|
摘要
|
This study utilizes the 46-day ECMWF ensemble forecasts to assess the subseasonal forecast skills of tropical cyclones (TCs) in the western North Pacific. The 20-year reforecasts from 2002-2021 are used to objectively detect TC formations and their subsequent tracks (Tsai et al. 2023). The evaluation of TC forecast skills is carried out by utilizing the TC strike probabilities in the 10-degree latitude-longitude boxes. The reliability diagrams (RDs) are utilized to evaluate the forecast skills in weeks 1-4, which show slight over-forecasting biases. The AUC (Area Under Curve) of the ROC (Receiver Operating Characteristic) curve is computed for week-1 to week-4 forecasts, which are 0.91, 0.85, 0.82, and 0.81, respectively.
The impacts of the Madden Julian Oscillation (MJO), Boreal Summer Intraseasonal Oscillation (BSISO), and Western North Pacific Summer Monsoon (WNPSM; Wang et al. 2001) on the subseasonal TC forecast skills are investigated. In addition to the RD and ROC curve, the Precision-Recall (PR) curve is used to reduce the impact of imbalanced data. The preliminary results indicate that the subseasonal TC forecast skills are influenced by the large-scale environments. For example, TC forecast skills are better when the forecasts are initialized in MJO phases 6 and 7. More details about the impacts of MJO, BSISO, and WNPSM on the subseasonal TC forecast skills will be presented in the meeting. Finally, TC precipitation forecast skill in the subseasonal timescale will also be evaluated to explore the potential application for the water resources management in Taiwan. |
GoogleAnalytics