Oecologia (2001) 128:85–93 DOI 10.1007/s004420000627 Michael Jensen · Anders Michelsen Menassie Gashaw Responses in plant, soil inorganic and microbial nutrient pools to experimental fire, ash and biomass addition in a woodland savanna Received: 9 February 2000 / Accepted: 10 December 2000 / Published online: 23 February 2001 © Springer-Verlag 2001 Abstract In order to investigate the effects of savanna fires on nutrient cycling a field experiment was carried out in an open woodland savanna of southwest Ethiopia. This involved manipulations of fire, fuel load and ash fertilisation in a fully factorial design, and recording of responses in plants, soil inorganic and microbial nutrient pools up to 1 year after the disturbances. As plant biomass nitrogen (N) was only 3.5% of that in topsoil the N loss in a single fire event was relatively small. The microbial N pool size in the topsoil was similar to the N pool size in the aboveground part of the plants. Soil microbial biomass carbon increased slightly 12 days after the low severity fire, but the effect was transient and was not accompanied by an increase in microbial N. Instead, the soil inorganic N concentration increased strongly 1 day after the fire, remained higher up to 3 months after the fire and probably caused the 40% higher grass biomass in burned than unburned plots, and the similar sized increase in grass nitrogen, phosphorus and potassium pools in the following rainy season. In contrast, broad-leaved herbs showed less strong increments in biomass and nutrient pool sizes. Fire interacted with fuel load, as burning of plots with double plant biomass led to reduced microbial biomass, plant nutrient pools and herb (but not grass) biomass. Low-severity-fire nutrient losses appear to be moderate and may be replenished from natural sources. However, in areas with frequent fires and high grass biomass (fuel) loads, or with late fires, nutrient losses could be much larger and non-sustainable to the persistence of the woodland savanna ecosystem. M. Jensen (✉) · A. Michelsen Botanical Institute, University of Copenhagen, Øster Farimagsgade 2 D, 1353 Copenhagen K, Denmark e-mail: micjen@vip.cybercity.dk Fax: +45-35322321 M. Gashaw Biology Department, Addis Ababa University, P.O. Box 3434, Addis Ababa, Ethiopia Keywords Carbon · Nitrogen · Minerals · Fire · Tropical ecosystem Introduction Large parts of African savannas are frequently burned. Today this is mainly caused by deliberate or accidental human activities. Although natural as well as man-made grass fires probably have occurred for hundreds or even thousands of years in Africa, they are much more frequent today (Bourlière and Hadley 1983; Erlich et al. 1995; Lock 1998) because of the steadily increasing human populations. The burning may therefore not maintain a “savanna equilibrium” any more, and the increasing frequency of fires may have long-term consequences for soils, vegetation and animal life. For example, Dumontet et al. (1996) found lower soil nutrient content and microbial biomass in burned than unburned sites 11 years after a Mediterranean forest fire and similar reduction of nutrient pools have been observed in savanna systems (Menaut et al. 1993). Fire may affect the organisms of the savanna directly or indirectly and the consequences appear to depend greatly on the intensity and timing of the fire (DeBano et al. 1998). The severity of a fire is largely a reflection of the fuel load as well as the water content of soil and vegetation at the onset of fire (Menaut et al. 1993). When live and dead biomass is burned, the decomposition pathway is circumvented and nutrients are abruptly released. Some nutrient elements, like nitrogen (N), are partly volatilised whereas others such as phosphorus (P), potassium (K) and various cations to a larger extent remain in the ash, or are lost with particles in the smoke (Menaut et al. 1993). It has been suggested that the soil microbial biomass, through nutrient immobilisation, acts as a buffer or “safety net” preventing the loss of nutrients, which would otherwise be leached out of the ecosystem (Singh et al. 1989, 1991), but this role of soil microbial biomass in fire-prone tropical ecosystems has not been fully investigated. Although nutrient loss and retention has 86 been studied in a number of fire-prone savanna systems at various stages prior to or following fire, very few studies (if any) have used an experimental approach permitting the simultaneous evaluation of the effects of fire on biomass and nutrient partitioning between the major ecosystem pools: plants, microbes and soil organic matter. In order to investigate how biomass load, ash fertilisation and fire affect the plant and microbial biomass and the nutrient partitioning between these compartments and, hence, savanna ecosystem function in longer terms, we examined changes in plant, soil inorganic and microbial nutrient pools during 1 year after full factorial experimental manipulations of fire, fuel load and ash fertilisation in a savanna woodland. Materials and methods Site description Fig. 1 Soil temperature (°C) at 