火灾作为热带湿润森林选择动力的研究新进展----中国科学院西双版纳热带植物园.pdf

Oecologia (2010) 164:841–849 DOI 10.1007/s00442-010-1764-4 CONSERVATION ECOLOGY - ORIGINAL PAPER Fire as a selective force in a Bornean tropical everwet forest J. W. Ferry Slik · Floris C. Breman · Caroline Bernard · Marloes van Beek · Charles H. Cannon · Karl A. O. Eichhorn · Kade Sidiyasa Received: 23 February 2010 / Accepted: 17 August 2010 / Published online: 2 September 2010 © Springer-Verlag 2010 Abstract Tree species rarely exposed to burning, like in everwet tropical forests, are unlikely to be Wre adapted. Therefore, one could hypothesize that these species are aVected equally by burning and that tree abundance changes are linked solely to Wre behavior. Alternatively, if species do react diVerentially to burning, abundance changes should be linked to tree habitat preference and morphology. Using tree inventories from old-growth and adjacent burned Bornean forest in combination with a database on tree morphology and habitat preference, we test these alternative hypotheses by (1) determining whether species speciWc abundance changes after Wre diVer signiWcantly from equal change, and (2) whether observed abundance changes are linked to species morphology and habitat preference. We found that of 196 species tested, 125 species showed an abundance change signiWcantly diVerent from that expected under our null model of equal change. These abundance changes were signiWcantly linked to both tree morphology and habitat preference. Abundance declines were associated with slope or ridge preference, thin barks, and limited seed dormancy. Abundance increases were associated with high light preference, small adult stature, light wood, large leaves, small seeds and long seed dormancy. While species habitat preference and morphology explained observed abundance increases well, abundance declines were only weakly associated with them (R2 » 0.09). This suggests that most tree mortality was random and everwet tropical tree species are poorly Wre adapted. As Wre frequencies are increasing in the everwet tropics, this might eventually result in permanently altered species compositions and even species extinctions. Communicated by John Silander. Keywords Fire mortality · Functional traits · Habitat niche · Life history · Tropical trees Electronic supplementary material The online version of this article (doi:10.1007/s00442-010-1764-4) contains supplementary material, which is available to authorized users. J. W. F. Slik (&) · C. H. Cannon Key Laboratory of Tropical Forest Ecology, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun 666303, Yunnan, China e-mail: ferryslik@hotmail.com F. C. Breman Royal Museum for Central Africa, Royal Belgian Institute for Natural Sciences, Tervuren, Belgium C. Bernard Nationaal Herbarium Nederland, Leiden University, Leiden, The Netherlands M. van Beek Forest Ecology and Forest Management Group, Centre for Ecosystem Studies, Wageningen University, Wageningen, The Netherlands C. H. Cannon Department of Biological Sciences, Texas Tech University, Lubbock, TX, USA K. A. O. Eichhorn Eichhorn Ecologie, Zeist, The Netherlands K. Sidiyasa Wanariset-Samboja Herbarium, FORDA, km38, Samboja, East Kalimantan, Indonesia 123 842 Introduction Several studies have shown that undisturbed old-growth forests are diYcult to burn due to their high air humidity and high moisture content of soil and litter, even after prolonged droughts (Uhl et al. 1988; Uhl and KauVman 1990; Ray et al. 2005). For this reason, forest Wres in the everwet tropics used to be a rare phenomenon, mostly occurring locally and at irregular intervals of several 100 years (Goldammer 1989; Cochrane 2003; Power et al. 2008). This has changed since the second half of the last century, when large tracts of old growth forest were opened up for logging, mining, infrastructure and agriculture (Bradshaw et al. 2009). Most tropical forests remaining today are either disturbed or fragmented and have become increasingly susceptible to droughts and Wre (Uhl and KauVman 1990; Cochrane et al. 1999; Siegert et al. 2001). Forest Wres are now a common, large-scale and almost continuous phenomenon in most everwet tropical regions of the world. This is worrisome since it is questionable whether everwet tropical tree species that have rarely been exposed to Wre in the past are adapted to cope with this new Wre disturbance regime. At the same time, it oVers an opportunity to investigate if forest Wres aVect tropical tree species abundances diVerentially, and if they do, how this is related to habitat preferences or plant functional traits. Insight into this might help us understand how tropical forests can evolve from one type into another under diVerent scenarios of future climatic change and associated Wre regimes. Understanding the impact of Wres on everwet tropical forests has become an important research topic during the last decades. Many studies have shown that Wres can lead to dramatic and lasting changes in forest structure and tree species composition (Cochrane 2003; Barlow and Peres 2008; Slik et al. 2008). Generally, Wre aVects small diameter trees most, with tree mortality rates nearing 100% [diameter at breast height (dbh) <10 cm] (Uhl and Buschbacher 1985; Isichei et al. 1986; Woods 1989; Slik and Eichhorn 2003). This is linked to the strong relationship between tree diameter and bark thickness with thinner stems having thinner barks to protect them against the heat radiation generated by Wre (Uhl and KauVman 1990; Barlow et al. 2003; van Nieuwstadt and Sheil 2005; Michaletz and Johnson 2007). Presence of buttresses, resin and smooth barks can also increase tree Wre mortality due to accumulation of litter fuel load between buttresses, Xammability of resins, and a thinner isolating air boundary layer during Wre exposure resulting in higher