DEVELOPMENT OF A FLOOD FORECASTING SYSTEM USING REGIONAL ENSEMBLE PREDICTION – APPLICATION TO THE KINUGAWA FLOOD IN 2015

Abstract

 Recent advances in regional ensemble weather prediction systems (EPS) can considerably improve the prediction of torrential rainfall and resulting severe floods with a longer lead time. To maximize this advantage, we developed an ensemble flood forecasting system, composed of an ensemble Kalman filter and a regional numerical weather prediction on the atmospheric part and the Rainfall-Runoff-Inundation (RRI) model on the hydrological part. Simple downscaling of global EPS was also conducted to compare the superiority of the developed system. We applied the system to the Kinugawa flood event in September 2015 and studied its feasibility for forecasting. Ensemble flood forecasting with a lead time of 9–15 h was successfully conducted in quantitative prediction of a flood discharge peak. Forecasting with a lead time of 21–39 h showed a possibility of flood occurrence though its probability was low. The flood forecasting system was superior to deterministic forecasting based on the Japan Meteorological Agency (JMA) operational Mesoscale Model (MSM) in the predictability of a flow peak. A simple EPS by downscaling global EPS had an advantage with a lead time of 57 h compared to the previous system. This study revealed the potential and limitations of the ensemble flood forecasting system in predicting high flood peaks.