Probabilistic 8-day rainfall estimates for the Sanaga basin, Cameroon

Abstract

A method has been developed to produce 8-day forecasts or estimates of future rainfall over the 132 000 km² Sanaga basin in Cameroon. The estimates provide the input to a real-time flow management system which determines optimum reservoir releases to achieve expected power demands, and thereby tries to make the most effective possible use of the water resources of the basin. Attempting to forecast deterministically for up to 8 days ahead was not thought to be practicable, and a probabilistic approach was taken instead. This means that each forecast is associated with a reliability or probability of exceedance. In the initial technique, based on an analysis of historic data, the forecasts are determined by the date only. An additional forecasting method was also developed which includes the current position of the FIT (the local name for the ITCZ) as a causative factor but still maintains the forecasts on a probabilistic basis. This uses the variation of the FIT from its usual position for the time of year to determine whether the forecast rainfall should be greater or less than the standard forecast for that date, and so includes some ability to take account of the variability of rainfall. The forecasting system is believed to be a novel approach to a problem which has not been tackled before. While far from providing a complete solution to the problem of rainfall forecasting in real-time basin management, it does illustrate an approach that can be attempted in the absence of reliable deterministic techniques.