對中小尺度天氣預報而言，利用系集資料同化提供數值預報模式初始場較四維變分資料同化法在同化開始初期受到限制。因為對系集資料同化方法而言，開始進行同化的背景場(初始猜測場)之準確度決定了同化分析場收斂至應有精準度所需的時間。如無法提供適當的系集背景場，此限制將導致觀測資料在中小尺度天氣現象之發展初期較難發揮效用，並使得資料同化及預報能力受限。本研究希望利用”局部系集轉換卡門濾波” (Local Ensemble Transform Kalman Filter) 建立中尺度數值模式之資料同化系統。 並針對颱風預報課題，利用系集同步加速法 (Running in place)，期望改善並提昇同化初期的準確度及預報度，並期增加觀測資料的使用價值。 ; The ensemble Kalman filter (EnKF) has a disadvantage during the transition (spin-up) period due to the requirement for reliable ensemble mean and the perturbations. It takes much longer time for the EnKF to spin-up to the level with a satisfied accuracy while the variational methods, like 4D-Var, do not have such concern. This disadvantage will cause serious limitations on the meso-small scale predictions for representing the developing stage of an event, when the assimilation process is just being initialized. Therefore, the influence from the observation is also limited. In this study, we plan to investigate the solution proposed by from Kalnay and Yang (2008) for typhoon prediction and to improve the prediction skill during the spin-up period. This method will be implemented within the framework of Local Ensemble Transform Kalman Filter in the Weather Research and Forecasting model. ; 研究期間 9711 ~ 9807
