2006-07-282024-05-17https://scholars.lib.ntu.edu.tw/handle/123456789/684088摘要：臺灣因經緯度位置因素，在夏秋兩季常受到颱風侵襲而且造成嚴重的洪水災害，為保護民眾的生命財產安全、減低洪水災害與損失，發展洪水預警系統是一項重要的研究課題。然而，預警系統之發展需根基於有效率及高精度的洪水預報模式才能及時獲知正確的資訊。大漢溪流域中之石門水庫集水區、北勢溪流域中之翡翠水庫集水區與基隆河流域中之員山子分洪集水區在北臺灣均屬於重要的集水區域，這些集水區的下游區域均為人口居住密集與經濟活動頻繁的地區，若是這些區域遭受洪水侵襲將造成十分嚴重的損失。因此，本計畫分別針對這三個集水區域建置洪水預報模式，這些集水區均為淡水河流域中之子集水區域，相對於整個淡水河流域屬於較小型的集水區域，這些較小型的流域通常集流時間較短，亦即洪峰到達時間極短，使得對於該種流域之洪水預報的難度大幅提高，而且，對於洪水預報的效率及精度之需求也遠大於大型流域。 非線性計算單元串聯模式（nonlinear computational units cascaded model, 簡稱NCUC模式）為一有能力的洪水預報模式，其改善了目前常用洪水預報模式的缺點，包括預報精度不足、無參數自動檢定法以及必須實測流量資料作為模式輸入等缺點，而且NCUC模式的架構可依據使用者不同的使用需求而調整，十分擁有彈性。為能建置有效率及高精度的洪水預報模式之目的，本計畫應用NCUC模式針對石門水庫集水區、翡翠水庫集水區與員山子分洪集水區分別架構出適用於三個集水區域之颱風洪水預報模式。由模式之實際驗證結果顯示，以NCUC模式為基礎所建置之洪水預報模式，在颱風期間對於洪水量之預報有良好的能力。本研究所得之成果可供各集水區管理單位作為發展集水區洪水預報系統之根據，並可提供相關決策者做為防災工作之重要參考。 <br> Abstract: Because of the geographical location, typhoons often invade Taiwan during summer and autumn, and cause severe flood damage. For protecting the civil life and reducing flood disaster, it is an important research to develop a flood warning system. A flood warning system should be developed based on a flood-forecasting model, which works efficiently and has a well performance. The Shihmen reservoir watershed, the Feitsui reservoir watershed, and the Yuan-Shan-Tsu diversion watershed are critical areas in northern Taiwan. At the downstream regions of these watersheds, the population is concentrated and economic activities thrive. If these regions are invaded by flood, severe loss of property and life will be caused. Therefore in this project, flood forecasting models are built for the three watersheds. The three watersheds are all subwatersheds of the Danshui river watershed and are small watersheds relative to the Danshui river watershed. For small catchments, the time to peak discharge is short owing to the small area. Therefore, it is difficult to forecast the flood in small catchments. In order to provide enough time for the decision makers to make the proper decision, an efficient and well-performed model is needed for small catchments. To improve the prediction of flood, a new flood-forecasting model based on the nonlinear computational units cascaded model (NCUC model) is proposed by Lin and Wang in 2007. The flood forecasting model based on NCUC model is a capable model. It overcomes the shortcomings of current flood forecasting model that are insufficient accuracy, without automated calibration method and input requirement of runoff data. Moreover, the pattern of the NCUC model can be adjusted according to the users’ requirement. The purpose of this research is to apply the NCUC model to forecast the flood in the Shih-men Reservoir catchment, the Feitsui Reservoir catchment, and the Yuan-shan-tzu catahment. The results show that the performance of the NCUC model is outstanding. It is concluded that the NCUC model is useful to forecast flood and the results of this project are helpful for the decision makers to support the decisions in disaster-mitigation.小流域洪水預報非線性計算單元串聯模式small catchmentsflood forecastingnonlinear computational units cascaded model