台灣位於西北太平洋地區颱風侵襲的主要路徑，平均而言，每年會遭遇數個颱風，其能為台灣帶來的豐沛雨量，但往往也造成嚴重災情。陳(2010)根據颱風位於相同位置時，降雨分佈會因地形效應導致相似降雨分佈之特性，藉以修正熱帶氣旋降雨潛勢技術(Tropical Rainfall Potential Technique,TRaP)，但其方法包含不同路徑(北行與西行)的颱風降雨分佈，因而會造成誤差。 因此，本研究目的即為針對地形作用與環境風場對西行侵台颱風降雨的影響，建立TRaP估算降雨與中央氣象局自動觀測站時雨量關係，再根據陳(2010)方法進行降雨分佈修正與降雨總量修正，以改進降雨潛勢預報的精準度。 由1997年至2008年的25個非獨立颱風個案所得結果顯示陳(2010)的方法已能大幅改善原始TRaP的精度。而本研究亦能改善原始TRaP估算之結果，使其平均相關係數由0.08改善至0.69，均方根誤差則由171.6mm降至120.3mm，平均偏差量則由-25.8mm下降至2.9mm。而在獨立個案2009年莫拉克颱風的結果顯示，本研究與陳(2010)的相關係數皆達0.9，但在均方根誤差則能由陳(2010)的305.7mm降至256.4mm，平均偏差量也由-129.8mm降至-74.3mm，可見有所改進。因此在台灣地區應用TRaP估算颱風降雨時，颱風路徑也是必須考量的因素之一。Taiwan is situated within one of the frequent tropical cyclone tracks around Pacific Ocean. Such storms not only bring forth abundant rainfall, but also cause serious damages. Using typhoon cases that shared relatively the same typhoon positions, which would yield similar rainfall distributions in Taiwan, Chen (2010) sought to improve the tropical rainfall potential technique (TRaP). But this method didn’t consider the different rainfall distribution between westward and northward-moving typhoons.This study mainly analyzes westward-moving typhoons causes that made landfall in Taiwan. Chen’s (2010) method is employed by further considering the terrain effect to improve TRaP in Taiwan. The results of this study obviously improved when compared to the original TRaP. For 25 dependent cases from 1997 to 2008, the mean correlation coefficient increased from 0.08 to 0.69. Meanwhile, the root mean square error decreased from 171.6mm to 120.3mm and the mean bias decreased from -25.8mm to 2.9mm. For typhoon Morakot in 2009, the result of this study was able to obtain a better performance over Chen (2010). Although the correlation coefficient was both equal to 0.9, the root mean square error could decrease from 305.7mm to 256.4mm and the mean bias could drop from -129.8mm to -74.3mm. Consequently, the typhoon tracks should be considered when seeking to improve the TRaP technique in Taiwan.
