycle, especially after felling and after bucking. Accidental delays can be caused by jamming of the saw bar, by lodging of the felled tree in a neighbouring tree, etc. Minor repairs often occurred as a result of the former. Hang-ups were treated separately, because they were handled in different ways (see Section 5.2.6 and below). The recorded personal rest and accidental delays (excluding hang-ups) together amounted to, on average, 4.8 minutes per single tree in the conventional operation and 4.4 minutes in the reduced impact operation. This delay was dependent on the diameter of the felled tree (t266 = 2.10, p < 0.05), but not on the felling method (t266 = 0.53, p = 0.60). Based on the diameter distribution of the conventional operation, an average delay of 4.5 minutes is used. As already mentioned, minor repairs were associated with accidental delays. The time involved in minor repairs was not significantly influenced by the sawing time (t26 = 0.427, p = 0.67), but differed significantly between felling methods. Minor repairs involved 6.2 minutes per day in the conventional and 1.2 minute in the reduced impact operation. Lodged trees were freed by the skidder in the conventional operation, taking an average 2.1 minute per hang-up (n = 24, sd = 3.0). In the reduced impact operation the crew tried to bring the tree down by wedging, not always successfully. The time spent averaged 19.8 minutes per hang-up (n = 28, sd = 10.5). It is assumed that in a normal operation hang-ups are solved by using skidder traction. The situation in the conventional operation that was monitored with the skidder operating at close range of the felling operation, however, is not considered a representative situation either. Therefore, a - debatable - time of 5 minutes was conjectured as needed to free a lodged tree. Beforehand, the use of directional felling methods was presumed to reduce the frequency of hang-ups. Out of 130 trees felled in the conventional operation, 24 (18.5%) lodged in a neighbouring tree, while 28 out of 147 trees (19.0%) had the same fate in the reduced impact operation; a negligible difference. This lack of difference can be attributed to clumped felling in the conventional operation. Unavoidable delays occurred more frequently in the reduced impact operation and took 5 times as much time. In the model, an average time for unavoidable delays is used. 166 Chapter 5: Logging efficiency 5.3.4 Felling time model Felling cycle The regression analyses described in the previous section were used to build a model for the time required by the felling of a single tree under equal circumstances. Unavoidable delays and delays caused by hang-ups were understood to be of the same magnitude in both operations. The time taken by the other work elements is computed by substituting the diameter distribution in reduced impact logging by the one found in the conventionally logged units. The values for conventional felling differ slightly due to Purpleheart trees being left out of the model. The corrected time study of the felling work cycle shows that introduction of a reduced impact logging system increases the time required to fell a single tree by 75% (Table 5.3). This difference is partly due to the regular spacing of the trees to be harvested as for instance reflected in the increased time to locate the next tree. The increase in tree location time is only 4% of the total increase of 11.6 minutes per felled tree and tree spacing thus plays only a minor role. Extra time is linked to felling - 62% of the total increase is due to this activity - and felling preparations: 21%. Table 5.3 Felling efficiency: corrected operative and supportive time elements required to accomplish felling of a single tree in conventional and reduced impact operations Work element Location of tree Felling preparations Felling Bucking Operative work time Mean time taken in felling method (min) Conventional Reduced Impact 0.6 1.1 2.1 4.7 2.6 10.1 1.9 2.8 7.2 18.7 Machine servicing Minor repairs Work delays Unavoidable delays Freeing hang-ups Supportive work time Felling cycle 3.0 0.4 4.5 0.2 0.9 9.1 3.7 0.2 4.5 0.4 0.9 9.7 16.3 28.4 The time involved in the operative elements of the conventional felling cycle is much lower than Winkler (1997) reports for Brazil; i.e. 7.2 minutes here as opposed to 14.2 minutes in Brazil. Supportive work time was higher, however, resulting in a comparable aggregate estimate of the felling work cycle time. For Suriname, Hendrison (1990) reported 8.4 minutes, which is close to the values in the present study. The total felling cycle plus work delays was found to be 10.5 minutes in Suriname and between 10 and 15 minutes in French Guiana (Castanet 1989), while it amounted to 12.7 minutes here. Differences in tree diameter, undoubtedly, play a role in these comparisons. It may therefore be assumed that the performance in conventional felling in Guyana can be rated as typical for the region. 167 Reduced impact logging in the tropical rain forest of Guyana The time involved in the operative work elements of reduced impact felling was similar to what was found in Brazil; i.e. 18.7 minutes here and 19.3 minutes in Brazil. Supportive work time, however, was much higher in Guyana, i.e. 9.7 against 2.1 minutes. For Suriname, a much lower average of 9.2 minutes was found for the operative working time in damage-controlled felling. If work delays are added to the operative time an amount of 11.0 minutes results, whereas these work elements together consumed 23.2 minutes in Guyana. Even when differences in tree diameter are taken into account, it is clear that the relatively high supportive work time diminished the performance of the improved logging method in Guyana. As regards the effective work time, the time spent on felling is relatively high. It is unknown whether this is related to the short training period of field personnel preceding the study, or whether it is intrinsic to the felling technique used. The felling technique used here is undoubtedly one of the safest felling methods. In addition, it leads to very little timber damage. In view of logging efficiency, the main point of interest is, whether the felling pattern resulted in an increased efficiency of the extraction phase (see Section 5.4.). Table 5.4 Felling efficiency: effective crew time in minutes and production statistics of a fair day’s work, based on the felling time models for the two logging methods. Note that organisational delays are not included in the model. Subject of observation Effective crew time (min) Operative machine time (min) Crew down time (min) Camp-to-camp time Conventional 385 108 95 480 Trees felled / day Volume / day (m3) Trees felled / effect. crew hour Volume / effect. crew hour (m3) dbh (cm) Recovered volume/tree (m3) Logging intensity (n/ha) 24 68 3.7 10.6 55 2.9 10 Logging method Reduced Impact 385 179 95 480 14 43 2.2 6.7 55 3.1 10 Output Based on the time involved in the felling of a single tree, the performance of ‘a fair day’s work’ is estimated (Table 5.4). This estimate builds upon an 8 hours working day, a commuting time as found in the reduced impact operation and a suggested lunch-break of 30 minutes. The total effective crew time then equals the remaining hours. All time elements are directly related to the number of trees that are felled, except the time elements servicing and tree location. Servicing is related to the operative machine time. Locating trees is related to a suggested logging intensity of 10 trees/ha and the number of trees that are felled per day minus one (searching for the first tree is included in commuting time). In order to calculate the number of trees 168 Chapter 5: Logging efficiency felled during the effective crew time, the iterative Microsoft Excel Solver algorithm is used. It appears that the implementation of reduced impact felling reduced felling performance by 37% (Table 5.4). In addition, the operative machine time was increased by 66%, which will further increase the cost of felling. How these changes affect the cost of felling will be dealt with in Section 5.5. 5.4 SKIDDING EFFICIENCY Skidding is the second phase of harvesting and includes collecting logs, composing a load, and skidding the load to the log landing, discharging and return to the stump area. The performance of skidding depends on many different factors such as log size, the ease with which logs are collected from the stump, the ease of travelling (hard versus smooth surface, grade of adverse slopes, etc.) and - most important - the travelling distance (see Section 5.2.7). Furthermore, the geometric efficiency of the skid trail pattern and the efficiency of the load collection procedure are important variables. The efficiency of the latter two can be improved by upgrading the skidding method. The felling efficiency was mainly determined by tree characteristics and the method used. In skidding, terrain characteristics and skidding distance weigh heavily on the result, and, hence, necessary corrections will reach much further. 5.4.1 Work study statistics The time study of reduced impact skidding took place during October and November 1994. The analysis concerns only those days that were monitored over the entire working day, which amounted to 10 days. Distributed over eight plots of the logging experiment (see Chapter 2), the extraction of 183 logs in 73 trips was monitored. A time study was conducted in a conventional operation during 13 days in March and April 1996. During these 13 days, 201 logs were extracted in 60 trips from two separate units (see Chapter 2). The trip from the stump area to the landing, unloading and the return trip were not measured separately on all trips, but for some trips just the time between the departure and return at the stump area was taken. In the reduced impact operation, 56 round-trips were timed in detail and 34 trips in the conventional operation. Based on the detailed measurements, the total round-trip time and the travelled distance, the times required to travel to the landing, to unload and to return were estimated for each trip. The average time spent during a day’s work is subdivided in one out of three classes: ‘effective crew time’, ‘crew down time’ or ‘organisational delay’. The distribution of these time classes is shown in Table 5.5, together with data concerning the average load that was skidded and the average travelled distance. In felling, an important portion of the effective crew time was absorbed by work elements during which the machine was idle. Only 30% to 40% of the time were spent on operating the saw; the 169 Reduced impact logging in the tropical rain forest of Guyana operative machine time was thus short compared to the effective working time of the crew. This was not the case in the skidding operation. After having carried out daily routine servicing of the machine, the operator normally operated the machine throughout the rest of his effective working time. The operative machine time thus equals the effective crew time less the machine’s down time; i.e. time for maintenance and refuelling. Poor organisation caused delays. In the conventional operation, the skidder operator frequently had to wait for trees to be felled. In both operations, decking of logs at the landing caused delays, because of limited storage capacity. The limited storage capacity of the landings was not provoked by a substandard size of the landings, but due to the fact that long-distance transport was postponed because of organisational problems. These delays were excluded from the comparison and are mentioned separately in Table 5.5. Table 5.5 Skidding efficiency: Average effective crew time and production statistics based on the raw, recorded data of the time studies of the two logging methods Subject of observation Effective crew time (min) Operative machine time (min) Machine servicing (min) Organisational delays (min) Crew down time (min) Camp-to-camp time Travelled distance (one way; m) Trips/day Logs/day Volume/day (m3) dbh (cm) Volume/load (m3) Logs/load Logging intensity (n/ha) Conventional 260 228 32 49 82 391 1160 4.6 15.5 44.6 56 9.7 3.4 10.0 Logging method Reduced Impact 296 269 27 15 64 375 450 7.3 18.3 64.6 57 8.9 2.5 8.7 For the conventional operation, an effective crew time of 4.3 hours/day was found and 4.9 hours/day for the reduced impact operation. The reasons behind these rather low scores differ between the operations. Except for the organisational delays mentioned above, the skidding output was limited by the output of the felling crew in the conventional operation. In the reduced impact operation, where felling was completed before skidding commenced, the average operative time was cut down by frequent machine break-downs. Most break-downs led to an early return to the camp, which meant that their effect is not incorporated in the study. In both cases, other delays of organisational nature - such as waiting for fuel to arrive, waiting for a spare tyre, etc. occurred as well. Technically speaking, all such delays should be included in organisational delays in Table 5.5. The present study, however, is intentionally limited to a comparison of working methods, and not of technical or organisational differences. The difference in crew down time is related to the accessibility of the 170 Chapter 5: Logging efficiency areas, just as in case of felling. The down time is less than in felling, because the crew hardly ever took a meal-break and because the locations where the time studies of skidding took place were closer to passable passenger vehicle roads than the ones where the time studies of felling took place. Excluding time for commuting, meal-breaks and the organisational delays described above, the time required per round-trip was, on average 49.3 minutes in the conventional and 36.8 minutes in the reduced impact skidding operation. The volume delivered at the landing per operative machine hour amounted to 11.8 m3/hr and 14.4 m3/hr, respectively. The round-trip is composed of charging, loaded travel, discharging and light travel. Terrain characteristics, distance and efficiency of the trail system influence the travelling times. The time required to collect a load is influenced by the number and size of the logs that compose the load. Various factors influence an operator’s decision on how many logs are picked up: the capacity of the machine, grade and condition of the trail, the distance to the market, etc. If the logs are located at short distance from the landing, the operator may decide to transport the logs one by one, instead of combining them into a load. Because these factors differed between the operations, load sizes differed (Table 5.5). In order to eliminate the effect of this disparity, the times required for the work elements that constitute the skidding work cycle are compared in detail. The uncorrected average times of work elements involved in a round-trip are shown in Table 5.6. Table 5.6 Skidding efficiency: actual round-trip times in conventional and reduced impact operations. The average travelling distance differed considerably (Table 5.8) Work element Return Approach logs Collecting Bunching Skidding Discharging Work delays Unavoidable delays Round-trip time Mean time taken in skidding method (min) Conventional Reduced Impact 10.1 3.7 3.1 4.8 11.2 9.1 3.2 4.7 12.2 7.0 1.1 0.9 5.4 2.1 3.0 4.5 49.3 36.8 The comparison of the recorded time distribution reveals that the time to charge a load (approach, collect and bunch) did not differ much between the operations, whereas travel times and work delays were substantially longer in the conventional operation. In order to apply the same standards to both operations, multiple regression models were used to predict the skidding cycle time. The standards that were set are: a skidding distance based on the average spacing of roads and landings in the concession, a load size distribution as observed in the conventional operation (9.7 m3 on average per trip), and a maximum adverse grade of 10%. Unavoidable delays are caused by such factors as malfunctioning of the winch, changing tyres, or other minor break-downs. They delayed each round-trip by an 171 Reduced impact logging in the tropical rain forest of Guyana average 4.5 minutes in the reduced impact operation. In the conventional operation, delays of technical nature took up 2.0 minutes, whereas 1.0 minute was spent on assisting the felling crew. 5.4.2 Collecting In the conventional operation, the skidder operator was directed by his assistant to the nearest log. Although the assistant started the search for other logs during the skidder’s trip to the market, the assistant had no proper overview of the stump area and was not able to guide the operator along the shortest route. Because logs were positioned at random and because no choker straps were used, logs had to be repositioned before hooking. The skidder consistently approached a log a few metres from the butt end and used its blade to push the log into the required position. In this manner an inefficient, winding trail system was produced (Fig. 5.9). This not only affected the speed of the skidder, but also increased the risk of damage to the logs. (For details of the conventional skidding method, see Section 3.2.2) In the reduced impact operation, the crew leader was directing the skidder. He indicated the skidder where to position itself to pick up the first log. The winch line was then hauled out and the pre-choked logs attached to the choker on the winch line. The log was hauled in and skidded to an assembly point that was being indicated by the crew leader. This procedure was repeated until sufficient logs were collected to compose a load. Establishment of the trail preceded log collection. The time required for skid trail establishment is incorporated in the approach time. These (secondary) trails ran parallel - where the topography allowed this - at a spacing of 80 metres. Because trails were straight the skidder could travel at maximum speed and risk of damage to logs was minimal. (For details of the reduced impact skidding method, see Chapter 2.3.5). Approaching logs The average recorded time required to approach the logs was 3.1 minutes per load in the conventional (6% of the skidding cycle time) and 4.8 minutes per load in the reduced impact operation (13%). The 4.8 minutes recorded for reduced impact skidding includes the time spent on establishing skid trails. This amounted to 1.8 minute per load, rendering an actual log approach time of 3.0 minutes. Because log extraction was hardly ever monitored over an entire plot, the established trails may also have served the extraction of logs that were not included in the present study. The following procedure made it possible to take care of this limitation. First, the total length of the trails was measured for each plot and the number of logs that were extracted along them. This produced an average of the trail length per extracted log for each plot, varying between 6.7 and 28.3 m per log. Secondly, the relation between the time to push a trail and its length was established. The following linear regression equation was derived: (time) = –0.01 + 37.85 ⋅ (distance), where time is in minutes and distance in km (R2 = 0.90, F1,25 = 226.08***). Using this equation, the trail establishment time was determined for a particular plot, then distributed over the 172 Chapter 5: Logging efficiency number of logs in that plot. This produced an adjusted estimate for the time spent on establishing skid trails of, on average, 1.2 minute, and, hence, an adjusted time for approaching logs of 4.3 minutes. The average number of logs in a load varied between logging methods. Multiple regression analysis with the logging method, the number of logs in a load and the total volume of a load as independent variables, showed a significant interaction between the logging method and the number of logs in the load. Volume had no effect and the best fit was obtained with the following model: (approach time) = 1.46ns – 1.56 ⋅ [M] + 0.50* ⋅ (logs) + 1.24*** ⋅ [M] ⋅ (logs). [M] is a dummy variable coded 0 for conventional and 1 for reduced impact skidding (R2 = 0.46). Separate tests for the logging methods showed that the number of logs did not influence the approach time in the conventional operation (Fig. 5.5). 12 Conventional Reduced impact Approach time (min) 10 8 6 4 2 0 1 2 3 4 5 1 2 3 4 5 Number of logs per load Figure 5.5 Average time required to approach logs that compose a load, based on 60 trips in a conventional and 73 trips in a reduced impact operation in Pibiri, Central Guyana. Number of logs per load is included as a co-variate. Boxes indicate standard errors and whiskers indicate standard deviations. The number of logs had barely any effect on the time in case of the conventional operation (Kruskal Wallis ANOVA: H4,60 = 8.57, p = 0.07), whereas it had a significant effect in case of the reduced impact operation (Kruskal Wallis ANOVA: H4,73 = 51.62, p < 0.001). The only explanation is that the logs were located close to each other. Figure 5.5 also indicates that the difference between the logging methods is negligible as long as a load contained two logs or less. This shows that felling in clusters was the decisive factor. Hence, the difference in felling pattern influenced the approach time more strongly than the difference in the way that the skid trails were established. Applying the multiple regression model to a load size distribution as found in the conventional operation, an approach time of 5.3 minutes was derived for reduced impact logging against 3.1 minutes for conventional logging. 173 Reduced impact logging in the tropical rain forest of Guyana Hooking and pre-skidding The average recorded time required to collect the logs that made up a load was 11.2 minutes (23% of the cycle time) in the conventional and 9.1 minutes (25%) in the reduced impact operation. The collecting procedure differed considerably between the two methods. In conventional logging, the collection phase consisted of positioning the logs, taking 4.3 minutes on average, hooking the logs, 2.0 minutes on average, and preskidding the logs before assembling them, 4.9 minutes on average. In reduced impact logging logs were not positioned with the blade, but were winched to the skidder. Hauling out the winch line and hooking took 3.0 minutes on average, winching 2.6 minutes on average, and pre-skidding 3.5 minutes on average. Just as with approaching, the number of logs and their volume certainly play a role. The analysis considers two phases during collecting: pre-skidding and hooking. The latter comprises both positioning and hooking or hauling out the winch line, hooking and winching. Multiple regression analysis produced the following model for hooking: (hooking time) = –0.56ns – 0.30ns ⋅ [M] + 0.60ns ⋅ (logs) + 1.15* ⋅ [M] ⋅ (logs) + 0.50** ⋅ (vol.) – 0.22ns ⋅ [M] ⋅ (vol.); R2 = 0.47. [M] is a dummy for method, (logs) is the number of logs in the load and vol. is the total volume (m3) of the load. Just as with approaching of the logs, there is a significant interaction between the number of logs and the logging method. In addition, the volume of the load plays a significant role. The outcome that volume has a significant effect is not unexpected, because big logs are more difficult to hook than thin ones. Separate analyses of each method showed that the number of logs - again - did not influence the time required in the conventional operation. The volume of the load proved to have a significant effect with both methods (Fig. 5.6). Apparently, the felling patterns, in clusters or regularly spaced, had a stronger influence on hooking time than the technique that was used. The corrected time for hooking in the reduced impact operation was derived by applying the regression equation to a distribution of load sizes as observed in the conventional operation: 7.4 minutes, which stands against 6.3 minutes in the conventional operation. Pre-skidding involved all manoeuvres that were carried out to get the logs at the assembly point where the logs were bunched for the trip to the landing. Because of the felling in clusters and the unsystematic search pattern in the conventional operation, bunching of logs often took place in steps. First small bunches were made, involving unhooking, decking and re-hooking, which were later combined at the final assembly point. Some logs were not pre-skidded but transported straight to the landing; i.e. 3 logs in the conventional and 21 logs in the reduced impact operation. Multiple regression analysis showed that pre-skidding time was only significantly influenced by the number of logs that were involved: (pre-skidding time) = –2.76ns – 0.07ns ⋅ [M] + 2.71*** ⋅ (logs) – 0.69ns ⋅ [M] ⋅ (logs) – 0.16ns ⋅ (vol.) + 0.29ns ⋅ [M] ⋅ (vol.); R2 = 0.36. 174 Chapter 5: Logging efficiency Reduced impact Conventional Vol < 10 m 16 3 Vol < 10 m 3 3 Vol >= 10 m 3 12 8 Hooking time (min) 4 0 Vol >= 10 m 16 12 8 4 0 1 2 3 4 5 1 2 3 4 5 Number of logs per load Figure 5.6 Average time required to hook logs that compose a load, based on 60 trips in a conventional and 73 trips in a reduced impact operation in Pibiri, central Guyana. Number of logs and volume per load are included as co-variates. Boxes indicate standard errors and whiskers indicate standard deviations. The number of logs had no effect on the time in case of the conventional operation: (time) = –0.56ns + 0.60ns ⋅ (logs) + 0.55** ⋅ (vol.), whereas it had a significant effect in case of the reduced impact operation: (time) = –0.86ns + 1.76*** ⋅ (logs) + 0.28* ⋅ (vol.). The corrected times for pre-skidding are, hence, the same for both operations. Removing all non-significant elements from the linear model, yields the following linear model that applies to both logging methods: (pre-skidding time) = –2.32* + 2.22*** ⋅ (logs), R2 = 0.34. By applying this model to an average load of 9.7 m3 - as in the conventional operation - and hence an average load of 3.1 logs, a pre-skidding time of 4.5 minutes was computed, which stands against 4.9 minutes in the conventional operation. The time required to move the logs from the stump area to the assembly point was, as anticipated, strongly influenced by the number of logs involved. The smaller average load size in the reduced impact operation resulted in a shorter collecting time (Table 5.6). Adjustment for this difference produces the opposite result, a longer collecting time in reduced impact logging; i.e. 12.0 minutes as opposed to 11.2 minutes. Clustered felling caused the higher efficiency in conventional logging and not the technique that was used. The time required to collect logs from the stump area and moving them to the assembly point was longer than was observed by Hendrison (1990) in Suriname. There, this phase required 9.5 minutes for conventional and 5.9 minutes for controlled logging, with a Caterpillar 518 skidder (89 kW). The load in that study was 4 tonnes on average. In the State of Amazonas, Brazil, the environmentally sound harvesting 175 Reduced impact logging in the tropical rain forest of Guyana system used a ‘shuttle’ system, whereby a Caterpillar D4 crawler tractor (70 kW) transported logs to concentration points along the main skid trail, from where they were transported by a Caterpillar 518 grapple skidder. Pre-skidding required 9.3 minutes for a load of 4.8 m3. In the traditional skidding operation studied by Winkler (1997), logs were skidded straight to the landing by skidder. Recalling that the load fetched in the conventional operation was 9.7 m3 (≈ 12 tonnes), concentration of loads was more efficient in Guyana than in the two other studies, regarding the conventional as well as the reduced impact operation. The difference in engine power between the Cat 528 on the one hand and the Cat 518 and Cat D4 on the other partly explains this difference Bunching Bunching involved unhooking of pre-skidded logs, aligning the butt ends (‘decking’) and choking the load. Both logging systems used the same procedure, The use of a bull-hook versus a choker bell as a tool for hooking as the only difference. Although not anticipated, bunching took 4.7 minutes - 2.3 minutes for unhooking and decking and 2.4 minutes for choking the load - in the reduced impact operation, whereas it took only 3.2 minutes in the conventional operation - 1.4 minute for unhooking and decking and 1.8 minute for hooking the load. As discussed before, the mean number of logs per load was higher in the conventional operation. Multiple regression analysis showed an interaction between correlations of logging method and of the number of logs with bunching time. The volume of the load had no significant effect and was removed from the model: (bunching time) = 1.42ns – 3.37ns ⋅ [M] + 0.55ns ⋅ (logs) + 2.20*** ⋅ [M] ⋅ (logs); R2 = 0.48. Separate Student t-tests for each load class; i.e. number of logs per load, showed no difference between the logging methods if two logs were bunched (t24 = –1.93, p = 0.06). However, bunching took more time in reduced impact operation in case of loads of 3 logs (t33 = –2.21, p < 0.05), 4 logs (t35 = –6.85, p < 0.001), or 5 logs (t9 = –5.04, p < 0.001). Ascribing these differences to the logging method or equipment per se is not plausible (see Sections 3.2.2 and 3.3.5). Ascribing them to a difference in skill and dexterity of the operator and his assistant is more convincing. Therefore it is assumed that the corrected bunching time in the reduced impact operation will equal the time for conventional logging; i.e. 3.2 minutes, when the personal skills of the reduced impact logging operators will have become routine 5.4.3 Travelling Reduced impact logging features the planning and marking of a designed skid trail system, whereas in conventional logging trails are opened after felling (see Sections 3.2 and 3.3). In both cases, the skidder itself is used to open trails. The first passage of a skidder results in a narrow track with little impact on soil and vegetation. The ground tractive conditions are still good and a coefficient of traction (grip of the tyres on the ground surface) of 0.55 applies under these conditions (Caterpillar 1986). The coefficient of traction together with the gross vehicle weight determine the usable pull for a machine; i.e. any excess rim pull above this usable pull is not converted into 176 Chapter 5: Logging efficiency useful work but serves only to spin the wheels. After repeated trips the topsoil is being smeared and compacted and ruts are formed in the trails. Thereby the coefficient of traction is reduced to 0.40 (Caterpillar 1986) and, hence, the maximum load is reduced. For this reason, it is not advisable to construct trails by crawler tractor. A crawler tractor has a wider blade, a larger weight and is capable of a higher tractive effort. Trails constructed this way will be wider, and the soil will be compacted and smeared before any log has been extracted, and hence the maximum load will be reduced, making skid trail construction by crawler tractor indeed counterproductive. The skidding distance is one of the most important variables determining the performance of an operation. The skidding distance - as the crow flies - is determined by the spacing of truck roads, the distance between landings and the terrain conditions. The optimum economical road spacing can be deduced from road construction cost and the skidding cost per unit output (Sundberg & Silversides 1988; FAO 1978). The wider the spacing of the roads, the longer are the off-road transport times and the larger is the area that each section of a road serves. Widely spaced roads hence lead to higher skidding cost and stronger ecological impact, but lower road construction cost. Light and loaded trips were monitored during the extraction of two conventionally logged units and of several experimentally logged plots. Only the trips from one conventionally logged unit were included in the study. The trail leading from the edge of the other unit to the landing followed a straight, cut line, and was, therefore, not considered representative of conventional logging in general. In order to eliminate differences in terrain conditions as much as possible, trips in the reduced impact operation were selected only if the terrain conditions corresponded to the one in the conventionally logged unit. This resulted in the selection of 19 light, 25 loaded trips for conventional and 38 light, 45 loaded trips for reduced impact logging. Loaded travel Besides skidding distance and terrain conditions, the skidder’s payload affects the travelling time of the loaded skidder. Travel time of the loaded skidder in the conventional operation was 7.9 minutes with an average travelled distance of 659 m and an average load of 9.9 m3. In the reduced impact operation with an average distance of 260 m and an average load of 7.0 m3, the loaded trip took 3.2 minutes on average. Due to the large difference in load and distance, these figures are not very informative. Multiple regression analysis produced the following equation: (travel time) = 0.16ns + 0.07ns [M] + 0.35** (vol.) – 0.29* [M] (vol.) + 6.49*** (distance) + 3.43ns [M] (distance); R2 = 0.88, F5,64 = 92.40; [M] is a dummy variable coded 1 in case of reduced impact and 0 in case of conventional logging, (vol.) in m3 and (distance) in km. There was no significant interaction between the correlation of logging method and of distance with the travelling time. After removing this variable from the model, the interaction between logging method and volume did not appear to be significant either. The final model hence does not involve any interaction terms: (travel time) = 0.87ns – 0.90* [M] + 0.15* (vol.) + 8.43*** (distance); R2 = 0.87, F3,66 = 142.50, (vol.) in m3 and (distance) in km (Fig. 5.7). 177 Reduced impact logging in the tropical rain forest of Guyana Having established that the logging method has a significant effect on the travel time, one would actually not expect this increase in travelling time to be constant with increasing distance as indicated by the final model. The difference is explained by the fact that trails are more tortuous for conventional skidding, resulting in a lower travelling speed. If this difference in speed applies along the entire trail, one would in fact expect an interaction between the effects of the logging method and the travelled distance on the required time. Assuming a travelled distance of 500 m and a load size of 9.7 m3 for both logging methods, the travel time of the loaded skidder becomes 6.5 minutes for conventional and 5.6 minutes for reduced impact logging. This corresponds to respective speeds of 4.6 km⋅h-1 and 5.4 km⋅h-1. Figure 5.7 Travel time for the loaded skidder trips in a conventional and a reduced impact operation in Pibiri, central Guyana. Relationship between skidding distance, payload, logging method and travel time. Extrapolation showed that the time increased when distance or volume was increased and is higher in case of conventional logging (CNV) than in case of reduced impact logging (RIL). Reduced impact logging featured mainly light loads and short distances whereas the opposite applied to conventional logging. This implies that the results should be interpreted cautiously. See text for explanation of the regression lines. Skidder performance can also be estimated using machine specifications and rules of thumb (Caterpillar 1986). A skidder’s travel speed, hence travel time, can be determined from rim pull graphs specific for a particular type of machine. The usable rim pull depends on traction factors, tyre size and equipped weight of the loaded skidder. The required rim pull is the pull needed to overcome the forces of resistance that act upon the machine. The total resistance comprises grade, rolling and skidding resistance forces (Caterpillar 1986; FAO 1977; Silversides & Sundberg 1989). Grade resistance depends on the per cent grade, gravity force and the gross vehicle weight. Rolling resistance depends on the firmness and smoothness of the surface and tyre penetration. Skidding resistance is calculated per log and determined by the weight of the log and grade. The gross vehicle weight is determined by the operating weight of the machine, weight of average log, the average number of logs and the percentage of the weight of the load that is transferred to the machine. 178 Chapter 5: Logging efficiency If the required rim pull exceeds the usable pull due to traction limitations, the load must be reduced. The required rim pull for a Cat 528 cable-arch skidder for a series of grades and numbers of logs is shown in Table 5.7. The required rim pull for a load size of three logs would vary between 5.2 and 8.4 tonnes depending on the slope of a skid trail section. The theoretical travelling speed is read from a machine-specific rim pull graph (Caterpillar 1986) and varies between 4 and 6 km/h, according to the slope of the terrain. Table 5.7 Required rim pull and usable pull for four different terrain conditions and four different loads for a Caterpillar 528 cable-arch skidder. The results are based on a machine operating weight of 13.2 t, a 50% transfer of load weight to the machine, a gross log volume of 2.9 m3, a wood density of 1200 kg/m3 (the green density of Chlorocardium rodiei is 1300 kg/m3 (Gérard et al. 1996), but allowance is made for bark) and 15 cm average tyre penetration. A fresh trail has a traction coefficient of 0.55 and a rutted one of 0.40 (Caterpillar 1986) Adverse grade (%) 0 5 10 20 Traction factors Fresh trail Rutted trail Required rim pull (t) in relation to number of logs (1 log = 3.5 t) 1 2 3 4 2.7 4.0 5.2 6.5 3.7 5.2 6.8 8.4 4.6 6.5 8.4 10.2 6.6 9.0 11.5 14.0 8.2 6.0 9.2 6.7 Usable pull (t) 10.1 7.4 11.1 8.1 These values agree with the empirical values found with the model, and indicate that operator and machine performed according to standard. Table 5.7 also shows that the payload chosen by the operator was close to the ceiling. If slopes steeper than 10% would have occurred on the trail, which they did not, he would have had to drop at least one log. The effect of the rutting of the trail on the maximum load size is also clear. On flat terrain, rutting does not pose any limitations on the number of logs that can be hauled in one go. If trail sections with a 5% adverse grade were encountered, rutting would bring the maximum load down to three logs, in case of 10% adverse grade to only one log, and in case of 20% a new trail would have to be established. In skid trail planning, slopes of over 20% thus should be completely avoided. It is recommendable to use an upper limit of 10% for adverse grades on main trails. Else, bypasses will be needed after some trips. Light travel Travel time of the unloaded skidder from the landing to the forest is mainly determined by the rated power of the machine, the distance and terrain conditions. Regarding the samples with comparable terrain conditions, travel time of the unloaded skidder was 6.7 minutes in the conventional operation with an average travelled distance of 671 m and 2.6 minutes in the reduced impact operation with a distance of 290 m. Correcting for the difference in distance by multiple regression analysis, produced the following equation: (travel time) = 2.52* – 2.70* [M] + 6.16*** (distance) + 3.55ns [M] 179 Reduced impact logging in the tropical rain forest of Guyana (distance); R2 = 0.79, F3,53 = 64.67. [M] is a dummy variable coded 1 in case of reduced impact and 0 in case of conventional logging; (distance) in km. The interaction between the logging method and the distance is not significant. After removing this variable from the model, it appeared that the travel time did not differ significantly between the logging methods: (travel time) = 0.81ns – 0.70ns [M] + 8.72*** (distance); R2 = 0.77, F2,54 = 91.89 (Fig. 5.8). Apparently, the greater tortuosity of the conventional trails affected the travel time of the loaded skidder, but not of the unloaded skidder. Indeed, it is well–known that the sweeping logs behind the skidder are the major cause of increased travel time. Figure 5.8 Travel time for the unloaded skidder in a conventional and a reduced impact operation in Pibiri, central Guyana. Relationship between skidding distance, logging method and travel time. The time increased when distance was increased but is not significantly influenced by the logging method. Reduced impact logging featured mainly short distances whereas the opposite applied to conventional logging. This implies that the results should be interpreted with caution. See text for explanation of the regression lines. It should be noted that the technical state of the machine used in the conventional operation was better than of the one used in the reduced impact operation (see Section 5.2.7). The effect this may have had on the results of the time study is unknown and so could not be discounted for. This inequality, the small number of observations and differences in soil type and skidding distances limit the validity of the results. A larger scale time study is evidently required. Assuming a travelled distance of 500 m for both logging methods, the travel time of the unloaded skidder becomes 4.8 minutes for both methods, corresponding to a speed of 6.3 km/h. Theoretically, a Cat 528 skidder has a top speed of 31 km/h (Caterpillar 1986), but this speed is only achieved on flat, paved roads. In the forest, the unloaded machine has to overcome grade and rolling resistance forces. Applying the same tyre penetration as in Table 5.7, produced a required rim pull of 1.5 tonnes on flat terrain, 180 Chapter 5: Logging efficiency and 2.8 tonnes with an adverse grade of 10%. These values correspond to speeds of 20 and 8 km/h, respectively. The speed of the unloaded skidder in the present study thus was comparatively low. This can be explained by the fact that trails were not straight and were partly covered with debris. Pre-constructing skid trails by crawler tractor indeed may have a negative effect on skidding production due to load size restrictions (see above), but this is probably compensated for by faster return trips. The ecological damage, however, is certainly higher. Travelled distance In Table 5.5 it was shown that the average travelling distance was much longer in the conventional than in the reduced impact operation that was monitored, which resulted in longer travel times (Table 5.9). In order to get a representative picture of skidding cost a model approach was followed. The road density in the DTL concession amounted to 1 km for 112 ha in 1987 (Klassen 1987), which represents a theoretical parallel road spacing of 1.12 km. It is assumed that the compartments that have been laid open after that date, feature a similar road density. Field observations led to the conclusion that landings lie 1 km from each other on average. For the model, road and landing spacing are assumed to be both 1 km. Using Peters’ (1978) equation, an average skidding distance (as the crow flies) of 383 m is obtained. In order to estimate the average travelled distance, this distance should be multiplied with the winding coefficient (see Section 5.2.7). This coefficient can vary between 1.1 and 1.6, depending on the efficiency of the trail system and terrain conditions (Sundberg & Silversides 1988). The travelled distance hence could vary between 420 and 610 m. The skid trail pattern created in conventional logging is a geometric network (Fig. 5.9 top), whereas the pattern created in reduced impact logging is a dendritic pattern. The latter is ramified whereas trails are joined by ‘knotting’ in the former. Geometrically speaking, there is only one way to get from point A to point B in a dendritic pattern, whereas there are more options in case of a network. The travelling distance (time) is thus determined by the lay-out of the pattern in case of a dendritic pattern and by the chosen route in case of a network. In theory, the skidder operator could always opt for the shortest route possible in case of a network. In practice this does not happen and the operator returns to the same ‘main’ trail all the time, for reason of convenience (no opening of new trails) and because he lacks a proper overview of the stump area. The actually travelled distance and as the crow flies distance to the point of exit of a plot was measured from digitised maps for 48 trips in the experimental plots. The same procedure was followed for the conventionally logged units where the distances for 77 trips were measured (Fig. 5.9). The degree to which the trails deviated from a straight line, expressed as the winding coefficient, was 1.04 (sd = 0.4) for reduced impact and 1.25 (sd = 0.13) for conventional logging. 181 Reduced impact logging in the tropical rain forest of Guyana Figure 5.9 Skid trail patterns formed in conventional (above) and reduced impact logging (below).The map shown for the conventional operation covers about half the area of one of the monitored logging units. For reduced impact logging, parallel trails were used in flat to rolling terrain (left) and forked trails in dissected terrain (right). The degree to which trails deviate from the straight-line distance was determined by measuring actually travelled distance in comparison to the distance as the crow flies from log concentration points to the point of entry of a unit. Since logs are transported to assembly points along the truck road (landings) that are assumed to be 1 km apart, the trails will have to meet either at the landing or join each other on the way to the landing. Three geometric trail patterns have been worked out; i.e. parallel trails, trails that join in angles of 45° and trails that join in angles of 30° (Fig. 5.10). The deviation from the straight-line distance appears to be within a range of 1.02 to 1.31. Assuming a 45° branching pattern and making allowance for deviations due to terrain conditions, a winding coefficient of 1.1 is applied to both conventional and reduced impact logging. Including the tortuosity of the trails 182 Chapter 5: Logging efficiency themselves, an overall coefficient of 1.38 is derived for conventional and of 1.14 for reduced impact logging. According to the model, the average travelled distance for a 1 km by 1 km drainage area (road passing through the centre) hence is 527 m for conventional and 436 m for reduced impact logging. Figure 5.10 Dendritic skid trail patterns for drainage areas measuring 1 km by 1 km (hauling to landing on both sides of the road) and a skid trail spacing of 100 m. From left to right patterns based on a 45°, a 90° and on a 30° ramification pattern. The winding coefficient was determined using a grid of 36 points, evenly distributed, and measuring distances as the crow flies and distances along the trails. Parallel (90° ramification) trails feature a winding coefficient of 1.31, 45° trails of 1.05 and 30° trails of 1.02. Total length of trails is smallest with parallel trails and increases with narrower angles. Obviously, terrain conditions force deviations from these geometric patterns, so the actual coefficient will be higher. Accommodating the effect of the efficiency of the trail system will affect both light and loaded travel times. For a travelling distance of 500 m, a loaded travel time of 6.5 minutes was obtained for conventional logging and 5.6 minutes for reduced impact logging, whereas light travel time amounted to 4.8 minutes in both cases. Assuming an average distance as the crow flies of 383 m and providing for differences in trail system efficiency, the following travel times are obtained. Loaded travel takes 6.7 minutes for conventional logging and 5.1 minutes for reduced impact logging. Light travel takes 5.0 minutes for conventional and 4.2 minutes for reduced impact logging. 5.4.4 Other operative work elements Two other work elements may be influenced by the logging method. These are discharging of the load at the landing and (avoidable) work delays. 183 Reduced impact logging in the tropical rain forest of Guyana Discharging Discharging of the load at the landing followed the same procedure in both operations. The skidder positioned the new load alongside the logs that were discharged before. The load was subsequently unhooked and the skidder aligned the logs with its blade. Stacking of logs occurred frequently, but is considered no routine activity and so is included in the organisational delays. Discharging consumed 1.1 minute per trip on average (2% of the cycle time) in the conventional and 0.9 minute (2%) in the reduced impact operation. In order to correct for different load compositions, multiple regression analysis was used. The following equation was derived: (discharging time) = 0.87ns – 0.81ns ⋅ [M] – 0.06ns ⋅ (logs) + 0.18ns ⋅ [M] ⋅ (logs) + 0.04ns ⋅ (vol.) + 0.05ns ⋅ [M] ⋅ (vol.); R2 = 0.11, F5,77 = 3.11. After removing the non-significant interaction terms, it appeared that the non-significant interaction term concealed that discharging time was significantly related to load volume: (discharging time) = 0.34ns + 0.08** ⋅ (vol.); R2 = 0.12, F1,81 = 10.91. Since the load is fixed at 9.7 m3 in both cases, a discharging time of 1.1 minute is applied to both methods. Work delays Short delays that occur during the skidding cycle, that are not caused by differences in organisation (see Sections 5.2.6 and 5.3.1), and thus form part of the effective machine-crew time, are referred to as work delays. Work delays played a much smaller role in skidding than in felling, because operating the skidder is less exhausting than operating the chain saw. Most work delays occurred during the log collection phase and especially during hooking. In the reduced impact operation, short delays occurred because the choker strap was not set yet. It occasionally happened that a (heavy) log had to be repositioned with the skidder blade, either because the nubbin of the choker strap could not pass underneath, or because it was too heavy to winch. In the conventional operation, delays occurred either because the skidder had approached the log in an awkward way and had to re-approach it from another point, or because the hook could not pass underneath the log and logs had to be repositioned. Such work delays absorbed significantly more time in the conventional operation; 5.4 minutes per trip, than in the reduced impact operation; 2.1 minutes per trip in the reduced impact operation (Mann-Whitney U-test; U = 670, p < 0.001). These values will be used in the skidding model. 5.4.5 Supportive work elements and delays Servicing and repairs Servicing of the skidder includes refuelling and regular checks of oil, bearings, joints, etc. It required 32 minutes per day in the conventional and 27 minutes in the reduced impact operation. Minor repairs occurred because certain parts of the machine malfunctioned. Since two machines were used of the same make and model, no differences should occur in this respect. The time consumed by these work elements was fixed for both methods at the proportion of the daily operative machine time as found with the conventional operation; minor repairs 1.3 minute and servicing 5.7 minutes per operative machine hour. 184 Chapter 5: Logging efficiency Crew down time and delays Crew down time differed between the two operations (Table 5.8). This difference is not related to the logging method. For the skidding model, a crew down time of 82 minutes was applied for both methods. Organisational delays and delays due to natural causes are bound to occur during a day’s work. This could involve changing tyres, occasional stacking of logs at the landing, waiting for heavy rain to stop, etc. It was assumed that these delays are related to the operative machine time. For both methods, these were set at the time for ‘normal’ delays as found in the conventional operation: 2.4 minutes per operative machine hour. Assisting the felling crew in freeing hang-ups was set at 22.5 minutes per day. The time spent on these delays per skidding cycle was deducted by using the Microsoft Excel Solver algorithm. 5.4.6 Skidding time model Skidding cycle The regression analyses described in the previous sections were used to build a model of the time required for the extraction of a load under ‘identical’ circumstances. The time taken by most work elements results from projections of the load size distribution found in the conventional operation. In one case this involved the number of logs in a load; i.e. approaching the logs in the stump area; in another case the volume of the load; i.e. discharging; and in still another case both the volume and the number of logs; i.e. hooking. Both loaded and unloaded travel are related to the travelled distance from stump area to the roadside landing. Two scenarios have been worked out: one in which an average travelled distance of 500 m applies and one in which the difference in travelled distance, indicated by the higher winding coefficient of conventional trails, applies. The corrected time study of the skidding work cycle shows that implementation of the reduced impact logging system reduced the time required to extract a load by 5%, as long as the efficiency of the trail system is not considered. If this is included the cycle time is reduced by 9% (Table 5.8). Table 5.8 Skidding efficiency: modelled round-trip times in conventional and reduced impact operations for a 1 × 1 km2 stumpage area with a standard load of 9.7 m3 per trip. This volume per trip corresponds to 3.4 logs in conventional and 3.1 logs in reduced impact logging. Two scenarios are defined: one ignoring difference in tortuosity of the trails between methods (average travelled distance 500 m) and one incorporating this difference (average distance as the crow flies: 383 m) Work element Return Approach logs Collecting Bunching Skidding Discharging Work delays Mean time taken in skidding method (min) exc. trail efficiency incl. trail efficiency Conventional Reduced Impact Conventional Reduced Impact 4.8 4.8 5.0 4.2 3.1 5.3 3.1 5.3 11.2 12.0 11.2 12.0 3.2 3.2 3.2 3.2 6.5 5.6 6.7 5.1 1.1 1.1 1.1 1.1 5.3 1.9 5.3 1.9 185 Reduced impact logging in the tropical rain forest of Guyana Unavoidable delays Round-trip time 4.5 39.7 4.0 37.8 4.6 40.2 3.9 36.5 The basic travel time is shorter in reduced impact logging, but this is counteracted by a longer (dis-)charging time: 21.5 as opposed 18.6 minutes. The latter is not so much related to the technique used as to the fact that trees were felled in clusters in the conventional operation. If the difference in approaching and hooking time is neglected, the conventional skidding cycle time becomes 44.9 minutes. In that case, reduced impact logging reduces the skidding cycle time by 19%. These reductions will become more pronounced over larger skidding distances as the crow flies, because the collecting time is a relatively fixed component of the cycle time, whereas the basic travel time is a variable component. Output Based on the time involved in a single round-trip, the output of ‘a fair day’s work’ is estimated (Table 5.9). This estimate is based upon an 8 hours working day, a commuting time as found in the reduced impact operation and a suggested lunch-break of 30 minutes. The total effective crew time then covers the remaining hours. The number of round-trips per day was determined by comparing the operative machine time with the time involved in a single round-trip (Table 5.8). Servicing was fixed at 30 minutes per day. It appears that the implementation of reduced impact logging increased skidding output by 5%, if the difference in the efficiency of the trail system is ignored (Table 5.9). Including this difference results in an increase in output by 10%. Collecting time was influenced by the felling pattern. Adjusting for this difference, by assuming similar times for approaching and hooking of logs in both operations, would result in skidding output of 12.9 m3 per operative machine hour for conventional logging. Compared to this figure, implementation of reduced impact logging, would result in an increase in skidding output by 23%. The way in which the different scenarios affect the cost of skidding will be examined Section 5.5. 186 Chapter 5: Logging efficiency Table 5.9 Skidding efficiency: effective crew time and production statistics based on the skidding time models of the two logging methods. Production figures apply to a stumpage area of 1 × 1 km2 area (extracting from two sides of the road), flat to rolling terrain and a logging intensity of 10 trees/ha. Subject of observation Effective crew time (min) Operative machine time (min) Machine servicing (min) Crew down time (min) Camp-to-camp time exc. Trail efficiency incl. Trail efficiency Logging method Conventional Reduced Impact Conventional Reduced Impact 398 398 398 398 368 368 368 368 40 40 40 40 82 82 82 82 480 480 480 480 Travelled distance (one way; m) Volume/day (m3) Volume/effect. machine hour (m3) Volume/load (m3) Logs/load 5.5 500 87 14.6 9.7 3.4 500 94 15.3 9.7 3.1 527 84 14.4 9.7 3.4 436 97 15.9 9.7 3.1 COST OF OPERATION The foregoing sections described how the hourly output of felling was strongly reduced by employing a directional felling method, which, however, benefited the performance of the skidding operation, of which the hourly output was increased. Performance and inputs expressed in time units, however, are meaningless at different cost of the inputs per time unit. Therefore, inputs should be brought over a common denominator, which per definition is the monetary value attached to the input in time units. Moreover, the comparison is not between a method of felling and a method of skidding, but rather between two logging systems. A logging system includes all the operations that are involved in reducing a standing tree into logs and the delivery of these at the final terminal. Essential is, that the various sub-operations are sequentially related, each depending upon the next one or the one before; i.e. an earlier operation influences a subsequent one, and similarly, a later operation may require some preceding jobs to have been carried out in a particular way (Sundberg & Silversides 1988; Silversides & Sundberg 1989). For example, the decision to fell a tree in a certain direction, can only be made after the inventory data have been processed and a skid trail alignment has been planned. The concept then is to consider the cost-effectiveness of the entire string of operations and not to consider the cost-effectiveness of each sub-operation in isolation. 