DOTSTAR投落送資料與其他衛星觀測資料對颱風 路徑模擬影響之探討 ─ 梅姬與卡努颱風個案研究

Abstract

為有效增進西北太平洋地區颱風周遭環境大氣資料的觀測，台灣自2003年起正式展開「侵台颱風之飛機偵察及投落送觀測實驗」，簡稱DOTSTAR (Wu et al. 2005)，希望藉由投落送觀測資料來增加對颱風環境流場的掌握，增進颱風模擬之準確度。除了投落送資料外，觀測範圍相當廣泛的衛星資料亦是增進對環境場掌握的重要資源，且亦有助於颱風模擬之改善。另外，在模式初始場中植入一虛擬渦旋將可重新建立颱風內部結構，亦有助於改善颱風強度及路徑之模擬。 本研究為探討植入渦旋、投落送資料、QuikSCAT風場及衛星雲導風場分別對於颱風路徑與強度之模擬有何影響，利用MM5 3DVAR同化系統，選取DOTSTAR所觀測的梅姬 (Megi 2004)與卡努(Khanun 2005)兩颱風，進行一系列資料同化與數值模擬實驗。此外，本研究亦針對觀測誤差之敏感性、植入渦旋大小之敏感性以及移除近颱風中心QuikSCAT風場等部分進行測試與探討。 研究結果顯示，未進行渦旋植入的各組實驗，其模擬的颱風強度皆太弱，颱風垂直結構相當鬆散。這也使得颱風受到駛流的深度和真實的情形不同，因此路徑模擬有較大的誤差。而當植入渦旋之後此一現象大為改善，兩颱風在有植入渦旋的五組實驗中，其路徑移動的掌握都較未植入渦旋的五組實驗佳，此一結果顯示渦旋植入於模擬實驗中扮演了關鍵性的角色。 在梅姬颱風個案中，植入渦旋後同化投落送資料的路徑模擬結果最佳，植入渦旋未同化任何資料實驗結果次之。而卡努颱風則以植入渦旋未同化任何資料的實驗颱風路徑模擬最佳。在敏感性實驗的部分，於MM5 3DVAR系統中對於觀測誤差的選取是有敏感性存在的，而對於植入之渦旋大小的敏感性則不顯著。在移除近颱風中心3度內QuikSCAT風場資料的實驗結果顯示當完全剔除誤差可能較大的近颱風中心風場資料後，其路徑模擬反而較差。故針對QuikSCAT風場資料的品質控制，仍需進行更多實驗與研究來找出最佳的資料品質控制方法。

Starting from 2003, a new typhoon surveillance program, DOTSTAR (Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region) was launched (Wu et al. 2005). It’s expected to enhance the number of observations in the environment of hurricanes threatening North-West Pacific area. When dealing with typhoon simulation data, it is obvious that both dropwindsonde data and satellite data are very important for the betterment of typhoon simulation. During the process, we found another method that might be able to approach to the real typhoon initial structure and improve the prediction at the same time－implanting the bogused vortex. According to the experiment results, the typhoon strength is going to be weaker and the typhoon vertical structure is easier to decay without considering the bogused vortex. All the results point out that bogused vortex plays an important role in typhoon simulation. The results reveal that the typhoon strength is weak and typhoon vertical structure is quite decay if the experiments are without implanting the bogused vortex. Therefore, this result presents that the bogused vortex in typhoon simulation plays a very important role. In the case study of the typhoon Magi (2004), we use both the dropwinsonde data and the bogused vortex in setting of the model initial field. And the result shows that it is a method in possibilities for improving the forecasting in typhoon intensity and possible tracks prediction. Even only consider the bogused vortex without assimilating any observation data, it still comes out a better track predictions. In the other case study of the typhoon Khanun(2005), the results point out that only consider the bogused vortex without assimilating any observation data also could comes out a better track simulation. For Megi(2004), appropriate combination of dropwinsonde data and the bogused vortex in the model initialization have the potential to improve both the track and intensity forecasts. Even only consider the bogused vortex without assimilating any observation data, it still comes out a better track predictions. In the other case study of the typhoon Khanun(2005), the results point out that only consider the bogused vortex without assimilating any observation data also could comes out a better track simulation. In sensitivity experiments, the simulation results reveal that MM5 3DVAR assimilation system has the sensitivity for observation inaccuracy (O) , but it’s not obvious for the sensitivity of the bogused vortex size. After removing the QuikSCAT winds near the typhoon center, the result shows the typhoon track becomes worse. More experiments and researches should be done in order to find out the best way for data quality control of QuikSCAT winds.