Spatial patterns of the NDVI–rainfall relationship at the seasonal and interannual time scales in an African savanna

Abstract
Climate change is predicted to affect both the mean annual rainfall and its seasonal distribution over the African continent. Understanding their respective influences on primary production, an ecosystem's key feature, is therefore a major challenge for rangeland ecologists. We have investigated the change in intra‐ and interannual Normalized Difference Vegetation Index (NDVI) in relation to rainfall in Hwange National Park, Zimbabwe. Two distinct NDVI time series were built using NOAA/AVHRR data for the period 1982–2002. Long‐term monthly means described the change in seasonal NDVI, whereas annually integrated NDVI related to year‐to‐year fluctuations. The rainfall–NDVI relationship was stronger along the seasonal course [with a lag of 1 month, Kendall's tau (τ) = 0.879] than when studied interannually (τ = 0.476). Principal component analysis (PCA) demonstrated that spatial patterns of the NDVI fluctuations differed when studied interannually or during the seasonal course. Field features such as topography or vegetation composition influenced seasonal NDVI values whereas only rainfall distribution played a role at the interannual time scale. Our results show that rainfall controls on primary production and their mitigation differ between time scales, and these findings bring insights on the future response of savannas to climate change.