颱風數值模擬之現況與挑戰－2009年莫拉克颱風

Typhoon Morakot (2009) and a Special Review on the Current Status and Future Challenge of Tropical Cyclone Simulation

吳俊傑(Chun-Chieh Wu) ； 黃清勇(Ching-Yuang Huang) ； 楊明仁(Ming-Jen Yang) ； 簡芳菁 (Fang-Ching Chien) ； 洪景山(Jing-Shan Hong) ； 顏自雄(Tzu-Hsiung Yen)

《大氣科學》 38卷2期 (2010/07) Pp. 99-134

摘要

過去30年來，隨著電腦計算速度與資源的進步，數值天氣預報呈現突飛猛進的發展，在颱風預報方面，透過資料同化及渦旋植入等技術，改進了模式初始分析場中的颱風結構，並配合大量非傳統（如衛星、飛機、雷達）觀測資料的運用，使得颱風路徑預報的準確度有大幅的進步。
然而，除移動路徑外，對於其他颱風相關預報，如強度、結構、降水等，在多種大氣尺度與物理因素的交錯影響下，仍具有一定的困難度。莫拉克颱風(2009)於侵台期間帶來將近3000公釐的豪雨，造成高屏地區的洪水氾濫以及土石流，導致超過700人的死亡，台灣南部居民之生命及財產損失十分慘重。莫拉克颱風降水預報低估的原因包括：模式無法確切掌握颱風移速，導致模擬颱風迅速通過台灣而造成預報的雨量不足、加上地形因素及廣大洋面上資料缺乏，造成環境流場的模擬與實際有很大差異等。
颱風數值預報的重點，包括路徑、強度與結構及降水和風雨分布等。颱風模式之選取則依研究目的而定，但不論是何種模式，颱風渦旋初始化及資料同化之發展都是非常重要的一環。此外，系集預報之發展、解析度之選取、改善模式的物理參數化方案、藉由不同模式之比較以改進現有之颱風模式、及進一步瞭解地形影響及海氣交互作用效應等，皆是值得繼續深入研究之重點議題。

關鍵字

颱風數值模擬 ； 莫拉克颱風 ； 渦旋初始化 ； 資料同化 ； 系集預報

並列摘要

In the past 30 years, with the progress in computational power, the numerical weather forecast has developed rapidly. The advancement in data assimilation and bogusing has tremendously improved the model initialization of tropical cyclone (TC) structures and together with the extensive application of remote sensing (satellite, airplane observation, and radar), the accuracy of the forecast of TC tracks has steadily improved.
Nevertheless, except for TC motion, the complex multiple-scale physical processes essentially limit the predictability of TC intensity, structure and precipitation. In 2009, Typhoon Morakot devastated Taiwan with torrential rain of nearly 3,000 mm which led to the flooding and mudslides in southern Taiwan (Kaohsiung and Pingtung County) and caused over 700 dead and the significant loss of properties. The underestimation of precipitation of Typhoon Morakot could be attributed in several factors. First, the incorrect forecast of translation speed before making landfall and an underestimation of rainfall related to topographic effect. Second, poor moisture content prediction in the forecast model may have taken another role. Third, sparse observation data over the ocean degraded the simulated environmental flow and contributed to a biased simulation.
It is recommended that the numerical forecast of TCs should focus on tracks, intensity and structure and the distribution of precipitation and gale force wind. Although the TC models chosen depends on the purpose of the research, the development of vortex initialization and data assimilation is an important step in every model. In addition, the advance in ensemble prediction, the selection of resolution and parameterization of physical processes are issues worth of further investigation. Systematic comparison of different models, better understanding of the effect of topography and the air-sea interaction could improve the present models and are subjects of primary research interest.

並列關鍵字

Numerical simulation ； Typhoon Morakot 2009 ； Vortex initialization ； Data assimilation ； Ensemble forecast