Africa is largely influenced by fires, which play an important ecological role influencing the distribution and structure of grassland, savanna and forest biomes. Here vegetation strongly interacts with climate and other environmental factors, such as herbivory and humans. Fire-enabled Dynamic Global Vegetation Models (DGVMs) display high uncertainty in predicting the distribution of current tropical biomes and the associated transitions, mainly due to the way they represent the main ecological processes and feedbacks related to water and fire. The aim of this study is to evaluate the outcomes of two state-of-the–art DGVMs, LPJ-GUESS and JSBACH, also currently used in two Earth System Models (ESMs), in order to assess which key ecological processes need to be included or improved to represent realistic interactions between vegetation cover, precipitation and fires in sub-Saharan Africa. To this end, we compare models and remote-sensing data, analyzing the relationships between tree and grass cover, mean annual rainfall, average rainfall seasonality and average fire intervals, using generalized linear models, and we compare the patterns of grasslands, savannas, and forests in sub-Saharan Africa. Our analysis suggests that LPJ-GUESS (with a simple fire-model and complex vegetation description) performs well in regions of low precipitation, while in humid and mesic areas the representation of the fire process should probably be improved to obtain more open savannas. JSBACH (with a complex fire-model and a simple vegetation description) can simulate a vegetation-fire feedback that can maintain open savannas at intermediate and high precipitation, although this feedback seems to have stronger effects than observed, while at low precipitation JSBACH needs improvements in the representation of tree-grass competition and drought effects. This comparative process-based analysis permits to highlight the main factors that determine the tropical vegetation distribution in models and observations in sub-Saharan Africa, suggesting possible improvements in DGVMs and, consequently, in ESM simulations for future projections. Given the need to use carbon storage in vegetation as a climate mitigation measure, these models represent a valuable tool to improve our understanding of the sustainability of vegetation carbon pools as a carbon sink and the vulnerability to disturbances such as fire. (Less)
Read full abstract