1 and 5 cm soil depth during the experimental fire in savanna woodland. Fire was set in plots with normal grass biomass (480 g m–2) and in plots with double biomass The experiment was carried out in the open woodland savanna zone of southwestern Ethiopia about 12 km south of the provincial capital Gambella, at 8°09′N, 34°34′E. This is at the eastern border of the “Sudanian Region”, a vegetation zone extending across the African (White 1983). The research area is situated at approximately 550 m above sea level (a.s.l.) and receives about 900–1000 mm rain annually, concentrated in the rainy season stretching from late May to October. The dominant tree in the area where the experiment was conducted was Anogeissus leiocarpus, with lower frequencies of Combretum spp., Grewia mollis, Lonchocarpus laxiflorus, Pterocarpus lucens, Strychnos sp., Tamarindus indica and Zizyphus mauritiana. Tree canopies covered about 60% of the ground and the average stem volume was about 95 m3 ha–1. The most frequent shrubs were Allophyllus rubifolius, Dichrostachys cinerea, Fluggea virosa and Harrisonia abyssinica. The grass vegetation consisted of Hyparrhenia confinis var. nudiglumis and smaller amounts of Loudetia arundinacea, Sporobolus pyramidalis and Pennisetum polystachyum. Grass biomass peaked around November at the end of the rainy season when grass covered 88% of the ground and had a biomass of about 4.8 t dry weight (DW) ha–1. Herbs were infrequent and made up 0.15 t ha–1 at this time. For a more detailed description of the vegetation in the area see Jensen and Friis (2001). The major part of the grass and herb biomass burns at the beginning of the following dry season, although the fire severity is probably lower in the woodland than in the more open wooded grassland, which is the most widespread vegetation type in the area (Jensen and Friis 2001). equivalent to the standing (dead) aboveground grass biomass (averaging 480 g m–2) was added to 24 of the 48 plots at day 0. Subsequently 24 plots were burned at day 0, including 12 with added biomass. After the burning, 12 unburned and 12 burned plots had 15 g m–2 fresh ash from recently burned grass plots adjacent to the experimental plots added, corresponding to the average amount of ash remaining after burning an area of the same plot size with normal biomass load. During the fire, which lasted for 2–6 min in individual plots, soil temperature was measured with buried thermocouples. In plots with added biomass (i.e. with a total of 960 g m–2) temperature increased from 25°C to 48°C at 1 cm depth and 32°C at 5 cm depth (Fig. 1), whereas in plots with the “normal” biomass of 480 g m–2, the soil temperature at 1 cm depth reached 37°C. Soil sampling was carried out on 4 occasions: 21 November 1997, 3 December 1997, 15 February 1998 and 25 June 1998, i.e. at days 1, 12, 90 and 210 after the start of the experiment, respectively. Soil (0–5 cm) was collected with an auger, sieved through a 2 mm mesh and stored fresh at 15°C in the field camp. After separation of subsamples for fumigation-extraction, the remaining soil and the plant material was air dried and upon return to Denmark dried in an oven to constant weight. Aboveground biomass of grasses and broad-leaved herbs were sampled on 15 February 1998 in the border zone around the inner 1-m2 part of the 2×2 m2 plots, and again on 20 June 1998 and 15 December 1998 in the inner part of the plots (i.e. on day 90, day 210 and after 1 year). Experimental design and sampling Chemical and physical analysis The effects of three main factors were investigated: fire, biomass addition (before burning) and ash fertilisation (after burning). Biomass was added in order to increase fire temperature and also served to increase the ash input in burned plots. Ash was added in order to study its fertilisation effect on the ecosystem. The experiment was laid out as a full factorial randomised block design with the eight possible factor combinations each with six replicates, i.e. a total of 48 plots. The treatment combinations were applied randomly within each of the six replicate blocks. Each of the 48 plots measured 2×2 m2 but all sampling was carried out in the central 1×1 m2, thus leaving a 0.5-m-wide border zone around the sampling area, except the first sampling of plant biomass. The grass biomass around the 2×2 m2 plots was cleared for fire control in 2-m-wide bands. The experiment was initiated in late November 1997, at the beginning of the dry season. Extra fuel load consisting of grass The water content of the soil was measured gravimetrically, and a part of the dried soil was ashed at 550°C for determination of losson-ignition. A 7-g sample of the fresh soil was fumigated 1–4 days after sampling with CHCl3 for 24 h in a vacuum desiccator to release the carbon (C) and N in the soil microbial biomass (Jenkinson and Powlson 1976; Brookes et al. 1985). After fumigation the soil was extracted for 1 h in 35 ml 2M KCl. Another 7 g of fresh soil was extracted as above but without fumigation in order to recover the soil inorganic nutrients. KCl (2M) was used as extractant