Accuracy of radar rainfall estimates for streamflow simulation

The aim of this work is to analyse the impact of errors in radar rainfall estimates on rainfall-runoff modelling. This issue is addressed through application of radar rainfall estimates for continuous, lumped rainfall-runoff modelling of the Brue catchment, a mid-sized basin in South-West England, over a two and a half-year period. The study is focused on radar rainfall errors associated with range-related bias due to the vertical profile of reflectivity and with mean-field bias due to systematic errors in the radar calibration and biased reflectivity-to-rainrate relationship. Streamflow is simulated through a conceptual hydrological model based on mean areal rainfall estimates obtained by using various radar rainfall processing scenarios. These simulations are evaluated and compared with corresponding streamflow simulations from a dense raingauge network. The comparisons show that radar errors may preclude the use of unadjusted radar estimates for runoff modelling. Radar rainfall adjustment significantly improves model results with simulation efficiency increasing up to 30% after adjustment. Comparison of radar-driven simulations with observed discharge data reveals a simulation efficiency of 0.75 for the lowest radar scan (adjusted), whereas simulation efficiencies are lower for higher radar scans. The results reveal the critical importance of using radar rainfall estimates as close as possible to the ground and the considerable impact that effects of vertical variability of reflectivity have on runoff simulation. (C) 2002 Elsevier Science B.V. All rights reserved.