為培養系集擾動的流場相依特性，系集卡曼濾波器(Ensemble Kalman Filter, EnKF)需一定spin-up時間使系集卡曼濾波器達到理想表現。針對颱風同化與預報課題，本篇在觀測系統模擬實驗(Observation System Simulation Experiment, OSSE)架構下探討加速就位法(Running In Place, RIP)法是否在同化早期可更有效使用QuikSCAT衛星與投落送觀測資料。在本研究中，利用模式WRF與LETKF-RIP的方法模擬梅姬颱風(2010)，同化資料包括sounding、Quikscat表面風場與飛機觀測。根據實驗結果顯示，若同化中使用RIP這個方法，可以在較早的時間就得到與真實場相近的預報路徑。同時在實驗中也針對系集數量進行探討，從理論上來看，系集的數量是越多越好，但是本研究中證實即使使用較少的系集數量，配合RIP方法的使用後亦可發揮不錯的效果。The ensemble Kalman filter (EnKF) needs a number of analysis cycles to develop reliable ensemble-based background error covariances for better analysis performance. Focusing on typhoon assimilation and prediction, this study carries out observation system simulation experiments (OSSEs) to investigate the effects of applying the Running In Place (RIP) method to the assimilation of QuikSCAT and dropsonde observation data at early analysis cycles, In these experiments, the Weather Research and Forecasting (WRF) model coupled with the local ensemble transform Kalman filter (LETKF) and RIP is used to simulate Typhoon Megi (2010) assimilating sounding, Quikscat surface wind and airplane observations. The results show that the track forcasts initialized at early time are better in the assimilation experiments with RIP. Besides, even though a larger ensemble size theoretically gives higher analysis accuracy, it is found that RIP is still beneficial when the ensemble size is smaller.
