Cloud Patch-based Rainfall Estimation Using a Satellite Image Classification Approach

The relationship between cloud and surface rain rate varies significantly from cloud patch to cloud patch. Therefore, a rainfall estimation model characterized by significant transience, heterogeneity, and variability is needed to associate rainfall with the extremely complex and still imperfectly understood precipitating processes to produce higher quality estimates. We have responded to this by developing a high-resolution precipitation estimation algorithm dubbed “CCS” (Cloud Classification System) at UC Irvine. The CCS uses computer image processing and pattern recognition techniques to develop a patch-based cloud classification and rainfall estimation system based on co-registered passive microwave and infrared images from Low Earth-orbiting and Geostationary satellites. Unlike the region-based approach, which establishes only one Tb-R function for all clouds, this technique classifies various patches into different clusters and then searches the best-matched nonlinear Tb-R mapping function for each patch. Therefore, CCS jumps out the deadlock of the assumption that colder cloud pixel must produce higher rain rates than warmer cloud pixel, which is not all-time-true but popularly used by some other statistical regression or histogram matching approaches. This design feature enables CCS to generate various rain rates at a given brightness temperature and variable rain/no-rain IR thresholds for different cloud patches, which overcomes the one-to-one mapping limitation of a single statistical Tb-R function for the full spectrum of cloud-rainfall conditions. In addition, the computational and modeling strengths of neural network enable CCS to cope with the nonlinearity of cloud-rainfall relationships by fusing multi-source data sets. We are operating this system with the goal to produce data at spatio-temporal resolution suitable for basin scale hydrological research and applications, along with the goal to provide high-quality precipitation analysis for GEWEX CEOP sites.