because the salt content prevented microbial growth in extracts stored at 15°C during the field work. Analysis of microbial N in extracts using 0.5M K2SO4 gave nearly identical results. The extracts were filtered through Whatman GF-D filters and kept cool until their NH4-N content was analysed with the indophenol method using a Hitachi U-2000 spectrophotometer. NO3-N was analysed with the cadmium reduction method on an Aquatec 5400 87 Table 1 Carbon, nitrogen and phosphorus partitioning between plants and soil pools. Note that the soil pools are from 0 to 5 cm depth, and that plant pools are estimated at peak biomass without taking root biomass into account. Due to high P fixation inorganic and microbial P were not detected. For total soil and plant C, N and P data are end of rainy season means±SE; n=6. For microbial and inorganic soil pools data are ranges of means (n=6) across the year Soil (0–5 cm depth) Carbon (g m–2) Nitrogen (g m–2) Phosphorus (g m–2) C in % of total soil C N in % of total soil N P in % of total soil P Aboveground herbaceous plants Total Microbial Inorganic 1420±124 71±10 10.65±0.9 100 100 100 10.9–41.0 0.36–2.50 – 0.8–2.9 0.5–3.5 – – 0.11–0.36 – – 0.08–0.5 – analyser. The pH of the soil extracts in 2M KCl was measured using a MeterLab 240 electronic pH-meter. PO4-P was analysed spectrophotometrically using the molybdenum blue method. Extracts (10 ml) from the fumigated and the non-fumigated soils were digested in H2SO4 with Se as a catalyst. The digestion of the fumigated samples mineralises the organic fractions of microbial N plus other suspended organic, nutrient holding constituents. Hence, the extractable microbial N content can be calculated by subtracting the N in digested, unfumigated extracts from that in digested, fumigated extracts. The microbial N was calculated by assuming an extractability of 40% (Jonasson et al. 1996). Microbial C was estimated as the content in fumigated samples minus that in unfumigated samples analysed with a Shimadzu TOC-5000A total organic C analyser. The microbial C was estimated assuming an extractability of 45% (Wu et al. 1990). We can not exclude that fire influenced the extractability of the microbial biomass through effects on the soil organic matter (SOM) composition, but as the effects of fire on soil temperature were moderate and no changes in SOM content were detected effects on extractability were probably minute. Samples of soil and plant biomass from each experimental plot were milled and digested in concentrated sulphuric acid with Se as a catalyst (Kedrowski 1983). Total N and P in the digest were measured by the indophenol method and molybdenum blue method respectively, and K was analysed by atomic absorption spectrophotometry. On 15 February, in the middle of the dry season, the plant biomass was very small and was not chemically analysed. Statistical analysis For each sampling period, data were tested statistically with analysis of variance (ANOVA) using the SAS procedure GLM (SAS Institute 1997). The model used was a four-way ANOVA, with fire, biomass, ash and block as main factors, and including all interactions between fire, biomass and ash. In some cases data were transformed in order to meet assumptions of normality and homogeneity of variance. Results Carbon, nitrogen and phosphorus partitioning between plants and soil pools The partitioning of carbon, nitrogen and phosphorus between the aboveground parts of plants and soil inorganic, microbial and total C, N and P pools were estimated based on information on soil bulk density (1.42 g cm–3), plant C content (45%), soil organic matter C content (50%) and nutrient pool sizes in control plots. Microbial 220±64 2.50±0.6 0.63±0.1 15.5 3.5 5.9 C and N made up 0.8–2.9% and 0.5–3.5% of the total soil C and N, respectively, with the lower concentrations in part of the dry season (Table 1). Inorganic N was only 0.08–0.5% of the total soil N, whereas above-ground plant C, N and P were 15.5, 3.5 and 5.9% of the size of the total soil pools, respectively. Effects on soil pH, total soil nutrient and organic matter content The soil was slightly acidic, with an average pH of about 6.2. All main treatment factors influenced soil pH significantly. Fire increased pH significantly (P<0.05) 12 and 210 days after burning and also had a near-significant effect (P<0.10) after 90 days. On average pH rose by 0.2 units across all fire treatments. Addition of biomass caused a slight pH increase after 12 and 210 days (P<0.05 and P<0.001, respectively), with disproportional strong effects in plots which also received ash, shown by the significant interaction between ash and biomass addition after 90 and 210 days (P<0.05 and P<0.01, respectively). Ash addition combined with fire or biomass addition resulted in a near-significant increase in pH by c. 0.2 pH units across treatments at day 1, probably because the pH of samples of fresh ash was high; ash collected in nearby areas had a pH of 11.2, appreciably higher than the soil pH. However, ash addition alone did not affect pH. The soil organic matter content was about 4.0% and there was no significant main effects of the treatments showing that fire caused relatively small C losses from the soil, probably due to the low fire temperature. The concentration of total soil N ranged between 0.75 and 1.5 mg g–1, and that of total P was about 0.15 mg g–1. Generally, there was no effect of any of the main treatments on the concentrations, although there was a significant (P<0.05) reduction in total N 90 days after ash addition mainly in plots which also had extra biomass (fuel load) added, shown by the significant (P<0.05) interaction between ash and biomass addition. 