Seasonal forecasting potential of maize yield in southern Africa

Large parts of southern Africa is subject to climatic extremes that often result in poor crop yields. A strong dependence upon agriculture, high population growth rates, and unstable economic conditions compound the sensitivity to climate variability. Of the crops grown in southern Africa, maize is the most important, accounting for up to 70% of total caloric intake but with a high sensitivity to water availability, it has a substantial year to year variability.

Much progress towards seasonal forecasting of rainfall in southern Africa has been made. Such knowledge may serve as one input to predict maize yield several months in advance. Rainfall is, however, only one of the climatic determinants of maize yield. In this paper we present a new seasonal forecasting scheme whereby water stress, the primary determinant of maize yield in southern Africa, are predicted directly. Instead of training rainfall on sea surface temperatures (SSTs) and atmospheric circulation predictors, we use maize water stress generated from a maize model forced by observed climate time series from 1961-1994. The maize model generates a time series of maize water stress variability as it would have occurred if climate where the only factor determining maize growth over the period 1961-1994. This model generated maize water stress time series, free of confounding socio-economically related factors, is then trained on SSTs and atmospheric circulation predictors from which a seasonal forecasting scheme is developed. For SST data the UKMO GISST data set is used while NCEP data provides the atmospheric circulation data. Since maize water stress serves as a multivariate climate index, the seasonal forecasting skill of this scheme is enhanced over schemes which aim to predict rainfall alone.