Committed changes in tropical tree cover under the projected 21st century climate change

International audienceWarming and drought pose a serious threat to tropical forest. Yet the extent of this threat is uncertain, given the lack of methods to evaluate the forest tree cover changes under future climate predicted by complex dynamic vegetation models. Here we develop an empirical approach based on the observed climate space of tropical trees to estimate the maximum potential tropical tree cover (MPTC) in equilibrium with a given climate. We show that compared to present-day (2000-2009) conditions, MPTC will be reduced by 1 to 15% in the tropical band under equilibrium future (2090-2099) climate conditions predicted by 19 IPCC climate models. Tropical forests are found to regress or disappear mainly in the current transition zones between forest and savanna ecosystems. This climate pressure on tropical forests, added to human-caused land use pressure, poses a grand challenge to the sustainability of the world's largest biomass carbon pool. T ropical forest is threatened by global climate changes 1,2 (but see ref. 3) as well as by land use changes induced by increasing food, energy, and development demand 2,4,5. Simulations from Dynamic Global Vegetation Models (DGVMs) run with prescribed climate fields, or coupled with General Climate Models (GCMs) consistently indicate that tropical forest, especially the Amazon forest, is likely to be replaced by savanna or C4 grasses in response to projected climate changes 6,7. However, the strength of the climate induced 'tropical forest dieback' greatly differs among different model simulations 7-9. This spread of the model results reflects different vegetation-climate relationships emerging from the complex equations of DGVM models. Improving the prediction of future climate-induced loss of tropical forest requires a more quantitative understanding of interactions between vegetation and climate 6. In this paper, we quantify the climate envelope of tropical forest by relating tree cover fraction with the observed evapotranspiration (ET). Evapotranspiration through tree crown is one major component of the tropical water balance 10. In a given climate envelop, by assuming the rate of evapotranspiration through unit area of treeless ground as constant b, and that through unit area of tree crown as constant (a 1 b), we are able to relate ET and satellite derived tree coverage (TC) with a linear function: ET 5 a 3 TC 1 b (1), and to estimate the parameters (a 1 b) and b which determine the ET demand for a unit tree crown. Note that here our climate envelops are constrained by annual mean air temperature (T) and annual precipitation (P). Radiation (R), which is one of the important factors affecting ET, is not directly included in the climate envelop (see Discussion). Secondly, on decadal scales when runoff and other water storage terms and loss terms can be neglected, in the maximum scenario all water acquired from precipitation (P) can be used for potential tree growth. The climate maximum potential tree coverage (MPTC) it can support is thus determined by P and parameters (a 1 b) and b estimated from Equation (1) (see Methods). This is called potential fraction, because other non-climate factors or indirect climate factors, such as terrain slope, soil fertility, herbivores, disturbance, may further reduce or enhance tree cover 11 , and because human-caused deforestation and degradation will also yield to future forest loss beyond climate effects. The same parameters of (a 1 b) and b are also applied to predict future potential MPTC in equilibrium with IPCC climate modeled by the end of the 21 st Century (2090-2099) 12. In addition, atmospheric carbon dioxide (CO 2) concentration is also projected to rise by the end of this century, which has a profound implication for plant transpiration through decreasing stomatal conductance and increasing water use efficiency. Hence, to estimate SUBJECT AREAS: BIOGEOCHEMISTRY ECOLOGICAL MODELLING CLIMATE-CHANGE ECOLOGY BIOGEOGRAPH