This study examines contemporaneous and lagged normalized difference vegetation index (NDVI) and precipitation time series in sub-Saharan Africa. Using the 22-year AVHRR-NDVI dataset from NASA's Global Inventory Mapping and Modeling Studies Group (GIMMS)and interpolated gauge/satellite rainfall estimates we quantify statistically the temporal co-evolution of vegetation and rainfall. The study seeks answers to the following questions: I) When and where can NDVI be satisfactorily modeled as time-integrated precipitation? II) Can lagged precipitation be used, in conjunction with other climate indicators, to predict future values of NDVI? And III) Can lagged NDVI be used, in conjunction with other climate indicators, to predict future values of precipitation?

There are several practical applications of this research, related to the authors' work with USAID’s Famine Early Warning System Network. Quantifying the regional/seasonal strength of the NDVI-rainfall relationship can help improve the interpretation of NDVI anomaly maps. NDVI anomalies in regions with strong relationships to rainfall can be monitored closely, since they are presumably tightly linked to variations in hydroclimatic extremes. The lagged NDVI-precipitation, precipitation-NDVI relationships may also be used to improve forecasts of these important variables, enhancing early warning relationships. We will examine the data, looking for empirical justification for early/late rainy season land surface interactions, hopefully demonstrating that vegetation is a ‘slowly evolving boundary’ that can be used to improve climate forecasts.