UC Irvine

Accurate measurement of rainfall distribution at various spatial and temporal scales is crucial for hydrological modeling and water resources management. In the literature of satellite rainfall estimation, many efforts have been made to calibrate a statistical relationship (including threshold, linear, or nonlinear) between cloud infrared (IR) brightness temperatures and surface rain rates (RR). In this study, an automated neural network for cloud patch-based rainfall estimation, entitled self-organizing nonlinear output (SONO) model, is developed to account for the high variability of cloud-rainfall processes at geostationary scales (i.e., 4 km and every 30 min). Instead of calibrating only one IR-RR function for all clouds the SONO classifies varied cloud patches into different clusters and then searches a nonlinear IR-RR mapping function for each cluster. This designed feature enables SONO to generate various rain rates at a given brightness temperature and variable rain/no-rain IR thresholds for different cloud types, which overcomes the one-to-one mapping limitation of a single statistical IR-RR function for the full spectrum of cloud-rainfall conditions. In addition, the computational and modeling strengths of neural network enable SONO to cope with the nonlinearity of cloud-rainfall relationships by fusing multisource data sets. Evaluated at various temporal and spatial scales, SONO shows improvements of estimation accuracy, both in rain intensity and in detection of rain/no-rain pixels. Further examination of the SONO adaptability demonstrates its potentiality as an operational satellite rainfall estimation system that uses the passive microwave rainfall observations from low-orbiting satellites to adjust the IR-based rainfall estimates at the resolution of geostationary satellites. Copyright 2005 by the American Geophysical Union.