SATELLITE IDENTIFICATION OF RAIN DAYS OVER THE UPPER NILE RIVER BASINUSING AN OPTIMUM INFRARED RAIN NO-RAIN THRESHOLD TEMPERATURE MODEL

As part of the U.S. Agency for International Development/National Ocea nic and Atmospheric Administration project to develop an improved moni toring, forecasting, and simulation system for the river Nile, the Rem ote Sensing Unit of the University of Bristol has been investigating a nd developing satellite infrared techniques for small-scale estimation of rainfall over the region of the upper Nile basin. In this paper, t he need for variable IR rain/no-rain temperature thresholds as a basis for reliable satellite identification of rain areas over small scales is explained, and the spatially and temporally variable nature of opt imum IR rain/no-rain threshold temperatures is examined. Meteosat IR d ata covering a period of 17 months have been analyzed along with daily rain gauge reports for calibration and validation. Analyses have been carried out on a monthly basis. Optimum IR rain/no-rain threshold tem peratures over the study area in the east Africa region are shown to h ave exhibited a marked seasonal trend, with an annual variation approa ching 40 K. Minimum threshold temperature values were found at the ons et of the summer wet season, and maximum threshold temperature values during the driest winter months. Generally, summer threshold temperatu res were low, around 230 K, and winter thresholds high, in the range o f 240-260 K. During the wet season, optimum IR rain/no-rain threshold temperatures exhibited a distinct pattern of spatial variation. This w as modeled as a function of pixel latitude, longitude, and surface ele vation. This threshold temperature model was then used to generate thr eshold temperature estimates at the pixel scale from an independent Me teosat dataset for 1992. Compared with the performance of spatially un iform threshold methods, marked improvements in rain-area classificati on accuracy were obtained. Optimum IR rain/no-rain threshold temperatu re variation is therefore seen to be a result of a complex interaction of climatology, meteorology, and topography, and as such the implicat ions of this for the design and use of regional-scale rainfall monitor ing techniques are discussed.