Hydrologic Modeling with NEXRAD Precipitation in Middle Tennessee

Abstract

The use of radar-based precipitation is investigated for possible improvement of spatially lumped continuous hydrologic modeling in two subbasins of the Cumberland River basin in Middle Tennessee. Hydrologic predictions of streamflow at the outlets of the subbasins are obtained using HEC-HMS and two precipitation inputs, Stage III radar-derived and gauge-only data. Model performance with each precipitation input is assessed by comparing predicted and measured streamflow at each subbasin outlet and calculating streamflow volume bias, root mean square difference, mean normalized peak error, and mean peak timing error. The results indicate that the Stage III precipitation suffers from systematic underestimation at both the point and subbasin scales and cannot be completely offset by model recalibration. Simulations driven by both the Stage III and gauge-only data periodically failed to reproduce observed flood peaks in both subbasins. But Stage III simulations are generally less accurate in prediction of streamflow volume as compared to gauge-only simulations and, although comparable to the gauge-only simulations in predicting the magnitude and time to peak, offer no apparent improvement in predicting these quantities either.