5.5.1 Planning costs In contrast to other studies comparing conventional and reduced impact logging, where the conventional logging is usually performed in an unplanned, haphazard manner (e.g. Hendrison 1990; Winkler 1997; Holmes et al. 1999) - and as such are not logging 187 Reduced impact logging in the tropical rain forest of Guyana systems in the proper sense of the word - , the conventional logging system used as a reference in the present study featured planning and preparation phases. The critical distinction between reduced impact logging and conventional logging in the present study is expressed by the detail and accuracy of the planning operations that were carried out. A higher degree of accuracy and additional activities are required for the accurate execution of subsequent operations, such as felling and skidding, so that the objectives of the logging system are met. Pre-harvest planning activities should begin at least 12 months prior to harvesting and include block layout, inventory, road planning, data processing and map making in both logging systems. In the reduced impact logging system, liana cutting is added. The main difference with the conventional system is that slope correction is applied during line cutting and that tree locations are mapped with a higher precision and at a larger scale. Table 5.10 shows the cost per hectare of pre-harvest planning for the conventional and the reduced impact system. For a detailed overview of the source of these costs see Tables C.3 and C.4 in Appendix C. Planning costs are incurred one year prior to harvesting and in principle the net present value (NPV) of these costs should be considered. Applying an interest rate of 16% (see Section 5.2), the NPV becomes US$ 5.90/ha and US$ 18.00 for conventional and reduced impact logging respectively. In total, pre-harvest planning cost of reduced impact logging exceeds conventional logging cost by $ 10.46/hectare (NPV = $ 12.15/ha). Harvesting operations hence must be more cost-effective in reduced impact logging, at least so much that this cost is recovered. Looking at this cost with a perspective to the final returns of the operation, clearly increasing the logging intensity can reduce the weight of this up-front-cost. Table 5.10 Pre-harvest planning cost for conventional and reduced impact logging in incurred one year prior to harvesting. Activity Block lay-out Inventory Liana cutting Data processing Map making Road planning Total US$/ha. Costs are Cost (US$/ha) Conventional 2.08 2.45 0.00 0.23 0.10 0.20 5.06 Reduced Impact 2.80 7.66 3.99 0.46 0.43 0.20 15.52 The reliability of these estimates was checked by comparing them with planning costs reported elsewhere. In such a comparison it is essential to compare costs per hectare and not per m3, because logging intensities may vary substantially. Landell-Mills (1997) reported that the cost of inventory for three logging companies in Guyana in 1996 - assumed to include line cutting, inventory, and data processing, but to exclude road planning - varied between $ 0.17 and $ 1.46 per m3. Flaming (1995) reported a cost of $ 5.81/m3 for DTL in 1995, while DTL (pers.comm.) itself recently reported a cost of $ 2.03/m3 for October 1998. Assuming a logging intensity of 5 m3/ha (Flaming 1995; DTL, pers.comm.; see also Section 5.6), this translates to a cost per hectare 188 Chapter 5: Logging efficiency between $ 0.86 and $ 10.14 ($ 29.06). The estimated planning cost for conventional logging obtained in the present study, falls within this range and appears to be quite reliable. For reduced impact logging, we have to resort to international comparisons. International comparisons should be handled with care, because differences in wage rates could strongly affect the cost of planning activities, since these consist for nearly 100% of manual labour. In a study comparing reduced impact with conventional logging in Paragominas, Brazil, Holmes et al. (1999) reported a planning cost (road planning excluded) of $ 29.10/ha. Because they state monthly cost of labour, their results can be adjusted for differences in wage rates. The total monthly cost of a tree spotter in Brazil being 194% of that of a tree spotter in Guyana and of a line cutter in Brazil being 184% of a line cutter in Guyana, this figure appears to match the cost found for reduced impact logging in the present study. Reporting on a financial study of reduced impact logging in La Mayronga, Ecuador, Montenegro (1996), found planning cost to come up to $ 38.23/ha. Montenegro (1996) and Holmes et al. (1999) also stated the output per crew per day (hour) for forest inventory. For La Mayronga an output of 11 ha/day and for Paragominas an output of 1.4 ha/hr is given. Both values appear to agree with the estimate of 10 ha/day for reduced impact logging in the present study (see Appendix C, Table C.4). 5.5.2 Harvest preparation costs Harvest preparation costs include the cost of road and landing construction and skid trail demarcation. In other studies on reduced impact logging (e.g. Montenegro 1996; Holmes et al. 1999) and in fact since over a century in Central European selection silviculture (Leibundgut 1984), tree marking is mentioned as an essential pre-harvest activity. Such activity was not carried out separately in the two logging systems studied here. In the conventional operation, all trees that were tagged and mapped during the inventory were intended to be felled. In the reduced impact operation, all trees with dbh ≥ 20 cm were tagged, mapped and classified as to species and stem quality. Trees to be felled were selected in the office and indicated as such on the stock map. The situation occurred where trees were identified as being harvestable by the tree spotters, but were condemned by the feller, because either he anticipated interior defects, or found previously unnoticed visible defects in a tree. The tree could also be considered to be located in an unmanageable position (for felling or for skidding). When this situation occurred in the conventional operation, sometimes an unmarked but harvestable tree (overlooked during the inventory) was selected instead. In the reduced impact operation, where, inherent to the intensity of the inventory, supernumerary harvestable trees were tagged as well, the felling crew selected a replacement from a list of possible substitutions. Tree marking as a separate activity was thus rendered superfluous. Road construction should start at least 6 months prior to harvesting in order for the roadbed to settle. Road construction practice was not considered to be in need of any improvement and no differences occur between the logging systems in this respect. 189 Reduced impact logging in the tropical rain forest of Guyana The only difference with the conventional system is that skid trails are demarcated during this phase in reduced impact logging. Table 5.11 shows the cost per hectare of the harvest preparation for the conventional and the reduced impact system. For a detailed overview of the source of these costs see Tables C.3 and C.4 in Appendix C. Harvest preparation costs are incurred six months prior to harvesting and in principle the net present value (NPV) of these costs should be considered. Applying an interest rate of 16% (see Section 5.2), the NPV becomes $ 16.02/ha and $ 18.57/ha for conventional and reduced impact logging respectively. Table 5.11 Cost of harvest preparation for conventional and reduced impact logging in US$/ha. The road density of primary roads is 2 m/ha and of secondary roads 10 m/ha. The area occupied by landings amounts to 50 m2/ha. Costs are incurred six months prior to harvesting. Activity Cost (US$/ha) Conventional Reduced Impact 6.42 4.61 3.84 6.42 4.61 3.84 2.37 17.24 Road construction Primary road Secondary road Landing construction Skid trail demarcation Total 14.87 Note: road and landing construction cost reported here is lower than reported for Guyana elsewhere (e.g. Flaming 1995; Landell-Mills 1997). This may be due to machine operating costs having been estimated according to handbook guidelines (FAO 1977 1978; Caterpillar 1986) and may underestimate real cost. Output was estimated using rules of thumb (Caterpillar 1986) and therefore may overestimate performance. In total, harvest preparation cost of reduced impact logging exceeds conventional logging cost by $ 2.37/hectare (NPV = $ 2.54/ha). Reduced impact logging thus increases up-front-cost by $ 12.83/ha (NPV = $ 14.69/ha), which have to be recovered by more efficient to harvesting operations.The validity of the road construction estimates is hard to check, because road construction cost vary with the width of clearance, difficulty of the terrain, carrying capacity of the soil, the number of bridges and culverts needed, and the applied road building standard. Consequently, stated costs per m3 are little informative. Besides logging intensity, road density starts to play a role. Even when the values of these variables are known, it is important that the cost of a certain total road length is distributed over the total volume yield of the area that has been opened up by these roads. This is more complex than it seems, because the total cost is accumulating over the years (months) by yearly (monthly) costs which are unlikely to be the same, and, therefore, not directly proportional to the yearly (monthly) volume yield. For comparison with other studies, the cost per ha was expressed in cost per linear m of road. This produced the following values: $ 3.21/m for primary roads and $ 0.46/m for secondary roads in the present study. In a study using deductive methods of logging costs of eight logging operations in Guyana, Landell-Mills (1997) found road construction costs to vary between $ 2.72 and $ 45.34/m3. In her study a yield per km of road of 2,000 m3/km was assumed for primary roads, of 1,000 m3/km for secondary 190 Chapter 5: Logging efficiency roads, and an output of 3 m and 5 m per effective crew hour respectively. The cost range of $ 2.72 to $ 45.15/m3 thus corresponded to a range of $ 2.48 to $ 41.21 for each linear metre of primary road and a range of $ 1.48 to $ 24.73 for each linear metre of secondary road, provided that output and road density are the same for each operation. FAO (1974) reported a cost of $ 3.70/m for primary roads and of $ 1.80/m for secondary roads in Ivory Coast. Assuming an inflation rate of 5%, this would translate to $ 7.50/m for primary roads and $ 3.65/m for secondary roads in 1998. Montenegro (1996) reported a road construction output of 27 m/hr for primary roads and 32 m/hr for secondary roads for reduced impact logging in La Mayronga, Ecuador. The estimated output in the present study was 73 m/hr for primary and 196 m/hr for secondary roads (see Appendix C, Table C.4). This indicates that the output was about three times lower for primary and six times lower for secondary roads than in Ecuador. These references suggest that the road construction estimates in the present study are comparatively low. Conversely, there are also reports that state a cost equal to or lower than the estimates in the present study. For example, Klassen (1987) stated a road construction cost (primary and secondary roads together) of G$ 9,611/km for the DTL concession in 1987. Correcting for changes in exchange rate (1987 exchange rate: 10.50 G$ = 1 US$) and for inflation (inflation on machines: US Consumer Price Index 1987-1997 = 3.6%; inflation on labour: estimated Guyana inflation rate 1987-1997 = 16%), this translates to $ 1.44/m for 1998. Holmes et al. (1999) reported extremely low road construction cost for secondary roads in Paragominas in 1996; i.e. $ 0.33/m for roads constructed in conventional logging and $ 0.23/m for roads constructed in reduced impact logging. The authors explain the latter difference by the greater amount of area cleared in the conventional operation, due to the omission of road planning, and the inefficient use of machines. The figures of Holmes et al. (1999) correspond closely to the estimates in the present study, however, when the difference in hourly machine cost is taken into account. In their study, a CAT D6 was used with an hourly machine operating cost of $ 34.06/hr, which is less than half of the estimated operating cost of the CAT D8 used in the present study (see Appendix C, Table C.2). This is further confirmed by the output reported by Holmes et al. (1999); i.e. 200 m/hr for reduced impact logging, which matches the output of 196 m/hr used in the present study for secondary roads. In summary, it can be concluded that the cost of road construction can vary enormously; i.e. for a secondary road, from as low as $ 0.23/m (Holmes et al. 1999) to as high as $ 24.73/m (Landell-Mills 1997) for a primary road. This variation in cost is mainly related to the output per hour, which varied between 5 m/hr to 200 m/hr for secondary roads. The estimates used in the present study are on the lower end of the reported range. Although the results in absolute terms are acceptable, their validity is limited. The same holds for road construction costs reported elsewhere for Guyana based on deductive methods. A detailed study of this phase of the operation is urgently needed. 191 Reduced impact logging in the tropical rain forest of Guyana Few references were found concerning the cost of skid trail demarcation. Montenegro (1996) reported a cost of $ 2.02/ha at an output of 49 ha/day, whereas Holmes et al. (1999) reported a cost of $ 7.80/ha at an output of 1.3 ha/hr. The estimate of 8 ha/day used in the present study (Appendix C, Table C.4) corresponds to what was found in Paragominas. 5.5.3 Harvesting cost and output Based on studies elsewhere, it was anticipated that the additional cost associated with planning and directional felling would be offset by an increased efficiency of the actual extraction - skidding - operation. In Section 5.3, it was shown that hourly felling output was reduced by 37% by implementing the reduced impact felling method. Hourly cost of felling and skidding differ widely, because of the large gap between machine operating cost and labour cost in Guyana. The effect of this difference on the aggregate cost of the harvesting phase can be read in Table 5.12. Table 5.12 Cost of felling, skidding and landing operations for conventional and reduced impact logging in US$/m3 and output expressed as m3 per effective crew hour (felling and landing operations) or effective machine hour (skidding). Figures apply to a drainage area of 1 km by 1 km (extracting from two sides of the road), flat to rolling terrain, a logging intensity of 10 trees/ha and a standard load of 9.7 m3 per trip. The average straight-line skidding distance is 383 m. Activity Felling and bucking Skidding Landing operations Total Conventional Cost (US$/m3) Output (m3/hr.) 0.60 10.6 4.30 14.4 0.34 19.3 5.24 Reduced Impact Cost (US$/m3) Output (m3/hr.) 1.16 6.7 4.10 15.9 0.32 20.9 5.58 Note: costs of felling, skidding and landing operations reported here are lower than reported for Guyana elsewhere (e.g. Flaming 1995; Landell-Mills 1997). This may be due to machine operating costs having been estimated according to handbook guidelines (FAO 1977, 1978; Caterpillar 1986) and therefore may be underestimating real cost (see text for details). Output was based on a logging intensity of 10 trees/ha. This is higher than what is usually reported for Guyana (2 trees/ha), and therefore may be overestimating actual performance. Finally, the skidding distance in other operations may be substantially longer. This table shows that the cost of felling and bucking per m3 increased nearly twofold in reduced impact logging as compared to conventional logging. This is caused by an hourly output that was 37% lower at an hourly cost that was 21% higher. The cost of skidding per m3 was only 5% lower for reduced impact logging than for conventional logging. The hourly skidder output had increased by 10%, but the hourly cost as well, by 5%. The latter was caused by expanding the number of workers in the skidding team. The increased wood recovery in the reduced impact operation was assumed to enhance the hourly output of trimming of logs at the landing as well, whereas the efficiency of scaling was not assumed to be enhanced. Overall, the cost of landing operations was thus reduced by 5%. (See Appendix C, Tables C.3 and C.4 for details). Contrary to what was anticipated based on other studies, reduced impact logging augmented the cost of harvesting by 7%. 192 Chapter 5: Logging efficiency The estimated costs of felling, skidding and landing operations were compared with results from other studies. Comparing stumpage cost of eight logging operation in Guyana, Landell-Mills (1997) reported felling cost varying between $ 0.56 and $ 5.76 per m3; skidding cost between $ 4.75 and $ 45.44 and landing operations between $ 1.26 and $ 11.23. The variation is huge, the lowest and highest records differing in a proportion of one to ten for all sub-operations! Flaming (1995) reported a felling cost of $ 3.95/m3, a skidding cost of $ 2.99/m3, while landing operations were not specifically dealt with. Logging intensities and skidding distances are not specified in these two studies, which limits their use as a reference. Another explanation why the cost of skidding was much higher in Landell-Mills’ study (median value: $ 17.00/m3) than in the present study is offered by the reported machine operating cost. Landell-Mills reported a median owning cost of $ 2.75/m3, compared to a owning cost per m3 of $ 1.94 in the present study (see Appendix C, Table C.2). Moreover, a cost of maintenance and repairs of $ 7.03/m3 is reported, compared to $ 1.12/m3 in the present study. The latter was equated with the cost of depreciation according to guidelines by FAO (1977, 1978). If maintenance cost would have been estimated according to Caterpillar’s (1986) guidelines, maintenance cost would have been as low as $ 0.66/m3. If the difference in owning cost is supposed to reflect a difference in average skidding distance, and a similar ratio is applied to maintenance cost, a large difference in maintenance cost is found. This may be related to the relatively high price for spare parts in Guyana, but Landell-Mills’ suggestion, that the use of machinery past its economic life and the lack of accurate records contributed to this, is also plausible. Montenegro (1996) reported felling costs of $ 1.50/m3 for conventional and $ 2.03/m3 for reduced impact logging and skidding cost of $ 8.38/m3 for conventional and $ 8.60/m3 for reduced impact logging in La Mayronga, Ecuador. More informative is the felling output of 39.6 m3⋅day-1 for conventional and 28.8 m3⋅day-1 for reduced impact logging. These figures are about half of the output obtained in the present study (Table 5.7). The skidding output was only about one-third of the results from the present study: 5.6 m3 per machine hour for conventional skidding at an average skidding distance of 800 m and 3.5 m3/hr for reduced impact logging with unspecified skidding distance. According to the author, performance of both felling and skidding in reduced impact logging was related to a lower density of commercial stock in the units subjected to reduced impact logging. Holmes et al. (1999) reported felling costs of $ 0.50/m3 for conventional logging at 25.4 m3/ha and $ 0.63/m3 for reduced impact logging at 25.1 m3/ha for Paragominas, Brazil. Skidding costs amounted to $ 2.12/m3 for conventional and $ 1.24/m3 for reduced impact logging and costs of landing operations (including loading) $ 2.00/m3 and $ 1.27/m3 respectively. The difference in cost of skidding in the Paragominas study is influenced by the use of a CAT D6 crawler tractor in the conventional operation against a CAT 525 grapple skidder in the reduced impact operation. According to the authors, the hourly cost of the grapple skidder is 14% less than that of the crawler, 193 Reduced impact logging in the tropical rain forest of Guyana while labour cost were also lower in the reduced impact operation; i.e. 30% less than in the conventional operation. Together this resulted in a reduction of the machine operating cost by 17%. In addition, the travelling speed of the skidder most likely was higher than that of the crawler (Caterpillar 1986). The cost reduction stated by Holmes et al. (1999) indeed is only partly associated with pre-harvest planning and directional felling. In summary, it can be concluded that the cost of the harvesting phase is similar to findings in Brazil and Ecuador, but much less so to findings in Guyana. Since the international data resulted from more detailed studies than the Guyanese studies, the estimates in the present study are judged to be realistic. 5.5.4 Main transport An evaluation of logging cost should include the transportation of logs to the delivery point and associated activities such as loading and unloading of the transport vehicle. In studies on cost of logging in Guyana, this phase was indeed included (Landell-Mills 1997 and Flaming 1995). Most studies that compare conventional and reduced impact logging, however, omitted this phase (e.g. Hendrison 1990; Winkler 1997; Holmes et al. 1999, but see Montenegro 1996). This may, however, overemphasise the negative or positive effect of implementing reduced impact logging on stumpage cost. The reason for excluding this phase most likely lies with the fact that the main transport system is regarded to remain unchanged. Unfortunately, it was not within the bounds of possibility in the present study to carry out a detailed study on the performance and cost of main transport. Therefore, they were assessed on the base of random field observations and rules of thumb (FAO 1978). This resulted in the estimates shown in Table 5.13 (for details see Appendix C, Tables C.3 and C.4). Table 5.13 Activity Loading Trucking Unloading Cost and performance in m3 per machine hour of trucking, loading and unloading for both logging operations. Estimates are based on a hauling distance of 50 km, a maximum truck-trailer load of 26 t (21.7 m3) and an average travelling speed of 36 km/hr. (for details see Appendix C, Tables C.3 and C.4) Cost (US$/m3) 3.43 6.17 2.58 Output (m3/hr.) 21.7 6.0 28.9 Because the performance of all activities in this phase are limited by the load capacity of the truck-and-trailer and the truck turnover time, and not by individual log size or stocks at the landing, it is assumed that the same cost and output rates apply to both operations. The validity of the estimates in Table 5.13 was checked by comparing them with costs reported elsewhere. Landell-Mills (1997) reported a cost of $ 4.06/m3 for loading and $ 9.79/m3 for trucking (median of 8 operations in both cases). Flaming (1995) advanced $ 0.98/m3 for loading and $ 8.94/m3 for trucking at DTL in 1994, whereas 194 Chapter 5: Logging efficiency DTL stated $ 4.81/m3 for trucking over October 1998. The estimates in Table 5.13 thus are not far from reported costs of actual operations in Guyana. 5.5.5 Support, logistics, supervision and overhead costs In addition to the direct costs of the logging operation, which was discussed in the foregoing sections, indirect (supportive) costs are incurred. These include road maintenance, transport of workers and supplies, supervision, administration cost and rent capture. The cost of road maintenance is calculated on the basis of the daily output, and not on hectare basis, because it depends on the length of the stretch of road to be maintained, accumulating over the years. As shown in Table 5.14, the increase in output by implementing reduced impact logging led to a reduction of indirect costs by 8%. Costs for support, supervision and logistics are being incurred during the planning, preparation, actual harvesting and trucking phases. This implies that part of these costs are being incurred one year prior to harvesting, part six months prior to harvesting and part at the time of harvesting. It was therefore decided to regard these costs as having been incurred six months prior to harvesting. The net present value of costs for transport of workers and supplies then becomes $ 3.10/m3 for conventional logging and $ 2.66/m3 for reduced impact logging; for supervision $ 0.28/m3 and $ 0.24/m3, respectively; and for cooks $ 0.25/m3 and $ 0.21/m3, respectively. The net present value of the indirect costs together then becomes $ 10.42/m3 for conventional logging and $ 9.62/m3 for reduced impact logging. Table 5.14 Cost per m3 and equivalent volume output per day of support, logistics and supervision. Estimates are based on a daily output per logging team of 84 m3 for conventional logging and 97 m3 for reduced impact logging, at logging intensities of 28.5 m3/ha and 31.0 m3/ha respectively. Base camp and administration costs are based on data provided by Demerara Timbers Ltd. (for details see Appendix C, Tables C.3 and C.4). Cost subject Road maintenance Logistics Supervisor Cooks Base camp Administration cost Royalty Area fee Total Conventional Cost (US$/m3) Output (m3/day) 1.91 167 2.88 84 0.26 84 0.23 84 1.66 0.33 2.68 0.21 10.17 Reduced Impact Cost (US$/m3) Output (m3/day) 1.64 195 2.48 97 0.22 97 0.20 97 1.66 0.33 2.68 0.19 9.40 The validity of the estimates in Table 5.14 was checked by comparing these with costs reported elsewhere. Landell-Mills (1997) reported a cost per m3 of $ 3.06 for road maintenance, $ 14.30 for camp operations, an overhead cost of $ 7.23 and a rent capture of $ 0.57 (pre-1998 royalty and area fee rates). This adds up to total indirect cost of $ 26.17 per m3 (median of 8 operations). Flaming (1995) reported a cost per m3 of $ 2.85 for road maintenance, $ 2.30 for camp operations, $ 2.23 for forest 195 Reduced impact logging in the tropical rain forest of Guyana administration cost and a rent capture of $ 1.48 (pre-1998 royalty and area fee rates) for DTL in 1994. Cost for logistics, supervision and support were not specified. DTL itself (pers.comm.) reported a road maintenance cost of $ 4.83/m3 for October 1998. Holmes et al. (1999) reported cost of support, logistics and supervision to amount to only $ 0.41/m3 in conventional logging and $ 0.32/m3 in reduced impact logging in Paragominas, Brazil. Although indirect costs are not clearly itemised in the Guyanese studies, it is almost certain that the higher indirect costs reported in those studies are caused by a lower daily output of logging teams and lower logging intensity. 5.5.6 Aggregate costs of logging The aggregate cost of logging, expressed in US$/m3, was similar in both operations, as shown in Table 5.15. Reduced impact logging required increased investments during the pre-harvest phases; i.e. an increase by 182% of planning and by 7% of preparation costs. Actual harvesting costs were increased too, by 7%. In contrast, however, indirect costs were reduced by 8% due to a higher daily output and a higher yield per hectare. The large difference in weight of these activities made that these differences balanced out in the aggregate cost (Fig. 5.11). Reduced impact logging bore higher up front costs; the weight of which should be adjusted by taking the net present values of these costs. The net present value (NPV) of the aggregate cost is $ 28.61 for the conventional and $ 28.56 for the reduced impact operation. Due to the low weight of the pre-harvest costs - on which interest is payable - the relation between the two aggregate costs is hardly affected. Table 5.15 Cost of logging in US$/m3 of conventional and reduced impact logging in Pibiri, central Guyana. Costs based on logging intensity of 10 trees/ha (conventional yield = 28.5 m3/ha; reduced impact yield = 31.0 m3/ha), feeder road density of 1 km/km2; average skidding distance as the crow flies of 383 m and hauling distance 50 km. Daily output based on skidding performance; i.e. 84 m3/day for conventional and 97 m3/day for reduced impact logging (for details see Appendix C and Sections 5.3 and 5.4). Cost (US$/m3) Activity Pre-harvest planning Harvest preparation Felling & bucking Skidding Landing operations Trucking, loading and unloading Road maintenance Support, logistics and supervision Other overhead costs Royalty and area fee Total cost Conventional 0.18 0.52 0.60 4.30 0.34 12.18 1.91 3.37 1.99 2.89 28.29 Reduced Impact 0.50 0.56 1.16 4.10 0.32 12.18 1.64 2.90 1.99 2.88 28.23 Note: total logging cost reported here is lower than reported for Guyana elsewhere (e.g. Flaming 1995; Landell-Mills 1997). This is partly due to a higher yield per hectare in the present study. Average yields in Guyana: 4 to 10 m3/ha (Flaming 1995). It may also be due to machine operating costs having been estimated following handbook guidelines (FAO 1977, 1978; Caterpillar 1986; Sundberg & Silversides 1988). Finally it may be due to performances of pre-harvest planning and preparation operations, loading, trucking and unloading having been estimated using general guidelines. 196 Chapter 5: Logging efficiency The indirect costs apparently play an important role in this comparison. These costs except for royalty and area fee - together with the cost of main transport were estimated following rules of thumb, and hence may not be accurate. Leaving out these cost factors, the direct cost of a cubic metre of timber at the landing was estimated at $ 5.94 (NPV = $ 6.01) for the conventional and $ 6.63 (NPV = $ 6.75) for the reduced impact operation; an increase by 12% (increase in NPV by 12%). Also the costs of pre-harvest planning and preparation were based on general guidelines, and may be higher than estimated. The latter especially applies to the cost of road construction. An increase of these costs will, however, lead to convergence, because they are fixed on a hectare basis. It can thus be safely concluded that the supernumerary cost of logging in reduced impact logging will not exceed 12% and most likely is far less than this amount when indirect and transport costs are taken into account too. Figure 5.11 Breakdown of logging cost in US$/m3 for a conventional and a reduced impact operation. The costs are based on a logging intensity of 10 trees/ha (conventional operation yield = 28.5 m3/ha and reduced impact operation yield = 31.0 m3/ha); a feeder road density of 1 km/km2 (resulting in an average skidding distance of 383 m); and an hauling distance of 50 km. Cost of activities characterised by manual labour, such as planning activities (line cutting, inventory, etc.), skid trail demarcation and landing operations form a small proportion of the total cost. Cost of planning activities, road construction and skid trail demarcation depend on a fixed cost per hectare. In case the logging intensity is reduced, the proportion of these costs in the total cost per m3 will increase accordingly. Trucking and indirect costs will also be affected, the former due to a decrease in daily output and the latter due to an increase of the hauling distance. The effect of logging intensity can be found in Section 5.7. 5.5.7 Effect of wood recovery and felling pattern The difference in efficiency between the two logging methods is determined by several factors. In the reduced impact logging operation, skidding efficiency was enhanced by a rational extraction system. However, certain sacrifices had to be made, such as an increase in detail during the planning phase and a reduction of the output during 197 Reduced impact logging in the tropical rain forest of Guyana felling, related to a more careful felling method. Moreover, the final output of the operation was also enhanced by a higher wood recovery rate of the felled trees. In principle, it would also have been possible to aim at a higher recovery rate alone without aiming to facilitate the skidding phase. Conversely, the conventional operation had an advantage over the reduced impact operation, because in this particular ecosystem felled trunks were located in clusters. The contribution of these two factors to logging cost have been examined separately, in order to get a filtered picture of the contribution of the rational extraction system itself. Wood recovery rate The average recovered volume per tree in the conventional operation amounted to 2.9 m3, whereas a volume of 3.1 m3 was recovered on average in the reduced impact operation. It could be argued that the increased utilisation of the forest resource is the basis for the initial extra investment during the felling phase to be earned back by lower indirect costs. In order to verify whether this is true or not, I attributed the extra investment during the felling phase to a higher wood recovery only. The financial value of this additional amount of timber recovered was estimated by substituting the cost for felling and bucking in the conventional operation by the cost found for this activity in the reduced impact operation. By increasing the utilisation of the standing timber, the yield per hectare is increased and, consequently, pre-harvest planning and preparation costs per m3 are decreased. In addition, the output of the skidding operation changes as well by a higher wood recovery. In the skidding time model (see Section 5.4.6), an average skidder load of 9.7 m3 was assumed for both logging systems. Applying the same load size here is assuming that the time required for loaded and unloaded travel and unloading does not change. The time to collect a load changes, however, because fewer logs are needed to compose a load of that size with the adjusted tree volume. Per round-trip the time required to collect and bunch a load is therefore decreased from 27.4 to 26.5 minutes. The hourly output is this way increased from 14.4 m3 to 14.8 m3. The daily output, which affects indirect cost items, is increased from 84 m3 to 86 m3. As shown in Table 5.16, the greater volume and value recovered per hectare and per day are outweighed by the higher cost of felling. Indeed, the extra investment during the felling phase in reduced impact logging was mainly benefiting the performance of the skidding phase and only marginally the higher wood recovery. Felling pattern As discussed in Section 5.4.6, the efficiency of the conventional operation benefited from clustered felling. This resulted in a relatively shorter time for approaching and collecting logs. Adjusting for this difference, by assuming similar times for approaching and hooking of logs in both operations, resulted in an hourly skidder production of 12.9 m3 for conventional logging. The financial value of this difference is given in Table 5.16, which indicates that clustered felling enhanced the efficiency by 4%. It is also shown that, once the effect of clustered felling is being eliminated, the implementation of reduced impact logging would reduce the total cost per m3 by 4%. 198 Chapter 5: Logging efficiency As mentioned in Section 5.5.6, the reduction of the indirect cost of logging plays an important role in these comparisons. Leaving out indirect costs and cost of main transport, the direct cost of a cubic metre of timber at the landing was estimated at $ 6.43 for the conventional with adjustment for the felling pattern and $ 5.94 for the clustered felling. This indicates a difference due to clustered felling in direct cost at the landing of 8%. Table 5.16 The effect of waste reduction and felling pattern on stumpage cost (US$/m3). The change in cost of each activity due to a higher recovery of usable timber if the reduced impact felling method would have been applied in the conventional operation (no waste) is illustrated by the differences between columns 1 (the observed case - see Table 5.15) and 2. The change in cost of each activity if the effect of clustered felling is eliminated is illustrated by the differences between columns 1 and 3. The cost of each activity for the reduced impact operation is given as a reference. Activity Conventional Observed Pre-harvest planning Harvest preparation Felling & bucking Skidding Landing operations Trucking, loading and unloading Road maintenance Support, logistics and supervision Other overhead costs Royalty and area fee Total cost 5.6 No waste 0.18 0.52 0.60 4.30 0.34 12.18 1.91 3.37 1.99 2.89 28.29 0.17 0.48 1.16 4.21 0.32 12.18 1.86 3.30 1.99 2.88 28.54 Reduced Impact No clustered felling 0.18 0.52 0.60 4.80 0.34 12.18 2.13 3.76 1.99 2.89 29.39 0.50 0.56 1.16 4.10 0.32 12.18 1.64 2.90 1.99 2.88 28.23 THE INFLUENCE OF LOGGING INTENSITY ON THE COST OF AN OPERATION In the previous sections, it has been stated several times that stumpage costs depend on the logging intensity. The study on logging efficiency was carried out in relatively small logging units, where logging intensity amounted to about 10 trees per hectare with a yield of 31.5 m3/ha for reduced impact logging. In Guyana, such yields can only be obtained in well-stocked areas. Guyana’s forests are characterised by a patchy occurrence of desirable species (see Section 4.1) and well-stocked areas are thus interspersed with commercially less interesting areas. Consequently, the average logging intensity is reportedly low. Flaming (1995) reported that logging intensity in Guyana varies between 4 and 10 m3 per hectare. Klassen (1987), in a report describing the first years of logging in the Mabura Hill area, mentions that an area of 39,600 ha had been laid open by mid-1987. By that time, 4,800 hectares of that area had been logged, whereas another 7,500 hectares had been identified as being of sufficient commercial interest to be harvestable as well. 199 Reduced impact logging in the tropical rain forest of Guyana The logging intensity achieved in the actually logged area was 26.8 m3/ha, so that the overall logging intensity was 3.3 m3/ha. Including the projections for further logging in the area made at that time would have resulted in a projected logging intensity of 8.6 m3/ha. This figure is based on an assortment of species that were (easily) marketable in 1987. Expansion of this assortment in the near future is imaginable, because of a shift in focus in recent years from the traditional sawn timber species to peeler species (see Sections 1.5.3 and 2.4.4). The logging intensity in the present study is thus (far) higher than what is commonly harvested. For this reason, the influence of the logging intensity has been worked out for the reduced impact operation. As shown in Table 5.17, the activities the cost of which is fixed per hectare, such as pre-harvest planning, road construction, demarcation of skid trails and the area fee, change in direct proportion to a change in logging intensity. The cost of felling and skidding changes because the hourly output is being changed. In case of felling the latter is a result of a change in the time required for locating trees, and in case of skidding a change in the time required for approaching logs. Because the time involved in these activities only forms a minor part of the total felling or skidding cycle times, the changes are relatively small. The costs of trucking, road maintenance and logistics are also influenced by the logging intensity, because of a change in hauling distance. The change in the aggregate logging cost per m3 is small (increase of 5%) when the logging intensity is reduced from 16 to 8 trees/ha (from 50 to 25 m3/ha). The cost rises more strongly, when the intensity is further reduced to 4 trees/ha (for 25 to 12.5 m3/ha) and even more when subsequently being reduced to 2 trees/ha (from 12.5 to 6.25 m3/ha) (Fig. 5.12). Table 5.17 Estimated stumpage cost in US$/m3 by activity for reduced impact logging at four different logging intensities. The results are based on the assumption that the logging compartment is situated 50 km from the point of delivery with an annual yield of 100,000 m3. The average log size is 3.1 m3. (See Appendix C, tables C.6 and C.7 for details of cost calculations) Activity Pre-harvest planning Harvest preparation Felling & bucking Skidding Landing operations Trucking, loading and unloading Road maintenance Support, logistics and supervision Other overhead costs Royalty and area fee Total cost Logging intensity (N/ha) 4 8 1.21 0.61 1.34 0.68 1.17 1.16 4.17 4.08 0.32 0.32 12.69 12.26 1.82 1.66 3.20 2.94 1.99 1.99 3.15 2.92 31.08 28.62 2 2.42 2.69 1.18 4.25 0.32 13.54 2.12 3.68 1.99 3.63 35.83 16 0.30 0.34 1.14 4.01 0.32 12.04 1.57 2.79 1.99 2.79 27.30 The changes in net present value of these costs are sharper; rising from $ 27.59/m3 to $ 28.99/m3 (an increase of 5%) when the intensity is reduced from 16 to 8 trees/ha, to $ 31.61/m3 (an increase of 9% compared to the former) when the intensity is reduced to 4 trees/ha, and $ 36.71/m3 (an increase of 16% compared to the former) when reduced 200 Chapter 5: Logging efficiency to 2 trees/ha. This is - among others - related to the fact that the cost of planning activities, which take place one year prior to harvest, are doubled each time the intensity is halved. The absolute financial values and differences shown in Table 5.17 should not be taken too seriously. The results are all based on projections from the observed situation with a logging intensity of 10 trees/ha. Some aspects were intentionally ignored for convenience of comparison. For instance, the cost of support and logistics is probably more strongly affected than assumed in the projections above, because by reducing the logging intensity either larger areas would have to be served from a particular camp site, or the camp would have to be shifted more frequently. Another simplification is, that trees were assumed to be evenly spaced over the area. As discussed in the beginning of this section, commercially interesting trees often occur in patches so this assumption is wide of the mark. It is unknown, if the clumped occurrence affects the performance of felling and skidding. A large-scale empirical study of the cost and benefits of logging at different intensities may provide an answer, but the cost of such a study would undoubtedly be prohibitive. This hypothetical comparison shows that the costs of pre-harvest planning, road construction, road maintenance, trucking and logistics go up markedly when the logging intensity is reduced, whereas the cost of felling and skidding only increase marginally. The comparison between conventional and reduced impact logging showed, that the cost of pre-harvest planning and felling were strongly increased by implementing the reduced impact logging, whereas the cost of skidding and overheads were reduced. These results apply to well-stocked areas, where a logging intensity of 30 m3/ha can be achieved. Going by previous experiences, well-stocked areas are interspersed with poor areas, resulting in a logging intensity of between 5 and 10 m3/ha (see top of this section). Assuming that the assortment of presently marketable species is larger than in the period 1985-1995, it seems reasonable to expect the average logging intensity to fall at the upper end of the reported range; i.e. 10 m3/ha. Table 5.18 Projected cost of logging in US$/m3 in conventional and reduced impact logging in Pibiri, central Guyana, for a logging intensity of 10 m3/ha. The costs are projected from the observed logging intensities. (See Tables 5.15 and 5.17 and Appendix C, Tables C.6 and C.7 for details of calculations) Cost (US$/m3) Activity Pre-harvest planning Harvest preparation Felling & bucking Skidding Landing operations Trucking, loading and unloading Road maintenance Support, logistics and supervision Other overhead costs Royalty and area fee Conventional 0.55 1.62 0.61 4.41 0.34 12.92 2.22 3.85 1.99 3.34 201 Reduced Impact 1.55 1.72 1.17 4.19 0.32 12.92 1.90 3.33 1.99 3.28 Reduced impact logging in the tropical rain forest of Guyana Total cost 31.85 32.40 As shown in Table 5.18, the application of a realistic logging intensity gives conventional logging a slight advantage over reduced impact logging. The increase in terms of percentage does not change for planning and preparation cost, neither does the decrease in terms of percentage of the indirect cost. The change in weight of these activities, however, results in an increase in the aggregate cost of 2% by implementing reduced impact logging. Leaving out the indirect cost and the cost of main transport results in a direct cost of a cubic metre of timber at the landing of $ 7.52 (NPV = $ 7.73) for the conventional and $ 8.97 ($ 9.34) for the reduced impact operation. This indicates an increase of direct cost at the landing by 19% (increase in NPV by 21%). 5.7 COMPARISON WITH OTHER STUDIES Several studies have shown that reduced impact logging can be an instrument to achieve a dual objective of damage reduction and efficiency enhancement. In these studies additional costs incurred because of planning and directional felling were more than compensated by increased efficiency of the actual extraction - skidding operation. Mattson-Mårn & Jonkers (1981) in Sarawak reported additional cost due to planning and pre-opening of skid-trails amounting to 4% of the cost of traditional skidding, whereas actual skidding cost was decreased by 27%. Interestingly, directional felling was achieved in Sarawak without any increase in cost in that study. Hendrison (1990) reported an increase in planning cost of 72% and in cost of felling by 85% by implementing the CELOS harvesting system in Suriname, whereas cost of skidding was decreased by 37%. Since the cost of skidding was about four times higher than the other activities combined, aggregate direct costs (up to the landing) were reduced by 16%. Barreto et al. (1998) reported that felling in a ‘planned’ operation was 18% less productive than felling in an ‘unplanned’ operation, whereas skidding performance was increased by about 27% in the Paragominas region in Brazil. The cost to plan logging operations was approximately US$ 1.87/m3 on a total cost at the sawmill gate of US$ 26.48/m3. Improvement of efficiency of road and landing construction and extraction and a marked reduction of wasted volume led to an estimated 35% greater net profit with planning than without. Holmes et al. (1999), on the other hand, found that pre-harvest planning and preparation costs - including line cutting, inventory, liana cutting and tree marking - in a reduced impact operation were elevenfold the costs incurred during this phase in a conventional operation, in which it included only scouting, and that the cost of felling was 26% higher. These investments outweighed a gain in skidding efficiency, which amounted to a cost reduction of 42%, resulting in an increase in the aggregate direct cost at the landing of 25%. In addition, Holmes et al.(1999) report that the costs of road construction, landing operations - trimming and loading together - and of 202 Chapter 5: Logging efficiency overheads were considerably reduced by implementing reduced impact logging. If these costs are included, an overall cost reduction by 5% is achieved in Paragominas. A study by Winkler (1997) in the State of Amazonas, Brazil, also resulted in a higher cost per m3 wood volume delivered at the landing site. The cost increased with 5% when recommended ‘environmentally sound harvesting system’ changes were implemented. Pre-harvest planning and preparation did not occur in the traditional operation; felling cost was 31% higher in the environmentally sound operation, whereas skidding cost was 33% lower. This resulted in the aggregate cost per m3 at the landing in the environmentally sound harvesting system being 9% higher than in the traditional operation. The author denoted that the improved harvesting system was not carried out according to the planned changes, in which case the cost of implementing the improved system, would only have been 1.5% higher. In La Mayronga, Ecuador, Montenegro (1996), reported that implementation of a reduced impact (‘rational’) logging system increased the direct cost of felling by 34% and skidding by 1%. The traditional logging method it was compared to featured no planning. These increases together result in a 55% higher cost of logs delivered at the landing. Inclusion of the cost of road construction, main transport, stumpage price, licences and other overhead cost in the latter comparison resulted in a weighed down increase of 21%. Training was also included as additional cost in the latter study and amounted to US$ 1.42 per m3. In the study in the State of Amazonas and the one in La Mayronga, the logging intensity in the reduced impact logging operations was conspicuously lower than in the respective traditional operations - a factor four and a factor two respectively. This inflated the cost per m3, mainly of road construction, planning and overheads, but also to some extent of felling and skidding. In the present study in Guyana, precision inventory maps allowed the felling crew to locate selected trees efficiently, but still were trees more quickly located in the conventional operation. This was undoubtedly due to selected trees occurring in clusters. Directional felling itself took substantially more time than the traditional felling method, partly caused by an increased preparation time, but mainly by increased felling time. The amount of time spent bucking felled trees was also longer for reduced impact logging due to increased efforts to recover usable wood, and also due to the more frequent encounter of internal decay (see Section 5.3). It was anticipated that the efficiency of the skidding operation would be higher for reduced impact logging, because directional felling oriented the downed boles in a manner that movements of the skidder in the stump area were minimised. Although trees were felled in clusters in the conventional operation, giving it an edge over reduced impact logging as regards efficiency, the output was increased by 5% by implementing the reduced impact logging system. An increase in performance was also expected during the round-trip to the landing. A planned, dendritic trail system was presumed to allow the skidder to move more efficiently (faster) between the stump area and the landing than the unplanned, winding trail network of conventional logging. It was also presumed that it would reduce the 203 Reduced impact logging in the tropical rain forest of Guyana actual distance covered. Unfortunately, the scale of the experiment precluded a proper field test of this hypothesis. Nevertheless, based on the tortuosity of the trails within the units where skid trail patterns were representative of actual practice and a theoretical model of the skid trail pattern of an entire logging block of 100 ha, it was assumed on empirical grounds that skidding efficiency would be augmented by 10% by implementing reduced impact logging. If, in addition, the advantage conceded to conventional logging because of clustered felling would be ignored, hence also matching the collecting time with the one of reduced impact logging, skidding efficiency would be augmented by 23% (see Section 5.4). Translation of these changes in performance into logging cost and incorporating changes in pre-harvest planning and preparation costs provided the following picture. Implementation of the reduced impact logging system led to a threefold increase in pre-harvest planning and preparation costs and a twofold increase in felling cost, which were only partly offset by lower skidding cost; i.e. a reduction by 5%. Because the cost of skidding was about six times the combined cost of felling and planning in the conventional operation, this resulted in an increase of the aggregate direct cost at the landing of only 12%. The direct logging cost in the reduced impact operation was thus higher than in the conventional operation, but the output per hectare and per day were importantly higher under the reduced impact regime. The higher output per hectare and per day reduced the cost per m3 of road construction as well as indirect cost factors, such as costs of logistics, support, area fees, etc. In the final result these gains neutralised the differences in the direct logging costs, and the reduced impact logging system is thus neither necessarily more expensive, nor cheaper than harvesting in the traditional way. It must be noted that this applies only under the conditions found in the operations that were monitored. These operations were monitored in stands with a high density of - often clustered - trees of commercial species of harvestable size. It is well known that such high densities are only found in scattered patches in Guyana. Why did the results of the present study differ from the ones of the other studies? Skidding efficiency was enhanced by on average 33% in most other studies (but see Montenegro, 1996), and felling efficiency declined by on average 30% (but see Hendrison 1990). In the present study, skidding efficiency was only marginally improved, whereas felling efficiency declined substantially. This has several reasons. ƒ Differences of technical and organisational nature existed between the logging operations compared elsewhere. For instance, in Hendrison’s study, the efficiency improvement was partly related to a higher level of organisation in the damagecontrolled operation, and partly to a lower maintenance standard of the machinery used in the conventional operation. In the study of Holmes et al. (1999), the employment of a more cost-effective machine in the reduced impact logging system, contributed to the enhancement of skidding efficiency (see Section 5.5.3). ƒ In the present study, a ‘best practice’ logging operation was used as a reference, whereas in the other studies ‘hit-and-miss’ logging operations were used. ƒ In the present study, the conventional operation featured felling in groups, whereas trees were scattered in the reduced impact operation. This feature gave the 204 Chapter 5: Logging efficiency ƒ ƒ conventional operation an advantage (see Section 5.5.7). It is assumed that such differences did not occur in the other studies. In the present study a special felling technique was employed in the reduced impact operation, which to our best knowledge was more sophisticated than the directional felling techniques used in the other reduced impact operations. Due to several constraints, training of felling and skidding crews was forcibly restricted to one month only. Especially in case of the felling method, the applied technology probably preceded the presence of the required technical competence, thereby to some extent ‘putting the cart before the horse’. Indeed, can operational efficiency be compared well at all among selective logging operations? Forest harvesting systems are more complex than most other production systems because of the necessity of moving to the object of interest - the tree to be felled or log to be extracted - passing through a welter of continuously varying conditions imposed by soil, topography and distances, resulting from a complex of site properties. In selective logging in tropical rain forest with its multitude of species and tree dimensions and variable distances between trees to be harvested, this complexity far surpasses that of temperate forests. In comparisons involving sites in different countries, another factor is introduced; i.e. the inequality of cultural and historical frameworks. The complexity of the work environment implies that ceteris paribus conditions can not apply. Van Rompaey (1993) proved that neither homogeneous sample units nor true replication could exist in a (tropical rain) forest. This appears to apply to studies on operational efficiency just as it applies to biological studies. Therefore, no widely valid differences can exist between financial values of inputs and outputs. This limitation applies within the borders of a single country, but applies even more strongly when borders are crossed. 5.8 REQUIRED PRECISION IN DIRECTIONAL FELLING In the present study, care was taken to direct a tree in the position prescribed by the planned skid trail system. This was achieved by applying the felling method described in Section 2.3.4 almost invariably, whereby the crew forced the tree to fall in an angle of about 135° with the transport direction, with the butt end of the bole leading. If felling in the prescribed direction was likely, either to damage future crop trees, the log itself, or to cause the tree to lodge in a neighbouring tree the direction was adjusted. In such cases - and in case of strong natural lean - it was first examined whether or not the tree could be felled within a range of ± 30° on either side of the intended direction. If this was not possible, the next alternative was to fell the tree in the opposite direction, at an angle of 315° with the trail. These directives were followed tenaciously, which often necessitated trees to be directed against their natural lean, sometimes requiring arduous wedging. Nevertheless, the obtained results were certainly no mean achievement of the felling crew. The deviation from the predetermined direction was recorded for 500 trees. The direction 205 Reduced impact logging in the tropical rain forest of Guyana was adjusted in 32% of the cases in order to avoid damage to the log itself or to growing stock. In 12% of the cases, the natural lean of the tree was such that the tree could not be steered, and, hence, was felled with the lean. Of the remaining 279 trees 85% fell within a range of ±30° with the intended direction and only 8% in an undesirable direction, i.e. making an angle between 30° and 150° with the intended direction (Van der Hout 1996). Field observations during the skidding operation challenged the necessity of such a high precision felling operation. It appeared that trees could be felled in plenty of directions and still could be managed conveniently by the skidder. Boles lying at an angle between 105° and 165° with the transport direction could be winched satisfactorily as long as the bole end was more than one tree length away from the skidder. Neither was there a big problem if the tree had been felled towards the trail. If it was in an angle between 195° and 245° with the planned transport direction, the skidder winched the bole by the top end and positioned it alongside the trail. Subsequently, the log was unhooked, the skidder turned around and the log hooked again at the butt end. For skidding with the butt forward, which is recommended to reduce skidding resistance, the combined allowable range covered thus 120°. It is generally not considered recommendable to skid with the top end leading (but see Dykstra 1996). First, it increases the skidding resistance and, second, the bole has to be approached through the crown zone. The latter problem was far smaller than expected, because the choker man usually could approach the log readily from the side. Although not recommendable, skidding with the top end forward proved to be feasible. This opportunity further extended the range of manageable directions of the lie, implying that a tree could be felled in directions together covering 180° of the compass rose, on condition that the distance to the trail was long enough. If the tree is located closer to the trail the manoeuvrability is less, and the range more restricted. Van Leersum (1984) made similar observations during a comprehensive study on the effect of the direction of the lie on the manageability for winching in Suriname. In sum, it appears that more lenience can be applied regarding the felling direction. However, in that case the skidder operator has to judge the handling of each log individually. The presence of a crew chief is of paramount importance, because he is the one that will have to act as a ‘logistic manager’. He will have to obtain information from the felling crew on the direction of the lie the logs and to transfer this information to the skidding crew. This information should not only include where the operator must reverse into the forest and where to halt, but also which actions must be taken next. This can be facilitated by letting the felling crew sketch the actual direction of the lie on the stock map, which is handed over to the skidding crew at the end of a working day. If greater flexibility of the felling direction is allowed, in principle, this can be short-circuited by the recommendations in the previous chapter. There, it was suggested to reduce damage to the remaining stand by forming dual treefall gaps when 206 Chapter 5: Logging efficiency appropriate. Caution should be exercised in this case, because, first, the formation of too overly large gaps should be avoided and, second, cross-felling should be avoided. 