DOTSTAR策略性觀測投落送資料對於颱風路徑模擬影響研究

Abstract

為增加對於颱風內部以及周圍環境大氣的觀測資料，除了有賴於觀測範圍較廣大的衛星資料外，具有能夠得到垂直解析度上較高之大氣參數的投落送亦是重要的觀測工具。台灣自2003年開始進行「侵台颱風之飛機偵查及投落送觀測實驗」，英文簡稱DOTSTAR，以取得接近台灣的颱風周圍大氣之投落送觀測資料，並且配合策略性敏感方法，以改善路徑預報與增進颱風分析研究。 DOTSTAR團隊在每一次飛行任務執行前，會參考4大性質不同的統計及動力架構之策略性敏感方法以規劃合適的觀測飛行路徑，而主要的取樣策略可分為颱風內圈與額外敏感區域兩種方式。為了探討全部、颱風內圈以及額外敏感區域等3種不同取樣方法的投落送資料對於路徑模擬的影響並得到統計分析，本研究利用MM5 3DVAR資料同化方法選取2004至2006年的DOTSTAR觀測個案進行模擬實驗與評估，並選取珊珊颱風(2006)做進一步個案之投落送資料的影響探討。此外也針對颱風內圈一半與颱風內圈四象限各1枚投落送資料的敏感性進行討論。 研究結果顯示同化全部的投落送資料在6至72小時平均能夠大幅改善路徑模擬達24%，而僅同化颱風內圈的投落送資料亦對路徑模擬有正面的改善，但僅同化額外敏感區域的投落送資料則改善路徑模擬的程度較不明顯。而僅同化颱風內圈與同化颱風內圈一半的投落送資料對於路徑模擬的結果相似；僅同化颱風內圈四象限各1枚相較於同化額外敏感區域的投落送資料更能有效改善路徑模擬。 由珊珊颱風個案的模擬則發現，僅同化額外敏感區域相較於同化颱風內圈的投落送資料更能改進路徑誤差，此結果與上述的統計分析相異。然而無論同化額外敏感區域或是颱風內圈的投落送資料皆無法反應珊珊颱風同化全部的投落送資料的後期路徑，唯有適當結合颱風內圈與額外關鍵的投落送資料才能有效的掌握此路徑特徵。

In order to increase the atmospheric observations of typhoons over the ocean region, besides the satellite data, the data from GPS (Global Positioning System) dropwindsonde lunched by the surveillance aircraft are also important. Under the support of the National Science Council (NSC) and Central Weather Bureau (CWB), the synoptic surveillance missions to improve TC track forecasts has been conducted by DOTSTAR (Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region) in the western North Pacific Ocean since 2003. Four different sensitivity methods have been employed as the guidance to design flight routes and deployment locations of GPS dropwindsondes for the typhoon‘s synoptic surveillance in DOTSTAR.. The two main strategies of the targted observations are ‘around-storm’ and ‘extra-targeted area’. In this research, the impact of dropwindsonde data on typhoon track simulations under different observing strategies is studied. MM5 3DVAR data assimilation system is adopted to assimilate the dropwindsonde data from DOTSTAR during 2004 to 2006, and to asses the statistical impact on typhoon track simulations. It is shown that inclusion of all dropwindsonde data in MM5 3DVAR can effectively reduce the 6 to 72-h track forecast error by about 24%. By only assimilating the around-storm dropwindsonde data, the simulation of typhoon tracks can also be improved. On the contrary, assimilating the extra-targeted dropwindsonde data shows slightly improvement of typhoon track simulation. Typhoon Shanshan (2006) is selected for the case study. Consistent with the results based on the statistics of multiple cases, assimilating all dropwindsonde data can effectively reduce the track forecast error. Overall, the improvement of track by assimilating the extra-targeted dropwindsonde data is more significant than that by assimilating the around-storm dropwindsonde data. Moreover, the combination of around-storm and appropriate extra-targeted dropwindsonde data shows the most track forecast improvement with the northward movement in the later period. Assimilating only half of the around-storm dropwindsonde data shows similar results to the experiments assimilating all around-storm dropwindsonde data. Comparing to the results of assimilating the extra-targeted dropwindsonde data, experiments by assimilating four of the around-storm dropwindsonde data in each quadrant can have more improvement on the track simulations.