2009-05-012024-05-14https://scholars.lib.ntu.edu.tw/handle/123456789/660752摘要：由於台灣地形多山，河川短而湍急，造成集流時間很短，也造成每遇挾帶大量降水的颱風侵襲，便極易使下游遭受嚴重的洪水災害。此外，臺灣本島位於西北太平洋颱風的主要路徑上，每年平均遭受颱風侵襲3-4 次。若能作好防災措施，颱風所帶來的降雨可以是最珍貴的水資源，否則颱風降雨將形成嚴重的災害。對臺灣而言，增進颱風洪水預報能力一直是洪水災害管理的重大任務，因此，如何改善颱洪預報，特別是如何提升長延時預報的準確性，以提供下遊民眾足夠的反應時間，對台灣而言更顯重要。但想要了解颱風現象與洪水災害兩者背後的物理機制，並藉此發展一套物理模式卻非常不容易。本計畫利用類神經網路來建構四個包含不同輸入項的颱風洪水預報模式：(a) ANN1 僅以前期流量作為輸入項；(b) ANN2以前期流量及前期雨量作為輸入項；(c) ANN3 的輸入項則整合前期流量、前期雨量以及颱風特徵; (d) ANN4 及以前期流量與颱風特徵作為輸入項。完成模式建構後，以實際颱風事件進行測試，並比較四個模式的結果，以了解颱風因子對於颱洪預報效果的影響。經由分析及比較，成果可顯示颱風特徵能明顯地改善洪水預報的效果。<br> Abstract: In Taiwan, floods caused by typhoons often result in serious disasters, because the time of concentration is short. In addition, the island of Taiwan is situated in one of the main paths of northwestern Pacific typhoons. On average, three to four typhoons attack the island each year. Typhoon rainfall can be a most valuable resource if proper disaster-mitigation measures are made. Otherwise, it can cause serious damage. Improving typhoon-flood forecasting is always an important task of flood management in Taiwan, especially for long-lead time forecasting. However, in order to develop a physically based mathematical model for typhoon-flood forecasting, one must know the behaviors of the physical process, which is not an easy task. In this project, artificial neural networks (ANNs) are used to construct typhoon-flood forecasting models. Four typhoon-flood forecasting models with different inputs are constructed. Four types of model inputs are: (a) antecedent runoff only (ANN1), (b) antecedent runoff and rainfall (ANN2), (c) antecedent runoff, rainfall and typhoon characteristics (ANN3), and (d) antecedent runoff and typhoon characteristics (ANN4). Then, four typhoon-flood forecasting models are applied to actual typhoon events. To investigate the influence of the typhoon characteristics on the flood forecasting, comparison of the model performance is performed. Finally, summary and conclusions are made to demonstrate that the typhoon characteristics can improve the flood forecasting.颱風因子類神經網路流量預報typhoon characteristicsartificial neural networkstyphoon-flood forecasting