2010-01-012024-05-15https://scholars.lib.ntu.edu.tw/handle/123456789/663312摘要：河川水位預報準確度通常是吾人用來評價預報模式優劣的指標，然而水位預報誤差實際上是「降雨預報不確定性」、「洪水預報模式不確定性」和「模擬條件不確定性」等各項誤差累加的綜合表現。其中，洪水預報模式與模擬條件不確定性，包括率定曲線不確定性、模式參數不確定性與河川斷面不確定性等，這部份誤差，可透過事先檢定與即時修正演算，將其影響控制在可接受的範圍內（大規模山崩和土石流除外）。此外，現有數值天氣預報技術仍多科學限制，在台灣集水區空間尺度、小時時間尺度，尚難降到與「洪水預報模式不確定性」和「模擬條件不確定性」相同的水準；氣候法颱風定量降雨模式推估的降雨產品，事件集水區總量準確度因颱風個案而異，在大氣物理條件欠缺的限制下，使其無法在小時時間尺度上準確。本計畫以短期間內無法排除定量降雨預報誤差為前提，擬研發機率式颱風降雨預報，將不同機率的降雨歷線輸入逕流、洪水演算模式，產出不同累積機率的河川水位預報，供防洪作業決策之參考。<br> Abstract: The accuracy of predicted water level is generally used to evaluate forecasting model. However, forecasting error is the aggregation of uncertainties which result from precipitation forecast error, flood model error and the unknowns of simulation conditions. The errors from flood model and the unknowns of simulation conditions, such as uncertain rating curve, inaccurate model parameters and deformed river bed, are manageable within acceptable deviation by performing statistical tests and real time adjustment, except for the errors from tremendous landslides and mudflows. Moreover, numerical weather models nowadays are limited by many scientific constrains, as is difficult to compete flood models which even contain the unknowns of simulation conditions. The quantitative typhoon rainfall estimations which are made from climate method are dominated by individual case, and are not accurate in 1-hour time scale because of the lack of atmospheric conditions. In this project, it is assumed that the quantitative rainfalls estimations remain frustrate in the short-term future, and a probabilistic typhoon rainfall forecast model will be developed. Different probabilistic rainfall histograms are input into run-off and flood models, as yields different accumulated probabilistic water levels, which could be referred by decision-making officers.河川水位預報水位預報誤差機率式颱風降雨預報forecasted water levelforecasting error of water levelprobabilistic typhoon rainfall forecast model