A Feature-Based Approach to Satellite Precipitation Monitoring Using Geostationary IR Imagery

Abstract

This paper describes a new geostationary infrared satellite precipitation monitoring methodology that draws on techniques from machine vision to develop a mathematical representation of cloud shapes and textures and the form of their embedding meteorological context. Quantitative feature descriptions derived from Geostationary Operational Environmental Satellite (GOES) band-4 infrared imagery were translated into 15-min precipitation estimates through the use of an artificial neural network. The network was trained using 3-hourly ground-radar data in lieu of a dense constellation of polar-orbiting spacecraft such as the proposed Global Precipitation Measurement (GPM) mission. The technique has been tested using data from the Texas?Florida Underflights Experiment (TEFLUN-B) and has demonstrated significant advantages over existing satellite precipitation monitoring techniques when applied to a summer precipitation regime in Florida.