The observed and model-simulated response of southern African vegetation to drought

Drought is a frequent disturbance in many regions of the globe and can have a particularly severe impact on vegetation. Over southern Africa, where drought is a regular occurrence, relatively little is known about how quickly vegetation responds to droughts. We characterized the meteorological drought occurrence in southern Africa from 1981 to 2005 and examined the impacts on vegetation productivity, as derived from satellite data. The spatio-temporal extent and severity of droughts were assessed at different timescales (1- to 18-month timescales) using the Standardized Precipitation Evapotranspiration Index (SPEI) and the Standardized Precipitation Index (SPI). Thereafter, we examined the impacts of droughts on southern African vegetation using the Normalized Difference Vegetation Index (NDVI). The results of this study show that southern African vegetation responds differently to drought over the different timescales. During the 1981-2005 period, droughts had extensive impacts over central parts of South Africa, Namibia and western areas of Botswana and the responses of vegetation varied according to season and biome, likely due to the differences in the levels of water needed by vegetation during various growth/phenological phases. In addition, the magnitude of the drought's impact on vegetation is sensitive to the type of drought index that is used to characterize its severity. We further investigated the response of vegetation as simulated by an Earth System Model (the Community Earth Systems Model, CESM) over the same period. The intensity of drought impacts on vegetation is underestimated by CESM, while the timescales at which vegetation responds to droughts are overestimated by it. Different model ensemble members show a substantial spread in the simulation of vegetation response across different biomes and seasons, which may be due to inherent errors and biases in the land component of the model (CLM4.5), or in the atmospheric model and/or the parameterizations. However, we demonstrate that the land component of the CESM (CLM4.5) does provide an adequate representation of the response of vegetation to drought and would provide useful information about the response of vegetation to future drought. If the uncertainties we identified could be reduced, a more realistic simulation of the vegetation response to drought could be realized.