bark temperatures, respectively (Barlow et al. 2003). Next to these morphological factors, non-random spatial patterns of Wre occurrence and intensity linked to fuel loads (Uhl and KauVman 1990; Barlow et al. 2003), topography and soil/litter moisture (Uhl and KauVman 1990; Slik and Eichhorn 2003; Nepstad 123 Oecologia (2010) 164:841–849 et al. 2004; Ray et al. 2005), and canopy openness (Uhl and Buschbacher 1985; Woods 1989) have been shown to aVect tree Wre mortality rates diVerentially throughout the landscape. Tree species recovery and community compositional changes after Wre are also strongly linked to survival and colonization patterns. Several studies have shown that resprouting and emergence from the soil seed bank can be important mechanisms by which tree species survive or even proWt from Wres (Riswan and Yusuf 1986; Isichei et al. 1986; de Rouw and van Oers 1988; Marod et al. 2002; Fensham et al. 2003; Vesk and Westoby 2004; Baker et al. 2008), while good dispersal ability in combination with early reproductive maturity form additional strategies enabling tree species to cope with a regular Wre regime (Uhl et al. 1981; HoVmann et al. 2003; Muller et al. 2007). Fire studies in the everwet tropics have mainly focused on general trends in tree Wre mortality and survival, indiscriminate of species identities, or when they did focus on species-speciWc survival rates, included only a few species. Therefore, it remains unclear how these general Wndings translate to species-speciWc abundance changes, and thus if and how Wres work as a selective force in everwet tropical forests. If Wre aVects all species equal, then Wre damage will be largely driven by spatial patterns in Wre behavior. However, if species are aVected diVerentially, Wre damage will result in more complicated spatial patterns based on interactions between Wre behavior, tree species habitat preferences and tree morphology. To resolve this issue, we tested whether (1) changes in species abundance after Wre are dependent on species identity, and (2) if so whether these changes are related to tree morphological traits and habitat preferences. Materials and methods Study site The Sungai Wain Protection Forest is located ca. 15 km north-west of the city of Balikpapan, East Kalimantan, Indonesia (Fig. 1). Two-thirds of this forest (ca. 8,000 ha) was burned in April 1998 after a prolonged drought that was associated with a severe El Nino Southern Oscillation (ENSO) event, while a ca. 4,000-ha core area was saved from Wre by man-made Wre breaks. The Wres were low intensity surface Wres that slowly moved through the area mainly fuelled by the thin litter layer. These Wres caused signiWcant tree mortality especially in the small diameter classes (dbh < 10 cm) (Slik and Eichhorn 2003; van Nieuwstadt and Sheil 2005). The area receives a mean of »2,400 mm precipitation annually, with rainfall exceeding evaporation in all months (Walsh 1996). Soils are relatively Oecologia (2010) 164:841–849 843 Detecting signiWcant species speciWc abundance changes after Wre Fig. 1 Map indicating the location of the research area sites in Borneo (arrow in left map). Two 450-ha plots were established in unburned (0) and adjacent burned (1) forest (right map) poor and sandy and the topography covers an elevation gradient of 10–120 m above sea level. The area includes freshwater swamps, coastal tidal areas, periodically inundated river valleys and low hill ridges. The vegetation composition and forest structure is typical for lowland coastal forests in eastern Borneo, with forests within a 30-km radius of the research area showing high Xoristic and structural similarity (Slik et al. 2003; Eichhorn 2006; Slik et al. 2009; Slik et al. 2010). Tree abundance data Tree abundance data were taken from a large Weld inventory carried out 2 years after the 1998 Wres in the Sungai Wain Protection Forest (Eichhorn 2006). The inventory was conducted in two north–nouth oriented 1.5 £ 3.0 km (450-ha) squares, one in burned old-growth and one in adjacent unburned old-growth forest (Fig. 1). Each square contained 80 randomly distributed plots of 10 £ 20 m (0.02 ha). Within each plot, all live trees taller than 1.3 m were identiWed [voucher specimens deposited in the Nationaal Herbarium Nederland, Leiden University, The Netherlands ([NHN-L)] and their dbh measured. A total of 28,388 individuals were found belonging to 695 (morpho)-species, representing about one-third of the known tree species in the Sungai Wain Protection Forest (http:\\www.nationaalherbarium.nl/sungaiwain). From these, we excluded species with less than 10 individuals and morpho-species (499 species excluded of which 235 were morpho-species, 41 of which had more than 10 individuals) resulting in a total of 196 species used in the analyses (Appendix 1, see supplementary material). Unfortunately, no pre-Wre data was available for the burned forest square; however, given the relatively uniform landscape, close proximity of the two plots, our use of only common species and limited Xoristic variability observed within a 30-km radius of our study site (Slik et al. 2003; Eichhorn 2006), we assume that both plots had comparable species composition and abundances. For the 196 species, we found 10,723 (76.6%) individuals in the unburned versus 3,283 (23.4%) in the burned plots. We used this overall abundance diVerence between unburned and burned forest to calculate expected abundances for each species in the burned forest. For each species, we tested whether observed abundance in burned forest was signiWcantly lower or higher than the expected abundance using a one-tailed Fisher’s Exact test. Unlike the Chi-square test, the Fisher’s Exact test can be reliably applied with categories containing less than Wve individuals. Even when abundance change after Wre would be random