Copyright © NISC Pty Ltd Ostrich 2007, 78(1): 31–36 Printed in South Africa — All rights reserved OSTRICH ISSN 0030–6525 doi: 10.2989/OSTRICH.2007.78.1.5.49 Habitat preferences of birds in a montane forest mosaic in the Bamenda Highlands, Cameroon Jirí Reif1,2*, Ondrej Sedlácek1, David Horák1,2, Jan Riegert3, Michal Pešata4, Záboj Hrázský5,6 and Štepán Janecek7 V V V V V 1 V V Department of Ecology, Faculty of Science, Charles University in Prague, Vinic ná 7, CZ-128 44 Praha, Czech Republic 2 Department of Zoology, Faculty of Science, Charles University in Prague, Vinic ná 7, CZ-128 44 Praha, Czech Republic 3 Department of Zoology, Faculty of Biological Sciences, University of South Bohemia, Branišovská 31, CZ-370 05 C eské Budejovice, Czech Republic 4 Department of Ecology, Faculty of Agriculture, University of South Bohemia, Studentská 13, CZ-370 05 C eské Budejovice, Czech Republic 5 Department of Botany, Faculty of Biological Sciences, University of South Bohemia, Branišovská 31, CZ-370 05 C eské Budejovice, Czech Republic 6 Academy of Sciences of the Czech Republic, Institute of Systems Biology and Ecology, Na Sádkách 7, CZ-370 05 C eské Budejovice, Czech Republic 7 Academy of Sciences of the Czech Republic, Institute of Botany, CZ-379 01, Tr ebon, Czech Republic * Corresponding author, e-mail: jirireif@yahoo.com V V V V V V V V V V V Although the high species richness and endemism of birds in the Bamenda Highlands has attracted ornithological research for decades, most studies have been restricted to bird communities of continuous montane forests. Instead, we focused on a mosaic landscape with montane forest remnants, where the habitat preferences of birds remain unknown. We performed an assessment of habitat associations of birds in the Bamenda Highlands in the Cameroon Mountains. Using a point count census method, we detected 71 species within the study area. The most abundant species were the Northern Double-collared Sunbird Cynniris reichenowi, the Oriole Finch Linurgus olivaceus, the Common Stonechat Saxicola torquata, the Thick-billed Seed-eater Serinus burtoni, the Black-crowned Waxbill Estrilda nonnula, the Brown-backed Cisticola Cisticola chubbi and the Yellow-breasted Boubou Laniarius atroflavus. Canonical correspondence analysis revealed that the most important environmental gradient structuring the bird community follows the forest coverage. We found that both endemic and non-endemic montane species are more closely associated with montane forest remnants, compared to widespread species. Endemic species are most closely dependent on continuous forest cover. However, some montane species did not show any clear habitat associations and thus can be viewed as local habitat generalists. This study shows that many restricted-range species (including endangered endemics) are able to live in fragmented landscapes, which cover a substantial part of the Bamenda Highlands. Therefore, conservation programmes should focus their action plans on these landscapes. Introduction The Cameroon Mountains, with extraordinarily high concentrations of endemic bird species (Stattersfield et al. 1998), contribute substantially to the species richness of West African forests, which are currently recognised as a biodiversity hotspot of global importance (Orme et al. 2005). Such high endemic species richness is probably caused by environmental stability and spatial isolation of local montane forests (Fjeldså and Lovett 1997) that form the only large area of this environment in West Africa (Mayaux et al. 2004). The Bamenda Highlands are formed by several mountain ranges in the central part of the Cameroon Mountains. In the Bamenda Highlands, endemic bird species richness is higher than that of southern parts of the Cameroon Mountains, probably because of longer isolation and the larger extent of montane forests during the Pleistocene (Smith et al. 2000). In spite of its uniqueness, the Cameroon Mountains represent a gap within the Pan-African network of protected areas because of the absence of legal protection (de Klerk et al. 2004). Unfortunately, a lack of conservation effort has meant that during past decades there have been few or no constraints on extensive deforestation and forest fragmentation in most parts of the Cameroon Mountains (Stuart 1986). Within the Bamenda Highlands, the Kilum-Ijim forest on Mount Oku represents the largest block of continuous montane forest (Thomas et al. 2000). Most ornithological research activities in the region