The National Weather Service (NWS) has been using hourly radar-derived precipitation estimates from multiple radars to create regional precipitation mosaics for use by River Forecast Centers (RFCs). The radar estimates along with hourly rain gage observations and hourly satellite-derived are the main input into the Multi-sensor Precipitation Estimator (MPE) which was implemented in AWIPS for operational use in July 2001. Here we describe the various components of the RFC-wide MPE which include: mean-field bias adjustment of individual radar estimates at multiple time scales, multi-radar mosaic, merging of radar and gage estimates, local bias adjustment, and use of satellite estimates in areas where radar coverage is poor and gage coverage is sparse.

MPE performance during heavy rainfall events will be discussed for several different regions of the United States illustrating the MPE improvement over raw radar estimates. In addition, examples of forecast river hydrographs using the MPE estimate as precipitation input will be shown and compared to forecasts which used raw radar estimates with no bias correction. Forecast river hydrographs using rainfall estimates derived only from rain gauge data will also be compared to those derived from the MPE estimate.