Subseasonal Typhoon Precipitation Forecast in Taiwan Area Using the ECMWF Reforecasts: Forecast Verification and Application

The main objective of this study is to assess typhoon precipitation forecast skill on the subseasonal timescale. The 20-year reforecasts from the ECMWF 46-day ensemble (ENS) are utilized to compare with gridded surface observations in Taiwan. The analysis focuses on the dates when typhoons affect Taiwan (117-129E and 19-28N). 15 ENS grids around Taiwan area are used with the grid size of 0.8 x 0.8 degree. Historical rainfall observations are provided by the Central Weather Administration (CWA), which the observations from the surface stations are interpolated into a resolution of 1km x 1km grid box. A comparison between the ENS forecast data and gridded CWA rainfall observations is performed by searching the optimal percentile rank (PR) of gridded CWA rainfall that has the smallest mean difference against the ENS data. The result reveals that the ENS can somewhat capture the rainfall contrast between the mountainous area and plain area, despite its relatively lower horizontal resolution. However, the difference between ENS rainfall forecasts and surface observations significantly increases for the forecasts beyond 72 hours, due to the model's coarser resolution and typhoon track forecast errors. The ENS typhoon track forecast errors in weeks 1-4 are analyzed by comparing the ensemble vortex tracks with the JTWC best tracks. The track forecast error is decomposed into the along-track (AT) and cross-track (CT) components. The analysis result shows negative mean AT errors, indicating slower translation speed biases in the model. The mean AT errors could reach up to 400 km for the 168 h forecasts after TC formations. Given the significant typhoon track forecast errors, using the raw ENS rainfall forecasts for the operational TC forecasting/outlook become challenging. In response, we have developed a statistical Quantitative Precipitation Forecast (QPF) model to predict typhoon rainfall, considering the track biases in the ENS forecasts. The forecast tools developed in this study will be integrated into CWAs subseasonal typhoon forecast system to support water resources management and disaster risk reduction.