88 Effects on soil inorganic N and microbial C and N concentration The concentration of soil nitrate varied from values close to the concentration of ammonium up to an order of magnitude higher than the soil ammonium concentration, depending on the season (no results presented). The soil inorganic N concentration generally increased continuously from the beginning of the dry season (November, day 0) to the beginning of the rainy season (June, day 210) (Fig. 2). Fire increased inorganic N strongly after 1, 12 and 90 days, but not after 210 days. After 210 days, plots with ash added had slightly but significantly lower levels of inorganic N, and plots with extra biomass added tended to have lower levels. We were unable to detect any inorganic P in the extracts, probably because of poor extractability with KCl due to high P-fixing capacity of the soil. Both microbial C and N varied strongly across the seasons. Microbial C peaked in the beginning of the rainy season in June (at day 210) (Fig. 2). Microbial N was lowest at day 90 in February and highest in June. At day 1 after fire there was a significant reduction in microbial C in plots with extra biomass added, but the effect was confined to the plots that were also burned. In contrast, fire alone led to an increase in microbial C after 12 days, but only in plots which had no extra biomass added, i.e. with relatively low fire temperature. Hence, the short-term effects of fire on the microbial biomass C seem to depend on the fire temperature. Moreover, the effect of fire on microbial C was of short duration and had disappeared after 90 days. The microbial biomass N was more variable and there was no main effect of fire. Ash addition tended to have a negative effect on microbial N at day 1, but the effect disappeared after day 12. Effects on plant biomass production and nutrient pool sizes Fig. 2 Effects of experimental treatments on soil inorganic nitrogen and soil microbial biomass carbon and nitrogen (treatments were: C control, A additional ash added after fire, B additional biomass added before fire, F fire). Treatments were applied alone and in combinations, all with 6 replicates. After the experimental fire, samples were harvested on days 1, 12, 90 and 210. For each day, significant effects of main factors and interactions are indicated (***P<0.001, **P<0.01, *P<0.05; near-significant †P<0.10) The herbaceous biomass was almost completely made up by grasses. Their biomass were about 7–15 times higher than the mass of broad-leaved herbs at the end of the rainy season (Fig. 3, 1 year sampling). The plant biomass in the middle of the dry season was very low due to the combined action of termites and fire. At that time, 90 days after the fire, the sparse new plant biomass production was stimulated in the burned plots. After 210 days, at the beginning of the rainy season, there was a significantly higher grass and total herbaceous biomass in plots subjected to fire as compared to treatments not including fire. After 1 year, when biomass was at its peak, there was still a tendency (P<0.10) towards higher biomass associated with fire reaching about 700 g m–2 compared to c. 500 g m–2 in unburned plots, calculated across treatments. After 210 days the broad-leaved herbaceous biomass responded negatively to treatments involving addition of extra biomass before burning, as did total herbaceous biomass after 90 days, but there was no effect of biomass addition on broad-leaved herbs in non- 89 burned plots, as shown by the significant interaction between fire and biomass. The negative effect on broad-leaved herbs, and lack of effect on grasses by higher fuel load suggest that increased fire intensity is relatively more damaging to broad-leaved herbs than to grasses. As there were no strong changes in the concentration of N, P and K in the plant biomass (M. Gashaw, A. Michelsen and M. Jensen, unpublished work) the nutrient pool sizes (Table 2) largely followed that of the total aboveground plant biomass. Nutrient concentrations were only slightly higher in broad-leaved herbs than in grasses after 210 days, but at peak biomass at the beginning of the dry season the concentration were much higher in broadleaved herbs. Hence, although the (still green) herb biomass was 7–15 times smaller, its nutrient pool sizes herein were only 1.5–6 times smaller than that in the (almost dry and wilted) grasses (Table 2). The grass pools of N, P and K after 210 days were significantly higher in plots subjected to fire and for P the pool