We examine the feasibility of ensemble streamflow prediction utilizing forecast rainfall by regional ensemble prediction system by Weather Research and Forecasting (WRF) model and ensemble Kalman filter. The target area is the Hiyoshi Dam catchment in Kyoto Pref., and the target events are typhoon TALAS (201112) and ROKE (201115). The WRF model with 3 km horizontal interval was run by initial and boundary condition made up with Japan Meteorological Agency-Mesoscale Model (JMA-MSM). The predicted rainfall was given to a distributed hydrological model, Rainfall-Runoff-Inundation (RRI) model. The ensemble streamflow predictions in their forecast lead time up to 6 hours improved those of deterministic forecasts in both typhoon events. On the other hand, forecast experiments up to 33 hours predicted reliable discharge in ROKE case, but not in TALAS case. In ROKE case, ensemble streamflow forecasts improved deterministic forecasts, and showed future discharge in fairly good accuracy independent from forecast initial time. In TALAS case, the forecasts were unstable depend on initial time. In case of stable ensemble streamflow forecasts are obtained such as in ROKE case, the ensemble streamflow forecasts could be useful for operational decision making. |