for all species, one would expect to Wnd 10 species with a signiWcantly larger or lower abundance change than expected using a 0.05 probability (p) value (0.05 £ 196 species). To determine whether our observed number of species with a signiWcant abundance deviation diVered from the expected number of 10, we again performed a Fisher’s Exact test. Ranking species according to abundance change We used three abundance change measures: (1) ‘absolute’ which ranked species from low to high according to diVerence between observed and expected number of individuals in burned forest; (2) ‘relative’ which ranked species from low to high according to percentage change between observed and expected number of individuals in burned forest; and (3) ‘combined’ which ranked species from low to high based on the average of their absolute and relative abundance rank. The absolute measure gives a direct indication of the number of individuals lost or gained per species and therefore gives more weight to species with high pre-Wre abundance since more abundant species can lose more individuals. The relative measure, on the other hand, gives more weight to less abundant species since loss or gain of an individual in a rare species counts for more than a single loss or gain in a common species. The combined abundance measure should give a more balanced indication of abundance change since it equalizes the overweighting of common and rare species. Species habitat preferences and morphological attributes Species habitat preferences were determined using a large database of tree inventories including 26,604 individual trees belonging to 1,469 taxa, of which 989 were identiWed to species level (including all 196 species used in our analyses). These inventories took place between 1997 and 2005 in small 10 £ 10 m (0.01-ha) plots and were spread over a large portion of East Kalimantan Province. They included old-growth forests, logged forests, burned forests, logged 123 844 and burned forests, and logged and thinned forests within an altitude range of 10–1,200 m. The following habitat variables were determined for each species in this data set: (1) the median canopy openness (determined from hemispherical canopy photographs taken in the center of each 0.01-ha plot) under which individuals were found to grow; (2) the median slope (determined with a clinometer for each 0.01-ha plot) on which individuals were found to grow; (3) the topographical class [alluvial (swamps, river valleys, Xood plains); hillside (sloped area between valleys and hill tops); ridge (hill tops)], based on the percentage of individuals found growing in each topographic class; and (4) soil texture (clay, silt, Wne sand, sand, coarse sand), based on the percentage of individuals found growing in each soil texture class. Morphological variables and elevation preference for each species were determined from a minimum of ten herbarium specimens per species in the NHN-L, and included: (1) minimum reproductive diameter (determined by ranking the dbh values scored from herbarium labels of fertile specimens from small to large, plotting them in a scattergraph (Y = dbh; X = rank), Wtting an exponential function through the graph, and scoring the Y-intercept of this function as the minimum reproductive dbh); (2) maximum adult diameter (determined in the same way as minimum reproductive diameter, but this time with the dbh values ranked from high to low); (3) maximum adult height (determined in the same way as maximum diameter, but then using estimated tree heights); (4) mean elevation above sea level; (5) maximum elevation above sea level; (6) leaf size based on the mean of length multiplied by width values; (7) leaf shape based on the mean length divided by width values; (8) fruit size based on the mean maximum fruit length; (9) seed size based on the mean maximum seed diameter; and (10) Xower size based on the mean maximum Xower diameter. We also determined mean oven-dry wood density of each species based on reported values in the literature (Oey 1990; Suzuki 1999; Osunkoya et al. 2007), or when a species had no recorded wood density, from the genus average (sensu Slik 2006). Median population bark thickness in old growth forest was determined for each species using the individuals found in the unburned forest plot. This was based on a diameter–bark thickness relationship measured in the Sungai Wain Protection Forest by van Nieuwstadt (2002): bark thickness = 10[¡1.21 + 0.669LOG10(dbh)]. van Nieuwstadt (2002) also presented a species sprouting frequency equation (percentage sprouting = 37.1–38.9 £ oven dry wood density), but since this relationship depends completely on the wood density values already included in our study, we did not use it here. As an indication of seed dormancy capacity of the tree species in our study, we used the maximum observed species germination times recorded by Ng (1991). If a tree species was not treated by Ng (1991), 123 Oecologia (2010) 164:841–849 we used the average maximum germination time of the genus to which it belonged. If a genus was not present in Ng (1991), we used the average maximum germination time of the phylogenetically closest related genus for which data was available. Variable selection for Wnal analyses Our selection of morphological and habitat variables was based on the assumption that they were related to tree Wre mortality, survival and/or colonization patterns. However, for our Wnal analyses, we wanted to exclude redundant variables, i.e. variables that were strongly correlated to other variables (correlation coeYcient >0.7). To do that, we performed a Pearson’s Rank correlation test for all pair-wise combinations of variables (Appendix 2, see supplementary material). This showed that the following variables were highly correlated: (1) ‘Alluvial’ and ‘Ridge’, (2) all ‘Soil Texture’ variables, (3) ‘Reproductive dbh’, ‘Maximum dbh’ and ‘Maximum height’, (4) ‘Mean elevation’ and ‘Maximum elevation’, and (5) ‘Fruit size’ and ‘Seed size’. Therefore, we excluded the following variables from the Wnal analyses: (1) ‘Ridge’, (2) ‘Clay’, ‘Fine sand’, ‘Sand’ and ‘coarse sand’, (3) ‘Reproductive dbh’ and ‘Maximum height’, (4) ‘Maximum elevation’, and (5) ‘Fruit size’. This left 14 variables for Wnal analysis. Explaining species abundance changes We used Akaike’s information criterion (AIC) to select the variables that best explained the observed species abundance changes using the Spatial Analysis in Macroecology (SAM) software developed by Rangel et al. (2006). This software calculates multiple regression models for all possible combinations of variables, i.e. for our 14 Wnal variables there were 16,383 possible combinations, and then ranks these models according to their AIC score. AIC selects models based on a compromise between least number of variables included, highest possible explanatory power and lowest residual variation. This analysis was performed to explain ‘Absolute abundance change’, ‘Relative abundance change’ and the ‘Combined abundance change’ for: (1) all species, (2) species that had declined in abundance after Wre, and (3) species that had increased in abundance after Wre. We did this to be able to identify not only the processes that drive the overall abundance change of the tree population after Wre but also the speciWc ones associated with tree species abundance declines and increases. Testing for phylogenetic autocorrelation An important aspect of species-speciWc analyses is the fact that due to their evolutionary descent, species cannot Oecologia (2010) 164:841–849 Results Of the 196 tree species tested, 150 had lower than expected abundance while 46 had higher than expected abundance after Wre (Fig. 2; Appendix 1, see supplementary material). For 125 of the 196 species, the observed abundance in burned forest was signiWcantly larger or lower than expected, which was signiWcantly higher (one-tailed Fisher’s Exact test p < 0.0001) than the expected value of 10 species if abundance change across species would have been equal (using a probability value of 0.05). Tree morphology and habitat preference had a signiWcant eVect on species abundance changes after Wre (Table 1). The three best models (for absolute, relative and combined abundance change) explained between 26.6 and 34.4% of data variance when all species were analyzed simultaneously. This analysis showed that a mix of tree species morphological characteristics and habitat preferences inXuenced 3 2.5 Log 10 expected abundance automatically be assumed to be independent sample units. Since independence of sample units is an important prerequisite of most statistical tests (dependence leads to inXated degrees of freedom in statistical testing and thus p values that cannot be trusted), it is important to check whether the diVerence between the observed and predicted values of models (residuals) are independent of species phylogenetic relationships. To test for this, we used a phylogenetic tree that was resolved up to genus level for most Bornean tree genera (Slik et al. 2009) as input phylogeny in the program PHYLOMATIC (http://www.phylodiversity.net/phylomatic/), and then pruned it down to contain only the genera present in our dataset (Appendix 3, see supplementary material). Using the BLADJ function in PHYLOCOM (Webb et al. 2008) in combination with estimated family ages (Wikstrom et al. 2001), we dated this phylogeny. We then calculated the phylogenetic distance in millions of years (‘phydist’ in PHYLOCOM) between all possible pairs of species in this phylogeny. The resulting matrix was entered in a Principal Coordinate Analysis (PCO) to summarize phylogenetic patterns along the main PCO axes. The Wrst PCO axis separated species from ‘primitive’ to ‘advanced’, while the second axis diVerentiated between the Eurosid1 and 2 clades. We used the coordinate of each species on the Wrst two PCO axes as the values indicating their phylogenetic position in relation to all other species. This data was used to calculate Moran’s I values for our model residuals in the program SAM (with X containing PCO axis 1 coordinate and Y containing the PCO axis 2 coordinate of each species). In this way, we were able to detect presence or absence of residual phylogenetic autocorrelation in our Wnal models. 845 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 Log 10 observed abundance Fig. 2 Plot of observed versus expected species abundance in burned forest. Line indicates no change; points below line are species that increased in abundance; points above line are species that decreased in abundance after Wre their abundance change after Wre. Tree species with heavy wood, large seeds, short seed dormancy, preferring hillsides and/or closed canopy were aVected most adversely by Wre, while tree species with light wood, small seeds, long seed dormancy, preferring alluvial sites and/or an open canopy were less aVected or increased in abundance after Wre. There was high agreement among selected variables for the three abundance change measures, with the only exception being seed dormancy, which was only important in explaining absolute abundance changes. Amount of variance explained by the models decreased markedly (between 6.2 and 11.3%) when only species were analyzed that had declined after Wre. The most important variables responsible for abundances change in declining species after Wre were: topographic habitat preference, bark thickness, and seed dormancy. Species with a preference for alluvial sites, having a population structure consisting of mostly large diameter trees with thick barks and/or long seed dormancy, showed limited abundance decline after Wre, while