size also remained higher after 1 year. The significant or near-significant interaction between fire and biomass addition on N, P and K pool sizes of broad-leaved herbs after 210 days was due to extra fuel load reducing tissue pool sizes in burned plots. The significant or near-significant interaction between fire and biomass addition on the N and K pool sizes of grasses after 1 year was due to the extra biomass added to non-burned plots promoting the growth of grass the following year. Discussion Overview of responses to fire, biomass and ash addition Fig. 3 Effects of experimental treatments on aboveground grass biomass, broad-leaved herb biomass and total herbaceous biomass (treatments were: C control, A additional ash added after fire, B additional biomass added before fire, F fire). Treatments were applied alone and in combinations, all with 6 replicates. After the experimental fire, samples were harvested on day 90 (total herbaceous biomass only), on day 210 and after 1 year. For each day, significant effects of main factors and interactions are indicated (***P<0.001, **P<0.01, *P<0.05; near-significant †P<0.10) This field experiment demonstrates that low severity woodland savanna fires have moderate and transient effects on soil microbial biomass but strong effects on soil nutrient availability and plant performance, such as longerterm increases in inorganic N availability, herb biomass and herb N, P and K pool sizes. Also, fire interacted with plant biomass (fuel) addition because fire temperature increased with increasing fuel load. This reduced the microbial biomass C, nutrient pools in plants and the biomass of broad-leaved herbs (but not the grasses) in burned plots with extra biomass added compared to those with normal fuel load. However, although part of the fire effect on soil and plants may be due to release of nutrients from ash, we found no or weak effects of ash addition only. This suggests that N release from soil organic matter and soil microbes in burned plots supplies more N than that released from ash. However, although we intended to add amounts of ashes similar to that produced by the experimental fire, we can not exclude that a small fraction was lost during the ash application, contributing to the weak responses. Fire, nitrogen loss and nitrogen availability The fate of plant biomass N after fire is critically dependent on the temperature reached during the fire. No N is ***P<0.001, **P<0.01, *P<0.05, near-significant †P<0.10; significant effects of main factors and interactions are indicated for each period of measurement FxB*, F×A×B† F×A×B** F*, F×A×B† F×A×B† F×B† 2429±94 721±158 717±146 161±10 3072±851 810±372 3912±306 561±18 708±188 75±7 4209±1123 463±82 3979±834 599±88 889±18 102±31 3504±343 678±179 3414±502 945±164 817±189 155±18 4960±913 912±321 4466±1000 705±35 913±153 100±37 4772±850 709±51 1323±641 821±33 210±106 94±16 2300±1208 571±108 1858±478 626±96 539±142 94±9 2292±204 417±72 2139±1456 488±241 377±280 140±19 3092±1747 356±207 127±39 27±8 35±12 5±2 350±116 69±32 314±159 40±16 48±15 5±2 508±198 47±19 222±64 212±63 46±11 47±32 404±106 597±413 206±60 164±77 50±19 24±12 425±135 321±178 94±24 35±11 22±6 6±3 158±48 62±24 93±58 49±24 27±19 8±3 178±127 88±48 33±14 44±13 7±2 7±3 57±22 84±24 79±25 39±12 16±4 8±3 116±24 77±24 Pools at 210 days (mg m–2) N in grasses N in broad-leaved herbs P in grasses P in broad-leaved herbs K in grasses K in broad-leaved herbs Pools at 1 year (mg m–2) N in grasses N in broad-leaved herbs P in grasses P in broad-leaved herbs K in grasses K in broad-leaved herbs –A –B +A –B –A +B +A +B –A –B +A –B –A +B +A +B F** F†, B*, F×B* F*** F×B† F*** B*, F×B† Treatment effects Fire No fire Table 2 Effects of experimental treatments on aboveground plant nutrient pools of nitrogen, phosphorus and potassium after 210 days and after 1 year. The eight treatments were: ±fire (F), ±additional ash (A), and ±additional biomass (B), in a fully factorial design with all treatment combinations; data are means±SE, n=6 90 volatilized at temperatures below 200°C, but at higher temperatures proportionally higher amounts are lost and at temperatures above 500°C all of the N is volatilized (DeBano et al. 1998). The amount of N lost is also strongly related to the proportion of biomass combusted, which again increases with the amount of biomass burned (McNaughton et al. 1998). We estimated our experimental fire to be of the “low-severity” type, with maximum temperatures up to 225°C at the soil surface, but only up to about 50°C at 1 cm depth, according to the classification of DeBano et al. (1998). Burning of the vegetation did not show any significant effects on total soil N within the time frame of the experiment and because the plant biomass N was only 3.5% of that in the topsoil (Table 1), the N loss in a single fire event is relatively small. However, annual burning of most of the aboveground herbaceous biomass may eventually reduce the organic matter input to the soil and lead to reduced levels of soil N (DeBano et al. 1998). Total aboveground herbaceous biomass was about 480 g m–2 (Fig. 3) and