福爾摩沙衛星七號(福衛七號)於 2019 年 6 月發射，並能提供一天超過 6000多筆的無線電掩星(RO)觀測資料，其中 RO 偏折角能夠使用 NCEP 偏折角反演法以及 Abel 轉換兩種技術，將偏折角反演成大氣剖面資訊如氣壓、氣溫以及水氣等等。本研究旨在使用氣象局的全球模式 TGFS 探討福衛七號 RO 觀測資料對於颱風預報的影響，模式解析度約為 25 公里，並選擇了 7 個颱風個案進行 83 組預報。每組預報分為兩個敏感度實驗，即包含福衛七號偏折角同化的實驗（WB）和不包含福衛七號偏折角同化的實驗（NB）。同化策略採用 GSI 的 Hybrid 4DEnVar 方法，首先進行 NB 實驗的同化循環，直到颱風生成前四天。在此時間點，將實驗分為 WB 和 NB 兩個敏感度實驗，每六小時執行一次同化，同化各自的觀測資料。在颱風生成後，進行每六小時一次的 120 小時長期預報，同時持續進行各自實驗的同化循環。統計結果顯示，在 83 個預報中，WB 在路徑誤差整體表現上較 NB 差，直到 60 小時後 WB 的強度誤差才低於 NB，此結果與先前針對 5 個颱風個案、42組預報的研究結果相反。梅花颱風個案分析中的掩星偏折角校驗結果顯示，三組實驗的趨勢基本相似，高層的掩星偏折角偏差較小，底層偏差較大，剃除 4 公里以下福衛七號資料後，路徑誤差在前期與 NB 結果差異不大，到後半段誤差則急劇增加。在全球校驗方面，NB 和 WB 在 U、V 分量風以及溫度場的表現呈現好壞交錯。然而，在水氣方面，WB 在對流層有明顯的改善，尤其在北半球改善效果最為明顯。 綜合而言，本研究結果顯示同化福衛七號掩星觀測對於颱風預報的正面影響 並不明顯。這可能與福衛七號在同化過程中存在的一些限制有關，尤其是與低層觀測的不確定性相關。;The FORMOSAT-7 (FS7) satellite, which was launched in 2019, has the capability to provide over 6000 Radio Occultation (RO) observation profiles per day. To conduct our analysis, we employed the CWB′s Taiwan Global Forecast System (TGFS) numerical weather prediction model, which features a fine grid resolution of 25 km. We simulated 7 Typhoon cases, consisting of 83 individual runs between 2021 and 2022. For each case, we conducted two experiments: one utilizing the FS7 RO bending angle data (WB), and the other without it (NB).To assimilate the RO bending angle data, we employed the GSI hybrid 4DEnVar data assimilation system. The assimilation process spins up for 4 days. This was followed by a 120-hour forecast.The statistical analysis of the total 83 runs reveals that the assimilation of FS7 RO bending angle data leads to a degradation in track forecast performance, with a slight positive impact on intensity forecast after 60-hour.The case study of Muifa exhibited a similar overall trend in track and intensity errors as observed in the overall statistical results. However, it demonstrated a more substantial improvement in intensity forecast beyond the 60-hour threshold. Additionally, we observed that the Root Mean Square Errors (RMSEs) of U, V, and T variables were comparable between the WB and NB. Notably, WB exhibited a notable positive impact on water vapor for each atmospheric layer. Despite the abundance of available FS7 RO data on a daily basis, the positive impact on Typhoon forecasting seems to be implicit. This could be attributed to the limited utilization of FS7 RO data, primarily influenced by observation errors in the lower levels of the atmosphere.