5.9 BENEFITS OF REDUCED IMPACT LOGGING So far, no attention has been paid to the financial benefits of the logging operations. This has several reasons. First, the absolute cost estimates were determined only for a small part by actual performance studies. Felling and skidding were studied in detail. Nevertheless, only the manoeuvres in the stump area were judged to be representative of a real skidding operation, because the performance assessment during the trips to and from the landing was partly based on projections of the performance within the stump area. The cost estimates of the activities that precede and succeed these two core activities were not obtained with real data but deducted using rules of thumb. Secondly, the performance studies of felling and skidding were carried out in well-stocked areas. In reality, well-stocked areas are islands in a sea of poorer areas, indicating that the costs in real terms are likely to be higher, especially those of pre-harvest planning, road construction and overheads. Thirdly, it has proved difficult to establish the true value of an extracted log. The domestic market is distorted by market failure and the increase in ‘chainsaw operations’, where trees are felled and converted at the stump into cants by chainsaw. These operators work with very small capital costs and pay few or no fees, enabling them to undercut the prices charged by conventional saw-millers (Flaming 1995; Palmer & Marshall 1996). Export prices also suffer from distortion because of government interventions, weak marketing strategies and, allegedly, transfer pricing (Flaming 1995; Colchester 1997; Landell-Mills 1997). The importance of the choice of price is stressed by Gray (1983), who indicated that a 10% error in price leads to a 20% deviation in stumpage value. Therefore, in order not to create any false impression of the potential stumpage value of a tree standing in the forest, this subject was not treated here so far. The implementation of the reduced impact logging system, however, had a substantial effect on the species composition of the cut. This is associated with the directive to discourage felling in groups, which leads to the formation of large gaps in the canopy and is regarded to be detrimental to the sustainability of timber production (see Sections 4.5.2 and 4.8.3). Indirectly, this directive led to a relief of the pressure on Greenheart, which by some authors is considered over-harvested (e.g. GNRA 1989; Ter Steege et al. 1996). In Section 4.4.1, it is mentioned that the harvest obtained in the conventional operation was for 90% composed of Chlorocardium rodiei, the remainder belonging to other hardwood species such as Peltogyne venosa (Purpleheart) and Hymenaea courbaril (Locust). In the reduced impact operation, Chlorocardium rodiei constituted only 47% of the total volume output (see Section 4.4.2). Other hardwoods made up 37% of the cut, peeler species (mainly Catostemma spp. - Baromalli) 16%. The average export price for Greenheart logs amounted to $ 322/m3 in 1997 (Anonymous 1998). The average export price for other hardwoods varied between $ 110/m3 and $ 150/m3, while peeler logs fetched $ 90/m3 on average (Mazaharally, 207 Reduced impact logging in the tropical rain forest of Guyana pers.comm.). To arrive at a sawmill gate price, the cost of transport to port and port handling charges should be deducted from these export prices. Landell-Mills (1997) employed a 30% deduction for transport and charges. The sawmill gate price is therefore estimated at $ 225/m3 for Greenheart logs, $ 90/m3 for hardwood logs and $ 63/m3 for peeler logs. Applying these prices to the species group distribution in Section 4.4.2 yields an average price for a m3 in the conventional operation of $ 212 and in the reduced impact operation of $ 149. The stumpage value, excluding profit allowance, is estimated by deducting the stumpage cost (Table 5.18) and amounts to $ 180/m3 for conventional logging and $ 117/m3 for reduced impact logging. The absolute values and their variations are not accurate, because it is unknown whether or not the costs per m3 reported in Table 5.18 are accurate, whether or not the species composition as found in the study area applies to the overall area and whether or not transport to port and port handling charges are accurate. It is also assumed in Section 5.6 that the cost of felling and the cost of skidding in a conventional operation would increase with decreasing logging intensity at the same rate as it did in the reduced impact operation. Although this assumption is questionable too, still, in the short term gross as well as net returns of the conventional logging system seem to be considerably higher than the ones obtained by the reduced impact operation studied at Pibiri. The above result is no surprise, because it is inherent in the philosophy behind the two logging systems. The reduced impact logging system is essentially a component of a silvicultural system, which the conventional system is not. It all boils down to the fundamental distinction between two classic forms of forestry; i.e. ‘timber mining’ versus silviculture. Enterprises exploiting large concession areas basically have as their prime, or even sole, objective to maximise profits on the output of their logging and saw-milling capacity, which measures up to short-term maximisation of the profit per unit of production (logging equipment, sawmill, etc.). Inherently, there is no interest in or incentive to use secondary species or to reduce waste, because this would affect the pace of output of - highly profitable - prime logs. This point of view can be understood as long as the forest resources are considered - for practical purposes - to be unlimited (Sundberg & Silversides 1988). Silviculture, which per definition aims at perpetual yields, has different objectives. Silvicultural profit maximisation is based on a long-term maximisation of the profit per unit area. The feasibility of this kind of maximisation wholly rests on safeguarding the long-term productive capacity of the forest. In that respect, sustained yield forestry (silviculture) considers the full economic value of the forest including the value of unmarketable species and environmental services provided by the forest, whereas timber mining only concerns the potential financial value of the forest. Since natural forest resources are limited in principle, timber mining in due course is bound to either be converted into a step in silviculture, or the forest sector - especially when based solely on Chlorocardium rodiei - will end up going bust. The conversion into silviculture can and will only be achieved if funds, logically part of the income obtained by ‘timber mining’ - are invested in the forest. 208 Chapter 5: Logging efficiency In the long term, several benefits are associated with reduced impact logging. Attempts to place a financial value on this are neither realistic nor even desirable for that very reason. As amply discussed by Leslie (1977, 1987a, 1987b), Fearnside (1989), Oldeman (1991) and Lande et al. (1994) standard financial appraisals based on the net financial return inevitably lead to decisions that favour short-term gains over long-term benefits. An obvious weakness in financial assessment is the omission of non-commercial benefits provided by the forests. The step from the financial appraisal underlying the standard evaluation to the economic one is now well recognised. It means that non-revenue aspects and external effects are brought into the analysis and that market values are replaced by social values. Although the derivation of monetary and social values for non-commercial benefits remains problematic, correcting for market imperfections and distortions can make a substantial improvement in the economic appraisal of forest management. Even so, the idea and application of compound interest alone are almost sufficient to undermine the corrected economic appraisal. The choice of the right rate of interest is thus quite crucial, but also quite subjective, as it is related to the object and processes that have to provide that interest. Several reasons are given by Leslie (1987a), Fearnside (1989) and Lande et al. (1994) for using lower than standard interest rates based on income potential from alternative investments, or even a zero interest rate, in the case of forestry, where inputs and outputs are spread over long periods of time. The example below illustrates how the choice of the interest rate - which earned the reputation as a tyrant in forestry (Leslie 1987a) - can lead to opt for timber mining (see also Kuper 1994). No definite information is available at present as to the rotation to be expected under a regime of reduced impact logging, because growth, recruitment and mortality rates have only been monitored for three years after logging (see Section 1.7.3). Some information is available on the prospects for Chlorocardium rodiei in conventional logging (Prince 1971a, 1973; Zagt 1997), and some scanty information for the entire commercial stand (Prince 1971b). Prince (1971a) suggested that it takes Greenheart about 150 years to grow from 5 cm dbh to 50 cm dbh when left untreated after logging, while advance growth would allow cutting cycles between 80 and 100 years (see Section 2.4.3). Zagt (1997) even predicted that Greenheart stocks would keep declining after the first cut during the first 100 years without substantial silvicultural interventions. Because all harvestable Greenheart was taken out in the conventional operation, the growth rate of the species is slow and advance regeneration is usually deficient (see e.g. Ter Steege et al. 1996), the second cut in conventionally logged forest most likely will have to rely on other species than Greenheart. This is not the case in forest stands subjected to reduced impact logging, where a large part of the larger individuals were retained as growing stock. The value of the next crop thus will most likely be higher in the stand subjected to reduced impact logging, as long as the price relationship between species remains constant, which obviously reflects the demand expected in the future. Future demand quantification is a well-known problem in long term management and nobody can say precisely how much of a certain product will be in demand in the future, let alone at which price (e.g. Oldeman 1991). 209 Reduced impact logging in the tropical rain forest of Guyana In a hypothetical case, I assumed the first harvest to be equal to what was found in the study area with stumpage values of $ 180/m3 for conventional logging and $ 117/m3 for reduced impact logging. Then I assumed an arbitrary mean annual volume increment of 1 m3/ha, which assumes that a second harvest is feasible in 30 years. This periodicity is of usual order in literature on tropical silviculture (e.g. De Graaf 1986). Computations are made using discount rates between 0% and 10%. Prices for different types of timber are assumed to inflate at an equal rate corresponding to an assumed, average inflation rate. The second cut in conventional logging can not contain any Greenheart and is assumed to supply 66% mixed hardwood species and 33% peeler species. The second cut in the reduced impact logging is assumed similar in composition to the first harvest. Table 5.19 Indicative net present value (NPV) calculations of stumpage values in US$ for conventional and reduced impact logging in Chlorocardium rodiei forest in central Guyana, over a range of discount rates. Logging system Conventional Reduced impact year 1 year 30 total year 1 year 30 total Net present value Discount rate 2% 6% 180 180 16 5 195 185 117 117 64 20 181 137 0% 180 28 208 117 117 233 10% 180 2 181 117 7 123 A simplified demonstration is constructed as follows. The stumpage values for each type of logging in present day timber prices are given in Table 5.19. If compound interest is ignored, reduced impact logging is favoured, but a switch to conventional logging occurs even before a discount rate of 2% is reached; i.e. at an interest rate of 1.1% per year. Following Fearnside (1989), it thus is quite logical that loggers make no effort to restrict the cutting of the prime species or to restrict damage to the future crop under the current system of economic decision rules. 5.10 CONCLUSION A financial appraisal of reduced impact logging in the Mabura Hill area in Guyana showed that reduced impact logging is cost competitive with conventional logging. The direct cost per m3 of timber at the landing may have been 12% higher, which was mainly related to higher planning and felling costs. There are, however, strong indications that the cost of felling will be reduced in the future, when the operators have gained more experience in using the felling method and directional felling is carried out with greater flexibility. In the present study, the additional costs were offset by a higher daily as well as areal output, which reduced road construction and indirect costs, such as logistics, supervision and support. 210 Chapter 5: Logging efficiency There are strong limitations to the absolute values reported in the present study, because of a large number of assumptions with respect to the cost of a number of activities. These activities preceded felling or succeeded skidding. Further limitations are imposed by the fact that the study was carried out in well-stocked areas, where a yield of 30 m3/ha could be achieved, whereas lower yields are reported elsewhere. This implies that there is an urgent need for a detailed performance study of an entire operation - from the planning phase to the delivery of the logs at the sawmill gate - on a scale representative of an actual operation, including both rich and poor silvicultural units. The net return, in the short run, for conventional logging is higher for conventional logging than for reduced impact logging due to a partial shift in species selected for harvesting. In the long term, however, reduced impact logging will lead to higher returns, as long as a low discount rate is applied. If, in addition, the step is made from financial to economic appraisal - discounting non-commercial benefits and recalculating the monetary values of all benefits and costs in social values - the advantage is wont to go easily to reduced impact logging. A realistic study on the social value of these non-commercial benefits and the associated problems in converting the social values into monetary values would be quite valuable in this respect. 211 Reduced impact logging in the tropical rain forest of Guyana 212 Chapter 6: Simulated silvicultural treatment 6. SIMULATED SILVICULTURAL CROWN ILLUMINATION 6.1 INTRODUCTION TREATMENT AND In the previous chapters it was shown that implementation of reduced impact logging can mitigate logging damage and is not necessarily more costly than conventional practice. However, sustainable forest management is not confined to sound harvesting practice alone. It appeared to be equally important that the intensity with which logging is carried out co-determines the failure or success of reduced impact logging. Furthermore, the target species in the conventional practice - in the present case, Chlorocardium rodiei (Greenheart) - occurs in even-sized groups (groves), and the current logging practice in Greenheart forest hence is characterised by felling in groups. This leads to the formation of large canopy openings, which favour the establishment of light-demanding, pioneer species and climber tangles. It is doubtful whether Chlorocardium rodiei seedlings that are already present can survive this ephemeral stage (Ter Steege et al. 1994; Zagt 1997). Hence, consideration also has to be given to restriction of the number of trees to be felled in these groves, to limit gap size. Still, implementation of a silvicultural system that incorporates reduced impact logging methods and considers limits to logging intensity does not guarantee that the objective of sustainable timber production can be achieved. If the objective is the sustenance of perpetual capability of a forest for self-regulation, a sustainable silvicultural system must be formulated (see Section 2.3.1). 6.1.1 Silvicultural operations After selective logging the overstorey left is impoverished with regard to trees of species that would be valuable for future harvests or as sources of seeds. Reduced impact logging may assist in reducing damage to advance growth to a minimum, but generally, sound harvesting alone is not sufficient to cause a uniform positive reaction from advance growth or natural regeneration of commercially desirable species (Hutchinson 1988). By felling, some suppressed trees may be released, but most of the commercially desirable saplings and juveniles are overtopped by overmature and non-commercial trees, that retard their growth. The unattended logged forest therefore usually shows only marginal volume increment of commercially desirable species (e.g. Synnott and Kemp 1976; De Graaf 1986). Moreover, the remaining canopy is stripped of seed trees belonging to commercial species, in the long term leading to progressive economic impoverishment of the forest. The desire to apply a silvicultural post-harvest treatment that ought to restore the balance in the species composition is obvious. Treatments can also aim at the release of advance growth and surviving growing stock, which is expected to lead to enhanced increment rates. Enhancement of the increment rate of advance growth can potentially 213 Reduced impact logging in the tropical rain forest of Guyana shorten the cutting cycle and hence increase the volume production per unit of land. Consequently, the hunger for land by logging companies and the stumpage cost will be reduced. The smaller the area required to produce the necessary timber, the greater will be the natural area that can, in theory, be preserved for conservation and environmental objectives (cf. Wyatt-Smith 1987). Obviously, this argumentation only holds when the demand for timber remains constant over the years. Finally, release will avoid much of the natural mortality of desirable species which would otherwise occur owing to wind-throw of overmature disfavoured trees or suppression. On the other hand, any further interventions in the canopy following logging will inevitably lead to increased light levels on the forest floor which in turn may provoke the establishment of silviculturally undesirable pioneer species. Additionally, the need for maintenance of biodiversity and of species of special importance, due to their value as food suppliers for frugivorous mammals and birds (Terborgh 1986; Bonnéhin in press), may put restrictions on the elimination of certain non-commercial species. Overmature, defective trees of commercial species can have an important function as seed trees in the mycorrhizal network supporting juvenile trees (Yasman 1995). Such trees also have an important function by providing breeding places for birds. Development of silvicultural guidelines thus will require decisions regarding trade-offs between activities that enhance timber production and those that reduce negative impacts on the biological integrity of the system (cf. Pinard et al. 1999). Removal of the unwanted trees in favour of the desirable trees could be carried out by felling, but this would open up the canopy drastically, creating large patches of high illumination and thereby favouring the germination and growth of light-demanding pioneer species. This can only be justified if the desirable species are light-demanding. This is not the case in Guyana - except for few species such as Goupia glabra and Jacaranda copaia - and the alternative is to eliminate non-desirable trees by girdling them. Girdling alone is not always effective and the application of arboricides is usually advocated (Dawkins 1958; De Graaf 1986; Mallet and Tuo 1989). The benefit of (frill-) poisoning is situated in the fact that trees die slowly inflicting little damage to surrounding trees. The poisoned tree is dismantled step by step, first of its leaves, subsequently of its minor branches, then larger branches, until, finally, the remaining trunk, denuded of its crown, tumbles down. This process may take up to several years, while dead wood once on the ground may take years to decompose and hence only marginal losses of sequestrated carbon are foreseen (Ter Steege, pers.comm.). The gradual decomposition of the fallen wood is suggested to give rise to a ‘fertiliser effect’ (Jonkers and Schmidt 1984). The selection of trees for elimination can be based on two opposing approaches: systematic and selective release: ƒ Systematic release eliminates trees exclusively because of their lack of commercial potential (trees of non-commercial species and defective individuals of commercial species), once they surpass a certain diameter threshold. Such a prescription is applied in De Graaf’s CELOS system (1986) and in the 214 Chapter 6: Simulated silvicultural treatment experiments established with the aid of CIRAD-forêt in Ivory Coast (Maître 1986; Dupuy et al. 1997), Central African Republic (Dupuy et al. 1998), and French Guyana (Schmitt and Bariteau 1990). This type of operation is often referred to as refinement (Boerboom 1964; Hutchinson 1988). A classic forestry term for this type of operation would be weeding. Smith et al. 1996, p. 147, say that weeding is: ‘the thorough removal of all plants competing with the crop species, regardless of whether their crowns are above, beside, or below those of the desirable trees’. ƒ Selective release eliminates non-commercial trees that compete for light with potential crop trees, i.e. reserved juvenile trees belonging to commercial or potentially commercially species of good quality. This type of intervention can be compared with classical crown thinning (Fr. éclaircie par la haut). Smith et al. 1996, p. 102, define this as: ‘trees are removed from the upper crown classes in order to open up the canopy and favour the development of the most promising trees of these same classes … overtopped trees that do not interfere with crop trees are not removed’. In tropical forests this method, referred to as ‘improvement thinning’ - was practised in Ghana between 1956 and 1970, having been abandoned under the pressure from the timber industry (Asabere 1987). On an experimental scale, it has been tried in the Philippines (Weidelt and Banaag 1982; Wöll 1988) and recently in Indonesia (Nguyen-Thé et al. 1999). The two approaches each have their own advantages and disadvantages. Systematic release has gained much support because of its simplicity (cf. Brasnett 1953; De Graaf 1986), but often leads to the elimination of many large trees in areas deprived of potential crop trees. This is not only a waste of time and funds, it also ignores possible changes in the timber market, and the useful regulatory role currently not merchantable trees may play in the stand. For instance, Couratari (Wadara) trees were poison-girdled in the trials carried out in Suriname in 1976, an established commercial species nowadays (De Graaf 1995, pers.comm.). This varies from their beneficial effect on height growth and self-pruning of trees of desirable species and on soil protection and nutrient cycling, while their presence inhibits the establishment and growth of climber tangles and light-demanding pioneers (Fox 1976). Mycorrhizal fungi die when the soil is heated over ca. 30°C (Smits 1995, p.158), so useful tree regeneration often depends on shaded soils. Selective release is closer to natural forest dynamics because poisoned stems occur in patches surrounding individual trees. The general canopy level remains high and the diversity of tree species is largely preserved. Therefore, it is highly flexible and can accommodate changes in market demand. However, this type of operation demands skilled labour. First, potential crop trees have to be identified, which is not necessary with the former method. Species identification of today’s commercial species may not pose a problem, but may become one when identifying potentially commercial species. In addition, the detection of hidden interior defects may pose a problem. Consequences of misjudgement are far-reaching in ‘tree-wise’ selection. If a defective tree is wrongly being reserved as a potential crop tree, its neighbours will be eliminated in vain. 215 Reduced impact logging in the tropical rain forest of Guyana Conversely, if a sound tree is wrongly deemed defective, it may be eliminated for the sake of a neighbouring potential crop tree. Secondly, the decision as to a greater or lesser potential value of a tree, may result in complicated preferential treatment prescriptions, that are often difficult to apply (Jonkers 1995, pers.comm.). Finally, although it may be possible to determine which neighbouring trees are shading a potential crop tree directly from above, it is usually quite difficult to determine which trees are shading laterally. The first constraint can be partly solved by formally training staff involved and by research into the development of tools for detecting interior decay (e.g. a suitable boring tool or sonic device). The second problem can be attacked by simplifying decision rules. A binomial system that opposes desirable and non-desirable trees seems to be most practical, implying, however, that when a desirable trees is being impeded by another desirable tree, both are left standing regardless whether one has a greater potential value than the other. The third constraint is often approached by using tree size related parameters. In the ‘liberation thinning’ method that was outlined by Wadsworth (1958) for Puerto Rico, a distance table is used whereby the distance within which a tree is to be eliminated depends on the diameter of that particular tree. This method was further developed by Hutchinson (1987) for Sarawak. Jonkers (1995, pers.comm.) claims that the use of the distance table often leads to confusion and advocated elimination of inferior ‘competitors’ in a fixed radius surrounding the potential crop trees (see also Gourlet-Fleury 1992, 1997). Both in Hutchinson’s liberation thinning and in Jonkers’ modified CELOS system (1987) not only the trees that are impeding the development of potential crop trees at any one moment, but also those trees that are expected to become impeders within 30 years later, are subject to elimination. In addition, both methods advocate the systematic elimination of large non-commercial overmature trees (‘relics’). Jonkers (1987) recommended poison-girdling of all those non-desirable trees greater than 40 cm dbh and in addition any of more than 20 cm dbh that are within ten meters of a desirable stem of more than 20 cm dbh. The reasoning behind this prescription was that in Suriname trees smaller than 40 cm dbh usually have a crown radius of less than four metres and are not likely to impede the desirable tree at a distance of more than ten metres. On the other hand, severe hindrance over such distance was indicated for trees with a diameter greater than 50 cm dbh. The threshold was put 10 cm lower presuming that an ‘impeder’ of 40 cm dbh would have reached a size of 50 cm dbh before the end of the projected cutting cycle. Hutchinson’s (1987) method prescribed the removal of all non-commercial trees of listed species 60 cm dbh and larger and of non-commercial species of 50 cm dbh and larger. In addition, non-commercial trees were eliminated that were found to impede reserved trees of more than 10 cm dbh, or that were considered to impede them in the future (decision based on the distance table). The method whereby trees are systematically eliminated within a fixed radius around potential crop trees can be seen as a hybrid form of systematic and selective release or as a mixed method (Gourlet-Fleury 1992; see also Kio et al. 1986). 216 Chapter 6: Simulated silvicultural treatment 6.1.2 Light conditions for potential crop trees In order to examine whether a release operation can improve the conditions for potential crop trees and, if so, which treatment rules will be most effective, the following approach was followed. In two plots subjected to reduced impact logging with an intensity of 8 trees per hectare, the amount of light received by selected trees was estimated by using SILVI-STAR, a three-dimensional forest model (Koop 1989; Koop and Bijlsma 1993; Koop and Sterck 1994). The two plots represent two common types of ‘Greenheart forest’; i.e. one dominated by Chlorocardium rodiei, Lecythis confertiflora (Wirimiri Kakaralli) and Catostemma fragrans (Sand Baromalli) and another by Mora gonggrijpii (Morabukea), Chlorocardium rodiei and Lecythis confertiflora. The relationship between the amount of light received by a tree crown and its status in the forest stand; its ‘social position’ (Kraft ex Houtzagers 1956, p. 257; see also Leibundgut 1958 and Mayer 1980) was examined. The parameters used to describe the social position of the trees were: its crown position according to Dawkins (1958) and the number of neighbouring trees that shade vertically or laterally (‘impeders’). In addition, crown and stem size parameters of favoured and disfavoured trees were examined. Better knowledge about these relationship will tell whether these parameters, most of which can be easily determined in the field, can be used as indicators for the social position of a selected tree and which neighbouring trees are likely to impede its development. Five different silvicultural treatments, removing non-desirable trees from the database according to different prescriptions, were simulated using SILVI-STAR. After each treatment, the amount of light received by selected trees was computed with the sub-programme FORFLUX. The simulated treatments were: ƒ systematic release by eliminating all non-commercial trees with two diameter limits; i.e. 20 cm dbh and 30 cm dbh; ƒ systematic release by eliminating non-commercial trees that are situated within a fixed radius around potential crop trees with 2 distances; i.e. 7 and 10 metres; ƒ a release with a fixed radius of 10 m around selected trees, supplemented by the systematic removal of all non-commercial trees with a diameter exceeding 40 cm dbh (modified CELOS system, Jonkers 1987). True selective release could not be simulated, because the trees that would qualify as impeders in the SILVI-STAR forest model were unknown beforehand. Practical restrictions also prevented this after the model was constructed, because this would require removal of potential impeders one by one and each time estimating the improvement in light conditions for the selected trees. The established relationships between the amount of light received and the two parameters indicating the social position of a tree, were used to determine which trees qualified as impeders. The different treatments were subsequently evaluated based on their efficiency in removing impeders: i.e. the percentage of all impeders actually eliminated; and the percentage eliminated trees that actually did not impede the development of selected trees. 217 Reduced impact logging in the tropical rain forest of Guyana 6.2 METHODS 6.2.1 Selection of potential crop trees and their social position Two plots subjected to reduced impact logging with an intensity of 8 trees per hectare in 1994; i.e. plot 4 and plot 13, were selected for simulation of silvicultural treatments. Trees over 20 cm dbh in these plots, that measure 140×140 m2 each, have been monitored since the latter part of 1993. All trees have been identified to the species (occasionally genus) level, have received a unique number and have their position recorded. Since 1993, diameter measurements have taken place in 1995, 1996 and 1997. In addition to the recording of the diameters, the Dawkins crown class and the stem quality of commercial and potentially commercial species have been recorded. Three broad categories of species are defined: 1. commercially desirable species, the timber of which has commercial value at present; 2. potentially commercial species, the timber of which is scarcely marketable at present but is known to have suitable properties and is projected to have higher value in the future; 3. other species, with neither current nor projected future value. The first two categories together comprise the class of listed species (Appendix A). In order to be selected as a potential crop tree, a tree should: 1. be a listed species; 2. have a diameter in the range 20.0 to 39.9 cm dbh; 3. be free of serious defects or damage; 4. contain or be projected to contain at least one length of sound, straight stem six metres long (after allowing for stump height); 5. not lean more than 15 degrees from the vertical. The diameter range was limited at the lower end at 20 cm dbh, because the release of smaller trees carries the risk that these are damaged by the falling branches of the poisoned neighbours. Another reason is that the majority of these trees will not attain merchantable size before the end of the projected cutting cycle (some 30 years). De Graaf (1986) did find an annual increment rate of 1.0 cm during a period of six years since treatment, but it is not likely that such a rate can be sustained over longer periods. Over an entire felling cycle of some 30 years, the average annual increment rate in treated forest will probably not exceed 0.4 cm (Gourlet-Fleury 1992; Favrichon 1997). It will take a 10 cm tree 75 years to reach 40 cm when growing at such a rate. Natural mortality will take its toll and only a reduced number will survive. In French Guyana Favrichon (1997) found an average mortality rate of 2% during the first seven years after silvicultural treatment. A similar rate was found by Van der Hout (1983 ex De Graaf 1986) in Suriname during the first six years after treatment. The mortality rate will most likely return to its background level of 1% some ten years after treatment (De Graaf, pers.comm.). Application of a negative exponential model (Sheil & May 1996) with these mortality rates indicates that less than 50% of these trees will reach the 218 Chapter 6: Simulated silvicultural treatment felling limit. The investment in releasing such small individuals thus does not appear justifiable. Light availability is critical to the growth of trees. It is also the only environmental factor that is quite easy to manipulate by the forester. The position of a tree crown relative to its neighbours in the forest canopy has been used in forestry for silvicultural decision making since the nineteenth century. Kraft (ex Houtzagers 1956, p. 257) devised a system of crown classification with four classes: dominant, co-dominant, intermediate and suppressed. The Kraft classification is intended for pure stands and in mixed uneven-aged forests, the Dawkins classification has become well established (Dawkins 1958): 5. Crown in full overhead and lateral light; 4. Crown in full overhead light, lateral shade; 3. Crown partially exposed to overhead light, lateral shade; 2. Crown without overhead light, partial lateral light; 1. Crown shaded on all sides, no direct light. Other measures of ‘competition’ are the number of neighbouring trees that shade the focal tree overhead or laterally (e.g. Gourlet-Fleury 1992) and the index of canopy closure of Lieberman et al. (1989). Wadsworth et al. (1988) and Gourlet-Fleury (1997) compare several ‘indices of competition’. Many indices, such as ‘size ratio distance indices’ (e.g. Steneker and Jarvis; 1963; Hamilton 1969), ‘polygon indices’ (e.g. Moore et al. 1973; Daniels et al. 1986) are designed for temperate forests and are difficult to transplant to mixed tropical forest (Gourlet-Fleury 1997). All these indices or classes can be used to determine the efficiency of a release operation. The question remains how well they are associated with the amount of light indeed received by the focal tree. In recent years, efficient tools have become available to determine the amount of incoming radiation at a certain point in a forest. Hemispherical photographs are recently suggested as an efficient tool (e.g. Whitmore et al. 1993). The use of hemispherical (or fisheye) photographs in estimating incoming radiation was introduced by Evans and Coombe (1959). Anderson (1964) described the formal procedure to calculate diffuse and direct light. Since then a number of methods have been described, ranging from manual analysis to computerised techniques (e.g. Bonhomme and Chartier 1972; Walter 1994; Walter and Torquebiau 1997; Ter Steege 1997). Obviously, it is not practical to attempt to take hemispherical photographs at the tops of selected potential crop trees, but, fortunately, alternatives are available. 6.2.2 The three-dimensional forest model SILVI-STAR The three-dimensional forest model SILVI-STAR (acronym for SILVIgenesis and Single-tree Three-dimensional ARchitecture), developed and described by Koop (1989) allows computation of the hemispherical cover for fixed points as an indication of the interception of diffuse light and changing oblique direct radiation. 219 Reduced impact logging in the tropical rain forest of Guyana SILVI-STAR attempts to bridge the gap between geometric transect analyses (e.g. Richards 1952; Oldeman 1972, 1974a/b; Whitmore 1975; Hallé et al. 1978) and computerised predictive forest models (e.g. Shugart 1984; Pacala et al. 1993). The model describes stylised volumes occupied by trees. For a certain tree this volume is described by the co-ordinates of eight points, representing tree base, position of first fork, top and base of tree crown and the four extremities (peripheral points) of the crown circumference in x- and y-direction. All z-co-ordinates are derived from height measurements while x- and y- co-ordinates are digitised from a hand-drawn map (Figs. 6.1 and 6.2). X- and y- co-ordinates of the fork and the top and the base of the crown are not measured in the field, but computed from the eccentricity of the tree base in the crown projection. The shape of the crown model is constructed by connecting the four peripheral points of the crown circumference with quarters of ellipses. (a) Figure 6.1 (b) (a) The three-dimensional tree model. T top of the tree; P periphery point; C crown base; F fork; B stem base. (b) Height measurements on a tree: ht top height, hp height of the greatest width of the crown periphery, hc height of the crown base, hf height of the first fork, hb height of the tree base to a reference height. (after Koop, 1989). The crown shapes are regarded as homogeneously transparent bodies with a specified extinction coefficient determined by the inner crown cover. The inner crown cover is estimated by visual assessment according to the decimal scale of Londo (1984). The data collection procedure described by Koop (1989) was used in this study. Height measurements were made with a Haglof Vertex hypsometer, which yields more accurate results than standard hypsometers, especially under the difficult circumstances of a tropical rain forest. Hand-drawn maps were digitised at the Dutch Institute of Forestry and Nature Research in Wageningen. SILVI-STAR uses the programme FOREYE, developed by Bijlsma (1990), for the purpose of computing the hemispherical crown cover. It computes hemispherical crown cover by scanning all trees in a 180° cone over a matrix of 360 azimuthal 220 Chapter 6: Simulated silvicultural treatment degrees by 90 zenithal degrees. All individual trees are scanned for each horizontal and corresponding vertical degree. The values resulting from those computations are processed with the sub-programme FORFLUX in a similar way as a hemispherical photograph. The underlying radiation model is based on standard functions of direct and diffuse light under clear sky and standard overcast sky in J/m/s against sun height (De Wit et al.1978 ex Koop 1989). With these programmes, diffuse and direct irradiation were computed for selected potential crop trees in the core area of each plot, which measured 100×100 m2. In order to keep computing time within acceptable limits, this was done for four days only; i.e. two days in the year that the sun is in zenith (around the equinocta: 21 March and 21 September) and two days that the sun is above either tropic (around the solstices: 21 December and 21 June). In addition, irradiation was computed for a point ‘outside’ the forest. For each tree, the percentage of the total irradiation (outside the forest) was calculated subsequently. This procedure was repeated for each of the six different modelled forest stands, the model of the forest as it was observed in the field and models after the five before-mentioned alterations. Computations were carried out by R.J. Bijlsma on the computer of the Institute for Forestry and Nature Research. P2 P3 P1 P4 Figure 6.2 The extremities of the tree crown projection in x- and y-direction (peripheral points) that are digitised from the hand-drawn map. In the model the periphery is indicated by a compound of quarter ellipses stretched up by these four points. (after Koop, 1989) 6.2.1 Assessment of social position In order to establish relationships between the percentage of total irradiation received by a focal tree and traditional indices, such as the Dawkins’ crown position and the number of neighbouring trees that seem to impede the development of the focal tree, assessments were made based on the three-dimensional forest model. The reason is that field assessments may lead to values different to those obtained from the model, which is related to the stylised form of the crowns and the perception of crowns as homogeneous bodies. In reality crown transparency is far from homogeneous (e.g. Oldeman 1974a, p. 133). On the other hand, the model allows a better view of the structure of the canopy and the relative positions of the crowns than an observer on the forest floor would have. Dawkins’ classes were assigned by interpreting plotted side and top views of overlapping 20 m wide transects (see Figs. 6.19 and 6.20). In addition, neighbouring tree crowns were identified that overtopped the focal tree in the model; i.e. preventing the focal tree from being assigned a Dawkins’ crown class ‘4’. 221 Reduced impact logging in the tropical rain forest of Guyana In determining which neighbouring trees impede the development of the favoured tree, it is clear that most shading is to be expected from adjacent trees of a stature similar as or taller than the one of the focal tree. Trees that are farther away do not offer any immediate threat to the dominance of the favoured tree, but may shade the tree laterally. One could take as a measure that neighbouring tree crowns that intersect an inverted cone with a certain angle with its apex resting on the top of the favoured tree are impeding the growth of the focal tree. Often an angle of 90° is used (e.g. Koop and Sterck 1994; Smith et al. 1996, p. 151). Koop and Sterck (1994) founded their decision for this angle on the fact that the angles near zenith are most important because most diffuse radiation can be expected from angles close to there. Direct radiation coming from angles below 45° is intercepted on a long path through a large number of tree crowns. The question is how strongly this intercepted radiation contributes to the final reduction in the total amount of radiation received by the favoured tree. Therefore, it was decided to compare two angles; i.e. counting neighbouring tree crowns that project upward into inverted cones having angles of 90° and 120° with their apex on top of the crown of the favoured tree. As mentioned above, neither the x- nor the y- co-ordinates of the top of the crowns in the three-dimensional forest model are based on field measurements. They are calculated from the eccentricity of the tree base in the crown projection en the relation between crown depth and stem length (Koop 1989; Bijlsma 1990). In order to determine which tree crowns protrude in the inverted cones, the x- , y- and zco-ordinates of the top of the crowns are needed. SILVI-STAR uses the following definitions. The eccentricity of the tree base in x- direction (eBx) is defined as the ratio of the distance between the tree base (Bx) and the centre point of the crown projection (Mx) and crown radius (Crx) as: eB x = with Crx = M x − Bx Crx P3x − P1x , and 2 M x = P1x + Crx where (see Fig. 6.2): P1x = crown projection extremity with smallest x - value P3x = crown projection extremity with largest x - value The eccentricity of the x- co-ordinate of the tree top (eTx) is then achieved by expressing it as the arctangent of the eccentricity of the tree base (eBx): 222 Chapter 6: Simulated silvicultural treatment eTx = 2 ⋅ arctan(eBx ) π A reduction factor for the relative stem length (R) reduces the eccentricity of the top of the crown when the ratio stem length ( Cz = height of the crown base) to total tree height (Tz) increases: R = 1− Cz Tz The x- value for the top of the crown is then calculated: Tx = M x + Crx ⋅ eTx ⋅ R Eccentricity along the y- axis is computed in an analogous way. It is not necessarily true that the top of an impeding tree crown is the first part of the crown to protrude into the inverted cone extending from the top of a favoured tree. The periphery or any other point between the periphery and the top may do so instead. In principle a tangent can be calculated along the vertical quarter ellipse that builds the part of the crown body that is closest to the top of the favoured tree when projected on the x- and y- plane. However, the values of a and b of the ellipse formula; ⎛ x2 y2 ⎞ ⎜ 2 + 2 = 1⎟ , a b ⎝ ⎠ shaping the quarter ellipse in a vertical plane, change constantly with the point on the crown periphery that the quarter ellipse is associated with. This would require a large number of calculations and it was decided to follow a simplified procedure. It was assumed that either the top or the periphery of the neighbouring tree, but no other point on the quarter ellipse between these two points, would intersect the inverted cone. The distance between the peripheral point that is closest to the top of the favoured tree was measured from the digitised crown projection map. This distance and the difference in height between the periphery of the neighbouring tree and the top of the favoured tree would then tell whether the crown of the neighbouring tree projected into the inverted cones. In fact, in determining which neighbours should be eliminated around each favoured tree, one should know the relative rates at which both each favoured and the undesirable trees surrounding it grow in terms of crown expansion. A tree that does not shade a favoured tree now, may eventually overtop it by expansion of its crown, either in height or laterally. On the other hand, lateral expansion of the crown of the potential crop tree may be hindered by neighbouring crowns imposing no impediment currently. Both Hutchinson (1987) and Jonkers (1987) suggested to provide for future hindrance. There are high correlations between crown diameter and stem diameter, as well as between crown expansion and diameter growth, but these relations vary widely between species. Correlations between crown and stem diameter were investigated for 223 Reduced impact logging in the tropical rain forest of Guyana some common desirable and non-desirable species in order to predict the amount of crown space required at maturity by either group. 6.2.2 Evaluation of treatments Treatments were evaluated on four accounts; i.e. the percentage of the potential irradiation received by the favoured trees in each case, the total basal area eliminated by each treatment, the effectiveness in removing impeders and the accuracy in elimination. Differences in the mean percentage of potential irradiation were tested by means of Repeated measures ANOVA (Zar 1984), followed by contrast analysis and carried out with SPSS 7.5. Three interpretations can be attached to the total basal area removed by a treatment. Firstly, it can be used as an index of the investment of material and human resources in the treatment, a true cost reflecting the effort needed to accomplish the treatment. Secondly, it can be interpreted as the biological cost reflecting a reduction in biomass, sequestrated carbon or biodiversity. Finally, it can be seen as a measure for the surge in nutrient availability; the ‘fertiliser effect’ (Jonkers and Schmidt 1984; Jonkers 1987). In this study the financial and biological cost are given a heavier weight than the expected increase in nutrient availability, and a higher eliminated basal area is regarded as a negative aspect. The effectiveness of a treatment is defined as the percentage of impeders that is actually removed by a treatment (following Gourlet-Fleury 1992): effectiveness = number of eliminated impeders total number of impeders This percentage will nearly always be less than 100% because a considerable number of impeders are favoured trees themselves or retained merchantable trees. The accuracy of a treatment is defined as the percentage of the total number of eliminated trees that were actually impeders (following Gourlet-Fleury 1992): accuracy = number of eliminated impeders total number of eliminated trees Other important measures to evaluate a treatment are the change in species diversity and the change in canopy openness. The first important aspect has not been investigated for the time being because of the small size of this experiment. The second aspect can theoretically be tackled by using SILVI-STAR. However, the database available at the time was not suited for this purpose, because only trees of which the diameter exceeded 20 cm were included. Meanwhile, relevant parameters of trees with a diameter in the range 10.0 to 19.9 cm dbh have been recorded and it is intended to carry out further computations in the future. 224 Chapter 6: Simulated silvicultural treatment 6.3 RESULTS AND DISCUSSION 6.3.1 Description of the tree populations Before logging, the forest in the experimental plots was rich in commercially desirable species (Table 6.1). Desirable species were relatively better represented in the larger than in the smaller diameter classes. They made up 56% and 64% of the total number of stems in the diameter range 20 to 40 cm dbh in the respective plots. Among the trees with a diameter of 40 cm dbh and over the percentage amounted to 74% and 87%. This difference is related to the fact that there are species among the non-desirable species that remain below this size at maturity. Table 6.1 Composition of the two experimental plots before logging. Number of stems and basal area per hectare for trees of 20 cm dbh and over by species group. Individuals of desirable species (currently and potentially marketable) were more numerous than individuals of non-desirable species. Desirable stems accounted for 60% and 72% of the total number of stems and 67% and 81% of total basal area. Species group desirable species non-desirable species total Plot 4 N/ha 115 76 191 Plot 13 BA/ha 12.9 6.4 19.3 N/ha 134 53 187 BA/ha 20.4 4.9 25.3 The forest in plot 4 consisted of pockets of young forest (‘aggradation phase’, Bormann and Likens 1979; Oldeman 1990) and small chablis in a matrix of mature forest. The canopy, although broken in some parts, reaches a height of 30 metres on average with a few emergent trees reaching up to 48 metres. The plot is situated on a slope in which two gullies descend in north-easterly direction. Slopes vary between 6% and 20% with a total difference in elevation of 21 metres. Chlorocardium rodiei (Greenheart) was the most abundant commercial species with 40 stems per hectare (corresponding to a basal area of 6.4 m2/ha), followed by Catostemma fragrans (Sand Baromalli), Swartzia leiocalycina (Wamara) and Carapa guianensis (Crabwood) with 18, 17 and 14 stems per hectare respectively (corresponding to basal areas of 1.1, 1.1 and 0.8 m2/ha respectively). Lecythis confertiflora (Wirimiri Kakaralli), Chaetocarpus schomburgkianus (Ruri) and Licania alba (Kautaballi) were the most abundant non-desirable species among the trees with a diameter of 20 cm dbh and over. The forest in plot 13 can be described as well-developed forest with a high average canopy height of 40 metres (maximum tree height 48 m) with pockets of overmature forest (‘degradation phase’) in a matrix of mature forest (‘steady-state’ or ‘bio-static phase’). Mora gonggrijpii (Morabukea) was the most abundant commercial species with 58 stems with a diameter of 20 cm dbh and above per hectare (corresponding to a basal area of 8.3 m2/ha), followed by Chlorocardium rodiei, Eperua falcata (Soft Wallaba) and Swartzia leiocalycina with 32, 10 and 9 stems per hectare respectively (corresponding to basal areas of 7.0, 1.4 and 1.0 m2/ha respectively). Lecythis confertiflora and Aspidosperma excelsum (Yaruru) were the most abundant non-desirable species among the trees with a diameter of 20 cm dbh and over. The plot 225 Reduced impact logging in the tropical rain forest of Guyana is situated on a gentle slope (8% but steady) that descends in north-westerly direction, with a total difference in elevation of 16 m. Several of the commercial individuals were defective. As shown in Figure 6.3, the proportion of defective trees increased with increasing diameter. In plot 4 the percentage defective trees amounted to 19% for trees in the range 20 to 40 cm dbh, 28% for trees in the range 40 to 60 cm, and as much as 58% for the largest trees. In plot 13 the percentages defective amounted to 24%, 33% and 37% for the diameter classes 20 to 40 cm dbh, 40 to 60 cm dbh and 60 cm dbh and over respectively. These values compare well with the study by De Milde (1970), who examined hidden defects by boring individual trees of several species. Figure 6.3 Diameter class frequency distributions of the two experimental plot before logging in 1993. Black bars represent sound individuals of commercially desirable species, hatched bars defective individuals. Open bars represent non-desirable species. Logging, which took place in 1994, removed only sound individuals of presently marketable species, while damage was inflicted indiscriminately (see Section 4.7.6). In both plots, eight trees were extracted per hectare, incidentally killing 13 trees per hectare in plot 4 and 19 in plot 13. In terms of number of stems, logging had a modest impact; 85% and 86% of the commercially desirable stems and 87% and 94% of the non-desirable trees survived the operation in plot 4 and plot 13 respectively. In terms of basal area, the impact was higher. It affected commercial species - the group of which trees were extracted - with a loss of basal area of 20% and 21% (plot 4 and plot 13 respectively) more seriously than non-commercial species with a loss of 12% and 10%. By 1996, the year tree assessments were made for the forest model, some ingrowth surpassing the 20 cm limit had occurred and some trees had risen to a next diameter class. The resulting stand structures are shown in Figure 6.4 and summarised in Table 6.2. Not unexpectedly, logging appears to have had a minor effect on the proportional representation of sound individuals of commercially desirable species in the diameter classes smaller than 40 cm dbh. In plot 4 the proportion of sound commercial trees had been reduced from 53% to 47% in the diameter class of 40 to 60 cm and from 31% to 21% in the diameter class of 60 cm and over. In plot 13 the reductions were from 56% 226 Chapter 6: Simulated silvicultural treatment to 52% and from 58% to 33% in the respective diameter classes. Apparently, a large number of trees of merchantable size are retained in the plots; i.e. 14 and 26 trees per hectare in plot 4 and plot 13 respectively. Continuity of timber supply thus may not to be at risk without treatment. Table 6.2 Composition of the two experimental plots two years after logging. Number of stems and basal area per hectare for trees of 20 cm dbh and over by species group. Species group desirable species non-desirable species total Figure 6.4 Plot 4 N/ha 104 78 182 Plot 13 BA/ha 10.5 5.9 16.4 N/ha 116 49 165 BA/ha 16.3 4.5 20.8 Diameter class frequency distributions of the two experimental plot two years after logging. Black bars represent sound individuals of commercially desirable species, hatched bars defective individuals. Open bars represent non-desirable species. However, the danger of this kind of reasoning has been touched upon in Section 2.1. It may lead to depletion of species one by one. For instance, eight Greenheart trees were cut per hectare in plot 4. A similar number of trees of merchantable size of this species were retained after logging. If a growth rate of 2 mm per year is assumed without silvicultural treatment (cf. Zagt 1997; Favrichon 1995), only trees between 34 and 40 cm will rise into the merchantable diameter class within the next 30 years. Allowing for an annual mortality of 1%, stand table projection indicates that only 3 trees will rise into this class, indicating that this level of cut will not be sustainable after the next cut (after which only 1 tree will be retained!). If a growth rate of 4 mm can be attained by releasing potential crop trees, it is indicated that 7 trees will survive to attain merchantable size in 30 years. This indicates that five trees may be retained after the following cut. Sustainability thus seems to come nearer, but is still not in reach. The level of cut has thus been too high to have a chance on sustainable production at that level by this species. However, it is clear that enhancement of growth and of survival of potential crop trees by silvicultural treatment may allow for higher cutting levels. Obviously, in deciding on allowable cutting levels, the structure of the population (positive stand table!) must be taken into account. In this respect, it is worthwhile to mention that population 227 Reduced impact logging in the tropical rain forest of Guyana structures of Greenheart varied widely between the 15 plots that have been established for this study, as be shown in Section 7.7. 6.3.2 Selection and social position of favoured trees The first question that must be answered is how many potential crop trees should be reserved per hectare. Cousens (1958), Dawkins (1958) and Wadsworth (1958) suggested a target of 100 ‘leading desirables’ per hectare. The concept of leading desirables aims to attain a regular distribution of commercial trees whereby the lower limit for a leading desirable can be as low as 2.5 cm dbh. In the selection approach used here, it is not a prominent goal to attain a regular distribution of commercial stems. Therefore, the problem is approached from the target stand as defined by the projected cutting level and felling cycle. Setting the cutting level at 8 trees/ha with a 30 year cutting cycle and regarding the retained commercial trees of merchantable size as a bonus or emergency supply, at least eight potential crop trees will have to attain merchantable size (40 cm dbh) during each cycle. To allow for flexibility in selecting crop trees, for instance to adapt to the marketability of the various species at that time or to cater for catastrophes, it is suggested to aim for 16 potential crop trees to survive to attain merchantable size before the end of the cycle instead. Based on a presumed diameter increment of 4 mm/yr., only trees that currently measure 28 cm dbh or more, will be able to surpass the 40 cm diameter limit within a period of 30 years. Trees between 20 and 28 cm dbh are thus destined for the subsequent cutting cycle. Allowing for a 2% annual mortality rate during the first ten years after treatment and 1% mortality thereafter, 24 trees between 28 and 40 cm dbh are needed. Trees between 20 and 28 cm dbh are reserved to be felled after 60 years. Applying the same growth and mortality rates and allowing for increased mortality during the next logging operation, indicates that 37 trees between 16 and 28 cm should be reserved to produce 16 stems of merchantable size in 60 years. The target is thus defined as 61 trees with a diameter of 16 cm and over per hectare. According to the selection rules (dbh≥ 20 cm; Section 6.2.1), 62 trees/ha qualified in plot 4 and 55 trees/ha in plot 13. The conclusion is thus that, in theory, a cutting regime with an extraction level of 8 trees/ha and a cycle of 30 years is feasible in this type of forest. Secondly, it implies that not all qualifying trees need to be selected to attain this objective. SILVI-STAR computations were limited to the central 100×100 m2, around which a buffer zone of 20 m was defined. According to Koop (1989), a core area where hemispherical cover is to be computed for points at ground level should be surrounded by a buffer zone with a width equal to the maximum tree height. Maximum tree height in both plots was 48 metres, but because hemispherical cover is computed for the top of the trees, the majority of those taller than 25 metres, a 20 m extension was judged sufficient. Within the borders of the central hectare, 58 selected trees were found in plot 4 and 48 in plot 13. 