Development of Global Precipitation Estimation System Using Artificial Neural Network Models

Abstract:

The PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network) system, developed at UC Irvine, is one unique source to estimate global precipitation in near real-time using infrared and passive microwave information from Geosynchronous Earth Orbital (GEO) and Low Earth Orbital (LEO) satellites. The algorithm uses an Artificial Neural Network to extract cold cloud pixels and neighboring features from GEO-satellites’ infrared images to generate rain rate. The precipitation estimates from the neural network are further adjusted by the PMW precipitation estimates produced using the data from LEO satellites. The operational PERSIANN system estimates global precipitation in near real-time. Data sources are also extended to the reconstruction of historical data for the past 30 years for hydroclimate studies. Continuing development of precipitation retrieval using artificial neural network models and advanced machine learning methods are ongoing. Studies including effective feature extraction from satellite multiple spectral imagery, integration of multiple satellite information, and merge of ground and satellite precipitation retrievals. Evaluation of PERSIANN precipitation and its application for catchment scale hydrologic simulation will be discussed.