tree species growing on hillsides and ridges, having a population structure consisting of mainly small trees with thin barks and/or short seed dormancy, declined the most. Again, there was high agreement among selected variables for the three abundance change measures, with seed dormancy as an exception. The models were able to explain high levels of data variance (between 32.0 and 45.9%) when species that increased after Wre were analyzed. The most important variables were canopy openness preference, maximum adult diameter, 123 846 Oecologia (2010) 164:841–849 Table 1 Modeling results (R2-adjusted, F-ratio, P value, n, and presence of phylogenetic autocorrelation of residuals) for the best model selected for: (1) all species, (2) species that declined in abundance after Wre, and (3) species that increased in abundance after Wre. Importance of variables (% presence in models with a delta-AIC score <2) and All species R2 adjust direction of correlation with abundance change are also added. Variables selected in the single best model (lowest AIC-score) are underlined. Total number of models with delta-AIC score ·2 indicated in the last table row Decline species Increase species Absolute Relative Combined Absolute Relative Combined Absolute Relative Combined 31.7 26.6 34.4 11.3 6.2 9.4 32.0 45.9 41.8 F ratio 15.6 14.6 20.0 6.6 4.8 7.6 7.1 12.5 8.2 P <0.001 <0.001 <0.001 <0.001 0.009 <0.001 <0.001 <0.001 <0.001 n 196 196 196 150 150 150 46 46 46 Phylo-autocor. No No Limited No No No No No No 100.0 (+) 100.0 (+) 100.0 (+) 46.7 (+) 0.0 21.7 (+) 90.0 (+) 100.0 (+) 100.0 (+) 5.3 (+) 0.0 0.0 0.0 15.4 (¡) 4.3 (¡) 30.0 (¡) 0.0 0.0 60.9 (+) 50.0 (¡) 0.0 22.2 (¡) 56.5 (¡) 0.0 0.0 0.0 0.0 90.0 (¡) 0.0 0.0 100.0 (¡) Variable importance Canopy openness Slope Alluvial pref. 100.0 (+) 62.5 (+) 100.0 (+) 100.0 (+) Hillside pref. 100.0 (¡) 0.0 100.0 (¡) 7.7 (+) 20.0 (¡) 7.7 (+) 76.9 (¡) Silt pref. 94.7 (¡) 31.6 (+) 26.7 (+) 15.4 (¡) 4.3 (+) Dbh max. 26.3 (¡) 18.8 (¡) 23.1 (¡) 6.7 (¡) 0.0 0.0 Elevation 5.3 (+) 0.0 0.0 (+) 0.0 7.7 (¡) 4.3 (¡) 0.0 16.7 (¡) 22.2 (¡) 100.0 (¡) 23.1 (+) 0.0 7.7 (¡) 4.3 (¡) 0.0 66.7 (¡) 11.1 (¡) 20.0 (¡) 7.7 (+) 4.3 (¡) 60.0 (+) 33.3 (+) 44.4 (+) Wood density Leaf size 100.0 (¡) 5.3 (+) 100.0 (¡) 25.0 (+) 50.0 (¡) Leaf shape 5.3 (¡) 0.0 7.7 (¡) 0.0 7.7 (+) 17.4 (+) 0.0 Seed size 43.8 (¡) 18.8 (¡) 61.5 (¡) 13.3 (¡) 0.0 0.0 20.0 (¡) 0.0 100.0 (¡) 0.0 55.6 (¡) Flower size 68.4 (¡) 10.5 (¡) 15.4 (¡) 13.3 (¡) 7.7 (¡) 13.0 (¡) 0.0 0.0 0.0 Bark thickness 10.5 (+) 6.3 (+) 7.7 (+) 100.0 (+) 100.0 (+) 100.0 (+) 0.0 0.0 0.0 Seed dormancy 89.5 (+) 6.3 (¡) 7.7 (+) 100.0 (+) 0.0 13.0 (+) 0.0 (+) 0.0 0.0 Model nr. 19 16 13 15 13 23 10 6 9 wood density, leaf size and seed size. Species preferring an open canopy, having a small adult stature, light weighted wood, large leaves and/or small seeds increased strongly after Wre. Models for absolute, relative and combined abundance change largely agreed on the responsible variables with the exception of wood density, which was only important for explaining changes in relative abundance after Wre. We detected only one case of phylogenetic autocorrelation in the model residuals (Appendix 4, see supplementary material). However, this was limited to one phylogenetic distance class with a very low value of autocorrelation (Moran’s I of ¡0.065, on a scale running from ¡1 to +1), and thus did not aVect model statistics seriously. Discussion Our analysis shows that Wre in an everwet tropical forest can aVect abundances of co-existing species diVerentially. This means that, even after a single low intensity surface Wre, species diversity and composition patterns will be 123 severely altered. This change was related both to species habitat preferences and species morphological traits. Topographic position, which is strongly linked to soil moisture content during drought (Daws et al. 2002; Gibbons and Newbery 2002) and the associated Wre risk and intensity (Slik and Eichhorn 2003), was a major factor aVecting tree Wre mortality, recruitment and survival, with tree species preferring hillsides and ridges generally showing higher abundance declines than tree species preferring alluvial sites. It was already known for our study area that lowest tree mortality occurred in river valleys and swamps, resulting in a patchily connected network of forest that was relatively unaVected by Wre (van Nieuwstadt 2002; Slik and Eichhorn 2003; Eichhorn 2006), but our study shows that this pattern directly translates into changes in species composition of burned forest due to species-speciWc habitat preferences. This means that the previously suggested importance of the remaining unaVected alluvial forest patches in the recovery of burned forest might be limited due to the fact that the species found there might not be able to perform well on hillsides and ridges. Even if they do manage to disperse and establish successfully onto hillsides Oecologia (2010) 164:841–849 and ridges, the species composition will remain diVerent from that of the pre-Wre forest. Species functional traits were also important determinants of tree species abundance changes after Wre. Our analysis shows that median population bark thickness and seed dormancy capacity are two main determinants of species abundance declines after Wre. Since bark thickness was calculated based on a correlation with stem dbh, this indicates that species characterized by a large standing stock of small diameter trees were aVected most negatively by Wre. These are generally late successional understorey species and canopy tree species with large standing stocks of saplings. These two species groups usually possess low seed dormancy capacity, which might explain the observed negative correlation between seed dormancy length and species abundance decline. Another reason why seed dormancy length was