228 Chapter 6: Simulated silvicultural treatment Mean relative illumination as the percentage of total irradiation amounted to 72% (sd 27%) in plot 4 and to 58% (sd 28%) in plot 13. The whole range from fully illuminated to fully shaded trees occurred in both plots. In plot 13, trees were evenly distributed over all illumination classes, but in plot 4, a large percentage of the favoured trees appeared to be almost fully illuminated (Fig. 6.5). The difference in mean illumination between the plots was significant (Student’s t-test on arcsin√p - transformed percentages; t104 = 2.97, p < 0.01). This difference can be explained by: ƒ the larger number of large trees remaining after logging in plot 13 (Fig. 6.4); ƒ plot 4 was situated on a steeper slope and had a more broken canopy than plot 13; and ƒ plot 13 contained Mora gonggrijpii trees, a species with a higher than average inner crown cover, whereas plot 4 did not. Figure 6.5 Distribution of illumination classes for favoured trees in the two experimental plots. Illumination is the total irradiation received by a favoured trees expressed as the percentage of the of the total irradiation received by a point outside the forest. The higher proportion of well illuminated trees was also expressed in the distribution of Dawkins’ crown classes (Fig. 6.6). Only 10% of the selected trees in plot 4 received no overhead light, while 31% were shaded only laterally. In plot 13, seventeen percent of the selected trees did not receive any overhead light, while 12% were not shaded overhead and 2% were fully exposed. The plots also differed significantly when assessing illumination according to Dawkins’ classification (χ2 = 10.83, df = 4, p < 0.05). Regarding overtopping trees and non-overtopping trees that cast vertical or lateral shade as impeders, it is shown that 76% of the favoured trees were shaded by at least one neighbouring tree in plot 4 and 94% of the favoured trees in plot 13 (Fig. 6.7). The median number of impeding, neighbouring trees differed significantly between the plots: i.e. a favoured tree was shaded by two neighbouring trees on average in plot 4 and three neighbouring trees on average in plot 13 (Median test, χ2 = 9.18, df = 1, p < 0.01). Only 47% of the impeding trees belonged to the non-desirable tree category (both plots pooled). These were regularly distributed because if a favoured tree was 229 Reduced impact logging in the tropical rain forest of Guyana shaded, there often was a non-desirable tree among the impeders. Fifty per cent of the selected trees in plot 4 were shaded by non-desirable trees and 71% of the selected trees in plot 13. Figure 6.6 Distribution of selected trees among Dawkins’ Crown Classes in the two experimental plots. Relative frequency is shown on top of bars. Figure 6.7 Distribution of the number of vertically or laterally shading neighbours among the selected trees in the two experimental plots. According to the percentage of potential irradiation received, it appears that 46% of the favoured trees in plot 4 and 28% in plot, did not really require release. Apparently, a large proportion of trees of this size was already firmly established in the canopy. The Dawkins’ classification indicates a smaller proportion to be firmly established in the canopy; i.e. 31% and 12% respectively. One should realise, however, that the Dawkins’ crown class ‘3’, is a very broad class. It ranges from trees that are shaded only marginally by one partly overtopping tree to trees that are almost fully shaded by several partly overtopping trees. The prospect of realising a reduction in impeders by a release operation is restricted because the majority of the impeding neighbours are favoured trees themselves or are 230 Chapter 6: Simulated silvicultural treatment retained desirable trees of merchantable size. In fifty per cent of the cases in plot 4, no improvement was possible, while in plot 13, improvement was possible in 71% of the cases. Of 89 trees being shaded vertically or laterally, only 15 can be completely liberated from impeders (Fig. 6.8). Obviously, the position of favoured trees can be improved if a selection is made for the trees with the highest potential value, eliminating desirable trees of lesser value. This approach is not considered here. Figure 6.8 Frequency distribution of the number of neighbours competing with favoured trees for light, subdivided in neighbours belonging to the group of desirable and to the group of non-desirable trees (pooled plots). 6.3.3 Relationship between social position and received irradiation The Dawkins’ Crown Classification was significantly associated with the percentage of total irradiation received (ANOVA arcsin√p - transformed percentages), F4,101 = 17.04; p < 0.001; Fig. 6.9). Multiple comparisons showed that tree crowns classified as Dawkins’ class ‘4’ received significantly more light on average than other classes, while class ‘3’ trees received more light than class ‘2’ trees (Table 6.3). The number of observations of crown classes ‘1’ and ‘5’ were too small to produce any significant differences. The relative irradiation was also correlated with the number of neighbouring trees that overtopped the selected tree, for trees receiving only lateral light and trees that received some overhead light. Linear regression analysis for selected trees that were not fully illuminated showed a significant effect of the number of overtopping trees (partial correlation r = –0.25, t79 = –2.30, p < 0.05; Fig. 6.10). The amount of explained variation is small which is easily explained by the fact that the size and inner cover of the shading crowns are not taken into consideration. This effect is clearly demonstrated by the fact that some trees that classified as ‘receiving some overhead light’ (crown class = 3), actually received a rather small amount of irradiation, because the crown of the overtopping tree extended over the top of the favoured tree. 231 Reduced impact logging in the tropical rain forest of Guyana Table 6.3 Relationship between Dawkins’ Crown Classification and received irradiation. Mean proportion of potential irradiation (arcsin√p - transformed percentages) and probabilities for multiple comparisons between Dawkins’ Crown Classes (Tukey’s Honest Significant Difference test for unequal N). Dawkins’ Crown Class ‘5’ was only observed once and therefore omitted in multiple comparisons. ANOVA: F3, 101 = 21.37; p < 0.001. Dawkins’ Crown Class Mean relative irradiation {1} {2} {3} {4} 24.6 32.7 55.3 74.6 Figure 6.9 Dawkins’ Crown Class {1} {2} {3} 0.956 0.226 0.012 0.005 0.000 0.001 Relation between relative irradiation and Dawkins’ Crown Classification. Median (points), minimum and maximum values (whiskers) and upper and lower quartiles (boxes) of the percentage of potential irradiation for selected trees with five Dawkins’ crown classes. Regarding ‘competition’ by considering trees that protrude into an inverted cone with a 90° angle with its apex on the top of a favoured tree, also showed a significant correlation with the relative illumination for trees that received at least some overhead light. Linear regression analysis showed a significant effect of the number of trees qualifying impeding according to the definition used (partial correlation r = –0.43, t101 = –4.71, p < 0.001; Fig. 6.11). Extension of the angle of the inverted cone to 120° did not improve the relationship as shown by a smaller partial correlation (r = –0.38). The three relationships reveal that the illumination of a favoured tree can be characterised by its crown class (Dawkins’ crown classes 1 to 3 inclusive) and, once they receive at least some overhead light, by the number of trees, of which the crown projects into a 90° inverted cone extending from the top of the favoured tree. 232 Percentage of potential irradiation Chapter 6: Simulated silvicultural treatment 100 80 60 40 20 0 1 2 3 4 Crown class: 1 5 1 2 3 4 Crown class: 3 5 1 2 3 4 Crown class: 2 5 100 80 60 40 20 0 Number of overtopping neighbours Figure 6.10 Relation between illumination, on the one hand, and number vertically shading neighbouring trees and crown class, on the other. Crown classes are: 1. no direct light, 2. lateral direct light only, 3. some overhead light. Linear regression analysis using two dummy variables for crown class produced the following best fitting model. p = 35.37** + 9.25ns ⋅ [M1] + 28.94* ⋅ [M2] – 4.31* ⋅ Ni ; with R2 = 0.27, where p = relative irradiation (arcsin√p - transformed). Ni = number of neighbouring trees casting overhead shade. M1 and M2 are dummy variables coded 0 or 1. If Crown Class = 1, then M1 = 0 and M2 = 0; if Crown Class = 2, then M1 = 1 and M2 = 0; and if Crown Class = 3, then M1 = 0 and M2 = 1. Hence, trees that either shade a favoured tree vertically or of which the crown projects into the inverted cone can be qualified as impeders and are used in this study to evaluate the efficacy of the different treatments. 6.3.4 Social position and tree characteristics The diameter of a favoured tree appeared to be poorly correlated with crown illumination in plot 13, but somewhat better in plot 4. A non-linear model of the form p = 1 / ( 1 + a ⋅ e ( b ⋅ d ) ) was fitted for both plots, with p = percentage of potential irradiation, d = diameter of favoured tree, and a and b fitted constants. Such a model was fitted by minimising variance through an iterative procedure (StatSoft 1994). The non-linear model explained 35% of variance in case of plot 4 and only 5% in case of plot 13 (Fig. 6.12). The diameter of the favoured tree thus appears to be a poor indicator of crown illumination. 233 Reduced impact logging in the tropical rain forest of Guyana 100 Percentage of potential irradiation 80 60 40 20 0 0 2 4 6 8 10 12 14 16 0 2 4 Crown class: 1 6 8 10 12 14 16 12 14 16 Crown class: 2 100 80 60 40 20 0 0 2 4 6 8 10 12 14 16 Crown class: 3 0 2 4 6 8 10 Crown class: 4 Number of Shading Neighbours (90° inverted cone) Figure 6.11 Relation between illumination, on one hand, and number of neighbouring trees that protrude into an inverted cone with 90° angle expanded from the top of selected trees and crown class, on the other. See Fig. 6.10 for crown classes 1-3. Class 4 = full overhead light. Linear regression analysis using three dummy variables for crown class produced the following best fitting model. P = 38.42*** + 3.04ns ⋅ [M1] + 22.44* ⋅ [M2] + 38.28*** ⋅ [M3] – 3.08*** ⋅ Ni ; with R2 = 0.50, and p = percentage of potential irradiation arcsin√p - transformed), M1, M2 and M3 dummy variables coded 0 or 1 and Ni = number of neighbouring trees qualifying as impeders. See Fig. 6.10 for coding of desing varaibles. Relationships between the number of impeding neighbours and characteristics of the favoured trees, such as diameter or total height, were examined by fitting a negative exponential model. The criteria for a neighbouring tree to qualify as an impeder are discussed in Section 6.3.3. Also in case of the correlation between diameter and the number of impeders only modest amounts of variance were explained (Fig. 6.13); i.e. 13% in case of plot 4 and 24% in case of plot 13. The elevation of the top of the tree (total tree height plus base height), however, appeared to be a much better indicator of the amount of impeding neighbours. A fitted negative exponential model explained 35% of variance in case of plot 4 and 70% of variance in case of plot 13. The examined relationships indicate that basing prescriptions for releasing potential crop trees on the size of these trees is not appropriate, but relative tree height (absolute elevation) is to be preferred. The applicability of distance tables (e.g. Wadsworth 1958) to determine which neighbouring trees should be eliminated was evaluated by examining distances between favoured trees and apparently impeding neighbours, by examining respective diameters and relationships between respective diameters and distances. The frequency 234 Chapter 6: Simulated silvicultural treatment distribution of distances between favoured trees and impeding neighbours differed significantly between the two plots (Kolmogorov-Smirnov two-sample test; d = 0.21, p < 0.01), but the magnitude of the difference was small. The mean distance was 8.3 m (median = 7.4 m) in plot 4 and 9.5 m (median = 8.4 m) in plot 13. Seventy-five per cent of impeding neighbours was within 11 m in plot 4 and within 12 m in plot 13, while 90% were within 16 m and 17 m respectively. Figure 6.12 Fitted non-linear models for relative illumination as a function of the diameter of favoured trees: p = 1 / (1 + 59.28 ⋅ e (– 0.193 ⋅ d) ) (plot 4) and p = 1 / (1 + 2.56 ⋅ e (– 0.043 ⋅ d) ) (plot 13). The amount of variance explained is modest in case of plot 4; i.e. 34.9% and negligible in case of plot 13; i.e. 5.4%. Figure 6.13 Relationship between the number of neighbouring trees competing for light (ni) and the diameter (d) of favoured trees (a), and between the number of neighbouring trees competing for light (ni)and the elevation (h) of the top of favoured trees (b). Elevation is defined as the top height of the tree plus the height of the base of the tree. plot 4: ni = 15.61 ⋅ e (– 0.078 ⋅ d) and ni = 62.41 ⋅ e (– 0.089 ⋅ h) ; plot 13: ni = 23.56 ⋅ e (– 0.073 ⋅ d) and ni = 142.44 ⋅ e (– 0.097 ⋅ h) For the purpose of determining treatment prescriptions, not all of these neighbours should be taken into account, because impeding trees shade more than one favoured tree at the density and spatial distribution of favoured trees in this study. Therefore, the distances between impeding trees and the nearest favoured trees was examined (Fig. 6.14). Focusing on nearest neighbours alone, 75% of impeding neighbours were within 8 m in plot 4 and within 9 m in plot 13, while 90% were within 11 m in both plots. Systematic elimination of impeding trees within a fixed radius of 10 m would affect 85% of the impeders in both plots at the density and spatial distribution of favoured 235 Reduced impact logging in the tropical rain forest of Guyana trees as found in this study. However, a prescription for elimination with a fixed radius of 10 m may miss out a considerable number of impeding neighbours when there are less favoured trees or when the distance between the favoured trees is large (clumping). Figure 6.14 Frequency distribution of the distance between trees competing for light and the nearest favoured tree for the two experimental plots. The main reasoning behind the use of distance tables is the assumption of a correlation between the size of the impeding tree and its distance to the favoured tree and, conversely, between the size of the favoured tree and its distance to its impeders. The correlation between the distance between impeding trees and the nearest favoured tree, on the one hand, and the respective diameter of the impeding tree, on the other, was weak but significant in plot 4 (Pearson r = 0.36, p < 0.01), and completely absent in plot 13 (Pearson r = 0.03, p = 0.76). There was no correlation between the distance between impeding trees and the nearest favoured tree, on the one hand, and the respective diameter of that favoured tree, on the other, in either plot (Pearson r = 0.21, p = 0.08 and Pearson r = –0.01, p = 0.95). These results indicate that the prospect for pinpointing distances of impeders based on the diameter of the favoured tree is poor in this forest. The use of the diameter of potentially impeding trees appears to be applicable in plot 4. There, tree configuration and dimensions appear to be comparable to what Jonkers (1987) found in Suriname; i.e. elimination in a 10 m radius knocks out the majority of the impeders with a diameter less than 50 cm, but not the larger ones. In plot 13, however, a large proportion of impeders in the diameter class between 30 cm and 40 cm escape elimination with such prescription (Fig. 6.15). Hence, Jonkers’ rule is sometimes but not always appropriate in these forests. In neither plot was the diameter of the favoured tree correlated with the diameter of the smallest impeding tree; i.e. Pearson r = 0.19 (p = 0.21) for plot 4 and Pearson r = 0.03 (p = 0.25) for plot 13. The mean diameter of the smallest impeder was 37.8 cm dbh and 236 Chapter 6: Simulated silvicultural treatment 38.0 cm dbh respectively, regardless of the size of the favoured tree. Absence of correlation and the fact that the mean diameter of the selected trees was 27.8 cm in plot 4 and 28.0 cm in plot 13, indicates that impeding trees were in general larger than favoured trees. The latter difference was greatest for the smallest of the favoured trees and gradually grew smaller with increasing diameter (Fig. 6.16). This indicates that prescriptions based on the assumption that such a correlation exists will ignore the fact that, once they are tall enough, relatively thin (younger) trees do impede relatively big (older) trees. On the other hand, this suggests that a lower limit of 20 cm for non-desirable trees to qualify as impeders is sufficiently low. Plot 4 Plot 13 25 distance to nearest favoured tree (m) distance to nearest favoured tree (m) 25 20 15 10 5 0 20-30 30-40 40-50 50-60 60-70 >= 70 20 15 10 5 0 20-30 diameter impeder (cm dbh) 30-40 40-50 50-60 60-70 >= 70 diameter impeder (cm dbh) Figure 6.15 Relationship between the diameter class of trees competing for light and the distance to the nearest favoured tree. Points indicate the median distance; boxes indicate the range of distances in which 80% of the trees in a particular diameter class fall and whiskers the total range of distances in a class. Figure 6.16 Relationship between the size of the favoured tree and its smallest impeder. The solid line signifies the correlation between the two: dimp = 26.85 + 0.40 ⋅ dpct. The intercept coefficient differs significantly from zero (t87 = 4.05, p < 0.001), but the slope coefficient does not (t87 = 1.71, p = 0.09), while the coefficient of determination is extremely small r2 = 0.03. The broken line signifies total correlation dimp = dpct . Most points in the scatterplot are situated above this line. 237 Reduced impact logging in the tropical rain forest of Guyana 6.3.5 Crown size of favoured trees and impeders The distance between the trunk of a favoured tree and the trunks of its impeders was less than 10 metres on average. However, it was found that an impeder could be up to 25 metres away and still cast shade on a favoured tree. Examination of tree diameters and distances did not produce any firm relationships on which to build prescriptions for elimination. Another way to approach this problem is to examine whether a firm relationship exists between crown size and tree diameter. An additional reason to examine this relationship is to establish the amount of crown space needed by potential crop trees during its development until reaching maturity. The relationship between bole diameter and crown radius varies from species to species according to there crown architecture, but also to the direction taken by crown reiteration processes, which by definition are unpredictable (Oldeman 1974a). Obviously, the artificial division into commercial and non-commercial species does not indicate a difference in crown architecture. For this particular purpose, however, the division into these two groups may be a useful guidance in determining the amount of crown space likely to be needed by potential crop trees and the amount of crown space likely to be occupied by their undesirable neighbours. Traditionally, polynomial regression models are fitted to relate crown size to bole diameter (e.g. Husch et al. 1982; Jonkers 1987). Polynomial equations were obtained by using crown radius (based on crown projections intended for use in the three-dimensional forest model) and diameter measurements for trees of which the crown was not damaged by felling (Table 6.4) Table 6.4 Relationship between crown radius (r) and bole diameter (2r) for commercially desirable and non-desirable species. Regression model R = b0 + b1 ⋅ d + b2 ⋅ d2, where R = crown radius (m) and d = bole diameter (m dbh). Symbols behind the coefficients indicate whether they differed significantly from zero; ns: p > 0.05; *: p < 0.05; **: p < 0.01; and *** p < 0.001 Species category Desirable non-desirable b0 0.30 ns 1.52 * Fitted coefficients b1 12.69 *** 9.61 ** b2 –4.50 * –1.61 ns N R2 399 225 0.65 0.40 The coefficient of determination is reasonable for the category of desirable species, but poor for the non-desirable species, indicating a variety in crown sizes among the latter group. This is no surprise, because the group contained species of different stature at maturity, varying from ‘lower-canopy’ trees to emergents, whereas desirable species all belong to the canopy or emerge above it at maturity. Crowns appeared to be larger than the sizes found by Jonkers (1987) in Suriname. A polynomial regression equation predicts y-values accurately within the range of measured tree sizes. It is, however, not very useful in predicting the expected maximum crown radius. A general logistic model will give a more accurate prediction of the maximum due to its asymptotic nature. General logistic models of the form R = a / (1 + b ⋅ e ( c ⋅ d ) ) (StatSoft 1994) were fitted for the two species groups to predict the maximum crown radius (Fig. 6.17). Fitting is done by minimising the residual sum 238 Chapter 6: Simulated silvicultural treatment of squares in an iterative way. In general, crowns of non-desirable species spread wider than crowns of desirable species, the maximum radius of the former is predicted as 9.4 m and of the latter as 8.4 m. The crown space for a potential crop tree can be regarded as the space it would need at a diameter of 45 cm dbh (felling limit for practical purposes). According to the non-linear model, the amount of crown space is indicated by a radius of 5.1 m. Consequently, if a fixed radius of 10 metres were used to decide on elimination, trees with a diameter ≥ 37 cm located beyond this radius, would prevent the favoured tree from obtaining the desired crown space. Figure 6.17 Relationship between crown radius and bole diameter for the categories of desirable and nondesirable species. The relationships are described by non-linear equations; desirable species: R = 8.39 / (1 + 5.26 ⋅ e (–4.68 ⋅ d ) ); (R2 = 0.65), and non-desirable species: R = 9.39 / (1 + 3.80 ⋅ e (–3.80 ⋅ d) ); (R2 = 0.40). Whether crowns of non-desirable species were significantly larger was tested by fitting a linear model with a dummy variable [C], coded 0 for desirable and 1 for non-desirable species. The following equation was fitted: R = 1.08 + 0.66 ⋅ [C] + 8.66 ⋅ d - 0.31 ⋅ [C] ⋅ d. The dummy variable [C] had no significant effect on the slope of the line (coefficient fourth term: t621 = –0.42, p = 0.67), but raised the intercept significantly (coefficient second term: t621 = 2.69, p < 0.01), implying that the crown radius of a non-desirable tree is an estimated 0.7 m wider than the one of a desirable tree. The division in desirable and non-desirable species categories does not really stand up, because overmature, defective trees of desirable species are viewed as non-desirable individuals as well. Therefore, it is more appropriate to investigate the crown radius / bole diameter relationship on the species level. General logistic models were fitted for the four most abundant species; i.e. Lecythis confertiflora, Chlorocardium rodiei, Mora gonggrijpii and Catostemma fragrans (Fig. 6.18). 239 Reduced impact logging in the tropical rain forest of Guyana Figure 6.18 Relationship between crown radius and bole diameter for four abundant species. General logistic models could be fitted for all species, except Lecythis, for which only a linear model could be fitted. Fitted equations are: Catostemma: R = 4.22 / (1 + 66.19 ⋅ e ( –21.12 ⋅ d ) ); n = 43, R2 = 0.60; Chlorocardium : R = 9.34 / (1 + 3.88 ⋅ e ( –4.07 ⋅ d ) ); n = 107, R2 = 0.63; Lecythis: R = 1.68 + 9.47 ⋅ d ; n = 112, R2 = 0.47; Mora: R = 8.02 / (1 + 5.17 ⋅ e ( –5.36 ⋅ d ) ); n = 91, R2 = 0.65. It is shown that all species except C. fragrans attain a crown radius of approximately 8 metres at maturity (60 to 80 cm dbh). C. fragrans generally does not attain a large diameter (its commercial potential being restricted thereby). It is indicated that the crown of this species remains relatively small. L. confertiflora and M. gonggrijpii many individuals of the latter species do not qualify as desirable trees because of a high percentage of hollow trees (see Section 4.4.2; De Milde 1970) - , attain a large crown size relatively quicker than C. rodiei. When reaching a diameter of 45 cm dbh L. confertiflora is predicted to have obtained a crown radius of 5.9 m, M. gonggrijpii of 5.5 m and C. rodiei of 5.1 m. If elimination is carried out within a 10 m radius alone, a C. rodiei tree may be hindered in obtaining desired crown space by M. gonggrijpii trees beyond this radius with a diameter of 39 cm dbh and L. confertiflora trees with a diameter of 34 cm dbh. C. fragrans appears to need less crown space; i.e. a predicted crown radius of 4.2 m when attaining a diameter of 45 cm dbh, and M. gonggrijpii trees with a diameter exceeding 49 cm or L. confertiflora trees with a diameter exceeding 44 cm may prevent the realisation of the desired crown space from beyond a 10 m radius. 240 Chapter 6: Simulated silvicultural treatment The significance of the differences in crown radius between the four species was tested by fitting a linear regression model with three dummy variables. Table 6.5 shows that C. fragrans indeed has a relatively smaller crown than the other species at a certain diameter. C. rodiei has a crown size similar to M. gonggrijpii and L. confertiflora when small, but the crowns of the latter two species spread out more widely when the diameter increases. L. confertiflora has a larger crown than M. gonggrijpii at a certain diameter. Coefficients of determination obtained with linear regression hardly differ from the ones obtained with non-linear regression, but it is not possible to predict the maximum crown expansion with the linear model. Table 6.5 Relationship between crown radius (r) and bole diameter (2r) for four desirable species. Regression model R = a + b ⋅ d, where a = a1 + a2 ⋅ [C] + a3 ⋅ [M] + a4 ⋅ [L] and b = b1 ⋅ d + b2 ⋅ [C] ⋅ d + b3 ⋅ [M] ⋅ d + b4 ⋅ [L] ⋅ d. R = crown radius (m), d = bole diameter (m dbh). [C] is a dummy variable for Catostemma fragrans, [M] is a dummy variable for Mora gonggrijpi, [L] is a dummy variable for Lecythis confertiflora, and ai and bi are fitted coefficients. Dummy variables are coded 1 if an individual belongs to that particular species and 0 if it does not belong to it. Chlorocardium rodiei is indicated when all dummy variables equal 0. Letters behind a coefficient refer to inter-specific comparisons (occurrence of the same letter with two species signifies no difference between the two species in regard to that coefficient). Symbols behind the coefficients indicate whether coefficients differed significantly from zero; ns: p > 0.05 and *** p < 0.001 Species Catostemma Chlorocardium Lecythis Mora Fitted coefficients a 0.41 ns 1.79 *** 1.68 *** 1.26 *** b 10.42 7.18 9.47 8.99 a b bc bd *** *** *** *** ab a b b N R2 43 107 112 91 0.54 0.63 0.47 0.64 Approaching competition by using the relationship between crown size and tree diameter appears to provide workable rules in contrast to what was found using actual distances and crown configuration. This is explained by the fact that the centre of a crown more often than not is not located directly above the base of the tree. Instead, the point of gravity of a crown is usually eccentric, while top and periphery may extend far and in any direction from there. Another reason is that tree height is not strongly related to tree diameter in the diameter range considered here. Apparently the majority of these trees has reached the canopy and has commenced or is commencing to expand their crown. Finally, the topography plays a role adding base height to tree height and thereby blurring the relation between competition and tree characteristics. 6.3.6 Evaluation of different prescriptions for release The effect of the simulated silvicultural treatments is visualised in Figures 6.19 and 16.20. It is clear that all types of systematic treatment are heavier than needed as shown by the result of selective elimination, which is based on the elimination of neighbouring trees competing for light. Systematic elimination with a diameter limit of 20 cm dbh (R20), within a radius of 10 m around favoured trees (L10) and the combined treatment with a systematic elimination of all non-desirable trees exceeding 40 cm dbh and of smaller non-desirables within a radius of 10 m, all appear to be unnecessarily heavy. The other two treatments are closer to the ideal treatment but the question is whether these removed the right trees. 241 Reduced impact logging in the tropical rain forest of Guyana 242 Chapter 6: Simulated silvicultural treatment Figure 6.19 Top view of plot 13 (trees with dbh ≥ 20 cm) as observed after logging without treatment and after five different treatments. Plot size is 140×140 m2. Grey crown projections denote favoured trees; white projections with solid border desirable trees of merchantable size and with broken border non-desirable trees.(a) situation two years after logging; (b) ‘selective release’ (crown thinning); (c) systematic elimination with diameter limit 20 cm; (d) as (c) diameter limit 30 cm; (e) elimination in fixed radius of 10 m; (f) as (e) radius of 7 m. Figure 6.20 Side view of a 20 m wide and 140 m long transect located in plot 13. See Fig. 6.19 for description of treatments, except (g) after systematic elimination with a diameter limit of 40 cm and systematic elimination within a 10 m radius around favoured trees with a diameter limit of 20 cm. Height of the frames is 60 m. For crown projections see Fig. 6.19, the transect covers x = 0 - 140 m and y = 80 - 100 m. 243 Reduced impact logging in the tropical rain forest of Guyana The effect of the different treatments on the average exposure of the crowns of the favoured trees in the central 100×100 m2 was tested by Repeated measures ANOVA using arcsin√p - transformed percentages of potential illumination, followed by pairwise comparisons. Figure 6.21 shows that the mean exposure of favoured trees differed between treatments (F5,530 = 35.40, p < 0.001). It also shows a considerable difference in the average illumination of favoured trees between the two plots (F1,106 = 7.03, p < 0.01) which is consistent with the different treatments (interaction plot × treatment; F5,530 = 1.97, p = 0.08). Absence of interaction between the plots and the treatments implies that treatments are not sensitive to differences in stand composition; e.g. the fact that plot 13 had twice the amount of retained merchantable trees did not affect differences between treatments. It also allows pooling of the plots for further analysis. Percentage illumination (transformed) 75 Plot 4 70 65 60 55 Plot 13 50 45 Control L7 R30 L10 Treatment R40+L10 R20 Figure 6.21 Effect of simulated treatments on the percentage of potential irradiation received by favoured trees. Treatments are systematic elimination with a diameter limit of 20 cm (R20) and 30 cm (R30), localised systematic treatment around favoured trees with a radius of 7 m (L7) and 10 m (L10) and a combined treatment; blanket elimination of non-desirables with dbh ≥ 40 cm and localised elimination of non-desirables with 20 cm ≤ dbh < 40 cm (R40+L10). Points and lines indicate mean percentages (after angular transformation), boxes standard errors and whiskers 95% confidence intervals. Pairwise comparisons showed that all treatments had a beneficial effect on the exposure of the favoured trees, but that there were few differences between treatments. Elimination in a fixed radius of 7 m (L7) resulted in a significantly lower mean exposure than the other treatments, except for systematic elimination with a diameter limit of 30 cm (R30) (Table 6.6). With regard to present 'competition', a localised treatment with a fixed radius of 10 m (L10) performs just as well as more intensive treatments such as R20 or L10 plus R40. It is thus not worthwhile to increase the intensity beyond this level. 244 Chapter 6: Simulated silvicultural treatment As shown in Table 6.6, the increase in mean exposure of favoured trees due to treatment is, though being significant, rather small. This is due to the fact that a large proportion of the crowns of favoured trees were already in a favourable position and the fact that a large proportion of the trees that compete with them for light are favoured trees themselves and hence were not removed. Table 6.6 Percentage of potential illumination received by favoured trees with five different simulated treatments. Median of raw percentages, mean (sd) of arcsin√p - transformed percentages with both plots pooled (n = 106). Probabilities produced by pairwise comparisons were adjusted with the Bonferroni method to limit the experiment-wise error rate (Sokal and Rohlf 1995). Letters indicate which pairwise comparisons produced significant differences at an experiment-wise error of α = 0.05. Treatment Median of percentages (%) 70.1 80.0 82.4 83.1 83.2 83.1 Control L7 L10 R30 R20 R40+L10 Percentage of potential illumination Mean transformed Standard deviation percentage (degrees) (degrees) 56.8 20.2 62.8 17.0 64.6 16.5 64.1 17.5 65.0 16.5 64.9 16.5 Pairwise comparisons a b c bc c c 70 L7 R30 L10 R40+L10 R20 Percentage improvement 60 50 40 30 20 10 0 0-20 20-40 40-60 60-80 Percentage illumination before treatment >=80 Figure 6.22 Improvement in illumination of crowns of favoured trees (added percentage illumination) by five simulated treatments in relation to the illumination status of these trees before treatment. Points are mean percentages, boxes are standard errors and whiskers are 95% confidence intervals. Values were obtained by computations using angularly transformed percentages and transforming them back into percentages. For treatment codes see Figure 6.21. Relative improvement of a tree’s crown position is marginal if the initial illumination was more than 60%. Treatments L7 performs less than the other treatments, except if the initial illumination was less than 20%. In that case, R30 performed the least. Figure 6.22 shows that the differences between the treatments manifest themselves mainly if the exposure before treatment was poor. The small differences between the 245 Reduced impact logging in the tropical rain forest of Guyana means are clearly a result from the fact that a large proportion of the favoured trees already enjoyed favourable conditions from the start. It is, therefore, concealed that the effect on the exposure of trees subjected to less favourable conditions, differed strongly between certain treatments. Treatments were also compared on account of their effectiveness in eliminating impeders and the amount of liquidated basal area (‘investment’). The effectiveness of the different treatments (number of eliminated impeders / total number of impeders) has a maximum of 53% (Table 6.7). This rather low value is due to a large proportion of the impeding trees actually being favoured trees themselves or retained merchantable trees. Fully releasing trees of potential highest value, as in the traditional selection system, is not considered an appropriate approach, because of the length of the cutting cycle (at least 25-30 years) that is foreseen. It is hard to predict changes in marketability of the different species over such a period and in addition, it would narrow down the amount of favoured trees to unacceptably low levels (see Section 6.1). Table 6.7 Efficiency of five different treatment prescriptions relative to the ideal treatment where all (non-desirable) impeding neighbours have been removed. The ‘investment’ refers to removed trees and is given in number of trees and corresponding basal area removed per hectare by each treatment. The investment is open to multiple interpretation (see text). Effectiveness = the number of impeding neighbours actually removed / the total number of impeding neighbours. Accuracy = the number of removed impeding neighbours / all eliminated trees. The opposite of accuracy ( = 100% - accuracy) is a measure of effort wasted on trees that actually were not impeders. Subject of evaluation Investment Method of elimination N/ha BA/ha reference selective 42 L7 56 R30 51 simulated treatments L10 R40+L10 85 87 R20 94 7.2 6.6 8.9 9.9 10.5 10.8 Effectiveness 53% 39% 41% 50% 52% 53% Relative effectiveness Accuracy 100% 74% 78% 95% 98% 100% N/ha 100% 54% 68% 48% 48% 44% BA/ha 100% 74% 77% 70% 69% 66% Relative to the ‘ideal’ method of release; i.e. selective release based on crown thinning principles, it is found that treatments L10, R40+L10 and R20, the latter by definition, come near to the desired result. In fact, these values are closely related to the results found for the percentage of maximum attainable illumination, because the criteria used to identify impeding neighbours are based on relative illumination. It is interesting to note that the expectation that a fair proportion of impeding trees (Section 6.3.4: 15%; Fig 6.14) would escape elimination if treatment is restricted to a 10 m radius is only found back as a marginal difference in this comparison. This is explained by the fact 246 Chapter 6: Simulated silvicultural treatment that most impeders that are situated beyond this limit are retained merchantable trees and are not removed in the reference treatment (crown thinning) either. Treatments L10, R40+L10 and R20, treatments that do not differ significantly in mean relative illumination, approached the highest attainable level of release, but did so at a cost of at least twice of that of the ideal treatment in terms of number of eliminated trees and at least 150% in terms of eliminated basal area. Treatment L7 requires a much smaller investment - even less than selective elimination in terms of basal area removed, but was ineffective. Treatment R30, which did not differ significantly from the more intensive treatments regarding improvement of the exposure of favoured trees (Table 6.6), appears to be ineffective when the number of impeding trees is used as a reference. This ostensible contrast suggests that trees smaller than 20 cm dbh were competing less vehemently for light than their larger counterparts. Therefore sparing them is reflected in a reduction of the effectiveness in regard to the number of eliminated trees, but not in a reduction of the mean relative exposure (cf. Fig. 6.22). This treatment is hence apparently sufficiently effective with regard to present competition, but ineffective with regard to future competition. The investment can be interpreted as a true cost reflecting the effort needed to accomplish the treatment. It can also be interpreted as the biological cost reflecting a reduction in biomass, sequestrated carbon or biodiversity. Conversely, it can reflect the increased availability of nutrients from which favoured trees may benefit. A higher ‘investment’ was associated with a lower accuracy. The most ‘expensive’ - and most effective - prescriptions also had the lowest accuracy; i.e. removing many trees to no avail. The ‘cheaper’ prescription of localised treatment within a radius of 7 m is rather wasteful as well. The blanket treatment with a diameter limit of 30 cm appeared to be the least wasteful. It is concluded that none of the examined prescriptions comes even near to the desired result. Treatment L7 dropped out because, although being the least costly in terms of eliminated basal area, it was significantly ineffective both in raising the average relative illumination and in reducing numbers of impeding neighbours. Treatment R40+L10 and R20 do not qualify as an alternative to selective elimination, because they were not significantly more effective than the L10 treatment, but required a higher investment and portrayed a lower accuracy than that treatment. This comparison applies to present competition. If future competition is considered the R40+L10 treatment may be preferred above the L10 treatment. Treatment L10 itself neither is an alternative for true selective release, because it is still highly wasteful. Treatment R30 may be the best alternative on account of accuracy, effectiveness in terms of mean relative illumination and investment. However, by leaving smaller impeding trees untouched, it does not cater for future competition. In general, ‘blanket’ prescriptions, such as R20, R30 or R40+L10 should not be advocated. The latter two may provide a good alternative in the situation that was studied, where density of 247 Reduced impact logging in the tropical rain forest of Guyana potential crop trees was high. In areas where potential crop trees are scarce, the accuracy of such treatments drops tremendously at an unwarranted biological cost. Ignoring future competition for the moment, localised elimination within a radius of 10 m gives the most promising result. However, the prescription has to be amended to increase its accuracy and reduce its financial and biological cost. Field observations have shown that it is not feasible to distinguish impeding neighbours properly from the ground, especially not those trees that are located further away. However, it has proved feasible to distinguish differences in top and periphery height for trees within a radius of 10 m. By not eliminating neighbouring, non-desirable trees, of which the total height or the height of the crown periphery is lower, the accuracy of the treatment can be increased considerably. The number of eliminated trees dropped to 59 trees per hectare with a corresponding basal area of 8.2 m2/ha, raising the accuracy by 40% to 67% in terms of the number of justifiably eliminated trees and by 15% to 81% in terms of justifiably eliminated basal area. 6.4 SILVICULTURAL CONSIDERATIONS 6.4.1 Why apply silvicultural treatment? Sustainable timber production from tropical rain forests can only be achieved if extraction levels are adapted to the potential productivity of these forests. Even when the logging method and cutting level are adapted (per species) to potential productivity, timber extraction will still deprive the canopy of commercial trees. Therefore, recovery of commercial resources will be slow, the more so because removal of seed trees of commercial species may lead to progressive impoverishment of the numerical importance of these species in the ecosystem. Silvicultural post-harvest treatments provide an opportunity to shorten the cycle in which the commercial resources are reconstituted. Silvicultural operations that release commercially desirable juveniles from the competition by non-desirable elements, such as climbers, trees belonging to non-desirable species and senescent trees of desirable species, serve the following specific purposes: ƒ they restore the initial balance of species in the canopy, which has become unbalanced by systematic removal of trees belonging to desirable species; ƒ they stimulate the growth of commercially desirable trees; especially those that will constitute the next crop; and ƒ they boost the recruitment of commercially desirable species. These strategies together are likely to accelerate the reconstitution of the exploitable stock in logged forest and will increase the chances on (higher levels of) sustainable timber production. In the present study, I concentrated on the potential effect of different silvicultural prescriptions on stimulating growth rates of potential crop trees by simulation of treatments. 248 Chapter 6: Simulated silvicultural treatment 6.4.2 Selection of the ‘best’ method: the empirical way Silvicultural treatments are usually directed at increasing the crown exposure of potential crop trees to light. In designing an appropriate silvicultural treatment, the essential question is how to evaluate its effectiveness. The approach followed in the present study was to classify the environment of potential crop trees according to diffuse and direct irradiation received by the top of the crown, relative to irradiation received by a point outside the forest. These amounts were estimated by constructing a three-dimensional forest model of two logged stands and subsequently simulating hemispherical photographs ‘touching’ the tops of the crowns. The programme SILVI-STAR was used for this purpose. The different treatments were evaluated on the amount of received irradiation, estimated from these simulated photographs. In determining the efficiency of a certain treatment, it is of equal importance to assess the proportion of impeding trees that was removed and the proportion of removed trees that in reality was not competing for light. In order to achieve this, impeders were identified. A number of environmental indicators were selected and their correlation with the estimated amount of received irradiation was investigated. In principle, environmental indicators should be linked to individual growth rates and thereby function as a growth indicator. The Kraft and Dawkins crown classifications have earned a reputation of being indicators that are associated with potential growth rates. There is sufficient evidence that the Dawkins crown classification is correlated with growth rates (e.g. Silva 1989; Favrichon 1991; Gourlet-Fleury 1997). Tree measurements that have been carried out since 1993 in the experimental plots in Pibiri, have been accompanied by assigning Dawkins crown classes. Relating individual mean annual increment over the three year period after logging in 1994 to the crown position of that individual, also provided evidence for using this crown classification for growth prediction (Van der Hout, unpublished data). Summarised results for selected commercial species are found in Table 6.8. This study showed that Dawkins crown classes are also significantly correlated with the actual amount of radiation received by the top of the focal trees. The crown class could therefore be used to identify neighbours that compete for light. If a tree received less than full overhead light, those trees that overtopped the focal tree were thus earmarked as impeders. It was further shown that, once a favoured tree received at least some overhead light, the number of trees of which the crown projected into a 90° inverted cone extending from its top was correlated with relative illumination as well. Having identified neighbouring trees that compete for light, a reference (‘ideal’) treatment - referred to as ‘selective release’ - was formulated. Five systematic release prescriptions (see Section 6.1.2) were tested against this ‘ideal’ treatment. The silvicultural treatment to be designed is linked with a reduced impact logging operation with an intensity of 8 trees per hectare. This level of extraction is considered potentially sustainable (see Section 4.7.7). In Section 6.3.2, it was deducted that in such a logging regime trees smaller than 28 cm dbh were not likely to be recruited into the felling limit diameter classes within a 30 year period, even when released from 249 Reduced impact logging in the tropical rain forest of Guyana shading neighbours. It is still considered worthwhile, however, to release desirable trees down to 20 cm dbh, because timber production per unit area will benefit from enhanced growth by these individuals. Table 6.8 Crown class Mean annual increment rate by crown position for five common commercially desirable species (dbh≥20 cm) in the Pibiri experimental plots (total area = 30 ha). Analysis of variance of the effect of crown position on increment rate with diameter as co-variable showed that increment rates differed significantly by crown position for all five species; Chlorocardium rodiei (F4,893 = 9.60; p < 0.001), Carapa guianensis (F4,182 = 9.40; p < 0.001), Mora gonggrijpii (F4,235 = 6.73; p < 0.001), Catostemma fragrans (F4,373 = 6.55; p < 0.001) and Swartzia leiocalycina (F4,150 = 2.79; p < 0.05). The table shows estimated marginal means (corrected for diameter) of increment rates in mm, with standard errors (SE) and number of observations per class by species (n). Letters behind the means indicate pairwise comparisons (experiment-wise error α = 0.05) mean 1 2 3 4 5 1.1 1.8 2.1 2.5 2.6 1 2 3 4 5 0.5 1.3 3.1 4.0 5.8 1 2 3 4 5 0.3 0.6 1.4 2.1 2.4 SE n Chlorocardium rodiei 0.3 17 0.1 101 0.1 537 0.1 239 0.5 5 Mora gonggrijpii 1.1 6 0.5 29 0.2 120 0.3 84 1.9 2 Swartzia leiocalycina 1.2 2 0.5 11 0.3 46 0.2 88 0.6 9 mean a ab b c abc 0.2 1.6 3.9 5.3 8.8 ab a bc c abc 0.4 1.3 1.9 2.4 3.8 SE n Carapa guianensis 2.7 1 0.7 14 0.3 89 0.3 81 1.5 3 Catostemma fragrans 0.6 8 0.3 46 0.1 172 0.2 143 0.6 10 abc a b cd d a a ab bc c a a a a a The first question that was answered was whether these trees are really in need of liberation. It was found that 28% to 46% of the favoured trees received more than 80% of the potential irradiation. Using the Dawkins crown class as a reference, it was found that 12% to 31% received at least full overhead light. This discrepancy has three origins. First, computation of irradiation was carried out for the top of the tree and hence disregards shading of other parts of the crown in cases where the tree providing the shade does not project beyond the top of the favoured tree. The Dawkins crown class does include this competition. Secondly, Koop and Sterck (1994) have shown that for application in structurally complex canopies SILVI-STAR may overestimate the actual amount of light. This is caused by for instance the omission of lianas and epiphytes in the model and the visual assessment of the inner crown cover. Thirdly, the Dawkins crown class ‘3’, which refers to trees that receive less than 100% overhead light, is a broad class. It includes tree crowns that are only marginally shaded by only one neighbouring tree as well as crowns that suffer almost complete overhead shade cast by several neighbouring trees. 250 Chapter 6: Simulated silvicultural treatment Another important observation was that about half of the identified impeders were favoured trees themselves or retained desirable trees of merchantable size. This obviously makes full release of favoured trees impossible. 6.4.3 Selection of the ‘best’ method: the deductive way The space occupied by a single tree during its lifetime is defined by Oldeman (1974b), with Whittaker’s (1970) term, as the tree ecotope. Morphological criteria for the observation of trees are integrated into architectural growth models, which can be considered as differentiated branching complexes (Hallé et al. 1978). Such a model can be seen as the genetic blueprint of growth of a tree. Before trees have reached the canopy their crown will extend and expand according to this branching pattern and a tree’s diameter, total height and crown dimensions are strongly related to each other during this phase. Once a tree has established itself in the canopy, it will start to expand its crown by complete or partial reiteration of the architectural model (Oldeman 1974a; Hallé et al. 1978). The expansion of the crown and its diameter growth are no longer associated with tree height at this point. Since reiteration processes are induced by a variety of exogenous influences, the direction of crown expansion becomes unpredictable. The ecotope of a tree hence is predictable only as long as a tree grows according to its architectural model. This blueprint for the development of the space required by a tree at a certain size, even before its formal architectural definition, has inspired silvicultural rules for many decades. Using the three-dimensional forest model, it was examined whether such rules apply to the potential crop trees and their impeders. Crown exposure of potential crop trees was poorly correlated with its diameter. Tree diameter explained 35% of variance in crown exposure in one plot and only 5% in the other. In addition, the number of impeders appeared to be poorly correlated with diameter of focal trees ( < 24% of variance explained). The total height (absolute elevation) of focal trees showed a much stronger correlation with the number of impeders (35% to 70% of variance explained). The mean distance between the bases of trees competing for light and the nearest favoured tree was 5.8 m in one plot and 7.3 m in the other. Only 15% of the impeders were located further than 10 m away for the nearest favoured tree. There was no correlation between this distance and the diameter of the favoured trees. In one plot there was a significant, but modest correlation between this distance and the size of the impeder (Pearson r = 0.36). The conclusion that can be drawn from these findings is that ecotopes of neither favoured nor impeding trees, can be deducted from the diameters of these trees. On the other hand, it is indicated that competition is restricted to trees within a certain distance (95% of impeders is located within a distance of 15 m). The lack of correlation with diameter of either favoured or impeding trees, prohibits the identification of impeders based on a combination of its distance to the focal tree and the diameter of either tree. Height of the favoured tree, on the other hand, was a stronger predictor of competition. 251 Reduced impact logging in the tropical rain forest of Guyana These observations raised the question to know whether tree ecotopes are determined by crown size alone or whether other factors played a role. Relationships between tree diameter and crown radius were examined for three common desirable species and one common non-desirable species. It appeared that crown radius could be fairly well predicted by tree diameter for these four species (amount of explained variance varied between 47% and 65%). For instance, the estimated crown radius of Chlorocardium rodiei at a merchantable size of 45 cm dbh was found to lie between 4.0 m and 6.0 m (95% confidence interval). If these relationships could be used to identify impeders, this would also allow prediction of future competition. However, the three-dimensional forest model more or less denied the predictive capacity of crown size - diameter relationships. Inspection of the forest model showed that the reasons behind this discrepancy are: ƒ crowns are generally eccentric relative to the position of the base of the tree; ƒ the vertical position of a crown’s extremity (periphery) varies widely between and within species; ƒ tree height as such is not strongly correlated with tree diameter in the diameter range considered here (dbh ≥ 20 cm); and ƒ the topography (slope) plays an important role. Apparently the majority of the favoured trees have reached the canopy and are rather ‘trees of the present’ than ‘trees of the future’ (sensu Oldeman 1974a; Hallé et al. 1978). This was already suggested by the large proportion of favoured trees that were well-exposed before treatment. This is caused by the fact that reiteration processes are rather guided by exogenous (light) than endogenous (architectural model) factors, as soon as a canopy opening is formed in the vicinity of the tree. 6.4.4 Conclusion The main conclusion of this study is that systematic silvicultural treatments, be it over the entire area or localised around favoured trees, are in general effective in reducing competition, but are not accurate and hence waste financial and biological resources. Blanket treatments are not recommendable because of the risk of increased wastage in areas poorly stocked with potential crop trees (cf. Jonkers 1987). Treatments cannot be designed based on relationships between tree diameter and crown dimensions, because the structure of the canopy depends largely on additional factors such as variation in tree height, variation of the direction of crown expansion as induced by exogenous influences and topography (cf. Gourlet-Fleury 1997). Tree height is a better predictor of competition than tree diameter. The most promising prescription that came out of the comparison was localised elimination of undesirable trees within a radius of 10 m around favoured trees. However, an amendment to this prescription is indicated because of its low accuracy. It is therefore recommended to include ‘crown thinning’ principles for the selection of trees for elimination. This implies that undesirable canopy trees should be removed in order to favour the development of the potential crop trees if, and only if, they interfere with them. 252 Chapter 6: Simulated silvicultural treatment Because it is almost impossible to judge which neighbouring trees are actually competing for light due to the poor visibility of tree crowns in the complex rain forest canopy, it is recommended to use tree height as a criterion. A treatment thus formulated was applied in the field in December 1996 and future tree measurements may point out whether this treatment is appropriate. The treatment formulated disregards future competition. Crown size - diameter relationships found in this study suggest that elimination should be extended to a radius of 15 m to tackle future impeders. The predictive capacity of such relationships for future competition, however, appears to be poor as explained above. The question to know whether or not silvicultural treatment is desirable cannot be answered at this point of time. Since the reaction to release is slow, results from the applied treatment will not come forth until at least five years after treatment. In the neighbouring countries Suriname and French Guyana it was shown that treatments can enhance growth rates of favoured trees significantly (e.g. De Graaf 1986; Jonkers 1987; Favrichon 1997). Whether this will also be the case in Guyana’s Greenheart forests, at a cutting level of 8 trees per hectare, is still uncertain. In Guyana’s Greenheart forests, prospects for releasing potential crop trees are dimmed by the fact that about 50% of neighbouring trees competing for light are favoured trees themselves and by the fact that a considerable proportion of the crowns of favoured trees are already well-exposed. 