RAINFALL AREA IDENTIFICATION USING GOES SATELLITE DATA

Geostationary operational environmental satellite (GOES) infrared (IR) images and rain gauge measurements are used to study five convective storm events in Florida. The study region is divided into a training area and a test area. An approach is initiated in this study to identify rainfall areas within GOES images using the coldest cloud-top temperature (CCTT) and the standard deviation of cloud-top temperatures (STD-DEV) within a group of pixels. A threshold value of satellite-derived cloud-top temperature is used to define the cloud-covered areas. GOES IR images are then divided into two categories: expanding cloud images and contracting cloud images. The Bayesian optimal classifier using CCTT and STD-DEV is implemented to further classify the cloud pixels in each category into precipitation-free cloud pixels and precipitating cloud pixels. Cutoff rain rates from the gauge measurements are used to define the actual precipitation-free and precipitating pixels. The results show that the classification accuracies are about 73% in the training area, and 63% in test area.