Validation and Intercomparison of Satellite Rainfall Estimation Algorithms over the Continental United States

Historically, three major classes of satellite rainfall estimation algorithms are those that utilize data from (1) visible or infrared frequencies, (2) microwave frequencies, and (3) multiple frequencies. It has been difficult to draw definitive conclusions on algorithm performances with respect to each other, as the spatial and temporal extent of the "ground truth" data used to validate the satellite estimates is usually limited and the accuracy of the data is often unknown.

The errors of satellite algorithms from each of the three classes mentioned above are quantified using radar and raingage data for the continental United States (CONUS) over a period of two years from 1998 to 2000. The National Centers for Environmental Prediction (NCEP) nationwide mosaic of Next Generation Weather Radar (NEXRAD) Hourly Digital Precipitation (HDP) estimates was collected on an approximately 4 km grid, and the collocated hourly rainfall estimates (also on an approximate 4 km grid) from Geostationary Operational Environmental Satellite (GOES) data were collected at more than 10,000 locations over the CONUS with 24-hr rain gages. In addition, instantaneous rainfall estimates from the Special Sensor Microwave/Imager (SSM/I) were collected with the corresponding HDP estimates for the locations of the 24-hr gage data and hours of the day corresponding to the SSM/I overpasses. The 24-hr gages are used to evaluate the temporally and spatially varying biases of the HDP estimates so that the satellite estimates may be compared with a quality-controlled bias-corrected hourly radar product whose uncertainty can be estimated.