253 Reduced impact logging in the tropical rain forest of Guyana 254 Chapter 7: Summary and general discussion 7. SUMMARY AND GENERAL DISCUSSION 7.1 DESIGN OF A REDUCED IMPACT LOGGING SYSTEM In this study the environmental problems of selective logging in Guyana’s Chlorocardium rodiei (Greenheart) - bearing forests were addressed. A reduced impact logging method was designed to redress these problems and the effect of introducing such measures on the cost of the logging operation was examined. Furthermore, the potential of silvicultural interventions after the first harvest was examined. Until 1993, when a market for plywood species emerged in Guyana, most selective logging focused on Greenheart-bearing forest stands. Forest composition in Guyana is quite variable (e.g. Ter Steege et al. 1993) and less than 20% of the forest is made up of Greenheart-bearing forest. Consequently, relatively small proportions of timber concessions were actually logged and average marketed volume per hectare was hence reported to be low; i.e. 5 m3/ha. Although the tree species composition of Greenheartstands varies across the area of natural distribution of the species (Fanshawe 1952; Ter Steege et al. 1993), the proportion of Chlorocardium rodiei in terms of volume is invariably high in these stands and not rivalled by any other commercial species, including plywood species. Prior to logging in the experimental area, which took place in 1994, 10% of all trees with a diameter of 10 cm and over, corresponding to 22% of the volume, belonged to Chlorocardium rodiei. On average, there were 21 harvestable trees per hectare; i.e. belonging to the currently commercial species, with a dbh ≥ 40 cm and containing a merchantable bole. Chlorocardium rodiei made up half of these stems and 49% of the harvestable volume. Including species that are potentially marketable, based on their wood properties, does not alter the situation much, because their density is low (which contributes to their present status of being unmarketable). It has been repeatedly reported that Chlorocardium rodiei is being overharvested, because the species share in the annual output from Guyana’s forests exceeds its relative abundance among commercial species (GNRA 1989; Ter Steege et al. 1996; Zagt 1997). This statement may be true, if one considers all of Guyana’s forest types, but, if only the Chlorocardium rodiei-bearing forests are considered, it must be concluded that: ƒ Chlorocardium rodiei is the most common commercial species in Greenheart forest and hence will remain the most important supplier of timber in these forests Whether or not the species is being overharvested should not be based on its share in annual output from Guyana’s forests, but on compliance with the silviculturally optimum harvest within the stands where it occurs naturally. The fact that Chlorocardium rodiei is the predominant commercial species in these stands does not imply that selective logging of Chlorocardium rodiei as practised nowadays is 255 Reduced impact logging in the tropical rain forest of Guyana sustainable. Three specific characteristics put serious constraints on sustainable timber production by the species. First, harvestable stems of the species occur in groves (locally named ‘reefs’), leading to large felling gaps. Secondly, the species’ growth rate is slow during its whole life cycle, and though being augmented by improved light conditions, many of its companion species show stronger responses to light (Zagt 1997). Thirdly, the natural size class distribution of the species shows a severe shortage in the medium size class range (10 cm ≤ dbh ≤ 30 cm). Matrix modelling showed that a second harvest couldn’t be expected within the next 100 years, if all currently harvestable stems were taken at the first cut (Zagt 1997). Conventional selective logging practice ignores silvicultural measures dictated by the constraints mentioned above. Typically, selective logging in these forests is taking all sound Chlorocardium rodiei stems above 45 cm dbh for sawn timber and, often in addition, stems down to 30 cm dbh for roundwood (to be used as pilings). Because of the gregarious distribution of Chlorocardium rodiei, conventional logging leads to a distinctly intermittent spatial distribution of large felling gaps connected by skid trails. Traces of skidder manoeuvring in felling gaps as a result of positioning of logs for hooking are common. Besides the absence of tree selection on a silvicultural basis, the following environmental problems are inextricably linked to conventional logging practice (Brouwer 1996; Ter Steege et al. 1996; Ek 1997). First, the size of the felling gaps is silviculturally undesirable. According to previous research, gaps should be kept small for the following reasons: ƒ to mitigate changes in micro-climate (Ter Steege et al. 1996, p.73); ƒ to provide small eco-units (Rossignol et al. 1998), needed by most commercial species for establishment, as these are non-pioneer species (Ter Steege et al. 1996, p.73); ƒ to allow the surrounding forest to buffer losses in nutrients and water through root absorption (Ter Steege et al. 1996, p.73); ƒ to allow the surrounding forest to buffer shortages in nutrients at a later stage through litter input (Ter Steege et al. 1996, p.73); ƒ to protect the vulnerable soil life from death by heating (Smits 1995); ƒ to discourage large gaps that favour the establishment of light-demanding trees and lianas, so slowing silvigenesis (‘liana blankets’; Kahn 1982; Oldeman 1990, his fig. 6.20). Secondly, haphazard skidding seriously affects established seedlings and saplings, and causes excessive soil disturbance, affecting forest recovery. Skidder movements should be particularly minimised because: ƒ skidding destroys seedlings, saplings and juvenile trees (Ter Steege et al. 1996, p.73); ƒ conditions for establishment and growth of seedlings are unfavourable on skid trails, because of: 256 Chapter 7: Summary and general discussion a. ƒ soil compaction (Brouwer 1996), which is a major handicap for root development and may lead to gully formation (Pinard et al. 1996; Munyanziza 1994); b. enhanced nutrient leaching, soil acidification and raised Aluminium concentration in the soil due to decreased root uptake of decomposing vegetation (Brouwer 1996), which strongly modifies the soil volume that roots can colonise and influences the diversity of soil biota (Keizer 1993; Munyanziza 1994; Nussbaum et al. 1995); c. heating and concomitant degradation of soil micro-life (Smits 1995); such disturbed soils barely allow other than germination of undesirable lightdemanding pioneer trees and lianas (Dalling et al. 1997; Guariguata & Dupuy 1997; Ek & Van der Hout, in prep.). A logging system was designed, inspired by the CELOS harvesting system (Hendrison 1990). It not only includes elements typical of reduced impact logging at large (e.g. Pinard & Putz 1996; Johns et al. 1996; Van der Hout & Van Leersum 1998) but also tree selection criteria. The main purpose of the latter innovation was to discourage the formation of large felling gaps. This was achieved by aiming at an even spacing of trees earmarked for felling. In contrast to the fixed legal felling limit of 34 cm dbh, a differential felling diameter limit was applied that was dependent on a species’ maximum diameter. If, after the first conditions had been fulfilled, more than one tree appeared to be eligible for felling, the abundance of the species in a plot (6 hectares) determined which trees should be felled. If this still left us with a choice between trees, those trees were selected that were overtopping future crop trees. It is realised that the tree selection criteria ignore aspects of scale. Species that are not abundant but portray a rather regular spatial distribution, such as Peltogyne venosa or Hymenaea courbaril, may rarely be selected for harvest using these criteria. On the other hand, gregariously growing species, such as Chlorocardium rodiei or Mora gonggrijpii, may be overly represented in the selection. On an operational scale, tree selection should therefore also include traditional allowable cut calculations that are based on forest inventory on a larger scale. As a reference for conventional logging, it was decided to use a ‘close to best practice’ operation instead of an unplanned, haphazard operation. The latter type of operation is usually poorly organised and equipped, and consequently inefficient. This certainly would interfere with the insight into the effects of the technological innovations. The differences between conventional and our environmentally sound logging are summarised in Table 7.1. The reduced impact logging system features a string of sequentially related activities, each depending upon the next one or the one before. Among these activities are a 100% enumeration and mapping of harvestable trees, liana cutting, skid trail planning and directional felling and winching. Directional felling can be carried out for a number of reasons. Felling a tree in a direction to ensure the safety of the felling crew and to prevent damage to the bole of the felled tree are the foremost reasons, but they do not specifically apply to reduced impact logging 257 Reduced impact logging in the tropical rain forest of Guyana alone. Three additional aims of directional felling can be distinguished that specifically belong to reduced impact logging, resulting in a total of five objectives: 1. to ensure the safety of the felling crew; 2. to prevent breakage of the felled bole, e.g. by falling on solid obstacles or downhill; 3. to avoid damage to potential crop trees (and crop trees); 4. to place logs in a position to facilitate extraction; 5. to make multiple treefall gaps in order to reduce damage to the residual stand. In most cases, directional felling cannot satisfy all these aims simultaneously. Field observations before logging had indicated that the density of potential crop trees (dbh ≥ 20 cm) was high, some 50 trees/ha, and that these trees were clustered around present crop trees. In view of the above considerations, the formation of multiple treefall gaps was considered undesirable. Therefore, it was decided to give priority to place logs in a suitable position to facilitate extraction (‘herring-bone’ felling pattern). Table 7.1 Key changes introduced in the reduced impact logging system (RIL) relative to the conventional operation (CNV) Phase of operation Logging method Pre-harvest planning Block lay-out Inventory, map scale Liana cutting Skid trail planning Tree selection system Annual allowable cut Felling pattern Tree selection criteria Skidding RIL yes 1 : 2,500 no no yes 1 : 1,250 yes yes no no no yes → species → no defect → legal diameter limit Harvest preparation Crop tree marking Marking potential crop trees Skid trail demarcation Harvest activities Felling CNV → → → → map no no → → → → → no directional felling high stumps convenient bucking no winching no choker straps species minor defect allowed differential diameter limit relative abundance map no yes → → → → → directional felling low stumps optimum recovery winching choker straps The reduced impact logging system is considered to fit into a silvicultural concept. The impact of a logging operation also depends on the intensity of the cut, which is another aspect of logging that needs to be incorporated. Hence it was important to establish 258 Chapter 7: Summary and general discussion how the implementation of reduced impact-logging techniques and tree selection criteria interact with logging intensity. 7.2 CANOPY FRAGMENTATION Total canopy loss due to felling did not differ between conventional and reduced impact operations at a logging intensity of 8 trees per hectare. Raising the logging intensity to 16 trees per hectare resulted in a relatively higher canopy loss in case of reduced impact logging. The presence of natural gaps or patches of aggrading forest prior to logging confounded the estimation of total canopy loss. If canopy loss is to be estimated by way of projection on the ground of felling gaps, it is therefore recommended that: ƒ Natural gaps should be mapped prior to logging in order to obtain an unbiased estimate of the amount of canopy loss due to felling. In the present study, circumstances dictated the use of an uncomplicated method. Prior to logging, the architectural mosaic of the plots that were to be submitted to reduced impact logging was assessed by dividing each plot in 20 x 20 m squares and estimating the proportion of each square occupied by aggrading forest. The blocks that were to be submitted to conventional logging had been logged before this assessment could be made. Comparison of the standing basal area prior to logging in conventionally and experimentally logged forest combined with field observations, indicated that the conventionally logged forest contained less patches of aggrading forest and less natural chablis than the experimentally logged forest. Careful analysis showed that the forest mosaic prior to logging had not influenced the interaction between logging intensity and logging method. Canopy loss increased proportionately less when the logging intensity was increased in case of conventional logging. This is plausibly explained by a difference in felling pattern and spacing between felled trees. In case of conventional logging, trees were felled in clusters, which led to overlapping felling gaps. In case of reduced impact logging, felled trees were evenly spaced and a herring-bone felling pattern was strictly applied. An increase of logging intensity meant an increase of the number of trees in a cluster in case of conventional logging; i.e. an increased overlap in felling gaps. In case of reduced impact logging, this meant only a marginal increase in overlap of felling gaps. ƒ Decrease in canopy cover due to felling is foremost determined by both the logging intensity and the architectural forest mosaic prior to logging, and furthermore by the felling pattern and the spacing between the felled trees, but hardly by the felling method. Because of an increase in overlap in felling gaps with an increase in logging intensity, the canopy loss per felled tree decreases with increasing intensity. This has prompted several authors to claim that if a certain volume of timber is to be harvested, it is better 259 Reduced impact logging in the tropical rain forest of Guyana to harvest a smaller area more intensively (Jonkers, 1987; Gullison & Hardner 1993; Whitman et al. 1997). However, such an approach overlooks that the forest mosaic resulting from such a logging regime risks to be strongly divergent from an evolutionary ‘fit’ between the natural forest mosaic and the regeneration strategies of the species characteristic for that forest. Therefore, it is important to compare changes in the forest mosaic between the two logging systems. In the present study, felling gaps have been assessed by applying Brokaw’s (1982a/b) definition rather loosely, thereby closely resembling the modified definition by Van der Meer & Bongers (1996a). Depicting a felling gap as a two-dimensional projection of a ‘hole’ in the canopy, which is delineated at the base has been proved to be useful in estimating abiotic conditions in a gap (Brown 1993; Whitmore et al. 1993). It must be realised that a delineated two-dimensional projection is a strong abstraction of the three-dimensional gap architecture as result of a treefall. First, inherently, there is no exact borderline between ‘felling gap’ and ‘surrounding forest’ (e.g. Lieberman et al. 1989; Vester, 1997; Rossignol et al. 1998) but rather a transition zone. Secondly, the effect of gap formation on seedling establishment, survival and growth is not restricted to the area immediately beneath a gap, but extends beyond the outline of the gap (e.g. Oldeman 1978, 1992, 1994; Orians 1982; Popma et al. 1988). Thirdly, differences in remnant and border vegetation, post-treefall debris and soil disturbance within the outline of and bordering a gap may affect the revegetation processes and gaps of similar size hence may develop differently (Oldeman 1978; Brokaw 1985a). The latter point is of great importance in comparing the two logging systems as shown here. A falling tree may form more than one opening in the canopy; i.e. the place where the crown of the tree has been removed from the canopy and the place where the crown of the tree lands - the epicentre of the treefall. At the epicentre, an opening may be formed because trees standing there are uprooted or decapitated. In case of multiple tree fall gaps epicentres may coincide or the crown of one tree may land near where the stump of another tree is located. Because these openings are usually separated by only a few remnant trees, borders of a canopy opening are ‘fuzzy’ and the effect of a canopy opening extends beyond its projected outline, it was decided to apply the chablis concept. By using the chablis concept, the connection between the number of trees felled in one gap can be included in the explanation of the forest dynamics involved. Not unexpectedly, it was found that multiple treefall chablis occurred much more often in the conventionally logged than in the experimentally logged areas. At a logging intensity of 8 trees per hectare, most chablis were single treefall chablis in case of experimental logging, whereas less than 40% of the chablis were caused by one treefall only in case of conventional logging. At a logging intensity of 16 trees per hectare, 50% of the chablis were still single treefall chablis in case of experimental logging, whereas this was the case in less than 25% of the chablis in conventional logging. In case of single, dual and triple treefall chablis, there was no difference between the logging methods in regard to the size of the projected canopy opening per chablis. If composite chablis containing more than three trees were formed, chablis were larger with experimental logging. 260 Chapter 7: Summary and general discussion The mean area of all chablis amounted to 265 m2 in case of experimental logging with an intensity of 8 trees per hectare. In case of conventional logging, it amounted to 430 m2 at this intensity. Reducing the logging intensity to an intensity of 4 trees per hectare (experimental logging only) did not reduce the mean chablis size any further. At 16 trees per hectare, the mean area of all chablis was 570 m2 for experimental logging and 780 m2 for conventional logging. Previous Tropenbos research has indicated that the maximum size of a canopy opening, that can be allowed to avoid adverse ecological and silvicultural effects, is 500 m2 (Brouwer, 1996), while Ek & Van der Hout (in prep.) found changes in species composition in gaps over 300 m2. In this respect, the following conclusions can be drawn from the results: ƒ ƒ ƒ Conventional logging of Greenheart-bearing forest stands leads to the formation of large canopy openings that involve high and undesirable silvicultural risks. Lowering the logging intensity below 8 trees per hectare in reduced impact logging does not reduce the average chablis size, while raising the intensity to 16 trees per hectare leads to silviculturally undesirable chablis sizes. In the present study, no difference was found in the size of single treefall chablis. This indicates that liana cutting and directional felling as carried out in this study did not reduce the amount of canopy lost. The latter conclusion need not be related to the felling method, but may also be related to a non-optimal period between the application of liana cutting and the moment of logging (six months) and/or to differences between the conventionally and experimentally logged areas in density of lianas and forest architecture. Field inspections showed that it is so difficult as to be virtually impossible to obtain a gap of a particular size at will (see also Van Dam & Rose 1997 and Mekkink & Nijmeijer 1998). Not only is it difficult to estimate the projected area of a canopy opening that will be formed by removing a certain crown from the canopy, it is even harder to predict whether or not the falling target tree will bring down any surrounding canopy trees, or, harder to predict still, tear big bush-ropes from the canopy. This fact, combined with the fuzzy nature of gap edges, indicates that focusing on a gap area with precise area specifications is not realistic. Still, a correlation was shown between the number of trees that form a certain chablis, and the size class of that chablis. This suggests that (cf. Webb 1998): ƒ It is better to focus on the mode of gap formation (i.e. single, dual, multiple tree falls) rather than on some precisely specified gap area, because the former is a more realistic and tangible objective when conceiving logging guidelines and at the same time refers to a biological unit. 261 Reduced impact logging in the tropical rain forest of Guyana 7.3 GROUND DISTURBANCE Proper planning of harvest activities, pre-alignment of skid trails, a herring-bone felling pattern and winching reduced the area traversed by the skidder considerably. At a logging intensity of 8 trees per hectare the area traversed by the skidder was reduced from 13% to 8% of the total area and at a logging intensity of 16 trees per hectare from 21% to 9%. The reduction could mainly be ascribed to an almost complete elimination of traces of skidder movements in felling gaps. The terrain in some of the experimentally logged plots was more dissected than the terrain in the conventionally logged area. If only the experimentally logged plots of which the terrain conditions were comparable to the ones found in the conventionally logged areas are considered, the area traversed by the skidder is reduced to 5% at a logging intensity of 8 trees per hectare and to 7% at 16 trees per hectare. Traversing of felling gaps was restricted to gaps along skid trails in the reduced impact operation, implying that traces of skidding occupied only 5% to 8% of the total felling gap area at logging intensities of 8 to 16 trees per hectare, respectively. In the conventional operation this amounted to as much as 30% to 36% of the total felling gap area. ƒ Implementation of reduced impact logging elements reduced the area traversed by the skidder by about two-thirds depending on the logging intensity, while skidder movements in felling gaps were even reduced by about three-quarters depending on the logging intensity. This difference is of great importance for the future tree species composition in these gaps. Successful establishment of shade-tolerant species, a group to which most commercial species belong, in gaps is depending on pre-existing seedlings (e.g. Hubbell & Foster 1983, 1986b; Brown & Whitmore 1992; Zagt & Werger 1998). Chlorocardium rodiei is locally very common as a seedling but its growth rate is slower than that of most pre-existing seedlings of other species. Zagt (1997) suggested, therefore, that it would only be successful in filling a canopy space if it has an advantage in size compared to close neighbours. Skidding in felling gaps steamrolls the few commercial seedlings that have this advantage, which means that recruitment either has to start from seeds or, if seedlings survive, by resprouting. In both cases, any advantage that a species may have had will be lost. To make matters worse, the machine scrapes and churns the top soil while it positions logs for hooking. This is an ideal treatment to favour the establishment of pioneer species (cf. De Foresta & Prévost 1986; Dalling et al. 1997; Guariguata & Dupuy 1997; Ek & Van der Hout, in prep.). It is also shown that the area traversed by the skidder per extracted tree is being reduced with an increase in logging intensity. This is intuitively expected, because a basic skid trail system is needed if one is going to extract a few logs or even one. The same system is used to extract any additional logs, and only some additional branch trails may be needed. This may be grist to the mill of proponents of high intensity 262 Chapter 7: Summary and general discussion small area logging (see above). In contrast to the case presented above against the excessive disruption of the forest mosaic resulting from high intensity logging, there are no silvicultural objections against this here. Indeed, though adopting the recommendation that logging intensity should be limited in order to prevent excessive disruption of the forest mosaic, here an argument is provided not to lower logging intensity any further than absolutely necessary. Just as with canopy fragmentation, the patterns of ground damage are related to the difference in technical aspects of the logging operations; i.e. directional felling, prealigned skid trails and winching. In addition, the change in tree selection criteria, and of those especially the change from group-wise to evenly spaced felling is certain to have had an effect on the outcome. If group-wise felling would not have been discouraged in the reduced impact operation, it is quite likely that less area would have been occupied by skid trails, because less branch trails would have been needed. It is more difficult to predict if it would have led to less skidding in felling gaps too. Conversely, if in the conventional operation trees eligible for felling would not have occurred in groups, it is most likely that more area would have been traversed by the skidder. This is because the conventional working method implies that each log is positioned by the blade of the machine, and hence less overlap would have occurred. Although I have no formal proof, I suggest on empirical grounds that group-wise felling most likely results in less ground area disturbed per tree than tree-wise felling. 7.4 DAMAGE TO THE RESIDUAL STAND Reduced impact logging damaged significantly fewer trees (dbh ≥ 10 cm) per extracted tree than conventional logging at both logging intensities. Regarding the damaged basal area, however, no significant difference was shown between the two logging systems. It would appear that these results are in disagreement with the outcome concerning total canopy loss, where reduced impact logging led to more canopy loss at a high logging intensity. However, this is not so. The - obvious - explanation is that conventional logging damaged relatively more small trees than reduced impact logging. This implicitly explains the fact that canopy loss and tree damage do not have to be mutually correlated, for small trees generally are subcanopy trees. The route chosen by the skidder follows places where seedlings and saplings are abound, because bigger trees are impediments. Skidding thus can be expected to mainly affect small trees, whereas felling can be expected to affect trees of all sizes. Based on the much larger area traversed by the skidder in the conventional operation, it thus would come as no surprise that the average size of the damaged trees was smaller after conventional logging. Separate analysis of felling and skidding showed a difference between the logging systems in the number of trees damaged by skidding, but not in terms of basal area damaged by skidding. In case of felling the reverse was found: no difference between logging methods in terms of number of damaged trees, but a difference in terms of 263 Reduced impact logging in the tropical rain forest of Guyana damaged basal area. Apparently, the difference in impact of the two logging systems is more intricate than suggested. The average size of all damaged trees was not only smaller because skidding damages smaller trees than felling and because more trees were damaged by conventional skidding than by experimental skidding, but also the size of the trees damaged by skidding was smaller. The opposite was found for trees damaged by felling. Reduced impact logging damaged relatively more trees by felling, but these trees were smaller than in case of conventional logging. In fact, experimental felling damaged more canopy trees (dbh ≥ 40 cm), less medium sized trees (20 cm ≤ dbh < 40 cm) and more small trees (10 cm ≤ dbh < 20 cm). The reasons for these differences lie partly with the difference in felling method and partly with the fact that the skidder destroyed trees in felling gaps in the conventional operation. In conventional - group-wise - felling, the number of neighbouring large trees that can potentially be damaged is smaller, because they have already been felled. The felling of trees criss-cross in one chablis, however, hits most standing mediumsized trees in that chablis. In case of directional, tree-wise felling, on the other hand, more neighbouring large trees can potentially be damaged. In addition, by directional felling, medium sized trees are intentionally spared at the cost of small trees. During skidding small trees were more often damaged in the conventional operation, because of movements in the stump area. During these manoeuvres small trees that survived felling were knocked down or grazed by the machine. ƒ In terms of residual stand damage, reduced impact logging diminished skidding damage - to small trees in particular – and shifted felling damage from medium sized trees to small trees. The breaking down of damage by size classes and logging phase showed that pooling of these categories could obscure important differences between logging methods. ƒ The use of global parameters, such as the total basal area or the total number of trees per hectare that have been damaged, is fundamentally inappropriate when comparing complex logging methods in complex ecosystems. Logging intensity, felling pattern, spacing of trees to be felled and directional felling all determined felling damage to the residual stand. Still these factors explained only part of the difference in damage to the residual stand. Forest inventory prior to logging had revealed differences in forest mosaic (see above) and also, partly related to the former, differences in the size class distribution of the stands and of the harvested trees in the various plots. Compensating for the size of the felled trees by relating damage to the basal area felled per hectare revealed differences that had remained hidden when relating damage to the number of trees felled per hectare. Obviously, differences in stand density and size class distribution of the residual stand have influenced the results. This is in agreement with Van Rompaey’s (1993) postulate that true replication cannot exist in complex ecosystems such as tropical rain forests. 264 Chapter 7: Summary and general discussion ƒ Van Rompaey’s (1993) postulate that neither homogeneous sample plots nor true statistical replication exists in complex ecosystems has been confirmed by this study. This implies that comparisons between different logging systems cannot produce unequivocal differences or widely valid reductions in damage due to implementation of certain methods. This also goes to show that comparison with studies carried out elsewhere - with different forests and different logging regimes - is a hazardous pursuit and should not be used for guidance. The reduced impact experiment was set up as a randomised block design. Three sets of five plots were regarded as three statistical blocks, based on their geographic distribution. In each block, logging intensities were assigned randomly to the five plots. It turned out that in each block the plots that were assigned the highest logging intensity were among the ones with the lowest stocking in harvestable trees and with the lowest stand basal area. Statistical analysis did not filter out the ‘random’ coincidence, which is probably related to the few degrees of freedom in the experiment. However, the result that high intensity logging led to proportionately more damage than high intensity conventional logging was statistically significant and thus statistically correct. Alder & Synnott (1992) warn that random sampling and the use of square sample plots ignore the ordination of vegetation along natural gradients, and propose the use of plots large enough to cancel out natural variation. For practical purposes; i.e. to estimate the percentage of canopy cover that is lost or the percentage of the area traversed by the skidder, one is forced to outline plots, square ones being the most practical. It is questionable, however, if a randomised block design is appropriate. It would most likely produce better answers if plots were first stratified according to e.g. stocking with harvestable stems or total basal area and treatments assigned accordingly. 7.5 THE COST OF INTRODUCING REDUCED IMPACT LOGGING The sustainability of conventional selective logging in Chlorocardium rodiei-forests has been questioned repeatedly, e.g. by GNRA (1989), Ter Steege et al. (1996) and Zagt (1997). Although current practice seems to be paying no attention to sustained yield principles, hence goes without investment into a future crop, only few logging companies are apparently operating at a profit, as concluded in studies by Flaming (1995) and Landell-Mills (1997). This is significant, because there is a strong belief among loggers in Guyana that logging cost will rise when reduced impact logging is implemented. The fact that profitability of operations is low combined with the impression that environmentally sound logging is more costly is used as an argument against a change away from habitual methods. Therefore, it is extremely important to quantify the economic returns effected by the two logging systems. The experimental set up was not entirely suited to carry out a detailed appraisal of the change in the cost-benefit ratio by introducing the reduced impact logging method 265 Reduced impact logging in the tropical rain forest of Guyana portrayed in table 7.1. The main reason is that the size of the experimentally logged plots was only 6 hectares, whereas the conventionally logged plots measured 11 hectares on average. Moreover, the plots were deliberately located in commercially rich stands. A complete, detailed picture of the logging operations could thus not be provided. Some essential elements, however; such as directional felling and log collection by using the winch and choker straps, could be appraised properly because they are independent of the size of the experiment. The cost involved in the remaining logging phases was estimated using standard methods published in handbooks (e.g. FAO 1977, 1978; Sundberg & Silversides 1988; Silversides & Sundberg 1989; Caterpillar 1986) and data provided by the timber company (Demerara Timbers Ltd., 1998 data of salaries, machine purchase prices, etc.). Logging costs and performance are affected by many different factors. The quantity and quality of the available labour and machinery, the characteristics (high or low wood density, presence or absence of silica or latex in wood and/or bark, buttressed or unbuttressed) and the size of the trees being harvested, topography, carrying capacity of soils, accessibility of the area to be logged, skidding distance and organisational efficiency all influence the costs and returns of a particular logging operation. The variation in stand characteristics, terrain, distances between trees implies that it is impossible to assume ceteris paribus conditions. A feller or skidder operator meets a steadily changing and challenging work environment and the work approach demanded by this environment is never the same. Such differences are inherent to the complex rain forest environment and, therefore, occurred also between the conventional and reduced impact operations that were studied here. By using regression models, it was attempted to filter out differences, particularly differences in tree sizes, topography, soils and distances to log markets. The models used a feeder road density of 1 km per 1 km2 and a main road density of 1 km per 5 km2 of logging area. Furthermore, results were based on an average logging intensity of 10 trees per hectare The mean recovered volume per tree in the conventional operation was 2.8 m3, and in the reduced impact operation 3.1 m3. One skidder load equalled a size of 9.7 m3. Several studies have shown that additional cost associated with planning and directional felling can be offset by an increased efficiency of the skidding operation (Mattson Mårn & Jonkers 1981; Hendrison 1990; Barreto et al. 1998; Holmes et al. 1999). In the present study, implementation of the reduced impact logging system led to a threefold increase in pre-harvest planning cost and a twofold increase in felling cost. Those increases were only partly offset by a reduction of skidding cost, which amounted to 5% only. The aggregate direct cost at the landing was increased by 15%. However, the output per day and per hectare was considerably higher under the reduced impact regime. This implied that the cost per m3 of road construction, road maintenance, supervision, logistics, etc. must be assumed lower. Including these indirect cost factors indicated that: 266 Chapter 7: Summary and general discussion ƒ ƒ ƒ There are strong indications that the reduced impact logging system is neither more expensive nor cheaper than logging in the customary way. However, it is also indicated that The scale of the model and the complexity of the forest environment, in terms of biotic and abiotic variability, make unequivocal statements impossible. Ceteris paribus conditions, inherently, do not apply and the impossibility of true statistical replication in tropical rain forest so applies just the same to studies on operational efficiency as it does to biological studies. Comparison of the results on operational efficiency with other sources provides little information, because of differences in research and logging methods, in biotic, abiotic and timber market circumstances and differences in history of the (logging) culture. The absence of ceteris paribus conditions prevents us from making precise statements relative to differences in cost-benefit ratios, amplified by the assumptions made and the size of the experiment. However, these limitations do not prevent us from sketching a clear picture of the crucial issues leading to differences in the total cost configuration. Before pinpointing the issues that are crucial to the financial result, it is important to differentiate between the components of the reduced impact guidelines that were used in the present study. As mentioned earlier when describing logging damage, the reduced impact logging system is composed of technical and tree selection components. The entire system is a silvicultural system, the logging method being part and parcel of the silvicultural system. The technical components included the ‘traditional’ reduced impact elements, such as pre-aligned skid trails, directional felling, winching, etc. The tree selection system included criteria based on the abundance of a species, differential diameter limits and a more or less even spacing of felling gaps. In the financial analysis, another technical component holds a prominent place; i.e. the difference in recovered volume per tree. This aspect is often included among the ‘traditional’ reduced impact elements, but there are reasons to single this aspect out. Higher wood recovery does not reduce the impact on the ecosystem. It neither is part of silviculture in the proper sense: i.e. the growing and tending of forest products and maintenance of forest services. It does belong, however, to sound forest management; i.e. optimising the output obtained from a parcel of forest. If the cost of a certain activity differed between the two logging systems, this usually could be ascribed to one of these three components or to a combination of two or even three of them. Cost differences are also related to a change of objectives and attitude of workers. In the conventional operation, the main objective of the workers was to reach a certain target in the least possible time, whereas in the reduced impact operation the emphasis was on quality and worker safety. This influenced the amount of time spent during nearly each activity. In the following list, differences that were established with 267 Reduced impact logging in the tropical rain forest of Guyana certainty are marked with an asterisk. In other cases, there were strong indications, although they remain without hard evidence, due to the limitations of the study. The introduction of technical elements led to changes in the cost of the following activities: 1* increase in the cost of pre-harvest planning and preparation; i.e. higher accuracy of tree enumeration and mapping, planning and demarcation of skid trails, pre-harvest climber cutting; 2* increase of time spent on felling preparations and actual felling. The time used for felling was augmented fourfold. This huge difference is presumably influenced by the short training period. This difference will most likely be reduced when the system will have become operational, because skills in the directional felling technique will have developed further. Directional felling was applied tenaciously during the experiment using the Brunberg (1984) method of sawing backward from the hinge. It appeared that more flexibility is allowed regarding the actual direction of the lie and regarding the felling technique used (often traditional proper felling techniques - Conway 1982; Klasson & Cedergren 1996 - will suffice); 3. reduction of time spent by the skidder moving back and forth between stump area and roadside landing. This is caused foremost by a shorter distance covered as a result of pre-alignment of skid trails and the higher speed that can be developed by the machine as a result of straighter trails and a dendritic trail pattern instead of a haphazard trail network; 4. because of the reduction in time of a skidder round-trip, the daily output of a skidding team is being increased. I claim that this leads to a reduction of cost per m3 of linked cost items, such as landing operations, support, supervision and logistics. The introduction of tree selection criteria led to changes in the cost of the following activities as well as perceived benefits, mainly due to a change-over from group-wise to tree-wise felling: 1* a shift in the species composition of the coupe away from high-value Chlorocardium rodiei to lower-value timber species; 2* time to approach and collect a load by the skidder; 3. time to locate trees for felling. This is also influenced by a change in attitude toward work. The pursuit of a higher timber recovery per tree affected the cost of following activities: 1* increase of time required for topping and bucking; 2* increase of volume obtained per felled tree, and hence a proportionate reduction of cost of all felling activities per m3; 3. increase of volume obtained for each skidded bole, hence a proportionate reduction of cost per m3 of skidding and subsequent activities, such as landing operations and loading; 268 Chapter 7: Summary and general discussion 4. increase of volume obtained per hectare, hence a reduction of all activities engaging costs per unit area, such as tree enumeration and mapping, road planning, construction and maintenance, support, logistics and overhead. Differentiation between the components of the reduced impact logging system tells us that introduction of the technical component alone most likely will lead to an increase of skidding efficiency. This is because the log collection phase will take less time without a difference in spacing between logs to be collected. The difference in cost of the pre-harvest and felling operations is not affected by the tree selection criteria. Proportionately, the skidding phase is much more expensive than the former two phases because of the high machine cost per m3. Therefore, there is little doubt that changes in the technical aspects alone will indeed improve efficiency of the logging operation. Higher timber recovery appeared to be an important issue in restoring the cost-balance between conventional and reduced impact logging. Although there is no direct evidence, modelling indicated that the advantageous effect of group-wise felling was larger than the effect of increased timber recovery. The introduced tree selection criteria reduced the efficiency of the operation. Partly this was related to the higher efficiency of group-wise felling, but the major factor is the change in species composition of the output. Undoubtedly, the aspect of foregone timber will be the most difficult one to reconcile with the interests of the logger. In the end, it depends on the forest policy of Guyana and on the expectancy that conventional tree selection methods will not qualify for certification. Implementing the tree selection criteria will bring good forest management closer and most likely will ensure access to privileged markets, maybe with bonus prices. Furthermore, conventional tree selection methods do not appear to allow a second high-value coupe in the near future, whereas the odds are much better when silvicultural tree selection criteria are implemented. The present financial value of future crops is debated widely. The use of interest rates for borrowed money is not encouraging in this respect (cf. Kuper 1994). ƒ ƒ The introduction of technical reduced impact elements, such as pre-harvest planning, pre-alignment of skid trails, directional felling and winching, will most likely result in a more cost-effective operation. The introduction of silvicultural tree selection criteria will probably result in a less cost-effective operation, because of a reduced efficiency of the skidding phase and reduced short-term financial returns (smaller proportion high-value timber available later). The silvicultural productivity of the ecosystem is much better maintained, however, and long-term benefits are markedly increased (larger proportion high-value timber available sooner). A limitation of the present study was that it was carried out within commercially rich stands. In fact, well-stocked stands are islands in a sea of commercially poor areas, 269 Reduced impact logging in the tropical rain forest of Guyana indicating that the costs in real terms are presumably higher, especially those of preharvest planning, road construction and maintenance and overheads. Breakdown of the total logging cost showed that skidding, loading, trucking and overheads are the most important cost factors, whereas pre-harvest planning and preparation, and felling were minor cost factors. It is hypothesised that in extrapolating the results of this study to an operational scale, skidding is the most sensitive issue. Furthermore, it is envisaged that the effect of group-wise felling, which gave conventional logging an edge over reduced impact logging will be toned down on an operational scale. Based on experience elsewhere, where wastage in the shape of overlooked logs appears to play an important part, it is also envisaged that timber recovery will be augmented on an operational scale. Finally, it is envisaged that the spatial pattern of logging will change when implementing the reduced impact guidelines. The conventional logging system will keep its intermittent, patch-wise character, whereas reduced impact logging will make better use of the resources. Volumes recovered per hectare are thus envisaged to be considerably higher. It is recommended to carry out a detailed study on an operational scale to verify these hypotheses. 7.6 THE PROSPECTS OF SILVICULTURAL INTERVENTION AFTER FIRST HARVEST Sustainable timber production can only be achieved if extraction levels are adapted to the potential productivity of the forest and hence if the forest mosaic after logging offers fitting biotopes to the regeneration strategy of the commercially desirable species. However, even after an environmentally sound first harvest, the overstorey is impoverished with regard to trees of species that would be valuable for future harvests or as sources of seeds. The unattended logged forest therefore often shows only marginal increment of commercially desirable species. Moreover, the remaining canopy is stripped of seed trees of these species, which may lead to progressive impoverishment of the forest in the long run (Synnott & Kemp 1976; De Graaf 1986; Wadsworth 1987; Hutchinson 1988). Numerous attempts have been made to design silvicultural systems throughout the tropics that would make logged forest productive again (Chapter 3). Silvicultural pre- and post-harvest treatments are usually directed at increasing the crown exposure of potential crop trees to light by releasing them from impeding commercially undesirable neighbours. In the Central European selection system (Leibundgut 1984; Schütz & Oldeman 1996) stand treatments are directed at the trees of potentially highest value, by thinning trees of lesser value that impede the development of the former. In Central Europe, a market exists for these thinnings, but this is usually not the case in tropical rain forests. Release operations are then carried out by killing commercially undesirable impeders. In order to examine whether a release operation can improve the conditions for potential crop trees in Greenheart-stands that are selectively logged according to the 270 Chapter 7: Summary and general discussion guidelines formulated above, the amount of total irradiation received by potential crop trees was estimated. This was done by constructing a three-dimensional forest model of two logged stands and subsequently simulating hemispherical photographs ‘touching’ the tops of their crowns. The programme SILVI-STAR (Koop 1989; Koop & Bijlsma 1993; Koop & Sterck 1994) was used for this purpose. Five different silvicultural prescriptions were evaluated in this way. In order to determine the efficiency of a certain prescription it is of importance to assess how many of the impeding trees it removes and how many trees that it removes do not impede the development of potential crop trees in reality. The five silvicultural prescriptions were: 1. systematic elimination of all commercially undesirable trees down to 20 cm dbh (R20); 2. as 1) down to 30 cm dbh (R30); 3. localised elimination of all commercially undesirable trees down to 20 cm dbh within a fixed radius of 7 m around favoured trees (L7); 4. as 3) with a fixed radius of 10 m (L10); 5. a combined treatment eliminating all undesirable trees with dbh ≥ 40 cm over the entire area and with 20 cm < dbh ≤ 40 cm within a fixed radius of 10 m around favoured trees (R40+L10). In order to be selected as a favoured tree, a tree should be a commercially desirable or a potentially desirable species, have a diameter between 20 cm and 40 cm dbh, be free of serious defect or damage, be projected to contain at least one length of sound, straight timber six metres long and lean no more than 15 degrees from the vertical. First, the illumination of potential crop trees and their Dawkins Crown Class were determined. The following emerged: ƒ a large proportion; i.e. 28% to 46%, of the tops of the focal trees received more than 80% of potential total irradiation, which can be considered to be sufficient; ƒ a considerable proportion; i.e. 12% to 21%, of the entire crowns of the focal trees received full overhead light; ƒ release of focal trees was restricted because the majority of the neighbouring, interfering trees were favoured trees themselves; no improvement was possible in 29% to 50% of the cases in the two plots, while complete release was only possible in 17% of all cases; ƒ neighbouring trees qualified as impeders up to 25 m away from a focal tree, but only 15% of the impeders were more than 10 m away from the nearest favoured tree (base to base distance). These findings imply that: ƒ Crown illumination of potential crop trees varies strongly according to the composition, architecture and structure of the remaining stand. 271 Reduced impact logging in the tropical rain forest of Guyana ƒ Augmenting crown illumination by elimination of undesirable shading trees after medium intensity reduced impact logging is constrained, because a large proportion of the interfering trees are either retained commercial trees or potential crop trees. Secondly, trees were identified that were actually competing for light. It appeared that the proportion of total irradiation received by potential crop trees was correlated with the Dawkins Crown Class of the potential crop trees as long as class ‘3’ was not exceeded: i.e. those trees receiving no direct light, trees receiving only lateral light and trees being partly shaded vertically. Furthermore, the proportion of total irradiation was correlated with the number of trees that cast lateral shade upon focal trees. So, neighbouring trees that were casting overhead shade or lateral shade were identified as impeders. ƒ The Dawkins Crown Classification and the number of laterally shading (Alder & Synnott 1992: ‘crowding’) trees are good indicators of the illumination of a focal tree and can be used to identify impeders. Having identified neighbouring trees that impede the development of potential crop trees, an ideal treatment was formulated, which I refer to as ‘selective release’ according the concept of crown thinning; i.e. removing suppressing neighbours of lesser value. Compared to the control (zero intervention) all examined treatments augmented the illumination of the tops of the favoured trees. All treatments but one improved the illumination to a similar degree. The localised treatment within a radius of 7 m appeared to be less successful. The maximum attainable effectiveness (= ‘selective release’), as expressed by the number of eliminated impeding trees divided by the total number of impeding trees, was only 53%. This is caused by the high density of retained commercial trees and potential crop trees. In terms of effectiveness, the treatment R20 (by default) scored just as good as the ‘selective release’ method, followed successively by R40+L10, L10, R30 and L7. Comparison with the ‘selective release’-method showed that all treatments removed more trees than necessary. Obviously, eliminating undesirable trees unnecessarily means that funds and resources are wasted unnecessarily. It also means unnecessary disruption of the canopy. Interventions in the canopy after logging inevitably lead to an additional increase in light levels on the forest floor, which in turn may provoke the establishment of silviculturally undesirable light-demanding pioneer trees and lianas. That risk and the need for maintaining biodiversity and keystone species (e.g. Terborgh 1986; Mills et al. 1993) demand that a treatment should be balanced against its potential negative impact on the biological integrity of the forest (cf. Pinard et al. 1999). In our case, this can be done by adapting silvicultural prescriptions so that they waste as little biomass as possible. Systematic elimination of all undesirable trees with a lower diameter limit of 30 cm dbh appeared to be the least wasteful while not being less effective than the other treatments. However, this type of treatment is a blanket treatment, implying that it is carried out over an entire area without paying attention to 272 Chapter 7: Summary and general discussion local variation in species composition. So, it may have been the least wasteful in this particular study with as many as 50 potential crop trees per hectare, but it will clearly be more wasteful than the localised treatments if applied to poorer areas. For the same reason, the combined treatment R40+L10 is not recommended. In my quest to come up with a less wasteful treatment, the localised treatment with a radius of 10 m will be the starting point. ƒ ƒ Systematic elimination of all undesirable commercial species down to 20 cm is not more effective in improving the illumination of potential crop trees than a localised treatment within a fixed radius of 10 m around potential crop trees. All treatments based on a diameter limit for elimination are wasteful because more trees are eliminated than necessary. Blanket treatments carried out over the entire area regardless of the density of potential crop trees are the most wasteful. Another way to formulate silvicultural prescriptions is to start from relationships between crown diameter and tree diameter, which differ between species. In theory, based on such relationships, one can distil the amount of crown space likely to be occupied by a tree at (financial) maturity, for practical purposes at a diameter of 45 cm dbh. Conversely, it can theoretically indicate the amount of crown space a neighbouring, undesirable tree will have at maturity. This principle has been translated into liberation schemes that are based on the diameter of a desirable tree, that of an undesirable neighbour and the distance between stem bases (e.g. Wadsworth 1958; Hutchinson 1988). Examination of crown diameter - tree diameter relationships revealed that strong relationships exist between these two parameters for at least the four most common species in the two studied 2-hectare plots. However, the three-dimensional forest model denied that this relationship could be used to predict whether a neighbouring tree is impeding or not. The explanation for this is as follows: ƒ crowns are generally eccentric relative to the position of the base of a tree; ƒ crown architecture varies between species, leading to a different height at which point a crown reaches its largest extension; ƒ crown architecture varies within species, because crowns in the canopy have expanded through reiteration, which is provoked by exogenous influences (light); ƒ tree height as such is not strongly correlated with tree diameter in the diameter range considered here (dbh ≥ 20 cm); ƒ the topography may confound any such relationships if indeed they exist. In conclusion: ƒ Whether or not a neighbouring tree will cast vertical or lateral shade on a favoured tree now or in the future cannot be determined by examining crown diameter - tree diameter and/or tree height - tree diameter relationships, even 273 Reduced impact logging in the tropical rain forest of Guyana when such relationships are shown to be strong, because other - crown architectural and exogenous - factors determine the shape and height of a crown and the direction in which it expands. Examination of the relationship between tree parameters, such as stem diameter, crown size and total height, and being a dominant or suppressed tree revealed that only total height had any predictive virtue. This fits in with the concept of ‘crown thinning’ (Fr.: éclaircie par le haut) where overtopped neighbouring trees that do not interfere with higher-valued trees are not removed. Field inspections showed that although it was most often unfeasible to determine from the ground which neighbouring trees were interfering with favoured trees, it did prove to be feasible to distinguish which trees were taller. Including this principle in our best treatment so far leads to the formulation of the following treatment: eliminate commercially undesirable trees within a fixed radius of 10 m around only if they are taller than the favoured tree. Still, this treatment is not ideal, because of the poor predictive value for interference of stem base to stem base distances and because future interference is disregarded. It would therefore be recommendable to eliminate any tree beyond 10 m that is visibly casting lateral or vertical shade. Whether or not the formulated treatment is desirable cannot be answered yet, because the effect on the increment rate of the favoured trees remains unknown. A treatment as formulated above was applied in the field in December 1996. Evidence from neighbouring Suriname (De Graaf 1986; De Graaf et al., in press) and French Guiana (Favrichon 1997; pers.obs. 1993) shows that systematic treatments (comparable to our R20) can enhance growth rates considerably, but also stimulates recruitment and growth rates of commercially undesirable pioneer species. Whether the localised treatment applied here will bring about similar increases in increment rates, while keeping a rampant growth of undesirable elements at bay, must remain unanswered for the moment. In this respect, it should be kept in mind that an investment to augment volume accumulation through silvicultural intervention, does not determine the total volume produced by a forest, but only the marginal accumulation (surplus) above the volume produced by a forest without investment (Kuper 1994). 7.7 SILVICULTURAL CONSIDERATIONS REGARDING GREENHEART FORESTS Selective logging in Chlorocardium rodiei-forests has been carried out according to a ‘log-and-leave’ principle. However, this is not the ‘log-and-leave’ management option as suggested by Poore et al. (1989), because there has been no conscious decision to manage the forest at that particular level. So, the reasonably good state of Guyana’s forests is not the result of intentional management, but rather the outcome of limited marketability of the resource and the absence of a demand from competing land uses due to the poor soils in the interior and Guyana’s small population. 274 Chapter 7: Summary and general discussion When starting this study, I had in mind to aim at a silvicultural system that was ‘close to nature’, a concept that has been practised successfully in Switzerland and Slovenia (Schütz & Oldeman 1996). By now, it could be called an ancient concept, since it already surfaced in Europe in the 19th century (Gayer 1880 ex Schütz & Oldeman 1996). Close-to-nature silviculture mimics the natural process of gap-phase succession using controlled single-tree selection (e.g. Baur 1964). The chablis formed by felled trees are similar to the natural chablis found in virgin rain forests and are often able to follow the same successional pathway to recovery (Whitmore 1991). Although the latter perception may hold, changes in the forest mosaic as a result of selective logging are not entirely congruent with natural forest dynamics. In the present study, it was found that the average chablis size resulting from logging is larger than the average natural chablis size. Moreover, a marked shift occurred in the chablis size distribution compared to the natural distribution, even at low intensity logging. The majority of the natural gaps are small (< 100 m2 - e.g. Van Dam & Rose 1997; Hammond et al., in prep.), with occasionally a very large gap (> 1000 m2 - e.g. Van Dam & Rose 1997; Uhl et al. 1988), resulting in a size distribution which is strongly skewed to the left. The size distribution of felling gaps tends to have more medium sized gaps, resulting in a more bell-shaped distribution. Another, equally important difference, is the synchronisation of gap formation. It is thus inevitable that there will be changes in the forest mosaic and some modification in species composition will inevitably occur and according to Smith et al. (1996, p. 414): ‘It must be recognised that there is no silvicultural magic that will enable (‘primary’) forests to be harvested and regenerated without altering them in some degree.’ The issue in question is rather whether it is desirable or less so to aim at maintaining the forest structure and composition in a pre-intervention state (Putz & Viana 1996). There is evidence of widespread human occupation and forest use in many tropical forests formerly considered ‘pristine’ (e.g. Sanford et al. 1985; Saldarriaga & West 1986; Gómez-Pompa & Kaus 1990; De Graaf & Hendrison 1993). In this respect, the example of the scarcity of regeneration in the vicinity of mature trees of Swietenia macrophylla, suggesting a relation with major disruptions in the past (Lamb 1966), has become well-known. The presence of charcoal in the soil in our study area, dated to stem from the not-so-distant past, suggests that also the Greenheart forests have suffered from rather recent impacts due to drought associated with the El Niño Southern Oscillation phenomenon (Cane 1983 ex Woods 1989) or human occupation (Ter Steege 1999, pers.comm.; see also Hammond & Ter Steege 1998; cf. CharlesDominique et al. 1998). There are also other reasons, related to the sustainable use of the forest resources. Modern silviculture does not focus on the use of the forest as a resource of timber alone, but more so on the multifunctionality of forests, including the sustenance of biodiversity, water quality and soil protection, and climate regulation. With a view to European forestry, Schütz & Oldeman (1996) unravelled the myth that only natural forest dynamics can really fulfil the present demands of the forest resource. They suggested that leaving stands unattended would lead to a uniform forest structure with 275 Reduced impact logging in the tropical rain forest of Guyana a regular, closed canopy. Such a structure eventually becomes susceptible to largescale impacts (e.g. storms, avalanches) and results in reduced biodiversity (see e.g. Zukrigl et al. 1963). An irregular canopy (Fr.: Futaie jardinée), on the other hand, is more stable, because the structure incites an individual stability of its components. Light penetrates deeper in an irregular canopy making way for a greater diversity of both flora and fauna. Irregular canopies are rare in virgin European forests (Fagus sylvatica forests in particular - Peters 1992) and seem to occur only during the senescent phase (Schütz & Oldeman 1996). The latter authors suggest to use natural forest dynamics as a source of information and inspiration to build silvicultural systems, such as the selection system (G.: Plenterwald, Fr.: Futaie jardinée), rather than to try to rebuild particular natural biotopes, that in fact represent merely a phase in natural forest dynamics. Obviously, there are great differences between European Fagus sylvatica forests and the Greenheart forests of Guyana. Still, there is a strong resemblance in stand structure between a mature Chlorocardium rodiei reef and the European Fagus sylvatica forest (cf. Van Baren & Hilgen 1984). Even the age at which the senescent phase is entered appears to be similar; i.e. some 300 years. Although direct evidence is lacking, I would suggest that (over-)mature Chlorocardium rodiei reefs at one time will enter an unstable senescent phase and will be prone to medium-scale overturning, where groupwise rather than tree-wise regeneration will occur. The relatively poor diversity of these forests has been reported by e.g. Ek (1997) and Ter Steege (1998). Zagt (1997) suggested that the secret behind the success of Chlorocardium rodiei is that it capitalises on chance by being very common as a seedling, and, in addition, capitalises on size once it has advanced to the sapling phase. Chlorocardium rodiei saplings would continue to grow slowly in the deep undergrowth, where other species would not grow or even die. In this way, it would gradually accrue a large size, which would give it an advantage over other faster growing but smaller individuals of other species when a chablis is formed. If this theory holds indeed, the making of an irregular canopy sensu the selection system, should be advantageous to the survival and growth of Chlorocardium rodiei saplings, on condition that they are protected from logging damage. Soil disturbance is an unavoidable consequence of ground-based log extraction; of which soil compaction has no natural equivalent, whereas churning and scraping of the topsoil may be compatible with treefall pits and mounds to some extent. Soil compaction has a dramatic effect on the infiltrability and sorptivity of the soil (e.g. Malmer & Grip 1990) and regeneration on such damaged areas is poor. Because of the latter and the importance of pre-existing seedlings and saplings as constituents of the future canopy, controlling skidding damage is of paramount importance. The present study has shown that the conventional logging practice in Greenheart forest is not likely to be sustainable. Reduced impact logging was successful in reducing skidding damage considerably, especially within gaps, and in building a forest mosaic that adheres to the evolutionary fit between the naturally occurring 276 Chapter 7: Summary and general discussion mosaic and the regeneration strategy of the species that are characteristic for these forests. The latter was only true if the logging intensity was restricted to a level of 8 to 10 trees per hectare. Conversely, reducing the logging intensity below this level did not lead to smaller chablis sizes. Otherwise, skidding damage and residual stand damage per extracted tree was, as anticipated, reduced with increasing logging intensity. Reconciling the increased disruption of the forest mosaic with a lower soil and stand damage per extracted tree suggests that logging intensity should not be lowered beyond the level recommended above. N/ha The reduced impact regime, which was applied in this study, consisted of traditional technical impact reducing elements, tree selection criteria and an increase of timber recovery per extracted bole. This resulted in an intricate blend of sometimes contrasting effects. The change from group-wise felling to tree-wise felling, one of the most important measures to secure the desired forest mosaic, was disadvantageous in other respects. Residual stand damage as well as probably skidding damage was increased proportionately. Also the efficiency of the skidding operation was affected, because the machine had to drive up and down the trail to collect logs, whereas these could be bunched in one go in case of group-wise felling. Group-wise felling thus had quite some advantages. It was also established that the size of dual treefall chablis remained within the limits of what is thought to be silviculturally acceptable. It is therefore not strictly necessary to restrict felling to single treefall gaps. (b) N/ha (a) (c) Figure 7.1 Diameter class frequency distributions for three common commercial species in Pibiri, Central Guyana prior to logging with balanced diameter distribution (‘de Liocourt’) curves. Carapa guianensis (Crabwood) approaches silvicultural “normalcy”; Catostemma fragrans (Sand 277 Reduced impact logging in the tropical rain forest of Guyana Baromalli) is underrepresented in the 10 to 25 cm classes; and in case of Chlorocardium rodiei (Greenheart) the classes below 35 are all underrepresented. Curves apply to the total area of fifteen 2-ha plots. Diameter class frequency of trees < 20 cm dbh is based on a 12.5% sample per plot. N/ha N/ha A basic tenet of silvicultural systems, which has not been dealt with explicitly so far, is that the rate of timber harvesting should not exceed the rate at which commercial timber volume accumulates per species or per stand (‘the annual allowable cut’). In other words, only the ‘interest’ on the forest ‘capital’ is harvested. Application of this concept is fraught with difficulty; one common problem is that few stands are composed of balanced mixtures of stand ages (Smith et al., 1996). Furthermore, in previously unlogged forests, a complication arises from the fact that some of the trees over the mean felling diameter are very large and very old - much older than the age of the largest trees that will be cut in the second cycle. If all these trees are removed in the first cycle, then the yield in the second cycle, inevitably, will be less (Dawkins & Philip 1998). Leaving those trees standing at the initial cut may risk that they die before they can be harvested. Figure 7.2 Diameter class frequency distributions of Chlorocardium rodiei (Greenheart) in four 2-ha plots in Pibiri, Central Guyana prior to logging. The species shows strong local variation in distribution; in some plots silvicultural “normalcy” is approached (plot 8); in others a bimodal distribution is indicated (plots 2 and 4), whereas the remaining plot 12 is deficient in the smaller diameters In case of Greenheart forest, but also in many other tropical forests, a lack of knowledge on growth and mortality rates until now has prevented calculations of allowable cut. A major complication is that growth and mortality rates vary along with 278 Chapter 7: Summary and general discussion the intensity with which a forest has been logged (Van der Hout, unpublished data). Such data have been collected since logging took place in the experiment under discussion, but due to the short period - 1994 to 1997 - no conclusions could be drawn from these measurements. Remeasurements are planned and preliminary results will be published in the near future. The problem of unbalanced stands also occurs in the forest stands that were studied. Before logging, some common commercial species had a balanced - Carapa guianensis, fig. 7.1a - or nearly balanced stand table - Catostemma fragrans, fig. 7.1b. The bimodal size class distribution of Chlorocardium rodiei is a well-known feature (e.g. Ter Steege 1990). Such a distribution was not found in the Pibiri area, although the distribution is certainly not balanced (fig. 7.1c). If the size class distribution of Chlorocardium rodiei is broken down per 2-hectare plot, a large variation in size distributions is found, some of them bimodal others regular (fig. 7.2). Some authors have pointed out that bimodal size distributions are a consequence of rare establishment conditions (Bongers et al. 1988; Poorter et al. 1996). The variation in size distributions found suggests that if this is caused by regeneration waves, these are certainly not a result of a large-scale phenomenon (see also Zagt 1997). Calculation of the allowable cut, once sufficient data are available, will need to consider this variation. It is hypothesised that balancing a calculated annual allowable cut against our tree selection criteria - the most important of which are even spacing of trees to be felled - and differential diameter limits, will ensure sustainability of timber production of the species that are characteristic of these forests. The problem of overabundance of trees above the mean felling diameter should not be solved by harvesting all of them at the first cut, because this will result in the same forest mosaic as was found after conventional logging. It is proposed to regard some of these trees - although often overmature - as savings for hard times. The chance that they will die before the next cut is not unreal, but in order to keep the forest structure in the desired state some of them may have to be sacrificed. Again we are confronted with the fact that implementation of a rational, sustainable management system, will force the logger to forego - for the time being - some of the timber he otherwise would have harvested. This issue will remain the most important barrier in introducing sound forest management. 279 Reduced impact logging in the tropical rain forest of Guyana 280 Samenvatting SAMENVATTING Tropische regenwouden Tropische regenwouden worden in een gordel rond de wereldbol aangetroffen, min of meer tussen de twee keerkringen gelegen. Het adjectief tropisch komt van het Griekse ‘tropein’, wat keren betekent. De term regenwoud geeft aan dat de bossen in natte klimaten te vinden zijn. Tropische regenwouden worden gekenmerkt door een gesloten complex van zowel verticaal als horizontaal gedifferentieerde levensgemeenschappen van voor het merendeel altijd groene vochtminnende loofboomsoorten en talrijke soorten kruiden, lianen, epifyten, hogere en lagere dieren en micro-organismen. De levensgemeenschappen zijn zeer complex met een ongekende verscheidenheid aan plantenen dierensoorten die in nauwe samenhang en onderlinge afhankelijkheid leven. Tropische regenwouden beslaan slechts 6% van het landoppervlak van de aarde, maar herbergen naar schatting meer dan de helft van de nu bekende organismen. De structuur (grootteklassenverdeling) en architectuur van de vegetatie is tevens zeer wisselend, zelfs wanneer er geen houtkap of andere menselijke ingrepen hebben plaatsgevonden. Hoogstammig, gesloten bos wordt met een variabele frequentie door openingen van verschillende grootte in het kronendak onderbroken, en hier en daar door lianenkluwens of door stukjes bos in stakenfase. Dit mozaïek van verschillende ontwikkelingsfasen is, door een voortdurende cyclus van sterfte en verjonging, aan voortdurende verandering onderhevig. Aantasting van tropische regenwouden Tropische regenwouden worden van oudsher door de mens gebruikt zonder veel inbreuk te doen op de integriteit van deze bossen. In de tweede helft van de twintigste eeuw vindt echter in toenemende mate ontbossing en degradatie plaats. Daaraan liggen directe en dieper liggende oorzaken ten grondslag. Naar schatting worden op dit moment zo’n 11 miljoen hectare per jaar ontbost en zo’n 4,5 miljoen hectare door commerciële houtexploitatie en brandhoutwinning aangetast. Behoefte aan land voor akkerbouw en veeteelt en aan brandhout zijn veruit de belangrijkste directe oorzaken van ontbossing en degradatie in de natte tropen met als belangrijkste achterliggende oorzaken overbevolking en heersende armoede. Hoewel houtwinning slechts voor een relatief beperkt percentage een directe oorzaak is van het verdwijnen van het tropisch regenwoud, gebeurt dit in bijna alle gevallen zonder acht te slaan op duurzame houtproductie en leidt dit vaak tot uitputting van de gezochte grondstoffen. Voorts baant de houtkap doorgaans de weg voor verdere kolonisatie van het geëxploiteerde areaal door mensen op zoek naar landbouwgrond. De meest algemene vorm van houtexploitatie in tropische regenwouden is middels (selectieve) uitkap. Deze kapmethode is selectief omdat slechts een beperkt aantal boomsoorten gekapt wordt en van deze soorten slechts die exemplaren die aan bepaalde minimum afmetingen en kwaliteitseisen voldoen. In het algemeen betreft het dan een relatief gering percentage van de aanwezige bomen. Commerciële uitkap van tro281 Reduced impact logging in the tropical rain forest of Guyana pisch hout vindt al meer dan twee eeuwen plaats, maar was voor de Tweede Wereldoorlog van bescheiden omvang. In het begin van de vijftiger jaren traden er echter radicale veranderingen op. Vergaande ontwikkelingen in de technologie in de houtverwerkende industrie en een stijgende mondiale behoefte aan houtproducten leidden tot een sterke groei in de wereldhandel in hout uit tropische regenwouden, met als gevolg een toename van het aantal bruikbare boomsoorten en een verlaging van de eisen die aan de dimensies en kwaliteit van de stammen gesteld werden. Samen met ontwikkelingen in de technologie in de houtkap, zoals de invoering van de motorzaag en in toenemende mate zwaardere en krachtigere machines voor wegenaanleg en de uitsleep van stammen, leidde dit tot een toenemende intensivering van de houtkap. Door deze ontwikkelingen heeft de term uitkap een misleidende betekenis gekregen. Het percentage van de aanwezige bomen die gekapt worden, is nog steeds niet erg groot, maar een groter percentage - met name bestaande uit jonge bomen - wordt tijdens de kap gedood of beschadigd, terwijl ook de bodem ernstig te lijden heeft. Effecten van de uitkap De schade die aan het bos wordt toegebracht bepaalt in grote mate de waarde van het resterende bos voor een toekomstige kap en de instandhouding van de regeneratieomstandigheden en de overige functies van het bos, en is dus van cruciale betekenis voor de duurzaamheid van het kapregime. Het vellen van de bomen en de uitsleep van de stammen zijn de twee belangrijkste oorzaken van schade aan het bos. De uitwerking van de kap op het bos komt in de verwijdering van een aanzienlijk deel van de volwassen bomen van commerciële soorten tot uitdrukking hetgeen een vermindering van zaadbomen inhoudt en tot een toenemende commerciële verarming van het bos kan leiden. Het gebruik van zware machines voor de extractie van de stammen leidt tot verdichting en verslemping van de bodem en tot ernstige schade aan alle vegetatie op het pad van de machine. Het vellen leidt tot gaten in het kronendak die vaak groter zijn dan de van nature gevormde gaten. Bovendien vindt er een synchronisatie van de vorming van openingen plaats. Deze veranderingen in het bosmozaïek kunnen vergaande consequenties voor de regeneratie processen hebben, zoals invasie door van nature weinig voorkomende ‘pionier’ boom- en lianensoorten. De schade die aan het bos wordt toegebracht varieert sterk met de wijze en intensiteit van de kap. Deze laatste twee factoren zijn onder meer afhankelijk van de dichtheid van de commercieel aantrekkelijke boomsoorten en dus afhankelijk van de bossamenstelling en marktsituatie ter plaatse. Verder speelt de schaal van de exploitatie en investeringsgraad van het houtbedrijf (type en aantal machines, kwaliteit management, machines en personeel) een belangrijke rol. Schadebeperkende houtoogstmethoden In een groot aantal landen begint men de gevaren van de bedreiging van het tropisch regenwoud in te zien. Op hoog politiek niveau komt dit tot uitdrukking in het kader van verdragen die in internationaal verband worden gesloten, zoals de Conferentie over Milieu en Ontwikkeling gehouden in Rio de Janeiro in 1992. De Food and Agriculture Organization van de Verenigde Naties (FAO) lanceerde het Tropical Forestry Action Plan in 1985. In 1990, stelde de Internationale Tropisch Hout Organisatie (ITHO) het 282 Samenvatting jaar 2000 als de deadline waarop de handel in tropisch hout zal worden beperkt tot hout dat uit duurzaam beheerde bossen is gewonnen. Gevoed door dergelijke initiatieven werd de interesse in geschiktere kapmethoden verhoogd. Dergelijke ‘schadebeperkende’ methodes werden eerst in Zuidoost-Azië en Australië en later in Zuid-Amerika geïntroduceerd. Deze methodes richten zich op een beperking van de oogstschade aan het ecosysteem, op maximalisering van de benutting van de gevelde stammen en op een verhoging van de doelmatigheid van met name de uitsleep en het wegtransport. Guyana Sinds het eind van de achttiende eeuw vindt commerciële houtwinning plaats in Guyana. Desondanks, heeft het land relatief een van de grootste bosarealen ter wereld en een verwaarloosbare mate van ontbossing. De goede staat van Guyana’s bossen is echter niet het resultaat van gericht beheer, maar eerder het gevolg van de - tot voor kort beperkte verkoopbaarheid van de in Guyana voorkomende houtsoorten en de afwezigheid van andere landgebruiksmogelijkheden voor het bosareaal, vanwege de lage vruchtbaarheid van de bodems en de extreem lage bevolkingsdichtheid (4 personen per km2). Tot voor kort hebben de ontwikkelingen in de mondiale houtindustrie en mechanisering van de houtwinning die in andere delen van de wereld voor een radicale omwentelingen hebben gezorgd, slechts een bescheiden invloed op de bosexploitatie in Guyana gehad, met name gezien het feit dat nog steeds slechts een handvol soorten aan de man gebracht konden worden en gezien de penibele economische situatie in het land. Deze situatie is sinds kort veranderd door de toevloed van internationale houtmaatschappijen, hetgeen samenhangt met houttekorten in met name Zuidoost-Azië en met de wijziging in het politiek bestel in Guyana, als gevolg waarvan buitenlandse investeringen gestimuleerd werden. De gebruikelijke wijze van houtwinning in Guyana verschilt weinig van de wijze waarop dit elders in de tropen gebeurt. Het grootste deel van het bosareaal is eigendom van de nationale overheid en bosexploitatie vindt plaats door particuliere ondernemingen die houtconcessies hebben die door de overheid worden uitgegeven. In de concessievoorwaarden worden aan de exploitatie eisen gesteld, die bedoeld zijn om de toekomstige houtwinning veilig te stellen, maar deze worden vanwege onvoldoende controle door een onderbemande bosdienst slecht nageleefd. Bovendien ontbreekt voldoende praktische kennis om het complexe ecosysteem te beheren. Eeuwenlang concentreerde de houtwinning zich op slechts één soort, Greenheart (Demerara Groenhart - Chlorocardium rodiei), die een houtsoort levert die befaamd is voor toepassingen in zout water. Al aan het eind van de negentiende eeuw werd bezorgdheid uitgesproken over de toekomstige beschikbaarheid van deze houtsoort en dit is sindsdien herhaaldelijk uitgesproken. Deze bezorgdheid werd nimmer bewaarheid, omdat keer op keer middels technologische ontwikkelingen nieuwe gebieden konden worden opengelegd. Momenteel lijkt de dreiging echter onafwendbaar. Ten eerste zijn de ongeëxploiteerde gebieden op een dergelijke afstand van exporthavens gelegen dat de oogstkosten - met name transportkosten - enorm zijn geworden en ten tweede dreigen er boycots voor hout dat uit niet duurzaam beheerde bossen gewonnen is. 283 Reduced impact logging in the tropical rain forest of Guyana Het Tropenbos-Guyana Programma De Nederlandse Stichting Tropenbos is in 1988 opgericht om het in 1986 door de Nederlandse overheid gestarte Tropenbos programma tot uitvoering te brengen. Tropenbos heeft als doel om bij te dragen aan het behoud en duurzaam gebruik en beheer van tropische regenwouden en het versterken van de wetenschappelijke basis en infrastructuur in landen met tropisch regenwoud. In nauwe samenwerking met Nederlandse universiteiten en instituten, en onderzoeksinstituten en overheden in een aantal tropenlanden zijn onderzoeksprogramma’s in een vijftal landen opgezet: Colombia, Guyana, Indonesië, Ivoorkust en Kameroen. Training en voorlichting zijn belangrijke elementen in de Tropenbos programma’s. Het Tropenbos-Guyana programma is in 1989 met een bilaterale overeenkomst tussen de regeringen van Guyana en Nederland van start gegaan. In het programma werken de Universiteit van Guyana, de Guyana National Resources Agency en de Universiteit Utrecht samen. De doelstelling van het programma is om kennis van het tropisch regenwoud in Guyana te verwerven en daarmee methoden van duurzaam bosgebruik en beheer te ontwikkelen. Hierbij wordt van een multidisciplinaire benadering gebruik gemaakt, waarbij onderzoek naar de water- en nutriëntenhuishouding, plantengroei, populatie dynamiek en biodiversiteit in ongekapt en gekapt bos wordt gedaan. Verder wordt onderzoek verricht naar de effecten van de grootte van kunstmatige openingen in het kronendak op een aantal biotische en abiotische kenmerken. Deze projecten spelen zich af in een houtconcessie in centraal Guyana, welke op dit moment in handen is van Demerara Timbers Limited. Recentelijk is het programma uitgebreid met studies op het gebied van niet-hout bosproducten, waarbij sociaal-economische, antropologische en populatie dynamische aspecten worden onderzocht. Dit deel van het programma wordt uitgevoerd in het North West District. Deze studie gaat over duurzame houtproductie van tropisch regenwoud, met name over schadebeperkende houtoogst. De doelstellingen van dit onderzoek waren: 1. het beschrijven en analyseren van de invloed van de conventionele houtoogst praktijk op het ecosysteem en de consequenties daarvan voor het herstel van het bos en de te verwachten toekomstige houtopbrengst; 2. het ontwikkelen van een schadebeperkend houtoogst systeem dat het bos in een zodanige staat achterlaat dat het bos zich kan ontwikkelen tot een staat die bosteeltkundig, ecologisch en economisch gewenst is; 3. het analyseren van het effect van de intensiteit van de kap op het herstel van het bos en het vaststellen van de intensiteit waarbij de voordelen van het gebruik van schadebeperkende houtoogst technieken, indien aanwezig, tenietgedaan worden; 4. het bestuderen van veranderingen in de kosten/baten verhouding bij een overgang van conventionele naar schadebeperkende houtoogst; 5. het bestuderen van de wenselijkheid van bosteeltkundige ingrepen na de kap met het oog op het herstel van de productiviteit van het bos. Het onderzoek werd op de Tropenbos onderzoekslocatie Pibiri nabij Mabura Hill in Centraal Guyana uitgevoerd, ongeveer 250 km ten zuiden van Georgetown. In 1993 284 Samenvatting zijn 15 permanente proefvlakken van 240×240 m2 uitgelegd, waarbinnen een meetvlak van 140×140 m2 is afgebakend. In deze meetvlakken zijn alle bomen met een dbh (diameter op borsthoogte) vanaf 20 cm gemeten, gemerkt en geïdentificeerd. Tevens werd de positie van deze bomen nauwkeurig bepaald. Bomen met een kleinere diameter zijn steekproefsgewijs gemeten. Aan de proefvlakken werden vijf verschillende behandelingen toegewezen, te weten een houtoogst met een intensiteit van 4 bomen per hectare, 8 bomen per hectare en 16 bomen per hectare; verder een houtoogst van 8 bomen per hectare te volgen door een bosteeltkundige behandeling na de oogst en een onbehandelde controle. Bij de kap in 1994 is een schadebeperkend houtoogstmethode gebruikt. Deze methode is in hoofdstuk 3 beschreven. Direct ten zuiden van de Tropenboslocatie zijn in 1995 drie stukken bos van 12 ha gemarkeerd, waarin een proefstrook van 500 × 20 m2 op dezelfde wijze is geïnventariseerd, als in de experimentele proefvlakken. Deze locaties zijn onderworpen aan ongecontroleerde, conventionele houtoogst in 1996. De proefopzet en de andere hierboven genoemde aspecten worden behandeld in hoofdstuk 1. Hoofdstuk 2 begint met een algemene beschrijving van de uitgangspunten voor het ontwerpen van een duurzaam bosbeheerssysteem voor houtproductie. Van duurzame houtproductie is sprake indien aan de voorwaarden van evenwicht tussen kap en bijgroei op een bepaalde oppervlakte wordt voldaan én de regeneratie-omstandigheden en de overige functies die het bos vervult, niet ondermijnd worden. Er zijn verschillende niveaus van intensiteit van beheer voor duurzame houtproductie. Dit kan variëren van uitkap zonder dat een bosteeltsysteem voor dat bos is ontwikkeld, een volledig gecontroleerde uitkap volgens een bepaald bosteeltsysteem, een volledig gecontroleerde uitkap gevolgd door bosteeltkundige ingrepen om het productieniveau van het bos te verhogen, tot een volledig gecontroleerde uitkap waarbij plaatselijk kunstmatig verjongd wordt (inboeten). Het spreekt voor zich dat hoe intensiever het beheer is, des te hoger de kosten zijn die aan een dergelijk beheer zijn verbonden. In de meeste gevallen richt het intensieve beheer zich op de potentieel meest waardevolle opstanden. Op dit moment is er onvoldoende kennis van de ecologische randvoorwaarden voor beheer. Tevens is er weinig bekend aangaande de groeisnelheden van de diverse soorten en hun populatie dynamiek in het algemeen en in het bijzonder hoe deze soorten na verschillende wijzen en intensiteiten van uitkap reageren. Dit hoeft echter geen belemmering te zijn om voor deze bossen een systeem voor duurzame houtproductie te ontwikkelen. Het bosteeltsysteem zou dan de waargenomen natuurlijke bosdynamiek moeten volgen - de ‘naturgemäße Waldbau’ die eind negentiende eeuw in Centraal Europa zijn intrede deed - waarbij men zich met name door het natuurlijke bosmozaïek van verschillende ontwikkelingsfasen laat sturen. De laatste twee decennia is er vrij veel onderzoek naar de natuurlijke bosdynamiek in de tropen en met name ten aanzien van de effecten van openingen in het kronendak op de vestiging en groei van zaailingen verricht. De voor een uitkapsysteem meest relevante conclusies die daaruit getrokken kunnen worden zijn: 285 Reduced impact logging in the tropical rain forest of Guyana ƒ ƒ overmatig grote gaten in het kronendak moedigen de vestiging van ongewenste lichtminnende boom- en lianensoorten aan, terwijl kleinere gaten zich doorgaans sluiten door zijdelingse uitbreiding van kronen van aangrenzende bomen; zaailingen die zich gevestigd hebben voordat de opening gevormd werd, spelen een belangrijke rol in de toekomstige samenstelling van het bos op die plek. Het is daarom van uitzonderlijk grote betekenis dat deze zaailingen intact gelaten worden. Hoofdstuk 2 vervolgt met een beschrijving van de ontwikkelingen in het beheer van natuurlijke tropische bossen wereldwijd en in Guyana in het bijzonder. De in de loop der tijd wereldwijd ontwikkelde bosteeltsystemen zijn in feite meestal op het Centraal Europese uitkapsysteem geënt, waarbij getracht wordt om de leeftijds- of grootteklassenverdeling van een heel bos of per soort om te vormen naar die van een bosteeltkundig ‘normaal’ bos. Normaliteit houdt in dat iedere leeftijdsklasse zodanig vertegenwoordigd is dat eindeloos een constante hoeveelheid hout uit het bos gewonnen kan worden, waarbij de frequentie in een bepaalde leeftijds- of grootteklasse telkens lager is dan die van de klasse daaronder om rekening te houden met natuurlijke sterfte. Om verscheidene redenen is dit systeem in de natte tropen nooit echt toegepast en het enige element dat van het traditionele uitkapsysteem is overgebleven, is de toepassing van een minimum diametergrens. Er zijn belangrijke bezwaren tegen het zich uitsluitend verlaten op een diametergrens. Zo kan dit tot gevolg hebben dat de meeste volwassen bomen van soorten met marktwaarde uit het bos verwijderd worden, welke bomen de zaailingen zouden moeten produceren voor de volgende rotatie. Verder leidt dit vaak tot negatieve selectie, waarbij slechtgevormde of rotte exemplaren overblijven. Mocht dit genetisch bepaald zijn dan zal dit de kwaliteit van de toekomstige opstand beïnvloeden. Tenslotte vertonen verschillende soorten al naar gelang hun temperament verschillende diameterklassenverdelingen en zou de selectie van de te kappen bomen daaraan aangepast moeten zijn. Hoofdstuk 2 eindigt met een beschrijving van de ontwikkeling van de bosexploitatie en het bosbeheer in Guyana door de eeuwen heen. Daarbij wordt met name aandacht aan de duurzaamheid van de productie van Greenheart geschonken en aan de vraag waarom de houtindustrie zich zolang op één soort heeft geconcentreerd. In hoofdstuk 3 worden de kenmerken van het schadebeperkende oogstsysteem beschreven. In het algemeen omvatten dergelijke systemen de volgende elementen: ƒ een 100% inventarisatie van de kapbare bomen en kartering daarvan en van het terrein voordat de kap begint; ƒ de planning van boswegen en uitsleeppaden voordat de kap begint; ƒ het kappen van lianen enige tijd voor de kap ter voorkoming van overmatige velschade; ƒ het bepalen van de velrichting voor de velling en het gebruik van bepaalde veltechnieken om deze gewenste velrichting daadwerkelijk aan te houden; 286 Samenvatting ƒ ƒ de maximale benutting van het per boom bruikbare hout door te hoge stobbes, het splijten tijdens het vellen en het te kort aftoppen te voorkomen, alsmede te voorkomen dat stammen in het bos worden achtergelaten; het lieren van stammen naar de uitsleeppaden in plaats van het ophalen bij de stobbe. Deze elementen maken ook deel uit van het schadebeperkende houtoogstsysteem dat in deze studie is toegepast. Bovendien is de selectie van de te kappen bomen in het systeem betrokken (dus niet gebaseerd op een integraal toegepaste minimum kapdiameter). De reden hiervan is dat eerder onderzoek in Guyana heeft aangetoond dat de openingen in het kronendak die bij de gebruikelijke kap in Greenheartbos worden gevormd, overmatig groot zijn. De hoofdoorzaak hiervan is dat Greenheart van nature in groepen voorkomt. Bij de gebruikelijke kap worden alle Greenhearts die aan de kwaliteitseisen voldoen geveld en worden dus vrij grote gaten gemaakt. Om de vorming van dergelijke grote gaten te ontmoedigen werd bepaald dat de te kappen bomen zo evenredig mogelijk verspreid over het proefvlak moesten voorkomen. Verder werd aandacht geschonken aan de schaarsheid van de diverse soorten in een proefvlak en werden minimum kapdiameters aangepast aan de maximum diameter die een soort kan bereiken. Gericht vellen In het algemeen zijn er een vijftal redenen om voor een bepaalde velrichting te kiezen: 1. om de veiligheid van de veller en zijn assistent te waarborgen; 2. om het breken van de gevelde stam te voorkomen; b.v. doordat deze over een obstakel (rots, andere gevelde stam) of over een kuil of bult in het terrein of hellingafwaarts wordt geveld; 3. om schade aan toekomstbomen te voorkomen; 4. om stammen in een positie te leggen die de uitsleep vergemakkelijkt; 5. om bomen in de richting van bestaande openingen in het kronendak te vellen, waardoor schade aan de overblijvende opstand wordt beperkt. Veelal kan niet aan al deze voorwaarden tegelijkertijd worden voldaan. Veldwaarnemingen voor de velling hadden aangegeven dat de dichtheid van toekomstbomen dermate hoog was dat het volledig vermijden van schade aan dergelijke bomen een onhaalbare kaart was. Het vellen van meer dan één boom in een gat werd ook afgeraden, omdat aangenomen werd dat de gaten dan te groot zouden worden. Daarom werd besloten om de velrichting aan te passen aan de ligging van de geplande sleepwegen, waarbij de bomen in een visgraatsgewijs patroon dienden te worden geveld. Gecontroleerde uitsleep Zoals gezegd, is één van de karakteristieken van de gebruikelijke houtoogst in Greenheartbos dat bomen groepsgewijs geveld worden. Er wordt daarbij amper rekening gehouden met de velrichting, ook al omdat de gebrekkige veltechniek dit nauwelijks toelaat. Zodoende worden de bomen doorgaans kriskras geveld, wat vaak resulteert in een wirwar van gevelde bomen. De skidder is weliswaar uitgevoerd met een lier maar 287 Reduced impact logging in the tropical rain forest of Guyana deze wordt meestal niet gebruikt voor het vergaren van stammen. Bovendien is de skidder uitgevoerd met een grote, zware haak die slechts met moeite onder een gevelde stam kan worden doorgeschoven. De gewoonlijke procedure is dan ook dat de skidder iedere gevelde stam met zijn blad verschuift - eventueel ergens op - om het aanhaken te vergemakkelijken. In het schadebeperkende oogstsysteem was de skidder voorzien van zogenaamde ‘choker-stroppen’. Deze aparte, lichte kabels worden aan de lierkabel gehaakt middels de zogenaamde ‘choker’ en een beslagring, een systeem dat werkt als een slot en sleutel. Stammen werden zo mogelijk gelierd. Hoofdstuk 4 behandelt de aan het bos toegebrachte schade bij de verschillende kapwijzen met verschillende kapintensiteiten. Vel- en uitsleepschade werden bepaald door het beschadigde bosoppervlak in kaart te brengen en door telling van beschadigde en gedode bomen. Velschade werd aan de hand van de grootte van de opening die in het kronendak ontstond door het wegvallen van de kroon van de gevelde boom en door kroonschade aan of ontwortelen van omringende bomen, geschat. Uitsleepschade werd bepaald door het oppervlak van de gebruikte uitsleeppaden en andere sporen van skidder manoeuvres (met name in de buurt van de stobbe) op te meten. Het totaal verlies aan kroonbedekking was bij beide methoden gelijk, mits de intensiteit tot 8 à 10 bomen per hectare beperkt bleef. Wanneer deze intensiteit overschreden werd, was er meer verlies aan kroonbedekking bij de schadebeperkende houtoogst. Nadere bestudering leidde tot de conclusie dat dit aan het verschil tussen groepsgewijze en gespreide kap te wijten was. In geval van groepsgewijze kap - conventionele methode - overlapten de gaten in het kronendak. Verder bleek de architectuur van het bos voor de kap een rol te spelen. Dit laatste werd mede veroorzaakt door de gevolgde opnamemethode. Voorafgaand aan de kap waren natuurlijke gaten (‘chablis’) niet zorgvuldig in kaart gebracht. Indien er veel natuurlijke ‘chablis’ waren, wat vaker het geval was in het bos waar de schadebeperkende methode was toegepast, kon het gebeuren dat de opening veroorzaakt door het vellen, gedeeltelijk overlapte met een reeds aanwezige opening. Bij het schatten van de grootte van opening kon niet duidelijk worden onderscheiden welk deel van de opening door het vellen veroorzaakt was. Bij toekomstige onderzoekingen is het dan ook raadzaam om natuurlijke chablis voorafgaand aan de kap zorgvuldig in kaart te brengen. De grootte van de individuele gaten in het kronendak lag bij de schadebeperkende methode gemiddeld rond 260 m2 bij een kapintensiteit van zowel 4 als 8 bomen per hectare en rond 570 m2 bij 16 bomen per hectare. Bij de gebruikelijke methode lag de gemiddelde grootte op 430 m2 bij een kapintensiteit van 8 bomen per hectare en op 780 m2 bij 16 bomen per hectare. Andere Tropenbos-Guyana studies hebben aangetoond dat de maximum toelaatbare gatgrootte rond de 300 m2 ligt (Ek & van der Hout in prep). Daarboven treden ongewenste bosteeltkundige en ecologische effecten op. Gezien de gemiddelde grootte van de individuele openingen in het kronendak is het schadebeperkende houtoogstsysteem aan te raden boven de gebruikelijke methode. De maximale kapintensiteit moet in dat geval onder de 10 bomen per hectare blijven. Het 288 Samenvatting is niet noodzakelijk om deze verder dan 8 bomen per hectare te verlagen, omdat dit niet tot kleinere openingen in het kronendak leidt. De gecontroleerde uitsleepmethode, waarbij stammen naar de skidder werden gelierd, leidde tot een onmiskenbare vermindering van het door de skidder bereden (en dus verdichtte) grondoppervlak. Bij de schadebeperkende methode bleef het beschadigde oppervlak, afhankelijk van de kapintensiteit, beperkt tot 5% tot 8% van het geoogste areaal, terwijl dit bij de gebruikelijke uitsleepmethode 30% tot 36% bedroeg. De belangrijkste verschillen in schade aan de blijvende opstand tussen de twee houtoogstsystemen was een vermindering van de schade veroorzaakt tijdens de uitsleep en van schade aan jonge bomen in het bijzonder door de invoering van het schadebeperkende systeem. Schade aan de blijvende opstand veroorzaakt door het vellen verschilde in de zin dat bij het schadebeperkende systeem relatief meer dikke (dbh ≥ 40 cm) en dunne (dbh < 20 cm) bomen werden geraakt, maar relatief minder (20 cm ≤ dbh < 40 cm) middelgrote bomen. Het feit dat relatief meer dikke bomen werden getroffen is gerelateerd aan het verschil in spreiding van de gekapte bomen. In geval van groepsgewijze kap zijn er simpelweg minder potentiële slachtoffers. Het gerichte vellen in het schadebeperkende systeem maakte het mogelijk om middelgrote bomen te ontwijken, wat ten koste ging van de dunne bomen. Groepsgewijze velling biedt zekere voordelen. Doordat er een overlap is in de aan de vegetatie toegebrachte schade, treedt er minder velschade aan de blijvende opstand op en is de reductie van de totale kroonbedekkingsgraad relatief minder. Het is daarom aan te bevelen om, waar mogelijk, twee bomen één opening te laten maken. Meer bomen per opening is niet gewenst, omdat de opening dan te groot wordt. In hoofdstuk 5 wordt op de verschillen in kosten en baten van de twee houtoogstsystemen ingegaan. Het is onlangs aangetoond dat maar weinig Guyanese houtbedrijven winst maken. Het schadebeperkend oogstsysteem vergt extra investeringen voorafgaand aan de kap, zoals het uitvoeren van een inventarisatie en andere planningsactiviteiten. Gezien deze situatie staan de bedrijven niet erg welwillend tegenover de invoering van een dergelijk oogstsysteem. Daartegenover staat dat studies elders hebben aangetoond dat deze extra investeringen tijdens de uitsleepfase terugverdiend worden. Om na te gaan of dit in de Guyanese situatie ook op gaat, is de doelmatigheid van de velling- en de uitsleepfase in detail bestudeerd. De totale oogstkosten behelzen natuurlijk ook de kosten die voorafgaand aan de kap worden gemaakt, zoals planning en wegenaanleg, en na de uitsleep, zoals het laden en lossen van trucks en wegtransport. Verder zijn er nog kosten verbonden aan wegenonderhoud, supervisie, logistiek en dergelijke. Deze laatste facetten konden in deze studie niet in detail bestudeerd worden, maar werden volgens een deductieve methode bepaald. Oogstkosten en opbrengsten worden door een groot aantal factoren beïnvloed. Deze zijn bijv. de staat van onderhoud en het type van de gebruikte machines, het opleidingsniveau en de ervaring van vellers en machinedrijvers, de hout- en basteigenschappen, diameter en vorm (wel of geen plankwortels) van de te vellen bomen, de 289 Reduced impact logging in the tropical rain forest of Guyana terreinomstandigheden en de afstand van het geoogste areaal tot de weg. Vanzelfsprekend waren er verschillen tussen de twee operaties in de afstand die afgelegd moest worden en in de soortensamenstelling en diameterklassenverdeling van de gekapte bomen. Deze verschillen werden er zo goed mogelijk uitgefilterd door regressiemodellen te gebruiken. In vergelijking met de gebruikelijke methode, leidde de schadebeperkende methode tot een verdrievoudiging van de voorbereidingskosten per m3 gewonnen hout en een verdubbeling van de velkosten per m3. Deze werden slechts ten dele gecompenseerd door een verlaging met 5% van de uitsleepkosten. De totale kosten per m3 hout op de landing bleek met 15% verhoogd. Daartegenover staat dat de opbrengst per hectare en per dag bij de schadebeperkende oogstmethode aanmerkelijk hoger was. Dit heeft te maken met een betere benutting van het bruikbare hout en een verhoging van de prestatie tijdens de uitsleepfase welke 10% hoger was dan bij de gebruikelijke methode. Door deze hogere opbrengstniveaus dalen de indirecte kosten die er per m3 gemaakt worden, zoals wegenaanleg en -onderhoud, supervisie, logistiek, etc. Het uiteindelijke resultaat was dat de invoering van het schadebeperkende oogstsysteem geen nadelige invloed op de kosten per m3 had. Hierbij moet worden opgemerkt dat het verschil in prestatie tijdens de uitsleepfase nog groter uitvalt, indien het effect van groepsgewijze velling wordt veronachtzaamd. Groepsgewijze velling biedt namelijk voordelen tijdens het verzamelen van te transporteren stammen. In hoofdstuk 6 wordt op de vooruitzichten voor eventuele bosteeltkundige ingrepen na de kap ingegaan. Duurzame houtproductie kan alleen gerealiseerd worden als de opbrengst van een kapcyclus in evenwicht is met de bijgroei en het bosmozaïek na de kap niet te ver afwijkt van de het natuurlijke bosmozaïek. Desalniettemin zal welk kapregime dan ook commerciële boomsoorten sterker benadelen dan niet-commerciële, omdat het aandeel van de eerstgenoemde in het kronendak sterker verlaagd wordt. Bosteeltkundige ingrepen hebben de volgende oogmerken: 1. het herstellen van de oorspronkelijke balans tussen commerciële en nietcommerciële soorten in het kronendak; 2. het stimuleren van de groei van toekomstbomen; 3. het bevorderen van de regeneratie van commerciële soorten. In de praktijk komt een dergelijke behandeling vaak neer op het verhogen van de blootstelling aan zonlicht van toekomstbomen door vrijstelling, ofwel het verwijderen van niet-commerciële naburige bomen die de toekomstbomen beschaduwen. Om uit te vinden of de lichtstelling van toekomstbomen door vrijstelling verbeterd kan worden, is een schatting gemaakt van de totale zonnestraling die door toekomstbomen na verschillende behandelingen ontvangen wordt. Daartoe is een driedimensionaal bosmodel voor twee proefvlakken van 2 hectare met het computerprogramma SILVISTAR geconstrueerd. Vijf verschillende behandelingen werden op deze wijze geëvalueerd: 290 Samenvatting 1. 2. 3. 4. 5. een systematische zuivering van alle niet-commerciële bomen vanaf 20 cm dbh; als 1) maar vanaf 30 cm dbh; vrijstelling van toekomstbomen met een dbh van 20 cm of groter door verwijdering van alle niet-commerciële bomen vanaf 20 cm dbh in een straal van 7 meter; als 3) maar dan in een straal van 10 meter; een gecombineerde behandeling waarbij alle niet-commerciële bomen vanaf 40 cm dbh over het gehele areaal zijn verwijderd en niet-commerciële bomen tussen 20 cm en 40 cm dbh in een straal van 10 meter rondom de toekomstbomen. In eerste instantie is de lichtstelling van de toekomstbomen in de proefvlakken zonder ingreep bekeken. Het bleek dat bijna de helft van de toekomstbomen al meer dan 80% van de maximaal haalbare straling ontving en dat de mogelijkheden tot vrijstelling beperkt waren, omdat ongeveer de helft van de beschaduwende bomen zelf toekomstbomen waren. Vervolgens is er bekeken welke niet-commerciële bomen in het proefvlak toekomstbomen beschaduwden. Alle ingrepen verbeterden de lichtstelling van toekomstbomen. Alle behandelingen deden dat in dezelfde mate behalve de vrijstelling in een straal van 7 meter, die minder succesvol was. Het bleek dat alle behandelingen meer bomen verwijderden dan noodzakelijk was, waarbij de systematische zuivering vanaf 30 cm dbh het minst verspillend was. Deze laatste ingreep is echter een ongedifferentieerde behandeling die geen rekening houdt met de dichtheid van de toekomstbomen. Het mag dan wel de minst verspillende ingreep in deze twee proefvlakken zijn geweest, maar deze ingreep zal in stukken bos met lagere dichtheden aan toekomstbomen zeker minder gunstig afsteken. Het is daarom aan te raden om zich tot een vrijstelling in een straal van 10 meter te beperken. Aangezien ook deze ingreep onnodig bomen elimineerde is enige aanpassing gewenst. De meest elegante en effectieve aanpassing is om alleen die naburige bomen te verwijderen die hoger zijn dan de vrij te stellen toekomstboom. Ten slotte is onderzocht of het mogelijk is om beschaduwende bomen aan te wijzen op grond van allometrische relaties, zoals de relatie tussen boomhoogte en stam diameter en tussen kroondiameter en stamdiameter. Alhoewel er sterke correlaties tussen deze dimensies bleken te bestaan, bleek dat dergelijke relaties niet gebruikt konden worden om aan te geven welke bomen toekomstbomen beschaduwden. De belangrijkste reden daarvoor was dat kronen in het bos doorgaans excentrisch zijn ten opzichte van de stamvoet. Conclusie Hoofdstuk 7 bevat een samenvatting en geïntegreerde discussie. De conclusies die daaruit volgden zijn hier weergegeven. Deze studie heeft aangetoond dat de gebruikelijke wijze van houtwinning uit Greenheartbos niet duurzaam is. Het ontwikkelde schadebeperkende houtoogstsysteem slaagde erin de uitsleepschade drastisch te verminderen, met name in de door de velling veroorzaakte openingen in het kronendak. Tevens resulteerde het systeem in een bosmozaïek dat beter aansluit bij het natuurlijke bosmozaïek en dus de natuurlijke regeneratie-omstandigheden. Dit laatste gold echter alleen dan wanneer de kapintensiteit beperkt was tot 10 bomen per hectare. 291 Reduced impact logging in the tropical rain forest of Guyana Het schadebeperkende houtoogstsysteem bestond uit de gebruikelijke schadebeperkende elementen - zoals bijv. planning van uitsleeppaden, lianenkap, gericht vellen - criteria voor het selecteren van de te vellen bomen en een verbetering van de benutting van het beschikbare hout op stam. Deze kenmerken resulteerden in een ingewikkeld samenspel van soms tegengestelde effecten. De wijziging van groepsgewijze naar gespreide kap was weliswaar een van de belangrijkste elementen om het gewenste bosmozaïek te bereiken, maar was in andere opzichten nadelig. Door de te vellen bomen te spreiden werd de schade aan de blijvende opstand per gevelde boom en het grondoppervlak bereden door de skidder per uitgesleepte stam vergroot. Tevens werd de doelmatigheid van de uitsleep benadeeld, omdat de machine heen en weer moest rijden om stammen op te halen, terwijl bij groepsgewijze velling een bundel in een keer vergaard kon worden. Het werd ook aangetoond dat het toelaatbaar is om twee bomen in één gat te vellen. Gezien de genoemde voordelen van groepsgewijze velling is het dan ook aan te raden dit, waar mogelijk, te doen. Hoewel het schadebeperkende oogstsysteem niet duurder is dan de gebruikelijke kap, zijn de opbrengsten lager. Dit is niet zozeer gelegen in een verschil in volume opbrengst, dan wel in de soortensamenstelling van dat volume. Door de overgang van groepsgewijze naar gespreide kap, zal de concessionaris enkele Greenhearts moeten laten staan, die hij anderszins zou oogsten. Het belemmeren van de vrijheid van de concessionaris in zijn keuze van de bomen die hij wil oogsten is dus de belangrijkste barrière bij het invoeren van bosteeltkundig en ecologisch verantwoord bosbeheer en dit zal waarschijnlijk voorlopig zo blijven. 292 References REFERENCES Abdulhadi, R., K. Kartawinata & S. Sukardjo. 1981. Effects of mechanized logging in the lowland dipterocarp forest at Lempake, East Kalimantan. The Malaysian Forester 44(2&3): 407-416. Alder, D. & T.J. Synnott. 1992. Permanent sample plot techniques for mixed tropical forest. Tropical Forestry Paper 25. Oxford Forestry Institute, Oxford, UK. Alexandre, D.Y. 1989. Dynamique de la régénération naturelle en forêt dense de Côte d'Ivoire. Études et thèses. Office de la Recherche Scientifique et Technique Outre-Mer, Paris, France. Alvarez-Buylla, E.R. & M. Martinez-Ramos. 1990. Seed bank versus seed rain in the regeneration of a tropical pioneer tree. Oecologia 84: 314-325. Anderson, M.C. 1964. Studies of the woodland light climate I and II. Journal of Ecology 52: 27-41 and 643-663. Anonymous. 1996. The productive sectors: forest management. in: Government of Guyana. National Development Strategy Volume 4, Chapter 30. http://www.guyana.org/NDS/chap30.htm. Anonymous. 1998. Forestry in Guyana; market summary 1997. Guyana Forestry Commission, Economics Unit, Policy and Planning Division, Georgetown, Guyana. Apel, K. 1993. Arbeitsstudien im Forstbetrieb; Terminologie und Gliederungsprinzipien. Allgemeine Forst und Jagd Zeitung 9/10: 157-162. Appanah, S. & F.E. Putz. 1984. Climber abundance in virgin dipterocarp forest and the effect of pre-felling climber cutting on logging damage. The Malaysian Forester 47(4): 335-342. Asabere, P.K. 1987. Attempts at sustained yield management in the tropical high forests of Ghana. in: Mergen, F. & J.R. Vincent (eds.). Natural management of tropical moist forests: 47-69. Yale University, School of Forestry and Environmental Studies, New Haven, CT, USA. Aubréville, A. 1938. La forêt coloniale: les forêts de l'Afrique occidentale française. Annales Académie Sciences Coloniale 9: 1-245. Augspurger, C.K. 1984. Light requirements of neotropical tree seedlings: a comparative study of growth and survival. Journal of Ecology 72: 777-795. Barreto, P., P. Amaral, E. Vidal & C. Uhl. 1998. Costs and benefits of forest management for timber production in eastern Amazonia. Forest Ecology and Management 108: 9-26. Barton, A.M. 1984. Neotropical pioneer and shade tolerant species: do they partition treefall gaps? Tropical Ecology 25: 196-202. Baur, G.N. 1964. The ecological basis of rainforest management. Forestry Commission of New South Wales, Sydney, Australia. Bernasconi, G. 1998. L'évolution de la sylviculture peut-elle contribuer à faire en sorte que notre civilisation se comprenne de mieux en mieux. Schweizerische Zeitschrift für Forstwesen 149(1): 41-49. Bertault, J.-G. & P. Sist. 1995. Impact de l'exploitation en forêt naturelle. Revue Bois et Forêts des Tropiques 245: 5-13. Bertault, J.-G. & P. Sist. 1997. An experimental comparison of different harvesting intensities with reduced-impact and conventional logging in East Kalimantan, Indonesia. Forest Ecology and Management 94: 209-218. Bijlsma, R.J. 1990. Het RIN-bosecologisch informatiesysteem SILVI-STAR; documentatie van FOREYEprogrammatuur en subprogramma's [in Dutch: The RIN-forest ecological information system SILVI-STAR; documentation of FOREYE software]. RIN rapport 90/1. Institute of Forestry and Nature Research, Wageningen [Leersum], The Netherlands. Biolley, H. 1930. Le bucheron artisan. in: Forêts de mon pays: 152-154. Editions Delachaux et Niestlé SA, Neuchâtel, Switzerland. Blate, G.M. 1997. Sustainable forest management in Brazil. The Tropical Forest Foundation's Low Impact Logging Programme integrates demonstration, training and research. ITTO Tropical Forest update 7(3): 14-15. Boerboom, J.H.A. 1964. De natuurlijke regeneratie van het Surinaamse mesophytische bos na uitkap [in Dutch: The natural regeneration of the Surinamese mesophytic forest after selective cut, with English summary]. Wageningen Agricultural University, Wageningen, The Netherlands. 293 Reduced impact logging in the tropical rain forest of Guyana Bongers, F. & J. Popma. 1988. Trees and gaps in a Mexican tropical rain forest; species differentiation in relation to gap-associated environmental heterogeneity. PhD thesis, Utrecht University, Utrecht, the Netherlands. Bonhomme, R. & P. Chartier. 1972. The interpretation and automatic measurement of hemispherical photographs to obtain sunlit foliage area and gap frequency. Israel Journal of Agricultural Research 22: 53-61. Bonnéhin, L. 1999 (in press). Domestication paysanne de quelques espèces fruitières de la forêt de Taï. PhD thesis. Wageningen Agricultural University, Wageningen, the Netherlands. Boomgaard, P. 1988. Forests and forestry in colonial Java, 1677-1942. in: Dargavel, J., K. Dixon & N. Sample (eds.). Changing tropical forests. Historical perspectives on today's challenges in Asia, Australasia and Oceania : 59-87. Centre for Resource and Environmental Studies, Australian National University, Canberra, Australia. Borhan, M., B. Johari & E.S. Quah. 1987. Studies on logging damage due to different methods and intensities of harvesting in a hill Dipterocarp forest of Peninsular Malaysia. The Malaysian Forester 50(2): 135-147. Bormann, F.H. & G.E. Likens. 1979. Pattern and process in a forested ecosystem. Springer-Verlag, Berlin-Heidelberg, Germany. Brandani, A., G.S. Hartshorn & G.H. Orians. 1988. Internal heterogeneity of gaps and species richness in Costa Rican tropical wet forest. Journal of Tropical Ecology 4: 99-119. Brasnett, N.V. 1953. Planned management of forests. George Allen & Uswin, London, UK. Brils, C.A.J. 1995. A time study of directional felling in a tropical rain forest in Guyana. M.Sc.thesis, Department of Forestry, Wageningen Agricultural University, Wageningen, the Netherlands. Brils, C.A.J. & E.A. Laan. 1995. Gaps and damage inflicted by directional felling in a tropical rain forest in Guyana. M.Sc.thesis, Department of Forestry, Wageningen Agricultural University, Wageningen, the Netherlands. Brokaw, N.V.L. 1982a. The definition of treefall gap and its effect on measures of forest dynamics. Biotropica 14: 158-160. Brokaw, N.V.L. 1982b. Treefalls, frequency, timing and consequences. in: Leigh, E.G., A.S. Rand & D.M. Windsor (eds.). The ecology of a tropical rain forest; seasonal rhythms and long-term changes: 101-108. Smithsonian Institution, Washington D.C., USA. Brokaw, N.V.L. 1985a. Treefalls, regrowth, and community structure in tropical forests. in: Pickett, S.T.A. & P.S. White (eds.). The ecology of natural disturbance and patch dynamics: 53-69. Academic Press, New York, USA. Brokaw, N.V.L. 1985b. Gap-phase regeneration in a tropical forest. Ecology 66(3): 682-687. Brokaw, N.V.L. & S.M. Scheiner. 1989. Species composition in gaps and structure of a tropical forest. Ecology 70(3) - Special Feature: 538-541. Brouwer, L.C. 1996. Nutrient cycling in pristine and logged tropical rain forest; a study in Guyana. PhD thesis, Utrecht University. Tropenbos-Guyana Series 1, Georgetown, Guyana. Brown, N.D. 1993. The implications of climate and gap microclimate for seedling growth conditions in a Bornean lowland rain forest. Journal of Tropical Ecology 9: 153-168. Brown, N.D. 1996. A gradient of seedling growth from the centre of a tropical rain forest canopy gap. Forest Ecology and Management 82: 239-244. Brown, N.D. & T.C. Whitmore. 1992. Do dipterocarp seedlings really partition tropical rain forest gaps? in: Marshall, A.G. & M.D. Swaine (eds.). Tropical rain forest: disturbance and recovery (Philosophical transactions of the Royal Society of London, Series B 335): 369-378. The Royal Society, London, UK. Bruijnzeel, L.A. & W.R.S. Critchley. 1994. Environmental impacts of logging moist tropical forests. Humid Tropics Programme Series 7. UNESCO / Man and the Biosphere: International Hydrological Programme, Paris, France. Brunberg, B., R. Gårdh & P. Lindgren. 1984. Felling manual. Forskningsstiftelsen ‘Skogsarbeten’ (Forest Operations Institute), Stockholm, Sweden. Brünig, E.F. 1983. Vegetation structure and growth. in: Golley, F.B. (eds.). Tropical rain forest ecosystem: structure and function (Ecosystems of the world 14A). Elsevier, Amsterdam, the Netherlands. Burgess, P.F. 1971. The effect of logging in hill dipterocarp forest. Malayan Nature Journal 24: 231-237. Burgess, J.C. 1993. Timber production, timber trade and tropical deforestation. Ambio 22: 136-143. 294 References Campbell, E.J.F. & D.M. Newberry. 1993. Ecological relationships between lianas and trees in lowland rain forest in Sabah. Journal of Tropical Ecology 9: 469-490. Canham, C.D. 1988a. Growth and canopy architecture of shade-tolerant trees: response to canopy gaps. Ecology 69(3): 786-795. Canham, C.D. 1988b. An index for understorey light levels in and around canopy gaps. Ecology 69(5): 1634-1638. Canham, C.D. 1989. Different responses to gaps among shade-tolerant tree species. Ecology 70(3) - Special Feature: 548-550. Canham, C.D., J.S. Denslow, W.J. Platt, J.R. Runkle, T.A. Spies & P.S. White. 1990. Light regimes beneath closed canopies and tree-fall gaps in temperate and tropical forests. Canadian Journal of Forest Research 20: 620-631. Castanet, P. 1989. Exploitation forestière en forêt dense guyanaise. Rapport technique. École Nationale du Genie Rural des Eaux et Forêts / Centre Technique Forestier Tropical, Kourou, Guyane française. Caterpillar. 1986. Caterpillar performance handbook. 17th edition. Caterpillar Tractor Co., Peoria, IL, USA. Catinot, R. 1997. L'aménagement durable des forêts denses tropicales humides. Association Technique Internationale des Bois Tropicaux, Paris, France. Cedergren, J. 1996. A silvicultural evaluation of stand characteristics, pre-felling climber cutting and directional felling in a primary dipterocarp forest in Sabah, Malaysia. PhD thesis, Swedish University of Agricultural Sciences, Umeå, Sweden. Cedergren, J., J. Falck, A. Garcia, F. Goh & M. Hagner. 1995. Impact of selective logging on silvicultural values in a mixed Dipterocarp forest in Sabah, Malaysia. in: Dykstra, D.P. (ed.). Proc. IUFRO workshop ‘Forest operations for sustainable forestry in the tropics’ (International Union of Forestry Research Organisations Subject Group S3.05-00): 39-45, Tampere, Finland. Chai, D.N.P. & M.P. Udarbe. 1977. The effectiveness of current silvicultural practice in Sabah. The Malaysian Forester 40(1): 27-35. Champion, H.G. 1936. The regeneration of tropical evergreen forests. Indian Forester 62(5): 289-294. Chandarpal, N. 1999. Sustainable development and exploratory forestry leases in Guyana. http://www.guyana.org.sustain.htm. Office of the President, Georgetown, Guyana. Chapman, C.A. & L.J. Chapman. 1997. Forest regeneration in logged and unlogged forests of Kibale National Park, Uganda. Biotropica 29(4): 396-412. Charles-Dominique, P., P. Blanc, D. Larpin, M.-P. Ledru, B. Riéra, C. Sarthou, M. Servant & C. Tardy. 1998. Forest perturbations and biodiversity during the last ten thousand years in French Guiana. Acta Oecologica 19(3): 295-302. Ciancio, O. & S. Nocentini. 1997. Forest management between ecology, economy and ethics. in: Ciancio, O. (ed.). The forest and man: 225-236. Accademia Italiana di Scienze Forestali, Florence, Italy. Clark, D.A. & D.B. Clark. 1992. Life history diversity of canopy and emergent trees in a neotropical rain forest. Ecological Monographs 62(3): 315-344. Clarke, E.C. 1956. The regeneration of worked-out Greenheart (Ocotea rodiaei) forest in British Guiana. Empire Forestry Review 35: 173-183. Clément, J. 1997. Le développement de la pensée et des actions forestières tropicales depuis 1946. Revue Bois et Forêts des Tropiques 252(2): 5-23. Colchester, M. 1994. The new sultans; Asian loggers move in on Guyana's Forests. The Ecologist 24(2): 45-52. Colchester, M. 1997. Guyana - fragile frontier; loggers, miners and forest peoples. Latin American Bureau / World Rainforest Movement, London, UK. Connell, J.H. & M.D. Lowman. 1989. Low-diversity tropical rain forests: some possible mechanisms for their existence. The American Naturalist 134(1): 88-119. Conway, S. 1982. Logging practices: principles of timber harvesting systems. Miller Freeman Publications, San Francisco, USA. Coudurier, T. 1992. Sur la place des lianes dans la forêt guyanaise. PhD thesis, Université de Montpellier, France. Cousens, J.E. 1958. A pilot sampling scheme in the regenerated forests of Perak. Malayan Forestry Research Pamphlet 23. 295 Reduced impact logging in the tropical rain forest of Guyana Crome, F.H.J., L.A. Moore & G.C. Richards. 1992. A study of logging damage in upland rain forest in north Queensland. Forest Ecology and Management 49: 1-29. Dalling, J.W., M.D. Swaine & N.C. Garwood. 1997. Soil seed bank community dynamics in seasonally moist lowland tropical forest, Panama. Journal of Tropical Ecology 13: 659-680. Daniels, R.F., H.E. Burkhart & T.R. Clason. 