correlated negatively with species abundance decline might be related to the fact that species with seed dormancy posses a seed bank in the soil. Since our data were collected 2 years after the Wre, this might already have included individuals that germinated from this soil seed bank thus resulting in less severe abundance declines after Wre for species possessing seed dormancy. Species that increased after Wre were characterized by high light preference, small stature, low wood density, large leaves and/or small seeds. These characteristics are typical for early successional tree species (Swaine and Whitmore 1988), although wood density has also been shown to correlate negatively with species sprouting capacity (van Nieuwstadt 2002), suggesting that increased sprouting capacity may also explain the observed abundance increases. The burned forest was indeed strongly dominated by several pioneer tree species, which is a common pattern found in burned forests across the tropics (Isichei et al. 1986; Barlow and Peres 2008; Slik et al. 2008). Many of the species that increased in abundance after Wre were very rare or even absent in the unburned forest, meaning that they either dispersed into the burned forest over distances of several kilometers or were already present in the soil seed bank. A study by van Nieuwstadt (2002) did Wnd many of the pioneer tree species germinating from soil samples collected in the unburned forest a few months after the 1998 Wres (thus before any serious seed input from the burned forest could have altered the seed composition of his soil samples), suggesting that the soil seed bank may have played a role in the establishment of these species. This would be in accordance with pioneer establishment in other tropical forests (Garwood 1989). It is noteworthy that our models were only able to explain »9% of data variance in tree species that declined in abundance after Wre suggesting that most of the Wre mortality was random and that selective species extinctions will only occur after repeated burning. However, the amount of 847 explained data variance might increase by inclusion of additional factors which were shown to be important in the Neotropics but for which we did not have enough data (Barlow et al. 2003). These are the presence of resin and buttresses, sprouting capacity, bark structure, mycorrhizal associations, and distribution of Wre fuel load in the forest. Additionally, data on two main Wre survival characteristics was lacking, namely species-speciWc sprouting capacity and bark thickness. Additional research will have to be carried out to measure these factors in the Weld to determine how this will aVect the results of our analyses. The changes in species-speciWc abundances and associated shifts in species composition in everwet forests after Wre contrast strongly with observations from the more seasonal forests of continental southeast Asia. In these seasonal forests, Wre had a limited impact on species diversity and composition and most species seemed to be Wre adapted, possessing relatively high sprouting rates and thick barks (Baker et al. 2008). Traditionally, the Xoristic transition from everwet to seasonal tropical forests has been ascribed to progressively increasing dry season length and resulting drought stress (Hely et al. 2006; ter Steege et al. 2006; Slik et al. 2009). However, since dry season length also aVects Wre risk and frequency (Van der Werf et al. 2008), part of this Xoristic shift might be caused by the selective pressure of forest Wres. This would mirror the transition from tropical seasonal forest to savannah where Wre has already been shown to be a main driving factor of species compositional and forest structural transition, next to the precipitation regime (HoVmann et al. 2003; Banfai and Bowman 2005; Beerling and Osborne 2006). In principal, it is possible to tease apart the eVects of precipitation and Wre on tree species composition and forest structure since droughts aVect tree composition diVerently than Wres. Droughts mostly aVect large-statured upper canopy and emergent evergreen trees with relatively low wood densities (Slik 2004; van Nieuwstadt and Sheil 2005). Therefore, decreasing amounts of rainfall and longer dry season length should lead to forests dominated by relatively short-statured heavy wooded and/or deciduous tree species, which corresponds to observations along increasing dry season length gradients in the Neotropics (Malhi et al. 2006). Fire on the other hand mostly aVects small trees with thin barks, heavy wood, low sprouting capacity, large seeds with short seed dormancy and late reproductive maturity (Barlow et al. 2003; Slik and Eichhorn 2003; van Nieuwstadt and Sheil 2005). Therefore, forests that are regularly burned should become dominated by trees with thick barks, high sprouting capacity, well-developed seed dormancy, light wood, small seeds and/or early reproductive age, such as are found in the seasonal forests of mainland Southeast Asia (Baker et al. 2008). The combined eVect of drought and Wre could result in a mosaic of forest types 123 848 diVering in species composition and structure depending on regularity and intensity of Wres and droughts. This corresponds to the co-occurrence and patchy distribution of diVerent forest types (seasonal dry evergreen, mixed deciduous, deciduous dipterocarp and many intermediates) in seasonal southeast Asia (Zhu et al. 2006; Baker et al. 2008). Conclusions Fire did aVect species abundances diVerentially, and this change was linked to species habitat preferences and morphology, although only »9% of species abundance declines could be explained in this manner. This indicates that most tree mortality due to Wre was random, although it cannot be ruled out that inclusion of more species attributes (like sprouting capacity and