R-CCS Cafe 第202回 第2部

Abstract: On September 9 and 10, 2015, River Kinu in Tochigi Prefecture had a catastrophic flooding due to torrential rainfalls over 600 mm in the 2-day period. This historic heavy rain occurred due to an active line-type rainband maintained for an extended period in the middle of two typhoons. This study explores a potential use of additional dense surface observations to better predict this rainfall event. To investigate the impact of these dense surface data that were not included in the operational weather forecasting systems at that time, we perform data assimilation (DA) experiments using the Local Ensemble Transform Kalman Filter (LETKF) with the SCALE regional numerical weather prediction model. Two DA experiments were performed: the control experiment (CTRL) at 4-km resolution with only hourly conventional observations that were included in the operational weather forecasting systems at that time, and the other TEST experiment with additional hourly dense surface observation data. CTRL showed general agreement with the observed weather patterns, although the track of Typhoon Etau was shifted westward. As a result, the heavy rainfall area was shifted to the west compared to the JMA analyzed precipitation based on the radar and gauge observations. By contrast, TEST showed stronger rainfall intensity, better matching with the observed precipitation likely due to an improvement of the track of the Typhoon. To evaluate the impact of the individual surface station data on the rainfall forecast, we also implemented the Ensemble Forecast Sensitivity to Observation (EFSO) technique with SCALE-LETKF. The EFSO showed that about 53 % of surface data contributed to improve the rainfall forecast. The results suggest that the dense surface DA have a potential to improve the forecast accuracy for severe rainfall events.