have been focused on this area because of its considerable importance for montane forest bird conservation (Dowsett-Lemaire and Dowsett 1998, McKay and Coulthard 2000). Recently, Forboseh et al. (2003) estimated densities of many bird species in several habitats within the Kilum-Ijim forest. Their study assessed the effectiveness of current conservation effort and identified future conservation priorities. Nevertheless, most of the Bamenda Highlands is covered by a mosaic of montane environments with small forest patches restricted 32 to steep slopes and creek valleys. The conservation value of such fragmented areas still remains unclear because of a lack of relevant data on bird distributions. Stuart (1986) performed several field surveys and described bird species composition in forest remnants at several localities. However, acquired information about species abundances and habitat preferences is in many respects insufficient. To our knowledge, there has been no attempt to assess the abundance of bird species in fragmented areas of the Bamenda Highlands. Therefore, it is unclear which species are restricted to remaining forest patches, which are able to survive in scrubland outside forests and which species of non-forest habitat specialists are living in fragmented landscapes. Without this information, it would not be possible to make effective conservation plans necessary for the protection of substantial parts of the Cameroon Mountains. Therefore, the aims of our paper are to assess habitat preferences of particular bird species, using accurate estimates of their abundances in a mosaic landscape of the Bamenda Highlands, and to compare these preferences between endemics of the Cameroon Mountains, afromontane non-endemic species, and widespread species. Study site The study was performed in the area named My Ogade in the Bamenda Highlands, North-West Province, Cameroon (geographical position: N 06°05’ 26”, E 10°18’ 09”; 2 200m asl). The area, which covered about 1km2, comprised: (a) high Hyparrhenia grasslands; (b) upper montane grasslands dominated by Sporobolus africanus; (c) intensive pastures dominated by Pennisetum cladestianum; (d) forest clearings dominated by Pteridium aquilinum; (e) speciesrich extensive pastures; (f) species-rich shrub vegetation with numerous Labiatae and Compositae; (g) Gnidia glauca woodlands; (h) stream mantel vegetation with numerous species from the family Acanthaceae, and (i) montane forests dominated mainly by Schefflera abysinica, S. manii, Bersama abyssinica, Syzygium staudtii, Carapa grandiflora and Ixora foliosa. Although the majority of the abovementioned vegetation types (except for Types c, d and e) are native, with the proportions of individual vegetation types progressively changing with increasing deforestation. The present-day vegetation cover is characterised by isolated forest fragments surrounded by vegetation types a–g. The montane forest was represented by two large patches (c. 20ha together) and several small fragments (0.1–1ha). The total cover of montane forest did not exceed more than 30% of the study plot area. The mean distance between fragments was c. 200m but shrubby corridors connected most forest fragments. Methods Bird census The bird census was carried out using a point count method (Bibby et al. 2000), which is recommended for areas with dense vegetation cover and high species richness (Gregory et al. 2003). To maximise sampling efficiency, we established two perpendicular transects where Reif, Sedlácbek,Horbák, Riegert, Pešata, Hrázský and Janecbek we located 50 census points at 100m-distances. The first transect contained 20 points and the second one 30 points. Transects were as straight as possible and covered all main habitat types within the study area. Elevation of both transects varied within a range of 200m. Changes in elevation between census points did not influence bird community structure (JR and co-workers, unpublished data). We performed bird censuses from 24 November–14 December 2003. We conducted three visits of each census point, recording all birds (both visually and acoustically) within a 50m-radius during 10min on each visit. We performed all visits during morning hours (between 06:00 and 10:00), changing the order of visited points to factor out the effect of daytime (see Reif et al. (2006) for a more detailed description of the bird census technique). The maximum counts recorded from all visits were taken as the species abundance at a particular