bark thickness) could increase the amount of explained abundance decline. Based on our current analysis, it seems likely that several Wres are needed to lead to selective species extinctions. On the other hand, our analysis very clearly identiWed species characteristics associated with abundance increases after Wre. These were mostly morphological traits associated with early successional life history strategies caused by the strong increase of pioneer species after Wre. Overall, our analysis indicates that everwet tropical tree species are not well adapted to Wre. The increase in Wre frequency in recent decades might therefore pose a serious threat to these forests, with decreasing tree diversity and species compositions starting to resemble those of Asia’s seasonal tropical forests. Acknowledgments We would like to thank the Wanariset-Samboja herbarium staV for the accurate identiWcation of the collected plant material. We also thank the motivated and enthusiastic local Weld crews from Samboja that helped us. We are grateful to the TropenbosBalikpapan Project for their logistical support. The Sungai Wain Management Body is thanked for granting us permission to work in the Sungai Wain area. We thank LIPI for providing us with the necessary research permits. This research was made possible thanks to Wnancial support by the Dutch Foundation for the Advancement of Tropical Research (NWO-WOTRO), Alberta Mennega Stichting and Treub Foundation. All experiments comply with the current laws of Indonesia. References Baker PJ, Bunyavejchewin S, Robinson AP (2008) The impacts of large-scale, low-intensity Wres on the forests of continental SouthEast Asia. Int J Wildl Fire 17:782–792 Banfai DS, Bowman DMJS (2005) Dynamics of a savanna-forest mosaic in the Australian monsoon tropics inferred from stand structures and historical aerial photography. Aust J Bot 53:185–194 Barlow J, Peres CA (2008) Fire-mediated dieback and compositional cascade in an Amazonian forest. Philos Trans R Soc Lond B 363:1787–1794 Barlow J, Lagan BO, Peres CA (2003) Morphological correlates of Wre-induced tree mortality in a central Amazonian forest. J Trop Ecol 19:291–299 123 Oecologia (2010) 164:841–849 Beerling DJ, Osborne CP (2006) The origin of the savanna biome. Glob Chang Biol 12:2023–2031 Bradshaw CJA, Sodhi NS, Brook BW (2009) Tropical turmoil: a biodiversity tragedy in progress. Front Ecol Env 7:79–87 Cochrane MA (2003) Fire science for rainforests. Nature 421:913–919 Cochrane MA, Alencar A, Schultze MD, Souza CM Jr, Nepstad DC, Lefebvre P, Davidson EA (1999) Positive feedbacks in the Wre dynamics of closed canopy tropical forest. Science 284:1832– 1835 Daws ML, Mulins CE, Burslem DFRP, Paton SR, Dalling JW (2002) Topographic position aVects the water regime in a semi-deciduous tropical forest in Panama. Plant Soil 238:79–90 de Rouw A, van Oers C (1988) Seeds in a rainforest soil and their relation to shifting cultivation in the Ivory Coast. Weed Res 28:373– 381 Eichhorn KAO (2006) Plant diversity after rain-forest Wres in Borneo. Blumea Suppl 18:1–140 Fensham RJ, Fairfax RJ, Butler DW, Bowman MJS (2003) EVects of Wre and drought in a tropical eucalypt savanna colonized by rain forest. J Biogeogr 30:1405–1414 Garwood NC (1989) Tropical soil seed banks: a review. In: Leck MA, Parker VT, Simpson RL (eds) Ecology of soil seed banks. Academic Press, San Diego, pp 149–202 Gibbons JM, Newbery DM (2002) Drought avoidance and the eVect of local topography on trees in the understorey of Bornean lowland rain forest. Plant Ecol 164:1–18 Goldammer JG (1989) Natural rain forest Wres in eastern Borneo during the Pleistocene and Holocene. Naturwissenschaften 76:518–520 Hely C, Bremond L, Alleaume S, Smith B, Sykes MT, Guiot J (2006) Sensitivity of African biomes to changes in precipitation regime. Glob Ecol Biogeogr 15:258–270 HoVmann WA, Orthen B, Vargas do Nascimento PK (2003) Comparative Wre ecology of tropical savanna and forest trees. Funct Ecol 17:720–726 Isichei AO, Ekeleme F, Imoh BA (1986) Changes in a secondary forest in southwestern Nigeria following a ground Wre. J Trop Ecol 2:249–256 Malhi Y, Wood D, Baker TR, Wright J, Phillips OL, Cochrane T, Meir P, Chave J, Almeida S, Arroyo L, Higuchi N, Killeen TJ, Laurance SG, Laurance WF, Lewis SL, Monteagudo A, Neill DA, Nunez-Vargas P, Pitman NCA, Quesada CA, Salomao R, Silva JNM, Torres-Lezama A, Terborgh J, Vasquez-Martinez R, Vinceti B (2006) The regional variation in aboveground life biomass in old-growth Amazonian forests. Glob Chang Biol 12:1107–1138 Marod D, Kutintara U, Tanaka H, Nakashizuka T (2002) The eVects of drought and Wre on seed and seedling dynamics in a tropical seasonal forest in Thailand. Plant Ecol 161:41–57 Michaletz ST, Johnson EA (2007) How forest Wres kill trees: a review of the fundamental biophysical processes. Scand J For Res 22:500–515 Muller SC, Overbeck GE, Pfadenhauer J, Pillar VD (2007) Plant functional types of woody species related to Wre disturbance in forestgrassland ecotones. Plant Ecol 189:1–14 Nepstad D, Lefebre P, Lopes da Silva U, Tomasella J, Schlesinger P, Solorzano L, Moutinho P, Ray D, Benito JG (2004) Amazon drought and its implications for forest Xammability and tree growth: a basin-wide analysis. Glob Chang Biol 10:704–717 Ng FSP (1991) Manual of forest fruits, seeds and seedlings. Malayan Forest Record No. 34, vol 1 and 2. Forest Research Institute Malaysia, Kuala Lumpur Oey DS (1990) Berat jenis dari jenis-jenis kayu Indonesia dan pengertian beratnya kayu untuk keperluan praktek (SpeciWc gravity of Indonesian woods and its signiWcance for practical use). Departemen Kehutanan Pengumuman nr. 13. Pusat Penelitian dan Pengembangan Hasil Hutan, Bogor Oecologia (2010) 164:841–849 Osunkoya OO, Sheng TK, Mahmud NA, Damit N (2007) Variation in wood density, water content, stem growth and mortality