We propose a new approach to statistical validation of maps of precipitation that are derived from satellite remote sensing observations. This approach requires availability of ground-based reference data and we plan to use NEXRAD observations. Our approach will answer the following question: "Given radar and satellite rainfall maps, what is the probability that the true rainfall amount exceeds the satellite rainfall estimate over an arbitrary region on the map" Posing the validation problem in a probabilistic context standardizes the notion of what is "good" or "bad" as there is a common scale (from 0 to 1) that is easy to interpret. Answering this question will greatly enhance the use of satellite data in flood forecasting and in operating water resources systems. Validation activities for GPM should directly assess the utility of rainfall estimates from the perspective of the drainage basin and at basin scales that are relevant to flood hazard problems that can be addressed by GPM rainfall products. Our focus in hydrologic studies will be on rainfall estimation for extreme floods at basin scales ranging from approximately 100 km2 to 100,000 km2. Hydrologic validation perspective requires use of a probabilistic model of the reference data uncertainty. We have developed such a model for hourly rainfall accumulation based on radar-rainfall estimates. It is known as Probabilistic Quantitative Precipitation Estimations. The model accounts for seasonal, scale, and range dependence of radar-rainfall uncertainties. The parameterization we have developed using a large set of empirical data addresses the multiplicative errors conditional on the estimated radar-rainfall amount. To address the validation issues from the perspective of flood hazards, we will perform hydrologic studies of the scale-dependent properties of flood-producing rainfall in the Potomac River basin and the Texas-Gulf drainage basins. These studies will center on analyses of more than 50 flood events in each of the study regions. The proposal addresses several topics of the NNH06ZDA001N-PMM, including (i) Development and application of statistical validation techniques; (ii) Development of metrics and methodologies for assessing satellite precipitation products in hydrological modeling; and (iii) Impact of future precipitation data sets on hydrological prediction, including methods for quantifying and using estimates of systematic and random precipitation measurement errors.
