The environmental drivers of tree cover in savanna-forest mosaic structure in Southeast Asia

Elise M. Pletcher and Naomi B. Schwartz, Department of Geography, University of British Columbia, Vancouver, BC, Canada, Carla Staver, Ecology and Evolutionary Biology, Yale University, New Haven, CT

Background/Question/Methods
Current global biogeographic maps identify large swaths of deciduous dipterocarp forest in SE Asia as seasonally dry tropical forest, despite their structural and functional similarities to savannas, i.e., a grassy understory and open canopy. Subsequently savannas in SE Asia have been left out of global analyses quantifying the determinants and distributions of savannas. This is a problem, since savannas in SE Asia are not only threatened by fragmentation, but also forest-centric management practices. In order to address this urgent need to identify the determinants and distributions of SE Asian savannas, we asked to what degree do forests and savannas represent alternative stable states in SE Asia? And how does spatial scale affect the interpretation of this distribution? To accomplish this, we used high resolution remotely sensed tree cover (30m) to determine if percent cover follows a bimodal distribution both across and within landscapes (0.05°). Secondly, we asked how do fire, climate, and soils influence both average tree cover and landscape-scale heterogeneity (tree cover patchiness)? In order to answer our second question, we used beta and linear regression models, respectively to quantify the relative effects of mean annual precipitation (MAP), fire frequency, precipitation seasonality, soil sand content.
Results/Conclusions
By evaluating distributions of tree cover for bimodality, we found that there is no evidence for bimodality in tree cover across landscapes, however, there is evidence for bimodality within landscapes; 17% of landscapes contained 2 or more modes. This evidence of distinct peaks in high vs. low tree cover demonstrates that discontinuities in high vs. low tree cover emerge and exist at fine scales. This suggests that the processes that maintain distinct forest and savanna boundaries may play out at finer scales in SE Asia, versus areas where savannas exist across large continuous expanses. Furthermore, we found that MAP, seasonality, and fire frequency were all significant predictors of tree cover. While climate does dictate patterns of tree cover, fire plays a major role in shaping vegetation structure across SE Asia. In addition, landscapes were patchier in drier and highly seasonal sub-regions. Put together, these results provide evidence that the same environmental factors important in delineating the distribution of savannas globally, play a significant role in the distribution of both tree cover and landscape heterogeneity across SE Asia. This work provides a framework for how we can include fine-scale savanna distributions in our broad scale, global analyses.