A continental-scale analysis of tree cover in African savannas

We present a continental-scale analysis that explores the processes controlling woody community structure in tropical savannas. We analyse how biotic and abiotic factors interact to promote and modify tree cover, examine alternative ecological hypotheses and quantify disturbance effects using satellite estimates of tree cover. African savannas. Tree cover is represented as a resource-driven potential cover related to rainfall and soil characteristics perturbed by natural and human factors such as fire, cattle grazing, human population and cultivation. Within this framework our approach combines semi-empirical modelling and information theory to identify the best models. Woody community structure across African savannas is best represented by a sigmoidal response of tree cover to mean annual precipitation (MAP), with a dependency on soil texture, which is modified by the separate effects of fire, domestic livestock, human population density and cultivation intensity. This model explains c. 66% of the variance in tree cover and appears consistent across the savanna regions of Africa. The analysis provides a new understanding of the importance and interaction of environmental and disturbance factors that create the broad spatial patterns of tree cover observed in African savannas. Woody cover increases with rainfall, but is modified by disturbances. These 'perturbation' effects depend on MAP regimes: in arid savannas (MAP < 400 mm) they are generally small (< 1% decrease in cover), while in semi-arid and mesic savannas (400-1600 mm), perturbations result in an average 2% (400 mm) to 23% (1600 mm) decrease in cover; fire frequency and human population have more influence than cattle, and cultivation appears, on average, to lead to small increases in woody cover. Wet savannas (1600-2200 mm) are controlled by perturbations that inhibit canopy closure and reduce tree cover by, on average, 24-34%. Full understanding of the processes determining savanna structure requires consideration of resource limitation and disturbance dynamics.