Statistical forecasting of the long rains of East Africa

A statistical forecast model was developed for the long rains (March-to-April) of East Africa. Two conclusions were drawn: the spring predictability barrier limited the lead time for forecasts and the skill of the model was much lower than for the short rains (October-to-December). Several papers have suggested that the long rainy season is much less homogeneous temporally than the short rains. For this reason, separate forecast models were developed for each of the months of the long rains. This dramatically improved forecast skill, although it was still lower than for the short rains. Notably, the greatest predictors were found not in surface features, such as SSTs, but in upper-level circulation. Zonal winds are particularly important. Moreover, the predictors are quite different for each month.