1986. A comparison of competition measures for predicting growth of loblolly pine trees. Canadian Journal of Forestry Research 16: 1230-1237. Davis, T.A.W. & P.W. Richards. 1933, 1934. The vegetation of Moraballi Creek, British Guiana; an ecological study of a limited area of tropical rain forest I and II. Journal of Ecology 21: 350-384 and 22: 106-155. Dawkins, H.C. 1958. The management of natural tropical high forest with special reference to Uganda. Institute Paper 34. Imperial Forestry Institute, University of Oxford, Oxford, UK. Dawkins, H.C. 1961. New methods of improving stand composition in tropical forests. Caribbean Forester 22: 12-20. Dawkins, H.C. & M.S. Philip. 1998. Tropical moist forest silviculture and management; a history of success and failure. CAB international, Wallingford, UK. De Foresta, H. & M.-F. Prévost. 1986. Végétation pionnière et graines du sol en forêt guyanaise. Biotropica 18(4): 279-286. De Graaf, N.R. 1986. A silvicultural system for natural regeneration of tropical rain forest in Suriname. Ecology and management of tropical rain forests in Suriname 1. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. De Graaf, N.R. 1988. The CELOS management system: a polycyclic method for sustained timber production in South American rain forest. in: Simpósio internacional sobre alternativas para o desmatamento, Belém, Pará, Brasil. De Graaf, N.R. & J. Hendrison. 1993. The quest for sustainability; silvicultural research in Suriname since 1900. in: Wood, P.J., J.K. Vanclay & Wan Razali Wan Mohd. (eds.). Proc. of the tropical silviculture workshop at the IUFRO centennial conference, Berlin, Germany: 101-111. International Union of Forestry Research Organisations, Vienna, Austria / Forest Research Institute, Malaysia. De Graaf, N.R., R.L.H. Poels & R.S.A.R. Van Rompaey, in press. Effect of silvicultural treatment on growth and mortality of rainforest in Surinam over long periods. Forest Ecology and Management De Milde, R. 1970. Assessment of hidden defect on standing trees during inventory work. FO: SF/GUY 9, Technical Report 4. United Nations Development Programme / Food and Agriculture Organization of the United Nations, Georgetown, Guyana. De Vries, P.G. 1986. Sampling theory for forest inventory. Springer-Verlag, Berlin-Heidelberg, Germany. Denslow, J.S. 1980. Gap partitioning among tropical rainforest trees. Biotropica 12 (Supplement): 47-55. Denslow, J.S. 1987. Tropical rainforest gaps and tree species diversity. Annual Review of Ecology and Systematics 18: 431-451. Denslow, J.S. 1996. Functional group diversity and responses to disturbance. in: Orians, G., R. Dirzo & J.H. Cushman (eds.). Biodiversity and ecosystem processes in tropical forests (Ecological Studies 122): 134-151. Springer-Verlag, Berlin-Heidelberg, Germany. Denslow, J.S. & G.S. Hartshorn. 1994. Treefall gap environments. in: McDade, L.A., K.S. Bawa, H.A. Hespenheide & G.S. Hartshorn (eds.). La Selva: ecology and natural history of a neotropical rain forest: 120-127. University of Chicago Press, Chicago, USA. Diggle, P.J. 1983. Statistical analysis of spatial point patterns. Mathematics in biology Academic Press, London, UK. Dupuy, B., J.-G. Bertault, F. Doumbia, A. Diahuissié, R. Brevet & K. Miezan. 1997. Régénération naturelle en forêt dense ivoirienne de production. Revue Bois et Forêts des Tropiques 254(4): 25-37. Dupuy, B., L. Durrieu de Madron & Y. Petrucci. 1998. Sylviculture des peuplements naturels en forêt dense humide africaine; acquis et recommandations. Revue Bois et forêts des Tropiques 257(3): 5-22. Durrieu de Madron, L. 1994. Mortalité des arbres en forêt primaire de Guyane française. Revue Bois et Forêts des Tropiques 239: 43-57. Dykstra, D.P. 1996. FAO model code of forest harvesting practice. Food and Agriculture Organization of the United Nations, Rome, Italy. Dykstra, D.P. & R. Heinrich. 1992. Sustaining tropical forests through environmentally sound harvesting practices. Unasylva 43: 9-15. 296 References Édelin, C. 1991. Nouvelles données sur l’architecture des arbres sympodiaux: le concept de plan d’organisation. Naturalia Monspeliensia, special issue “L’Arbre. Biologie et Développement":127-154. Ek, R.C. 1997. Botanical diversity in the tropical rain forest of Guyana. PhD thesis, Utrecht University, Tropenbos-Guyana Series 4, Georgetown, Guyana. Ek, R.C. & H. ter Steege. in prep. The flora of the Mabura Hill area. Ek, R.C. & P. van der Hout. in prep. Botanical diversity of Greenheart dominated mixed rain forest near Mabura Hill, Guyana. Ek, R.C., C.A. van der Kooij & J.A.C. van Dam. in prep. The effects of logging on the liana vegetation in Greenheart dominated tropical rain forest in Guyana. Estève, J. 1983. La destruction du couvert forestier consécutive à l'exploitation forestière de bois d'oeuvre en forêt tropicale humide africaine ou américaine. Revue Bois et Forêts des Tropiques 201: 77-84. Evans, G.C. & D.E. Coombe. 1959. Hemispherical and woodland canopy photography and the light climate. Journal of Ecology 47: 103-113. Fanshawe, D.B. 1944a. Improvement operations, Labakabra-Tiger creek area, 1937-1941. Internal report. Forestry Department, Georgetown, Guyana. Fanshawe, D.B. 1944b. Improvement operations, Moraballi-Seba, 1938-1943. Internal report. Forestry Department, Georgetown, Guyana. Fanshawe, D.B. 1952. The vegetation of British Guiana; a preliminary review. Institute Paper 29. Imperial Forestry Institute, University of Oxford, Oxford, UK. Fanshawe, D.B. 1954. Forest types of British Guiana. Caribbean Forester 15: 73-111. Fanshawe, D.B. 1961. Forest products of British Guiana I: principal timbers. Forestry Bulletin (new series) 1. Forest Department, British Guiana. FAO. 1974. Logging and log transport in tropical high forest. FAO Forestry Development Paper 18. Food and Agriculture Organization of the United Nations, Rome, Italy. FAO. 1977. Forest roads and harvesting systems. FAO Forestry Paper 2. Food and Agriculture Organization of the United Nations, Rome, Italy. FAO. 1978. Assessment of logging costs from forest inventories in the tropics. FAO Forestry Paper 10. Food and Agriculture Organization of the United Nations, Rome, Italy. FAO. 1981. Tropical Forest Resources Assessment Project. Forest resources of tropical America. Technical report 3. Food and Agriculture Organization of the United Nations, Rome, Italy. FAO. 1997. State of the world's forest. Food and Agriculture Organization of the United Nations, Rome, Italy. Favrichon, V. 1991. Sur quelques relations entre la croissance des arbres et la structure du peuplement en forêt semi-décidue (République Centreafricaine). Mémoire technique DEA, Laboratoire de Biométrie, Génétique et Biologie des Populations, Université de Paris VI, Paris, France. Favrichon, V. 1995. Modèle matriciel déterministe en temps discret - Application à l'étude de la dynamique d'un peuplement forestier tropical humide (Guyane française). PhD thesis, Université Claude Bernard Lyon I, Lyon, France. Favrichon, V. 1997. Réaction de peuplements forestiers tropicaux à des interventions sylvicoles. Revue Bois et Forêts des Tropiques 254(4): 5-24. Fearnside, P.M. 1989. Forest management in Amazonia: the need for new criteria in evaluating development options. Forest Ecology and Management 27: 61-79. Flaming, L. 1995. An economic analysis of the timber industry in Guyana; prospects for strengthening state capacity and private incentives for sustainable forest management. Harvard University, Cambridge, Mass., USA. Flemmich, C.O. 1963. Report to the Government of British Guiana on the marketing of wood and wood products with particular reference to the export of timber. Report No. 1737. Food and Agriculture Organization of the United Nations, Rome, Italy. Florence, J. 1981. Chablis et sylivgénèse dans un forêt dense humide sempervirente du Gabon. Thèse de Spécialité, Université de Strasbourg, Strasbourg, France. Fox, J.E.D. 1968a. Logging damage and the influence of climber cutting prior to logging in lowland dipterocarp forest of Sabah. The Malayan Forester 31(4): 326-347. Fox, J.E.D. 1968b. Defect, damage and wastage. The Malayan Forester 31(3): 157-164. 297 Reduced impact logging in the tropical rain forest of Guyana Fox, J.E.D. 1976. Constraints on the natural regeneration of tropical moist forest. Forest Ecology and Management 1: 37-65. Fröhlich, M. & H.D. Quednau. 1995. Statistical analysis of the distribution pattern of natural regeneration in forests. Forest Ecology and Management 73: 45-57. Gérard, J., R.B. Miller & B.J.H. Ter Welle. 1996. Major timber trees of Guyana; timber characteristics and utilization. Tropenbos Series 15. The Tropenbos Foundation, Wageningen, the Netherlands. Gillman, G.P., D.F. Sinclair, R. Knowlton & M.G. Keys. 1985. The effect on some soil chemical properties of the selective logging of a North Queensland rainforest. Forest Ecology and Management 12: 195-214. GNRA. 1989. National forestry action plan 1990-2000. Guyana Natural Resources Agency / Guyana Forestry Commission / Canadian International Development Agency, Georgetown, Guyana. Gordon, W.A. 1961. Forest management in the Caribbean. Caribbean Forester 22: 21-25. Gourlet-Fleury, S. 1992. Simulation d'éclaircies sur le dispositif sylvicole de Paracou (Guyana française): recherche de méthodes d'intervention en forêt, après exploitation. Mémoire technique DEA, Laboratoire de Biométrie, Génétique et Biologie des Populations, Université Claude Bernard Lyon I, France. Gourlet-Fleury, S. 1997. Modélisation individuelle spatialement explicite de la dynamique d'un peuplement de forêt dense tropicale humide. PhD thesis, Université Claude Bernard Lyon I, France. Gómez-Pompa, A. & A. Kaus. 1990. Traditional management of tropical forests in Mexico. in: Anderson, A.B. (ed.). Alternatives to deforestation: steps towards sustainable use of the Amazon rain forest: 45-64. Colombia University Press, New York, NY, USA. Gray, J.W. 1983. Forest revenue systems in developing countries; their role in income generation and forest management strategies. FAO Forestry paper 43. Food and Agriculture Organization of the United Nations, Rome, Italy. Guariguata, M.R. & J.M. Dupuy. 1997. Forest regeneration in abandoned logging roads in lowland Costa Rica. Biotropica 29(1): 15-28. Gullison, R.E. & J.J. Hardner. 1993. The effects of road design and harvest intensity on forest damage caused by selective logging: empirical results and a simulation model from the Bosque Chimanes, Bolivia. Forest Ecology and Management 59: 1-14. Hallé, F., R.A.A. Oldeman & P.B. Tomlinson. 1978. Tropical trees and forests; an architectural analysis. Springer-Verlag, Heidelberg, Germany. Hamilton, G.J. 1969. The dependence of volume increment of individual trees on dominance, crown dimensions, and competition. Forestry 42: 133-144. Hammond, D.S. in prep. Guianan forest dynamics: action and reaction in contemporary, historical and evolutionary time. in: Ter Steege, H. & D.S. Hammond (eds.). The rainforests of the Guianas.(Geobotany Series). Kluwer Scientific Publications, Dordrecht, the Netherlands. Hammond, D.S. & V.K. Brown. 1992. The ecological basis of recruitment and maintenance of timber species in the forests of Guyana. Interim Group Report DSH5. Department of Biology, Imperial College, Silwood Park, Ascot, UK. Hammond, D.S. & V.K. Brown. 1995. Seed size of woody plants in relation to disturbance, dispersal, soil type in wet neotropical forest. Ecology 76(8): 2544-2561. Hammond, D.S. & H. Ter Steege. 1998. Propensity for fire in Guianan rainforests. Conservation Biology 12(5): 944-947. Hammond, D.S., S. Gourlet-Fleury, P. van der Hout, H. ter Steege & V.K. Brown. 1996. A compilation of known Guianan timber trees and the significance of their disposal mode, seed size and taxonomic affinity to tropical rain forest management. Forest Ecology and Management 83: 99-116. Hammond, D.S., R.J. Zagt, I.A. Welch & P. van der Hout. in prep. The ecology, utilisation and management of Greenheart and Greenheart-bearing mixed forests in Guyana. Tropenbos Series. The Tropenbos Foundation, Wageningen, the Netherlands. Hart, T.B., J.A. Hart & P.G. Murphy. 1989. Monodominant and species-rich forests of the humid tropics: causes for their co-occurrence. The American Naturalist 133(5): 613-633. Hartshorn, G.S. 1978. Tree falls and tropical forest dynamics. in: Tomlinson, P.B. & M.H. Zimmermann (eds.). Tropical trees as living systems : 617-638. Cambridge University Press, Cambridge, UK. Hartshorn, G.S. 1980. Neotropical rain forest dynamics. Biotropica 12 (Supplement): 23-30. 298 References Hawthorne, W.D. 1993. Forest regeneration after logging: findings of a study in the Bia South Game Production Reserve, Ghana. ODA Forestry Series 3. Natural Resources Institute, Chatham Maritime, UK. Heinrich, R. 1995. Environmentally sound harvesting to sustain tropical forests. Commonwealth Forestry Review 74(3): 198-203. Hendrison, J. 1990. Damage-controlled logging in managed tropical rain forest in Suriname. Ecology and management of tropical rain forests in Suriname 4. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Hohenkerk. 1923. A review of the timber industry of British Guiana. The Journal of the Board of Agriculture of British Guiana 16(1): 2-22. Holmes, T.P., G.M. Blate, J.C. Zweede, R. Pereira Jr & F. Boltz. 1999. Financial costs and benefits of reduced-impact logging relative to conventional logging in the eastern Amazon. Forestry Private Enterprise Initiative Working Paper. Tropical Forest Foundation / Fundação Floresta Tropical / USDA Forest Service. Southeastern Center for Forest Economics Research, Research Triangle Park, NC, USA. Houtzagers, G. 1956. Houtteelt der gematigde luchtstreek, volume 2 [in Dutch: Silviculture in the temperate zone]. Tjeenk Willink, Zwolle, the Netherlands. Hubbell, S.P. & R.B. Foster. 1983. Diversity of canopy trees in a neotropical forest and implications for conservation. in: Sutton, S.L., T.C. Whitmore & A.C. Chadwick (eds.). Tropical rain forest: Ecology and management (British Ecological Society, Special Publication 2): 25-41. Blackwell Scientific Publications, London, UK. Hubbell, S.P. & R.B. Foster. 1986a. Canopy gaps and the dynamics of a neotropical forest. in: Crawley, M.J. (ed.). Plant ecology: 77-98. Blackwell Scientific Publications, Oxford, UK. Hubbell, S.P. & R.B. Foster. 1986b. Biology, chance, and history and the structure of tropical rain forest tree communities. in: Diamond, J. & T.J. Case (eds.). Community ecology: 314-329. Harper & Row Publishers, New York, USA. Husch, B., C.I. Miller & T.W. Beers. 1982. Forest mensuration. 3rd edition. John Wiley & Sons, New York, USA. Hutchinson, I.D. 1987. Improvement thinning in natural tropical forests: aspects of institutionalization. in: Mergen, F. & J.R. Vincent (eds.). Natural management of tropical moist forests: 113-131. Yale University, School of Forestry and Environmental Studies, New Haven, CT, USA. Hutchinson, I.D. 1988. Points of departure for silviculture in humid tropical forests. Commonwealth Forestry Review 67(3): 223-230. ITTO. 1990. Guidelines for the sustainable management of natural tropical forests. ITTO Technical Series 5. International Tropical Timber Organization, Yokohama, Japan. Jacobs, M. 1976. The study of lianas. Flora Malesiana Bulletin 29: 2610-2618. Jacobs, M. 1988. The tropical rain forest; a first encounter. Springer-Verlag, Heidelberg, Germany. Jans, L., L. Poorter, R.S.A.R. van Rompaey & F. Bongers. 1993. Gaps and forest zones in tropical moist forest in Ivory Coast. Biotropica 25(3): 258-269. Jetten, V.G. 1994. Modelling the effects of logging on the water balance of a tropical rain forest. A study in Guyana. PhD thesis, Utrecht University, Tropenbos Series 6. The Tropenbos Foundation, Wageningen, the Netherlands. John, C.A. 1961. The rate of growth of Virola surinamensis Warb. Unpublished Departmental Report. Forestry Department, Georgetown, Guyana. Johns, A.D. 1988. Effects of ‘selective’ timber extraction on rain forest structure and composition and some consequences for frugivores and folivores. Biotropica 20(1): 31-37. Johns, A.D. 1992. Species conservation in managed tropical forests. in: Sayer, J.A. & T.C. Whitmore (eds.). Realistic strategies for tropical forest conservation: 15-53. International Union for the Conservation of Nature, Gland, Switzerland. Johns, J.S., P. Barreto & C. Uhl. 1996. Logging damage during planned and unplanned logging operations in the eastern Amazon. Forest Ecology and Management 89(1): 59-78. Jonkers, W.B.J. 1987. Vegetation structure, logging damage and silviculture in a tropical rain forest in Suriname. Ecology and management of tropical rain forests in Suriname 3. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. 299 Reduced impact logging in the tropical rain forest of Guyana Jonkers, W.B.J. & P. Schmidt. 1984. Ecology and timber production in tropical rain forest in Suriname. Intersciencia 9: 290-298. Jonkers, W.B.J. & J. Hendrison. 1987. Prospects for sustained yield management of tropical rain forest in Suriname. in: Figueroa Colon, J.C. (ed.) Management of the forests of Tropical America. USDA Forest Service, San Juan, Puerto Rico. Jonsson, T. & P. Lindgren. 1990. Logging technology for tropical forests - for or against? Report for ITTO. Forskningsstiftelsen ‘Skogsarbeten’ (Forest Operations Institute), Stockholm, Sweden. Kahn, F. 1982. La reconstitution de la forêt tropicale humide. Mémoires ORSTOM 97. Office de la Recherche Scientifique et Technique Outre-Mer, Paris, France. Keizer, P.-J. 1993. The ecology of macromycetes in roadside verges planted with trees. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Kennedy, D.N. & M.D. Swaine. 1992. Germination and growth of colonizing species in artificial gaps of different sizes in dipterocarp rain forest. in: Marshall, A.G. & M.D. Swaine (eds.). Tropical rain forest: disturbance and recovery (Philosophical transactions of the Royal Society of London, Series B 335): 357-366. The Royal Society, London, UK. Kerr, B. 1993. Iwokrama: the Commonwealth Rain Forest Programme in Guyana. Commonwealth Forestry Review 72: 303-309. Kio, P.R.O., S.A. Ola-Adams, B.A. Lapido & F.O.C. Nwonwu. 1986. Natural regeneration technology in use in African High Forests. in: Proc. of the XVIII IUFRO World Congress, Ljubljana, Yugoslavia, Division 1, Vol. 2: 424-437. International Union of Forestry Research Organisations, Vienna, Austria. Klassen, A.W. 1987. Logging development plan Mabura Concession - 1987. Forest Department, Demerara Woods Limited, Mabura Hill, Guyana. Klasson, B. & J. Cedergren. 1996. Felling the right way; some hints on the art and science of directional felling. ITTO Tropical Forest Update 6(3): 5-9. Kleine, M. & J. Heuveldop. 1993. A management planning concept for sustained yield of tropical forests in Sabah, Malaysia. Forest Ecology and Management 61: 277-298. Koop, H. 1989. Forest dynamics; SILVI-STAR: a comprehensive monitoring system. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Springer-Verlag, Heidelberg, Germany. Koop, H. & R.J. Bijlsma. 1993. The SILVI-STAR link to a geographical information system; a tool for spatial analysis in digitally recorded forest reserves. in: Broekmeyer, M.E.A., W. Vos & H. Koop (eds.). European forest reserves: 145-152. Centre for Agricultural Publishing and Documentation, Wageningen, the Netherlands. Koop, H. & F.J. Sterck. 1994. Light penetration through structurally complex forest canopies: an example of a lowland tropical rainforest. Forest Ecology and Management 69: 111-122. Kramer, F. 1933. De natuurlijke verjonging in het Goenoeng Gedeh complex [in Dutch: The natural regeneration in the Goenoeng Gedeh complex (Java, Indonesia)]. Tectona 26: 156-185. Kuper, J.H. 1994. Sustainable development of Scots Pine forests. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Lamb, F.B. 1966. Mahogany of Tropical America; its ecology and management. University of Michigan, Ann Arbor, MI, USA. Lande, R., S. Engen & B.E. Saether. 1994. Optimal harvesting, economic discounting and extinction risk in fluctuating populations. Nature 372: 88-90. Landell-Mills, N. 1997. Stumpage value appraisal for log extraction in Guyana; report based on eight case studies. Internal report. Guyana Forestry Commission, Georgetown, Guyana. Lanly, J.P. 1982. Tropical forest resources. FAO Forestry Paper 30. Food and Agriculture Organization of the United Nations, Rome, Italy. Lee, H.S. 1982a. The development of silvicultural systems in the hill forests of Malaysia. The Malaysian Forester 45(1): 1-9. Lee, H.S. 1982b. Silvicultural management in Sarawak. The Malaysian Forester 45(4): 485-496. Leibundgut, H. 1958. Empfehlungen für die Baumklassenbildung und Methodik bei Versuchen über die Wirkung von Waldpflegemaßnahmen. in: Proc. XII IUFRO World Congress, Vol. 2 Section 23: 92-118. Oxford, UK. International Union of Forestry Research Organisations, Vienna, Austria. Leibundgut, H. 1970. Der Wald; eine Lebensgemeinschaft. 2nd edition. Huber, Frauenfeld, Switzerland. Leibundgut, H. 1984. Die Waldpflege. 3rd edition. Verlag P.Haupt, Bern, Switzerland. 300 References Leslie, A.J. 1977. Where contradictory theory and practice co-exist. Unasylva 29(115): 2-40. Leslie, A.J. 1987a. A second look at the economics of natural management systems in tropical mixed forests. Unasylva 39(155): 46-58. Leslie, A.J. 1987b. The economic feasibility of natural management of tropical forests. in: Mergen, F. & J.R. Vincent (eds.). Natural management of tropical moist forests: 177-198. Yale University, School of Forestry and Environmental Studies, New Haven, CT, USA. Lieberman, D., M. Lieberman, G. Hartshorn & R. Peralta. 1985. Growth rates and age-size relationships of tropical wet forest trees in Costa Rica. Journal of Tropical Ecology 1: 97-109. Lieberman, D., M. Lieberman & R. Peralta. 1989. Forests are not just Swiss Cheese: canopy stereogeometry of non-gaps in tropical forests. Ecology 70(3) - Special Feature: 550-552. Lieberman, M., D. Lieberman, R. Peralta & G.S. Hartshorn. 1995. Canopy closure and the distribution of tropical forest tree species at La Selva, Costa Rica. Journal of Tropical Ecology 11: 161-177. Liew, T.C. 1973. The practicability of climber cutting and tree marking prior to logging as a silvicultural tool in the management of dipterocarp forests of Sabah. The Malaysian Forester 36(2): 5-19. Liew, T.C. & F.O. Wong. 1973. Density, recruitment, mortality and growth of dipterocarp seedlings in virgin and logged-over forest in Sabah. The Malaysian Forester 36(1): 3-15. Lindeman, J.C. & A.M.W. Mennega. 1963. Bomenboek voor Suriname; herkenning van Surinaamse houtsoorten aan hout en vegetatieve kenmerken [in Dutch: Treebook for Suriname; recognition of Surinamese timber species through wood anatomy and vegetative characteristics]. Dienst 's Lands Bosbeheer, Paramaribo, Suriname. Londo, G. 1984. The decimal scale for relevés of permanent quadrats. in: Knapp, R. (ed.). Sampling methods and taxon analyses in vegetation science: 346-370. Junk, The Hague, the Netherlands. Lowe, R.G. 1978. Experience with the Tropical Shelterwood System of regeneration in natural forest in Nigeria. Forest Ecology and Management 1: 193-212. Luning, H.A. 1987. The need for tropical rain forests and their products. in: Van Beusekom, C.F., C.P. van Goor & P. Schmidt (eds.). Wise utilization of tropical forest lands (Tropenbos Scientific Series 1): 15-29. The Tropenbos Programme, Wageningen [Ede], the Netherlands. Magnusson, W.E., O.P. de Lima, F. Quintiliano Reis, N. Higuchi & J. Ferreira Ramos. 1999. Logging activity and tree regeneration in an Amazonian forest. Forest Ecology and Management 113: 67-74. Maini, J.S. 1992. Sustainable development of forests. Unasylva 43: 3-7. Maître, H.-F. 1986. Dynamique et production des peuplements naturels de forêt dense humide en Afrique. Revue Bois et Forêts des Tropiques 213(3): 3-12. Mallet, B. & N. Tuo. 1989. Évolution recente des techniques de dévitalisation des essences secondaires de la forêt naturelle en Côte d'Ivoire. in: Séminaire IUFRO sur herbicides et foresterie, Nancy, France. International Union of Forestry Research Organisations, Vienna, Austria. Malmer, A. & H. Grip. 1990. Soil disturbance and loss of infiltrability caused by mechanized and manual extraction of tropical rain forest in Sabah, Malaysia. Forest Ecology and Management 38: 1-12. Marshall, A.G. 1992. The Royal Society's South-east Asian rain forest research programme: an introduction. in: Marshall, A.G. & M.D. Swaine (eds.). Tropical rain forest: disturbance and recovery (Philosophical transactions of the Royal Society of London, Series B 335): 327-330. The Royal Society, London, UK. Martinez-Ramos, M., E.Alvarez-Buylla & J.Sarukhan. 1989. Tree demography and gap dynamics in a tropical rain forest. Ecology 70(3) - Special Feature: 555-558. Maser, C. 1992. Global imperative; harmonizing culture and nature. Stillpoint Publishing, Walpole, NH, USA. Mattsson Mårn, H. & W.B.J. Jonkers. 1981. Logging damage in tropical high forest. FO:MAL/76/008. Working Paper 5. UNDP/FAO Forestry Development Project Sarawak, Kuching, Sarawak. Mayer, H. 1980. Waldbau auf soziologisch-ökologischer Grundlage. 2nd edition. Fischer, Stuttgart, Germany. McTurk, M., W. Russell & G.M. Peake. 1882. Notes on the forests of British Guiana. Timehri (Guyana) 1: 173-215. Meijer, W. 1970. Regeneration of tropical lowland forest in Sabah, Malaysia, forty years after logging. The Malayan Forester 33(3): 204-229. Mekkink, M.J. & R.G. Nijmeijer. 1998. Microclimatological and hydrological ecosystem components within man-made gaps in tropical rain forest of Guyana. M.Sc.thesis, Department of Physical Geography, Utrecht University, the Netherlands. 301 Reduced impact logging in the tropical rain forest of Guyana Mennega, E.A., W.C.M. Tammens-de Rooij & M.J. Jansen-Jacobs (eds.). 1988. Check-list of woody plants of Guyana [based on D.B. Fanshawe's (1949) ’Check-list of the indigenous woody plants of British Guiana’]. Tropenbos Technical Series 2. The Tropenbos Foundation, Wageningen [Ede], the Netherlands. Mills, L.S., M.E. Soulé & D.F. Doak. 1993. The keystone species concept in ecological conservation. Bioscience 43: 219-224. Mitchell, P.L. & T.C. Whitmore. 1993. Use of hemispherical photographs in forest ecology: calculation of absolute amount of radiation beneath the canopy. Occasional Paper 44, Oxford Forestry Institute, Oxford, UK. Miyata, E.S. 1980. Determining fixed and operating costs of logging equipment. General Technical Report North-Central 55. USDA Forest Service, North Central Forest Experiment Station, St. Paul, Minnesota, USA. Moir, W.H. & H.T. Mowrer. 1995. Unsustainability. Forest Ecology and Management 73(1): 239-248. Montenegro, S.F. 1996. Low impact harvesting in La Mayronga: comparative analysis of its techniques, environmental impact, natural regeneration and costs versus those under traditional extraction. International Timber Trade Organization; Fundación Forestal Juan Manuel Durini, Quito, Ecuador. Moore, J.A., C.A. Budelsky & R.C. Schlesinger. 1973. A new index presenting individual tree competitive status. Canadian Journal of Forest Research 3: 495-500. Munyanziza, E. 1994. Miombo trees and mycorrhizae: ecological strategies, a basis for afforestation. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Mutoji-A-Kazadi. 1977. Notes de sylvigénèse pour la Guyane; transect et photographies aériennes. Mémoire technique DEA, Laboratoire de Écologie terrestre, Botanique tropicale, Université des Sciences et Techniques du Languedoc, Montpellier, France. Myers, N. 1984. The primary source: tropical forests and our future earth. W.W. Norton & Company, New York, USA. Neil, P.E. 1981. Problems and opportunities in tropical rain-forest management. Occasional Paper 16, Commonwealth Forestry Institute, Oxford, UK. Nguyen-Thé, N., V. Favrichon, P. Sist, L. Houdé & N. Fauvet. 1999. Dynamique de la forêt mixte à diptérocarpacées de basse altitude avant et après traitement sylvicole. Revue Bois et Forêts des Tropiques 259(1): 25-43. Nicholson, D.I. 1958. An analysis of logging damage in tropical rain forest, North Borneo. The Malayan Forester 21: 235-245. Nicholson, D.I. 1965. A review of natural regeneration in the dipterocarp forest of Sabah. The Malayan Forester 28: 4-24. Nicholson, D.I. 1979. The effects of logging and treatment on the mixed dipterocarp forests of South East Asia. FO: MISC/79/8. Food and Agriculture Organization of the United Nations, Rome, Italy. Nussbaum, R., J. Anderson & T. Spencer. 1995. Factors limiting the growth of indigenous tree seedlings planted on degraded rainforest soils in Sabah, Malaysia. Forest Ecology and Management 74: 149-160. Oberbauer, S.F., D.B. Clark, D.A. Clark & M. Quesada. 1988. Crown light environments of saplings of two species of rain forest emergent trees. Oecologia 73: 207-212. Ogawa, H., K. Yoda, K. Kira, K. Ogino, T. Shidei, D. Ratanawongse & C. Apasutaya. 1965. Comparative ecological study of three main types of forest vegetation in Thailand. I. Structure and floristic composition. Nature and Life in Southeast Asia 4: 13-48. Oldeman, R.A.A. 1972. L'architecture de la végétation ripicole forestière des fleuves et criques guyanais. Adansonia (N.S.) 12(2): 253-265. Oldeman, R.A.A. 1974a. L'architecture de la forêt guyanaise. Mémoires ORSTOM 73. Office de la Recherche Scientifique et Technique Outre-Mer, Paris, France. Oldeman, R.A.A. 1974b. Écotopes des arbres et gradients écologiques verticaux en forêt guyanaise. Revue d'Écologie (La Terre et La Vie) 28(4): 487-520. Oldeman, R.A.A. 1978. Architecture and energy exchange of dicotyledonous trees in the forest. in: Tomlinson, P.B. & M.H. Zimmermann (eds.). Tropical trees as living systems: 535-560. Cambridge University Press, Cambridge, UK. 302 References Oldeman, R.A.A. 1983. Tropical rain forest, architecture, silvigenesis and diversity. in: Sutton, S.L., T.C. Whitmore & A.C. Chadwick (eds.). Tropical rain forest: Ecology and management (British Ecological Society, Special Publication 2): 139-150. Blackwell Scientific Publications, London, UK. Oldeman, R.A.A. 1989. Dynamics in tropical rain forests. in: Holm-Nielsen, L.B., I.C. Nielsen & H. Balslev (eds.). Tropical forests: botanical dynamics, speciation and diversity: 3-21. Botanical Institute, Aarhus University. Academic Press, London, UK. Oldeman, R.A.A. 1990. Forests: elements of silvology. Springer-Verlag, Berlin-Heidelberg, Germany. Oldeman, R.A.A. 1991. The paradox of forest management. in: Proc. of X World Forestry Congress, Paris, France. (Revue Forestière Française, hors séries 4): 153-163. Oldeman, R.A.A. 1992. Architectural models, fractals and agroforestry design. Agriculture, Ecosystems and Environment 41: 179-188. Oldeman, R.A.A. 1994. Sur les écosystèmes forestiers: quatre principes sylvologiques. L'Italia Forestale e Montana 49(1): 1-16. Oldeman, R.A.A. & P.J. Van der Meer. 1988. Diagnosis of Northern Queensland rain forests: the impact of selection silviculture. Contract research report. Department of Forestry, Wageningen Agricultural University, Wageningen, the Netherlands. Oldeman, R.A.A. & J. Van Dijk. 1991. Diagnosis of the temperament of tropical rain forest trees. in: Gómez-Pompa, A., T.C. Whitmore & M. Hadley (eds.). Rain forest regeneration and management (Man and the Biosphere Series 6): 21-65. UNESCO, Paris, France / Parthenon, Carnforth, UK. Orians, G.H. 1982. The influence of tree-falls in tropical forests in tree species richness. Tropical Ecology 23(2): 255-279. Osunkoya, O.O., J.E. Ash, A.W. Graham & M.S. Hopkins. 1993. Growth of tree seedlings in tropical rain forests of North Queensland, Australia. Journal of Tropical Ecology 9: 1-18. Pacala, S.W., C.D. Canham & J.A. Silander Jr. 1993. Forest models defined by field measurements; I: the design of a northeastern forest simulator. Canadian Journal of Forest Research 23: 1980-1988. Palmer, J.R. & M. Marshall. 1996. Forest revenue systems and allocation of harvesting permits in Guyana. Guyana Forestry Commission support project, Georgetown, Guyana. Parren, M. & F. Bongers. 1999. Forest lianas and pre-felling climber cutting in southern Cameroon: a silvicultural evaluation. in: Proc. of CIFOR/FORAFRI conference, Libreville, Gabon. Paul, D.K. 1977. The potential for the establishment of long fibre pulp plantations in Guyana. UNDP/FAO, Georgetown, Guyana. Peters, P.A. 1978. Spacing of roads and landings to minimize timber harvest cost. Forest Science 24(2): 209-217. Peters, R. 1992. Ecology of beech forests in the northern hemisphere. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Pickett, S.T.A. 1983. Differential adaptation of tropical tree species to canopy gaps and its role in community dynamics. Tropical Ecology 24(1): 68-84. Pickett, S.T.A. & P.S. White (eds.). 1985. The ecology of natural disturbance and patch dynamics. Academic Press, New York, USA. Pinard, M.A. & F.E. Putz. 1996. Retaining forest biomass by reducing logging damage. Biotropica 28(3): 278-295. Pinard, M.A., F.E. Putz, J. Tay & T.E. Sullivan. 1995. Creating timber harvest guidelines for a reduced-impact logging project in Malaysia. Journal of Forestry 93(10): 41-45. Pinard, M.A., B. Howlett & D. Davidson. 1996. Site conditions limit pioneer tree recruitment after logging of Dipterocarp forests in Sabah, Malaysia. Biotropica 28(1): 2-12. Pinard, M.A., F.E. Putz, D. Rumíz, R. Guzmán & A. Jardim. 1999. Ecological characterization of tree species for guiding forest management decisions in seasonally dry forests in Lomerío, Bolivia. Forest Ecology and Management 113: 201-213. Plumptre, A.J. 1996. Changes following 60 years of selective timber harvesting in the Budongo Forest Reserve, Uganda. Forest Ecology and Management 89(1): 101-114. Poore, D., P. Burgess, J. Palmer, S. Rietbergen & T. Synnott. 1989. No timber without trees; sustainability in the tropical forest. Earthscan Publications Ltd., London, UK. Poorter, L. 1998. Seedling growth of Bolivian rain forest tree species in relation to light and water availability. PhD thesis, Utrecht University, Utrecht, the Netherlands. 303 Reduced impact logging in the tropical rain forest of Guyana Poorter, L., F. Bongers, R.S.A.R. Van Rompaey & M. De Klerk. 1996. Regeneration of canopy tree species at five sites in West African moist forest. Forest Ecology and Management 84(1): 61-70. Popma, J. & F. Bongers. 1988. The effect of canopy gaps on growth and morphology of seedlings of rain forest species. Oecologia 75: 625-632. Popma, J., F. Bongers, M. Martínez-Ramos & E. Veneklaas. 1988. Pioneer species distribution in treefall gaps in Neotropical rain forest; a gap definition and its consequences. Journal of Tropical Ecology 4: 77-88. Prince, A.J. 1971a. The rate of growth of Greenheart (Ocotea rodiei Schomb.). Unpublished Departmental Report. Forestry Department, Georgetown, Guyana. 60 p. Prince, A.J. 1971b. Moraballi Forest Reserve; report on sample plots in improved natural forest. Unpublished Departmental Report. Forest Department, Georgetown, Guyana. 14 p. Prince, A.J. 1973. The rate of growth of Greenheart (Ocotea rodiei Schomb). Commonwealth Forestry Review 52: 143-146. Putz, F.E. 1983. Treefall pits and mounts, buried seeds, and the importance of soil disturbance to pioneer trees on Barro Colorado Island, Panama. Ecology 64(5): 1069-1074. Putz, F.E. 1984. The natural history of lianas on Barro Colorado Island, Panama. Ecology 65(6): 1713-1724. Putz, F.E. & N.V.L. Brokaw. 1989. Sprouting of broken trees on Barro Colorado Island, Panama. Ecology 70(2): 508-512. Putz, F.E. & V. Viana. 1996. Biological challenges for certification of tropical timber. Biotropica 28(3): 323-330. Putz, F.E., H.S. Lee & R. Goh. 1984. Effects of post-felling silvicultural treatments on woody vines in Sarawak. The Malaysian Forester 47(3): 214-226. Quicke, E., K.F.S. King, D. Lens & A. De Coene. 1972. Report of the Guyana Forest Industries Development Project; identification mission. Report No 31/71 GUY.2. IBRD/FAO, Rome, Italy. Quicke, B.E. 1996. Demerara Timbers Limited; forest management plan 1997-2001. Demerara Timbers Ltd., Mabura Hill, Guyana. Raich, J.W. 1989. Seasonal and spatial variation in the light environment in a tropical dipterocarp forest and gaps. Biotropica 21: 299-302. Raich, J.W. & W.K. Gong. 1990. Effects of canopy opening on tree seed germination in a Malaysian dipterocarp forest. Journal of Tropical Ecology 6: 203-217. Repetto, R. 1988. The forest for the trees? Government policies and the misuse of forest resources. World Resources Institute, Washington D.C., USA. Rich, P.M., D.B. Clark, D.A. Clark & S.F. Oberbauer. 1993. Long-term study of solar radiation regimes in a tropical wet forest using quantum sensors and hemispherical photography. Agricultural and Forest Meteorology 65: 211-230. Richards, P.W. 1952, 1996. The tropical rain forest; an ecological study. 1st and 2nd edition. Cambridge University Press, Cambridge, UK. Riéra, B. 1982. Observations sur les chablis, Piste St. Elie, Guyane. Bulletin de liaison du groupe de travail sur l'écosystème forestier guyanais. Office de la Recherche Scientifique et Technique Outre-Mer, Cayenne, Guyane française. Riéra, B. 1985. Importance des buttes de déracinement dans la régénération forestière en Guyane française. Revue d'Écologie (La Terre et La Vie) 40: 321-329. Rijks, M.H., E.-J. Malta & R.J. Zagt. 1998. Regeneration through sprout formation in Chlorocardium rodiei (Lauraceae) in Guyana. Journal of Tropical Ecology 14: 463-475. Rollet, B. 1983. La régénération naturelle dans les trouées; un processus général de la dynamique des forêts tropicales humides. Revue Bois et Forêts des Tropiques 201: 3-35, 202: 19-34. Rose, S.A. 1997. Long-term seedling demography after gap creation in the tropical rain forest of Guyana. Tropenbos-Guyana Reports 97-1. Tropenbos-Guyana Programme, Georgetown, Guyana. Rossignol, M., L. Rossignol, R.A.A. Oldeman & S. Benzine-Tizroutine. 1998. Struggle of life or the natural history of stress and adaptation. Treebook 1. Treemail, Heelsum, the Netherlands. Runkle, J.R. 1981. Gap regeneration in some old-growth forest of the eastern United States. Ecology 62(4): 1041-1051. 304 References Runkle, J.R. 1982. Patterns and disturbance in some old-growth mesic forests of eastern North America. Ecology 63(5): 1533-1546. Runkle, J.R. 1992. Guidelines and sample protocol for sampling forest gaps. General Technical Report PNW-GTR 283. US Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon, USA. Saldarriaga, J.G. & D.C. West. 1986. Holocene fires in the northern Amazon basin. Quaternary Research 26: 358-366. Samset, I. 1972. Cost calculation of forest machines [in Norwegian with English summary]. Tidsskrift for Skagbruk 80(4): 357-374. Sanford Jr, R.B., J.G. Saldarriaga, K. Clark, C. Uhl & R. Harrier. 1985. Amazon rain-forest fires. Science 227: 53-55. Sarre, A. 1996. What foresters can do. ITTO Tropical Forest Update 6(3): 1. Schmidt, R.C. 1987. Tropical rain forest management; a status report. Unasylva 39(156): 2-17. Schmitt, L. 1989. Étude des peuplements naturels en forêt dense guyanaise; compte rendu de mise en application des traitements sylvicoles sur le dispositif de Paracou. Internal Report. Centre Techique Forestier Tropical, Kourou, Guyane française. Schmitt, L. & M. Bariteau. 1990. Gestion de l'écosystème forestier guyanais. Étude de la croissance et de la régénération naturelle au Dispositif de Paracou. Revue Bois et Forêts Tropiques 220: 3-23. Schulz, J.P. 1960. Ecological studies on rain forest in Northern Suriname. Mededelingen van het Botanisch Museum en Herbarium van de Rijksuniversiteit te Utrecht 163. Noord-Hollandsche Uitgevers Maatschappij, Amsterdam, the Netherlands. Schupp, E.W., H.F. Howe, C.K. Augspurger & D.J. Levey. 1989. Arrival and survival in tropical treefall gaps. Ecology 70(3) - Special Feature: 562-564. Schütz, J.-P. & R.A.A. Oldeman. 1996. Gestion durable par automation biologique des forêts. Revue Forestière Française XLVIII (special issue): 65-74. Sheil, D. & R.M. May. 1996. Mortality and recruitment rate evaluations in heterogeneous tropical forests. Journal of Ecology 84: 91-100. Shigo, A.L. 1989. A new tree biology. 2nd edition. Shigo and Trees, Associates, Durham, NH, USA. Shugart, H.H. 1984. A theory of forest dynamics; the ecological implications of forest succession models. Springer-Verlag, Berlin-Heidelberg, Germany. Silva, J.N.M. 1989. The behaviour of the tropical rain forest of the Brazilian Amazon after logging. Ph.D. thesis, University of Oxford, UK. Silva, J.N.M., J.O.P. de Carvalho, J. do C.A.Lopes, B.F. de Almeida, D.M.H. Costa, L.C. de Oliveira, J.K. Vanclay & J.P. Skovsgaard. 1995. Growth and yield of a tropical rain forest in the Brazilian Amazon 13 years after logging. Forest Ecology and Management 71: 267-274. Silversides, C.R. & U. Sundberg. 1989. Operational efficiency in forestry. Volume 2: Practice. Kluwer Academic Publishers, Dordrecht, the Netherlands. Sist, P., T. Nolan, J.-G. Bertault & D. Dykstra. 1998. Harvesting intensity versus sustainability in Indonesia. Forest Ecology and Management 108: 251-260. Sizer, N. 1996. Profit without plunder: reaping revenue from Guyana's tropical forests without destroying them. World Resources Institute, Washington D.C., USA. Sizer, N. & R. Rice. 1995. Backs to the wall in Suriname: forest policy in a country in crisis. World Resources Institute, Washington D.C., USA. Smith, D.M., B.C. Larson, M.J. Kelty & P.M.S. Ashton. 1996. The practice of silviculture; applied forest ecology. 9th edition. John Wiley & Sons, New York, USA. Smits, W.T.M. 1994. Dipterocarpaceae: mycorrhizae and regeneration. PhD thesis, Wageningen Agricultural University, Tropenbos Series 9. The Tropenbos Foundation, Wageningen, the Netherlands. Sokal, R.R. & F.J. Rohlf. 1995. Biometry: the principles and practice of statistics in biological research. 3rd edition. W.H. Freeman and Co., New York, USA. Steneker, G.A. & J.M. Jarvis. 1963. A preliminary study to assess competition in a white spruce - trembling aspen stand. Forestry Chronicle 39: 334-336. Sundberg, U. & C.R. Silversides (eds.). 1988. Operational efficiency in forestry. Volume 1: Analysis. Kluwer Academic Publishers, Dordrecht, the Netherlands. 305 Reduced impact logging in the tropical rain forest of Guyana Swabey, C. 1950. Note on the development of forest policy in British Guiana. Timehri (Guyana) 29: 27-35. Swaine, M.D. & T.C. Whitmore. 1988. On the definition of ecological species groups in tropical rain forests. Vegetatio 88: 31-51. Synnott, T.J. & R.H. Kemp. 1976. Choosing the best silvicultural system. Unasylva 28(1): 74-79. Ter Steege, H. 1990. A monograph of Wallaba, Mora and Greenheart. Tropenbos Technical Series 5. The Tropenbos Foundation, Wageningen [Ede], the Netherlands. Ter Steege, H. 1997. Winphot 5.0; a programme to analyze vegetation indices, light and light quality from hemispherical photographs. Tropenbos-Guyana Reports 97-3. The Tropenbos-Guyana Programme, Georgetown, Guyana. Ter Steege, H. 1998. The use of forest inventory data for a National Protected Area Strategy in Guyana. Biodiversity and Conservation 7: 1457-1483. Ter Steege, H., V.G. Jetten, A.M. Polak & M.J.A. Werger. 1993. Tropical rain forest types and soils of a watershed in Guyana, South America. Journal of Vegetation Science 4: 705-716. Ter Steege, H., C. Bokdam, M. Boland, J. Dobbelsteen & I. Verburg. 1994. The effects of man-made gaps on germination, early survival and morphology of Chlorocardium rodiei in Guyana. Journal of Tropical Ecology 10: 245-260. Ter Steege, H., R.G.A. Boot, L.C. Brouwer, D.S. Hammond, P. van der Hout, V.G. Jetten, Z. Khan, A.M. Polak, D. Raaimakers & R.J. Zagt. 1995. Basic and applied research for sound rain forest management in Guyana. Ecological Applications 5(4): 904-910. Ter Steege, H., R.G.A. Boot, L.C. Brouwer, J.C. Caesar, R.C. Ek, D.S. Hammond, P. Haripersaud, P. van der Hout, V.G. Jetten, A. van Kekem, M. Kellman, Z. Khan, A.M. Polak, T.L. Pons, J. Pulles, D. Raaimakers, S.A. Rose, J.J. van der Sanden & R.J. Zagt. 1996. Ecology and logging in a tropical rain forest in Guyana; with recommendations for forest management. Tropenbos Series 14. The Tropenbos Foundation, Wageningen, the Netherlands. Terborgh, J. 1986. Keystone plant resources in the tropical forest. in: Soulé, M.E. (ed.). Conservation biology. The science of scarcity and diversity: 330-344. Sinauer, Sunderland, MA, USA. Torquebiau, E.F. 1981. Analyse architecturale de la forêt de Los Tuxtlas (Veracruz), Mexique. Thèse 3me Cycle, Université de Montpellier, Montpellier, France. Torquebiau, E.F. 1986a. Tropical rain forest regeneration patterns as recognized by laying down forest mosaic maps. in: Proc. of BIOTROP conference ‘Forest regeneration in South East Asia’: 139-150, Bogor, Indonesia. Torquebiau, E.F. 1986b. Mosaic patterns in dipterocarp rain forest in Indonesia, and their implications for practical forestry. Journal of Tropical Ecology 2: 301-325. Tropenbos. 1998. http://www.tropenbos.nl/tropenbos/tropenbos-home.html. The Tropenbos Foundation, Wageningen, the Netherlands. Uhl, C. & P.G. Murphy. 1981. Composition, structure and regeneration of a tierra firme forest in the Amazon basin of Venezuela. Tropical Ecology 22(2): 219-237. Uhl, C. & I.C.G. Vieira. 1989. Ecological impacts of selective logging in the Brazilian Amazon: a case study from the Paragominas Region of the State of Pará. Biotropica 21(2): 98-106. Uhl, C., K. Clark, N. Dezzeo & P. Maquirino. 1988. Vegetation dynamics in Amazonian treefall gaps. Ecology 69(3): 751-763. Uhl, C., A. Veríssimo, M. Mattos, Z. Brandino & I.C.G. Vieira. 1991. Social, economic and ecological consequences of selective logging in an Amazon frontier: the case of Tailândia. Forest Ecology and Management 46: 243-273. Van Baren, B. & P. Hilgen. 1984. Structuur en dynamiek in La Tillaie, een ongestoord beukenbos in het bosgebied van Fontainebleau [in Dutch: Structure and dynamics in La Tillaie, an undisturbed beech forest in the forest area of Fontainebleau, with French summary]. M.Sc.thesis, Departments of Nature Management / Silviculture, Wageningen Agricultural University, Wageningen, the Netherlands. Van Beusekom, C.F., C.P. van Goor & P. Schmidt (eds.). 1987. Wise utilization of tropical forest lands. Tropenbos Scientific Series 1. The Tropenbos Programme, Wageningen [Ede], the Netherlands. Van Dam, O. & S.A. Rose. 1997. Natural tree fall gaps in the Pibiri Gap Study area, Guyana. Internal Report. The Tropenbos-Guyana Programme, Georgetown, Guyana. 306 References Van Dam, O., S.A. Rose, N.C. Houter, D.S. Hammond, T.L. Pons & H. ter Steege. 1999. The Pibiri Gap Experiment. A study of the effects of gap size on microclimate, edaphic conditions, seedling survival and growth, ecophysiology and insect herbivory. Site description, methodologies and experimental set-up. Tropenbos-Guyana Reports 99-1. The Tropenbos-Guyana Programme, Georgetown, Guyana. Van der Hout, P. 1996. Effects of logging with different intensities of low impact harvesting. TropenbosGuyana Reports 96-1. The Tropenbos-Guyana Programme, Georgetown, Guyana. Van der Hout, P. & G.J.R. van Leersum. 1998. Reduced impact logging: a global panacea? in: Tropenbos. Research in tropical rain forests: its challenges for the future: 185-202. The Tropenbos Foundation, Wageningen, the Netherlands. Van der Meer, P.J. 1995. Canopy dynamics of a tropical rain forest in French Guiana. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Van der Meer, P.J. 1997. Vegetation development in canopy gaps in a tropical rain forest in French Guiana. Selbyana 18(1): 38-50. Van der Meer, P.J. & F. Bongers. 1996a. Pattern of tree-fall and branch-fall in a tropical rain forest in French-Guiana. Journal of Ecology 84: 19-29. Van der Meer, P.J. & F. Bongers. 1996b. Formation and closure of canopy gaps in the rain forest at Nouragues, French Guiana. Vegetatio 126: 167-179. Van der Meer, P.J., F. Bongers, L. Chatrou & B. Riéra. 1994. Defining canopy gaps in a tropical rain forest: effects on gap size and turnover time. Acta Oecologica 15(6): 701-714. Van der Meer, P.J., F.J. Sterck & F. Bongers. 1998. Tree seedling performance in canopy gaps in a tropical rain forest at Nouragues, French Guiana. Journal of Tropical Ecology 14: 119-137. Van Gardingen, P.R., M.J. Clearwater, T. Nifinluri, R. Effendi, W. Rusmantoro, M. Noor, P.A. Mason, K. Ingleby & R.C. Munro. 1998. Impacts of logging on the regeneration of lowland dipterocarp forest in Indonesia. Commonwealth Forestry Review 77(2): 71-82. Van Kekem, A., J.H.M. Pulles & Z. Khan. 1996. Soils of the rainforest in central Guyana. TropenbosGuyana Series 2. The Tropenbos-Guyana Programme, Georgetown, Guyana. Van Leersum, G.J.R. 1984. Winchsystems and shortwood logging. Work-study Report, Centre for Agricultural Research in Suriname, Paramaribo, Suriname / Wageningen Agricultural University, Wageningen, Netherlands. Van Leersum, G.J.R. in prep. Logging research for the design of a forest management plan in south Cameroon. in: Proc. of a Tropenbos workshop ‘Monitoring the progress of the Dutch contributions to the Tropenbos Cameroon Programme towards planning sustainable forest management’, 1 October 1998, Wageningen, the Netherlands. Van Loon, T. 1997. Regeneration in logging gaps in the moist forest of South Cameroon. M.Sc.thesis, Department of Forestry, Wageningen Agricultural University, Wageningen, the Netherlands. Van Rompaey, R.S.A.R. 1993. Forest gradients in West Africa; a spatial gradient analysis. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Van Steenis, C.G.G.J. 1958. Rejuvenation as a factor for judging the status of vegetation types: the biological nomad theory. in: Proc. of the Kandy symposium ‘Study of tropical vegetation’, Kandy, Sri Lanka: 212-218. UNESCO, Paris, France. Vanmechelen, E. 1994. Vegetatiestructuur en lichtklimaat van een Greenheartbos in Guyana [in Dutch: Vegetation structure and light climate of a Greenheart-forest in Guyana]. M.Sc.thesis, Afdeling Land- en Bosbeheer, Katholieke Universiteit Leuven, Louvain, Belgium. Vázquez-Yanes, C. & A. Orozco-Segovia. 1984. Patterns of seed longevity and germination in the tropical rainforest. Annual Review of Ecology and Systematics 24: 69-87. Veríssimo, A., P. Barreto, M. Mattos, R. Tarifa & C. Uhl. 1992. Logging impacts and prospects for sustainable forest management in an old Amazonian frontier: the case of Paragominas. Forest Ecology and Management 55: 169-199. Veríssimo, A., P. Barreto, R. Tarifa & C. Uhl. 1995. Extraction of a high-value natural resource in Amazonia: the case of mahogany. Forest Ecology and Management 72: 39-60. Vester, H.F.M. 1997. The trees and the forest. The role of tree architecture in canopy development; a case study in secondary forests (Araracuara, Colombia). PhD thesis, University of Amsterdam, the Netherlands. 307 Reduced impact logging in the tropical rain forest of Guyana Vieira, V.S. 1967. Pinus caribaea on white sand soils in Guyana. Commonwealth Forestry Review 46: 317-322. Vieira, V.S. 1980. Logging in Guyana and considerations for improvements. Guyana Forestry Commission, Georgetown, Guyana. Von Segebaden, G. 1964. Studies of cross-country transport distances and road net extensions. Studia Forestalia Suecia 18. Skoghögskolan, Stockholm, Sweden. Von Segebaden, G. 1969. Studies on the accessibility of forests and forest land in Sweden. Studia Forestalia Suecia 76. Skoghögskolan, Stockholm, Sweden. Vooren, A.P. 1987. Development versus conservation: avoiding a conflict in the Taï region (Ivory Coast). in: Van Beusekom, C.F., C.P. van Goor & P. Schmidt (eds.). Wise utilization of tropical forest lands (Tropenbos Scientific Series 1): 130-137. The Tropenbos Programme, Wageningen [Ede], the Netherlands. Vooren, A.P. 1992. Harvest criteria for tropical forest trees. in: Cleaver, K., M. Munasinghe, M. Dyson, N. Egli, F. Peuker & F. Wencélius (eds.). Conservation of West and Central African rainforests (World Bank Environment Paper 1): 134-140. World Bank, Washington D.C., USA. Wadsworth, F.H. 1958. Tropical rain forest. in: Proceedings, IV World Forestry Congress, 11-22 December 1954, Dehra Dun, India. Vol. 3: 119-129. Wadsworth, F.H. 1987. Applicability of Asian and African silvicultural systems to naturally regenerated forests of the Neotropics. in: Mergen, F. & J.R. Vincent (eds.). Natural management of tropical moist forests: 93-111. Yale University, School of Forestry and Environmental Studies, New Haven, CT, USA. Wadsworth, F.H. 1997. Forest production for Tropical America. Agriculture Handbook 710. United States Department of Agriculture (Forest Service), Washington D.C., USA. Wadsworth, F.H., B.R. Parresol & J.C. Figueroa Colon. 1989. Tree increment indicators in subtropical wet forest. in: Wan Razali Wan Mohd., H.T. Chan & S. Appanah (eds.) Proc. of IUFRO seminar on growth and yield in tropical mixed moist forests, 20-24 June, 1988, Kuala-Lumpur, Malaysia: 205-212. International Union of Forestry Research Organisations, Vienna, Austria / Forest Research Institute, Kuala-Lumpur, Malaysia. Walter, J.M.N. 1994. CIMES: a package of programmes for assessing canopy geometry and light transmission by hemispherical photographs. Centre National de la Recherche Scientifique / Université Louis Pasteur, Strasbourg, France. Walter, J.-M.N. & E.F. Torquebiau. 1997. The geometry of the canopy of a dipterocarp rain forest in Sumatra. Agricultural and Forest Meteorology 85: 99-115. Ward, J.P. & P.J. Kanowksi. 1985. Implementing control of harvesting operations in north Queensland rainforests. in: Shepherd, K.R. & H.V. Richter (eds.) Managing the tropical forest, Proceedings from a workshop held at Gympie, Queensland, Australia: 165-186. Australian National University, Canberra, Australia. Watt, A.S. 1947. Pattern and process in the plant community. Journal of Ecology 35: 1-22. Webb, E.L. 1997. Canopy removal and residual stand damage during controlled selective logging in lowland swamp forest of northeast Costa Rica. Forest Ecology and Management 95: 117-129. Webb, E.L. 1998. Gap-phase regeneration in selectively logged lowland swamp forest, northeastern Costa Rica. Journal of Tropical Ecology 14: 247-260. Weidelt, H.J. & V.S. Banaag. 1982. Aspects of management and silviculture of Philippine dipterocarp forests. Schriftenreihe der GTZ 132. Deutsche Gesellschaft für Technische Zusammenarbeit, Eschborn, Germany. Welch, I.A., L. Harry & W.T. Lord. 1975. A short history of the Guyana Forest Department 1925-1975. Forestry Department, Georgetown, Guyana. Welden, C.W., S.W. Hewett, S.P. Hubbell & R.B. Foster. 1991. Sapling survival, growth, and recruitment: relationship to canopy height in a neotropical forest. Ecology 72(1): 35-47. Werger, M.J.A. 1992. De aantasting van het regenwoud [in Dutch: The degradation of the tropical rain forest]. in: Hoogmoed, M.S. & R. De Jong (eds.). Tropisch regenwoud; schatkamer van biodiversiteit [in Dutch: Tropical rain forest; treasury of biodiversity]. Nationaal Natuurhistorisch Museum, Leyden, the Netherlands. White, L.J.T. 1994. The effects of commercial mechanised selective logging on a transect in lowland rainforest in the Lopé Reserve, Gabon. Journal of Tropical Ecology 10: 313-322. 308 References Whitman, A.A., N.V.L. Brokaw & J.M. Hagan. 1997. Forest damage caused by selection logging of mahogany (Swietenia macrophylla) in northern Belize. Forest Ecology and Management 92: 87-96. Whitmore, T.C. 1975, 1984. Tropical rain forests of the Far East. 1st and 2nd edition. Clarendon Press, Oxford, UK. Whitmore, T.C. 1978. Gaps in the forest canopy. in: Tomlinson, P.B. & M.H. Zimmermann (eds.). Tropical trees as living systems: 639-656. Cambridge University Press, Cambridge, UK. Whitmore, T.C. 1982. On pattern and process in forests. in: Newman, E.J. (ed.). The plant community as a working mechanism: 45-57. Blackwell, Oxford, UK. Whitmore, T.C. 1989. Canopy gaps and the two major groups of forest trees. Ecology 70(3) - Special Feature: 536-538. Whitmore, T.C. 1990. An introduction to tropical rain forests. Clarendon Press, Oxford, UK. Whitmore, T.C. 1991. Tropical rain forest dynamics and its implications for management. in: Gómez-Pompa, A., T.C. Whitmore & M. Hadley (eds.). Rain forest regeneration and management (Man and the Biosphere Series 6): 67-89. UNESCO, Paris, France / Parthenon, Carnforth, UK. Whitmore, T.C. 1996. A review of some aspects of tropical rain forest seedling ecology with suggestions for further enquiry. in: Swaine, M.D. (ed.). The ecology of tropical forest tree seedlings (Man and the Biosphere Series 17): 3-30. UNESCO, Paris, France / Parthenon, Carnforth, UK. Whitmore, T.C. & J.A. Sayer (eds.). 1992. Tropical deforestation and species extinction. International Union for the Conservation of Nature, Gland, Switzerland / Chapman and Hall, London, UK. Whitmore, T.C. & N.D. Brown. 1996. Dipterocarp seedling growth in rain forest canopy gaps during six and a half years. Philosophical Transactions of the Royal Society London B 351: 1195-1203. Whitmore, T.C., N.D. Brown, M.D. Swaine, D. Kennedy, C.I. Goodwinbailey & W.K. Gong. 1993. Use of hemispherical photographs in forest ecology: measurement of gap size and radiation totals in a Bornean tropical rain forest. Journal of Tropical Ecology 9: 131-151. Winkler, N. 1997. Environmentally sound forest harvesting; testing the applicability of the FAO Model Code in the Amazon in Brazil. Forest Harvesting Case Study 8. Food and Agriculture Organization of the United Nations, Rome, Italy. Wittering, W.O. (ed.). 1973. Work study in forestry. Forestry Commission Bulletin 47. HRM Stationary Office, London, UK. Woods, P. 1989. Effects of logging, drought, and fire on structure and composition of tropical forests in Sabah, Malaysia. Biotropica 21(4): 290-298. Wöll, H. 1988. Struktur und Wachstum von kommerziell genutzten Dipterocarpaceenmischwäldern und die Auswirkungen von waldbaulicher Behandlung auf deren Entwicklung, dargestellt am Beispiel von Dauerversuchsflächen auf den Philippinen. PhD thesis, Universität Hamburg, Germany. Wyatt-Smith, J. 1987. Problems and prospects for natural management of tropical moist forests. in: Mergen, F. & J.R. Vincent (eds.). Natural management of tropical moist forests: 4-22. Yale University, School of Forestry and Environmental Studies, New Haven, CT, USA. Wyatt-Smith, J., W.P. Panton & B.A. Mitchell. 1963. Manual of Malayan silviculture for inland forest. Malayan Forest Records 23. Forest Research Institute, Kuala Lumpur, Malaysia. 2 vols. Yasman, I. 1995. Dipterocarpaceae: tree-mycorrhizae-seedling connection. PhD thesis, Wageningen Agricultural University, Wageningen, the Netherlands. Zagt, R.J. 1997. Tree demography in the tropical rain forest of Guyana. PhD thesis, Utrecht University, Tropenbos-Guyana Series 3, Georgetown, Guyana. 251 pp. Zagt, R.J. & M.J.A. Werger. 1998. Community structure and demography of primary species in tropical rain forest. in: Newberry, D.M., H.T.T. Prins & N.D. Brown (eds.). Dynamics of tropical communities (Symposium of The British Ecological Society 37): 193-219. Blackwell Science, London, UK. Zar, J.H. 1996. Biostatistical analysis. 3rd edition Prentice-Hall, Upper Saddle River, NJ, USA. Zöhrer, F. 1980. Forstinventur; ein Leitfaden für Studium und Praxis. Paul Parey, Hamburg-Berlin, Germany. Zukrigl, K., G. Eckhart & J. Nather. 1963. Standortskundliche und waldbauliche Untersuchungen in Urwaldresten der niederösterreichischen Kalkalpen. Mitteilungen der forstlichen Bundesversuchanstalt Mariabrunn 62. Österreichischer Agrarverlag, Vienna, Austria. 