Dynamic response of grass cover to rainfall variability: implications for the function and persistence of savanna ecosystems

Savanna grass cover is dynamic and its annual extent resonates with wet season rainfall, as shown by satellite observations of normalized difference vegetation index (NDVI) time series for the Kalahari Transect (KT) in southern Africa. We explore the hydrological significance of the dynamic grass cover by applying a soil moisture model to the water-limited portion of the KT, which spans a north-south gradient in mean wet season rainfall, (r) over bar, from approximately 700 to 300 mm. Satellite-derived tree fractional cover, x(t), is shown to be highly correlated with ground meteorological measurements of (r) over bar (R-2 = 0.94) in this region. By implementing a simple expression for grass growth and decay in the model that factored in only x(t) and near-surface soil moisture, we were able to effectively reproduce the satellite-derived fractional grass cover, x(g), along the transect over a 16-year period (1983-1998). We compared the results from dynamic grass model with those yielded by a static grass cover model in which x, was set to its 16-year average for each simulation. The dynamic quality of the grass was found to be important for reducing tree stress during dry years and for reducing the amount of water that is lost from the overall root zone during the wet years, relative to the static grass case. We find that the dynamic grass cover acts as a buffer against variability in wet season precipitation, and in doing so helps to maximize ecosystem water use. The model results indicate that mixed tree/grass savanna ecosystems are ideally suited to reach a dynamic equilibrium with respect to the use of a fluctuating limiting resource (water) by having functional components that respond to variability in rainfall over long timescales (trees) and short timescales (grasses). (C) 2004 Elsevier Ltd. All rights reserved.