point. The abundance of each species in the study plot was calculated as the sum of point abundances. We sorted all species into three categories: (i) endemics of the Cameroon Mountains, (ii) montane species non-endemic to the Cameroon Mountains, and (iii) species widespread throughout Africa, occurring in both lowlands and highlands. Species whose ranges occur in elevations mostly above 1 200m asl were considered to be montane species (sensu Graham et al. 2005). (See Reif et al. (2006) for the abundance and category classification of particular species). For data analyses, we excluded eight species of aerial feeders and raptors (the Black Sparrowhawk Accipiter melanoleucos, the Rednecked Buzzard Buteo auguralis, the Western Marsh Harrier Circus aeruginosus, the White-backed Vulture Gyps africanus, the Lanner Falcon Falco biarmicus, the Redthroated Cliff Swallow Hirundo fuligula, the Barn Swallow Hirundo rustica, and the Black Saw-wing Psalidoprocne pristoptera) because of the high probability of counting the same individuals at more than one census point. Vegetation sampling We estimated the relative coverage of particular vegetation layers in a 50m-radius around each census point. We distinguished five vegetation layers: up to 1m, 1–3m, 3–5m, 5–10m and >10m above the ground. We also estimated the degree of continuity of shrub and forest, respectively, on a scale from one (solitary trees or bushes) to five (one continuous block). These vegetation characteristics were summarised into eight habitat variables (i.e. coverage of each of five vegetation layers, number of vegetation layers, forest continuity and shrub continuity), describing each census point. These habitat variables were used for further analysis, to explain spatial changes in the bird community structure. Data analysis We used canonical correspondence analysis (CCA) to relate the data on bird abundance to habitat variables. CCA is a multivariate direct-gradient analysis technique, which is able to detect the patterns of variation in bird community composition that can be explained by the set of environmental variables. CCA ordinates the samples (census points) and variables (bird species and habitats) along axes Ostrich 2007, 78(1): 31–36 33 the order of variables tested can affect test results (Crawley 2003), each variable was tested independently during the manual selection. Autocorrelation is inevitably present in all data describing any spatial structure (Legendre and Legendre 1998). We controlled for the overall spatial trend in our transect data by adding two variables (first- and second-order polynomial functions, respectively) describing the position of a census point along the transect. Differences in species habitat use (expressed by CCA species scores along first two canonical axes) were tested by one-way ANOVA. 0.3 1.5 (b) eus dyb 0.0 str sem lan mac 0.6 0.0 –2.0 –2.0 1.5 (d) nes sch tau ban cry rei Axis 1 0.0 lan atr bra ban and mon plo ban col sjo and tep cos isa uro epi Axis 1 plo ins elm alb 0.0 3.0 0.0 cya ori eup cap pog cor cis chub cin rei lin oli ser bur apa pul apa cin par alb apa jac pse aby plo mel –2.0 –2.0 Axis 1 mot fla mus adu 0.0 alc leu jyn ruf Axis 2 (c) cor cri lag rub cin bou vid mac cor alb col str ser mos fra squ den goe phy tro plo bag est non tur tym cis bru pyc bar den fus bat min emb tah sax tor tur pel zos sen cos niv pog bil ant cin chlo nat ant tri herbs up to 1 m 1.5 ori nig est ast forest continuity trees above 10 m trees 5–10m –0.4 -0.4 Axis 2 Species composition We detected 71 species within the My Ogade area: 28 of them were montane species and 43 were widespread species. The montane species group comprises 11 shrub continuity shrub and trees 3–5 m number of vegetation layers Axis 1 0.0 Results and discussion Axis 2 (a) shrub 1–3 m Axis 2 such that the differences between species and samples, respectively, are maximised. Each ordination axis represents an environmental gradient along which the centroids of individual variables and samples are distributed so as to maximise differences between them (Storch et al. 2002). CCA is based on the assumption that species distribution is unimodal along environmental gradients. The species score is proportional to the mean of sample scores weighted by abundance of respective species, and indicates the centre of the distribution of the species. CCA was performed in CANOCO for Windows (ter Braak and Šmilauer 2002). We present the results of CCA as a plot depicting scores of particular bird