Modelling the future range and productivity of African tree species. Perspectives and limits

There remains a lack of information on the future of plant species in many parts of Africa under the threads of climate change with the exception of the mountainous areas. Models are valuables tools to examine this problem because they permit to extrapolate basic information as simple as species occurrence coming from a restricted number of localities to the entire continent. Niche-based models, like logistic regression or MaxEnt, easily allow fitting empirical relationships between environmental variables related to climate and possibly to soil properties. They produce probabilities of occurrence for the present with good accuracy (calibration phase). Projections for the future are made by switching the explanatory data set with future conditions. These models however are limited by the fact that it is difficult to integrate physiological response to increasing CO2 air concentration. Dynamic vegetation models (DVMs) are process-based models that simulate plant environment (soil water, light intensity at various heights, etc.) and plant physiology (transpiration, CO2 fixation, photosynthesis, respiration, carbon allocation, etc.) from climate variables, soil properties, and elevation. They could be run at various scales, from global to regional or even local scale, and simulate the growth of plant functional types (PFTs), of biological affinity groups (BAGs) or of species. A model like CARAIB is able to simulate PFTs and BAGs growth (occurrence and productivity) with rather good accuracy for Western Europe. For the future, the simulations confirm that the physiological effect of CO2 concentration change is dramatic but not easily foreseeable because it depends on overall fertility of the sites (Dury et al., iForest – Biogeosciences and Forestry, 4:82-99, 2011). From this conclusion, spatial and temporal variations of fertility would have to be introduced in modelling studies to reach more operational conclusions. Questions arising about the future of ecosystem services in tropical countries highlight particular plant species (BIOSERF project funded by the Belgian Science Policy: Sustainability of tropical forest biodiversity and services under climate and human pressure). In this study, we model a set of 11 selected African tree species including several Congolese species with logistic regression, MaxEnt and CARAIB models. The two niche-based-models rather properly simulate the ranges obtained with the alpha-hull polygon method. CARAIB correctly simulates the range of the evergreen species but not of the deciduous trees. We examine how physiological knowledge could be use to improve the model. IN particular, we conclude that bud dormancy breaking representation has to be upgraded in the model because this process is likely to control the range of the species. It should act in combination with the specific bioclimatic constants controlling the hydrological and thermal stress and the germination. Additionally, we examine the evolution of the ranges at the 2050 horizon using one of the most recent socio-economic scenarios