為了提升數值天氣預報準確度，改善模式初始場是一項重要關鍵。資料同化技術能結合更多額外的觀測資料，使得天氣預報模式的初始場更接近大氣的實際狀態。本研究使用三維變分同化(Three-Dimensional Variational Data Assimilation, 3DVAR)方法及四維變分同化(Four-Dimensional Variational Data Assimilation, 4DVAR)方法，將福爾摩沙衛星三號(FORMOSAT-3/COSMIC)的衛星掩星觀測(GPS RO)資料同化於中尺度模式，以了解同化衛星掩星資料對於改善颱風預報準確度的影響。 選取的颱風個案為2009年中度颱風莫拉克(Morakot)，共分成兩組實驗。第一組實驗是討論虛擬渦旋資料的同化方式與颱風初始結構的關係，結果顯示同化虛擬渦旋資料有效地改善颱風初始強度、結構與中心定位，若使用三小時的同化窗區，初始颱風能出現較強較高的眼牆，且具有不對稱性的風場。第二組實驗則是探討同化FORMOSAT-3/COSMIC的GPS RO資料對大氣環境的改善程度及討論對颱風模擬的影響，同時同化GPS RO和虛擬渦旋資料使得駛流場和颱風本身更接近真實，所以其模擬結果的颱風路徑、強度及降雨量都有所改善。 另外，本研究使用觀測系統模擬實驗(Observing Systems Simulation Experiment, OSSE)，評估未來COSMIC-II及其他GPS RO計畫(如：伽利略全球衛星定位系統)對侵臺颱風預報的影響。改良後的二維射線追蹤模式能有效降低非局地大氣折射率的誤差，提高OSSE實驗中資料的真實性。觀測系統模擬實驗的結果顯示模式範圍內的GPS RO資料數量越多，就能得到更多的修正，進而降低模式初始場的誤差，對模式預報準確度有正面的影響。 Improving the model initial analysis is crucial to the prediction of weather forecast model. Assimilation with the observations may help reduce the discrepancies between the model initial guess and the true atmospheric state, thus improving the model initial conditions. This study uses three- and four-dimensional variational method, namely WRF 3DVAR and MM5 4DVAR to assimilate FORMOSAT-3/ COSMIC GPS radio occultation (RO) refractivity data to understand their impacts on typhoon prediction. Typhoon Morakot (2009) is selected in this study. Two sets of experiments have been designed. The first set of experiment discusses the relationship between the bogus-vortex data assimilation (BDA) method and the initial structure of Morakot. The results have shown that the BDA using MM5 4DVAR is robust and can significantly improve the initial intensity, the structure, and the center position of the Morakot. With a longer three-hour data assimilation window, the initial typhoon can form a stronger and higher eye wall with asymmetric wind. The second set of experiment investigates the impact of assimilating GPS RO data on the improvement of model initial analysis and typhoon prediction. Assimilating both GPS RO and bogus vortex data makes the steering flow and typhoon much closer to the observed. Hence, the typhoon track, intensity and the associated rainfall in Taiwan have been improved in this simulation. This study also aims at the potential impacts of prospective RO measures from US GPS (COSMIC-II) and other navigation satellites (GALILEO) through the Observing Systems Simulation Experiment (OSSE) to evaluate their impacts Morakot typhoon prediction based on WRF 3DVAR and MM5 4DVAR. A modified version of two- dimensional ray tracing model effectively reduces the error of the derived RO refractivity for use in OSSE. Results of both OSSEs with WRF 3DVAR and MM5 4DVAR clearly illustrate that more GPS RO data (e.g., COSMIC-II plus GALILEO) can provide more amendments to the model initial states, thereby reducing their errors and leading to a positive impact on typhoon forecast.
