Validation and Intercomparison of SSM/I Rain-Rate Retrieval Methods over the Continental United States

Abstract

An important source of error or ambiguity in the satellite passive microwave detection and estimation of precipitation rate over land is variable background emission, reflecting differences in surface temperature and moisture, soil type, and vegetation cover. Three experimental algorithms for the Special Sensor Microwave/Imager (SSM/I) are described that attempt to improve the precipitation signal-to-noise ratio by selectively responding to transient brightness temperature perturbations relative to maps of seven-channel monthly mean radiances. These algorithms are validated and intercompared along with two quasi-standard SSM/I algorithms developed by Grody and Ferraro and by Adler and Huffman. For ground truth, nine months of 10-cm radar data taken at six sites and of hourly rain gauge reports from approximately 2700 locations in the United States were used. The radar data were carefully quality controlled and calibrated against coincident gauge reports. The required calibration adjustment of the radar rain rates was found to be as large as a factor of 2?5. The satellite estimates from snow-free pixels are validated both against the calibrated radar data within the zones of radar coverage and directly against the gauges for the entire lower 48 states. An adaptation of the Heidke skill score is introduced as a statistical calibration and validation tool that may avoid some of the pitfalls of standard validation statistics. Under the relatively homogeneous background conditions characteristic of the study region, the five algorithms are found to yield similar results, and no algorithm emerges as clearly superior. The hypothesis that maps of monthly mean surface emission can aid in the detection of light precipitation requires further testing over more varied land surfaces.