台灣位處於西太平洋熱帶氣旋發展最旺盛的區域，每年颱風季節所造成的豪大雨雖然為台灣帶來賴以為生的水資源，但是時常的嚴重的人員傷亡以及經濟上的損失。一般水文模式模擬流量是使用實際觀測雨量搭配降雨內插方式輸入雨量資料結果，其空間分佈並不一定代表當時大氣降雨分佈情況，且觀測雨量的時間解析度較低，可能會因此造成誤差。因此本論文研究中使用雷達參數搭配降雨關係式做降雨推估，並且輸入至水文模式(CASC-2D)模擬石門水庫霞雲集水區河川逕流。 2001年納莉颱風個案使用中央氣象局五分山雷達站雷達回波資料搭配回波-降雨關係式推估降雨，降雨推估結果與觀測雨量比較有低估的現象，經由雷達修正參數(修正值3.77)修正後降雨推估結果有明顯改善。另外將雷達推估降雨輸入至水文模式模擬流量結果可以看出越上游模擬流量相較於觀測雨量搭配內插方式有較好的結果，使用雷達推估降雨顯示於空間分佈上有較好的表現。 2005年龍王颱風及2006年碧利斯颱風個案使用中央大學大氣物理研究所雙偏極化雷達資料，利用多種雷達參數推估降雨並且輸入至水文模式模擬流量。於2005年龍王颱風個案使用KDP雷達參數推估降雨降雨峰值雖然與觀測雨量比較低估40％，但相較於其他雷達參數推估降雨有較好的雨量推估結果，且輸入至水文模式模擬流量結果KDP雷達參數推估降雨較接近實際觀測流量。2006年碧莉斯颱風使用KDP、 ZH、DSD等雷達參數推估降雨皆明顯低估降雨，而ZDR雷達參數推估降雨則較接近於實際觀測雨量值。利用ZDR雷達參數推估降雨模擬流量結果，洪峰流量值高估實際觀測流量約30?40％。 The heavy rainfalls associated with typhoons often induce sereve flooding, and cause a great lose of human lives in Taiwan. The hydrological model usually takes the observational rainfall data as an input, but the spatial distribution of rainfall is not the same as the distribution of atmospheric circulation, and lower time resolution of observed rainfall data may cause errors. In this thesis research, we use the rainfall data estimates by the radar-rainfall relationship to drive the hydrological model (CASC-2D) to simulate the river-runoff. We have used the radar data of Typhoon Nari (2001) observed by the Wu-fen Shan WSR-88D weather radar. Results show that the rainfall estimated by radar-rainfall relationship is lower than the observed rainfall. A correction factor of 3.77 for the radar-estimated rainfall is suggested to use to improve rainfall estimation. The simulated river runoff using the radar-estimated rainfall is in better agreement with observed runoff in the upstream region of the Shinmen watershed area. For Typhoons Longwong (2005) and Bilis (2006), the now dual-polarmetric radar data is used, and several radar-observed parameters are used to estimate rainfall and the corresponding runoff. For Typhoon Longwong, the KDP-estimated rainfall is in better agreement with the observed rainfall, compared to estimated rainfalls by other radar parameters, although it is still under-estimated by 40%. As a result, the predicted river runoff by the KDP-estimated rainfall is in better agreement with the observed runoff. For Typhoon Bilis, the ZDR-estimated rainfall is closed to the observed rainfall, and the estimated rainfalls by other radar parameters (KDP, ZH, and DSD) are all significantly lower than the observation. The predicted peak runoff by the ZDR-estimated rainfall is higher than the observed peak runoff by 30-40% in Typhoon Bilis.
