Estimating Watershed-Scale Precipitation by Combining Gauge- and Radar-Derived Observations

Watershed modeling requires accurate estimates of precipitation; however, in some cases it is necessary to simulate streamflow in a watershed for which there is no precipitation gauge records within close proximity to the watershed. For such cases, we propose an approach to estimating watershed-scale precipitation by combining (or fusing) gauge-based precipitation time series with radar-based precipitation time series in a way that seeks to match input precipitation for the watershed model with observed streamflow at the watershed outlet. We test the proposed data fusion approach through a case study where the Soil and Water Assessment Tool (SWAT) model is used to simulate streamflow for a portion of the Eno River Watershed located in Orange County, North Carolina. Results of this case study show that the proposed approach improved model accuracy (E=0.60; R2=0.74; PB=−10.2) when compared to a model driven by gauge data only (E=0.50; R2=0.54; PB=−25.5) or radar data only (E=0.33; R2=0.61; PB=−13.7). While this result is limited to a single watershed case study, it suggests that the proposed approach could be a useful tool for hydrologic engineers in need of retrospective precipitation estimates for watersheds that suffer from inadequate gauge coverage.