species and habitat variables, respectively, within two-dimensional space of the first two canonical axes (Figure 1). The origin of both axes lies in the centre of environmental gradients expressed by the axes. We used Monte Carlo permutation tests (500 runs) to test the explanatory power of particular habitat variables, using both manual and automatic selection of variables. Because cor cae 3.0 –2.0 –2.0 0.0 3.0 Figure 1: Positions of bird species (triangles) and habitats (arrows) in the space of the first two canonical axes expressing the most important environmental gradients within a bird community in the Bamenda Highlands, calculated by canonical correspondence analysis. Arrows indicate directions of increasing importance of particular habitat variables, and triangles are centroids of species distributions. The closer the distance of a species centroid from the end of a habitat arrow or a canonical axis, the tighter the bird-habitat association. See Methods section for further explanation of canonical correspondence analysis. For better readability, we provide separate graphs for: (a) habitats, (b) widespread species, (c) montane non-endemic species, and (d) endemics of the Cameroon Mountains. Note that the variable ‘number of vegetation layers’ does not correlate with any of the two axes. See Appendix 1 for species abbreviations 34 endemics of the Cameroon Mountains and one endemic (Bannerman’s Turaco Tauraco bannermani) of the Bamenda Highlands. The commonest species was the Northern Double-collared Sunbird Cynniris reichenowi, detected at all census points. Other dominant species (i.e. species over 5% of the total number of individuals in the bird community) were the endemic Yellow-breasted Boubou Laniarius atroflavus, three montane non-endemic species (the Oriole Finch Linurgus olivaceus, the Thick-billed Seed-eater Serinus burtoni and the Brown-backed Cisticola Cisticola chubbi, and three widespread species (the Black-crowned Waxbill Estrilda nonnula, the Common Stonechat Saxicola torquata, and the Orange-Tufted Sunbird Cynniris bouvieri. Despite the quite high number of species detected, we did not record several montane species, which are reported from other upper montane forests in the region by Stuart (1986) and Forboseh et al. (2003): the Bar-tailed Trogon Apaloderma vittatum, the Cameroon Olive Greenbul Phyllastrepus poensis, the Banded Wattle-eye Platysteira laticincta, and the Green-breasted Bush-Shrike Malaconotus gladiator. Because of the short time span of the study (i.e. three weeks in total), it is difficult to decide whether or not these species were present in the study area. We could have failed to detect them, or they could have been absent due to unsuitable habitat conditions or possible altitudinal movements. Presumably, the extent of montane forest in the study area may be too small for the occurrence of some of these species. This factor can also cause remarkably low abundance of some montane forest species (e.g. the endemic Green Longtail Urolais epichlora) which are reported to be common in non-disturbed areas (Serle 1965, Forboseh et al. 2003). Thus, these species seem to be under the highest extinction risk. The absence of other montane species listed in Fotso (2001) for the Mount Oku area (the Grey-chested Illadopsis Kakamega poliothorax, Füllerborn’s Boubou Laniarius poensis, Bamenda Apalis Apalis bamendae) can be explained by their preference of lower altitude forests, up to 2 200m asl (Stuart and Jensen 1986). Main gradients in the bird community structure The most important habitat factors influencing the spatial structure of the bird community were the coverage of trees above 10m height and the coverage of herbs under 1m, as indicated by results of CCA with Monte Carlo permutation tests of particular environmental variables (Table 1). These variables also explain two most important gradients in habitat requirements of the bird community expressed by the first two canonical axes (Figure 1a). The first axis can be viewed as a gradient of increasing forest coverage (7.5% variation of bird species data explained) and the second axis depicts the gradient from scrubland to herbaceous vegetation (3.4% variation of bird species data explained). Species with the highest scores at the first axis are the most closely confined to the forest habitat (Figures 1b–d) e.g. the Green Longtail Urolais epichlora, the Red-faced Crimsonwing Cryptospiza reichenowi and the White-bellied Kingfisher Alcedo leucogaster. Such