its N pool was about 3.5 g m–2 (Table 2). During low severity fires 0.88–1.75 g m–2 (8.8–17.5 kg ha–1) of this N (25–50%, according to DeBano et al. 1998) is likely to be volatized and the rest remains in the ash or in incompletely combusted plant parts. In addition to plant uptake, a large part of the N leached from the ash or soil organic matter could potentially be absorbed by the micro-organisms, which contained a similar amount of N as that in the plants (Table 1). However, we did not detect an increase in microbial N. Some inorganic N may be lost in leaching since rain showers can be very intense despite the low total rainfall, and mobile nitrate was the predominant form of N. Lost N could be replenished from N deposition and from nitrogen fixation (Menaut et al. 1993). The tree species Pterocarpus lucens and Acacia sp. and the herbs Thylosema fassoglensis and Indigofera garkeana are potentially able to fix nitrogen symbiotically. The crown cover of the leguminous trees alone was about 30% in the area, so it is not unreasonable to expect that a substantial part of the annual net loss of N at low fire severity level can be replenished by N2 fixation. However, a doubling of the fuel load to 960 g m–2 led to higher soil temperatures. In other widespread vegetation types of the research area, tree cover is lower and the annual grass biomass production is up to 1700 g m–2 (Jensen and Friis 2001). In such areas the high fuel load would lead to higher net N depletion in the resulting warmer fires. We did not observe any significant increase in soil NH4-N level in burned plots, but NH4-N was apparently quickly nitrified, because NO3 increased significantly following the fire. The increase in inorganic N up to 90 days following fire (Fig. 2) is consistent with previous reports by Singh et al. (1991), Franco-Vizcaíno and Sosa-Ramirez (1997), Ross et al. (1997) and PrietoFernández et al. (1998), and shows that the N sink source both in soil microbes and plants was insufficient to fully prevent potential fire-induced losses of N from the top- 91 soil. The observed larger amounts of nitrate N compared with ammonium N is uncommon in savannas, and fireinduced increase in nitrification has not been observed previously (Abbadie et al. 1992). The dominant grass Hyparrhenia confinis, hence, did not seem to suppress nitrification, although this has often been found for other Hyparrhenia species (Lee and Stewart 1978). Responses in plant functional groups, their biomass and nutrient pools to fire Although grass biomass increased strongly in response to fire, there was no such effect on the broad-leaved herbs (Fig. 3). Apparently, grasses were in a better position to benefit from the increased inorganic N availability, perhaps because of higher capacity for nitrate reductase activity in grasses than slower-growing broadleaved herbs (Lee and Stewart 1978), and perhaps because grass meristems are not damaged as much as those of herbs by fire. Blair (1997) also found a biomass increase, mainly of grasses, in response to infrequent fires in prairie vegetation. In plots burned with extra biomass added and, hence, higher fire intensity, broad-leaved herb biomass was much smaller after 210 days than in the lower fire temperature plots. This suggests that herb regeneration in savannas may be inhibited by high fire temperatures, most likely because of killing of growing points, and perhaps also because of increased seed mortality. The increase in grass N and P pools largely followed that of the biomass in burned plots, whereas the K pool increased slightly more than explained by the increase in grass biomass after 210 days, suggesting that additional K was absorbed after fire (Table 2). However, after 1 year this surplus uptake ceased to be significant. In the broad-leaved herbs, there was no significant response in N, P or K pools to fire treatment, and pool sizes followed that of the biomass. Both the N and K pools in the broadleaved herbs were reduced 210 days after addition of biomass, most strongly in burned plots (i.e. at high fire temperature), and the P pool also showed a tendency to decline. Hence, high fire temperature reduced both the biomass and the nutrient accumulation of herbs, but not of grasses. These patterns probably reflect the fact that N and P are available in near-limiting amounts for the grasses which tolerate the frequent fires, and that availability keeps pace with the growth, whereas K, and probably other cations as well, are in surplus, partly supplied by ash deposits. Indeed, Blair (1997) found that plant tissue N declined in prairie grasslands following repeated fires. The role of microbial biomass in fire-prone savanna ecosystems Although estimates of microbial C and N pools in savanna ecosystems are relatively few, our data on the propor- tion of microbial to total soil C and N are consistent with previously published estimates from similar and different systems (Wardle 1992; Dumontet et al. 1996; Jonasson et al. 1996). Even though soil temperatures did not increase dramatically in this experiment, the stronger fire in plots