A satellite-based rainfall estimation method for the Lake Victoria basin

This article describes a method for estimating precipitation over the Lake Victoria basin in East Africa using Meteosat images. The method identifies infrared and visible signals characteristic of convective storms and associates their features with ground rainfall measurements. These temporal patterns are recognized at the pixel level via a neural network model, while mean areal precipitation (MAP) is estimated at larger spatial scales. The method successfully identifies convective cells and provides reliable MAP for daily time steps, and even better for longer time intervals. The procedure is implemented within the Lake Victoria Decision Support System (LVDSS), which addresses planning as well as operational applications for water resources, agriculture and hydropower management in the Lake Victoria basin. LVDSS is presently used by several water resources agencies in Kenya, Tanzania and Uganda.