pronounced habitat preferences could, however, be partly caused by the low number of census points where these species were Reif, Sedlácbek,Horbák, Riegert, Pešata, Hrázský and Janecbek detected. Generally, montane species (both endemic and non-endemic) had higher scores at the first axis than the widespread species (ANOVA: F (2, 60) = 7.17, p = 0.002; Figure 2). In other words, montane species are more forestdependent than widespread species. Among montane species, endemics tend to be even more closely connected with continuous forest than non-endemic montane species (Figures 1 and 2), but this pattern is not statistically significant (Tukey test: p = 0.562). These bird-habitat associations generally concur with findings of Forboseh et al. (2003) from the Kilum-Ijim Forest in the Bamenda Highlands. Similarly, studies listing bird communities in other parts of the Cameroon Mountains, e.g. Bowden (2001) from Mount Kupe and Smith and McNiven (1993) from Tchabal Mbabo, reported the forest dependence of montane species. Despite this general pattern, our results reveal several exceptions to the habitat preferences of widespread species. The African Thrush Turdus pelios, the Great Blue Turaco Corythaeola cristata, and the Tambourine Dove Turtur tympanistria were not confined to deforested areas, but occupied montane forest patches (Figure 1). These species are presumably able to live in a wide array of forest types, irrespective of altitude (Borrow and Demey 2001). The second axis described the gradient from dense scrubland to open areas with scattered trees. Species clustered in the positive part of the second axis (e.g. the Cameroon Blue-headed Sunbird Cyanomitra oritis, the Black-winged Oriole Oriolus nigripennis, and the Common Waxbill Estrilda astrild) occur mostly along shrubby corridors that connect particular forest remnants. The negative part of the second axis separates species living in open grassy habitats, including Palaearctic migratory species such as the Tree Pipit Anthus trivialis or the Yellow Wagtail Motacilla flava. Figure 1 indicates that endemic species (Figure 1d) would have higher scores at the second axis than the non-endemic montane species (Figure 1c) and the widespread species (Figure 1b). However, this result is not significant (ANOVA: F (2, 60) = 1.455, p = 0.242). Interestingly, scores of many montane species (both endemic and non-endemic) are located near the origin of both canonical axes (Figures 1c–d). We propose two explanations for this pattern. First, distribution of these species along the census transect was not correlated with distribution of habitats recognised in our study. Although we cannot exclude this possibility using our data, we offer a second explanation with higher ecological relevance. These species seem to be tolerant of spatial environmental variation at the study site. For instance, they are present at census points with a variable proportion of trees above 10m as well as with various extents of open areas. Although some of these species were previously thought to be able to tolerate disturbed habitat (e.g. Bannerman’s Weaver Ploceus bannermani, Cisticola chubbi and Laniarius atroflavus: Stuart 1986, Forboseh et al. 2003), other representatives of the group clustered near the axes’ origin are — according to Forboseh et al. (2003) — classified as true forest species (e.g. the African Hill Babbler Pseudoalcippe abyssinica, the Grey Apalis Apalis cinerea or the Black-collared Apalis Apalis pulchra). High habitat plasticity of these species perhaps affords opportunity for Ostrich 2007, 78(1): 31–36 35 Table 1: Results of Monte Carlo permutations testing the significance of influence of individual habitat variables on spatial structure of a bird community in the Bamenda Highlands. Permutations were performed using both automatic forward selection and manual independent testing of particular variables. Variables that were significant (P < 0.05) in both procedures are in bold Manual selection Variable name SCORE AT THE FIRST CCA AXIS (mean ±SE) Trees above 10m Trees 5–10m Shrubs 3–5m Shrubs 1–3m Herbs under 1m Number of veg. layers Forest continuity Shrub continuity 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 Variation explained 9% 3% 2% 3% 5% 3% 2% 2% Automatic selection F P 2.96 0.94 0.98 1.15 1.79 0.87 1.11 0.97 0.002 0.536 0.496 0.254 0.004 0.656 0.272 0.538 Variation explained 9% 4% 3% 2% 4% 3% 2% 2% F P 2.96 1.19 0.98 0.88 1.61 0.98 0.69 0.82 0.002 0.21 0.49 0.7 0.02 0.464 0.942 