Development and Planned Operational Applications Of A Conditional Error Model For Radar Rainfall Estimates

A model for the statistical distribution of radar rainfall estimate errors has been developed empirically from WSR 88D and rain gauge data. The model expresses the expected value of actual rainfall and the mean and standard deviation of the multiplicative error in the radar estimate as functions of the radar estimate itself and several power law parameters derived from an historic sample of rain gauge/radar pairs. The model enables an end user of the radar estimates to determine the expected value of point rainfall, the probability that rainfall is less than or greater than a given value, and the probability that the true rainfall is within a given interval. Experiments with data from several WSR 88D umbrellas indicate that the basic form of the model is valid at most or all locations within the conterminous United States, though parameter adjustments for long-term radar bias and rainfall climatology must be made. The error model has several potential applications in radar hydrology, such as determining the probability that rainfall exceeds flash flood-producing thresholds, quality control of real-time rain gauge estimates, and construction of ensembles of rainfall fields. The development methodology and examples of operational applications will be presented.