Western part of Japan experienced record-breaking heavy rainfall and flood disaster in July 2018. The rainfall events were embedded in a Baiu front, a synoptic scale stationary front which often causes long lasting rainfall in June-July in Japan. In 2018, however, a preceding typhoon provided abundant moisture into the front, resulted in record-breaking heavy rainfall many places in the western Japan. We have been developing an ensemble flood forecasting system composed of a regional ensemble prediction system (EPS) by the weather research and forecasting model (WRF) and a local ensemble transform Kalman filter (LETKF), and a distributed hydrological model, the rainfall-runoff-inundation (RRI) model. We applied this system to one of the disastrous flood events in 2018 for Takahashi River in Okayama Pref. where more than 50 people were killed. The rainfall forecast system showed promising results with 30 h forecast lead time. Operational deterministic mesoscale forecasts by the Japan Meteorological Agency showed successful results as well with the same forecast lead time. However, our ensemble forecast system showed promising results up to 54 h lead time, and it showed possibilities of disastrous flood up to 78 h lead time. In our previous application to a mesoscale band-shaped rain system, the forecasting system showed promising results only for 18 h lead time. The difference of the predictabilities is likely come from the type of the rainfall system, a synoptic Baiu front or a mesoscale rainband. We need to be aware the predictability of the system is highly dependent on the type of rainfall system. |