with biomass added seemed to kill soil microbes in the topsoil, as also seen by Prieto-Fernández et al. (1998). In these plots the microbial biomass C was lower than in plots which were burned without extra biomass added. Generally, microbial activity in the upper horizon may be impeded in the long run by the direct effect of strong, regular fires, the drying-out of the bare surface during the dry season and the deprival of organic substrates consumed by the fires (Henrot and Robertson 1994; Diaz-Ravina et al. 1996; Dumontet et al. 1996; Ross et al. 1997). Deeper in the soil, conditions are more favourable, with root decay providing an organic matter input, and with fluctuations in temperature and water conditions being less dramatic and hence less inhibiting for microbes. The contrasting effect of relatively low and moderately high fire temperature suggests, firstly, that stimulation of microbial growth at low fire intensity may result from increased nutrient availability, as observed by Singh et al. (1991), and secondly that the effect of fire on the microbial community is strongly dependent upon the fuel load, i.e. fire temperature. The relatively short-lived effect of fire on microbial C and the lack of significant main effects on microbial N does not lend experimental support to the suggestion by Singh et al. (1989, 1991) that microbes provide a very important nutrient sink in the period between fire and vegetation regrowth. With low intensity fires there seems to be a transient stimulatory effect on the microbial community, in which the observed increase in microbial biomass may serve to retain some nutrients within the ecosystem. The much stronger fluctuation in microbial N during the dry season than that caused by the treatments suggests that other factors than fire such as soil water contents, availability of labile C and grazing may have a strong effect on microbial biomass and its uptake or release of N. Effects of fire on soil pH and phosphorus The observed increase in soil pH following fire is consistent with earlier observations (e.g. Strømgaard 1991; DeBano et al. 1998) and can be ascribed to the high content of cations in the ash. The effect was still seen after 7 months and elsewhere has lasted for 1–2 years (Strømgaard 1991; DeBano et al. 1998). The increase by about 0.2 pH units was relatively small, as compared to the ash pH of 11.3. The soil pH influences the P availability; often with a decrease in availability with increasing alkalinity, but we were unable to detect available or microbial P in the soil with the extractant used, most likely because P was strongly fixed in the soil. Burning of a dry tropical savanna in India increased inorganic P 92 by 35% (Singh et al. 1991). In our study the increase in the grass pool of P in burned plots was similar to the increase in the plant biomass suggesting that more P became available and was taken up by the grasses following the fire, similar to the results of Ross et al. (1997). Conclusion The experimental fire early in the dry season was of a low-severity type and did not result in dramatic losses of nutrients from the ecosystem. The N losses were no greater than they may be replenished through wet and dry deposition and N fixation by leguminous plants. The regrowth of grasses was promoted by fire, probably as a result of increased availability of N for plants, whereas broad-leaved herbs seemed less capable of benefiting from this, possibly because fire caused greater damage to their meristems. Although there was a short-term increase in the microbial biomass C in response to low temperature fire, the microbial biomass did not immobilise significantly higher amounts of N in burned than in unburned plots, and seasonal changes in microbial biomass were greater than those induced by treatments. As plant growth and N uptake also was low during the dry season there was a surplus of inorganic N during the first 210 days after the fire in the burned plots. In wooded grassland with larger grass biomass, the higher fuel load may result in higher fire temperatures increasing the fire severity. In such cases it is less likely that N losses can be replaced from natural sources and, consequently, the nutrient status will decline with time. Although we found an increase in N availability in the first year after fire, long-term studies have shown a decline in soil nutrients as a result of regular fires (Menaut et al. 1993; King et al. 1997). In the Ethiopian savanna this might also be the case if the fire takes place later in the dry season when the water content of soil and vegetation is lower (M. Jensen, A. Michelsen and M. Andersson, unpublished work). If fires are to be set this should take place very early in the dry season in order to minimise nutrient losses. Within the 3-year study period of this project deliberate, but unstructured onset of early fires appeared to be the general practice in the area, although later fires may also occur. Acknowledgements This work was funded by the Danish Council for Development Research as part of the Fire in Tropical Ecosystems programme. We