- 期刊
芙蘿颱風(1990)的數值模擬研究：初始資料及初始化方法影響之探討
Numerical Simulation of Typhoon Flo (1990) Using a Non-Hydrostatic Mesoscale Model: The Impact of Initial Data and Initialization Schemes
摘要
本研究以美國國家大氣科學研究中心與賓州大學合作發展的第五代中尺度靜力/非靜力模式MM5為模擬工具，採用不同來源的初始資料，包括日本氣象廳(JMA)、美國國家環境預報中心(NCEP)及歐洲中期天氣預報中心(ECMWF)的網格分析資料，針對1990年的芙蘿颱風(Flo)進行數値模擬。由於全球分析資料所解析的颱風範圍過大且強度太弱，在模擬前須透過初始化方法植入渦旋。本研究採用不同初始化方法針對芙蘿颱風進行模擬，包括：GFDL颱風模式初始化方法(Kurihara et al.,1995)、Kuo and Wang(1997)植入方法，前者的優點在於定義較佳的環境場，後者的優點在於由結構簡單的渦旋經預先模擬而產生與模式動力一致的渦旋。本研究亦結合前兩者優點而設計一系列的新方法，針對不同資料、不同初始化方法及不同初始時間進行模擬測試，以瞭解不同初始條件對於模擬結果的影響。 模擬結果顯示不同初始資料及初始化方法所得之模擬路徑有明顯的差異性，其中採用不同初始資料（JMA與ECMWF）及不同初始化方法，因初始場及模擬時對太平洋副高的趨勢掌握不同，導致對Flo颱風路徑的模擬有較大的差異。而使用相同初始資料(ECMWF)但採用不同的初始化方法，由於初始時間採用不同的渦旋植入過程，造成颱風中心附近流場改變，模擬的路徑亦有相當程度的差別。大部分的模擬結果顯示，因事先已濾除分析場中不正確之颱風訊息，故以結合渦旋與經濾除颱風分量的環境場之模擬結果較佳。又所植入渦旋必須經由預先模擬過程而產生，否則開始模擬時會因渦旋結構與模式動力不一致而需一段調整期，導致模擬路徑不佳。至於強度模擬方面，雖然模擬實驗可以掌握高層冷心低壓接近芙蘿颱風的過程，並與Wu and Cheng (1999)的分析結果相似，但是各模擬實驗皆無法反映芙蘿颱風迅速增強的情況，此結果正凸顯出現階段模式掌握颱風强度演變所面臨的難題。 總之，模擬路徑對於不同初始資料及初始化方法有相當程度的敏感性，因此模擬時需採用適當的初始化方法，才能得到合理的初始場。本研究所發展的初始化方法應可提供做為日後進行颱風模擬研究的參考。
並列摘要
In this study, Typhoon Flo (1990)was simulated using the Penn State-NCAR nonhydrostatic, two-way interactive, triply nested grid mesoscale model (MM5). The initial conditions for the experiments were analyzed from three operational centers, including JMA, NCEP, and ECMWF. To improve the representation of Flo, different initialization schemes were adopted, including the initialization scheme of GFDL hurricane model, KW bogusing scheme (Kuo and Wang 1997),and four new schemes designed in the study. In this paper, we performed numerical simulations to investigate the impact of initial data and initialization schemes. Results from the comparison of the model experiments indicated that using different initial conditions and initialization schemes led to different simulated track performance. The deviation in the experiments of using JMA and ECMWF analysis data as initial condition was due to the different simulated tendency of Pacific sub-high, and that of using the same initial condition(ECMWF)but different initialization schemes was resulted from the initially modified flow at the center of the typhoon by different initialization schemes. Most experiments indicated that simulated track performance was improved by merging the vortex with the environment flow that has removed unrealistic typhoon circulation resolved by the global analysis data. Moreover, the bogusing vortex has to be generated from pre-run process to ensure dynamical consistency in the vortex structure, as well as the compatibility of the vortex to the resolution and physics of the model. Although the cold core low in the vicinity of Flo could be reproduced which is consistent with Wu and Cheng (1999), the model failed to capture the intensity evolution of Flo. The result suggested that the simulation of tropical cyclone intensity remains a highly challenging task.