On the use of real-time radar rainfall estimates for flood prediction in mountainous basins

This paper investigates the effect of systematic mean-field and range-dependent radar rainfall errors on the accuracy of runoff simulation in mountainous basins. Statistical analysis of radar rainfall and runoff simulation error is performed on six flood events for two medium size watersheds in northern Italy, located at 38 and 60 km basin-to-radar distances, respectively. We show significant range-related rainfall biases, which are due to the high elevation radar scans used to minimize the interception of the radar beam with the topography. These biases are corrected by converting radar reflectivity measurements at a given altitude into their equivalent surface values, using real-time identification of the mean vertical reflectivity profile. The mean-field bias is adjusted using a multiplicative factor determined based on real-time radar-rain gauge comparisons. The impact of the above radar rainfall biases, and the improvements obtained from the proposed corrections, on areal-rainfall estimation and runoff simulation are evaluated. It is shown that radar rainfall biases magnify through the rainfall-runoff transformation and preclude the accurate simulation of runoff, particularly for the distant basin. The combined correction procedure results in significant reduction of the hydrologic prediction error, especially at the distant basin.