Regionalisation of rainfall models in Eastern Africa using METEOSAT-REAL-TIME-WINDOW-DATA

Two well known rainfall retrieval techniques (GPI,CST) based on satellite infrared data (METEOSAT) have been used to estimate monthly rainfall in the area of Eastern Africa (Kenya) and the adjacent Indian Ocean on the basis of 15 km/spl times/21 km (=315 km/sup 2/) areas during a regional drought period and a regional rainy season. Comparison of the satellite estimates with up to 129 raingauge measurements in this area show ambiguous results in the performance of the techniques, where at both techniques reveal similar results. A regionalisation of this comparison shows an improvement of the respective correlation results and reveals the fact that ambiguous or bad performances of the two used rainfall retrievals can be caused by a wrong compilation of ground truth data sets.