Regionalization of precipitation models in EastAfrica using meteosat data

Abstract

The following article outlines the possibilities, prospects and limitations of estimating precipitation on the basis of high temporal resolution METEOSAT Real Time Window satellite data for the region of East Africa and the adjacent Indian Ocean. We regionalized the study area for both the typical dry season (September 1993) and the rainy season (April 1994) using two distinct precipitation models. These are the statistical METEOSAT Precipitation Index (MPI) model and the physical Convective Stratiform Technique (CST) model. The variation of input values for the respective models were first examined using the threshold temperature of cloud surfaces for the MPI model or the dividing straight line for the classification into either convective or stratiform clouds for the CST model. In addition, we discuss natural conditions of topography, land-water ratio and the interrelation between highland and lowland areas within the study area as well as climatic/ geographic factors of air-mass transport and cloud systems and we evaluate the effects of these variables. By using regression analysis, the relationship between estimated MPI model index values and monthly precipitation is analysed for a maximum of 129 rainfall stations. Subsequent correlation coefficients vary between r=0·87 for September 1993 and r=0·49 for April 1994. For the physical CST model, the correlation coefficients reach values of 0·83 for September 1993 and 0·36 for April 1994. In addition, the study area was segmented on the basis of precipitation/ climatological considerations. By using this procedure, we developed four subregions: 1) WEST, the area around Lake Victoria and the Kenyan highlands; 2) CENTER, the lowlands of northern and eastern Kenya; 3) EAST, the coastal area along the Indian Ocean; and 4) SEA, from the Indian Ocean coastline eastward to longitude 56°E. Although this division leads to a dramatic improvement of the precipitation-index-relation within the WEST region for April 1994, it also results index-relation within the WEST region for April 1994, it also results in a further reduction of the already weak values, as seen in the extremely dry core CENTER region during September 1993. © 1997 The Royal Meteorological Society.