Spatio-temporal-dependent errors of radar rainfall estimates in flood forecasting for the Nam River Dam basin

Abstract

In flood prediction, radar rainfall estimates have been widely used for calculating approximate rainfall amounts and predicting flood risks. However, radar rainfall estimates have a number of errors, such as beam blockage and ground clutter, that hinder their application in hydrological flood forecasting. Even though previous studies have focused on removing radar data error, it is important to evaluate runoff volume, which is primarily influenced by such radar errors. Furthermore, correlation structures including temporal as well as spatially dependent errors have not been thoroughly studied. Therefore, how those radar rainfall estimate errors influence flood forecasting was tested in this paper by using a spatio-temporal error model (STEM), which is a cross-correlation model of radar error. Rainfall-runoff simulations were conducted with radar rainfall events generated using the STEM to investigate the response in a basin. The Nam River basin, South Korea, was the focus of the study. A distributed rainfall runoff model, Vflo, was used for runoff simulation. The spatial and temporal resolutions were 0.5km and 10min, respectively. The results indicated that a strong correlation caused a higher variation in the peak discharge. It is concluded that the STEM effectively represents the radar rainfall profile. The spatio-temporal error (STE) of the radar rainfall estimates caused much higher uncertainty in rainfall runoff than only spatial error. To reduce uncertainty in flood prediction using radar rainfall estimates, it is important to eliminate the STE from those estimates.