among twenty-seven tree species in a tropical rain forest on Borneo island. Aust Ecol 32:191–201 Power MJ, Marlon J, Ortiz N, Bartlein PJ, Harrison SP, Mayle FE, Ballouche A, Bradshaw RHW, Carcaillet C, Cordova C, Mooney S, Moreno PI, Prentice IC, Thonicke K, Tinner W, Whitlock C, Zhang Y, Zhao Y, Ali AA, Anderson SS, Beer R, Behling H, Briles C, Brown KJ, Brunelle A, Bush M, Camill P, Chu GQ, Clark J, Colombaroli D, Connor S, Daniau AL, Daniels M, Dodson J, Doughty E, Edwards ME, Finsinger W, Foster D, Frechette J, Gaillard MJ, Gaving DG, Gobet E, Haberle S, Hallett DJ, Higuera P, Hope G, Horn S, Inoue J, Kaltenrieder P, Kennedy L, Kong ZC, Larsen C, Long CJ, Lynch J, Lynch EA, McGlone M, Meeks S, Mensing S, Meyer G, Minckley T, Mohr J, Nelson DM, New J, Newnham R, Noti R, Oswald W, Pierce J, Richard PJH, Rowe C, Sanchez-Goni MF, Shuman BN, Takahara H, Toney J, Turney C, Urrego-Sanchez DH, Umbanhowar C, Vandergoes M, Vanniere B, Vescovi N, Walsh M, Wang X, Williams N, Wilmhurst J, Zhang JH (2008) Changes in Wre regimes since the last Glacial Maximum: an assessment based on a global synthesis and analysis of charcoal data. Clim Dyn 30:887–907 Rangel TFLVB, Diniz-Filho JAF, Bini LM (2006) Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Glob Ecol Biog 15:321–327 Ray D, Nepstad D, Moutinho P (2005) Micrometeorological and canopy controls of Wre susceptibility in a forested Amazon landscape. Ecol Appl 15:1664–1678 Riswan S, Yusuf R (1986) EVects of forest Wres on trees in the lowland dipterocarp forest of East Kalimantan, Indonesia. Forest regeneration in Southeast Asia (ed. S. S. Tjitrosomo). SEAMEO-BIOTROP, Bogor Siegert F, Ruecker G, Hinrichs A, HoVmann AA (2001) Increased damage from Wres in logged forests during droughts caused by El Nino. Nature 414:437–440 Slik JWF (2004) El Nino droughts and their eVect on tree species composition and diversity in tropical rain forests. Oecologia 141:114–120 Slik JWF (2006) Estimating species-speciWc wood density from the genus average in Indonesian trees. J Trop Ecol 22:481–482 Slik JWF, Eichhorn KAO (2003) Fire survival of lowland tropical rain forest trees in relation to stem diameter and topographic position. Oecologia 137:446–455 Slik JWF, Poulsen AD, Ashton PS, Cannon CH, Eichhorn KAO, Kartawinata K, Lanniari I, Nagamasu H, Nakagawa M, van Nieuwstadt MGL, Payne J, Purwaningsih Saridan A, Sidiyasa K, Verburg RW, Webb CO, Wilkie P (2003) A Xoristic analysis of the lowland dipterocarp forests of Borneo. J Biog 30:1517–1531 Slik JWF, Bernard CS, Beek M, van Breman FC, Eichhorn KAO (2008) Tree diversity, composition, forest structure and aboveground biomass dynamics after single and repeated Wre in a Bornean rain forest. Oecologia 158:579–588 Slik JWF, Raes N, Aiba SI, Bearley FQ, Cannon CH, Meijaard E, Nagamasu H, Nilus R, Paoli G, Poulsen AD, Sheil D, Suzuki E, Valkenburg JLCH, van Webb CO, Wilkie P, WulVraat S (2009) 849 Environmental correlates for tropical tree diversity and distribution patterns in Borneo. Divers Distrib 15:523–532 Slik JWF, Aiba SI, Brearley FQ, Cannon CH, Forshed O, Kitayama K, Nagamasu H, Nilus R, Payne J, Paoli G, Poulsen AD, Raes N, Sheil D, Sidiyasa K, Suzuki E, van Valkenburg JLCH (2010) Environmental correlates of tree biomass, basal area, wood speciWc gravity and stem density gradients in Borneo’s tropical forests. Glob Ecol Biogeogr 19:50–60 Suzuki E (1999) Diversity in speciWc gravity and water content of wood among Bornean tropical rainforest trees. Ecol Res 14:211– 224 Swaine MD, Whitmore TC (1988) On the deWnition of ecological species groups in tropical rain forests. Vegetatio 75:81–86 Ter Steege H, Pitman NCA, Phillips OL, Chave J, Sabatier D, Duque A, Molino JF, Prevost MF, Spichiger R, Castellanos H, von Hildebrand P, Vasquez R (2006) Continental scale patterns of canopy tree composition and function across Amazonia. Nature 443:444–447 Uhl C, Buschbacher R (1985) A disturbing synergism between cattle ranch burning and selective tree harvesting in the Eastern Amazon. Biotropica 17:265–268 Uhl C, KauVman JB (1990) Deforestation, Wre susceptibility, and potential tree responses to Wre in the eastern Amazon. Ecology 71:437–449 Uhl C, Clark K, Clark H, Murphy P (1981) Early plant succession after cutting and burning in the upper Rio Negro region of the Amazon basin. J Ecol 69:631–649 Uhl C, KauVman JB, Cummings DL (1988) Fire in the Venezuelan Amazon 2: environmental conditions necessary for forest Wres in the evergreen rainforest of Venezuela. Oikos 53:176–184 Van der Werf GR, Randerson JT, Giglio L, Gobron N, Dolman AJ (2008) Climate controls on the variability of Wres in the tropics and subtropics. Glob Biogeogr Cycles 22:GB3028 van Nieuwstadt MGL (2002) Trail by Wre: postWre development of a tropical dipterocarp forest. PhD thesis, Utrecht University, Utrecht van Nieuwstadt MGL, Sheil D (2005) Drought, Wre and tree survival in a Bornean rain forest, East Kalimantan. Indones J Ecol 93:191–201 Vesk PA, Westoby M (2004) Sprouting ability across diverse disturbances and vegetation types worldwide. J Ecol 92:310–320 Walsh RPD (1996) Drought frequency changes in Sabah and adjacent parts of northern Borneo since the late nineteenth century and possible implications for tropical rain forest dynamics. J Trop Ecol 12:385–407 Webb CO, Ackerly DD, Kembel SW (2008) Phylocom: software for the analysis of phylogenetic community structure and trait evolution. Bioinformatics 24:2098–2100 Wikstrom N, Savolainen V, Chase MW (2001) Evolution of angiosperms: calibrating the family tree. Proc R Soc Lond B 268:2211–2220 Woods P (1989) EVects of logging, drought, and Wre on structure and composition of tropical forests in Sabah, Malaysia. Biotropica 21:290–298 Zhu H, Cao M, Hu H (2006) Geological history, Xora, and vegetation of Xishuangbanna, southern Yunnan, China. Biotropica 38:310–317 123