A Satellite Algorithm for Predicting the Distribution of Rainfall Rate

An algorithm for predicting one to two hours in advance the spatial distribution of rainfall rate is introduced in this work. The suggested algorithm uses satellite rainfall estimates to predict the rainfall field. The algorithm predicts first the most likely rainy areas and then predicts the expected amount of rainfall rate in each rainy pixel. The algorithm identifies the rainy cloud cells and determines the cloud motion vector of each cell. The motion vector is used to advect the rainy area and to identify the potential predictors. The potential predictors are the previous observations of GOES brightness temperature located in a neighborhood region with center on a predicted pixel. The forward selection algorithm is used to eliminate irrelevant pixels and determine the best predictors for each region. Rainfall prediction is derived after evaluating the empirical models over the persistence and advected rainy pixels. The weather research and forecasting (WRF) model is used to compare the performance of the proposed algorithm. Prediction errors of the proposed algorithm were smaller than the forecast from the WRF model, which indicates that the proposed algorithm is a potential tool to predict rainfall spatial variability at short time interval. The algorithm was implemented to handle data from satellite and from weather radar. The proposed prediction algorithm will be couple with a hydrological numerical model to predict the flash flood events.