0.774 light of our findings, we argue that it is necessary to start conservation actions in fragmented areas throughout the region, because they are able to host high numbers of restricted-range species. Further research should focus on factors responsible for long-term survival of these species living in such disturbed landscapes. WIDESPREAD MONTANE ENDEMIC Figure 2: Differences in scores along the first canonical axis between widespread species (labelled ‘widespread’), montane non-endemic species (‘montane’) and endemics of the Cameroon Mountains (‘endemic’) in a bird community in the Bamenda Highlands. ANOVA: F (2, 60) = 7.17, p = 0.002 Acknowledgements — We wish to thank Dáša Bystøická, Michael Bartoš and Jakub Brom, who helped us in the field. David Storch provided valuable comments on statistical analyses. Matthias Waltert and an anonymous referee provided useful suggestions, which helped us to improve the manuscript. We are grateful to the Bamenda Highlands and Kilum-Ijim Forest Projects, especially to Michael Boboh Vabi, for enabling us to perform the research in the Bamenda Highlands. The study was performed with the kind permission of Ndawara-Belo Ranch. We thank the entire KedjomKeku community, and particularly Ernest Vunan and Devine Chikelem from Satec NGO, for their kind reception in Big Babanki village. The research was funded by the Grant Agency of the Czech Republic (GACR 206/03/H034, IAA601410709), the Academy of Sciences of the Czech Republic (AVOZ60050516, KJB601110703) and the Ministry of Education of the Czech Republic (MSM 0021620828, MSM 6007665801, LC06073). References their survival in mosaic montane landscapes. Consequently, not only primeval montane forests but also disturbed areas such as our study plot, should be the subject of conservation actions, as they host many restricted-range and globally-threatened montane bird species. Conclusions We found that endemics and montane non-endemic species are more closely confined to montane forest habitat than widespread species in an afromontane forest mosaic. Some montane species, including several endemics and globally-threatened species, showed surprisingly high habitat flexibility: they occurred in both large forest blocks and small forest fragments, as well as in forest edges and scrubland. Such tolerance of some forest bird species to deforestation was also observed in lowland areas of Cameroon (Waltert et al. 2005a) and in Indonesia (Waltert et al. 2005b). Despite quite high resistance of montane species to habitat fragmentation and alteration, some species reported from mature forests in the Bamenda Highlands did not occur in the study area. However, in the Bibby CJ, Burgess ND, Hill DA and Mustoe SH 2000. Bird Census Techniques (2nd edn). Academic Press, London Borrow N and Demey R 2001. Birds of Western Africa. Christopher Helm, London Bowden CGR 2001. The birds of Mount Kupe, south-west Cameroon. Malimbus 23: 13–31 Crawley MJ 2003. Statistical Computing. An Introduction to Data Analysis using S-Plus. John Willey and Sons Ltd, Chichester de Klerk HM, Fjeldså J, Blyth S and Burgess ND 2004. Gaps in the protected area network for threatened Afrotropical birds. Biological Conservation 117: 529–537 Dowsett-Lemaire F and Dowsett RJ 1998. Surveys of Oku Mountain and other IBAs in NW Province (Cameroon), February–March 1998. Report for BirdLife International. Tauraco Press, Liège Fjeldså J and Lovett JC 1997. Geographical patterns of old and young species in African forest biota: the significance of specific montane areas as evolutionary centres. Biodiversity and Conservation 6: 325–346 Forboseh PF, Keming EC, Toh CL and Wultof INB 2003. Monitoring of Kilum-Ijim Forest bird communities: initial findings. Bird Conservation International 13: 255–271 Fotso RC 2001. A contribution to the ornithology of Mount Oku Forest, Cameroon. Malimbus 23: 1–12 36 Reif, Sedlácbek,Horbák, Riegert, Pešata, Hrázský and Janecbek Graham CH, Smith TB and Languy M 2005. Current and historical factors influencing patterns of species richness and turnover of birds in the Gulf of Guinea highlands. Journal of Biogeography 32: 1371–1384 Gregory RD, Gibbons DW and Donald PF 2003. Bird census and survey techniques. In: Sutherland WJ, Newton I and Green RE (eds) Bird Ecology and Conservation. A Handbook of Techniques. pp 17–55. Oxford University Press, Oxford Legendre P and Legendre L 1998. Numerical Ecology. Developments in Environmental Modelling. Elsevier Science, Amsterdam Mayaux P, Bartholome