wish to thank Dr. Ib Friis, Dr. Sebsebe Demisew, Dr. Sven Jonasson, Michael Andersson, Malgorzata Sylvester and Esben Nielsen for their assistance in the work, The National Herbarium, Addis Ababa University for institutional help during the field work, and three anonymous referees for constructive comments on the manuscript. References Abbadie L, Mariotti A, Menaut J-C (1992) Independence of savanna grasses from soil organic matter for their nitrogen supply. Ecology 73:608–613 Blair JM (1997) Fire, N availability, and plant response in grasslands: A test of the transient maxima hypothesis. Ecology 78:2359–2368 Bourlière F, Hadley M (1983) Present-day savannas: an overview. In: Bourlière F (ed) Ecosystems of the world 13. Tropical savannas. Elsevier, Amsterdam, pp 1–17 Brookes PC, Landman A, Pruden G, Jenkinson DS (1985) Chloroform fumigation and the release of soil nitrogen: a rapid direct extraction method to measure microbial biomass nitrogen in the soil. Soil Biol Biochem 17:837–842 DeBano LF, Neary DG, Ffolliot PF (1998) Fire's effects on ecosystems. Wiley, New York Díaz-Raviña M, Prieto A, Bååth E (1996) Bacterial activity in a forest soil after soil heating and organic amendments measured by the thymidine and leucine incorporation techniques. Soil Biol Biochem 28:419–426 Dumontet S, Dinel H, Scopa A, Mazzatura A, Saracino A (1996) Post-fire soil microbial biomass and nutrient content of a pine forest soil from a dunal Mediterranean environment. Soil Biol Biochem 28:1467–1475 Erlich D, Grégoire J-M, Eva H, Janodet E, Koffi B (1995) Fire detection, land-cover characterisation and burnt area estimation in the savannah-forest transition zone of central Africa. Assembling a multiscale dataset. Institute for Remote Sensing Applications, Joint Research Centre, European Commission, Brussels Franco-Vizcaíno E, Sossa-Ramirez J (1997) Soil properties and nutrient relations in burned and unburned Mediterranean-climate shrublands of Baja California, Mexico. Acta Oecol 18:503–517 Henrot J, Robertson P (1994) Vegetation removal in two soils of the humid tropics: effect on microbial biomass. Soil Biol Biochem 26:111–116 Jenkinson DS, Powlson DS (1976) The effect of biocidal treatments on metabolism in soil-V. A method for measuring soil biomass. Soil Biol Biochem 8:209–213 Jensen M, Friis I (2001) Fire regimes, floristics, diversity, life forms and biomass in wooded grassland, woodland and dry forest at Gambella, SW Ethiopia. K Dan Vidensk Selsk Biol Skr Jonasson S, Michelsen A, Schmidt I K, Nielsen EV, Callaghan TV (1996) Microbial biomass C, N and P in two arctic soils and the responses to addition of NPK fertiliser and carbon: implications for plant nutrient uptake. Oecologia 106:507–515 Kedrowski RA (1983) Extraction and analysis of nitrogen, phosphorus and carbon fractions in plant material. J Plant Nutr 6:989–1011 King J, Moutsinga JB, Doufulon G (1997) Conversion of anthropogenic savanna to production forest through fire-protection of the forest-savanna edge in Gabon, Central Africa. For Ecol Manage 94:223–247 Lee JA, Stewart GR (1978) Ecological aspects of nitrogen assimilation. Adv Bot Res 6:1–43 Lock M (1998) Aspects of fire in tropical African vegetation. In: Huxley CR, Lock LM, Cutler DF (eds) Chorology, taxonomy and ecology of the floras of Africa and Madagaskar. Royal Botanic Gardens, Kew, pp 181–189 McNaughton SJ, Stronach NRH, Georgiadis NJ (1998) Combustion in natural fires and global emissions budgets. Ecol Appl 8:464–468 Menaut J-C, Abbadie L, Vitousek PM (1993) Nutrient and organic matter dynamics in tropical ecosystems. In: Crutzen PJ, Goldhammer JG (eds) Fire in the environment: the ecological, atmospheric, and climatic importance of fires. Wiley, Chichester, pp 215–231 Prieto-Fernández A, Acea MJ, Carballas T (1998) Soil microbial and extractable C and N after wildfire. Biol Fert Soils, 27:132–142 Ross DJ, Speir TW, Tate KR, Feltham CW (1997) Burning in a New Zealand snow-tussock grassland: effects on soil microbial biomass and nitrogen and phosphorus availability. N Z J Ecol 21:63–71 93 SAS Institute (1997) SAS/STAT users guide, release 6.12. SAS Institute, Cary Singh JS, Raghubanshi AS, Singh RS, Srivastava SC (1989) Microbial biomass act as a source of plant nutrients in dry tropical forest and savanna. Nature 338:499–500 Singh RS, Srivastava SC, Raghubanshi AS, Singh JS, Singh SP (1991) Microbial C, N and P in dry tropical savanna: effects of burning and grazing. J Appl Ecol 28:869–878 Strømgaard P (1992) Immediate and long-term effects of fire and ash-fertilisation on a Zambian miombo woodland soil. Agric Ecosyst Environ 41:19–37 Wardle DA (1992) A comparative assessment of factors which influence microbial biomass carbon and nitrogen levels in soil. Biol Rev 67:321–358 White F (1983) The vegetation of Africa. A descriptive memoir to accompany the UNESCO/AETFAT/UNSO vegetation map of Africa. Unesco, Paris. Nat Resources Res 20:1–356 Wu J, Joergensen RG, Pommerening B, Chaussod R, Brookes PC (1990) Measurement of soil microbial biomass C by fumigation-extraction, an automated procedure. Soil Biol Biochem 22:1167–1169