Abstract
Recent advances of regional ensemble weather predictions have increasingly improved predictability of rainfall forecasts, which would be useful for streamflow forecasts. We have developed an ensemble flood forecasting system, composed of an ensemble Kalman filter and regional model WRF (Weather Research and Forecasting) in atmospheric part, and Rainfall-Runoff-Inundation (RRI) model in hydrological part. We examined feasibility of the flood forecasting system on the Niigata-Fukushima torrential rainfall event in July 2011. The ensemble flood forecasting system showed improvements compared to deterministic forecasts in average accuracies and predictability of peak flow. However, the forecasting system failed to predict the largest flow peak in this event, because it could not predict the third of the main three rainfall peaks. It could improve the accuracy of JMA forecasts. However, at the same time, it depended on JMA forecasts accuracy. This study showed the advantages and limits of this ensemble flood forecasting system.
