This study investigated the response of hydrologic calibration to replacing gauge-based with radar-based precipitation data in the USEPA Chesapeake Bay Program (CBP) Watershed (CBW) model over the Potomac River Basin. Specific objectives were to (1) compare gauge-based and NEXRAD radar-based (Multisensor Precipitation Estimator, MPE) data at the (a) point-pixel and (b) spatially aggregated level; (2) evaluate the model's calibration accuracy using the different precipitation data sets; and (3) examine the response of model hydrology.

Hourly gauge-point and MPE-pixel data were compared at 80 locations.  The CBP's interpolated and aggregated precipitation data at the model unit (county) level were compared with MPE data aggregated to the same 114 county-based spatial segments.  The model calibration followed the CBP's automated approach, using observed streamflow at 37 gauge stations.  Model performance was evaluated using calibration and hydrologic statistics, and GIS-aided spatial information.  Calibrated parameters and model hydrologic fluxes were compared.

   The average annual gauge-point and MPE-pixel values (excluding hours when either was missing) agreed well.  Differences in average annual values between the spatially aggregated data sets were, however, significant in parts of the study area.

When parameter constraints were relaxed to allow calibration to adjust to the smaller volume of precipitation, the model using MPE outperformed the model calibrated to CBP precipitation data at 65% of the 37calibration sites.  The model response was controlled largely by the seasonal difference in precipitation inputs: (1) calibration process could not compensate for large differences in seasonal flow bias caused by the seasonal volume of precipitation; (2) seasonal flow bias affected the lower zone nominal soil moisture storage parameter (LZSN), mainly affecting interflow and groundwater flow.  The surface flow component was generally the same for the different precipitation inputs.  The two precipitation data types can be used interchangeably to simulate surface-flow dominated processes, but care must be taken in simulations where subsurface pathways and residence times are important.  

    MPE is a strong alternative to gauge-based precipitation data because of its spatiotemporal coverage and rare missing records.  Using MPE in hydrologic modeling is appealing because of the improved calibration accuracy of the CBW model demonstrated in this study.