本論文之主要研究目的為建構耦合系集颱風同化與預報系統，並以此探索耦合資料同化與預報在颱風應用中之相關科學議題。本研究中所建構之耦合系集資料同化與預報系統是由高解析度海氣耦合模式UWIN-CM (Unified Wave INterface-Coupled Model)與系集資料同化系統LETKF (Local Ensemble Transform Kalman Filter)所組成，在UWINCM-LETKF之架構下針對2010年的Fanapi颱風進行個案研究，以探討颱風資料同化與耦合預報中的問題。 從颱風系集同化系統中颱風位置不確定性的問題切入，我們先展示了颱風位置不確定性對同化系統表現的負面影響。接著，我們使用了前人研究中提出的颱風中心同化架構(TC-Centered assimilation framework, TCC)作為解方，並首次將此方法使用於真實颱風研究中。實驗結果顯示，使用TCC架構下，分析場中之颱風結構與觀測資料變得更為接近。而以此分析場進行預報後發現，此方法可以緩解預報初期模式的動力不平衡問題，但對颱風強度預報的影響較不一致，包含了較好的中心氣壓預報與預報風速過強的問題。 接著我們以UIWN-CM的系集預報來探討海氣交互作用對颱風預報之影響。實驗結果顯示了加入海氣交互作用後模擬的颱風變小、減弱與更為不對稱，同時路徑有北偏的情況。分析海洋與大氣變數間的耦合相關性讓我們對耦合模式的特性能有更深刻的了解，並對接下來進行耦合資料同化有很大的幫助。 最後，我們則對本研究中建立的UWINCM-LETKF系統進行檢驗，我們先討論在大氣的同化中使用耦合模式的影響。實驗結果與位於相同位置的海洋與大氣觀測比較後發現，使用耦合模式得到的背景場在TCC的架構下，海洋與大氣在交界面的增量較為一致。此結果除顯示了進行耦合同化的潛力，同時也告訴我們不只在大氣的同化中需要使用TCC架構，在海洋的同化上也應該使用。而我們也在此架構下進行耦合資料同化的實驗，嘗試使用大氣的溫度修正海溫，初步的研究結果顯示雖然在不同位置有不同的結果，但整體來說在颱風後側的海溫多能得到改善。 ;The main goal of this dissertation is to construct a coupled ensemble TC assimilation and prediction system to explore the challengess in regional coupled data assimilation for TC analysis and prediction. The coupled ensemble assimilation and prediction system is constructed by coupling a high-resolution coupled model UWIN-CM (Unified Wave Interface-Coupled model) and an ensemble data assimilation system LETKF (Local Ensemble Transform Kalman Filter). Under the UWINCM-LETKF framework, issues in TC ensemble data assimilation (EDA) and coupled model prediction are explored with a real TC study of Fanapi (2010). The first part investigates the problem of TC position uncertainty in current ensemble TC assimilation system. We have demonstrated the detrimental impact of TC position uncertainty on ensemble TC assimilation. The TC-centered (TCC) assimilation framework is adopted as a solution and evaluated with a real case study of Fanapi. Results show that with the TCC framework, the analyzed TC structure is in better agreement with independent observations. The improved TC analysis has alleviated the model shock during the early period of forecast, but the impact on intensity prediction is mixed with a better minimum sea level pressure and overestimated peak winds. We also examined the impact of two-way TC-ocean interaction on TC prediction. based on the coupled ensemble forecast from UWIN-CM. Results have demonstrated that TC-ocean coupled effect has led to weaker, smaller, more asymmetry TC, and have a northward track deflection. Analyze the coupled correlation between atmosphere and ocean variables provided us some insight of coupled model behavior in preparation for performing coupled data assimilation. In the end, the capability of UWINCM-LETKF on TC analysis is evaluated. The impact of adopting a coupled model in the forecast-analysis cycle during the atmosphere data assimilation is first discussed. Verified against the collocated atmosphere and ocean observations, the SST and near-surface temperature innovation can be more consistent when using the coupled model forecast (background field) under the TCC framework. This result has highlighted the potential of strongly coupled data assimilation, and also suggest that not only the TC assimilation but also the ocean analysis update should be performed under the TC-centered framework. The preliminary result of strongly coupled data assimilation, in which atmosphere observation is used to update the HYCOM temperature, has shown the mixed result, but the improvement can be identified in rear side of TC.