E, Fritz S and Betward A 2004. A new landcover map of Africa for the year 2000. Journal of Biogeography 31: 861–877 McKay CR and Coulthard ND 2000. The Kilum-Ijim Forests IBA in Cameroon: monitoring biodiversity using birds as indicators. Ostrich 71: 177–180 Orme CDL, Davies RG, Burgess M, Eigenbrod F, Pickup N, Olson VA, Webster AJ, Ding TS, Rasmussen PC, Stattersfield AJ, Bennett PM, Blackburn TM, Gaston KJ and Owens IPF 2005. Global hotspots of species richness are not congruent with endemism or threat. Nature 436: 1016–1019 Reif J, Hoøák D, Sedláèek O, Riegert J, Pešata M, Hrázský Z, Janeèek Š and Storch D 2006. Unusual abundance-range size relationship in an Afromontane bird community: the effect of geographic isolation? Journal of Biogeography 33: 1959–1968 Serle W 1965. A third contribution to the ornithology of the British Cameroons. Ibis 107: 230–246 Smith TB and McNiven D 1993. Preliminary survey of the avifauna of Mount Tchabal Mbabo, west-central Cameroon. Bird Conservation International 3: 13–19 Smith TB, Holder K, Girham D, O’Keffe K, Larison B and Chan Y 2000. Comparative avian phylogeography of the Cameroon and Equatorial Guinea mountains: implications for conservation. Molecular Ecology 9: 1505–1516 Stattersfield AJ, Crosby MJ, Long AJ and Wege DC 1998. Endemic Bird Areas of the World. BirdLife International, Cambridge Storch D, Gaston KJ and Cepák J 2002. Pink landscapes: 1/f spectra of spatial environmental variability and bird community composition. Proceedings of the Royal Society London B 269: 1791–1796 Stuart SN (ed) 1986. Conservation of Cameroon montane forests. International Council for Bird Preservation, Cambridge Stuart SN and Jensen FP 1986. The status and ecology of montane forest bird species in Western Cameroon. In: Stuart SN (ed.) Conservation of Cameroon Montane Forests. pp 38–105. International Council for Bird Preservation, Cambridge ter Braak CJF and Šmilauer P 2002. CANOCO reference manual and CanoDraw for Windows User ’s guide: Software for Canonical Community Ordination (Version 4.5). Microcomputer Power, Ithaca Thomas DHL, Anders S and Penn NJ 2000. Conservation in the community: the Kiluj-Ijim Forest project, Cameroon. Ostrich 71: 157–161 Waltert M, Bobo KS, Sainge NM, Fermon H and Muhlenberg M 2005a. From forest to farmland: habitat effects on afrotropical forest bird diversity. Ecological Applications 15: 1351–1366 Waltert M, Mardiastuti A and Muhlenberg M 2005b. Effects of deforestation and forest modification on understorey birds in Central Sulawesi, Indonesia. Bird Conservation International 15: 257–273 Appendix 1: List of species abbreviations used in Figure 1: Abbr. = Abbreviations; Sp. name = species name Abbr. Sp. name Abbr. Sp. name Abbr. Sp. name alc leu and mon and tep ant cin ant tri apa cin apa jac apa pul bat min bra ban chl nat cin bou cin rei cis bru cis chu col sjo col str cor alb cor cae cor cri cos isa Alcedo leucogaster Andropadus montanus Andropadus tephrolaemus Anthus cinnamoneus Anthus trivialis Apalis cinerea Apalis jacksoni Apalis pulchra Batis minor Bradypterus bangwaensis Chloropeta natalensis Cinnyris bouvieri Cinnyris reichenowi Cisticola brunnescens Cisticola chubbi Columba sjostedti Colius straitus Corvus albus Coracina caesia Corythaeola cristata Cossypha isabellae cos niv cry rei cya ori den fus den goe elm alb emb tah est ast est non eup ard eup cap eus dyb fra squ jyn ruf lag rub lan atr lan mac lin oli mot fla mus adu nes she Cossypha niveicapilla Cryptospiza reichenowi Cyanomitra oritis Dendropicos fuscescens Dendropicos goertae Elminia albiventris Emberiza tahapisi Estrilda astrild Estrilda nonnula Euplectes ardens Euplectes capensis Euschistospiza dybowskii Francolinus squamatus Jynx ruficollis Lagonosticta rubricata Laniarius atroflavus Lanius mackinnoni Linurgus olivaceus Motacilla flava Muscicapa adusta Nesocharis shelleyi ori nig par alb phy tro plo bag plo ban plo ins plo mel pog bil pog cor pse aby pyc bar sax tor ser bur ser moz str sem tau ban tur pel tur tym uro epi vid mac zos sen Oriolus nigripennis Parus albiventris Phylloscopus trochilus Ploceus baglafecht Ploceus bannermani Ploceus insignis Ploceus melanogaster Pogoniulus bilineatus Pogoniulus coryphaeus Pseudoalcippe abyssinica Pycnonotus barbatus Saxicola torquata Serinus burtoni Serinus mozambicus Streptopelia semitorquata Tauraco bannermani Turdus pelios Turtur tympanistria Urolais epichlora Vidua macrorua Zosterops senegalensis Received May 2006, accepted October 2006 Editor: M Louette