Predicting the regional onset of the rainy season in West Africa

摘要

Particularly in regions, where precipitation is limited to a few months per year only, reliable determination of the onset of the rainy season and the start of the sowing time is of crucial importance to sustainable food production. Especially since the mid-1980s, an increasing delay of onset dates in the Volta basin of West Africa has been suspected by local farmers. To investigate this speculation and develop a reliable tool to find the optimal sowing date, the onset of the rainy season in the region was analysed by means of several statistical techniques. The focus was put on the region of the Volta basin in Ghana and Burkina Faso. In a first step, two fuzzy logic based definitions of the onset were developed using daily precipitation data and additionally accounting for important plant physiological aspects. In this context, only one definition is potentially useful to judge whether the current onset of the rainy season has already begun. In a second step, methods for predicting the onset date of the ongoing season were investigated. In this context, the detection of onset controlling variables plays a major role. Two strategies are investigated and evaluated for the prediction of the monsoon's onset dates: 1) A combination of regionalized synoptic rainfall data by means of principal component analysis (PCA) in a spatial mode and linear discriminant analysis in order to detect reliable prediction parameters and allow for a classification of the rainy season, dry season, and the onset of the rainy season using current rainfall data. 2) Linear regression models were generated to estimate the onset of the rainy season for certain regions using the onset dates of regions, where the onset has already begun. To enhance the predictability, optimized definition parameterisation in the field of both strategies was applied. Copyright (c) 2007 Royal Meteorological Society.