An approach for improving the capability of a coupled meteorological and hydrological model for rainfall and flood forecasts

Abstract

This study proposes a method for improving the capability of rainfall and flood forecasts by generating ensemble precipitation predictions (EPPs) associated with radar-based rainfall prediction by considering spatial rainfall errors. The EPPs are generated as a kind of Monte-Carlo simulation based on the performance of the numerical weather prediction (NWP) model in the previous time step window. The generated EPPs continue to be blended with radar-based rainfall predictions to produce hybrid rainfall forecasts that perform better than each system could perform individually. The hybrid forecasts are then adjusted by reducing the spatial rainfall errors, which have considerable contributions to the accuracy of the flood forecasts. To validate the performance of the proposed method, this method was applied to improve the capability of the coupled Local Data Assimilation and Prediction system (LDAPS) and Sejong University Rainfall – Runoff (SURR) model for rainfall and flood forecasts during two flood events that occurred in 2013 and 2016 in the Yeongwol watershed. EPPs were generated from the deterministic LDAPS rainfall events and were then blended with the McGill algorithm for precipitation nowcasting by Lagrangian extrapolation (MAPLE) rainfall predictions to produce hybrid rainfall forecasts. Each ensemble member of the hybrid model was then corrected before it was forced as an input for the SURR model to obtain ensemble streamflow predictions. The results showed that the capability of the coupled model was improved sustainably step by step and exhibited the best skills after applying the final step. For rainfall forecasts, the proportion correct (PC), root mean square error (RMSE), correlation coefficient (CC), and Brier score (BS) were improved dramatically by 59%, 32%, 32% and 48% for flood events in 2013 and by 39%, 15%, 33% and 37% for flood events in 2016, respectively. For flood forecasting, the Nash – Sutcliffe efficiency (NSE) and absolute relative error in volume (AREV) were improved substantially by 63% and 24% for flood events in 2013 and by 42% and 53% for flood events in 2016, respectively. The success of this case study proved the viability of the method proposed in this study.