Rainfall characteristics and their implications for rain-fed agriculture: a case study in the Upper Zambezi River Basin

Abstract

This study investigates rainfall characteristics in the Upper Zambezi River Basin and implications for rain-fed agriculture. Seventeen indices describing the character of each rainy season were calculated using a bias-corrected version of TRMM-B42 v6 rainfall estimate for 1998-2010. These were correlated with maize yields obtained by applying a SVATmodel. Finally, a self-organizing map (SOM) was trained to examine multivariate relationships. The results reveal a significant spatio-temporal variability of rainfall indices and yields, with a gradient from north to south. Yields greater than 1 t/ha are found to be only achievable with rainy seasons longer than 160 days. For shorter durations, the interplay of total rainfall, dry spell frequency and maximum dry/wet spell durations determines agricultural success. Using total rainfall alone or wet day frequency as estimators for yields is insufficient. Alternating wet and dry spells affect yields most negatively. The results have significance in the context of agricultural planning under changing climatic conditions and agricultural planning, as well as for the development of forecasting mechanisms.