309 Reduced impact logging in the tropical rain forest of Guyana 310 Appendix A: List of species APPENDIX A: LIST OF SPECIES List of species occurring in 15 two-hectare plots of Greenheart forest at Pibiri, Central Guyana. Division in currently and potentially commercial timber species and commercially non-desirable - canopy and subcanopy - species. Species names follow Mennega et al. (1988) and Ek & Ter Steege (in prep.). Vernacular names in quotation marks do not occur in Mennega et al. (1988), but follow names used by tree spotters. Uses follow Fanshawe (1961) and Gérard et al. (1996). International trade names, if any and different from the local name, are given in brackets behind the local name. Species Currently marketable species ANACARDIACEAE Anacardium giganteum Hancock ex Engl. APOCYNACEAE Aspidosperma album (Vahl) Benoist ex Pichon Aspidosperma cruentum Woodson Parahancornia fasciculata (Poir.) Benoist ex Pichon BIGNONIACEAE Jacaranda copaia (Aubl.) D. Don subsp. copaia BOMBACACEAE Catostemma commune Sandw. Catostemma fragrans Benth. BURSERACEAE Protium decandrum (Aubl.) Marchand Trattinickia demerarae Sandw. Trattinickia rhoifolia Willd. CAESALPINIACEAE Eperua falcata Aubl. Hymenaea courbaril L. var. courbaril Vernacular (trade) name Uses Ubudi (Caju-açu) Plywood, general carpentry (interior) Shibadan, smooth leaf (Marfim falso) Shibadan, ‘boat’ leaf (Marfim falso) Dukali (Amapa) Heavy construction, general carpentry Heavy construction, general carpentry Interior joinery, plywood Futui (Gobaja) Interior joinery, plywood Baromalli, swamp Baromalli, sand Plywood, interior joinery Interior joinery, plywood Kurokai Cabinet work, interior joinery Plywood, interior joinery Plywood, interior joinery Ulu, rough skin Ulu, smooth skin (Amesclao) Wallaba, soft Locust (Jatobá) Mora excelsa Benth. Mora Mora gongrijpii (Kleinh.) Sandw. Morabukea Peltogyne venosa (Vahl) Benth. subsp. densiflora Purpleheart (Amarante) (Spruce ex Benth.) M.F. Silva Posts, shingles, staves Heavy construction, cabinet work Heavy construction Heavy construction Cabinet work, heavy construction CELASTRACEAE Goupia glabra Aubl. Kabukalli (Cupiuba) Heavy construction, general carpentry FABACEAE Clathrotropis brachypetala (Tul.) Kleinh. Aromata Heavy construction, joinery Cabinet work, general carpentry Cabinet work Diplotropis purpurea (Rich.) Amsh. var. purpurea Swartzia leiocalycina Benth. LAURACEAE Aniba hypoglauca Sandw. Chlorocardium rodiei (Schomb.) Rohwer, Richter & v.d. Werff Licaria cannella (Meissn.) Kosterm. Tatabu (Sucupira) Wamara Silverballi, yellow Greenheart Silverballi, brown 311 Boat building, joinery Heavy construction (marine), general carpentry Cabinet work, boat building Reduced impact logging in the tropical rain forest of Guyana Species Currently marketable species (cont.) LAURACEAE Ocotea canaliculata (L.C. Rich.) Mez Ocotea floribunda (Swartz) Mez cf. Ocotea puberula (Rich.) Nees LECYTHIDACEAE Couratari guianensis Aubl. Lecythis zabucajo Aubl. MELIACEAE Carapa guianensis Aubl. MORACEAE Brosimum guianense (Aubl.) Huber MYRISTICACEAE Virola sebifera Aubl. SAPOTACEAE Chrysophyllum pomiferum (Eyma) T.D. Penn. Chrysophyllum sanguinolentum (Pierre) Baehni Pouteria speciosa (Ducke) Baehni SIMAROUBACEAE Simarouba amara Aubl. STERCULIACEAE Sterculia pruriens (Aubl.) K. Schum. Sterculia rugosa R. Brown VOCHYSIACEAE Ruizterania albiflora (Warm.) Marcano-Berti Vochysia surinamensis Stafl. var. surinamensis Potentially marketable species ANACARDIACEAE Tapirira obtusa (Benth.) J.D. Mitchell CAESALPINIACEAE Peltogyne paniculata Benth. subsp. pubescens (Benth.) M.F. Silva Sclerolobium guianense Benth. var. guianense Vouacapoua macropetala Sandw. COMBRETACEAE Terminalia amazonia (J. Gmelin) Exell EUPHORBIACEAE Glycydendron amazonicum Ducke FABACEAE Dipteryx odorata (Aubl.) Willd. Hymenolobium flavum Kleinh. cf. Hymenolobium petraeum Ducke cf. Ormosia coccinea (Aubl.) Jacks. Ormosia coutinhoi Ducke Swartzia benthamiana Miq. var. benthamiana FLACOURTIACEAE Laetia procera (Poepp.) Eichl. Vernacular (trade) name Uses Silverballi, sawari skin (Louro preto) Silverballi, ‘pea's leaf’ kereti (Louro preto) Silverballi, kereti (Louro preto) Interior joinery, plywood Interior joinery, plywood Interior joinery, plywood Wadara (Tauari) Monkey pot (Sapucaia) Plywood, interior joinery heavy construction Crabwood (Andiroba) Cabinet work Tibokushi (Letterwood) Inlay, turnery Dalli, like (Virola) Plywood Paripiballi Barataballi Heavy construction General carpentry (interior) Plywood, general carpentry (interior) Suya Simarupa (Marupa) Plywood, interior joinery Maho, smooth leaf (Kobe) Plywood, interior joinery Maho, rough leaf (Kobe) Plywood, interior joinery Muneridan (Gonfolo) Iteballi (Quaruba) General carpentry, plywood Interior joinery, plywood Duka (Tatapiririca) Plywood Purpleheart, saka Heavy construction Kaditiri (Tachí) Sarebebeballi Plywood Cabinet work, flooring Fukadi (Nargusta) General carpentry Devildoor tree General carpentry Tonka bean (Cumaru) Darina (Angelim) Koraroballi (Angelim) Barakaro (Tento) Heavy construction General carpentry General carpentry General carpentry, plywood Korokororo (Tento) General carpentry, plywood Itikiboroballi (Saboarana) Cabinet work, joinery Warakairo 312 Interior joinery, plywood Appendix A: List of species Species Potentially marketable species (cont.) HUMIRIACEAE Sacoglottis guianensis Benth. var. guianensis LAURACEAE Aniba citrifolia (Nees) Mez Aniba excelsa Kosterm. Ocotea acutangula (Miq.) Mez Ocotea guianensis Aubl. var. guianensis LECYTHIDACEAE Eschweilera coriacea (A.P. DC.) Mori and E. decolorans Sandw. Eschweilera sagotiana Miers LOGANIACEAE Antonia ovata Pohl MIMOSACEAE Aberema jupunba (Willd.) Britton & Killip var. trapezifolia Barneby & Grimes Enterolobium barnebianum Mesquita cf. Enterolobium cyclocarpum (Jacq.) Griseb. cf. Inga alba (Sw.) Willd. Parkia pendula (Willd.) Benth. ex Walp. Parkia ulei (Harms) Kuhlm. var. surinamensis Kleinh. Pseudopiptadenia suavolens (Miq.) Grimes Zygia racemosa (Ducke) Barneby & Grimes MORACEAE Brosimum rubescens Taub. SIMAROUBACEAE Simaba multiflora A. Juss. Commercially non-desirable canopy species ANNONACEAE Bocageopsis multiflora (Mart.) R.E. Fries Guatteria sp. Unonopsis sp. APOCYNACEAE Ambelania acida Aubl. Aspidosperma exselsum Benth. Geissospermum sericeum (Sagot) Benth. & Hook.f. Himathanthus articulatus (Vahl) Woodson AQUIFOLIACEAE Ilex martiniana D. Don BORAGINACEAE Cordia exaltata Lam. var. melanoneura I.M. Johnston BURSERACEAE Protium tenuifolium (Engl.) Engl. CAESALPINIACEAE Cassia cowanii Irwin & Barneby var. guianensis (Sandw.) Irwin & Barneby Tachigali rusbyi Harms cf. CARYOCARACEAE Caryocar nuciferum L. Vernacular (trade) name Uses Dukuria Heavy construction Gale, almond Gale, greenheart Silverballi, pear leaf Silverballi, shirua Joinery Joinery Joinery, plywood Joinery, plywood Kakaralli, smooth leaf Piling Kakaralli, black Piling Inyak Plywood Huruasa Plywood Devil ‘grandfather's’ ear Maporokon (Inga) Hipanai Uya Heavy construction, general carpentry Heavy construction, general carpentry General carpentry Plywood Plywood Manariballi , like Tureli (Snakewood) General carpentry Cabinet work, joinery Dukaliballi (Satiné) Cabinet work, joinery Hachiballi Plywood, interior joinery Devil's ear Arara, fine leaf Arara, smooth skin Arara, broad leaf Makoriro Yaruru Manyokinaballi Mabwa Kakotaro Table tree Haiawaballi Warua Yawaredan Sawari 313 Reduced impact logging in the tropical rain forest of Guyana Species Commercially non-desirable canopy species (cont.) CECROPIACEAE Pourouma guianensis Aubl. subsp. guianensis CELASTRACEAE Maytenus myrsinoides Reissek CHRYSOBALANACEAE Couepia exflexa Fanshawe & Maguire cf. Couepia sp. Licania alba (Bernouilli) Cuatrec. Licania heteromorpha Benth. var. heteromorpha Licania heteromorpha Benth. var. perplexans Sandw. Licania hypoleuca Benth. Licania majuscula Sagot Licania micrantha Miq. cf. Licania sp. Licania sp. Parinari campestris Aubl. ELEOCARPACEAE Sloanea eichleri K. Schum cf. Sloanea guianensis (Aubl.) Benth. EUPHORBIACEAE Chaetocarpus schomburgkianus (Kuntze) Pax & K. Hoffm. Hyeronima alchorneoides Allemão var. alchorneoides FABACEAE Swartzia jenmanii Sandw. Swartzia oblanceolata Sandw. HIPPOCRATEACEAE Cheiloclinium cognatum (Miers) A.C. Sm. ICACINACEAE Emmotum fagifolium Desv. LAURACEAE Ocotea glomerata (Nees) Mez Ocotea oblonga (Meissner) Mez LECYTHIDACEAE Lecythis confertiflora (A.C. Smith) Mori Lecythis corrugata Poit. subsp. corrugata Lecythis holcogyne (Sandw.) Mori LINACEAE Hebapetalum humiriifolium (Planch.) Benth. MELASTOMATACEAE Miconia lepidota DC. Mouriri sp. MIMOSACEAE Balizia pedicellaris (DC.) Barneby & Grimes MORACEAE Helicostylis tomentosa (Poepp. & Endl.) Rusby NYCTAGINACEAE Neea sp. OLACACEAE Minquartia guianensis Aubl. PROTEACEAE Panopsis sessilifolia (Rich.) Sandw. SAPINDACEAE Talisia squarrosa Radlk. SAPOTACEAE Pouteria caimito (Ruiz et Pavon) Radlk. Pouteria cladantha Sandw. 314 Vernacular name Buruma Kaiarima ‘Bokotokon’, Aruadan (Couepia) Kauta, fine leaf Kautaballi Buruburuli Kairiballi Unikiakia Kautaballi, fine leaf Marishiballi, high bush Kauta Konoko Burada Aruadan, broad rough leaf Aruadan Ruri Suradan Parakusan Serebedan Monkey syrup Manobodin Silverballi, kurahara Silverballi, poor kereti Wirimiri Wina Haudan Shiballidan Waraia, lepidota Mamuriballi Manariballi, common Ituri-ishi-lokodo Mamudan Wanania Mahoballi Moroballi Asepokoballi, fine leaf Aiomorakushi Appendix A: List of species Species Commercially non-desirable canopy species (cont.) SAPOTACEAE Pouteria coriacea (Pierre) Pierre Pouteria egregia Sandw. Pouteria filipes Eyma Pouteria guianensis Aubl. Pouteria reticulata (Engl.) Eyma subsp. surinamensis Penn. Vernacular name Moraballi Kokoritiballi, fine leaf Kamahora, fine leaf Asepoko Kokoritiballi, common Commercially non-desirable sub-canopy species and short-lived pioneers ANNONACEAE Anaxagorea dolichocarpa Sprague & Sandw. Kurihikoyoko Duguetia negelecta Sandw. Yari yari Duguetia yeshidan Sandw. Yeshidan Oxandra asbeckii (Pulle) R.E. Fries Karishiri Xylopia nitida Dunal Kuyama, white Xylopia pulcherrima Sandw. Kuyama, red APOCYNACEAE Tabernaemontana spp. Pero-ishi-lokodo ARECACEAE Oenocarpus bacaba Mart. Lu BORAGINACEAE Cordia nodosa Lam. Huruereroko BURSERACEAE Protium guianense (Aubl.) March. var. guianense Haiawa CECROPIACEAE Cecropia obtusa Trécul. Congo pump, ‘common’ Cecropia sciadophylla Matt. Congo pump, ‘big leaf’ CHRYSOBALANACEAE Hirtella davisii Sandw. ‘Fire tree’ Licania canescens Benoist cf. Marishiballi CLUSIACEAE Rheedia benthamiana Planch. & Triana Asashi Tovomita spp. Awasokule DICHAPETALACEAE Tapura guianensis Aubl. Waiaballi EUPHORBIACEAE Alchornea sp. Kanakudiballi Aparisthmium cordatum (Adr.Juss.) Baill. Mababalli Sandwithia guyanensis Lanj. Unknown ‘A’ FABACEAE Swartzia spp. Itikiboroballi, others FLACOURTIACEAE Ryania speciosa Vahl var. tomentosa (Miq.) Monach. Kibihidan LAURACEAE Aniba kappleri Mez Gale, ginger Endlicheria punctulata (Mez) C.K. Allen Yekoro MALPIGHIACEAE Byrsonima aerugo Sagot Arikadako Byrsonima stipulacea A. Juss. Kanoaballi MELASTOMATACEAE Miconia hypoleuca (Benth.) Triana Wakaradan Miconia punctata (Dresr.) D. Don Waraia, punctata MELIACEAE Trichilia sp. Yuriballi MIMOSACEAE Inga lateriflora Miq. Shirada 315 Reduced impact logging in the tropical rain forest of Guyana Species Vernacular (trade) name Commercially non-desirable sub-canopy species and short-lived pioneers (cont.) MIMOSACEAE Inga spp. Warakosa Pentaclethra macroloba (Willd.) Kuntze Trysil MONIMIACEAE Siparuna sp. Muniridan MORACEAE Trymatococcus paraensis Ducke Pasture tree MYRTACEAE Calyptranthes forsteri O. Berg Kakirio Eugenia patrisii Vahl. Hichu Marliera schomburgkiana O. Berg Akarako Myrcia sp. Ibibanaro OLACACEAE Chaunochiton kappleri (Sagot ex Engl.) Ducke Hiwaradan Dulacia guianensis (Engl.) Kuntze Hishirudan Heisteria cauliflora Sm. Makarasali QUIINACEAE Quiina guianensis Aubl. Okokonshi, fine leaf Quiina obovata Tul. cf. Okokonshi Touroulia guianensis Aublet Fern tree RUBIACEAE Amaioua guianensis Aubl. var. guianensis Komaramaraballi Duroia eriophila L.f. var. eriophila Komaramara Faramea quadricostata Bremek Kamadan, green flower Ixora aluminicola Steyerm. ‘Coffeeballi’ Palicouria crocea (Swartz) Roem. & Schult. Kamadan, red/orange flower Palicouria guianensis Aublet subsp. occidentalis Steyerm. Kamadan, yellow flower Psychotria capitata Ruiz Lopez & Pavon subsp. inundata Kamadan, white flower (Benth.) Steyerm. RUTACEAE Zanthoxylum apiculata (Sandw.) Waterman Sada SAPINDACEAE Cupania hirsuta Radlk. Kulishiri, hairy black Cupania scrobiculata Rich. Kulishiri, black Matayba sp. Kulishiri, white SAPOTACEAE Micropholis venulosa (Mart. & Eichl.) Pierre Kudibiushi Pouteria sp. Kamahora, medium leaf Pouteria venosa (Mart.) Baehni subsp. amazonica T.D. Penn. Kamahora, broad leaf SIMAROUBACEAE Picramnia guianensis (Aubl.) Jansen-Jacobs Hikuribianda, like TILIACEAE Apeiba petoumo Aubl. Duru VIOLACEAE Paypayrola guianensis Aubl. and P. longifolia Tul. Adebero 316 Appendix B: Estimation of tree volumes APPENDIX B: ESTIMATION OF TREE VOLUMES Measuring stem volumes individually on standing trees in tropical rain forest is difficult and time consuming. Stem volumes can also be estimated from relationships between certain stem dimensions and stem volume. Diameter, height and form are the independent variables that are commonly used to determine the values of the dependent variable - stem volume. The final result is presented in equation or table form, which then gives the average volume of individual trees in terms of one or more of the previously mentioned stem dimensions. Volume tables or equations can be classified into three categories: i.e. standard volume tables, where the volume is a function of diameter at breast height and height (e.g. equation 1 below); local volume tables, where the volume is a function of dbh only (e.g. equation 2 below), and form class volume tables where dbh, height and a measure of tree form (taper) are the independent variables (e.g. equation 3 below - e.g. Husch et al 1982). Standard volume tables are more precise than local volume tables, but not often used for tropical forests because tree height is normally difficult and time consuming to measure. (1) v = b0 + b1 ⋅ d 2 ⋅ h (2) v = b0 + b1 ⋅ d + b2 ⋅ d 2 (3) v = b0 + b1 ⋅ d ⋅ di ⋅ h where: d h di = = = diameter at breast height height (crown point) upper stem diameter The logging experiment enabled measurement of the stem length of felled trees, and hence the establishment of standard volume equations for the experimentally logged area. However, for the conventionally logged area, no stem length data were available and local volume equations were needed. In order to obtain the data to establish volume equations or tables, diameter measurements at various intervals along the stem are collected from felled sample trees. The stem is normally divided in short sections between four and ten, depending on the length of the bole - and each section is measured according to a standard formula; e.g. Huber’s, Smalian’s or Newton’s formula (Husch et al 1982; Zöhrer 1980). In the present study, the basic data for the development of volume equations were collected during the reduced impact logging operation. Because of the nature of the experiment - boles had to be extracted and remained the property of the timber company -, the boles could not be cut into sections. Therefore, volumes over bark of 317 Reduced impact logging in the tropical rain forest of Guyana the individual trees were taken along the entire useful bole length using Newton’s formula: vob = gb + 4 × g m + g t 6 where vob gb gm gt = = = = tree volume over bark in m3 basal area of butt end (m2) basal area in middle of log (m2) basal area of top end (m2) There are several equations that can serve as models for the construction of local volume tables (ex Zöhrer 1980): v = b0 + b1 ⋅ d 2 v = b0 + b1 ⋅ d + b2 ⋅ d 2 log v = b0 + b1 ⋅ log d (following Kopezki-Gehrhardt) (following Hohenadl-Krenn) (following Husch) Each of these models was tested for seven species. The criterion adopted for the selection of the best equation was the value of the coefficient of determination. Table B.1 One entry volume equations for seven common commercial species in Pibiri, Guyana. v = bole volume over bark in m3 and d = diameter (m dbh) Species Chlorocardium rodiei Catostemma spp. Eperua falcata Peltogyne venosa Swartzia leiocalycina Goupia glabra Carapa guianensis Number of sample trees 273 52 34 25 20 19 15 Equation r2 v = 0.0274 + 9.342⋅d2 log v = 1.094 + 2.051⋅log d log v = 1.008 + 2.120⋅log d log v = 1.057 + 2.031⋅log d v = 0.614 + 6.615⋅d2 log v = 1.042 + 2.369⋅log d log v = 1.047 + 2.243⋅log d 0.76 0.86 0.83 0.93 0.62 0.88 0.77 Despite the small number of data for most of the species, the coefficients of determination are rather high. The volume equation derived for Greenheart based on data collected in Pibiri gives lower estimated volumes than an equation based on records of the Guyana forestry Commission for 4714 sample trees in the Bartica Triangle (Ter Steege 1990): vob = –0.592 + 11.29⋅d2 (r2 = 0.94) This is likely related to a difference in branch-free bole length and may indicate a difference in forest architecture between the two forest areas. 318 Appendix B: Estimation of tree volumes The linear coefficients for the species for which a logarithmic model was fitted are rather similar which suggests that the volumes could be estimated by one single volume equation. Differences between species were tested by fitting a model using dummy variables to identify species. For the sake of comparison a logarithmic model was also used for Chlorocardium (log v = 0.955 + 1.960⋅log d; r2 = 0.73) and Swartzia (log v = 0.816 + 1.551⋅log d; r2 = 0.57). This resulted in the following best fitting model: log v = 0.965 + 0.114⋅[B] + 0.89⋅[P] + 2.000⋅log d (R2 = 0.84), where [B] is a dummy variable for Catostemma and [P] for Peltogyne. Dummy variables are coded 1 if the tree belonged to that species and coded 0 if it did not belong to it. These results indicate that Catostemma and Peltogyne have either longer or less tapered boles than the other species. Field observations confirm that Catostemma has a relatively higher bole volume because of a cylindrical stem form and Peltogyne, because of longer bole length, which would be expected from an emergent species. It must be noted that the basic data that where used to obtain the equations included only trees with a dbh ≥ 40 cm. Their use for estimating volume increment in growth and yield analysis is therefore limited. It is recommended to include data on small trees (preferably measured on felled trees). The number of sample trees need not be very high, provided that the basic data are collected as accurately as possible and the different size classes are well represented (cf. Silva 1989). 319 Estimates of daily and effective hourly cost (US$) of each type of labour used in the two logging systems. The scheduled operation time per year counts 250 days. The costs are based on 1998 wage and taxation rates. ECH = effective crew hour. Designation 320 Compass Man Line Cutter Inventory Crew-chief Tree Spotter Office Technician Feller Feller assistant Skidder / Crawler Operator Skidder Assistant Log Scaler Log Scaling Assistant Log Loader / Motor Grader Operator Truck Driver Truck Assistant Cook Supervisor Basic Hourly Wages1 0.54 0.50 0.00 0.70 0.70 0.88 0.50 0.88 0.50 0.57 0.50 0.79 0.70 0.50 0.57 0.00 Basic Monthly Salary2 107.68 98.94 237.93 139.70 139.70 174.62 98.94 174.62 98.94 114.97 98.94 157.16 139.70 98.94 114.97 303.45 Legal Burden3 12.92 11.88 53.22 21.79 21.79 32.96 11.88 32.96 11.88 13.87 11.88 27.37 21.79 11.88 13.87 74.19 Indirect Payments4 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 25.60 51.20 Total Monthly Cost5 146.20 136.41 316.75 187.08 187.08 233.18 136.41 233.18 136.41 154.43 136.41 210.13 187.08 136.41 154.43 428.83 Food & Field Allowance 6 2.17 2.17 2.17 2.17 1.21 2.17 2.17 2.17 2.17 2.17 2.17 2.17 1.21 1.21 2.17 1.21 Daily Cost7 9.19 8.72 17.38 11.15 10.19 13.37 8.72 13.37 8.72 9.59 8.72 12.26 10.19 7.75 9.59 21.79 ECH per Cost per day ECH8 5.0 5.0 5.0 5.0 6.5 6.5 6.5 6.8 6.8 6.8 6.8 6.8 7.6 7.6 8.0 6.8 1.84 1.74 3.48 2.23 1.57 2.06 1.34 1.97 1.28 1.41 1.28 1.80 1.34 1.02 1.20 3.21 Reduced impact logging in the tropical rain forest of Guyana Table C.1 Appendix C: cost calculations APPENDIX C: COST CALCULATIONS 1 Basic hourly wage rates (fortnightly workers) and monthly salary rates (monthly workers) are rates for October 1998, provided by Demerara Timber Ltd. (M. Shahabudeen, pers.comm.). 2 Most salaries of forest workers in Guyana are based on a fortnight system. This implies that every other Saturday is a working day. Saturdays are paid at overtime rate; i.e. 150%. Counting a Saturday for 1.5 weekday, the number of working days amounts to 25 days per month on average. Each working day counts 8 hours. 3 Legal burden consists of taxes as well as social security and health insurance (NIS). The 1998 Guyana tax system distinguished three monthly income brackets: personal tax allowance US$ 114.60 (0%), first tax bracket: $ 114.60 to $ 201.15 (20%), second bracket: > $ 201.15 (33%). The National Insurance System distinguished two monthly income brackets with a limit $ 317.25. Incomes in the first bracket are charged 12% (employer and employee levies together). 4 Indirect payments include housing and medical and education facilities in the Mabura Hill Township. The cost of the township is US$ 2,345 per month, with a total number of 458 workers - office, forest and sawmill workers - (M. Shahabudeen, pers.comm.). Since most forest workers share their quarters with others, indirect payments were estimated at $ 25.60 for forest workers. Holiday bonuses and superannuating schemes do not apply. 5 Monthly costs are the sum of basic salary, legal burden and indirect payments. 6 Food and field (camp) allowances are paid on actually worked days and amount to $ 1.20 per day and $ 1.00 per night respectively. 7 Daily costs were calculated by multiplying the monthly costs by 12 and dividing by 250 working days per year. The number of 250 days was obtained by subtracting 115 non-working days per year. Nonworking days are: 52 Sundays, 26 non-working Saturdays, 13 public holidays, 12 days paid leave and 12 rain days. 8 Hourly costs were calculated by dividing the daily costs by the effective work hours per day. Effective working hours are those hours of the scheduled working hours (8) the worker is actively engaged in work. It equals the scheduled time minus time for meal breaks and travelling to and from the work site. The percentage of the scheduled operation time varies with the activity. 321 Estimates of the hourly operation cost (US$) of the pieces of equipment used in the two logging systems. The scheduled operation time per year counts 250 days (see Table C.1). The costs are based on 1998 prices. Chain saw Chain saw Skidder11 Manufacturer1 Model Stihl AV066 Stihl 070 Gross engine power (kW)2 Initial investment cost Economic life (hr)3 Effect. Crew time (hr/yr.) 4 Effect. Machine time (hr/yr.) 5 5 $ 1,174 2000 1625 650 6 $ 873 2000 1625 650 $ 0.53 $ 0.19 $ 0.39 $ 0.14 $ 0.72 $ 0.54 $ $ $ $ $ $ $ $ Description and data 322 Fixed cost of ownership Depreciation Interest (16%) Insurance (3%) Sub-total Variable cost of operation Fuel6 Lubricants and filters7 Maintenance and repairs8 Special wear items9 Tyres/Undercarriage10 Sub-total Total hourly cost 0.57 0.99 0.53 1.21 $ 3.30 $ 4.03 0.69 1.37 0.39 1.39 $ 3.84 $ 4.37 Log Loader Caterpillar 528 Crawler Tractor Caterpillar D8R 130 $ 180,000 10000 1700 1445 250 $ 388,000 18000 1700 1445 19.40 23.21 0.70 43.30 Caterpillar 966D Motor Grader Caterpillar 12G Logging Truck12 Mercedes 2636 149 $ 275,000 12000 1700 1445 101 $ 200,000 15000 1900 1615 260 $ 145,000 15000 1900 1710 $ $ $ $ 20.63 17.06 0.51 36.88 $ $ $ $ 12.00 10.97 0.33 23.30 $ $ $ $ 8.70 7.56 0.23 16.48 Four wheel drive vehicle Land Rover Defender 110 Tdi 83 $ 30,000 6000 1900 1710 $ $ $ $ 4.50 1.81 0.06 6.36 Service Truck Mercedes 915 112 $ 75,000 12000 1900 1710 $ $ $ $ 16.20 11.41 0.34 27.94 $ $ $ $ $ $ $ $ 5.63 4.01 0.12 9.75 $ $ $ $ $ $ $ 6.46 0.49 16.20 0.88 6.43 30.47 58.41 $ 12.41 $ 1.01 $ 19.40 $ 7.40 $ 0.73 $ 20.63 $ 5.01 $ 0.54 $ 12.00 $ 9.93 $ 0.24 $ 6.69 $ 1.59 $ 0.13 $ 3.46 $ 4.28 $ 0.18 $ 4.32 $ 8.10 $ 40.92 $ 84.23 $ 5.36 $ 34.12 $ 72.31 $ 6.43 $ 23.99 $ 47.28 $ 1.03 $ 17.90 $ 34.38 $ 0.48 $ 5.66 $ 12.01 $ 0.54 $ 9.32 $ 19.08 Reduced impact logging in the tropical rain forest of Guyana Table C.2 Appendix C: cost calculations 1 Information provided by Demerara Timbers Ltd. (personal communication and Forest Management Plan 1997-2001) 2 Information obtained from Stihl (http://www.stihl.de/produkte), Caterpillar (Caterpillar Performance Handbook, 1986), and Mercedes Benz (http://dialog.mercedes-benz.com/d/ecars/trucks). 3 The economic life (years of service) of a piece of equipment depends on the intended use, operating conditions and maintenance practices (Caterpillar, 1986). Caterpillar distinguishes three categories (zones) of application and operating conditions: moderate, average and severe. For all equipment, category B was considered applicable. Estimates of economic life of non-Caterpillar machinery, e.g. logging trucks, were based on the estimated economic lifetime mileage given by the manufacturer. Estimated years of service of chainsaws vary widely. FAO (1978) suggests 800 hours, Miyata (1980) 500 hours, Tsoumis (1992) 2500 hours and Holmes et al. (1999) 2000 hours. The latter value was applied. 4 Effective crew time is the part of the scheduled operation time the crew is actively engaged in work. It equals the labour time minus time for meal breaks and travelling to and from the work site. The percentage of the scheduled operation time varies with the activity (see Table C.1). 5 Effective machine time is that part of the effective crew time that a machine actually operates. Time for machine servicing and time lost due to delays such as mechanical breakdowns, waiting for supplies, weather, etc. are deducted from the effective crew time. Machine utilisation rates vary with the type of equipment. FAO (1978) utilisation rates were used: chainsaw: 40%; skidder, crawler tractor, log loader and motor grader: 85%; and trucks and four-wheel drive: 90%. 6 Price of two-stroke gasoline $ 0.46/l (chainsaw) and of diesel $ 0.28/l (all other machines). Hourly consumption was calculated according to the formula given in Section 5.2. Consumption rates of Caterpillar machines thus calculated agree with rates indicated by the manufacturer for zone B. 7 Hourly cost of lubricants and filters was initially calculated with the formula given in Section 5.2. The price of engine oil is $ 1.60/l. Comparison with manufacturers’ estimates showed that the formula tends to underestimate the cost of lubricants and filters for Caterpillar machinery and to overestimate the cost for Mercedes Benz trucks. In both cases, the manufacturer’s estimate has been used instead. The hourly cost of lubricants for the chainsaws was grossly overestimated by the formula. Chain oil is consumed in proportion to the fuel consumption rate. The price of chain oil is $ 2.00/l. The hourly cost of crankcase oil is negligible in case of chainsaws. 8 Maintenance and repairs are estimated at 100% of the hourly depreciation cost as recommended by FAO (1977, 1978). Miyata (1980) distinguishes different percentages for different types of equipment, for example skidder 50% of depreciation cost, crawler tractor 100%. Manufacturers’ estimates of maintenance and repair cost are markedly lower, for example $ 10.00 per hour for a skidder (60% of the cost applied in this study). The manufacturers’ estimates are based on North-American experience with less severe working conditions and higher skilled labour, and FAO’s recommendations were used instead. 9 Special wear items are the chainsaw bar and chain, and files in case of chainsaws and winch line, hooks, (and straps) in case of the skidder. Prices of chainsaw bars are $ 86.70 for a Stihl 070 (30′′) and $ 78.00 for a Stihl AV066 (25′′). Prices of chains are $ 17.75 and $ 14.80 respectively. Price of both flat and chain files is $ 1.45. Economic life of a bar is 120 hours, of a chain 30 hours, of a flat file 60 hours and a chain file 30 hours. The price of wire rope is $ 1.20 per ft. Wire rope length is 150 ft. The price of an arch hook with shackle is $ 76.60 and of a Bardon® choker hook with wedge-type ferrule $ 66.40. Straps used in the reduced impact operation cost $ 6.00 each. Expected life of a winch line (and straps) is 250 hours and of hooks 500 hours. 10 Price of a set of skidder or log loader tyres is $ 14,300; motor grader tyres $ 21,450; logging truck tyres $ 4,500; Land Rover tyres $ 825; and service truck tyres $ 1,900. For calculation of hourly cost, see Section 5.2. Hourly cost of undercarriage maintenance of the crawler tractor is based on Caterpillar (1986). 323 Reduced impact logging in the tropical rain forest of Guyana 11 The hourly operating cost in this column applies to the skidder in the conventional operation, which was equipped with 150-ft wire rope and arch hook and shackle. Special wear-items of the skidder used in the reduced impact operation included six choker hooks and ferrules. The hourly cost of this equipment was $ 1.70, rendering an hourly operating cost of $ 59.20. 12 For trucks and the like, the effective machine time (‘round-trip time’) is divided into ‘standing time’ which is the part of the round-trip time when the truck is at the landing or at the delivery point, either waiting to be loaded or unloaded or in the process of being loaded or unloaded and ‘travelling time’ which is the time for loaded and unloaded travel. Fixed costs of ownership are based on the entire roundtrip time, whereas variable costs of operation are based on travelling time only. Table C.3 Estimated cost by activity for conventional and reduced impact logging. Depending on the type of available information costs are either hourly or daily costs. Hourly costs are per effective machine hour except for felling, trimming and scaling, where hourly costs are per effective crew hour. Cost for each activity is in US$ and consists of labour, equipment and materials. These component costs are listed in the footnotes underneath the table. Labour cost rates per effective crew hour are given in Table C.1. Labour cost per effective machine hour can be derived by multiplying by the ratio of the effective crew time to the effective machine time for each activity (Table C.2). Activity Pre-harvest ($/day) Block lay-out1 Inventory2 Liana cutting3 Data processing4 Map making5 Road planning6 Harvest preparation Main road construction7 ($/hr) Feeder road construction8 ($/hr) Landing construction9 ($/hr) Skid trail marking10 ($/day) Harvest ($/hr) Felling and bucking11 Skidding12 Landing operations13 Main transport ($/hr) Loading14 Trucking15 Unloading16 Overhead Road maintenance17 ($/day) Logistics18 ($/day) Supervision ($/day) Cooks ($/day) Base camp19 ($/m3) Administration cost20 ($/m3) Royalty21 ($/m3) Area fee22 ($/ha) 1 Conventional operation Reduced impact operation 18.93 122.42 11.41 10.61 30.19 19.11 76.52 19.95 11.41 10.61 30.19 235.85 90.31 90.31 - 235.85 90.31 90.31 18.92 6.37 62.03 6.59 7.72 65.00 6.59 74.43 37.01 74.43 74.43 37.01 74.43 319.17 240.77 21.79 19.17 1.66 0.33 2.68 6.07 319.17 240.77 21.79 19.17 1.66 0.33 2.68 6.07 CNV costs: Labour = $ 17.90/day (1 compass man and 1 line cutter); materials = $ 1.00/day (includes surveyor’s rope, compass, 2 cutlasses, cutlass file, flagging tape, tarpaulin, 2 hammocks). RIL costs: Labour as in CNV: materials = $ 1.20/day (includes items as listed in CNV plus clinometer). 324 Appendix C: cost calculations 2 CNV costs: Labour = $ 62.00/day (1 inventory crew chief, 4 tree spotters); materials = $ 60.45 (includes 5 diameter tapes, aluminium tags (25/ha), 5 cutlasses, 2 files, 1 sheet holder, waterproof paper (1 sheet/10ha), tarpaulin, 5 hammocks). Costs of items that are consumed on a hectare basis - e.g. tags have been converted by multiplying ha-values with daily output (ha/day). RIL costs: labour as in CNV: materials = $ 14.55/day (includes items listed for CNV and, in addition, 2 compasses, 2 measuring tapes (30m)). 3 Lianas were cut in RIL only. Costs: Labour = $ 19.85/day (1 tree spotter, 1 line cutter): materials = $ 0.08 (2 cutlasses, 2 files). 4 CNV = RIL costs: Labour = $ 10.20/day (1 technician); materials = $ 1.25/day (personal computer) 5 CNV = RIL costs: Labour = $ 10.20/day (1 technician); materials = $ 0.45/day (drafting pens, millimetre paper (1 sheet/100 ha in CNV, 4 sheets 1 sheet/6 ha in RIL). 6 Road planning consists of office work (aerial photo interpretation and map making) and fieldwork. CNV = RIL costs: Office work: Labour = $ 10.20/day (1 technician): materials = $ 0.20/day (aerial photographs, topographic maps, drafting pens). Field work: labour = $ 17.90/day (1 compass man and 1 line cutter); materials = $ 1.90/day (includes surveyor’s rope, compass, 2 cutlasses, 2 cutlass files, flagging tape, tarpaulin, 2 hammocks). 7 CNV = RIL costs: Labour = $ 12.95/hr (1 supervisor, 4 machine operators, 1 assistant); Equipment = $ 222.90 (includes the hourly operating cost of a CAT D8 crawler tractor, CAT 966 bucket loader, MB 915 dump truck, and CAT 12G motor grader; see Table C.2). 8 and 9 CNV = RIL costs: Labour = $ 6.10/hr (1 supervisor, 1 crawler operator); Equipment = $ 84.20 (includes the hourly operating cost of a CAT D8 crawler tractor; see Table C.2). 10 Marking of skid trails in RIL only. Costs: Labour = $ 17.90/day (1 compass man and 1 line cutter); materials = $ 1.00/day (includes surveyor’s rope, compass, 2 cutlasses, 2 cutlass files, flagging tape, tarpaulin, 2 hammocks). 11 CNV costs per effective crew hour: Labour = $ 5.10/hr (1 chainsaw operator, 1 logging clerk and 1 assistant); equipment = $ 1.20 (hourly operating cost of a Stihl 070 (Table C.2) times 0.28 (effective machine time per day/effective crew time per day (Section 5.3)); materials = $ 0.05/hr (cutlass). RIL costs per effective crew hour: Labour = $ 5.45/hr (1 chainsaw operator, 1 crew leader and 1 assistant); equipment = $ 1.85 (hourly operating cost of a Stihl AV066 (Table C.2) times 0.46 (effective machine time per day/effective crew time per day (Section 5.3)); materials = $ 0.40/hr (cutlass, felling wedges, sledgehammer). 12 CNV costs: Labour = $ 3.60/hr (1 skidder operator and 1 assistant); equipment = $ 58.40 (hourly operating cost of a CAT 528 skidder equipped with bull-hook; see Table C.2). RIL costs: Labour = $ 5.80/hr (1 skidder operator, 1 crew chief and 1 assistant); equipment = $ 62.00 (hourly operating cost of a CAT 528 skidder equipped with Bardon choker hooks; see Table C.2) 13 Landing operations involve cross-cutting and trimming of the extracted rough logs and scaling of the trimmed logs. CNV costs are assumed to equal RIL costs. Debarking (estimated cost = $ 0.12 per m3) and other log protective measures are in fact required for non-durable (peeler) species. Difference in species composition will also affect the royalty rate and log price. Because reliable determination of the latter is fraught with difficulties (see Section 5.5), it was decided to ignore difference in species composition. Cross-cutting and trimming: Labour = $ 1.95/hr effective crew time (1 chainsaw operator); equipment = $ 1.70/hr (hourly operating cost of a Stihl 070 (Table C.2) times 0.40 (effective machine time per day/effective crew time per day (Section 5.3)); materials = $ 0.03/hr (cutlass). Scaling Labour = $ 2.68/hr effective crew time (1 log scaler and 1 assistant): materials = $ 0.21/hr (measuring tape (30m), diameter tape, branding hammer, stationary, chalk). 14 and 16 CNV = RIL costs: Labour = $ 2.12/hr; Equipment = $ 72.30/hr (includes the hourly operating cost of a CAT 966D front-end loader; see Table C.2). 325 Reduced impact logging in the tropical rain forest of Guyana 15 CNV = RIL costs: Labour = $ 2.62/hr; equipment = $ 34.40/hr (includes the hourly operating cost of a MB 2636 logging truck; see Table C.2). 17 CNV = RIL costs. It is assumed that main roads are maintained on a daily basis and that two logging teams operate along the road. It is also assumed that the road is only used for logging purposes and costs are not divided among other activities than logging. Cost of road maintenance per day. Labour = $ 13.70 (1 motor grader operator); equipment = $305.50/day (hourly operating cost of a CAT 12G motor grader; see Table C.2) 18 Transport of workers (pre-harvest planning, harvest preparation and harvesting crews) from field camp to work site, from base camp to field camp or from base camp to work site, is carried out by a four-wheel drive vehicle or a MB 915 truck. Supplies for logging machines (fuel, spares) are delivered by a MB 915 service truck. Several vehicles are employed simultaneously to carry out these transports. For convenience of comparison, it is assumed that these trips by different vehicles add up to a full day’s work for both one service truck and one four-wheel drive vehicle. Cost of labour = $ 28.15/day (2 drivers and 1 assistant); equipment = $ 212.65/day (daily operating cost of a Land Rover Defender 110 Tdi and a MB 915; see Table C.2) 19 and 20 No detailed data were available on the cost of the base camp (including workshop, stores, etc.) and administration cost (administration of inputs and outputs, personnel administration). Costs of these facilities in $/m3 for October 1998 were provided by Demerara Timbers Ltd. (M. Shahabudeen, pers.comm.). 21 and 22 The Forest Act of Guyana (Cap. 67:01 of the laws of Guyana, 2 may 1953) distinguishes several forest charges. The Forest Act was revised only once in 1973, but forest charges have been updated several times since. Current charges include application and licence issue fees, security bonds, export commission, royalties on timber and area fees. Only the latter two charges are considered here. Area fees are differentiated by class of forest harvesting permit and by area. Since 1998, a rate of US$ 0.15 per acre is being charged for Timber Sales Agreements larger than 300,000 acres. Four royalty rates for timber as logs are differentiated, charged per Hft: special category timbers (greenheart and Purpleheart), and Class I, Class II and Class III timbers. Most species felled in the operations studied here either belong to the special category or Class I timbers. Since 1996 the following rates apply: $ 0.10/Hft for special category and $ 0.06/Hft for Class I. Reduced impact logging may lead to a shift in species composition of the output. A shift in composition affects handling costs (e.g. debarking), royalty rates, but foremost the price of logs delivered at the sawmill or other delivery point. The effect of a shift in species composition is treated separately in the text. 326 Appendix C: cost calculations Table C.4 Estimated output by activity for conventional and reduced impact logging. Depending on the type of activity and available information output is in hectares or cubic metres, on either an hourly or a daily basis. Output per hour is per effective machine hour except for felling, trimming and scaling, where output is per effective crew hour. Activity Pre-harvest (ha/day) Block lay-out1 Inventory2 Liana cutting3 Data processing4 Map making5 Road planning6 Harvest preparation Main road construction7 (ha/hr) Feeder road construction8 (ha/hr) Landing construction9 (ha/hr) Skid trail marking10 (ha/day) Harvest (m3/hr) Felling and bucking11 Skidding12 Landing operations13 Main transport (m3/hr) Loading14 Trucking15 Unloading16 Overhead (m3/day) Road maintenance17 Logistics18 Supervision18 Cooks18 1 2 Conventional operation Reduced impact operation 9 50 7 10 5 25 25 100 50 100 100 37 20 24 37 20 24 8 10.6 14.4 19.3 6.7 15.9 20.9 21.7 6.0 28.9 21.7 6.0 28.9 167.2 83.6 83.6 83.6 194.7 97.3 97.3 97.3 Logging compartments are divided in square logging blocks of 100 ha. In both conventional and reduced impact logging 50 m forest inventory strips were cut. Each 100 ha block is served by 22 km of cut lines (19 × 1 km inventory lines and 3 × 10 km perimeter lines; the fourth perimeter line is shared with the adjacent block). For each hectare, 220 m cut line is thus needed. The average daily output, based on random field observations, was estimated at 2 km per day in the conventional (without slope correction) and 15 km per day in the reduced impact system (with slope correction). In conventional inventory, a 50 m wide strip, with 50 m × 100 m recording units, was browsed by four tree spotters (cf. Hendrison, 1990). Once a spotter finds a harvestable tree, its position is sketched on a map. In the reduced impact operation, strips were divided in 50 m × 50 m recording units. Position of trees was recorded by taking its polar co-ordinates (compass bearing and distance to one of the corners of a recording unit). Random field observations indicated an output of 50 ha per day in CNV. For RIL, the inventory of five 100-ha blocks was monitored in detail, and an average output of 10 ha per day was found. 3 Liana cutting was monitored over 17 hectares in the RIL experimental plots, and had an output of 5 hectares per day for a two men crew. 4 In the conventional inventory 25 trees were recorded per hectare on average. One person can process data for 750 trees per day on average. In the reduced impact operation, more trees were recorded per hectare, because trees with defects were included as well. In addition, more data had to be processed per tree. 327 Reduced impact logging in the tropical rain forest of Guyana 5 The conventional operation used maps with a scale of 1:2500, whereas the reduced impact operation used maps with a scale of 1:1250. One map sheet, with the size of four ‘letter’-size sheets covered 100 ha in the conventional operation. In the reduced impact operation ‘letter’-size sheets were used covering about 6.25 hectares each. 6 Road planning starts with planning from a 1:50,000 topographic map, assisted by aerial photo interpretation (1:41,000). The fieldwork consists of staking the alignment on the ground and adjusting it where necessary. A topographic map sheet covers approximately 100,000 ha and it is assumed that a draftsman can complete 10,000 ha per day. Assuming a road density of 1 km feeder road and 200 m main road for each 100 ha block (see Section 5.2; FAO 1978), 1200 m of road alignment have to be marked per 100 ha. Because adjustments have to be made to the planned alignment the daily output of the road alignment crew is assume to be lower than that of block and inventory line establishment. 7 For both RIL and CNV, main road construction involves clearing of the area where the road is to be built, piling of the uprooted trees (accumulating trees in windrows), and forming and grading the roadbed. A certain amount of surfacing material should be placed on a forest road, the depth of which depends on the characteristics of the soil on which the road is constructed. Concerning clearing and piling, the performance of a bull-dozer can be estimated by formulas developed by Rome Industries (Caterpillar, 1986). The time required to uproot trees per hectare is estimated by: T = X·[ A·(B) + M1·N1 + M2·N2 + M3·N3 + M4·N4 + D·F] , where: T X A B M N D F = = = = = = = = Time per hectare in minutes Wood density factor affecting total time Density or vine presence factor affecting base time Base time for a specific tractor per hectare Minutes per tree in each diameter range for a specific tractor Number of trees per hectare in each diameter range Sum of diameter in 30 cm of all trees per hectare above 180 cm in diameter at ground level Minutes per 30 cm for trees above 189 cm in diameter Diameter ranges start from 31 cm in diameter at ground level: M1 = 31-60 cm, M2 = 61-90 cm, M3 = 91120 cm and M4 = 121-180 cm. The number of trees in each diameter range is based on the diameter class frequency distribution of the experimental plots, whereby it was assumed that the diameter at ground level = 1.25 × diameter at breast height. For the forest in the research area the following time was derived: Tclearing = 1.3[2(45)+0.2(112.0)+ 1.3(23.1)+2.2(3.1)+6(0.6)] = 199 min/ha A similar procedure has been developed for estimating piling output for a tractor equipped with a KG blade (according to Caterpillar, 1986). Tpiling = 111+0.1(112.0)+0.5(23.1)+1.8(3.1)+3.6(0.6) = 142 min/ha For forming and grading an additional 50% of the time required for clearing and piling is recommended (FAO 1977). This implies that the hourly output of the bulldozer is 0.12 ha. The width of clearing of a main road being 16 m (DTL Forest Management Plan 1997-2001), the hourly output in linear meter of road = 73.4 m/hr. Since each 100 ha logging block is served by 200 linear m of main road, 37 ha are opened up per machine hour. Because the output rate of the subsequent sub-operations - grading, surfacing, etc. - is limited by the bulldozer output, it is assumed that the output of the other machinery motor grader, bucket loader and dump truck - equals the bulldozer output. 8 For both RIL and CNV, the construction of a feeder road requires the same bulldozer input as the construction of the main road. The width of clearing of a feeder road being 6 m (DTL Forest management Plan 1997- 2001), the hourly output in linear meter of road becomes 195.9 m. This figure is confirmed by field observations made during the opening up of the experimental area in 1994, where five kilometres of feeder road were established in three days. Since each 100 ha logging block is served by 1000 linear m of feeder road, 20 ha of forest are opened up per machine hour. 328 Appendix C: cost calculations 9 For both RIL and CNV, two landings totalling 5,000 m2 (one, measuring 50 m × 50 m, on either side of the feeder road) are assumed to serve 100 ha. Applying the same bulldozer output rate as with road construction, it takes 4.3 hours to clear, pile and form the total landing area, implying an output of 24 ha of opened up forest per machine hour. 10 It is assumed that 12,500 linear m of skid trail have to be marked on the ground per 100 ha (see Fig.5.10a), with an output of 1000 m per day. This implies an output of 8 ha of forest flagged per day. 11 Felling and bucking output was determined in the time study (see Section 5.3) 12 Skidding output was determined in the time study (see Section 5.4; output Scenario II, including the effect of trail system efficiency). 13 FAO (1978) suggests that the effective crew time required to cross-cut and trim one extracted log - with an average diameter between 40 and 60 cm - is 8 minutes. Furthermore it is suggested to adjust the hourly output volume by multiplying by 0.90 because of strain on the worker. The size of the average extracted log was larger in the reduced impact operation than in the conventional operation (see Section 5.4). This was mainly related to the length of the extracted log. It was therefore assumed that the time required to trim a log was the same in both operations. The hourly performance of log scaling was set at 20 m3/hr for both operations (FAO (1978). 14 The hourly performance of a loader can be calculated based on the rated capacity of the machine. Assuming a truckload of 21.7 m3 (see Note 15), a load would be composed of 7 logs on average. If logs are loaded one by one the grapple load is 3.1 m3. According to FAO (1977) the loading time per trucktrailer load is composed of a fixed time of 5 minutes (waiting for hauling rigs to be prepared, to rearrange load and to drive away after loading has been completed). The average loader cycle time is 1.6 minutes. This results in an estimated loading time of 16.2 minutes per truck-trailer load, or an hourly performance of 80 m3. This figure is, however, misleading, because the feasible loading capacity of the loader and truck turnover are hardly ever co-ordinated in such a way that frequent interruptions in the actual loading time are avoided (FAO 1978). The loader is required, however, to carry out auxiliary work too, such as turning and arranging logs for cross-cutting and sorting and stacking of trimmed logs. FAO (1978) suggests that, in fact, the loader works throughout the feasible machine time and for reasons of convenience the total charge is made to loading. For both RIL and CNV the work performance per effective machine hour is then considered equal to one truck load (see Note 15). 15 The output determining time factors for trucking are the round-trip time and the volume per load. Based on a hauling distance of 50 km, an average empty speed of 45 km/hr and an average loaded speed of 30 km/hr, a travel time of 2.8 hours was derived. Standing time is estimated at 50 minutes (20 minutes loading, 10 minutes unloading and 10 minutes idle at each terminal). The round-trip time is then estimated at 3.6 hours. The load capacity of the truck and trailer is 26 tonnes. Assuming a specific weight of 1.2 t/m3, the truck-trailer load becomes 21.7 m3, and the truck-trailer’s performance 6.0 m3/hr. 16 The hourly performance of a loader in unloading is usually higher than in loading (FAO 1977). Furthermore, the feasible unloading capacity is better utilised because logging usually is taking place at more than one site, which increases truck turnover time. Conversely, it is likely that the machine is required to carry out more auxiliary work at the log yard, such as sorting and stacking of logs, than at the landing. FAO (1978) suggests that the hourly cost of unloading is 75% of the cost of unloading. Sine the same make and model of machine is used, this translates to an increased output of 133%. 17 Road maintenance is carried out on a daily basis and, as a matter of simplification, is assumed to be performed by one motor grader only. It is also assumed that two logging teams operate in the same logging compartment. Assuming that the road is only used for logging purposes, the cost of road maintenance has to be borne by the daily performance of two logging teams. Daily output of a logging team is limited by skidding performance and amounted to 84 m3/day in CNV and 97 m3/day in RIL. 18, 19 and 20 The same principle holds for these overheads as for road maintenance. It is, however, assumed that each logging team features one supervisor, two cooks (one for block lay-out, inventory, road planning crews, and one for felling, skidding and landing operations crews), and makes use of one four- 329 Reduced impact logging in the tropical rain forest of Guyana wheel drive vehicle and service truck. The daily output is thus determined by the daily output of the skidder. Table C.5 Summary of stumpage cost in US$/m3 in a conventional (yield = 28.5 m3/ha) and a reduced impact operation (yield = 31.0 m3/ha) both with logging intensity of 10 trees/ha. Average skidding distance = 383 m and hauling distance = 50 km. To calculate the cost per m3 the hourly (daily) cost (see Table C.3) and corresponding hourly (daily) output (see Table C.4) was determined for each activity. If performance was in m3/hr or m3/day, the hourly (daily) cost was divided by the corresponding performance to obtain cost per m3. If output was in ha/hr or ha/day, the hourly (daily) cost was first divided by the output and this result ($/ha) was then divided by the yield per hectare in the particular operation (see top of caption). Activity Pre-harvest Block lay-out Inventory liana cutting Data processing Map making Road planning Harvest preparation Main Road Construction Feeder Road Construction Landing Construction Skid trail marking Harvest Felling Skidding Landing operations Main transport Loading Trucking Unloading Overhead Road maintenance Logistics Supervision Cooks Base camp Admin. Cost Royalty Area fee Total Cost (US$/m3) Conventional Reduced impact operation operation NPV (US$/m3) Conventional Reduced Impact operation operation 0.08 0.08 0.01 0.01 0.01 0.09 0.25 0.13 0.01 0.01 0.01 0.08 0.10 0.01 0.01 0.01 0.10 0.29 0.15 0.01 0.01 0.01 0.23 0.16 0.14 - 0.21 0.15 0.12 0.08 0.24 0.17 0.14 - 0.22 0.16 0.13 0.08 0.60 4.30 0.33 1.16 4.10 0.33 0.60 4.30 0.33 1.16 4.10 0.33 3.43 6.17 2.58 3.43 6.17 2.58 3.43 6.17 2.58 3.43 6.17 2.58 1.91 2.88 0.26 0.23 1.66 0.33 2.68 0.21 28.28 1.64 2.48 0.22 0.20 1.66 0.33 2.68 0.19 28.23 1.91 3.10 0.28 0.25 1.66 0.33 2.68 0.21 28.60 1.64 2.66 0.24 0.21 1.66 0.33 2.68 0.19 28.57 Note: total logging cost reported here is lower than reported for Guyana elsewhere (e.g. Flaming 1995; Landell-Mills 1997). This is partly due to a higher yield per hectare in the present study. Average yields in Guyana: 4 to 10 m3/ha (Flaming 1995). It may also be due to machine operating costs having been estimated following handbook guidelines (FAO 1977, 1978; Caterpillar 1986; Sundberg & Silversides 1988). Finally it may be due to performances of pre-harvest planning and preparation operations, loading, trucking and unloading having been estimated using general guidelines. 330 Appendix C: cost calculations Table C.6 Estimated output by activity for reduced impact logging for four different logging intensities. For activities of which the output is measured in hectares the same values apply as in Table C.4, and therefore these activities are not repeated here. Output per hour is in m3 per effective machine hour except for felling, trimming and scaling, where output is per effective crew hour. Activity Felling and bucking1 Skidding2 Landing operations Loading Trucking3 Unloading Road maintenance4 Logistics, support and supervision5 Logging intensity (N/ha) 4 8 m3/hr m3/hr 6.6 6.7 15.6 15.9 20.9 20.9 21.7 21.7 5.6 5.9 28.9 28.9 m3/day m3/day 191.4 195.2 95.7 97.6 2 m3/hr 6.5 15.3 20.9 21.7 5.0 28.9 m3/day 187.8 93.9 16 m3/hr 6.7 16.2 20.9 21.7 6.1 28.9 m3/day 199.1 99.6 1 Comparison of the time required by each sub-operation in the felling cycle in plots with different logging intensity showed that only the time required to locate the next tree was influenced by the logging intensity. Tree locating time differed significantly (t124=-3.819, p < 0.0005). Regression analysis produced the following estimates: 2.0, 1.6, 1.2 and 0.8 minute for 2, 4, 8 and 16 trees felled per hectare respectively. 2 Comparison of the time required by each sub-operation in the skidding cycle showed that the time required to collect logs at the stump area was significantly different (t70 = –2.370, p < 0.05). Regression analysis produced the following estimates: 27.6, 26.8, 26.1 and 25.4 minutes for 2, 4, 8 and 16 trees felled per hectare. 3 In the comparison between conventional and reduced impact logging the actual hauling distance of 50 km was used. A change in logging intensity has as its most important consequence that more area is needed to obtain a certain output. Consequently, hauling distances will change with a change in logging intensity. The timber company projected to achieve an annual log output of 100,000 m3 during the period 1997 - 2001 (DTL, 1996). At a logging intensity of 2 trees/ha (= 6.3 m3/ha) such a yield would imply a harvested area of 15,800 ha/yr. At a main road density of 1 km/500 ha, the average main road distance to be covered within this harvested area would amount to 16 km. Every time the logging intensity is doubled; this distance will be halved. In the base case the logging intensity was 10 trees/ha (31.5 m3) and the actual hauling distance 50 km. To achieve a yield of 100,000 m3 an area of 3,200 ha would be required at this logging intensity and the average main road distance within the area would amount to 3.2 km. The adjusted average hauling distance is then 63 km in case of 2 trees/ha, 55 km in case of 4 trees/ha, 51 km in case of 8 trees per ha and 49 km in case of 16 trees/ha. Based an average empty speed of 45 km/hr, an average loaded speed of 30 km/hr and a standing time of 50 minutes, round trip times of 4.3, 3.9, 3.6 and 3.5 hours are derived. 4 Just as in the base case, it is assumed that the cost of road maintenance has to be borne by the daily performance of two logging teams. Daily output of a logging team is limited by skidding performance and amounted to 93.9, 95.7, 97.6 and 99.6 m3/day for the four logging intensities. 5 The same principle holds for these overheads as for road maintenance. It is, however, assumed that each logging team features one supervisor, two cooks (one for block lay-out, inventory, road planning crews, and one for felling, skidding and landing operations crews), and makes use of one four-wheel drive vehicle and one service truck. The daily output is thus determined by the daily output of one skidder. 331 Reduced impact logging in the tropical rain forest of Guyana Table C.7 Estimated cost by activity for reduced impact logging at four different logging intensities in US$/m3. Average skidding distance = 383 m and yield per log = 3.1 m3. For cost per m3 for each logging intensity, the hourly or daily cost for each activity as shown in Table C.3 was used (exceptions in notes below). If performance was in m3/hr (m3/day), the hourly (daily) output for each activity is shown in Table C.6. If output was in ha/hr or ha/day, the data from Table C.4 were adjusted in proportion to the change in logging intensity. Activity Logging intensity (N/ha) 4 8 2 Pre-harvest Block lay-out Inventory liana cutting Data processing Map making Road planning Harvest preparation Main Road Construction Feeder Road Construction Landing Construction Skid trail marking Harvest Felling1 Skidding Landing operations Main transport Loading Trucking2 Unloading Overhead Road maintenance3 Logistics4 Supervision Cooks Base camp Admin. Cost Royalty Area fee Total 1 16 0.43 1.19 0.62 0.07 0.07 0.03 0.22 0.60 0.31 0.03 0.03 0.01 0.11 0.30 0.16 0.02 0.01 0.01 0.06 0.15 0.08 0.01 0.01 0.01 1.01 0.72 0.60 0.37 0.50 0.36 0.30 0.19 0.25 0.18 0.15 0.09 0.12 0.09 0.08 0.05 1.18 4.25 0.32 1.17 4.17 0.32 1.16 4.08 0.32 1.14 4.01 0.32 3.43 7.53 2.58 3.43 6.68 2.58 3.43 6.25 2.58 3.43 6.03 2.58 2.12 3.23 0.23 0.21 1.66 0.33 2.68 0.94 35.83 1.82 2.77 0.23 0.20 1.66 0.33 2.68 0.48 31.08 1.66 2.52 0.22 0.19 1.66 0.33 2.68 0.23 28.61 1.57 2.38 0.22 0.19 1.66 0.33 2.68 0.12 27.30 The time to locate trees increases with decreasing intensity. Therefore, the ratio of the effective machine time to the effective crew time varies, affecting the cost per effective crew hour. Hourly cost is estimated at $ 7.67, $ 7.70, $ 7.72 and $ 7.74 for the four successive logging intensities. 2 Truck travel times increase with decreasing intensity, whereas the standing time remains unaffected. Varaible operating costs of trucks are based on travel times. They are estimated at $ 37.86, $ 37.36, $ 37.06 and $ 36.90 for the successive logging intensities. 3 Realising that the hauling distance is depending on the logging intensity, adjustments were made for the difference in length of road that has to be maintained. This was done by increasing the base cost (Table C.3) by multiplying it by the proportional increase in the hauling distance relative to the hauling distance in the base case. 4 The same principle holds for the transport of goods and people as for road maintenance, and the base cost (Table C.3) was multiplied by the proportional increase in hauling distance relative to the distance used in the base case. 332 Appendix C: cost calculations 333 Reduced impact logging in the tropical rain forest of Guyana 334 CURRICULUM VITAE Pieter van der Hout werd op 30 maart 1957 geboren te Rotterdam. In 1976 behaalde hij het Gymnasium-B diploma aan het Johannes Calvijn Lyceum te Rotterdam. In hetzelfde jaar begon hij met de studie bosbouw (oriëntatie tropisch) aan de toenmalige Landbouwhogeschool te Wageningen. Het kandidaatsexamen werd in 1980 behaald. Een zesmaands stage werd verricht bij het Centrum voor Landbouwkundig Onderzoek in Suriname in 1982, waar hij veldwerk verrichtte voor Reitze de Graaf en John Hendrison. Verder werden twee afstudeervakken gedaan. Het eerste betrof een literatuurstudie naar de sociaaleconomische aspecten van het agroforestry systeem ‘taungya’ op Java en een boswaarderingsonderzoek in opdracht van de gemeente Putten (Gelderland). Deze onderwerpen vormden samen een hoofdvak bij de Vakgroep Boshuishoudkunde. Het tweede betrof een verzwaard hoofdvak tropische houtteelt bij de Vakgroep Bosteelt, waarvoor onderzoek naar de regeneratie na verschillende bosteeltkundige ingrepen in semi-loofverliezend bos in Ivoorkust werd gedaan. Het doctoraalexamen in de Landbouwwetenschappen in de studierichting Bosbouw werd in mei 1985 behaald. Van 1989 tot 1992 was hij onder auspiciën van de Britse ontwikkelingsorganisatie ‘Voluntary Service Overseas’ werkzaam bij de Guyana Forestry Commission. Hij was daar hoofd van de afdeling bosinventarisatie, met als belangrijkste taak het trainen van staf van de bosdienst en van houtconcessies in bosinventarisatietechnieken. Tevens werd een aanzet gemaakt voor een bosbeheersplan voor het oosten van Guyana (Corentyne regio). In januari 1992 volgde een aanstelling voor zes jaar als toegevoegd onderzoeker bij de Vakgroep Botanische Oecologie en Evolutiebiologie van de Universiteit Utrecht, gefinancierd door de Stichting Tropenbos. Gedurende deze periode werd in totaal vier jaar veldwerk verricht in het regenwoud van Guyana voor het onderzoek dat in dit proefschrift wordt beschreven. Vanaf oktober 1999 gaat hij beginnen aan een vervolgstudie bij het Tropenbos-Guyana Programma. 335 ISBN 90-393-2185-X