On Hydrologic Validation of Remote Sensing Precipitation Estimates

Quantitative evaluation of precipitation maps based on satellite-based information is a critical step in improving the retrieval (estimation) algorithms and the accuracy of the products. This evaluation, often referred to as validation, can take the form of statistical, physical, or application-related analysis. Statistical validation requires the use of reference data with known error characteristics; physical validation checks the physical consistency of the observed quantities; and the application-based validation establishes the value of a specific product for a given application. While the authors discuss all three approaches to satellite-based precipitation products, their emphasis is on the hydrologic applications of flood prediction using rainfall-runoff models. They demonstrate that to be useful in hydrologic validation of space-based rainfall products the hydrologic models should satisfy certain conditions, including the ability to make skillful predictions when forced by accurate data. The authors provide examples that illustrate how different hydrologic processes may compensate for errors in the rainfall products (model input) leading to erroneous inferences about the quality of the forcing. The presentation concludes with a proposed hydrologic validation framework that combines statistical and physical aspects.