Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/13266
標題: 濁水溪流域短期降雨預測模式之探討
A Short-Term Rainfall Prediction Model on Jhuoshuei River Basin
作者: 陳文正
Chen, Wen-Cheng
關鍵字: mudflows;土石流;short-term rainfall;analogical precipitation;Grey theory;短期降雨;類似降雨;灰色理論
出版社: 土木工程學系
摘要: 
近年來經常因為土石流與河川洪汛之肆虐,造成國人生命及財產之重大損失,因此對於導致土石流與洪汛之暴雨預測顯得更加重要。有關降雨之預測,2002年經濟部水利署建置「濁水溪流域逕流測預報系統」,對短期降雨之預測,採用類似降雨查詢模式及灰色理論預測模式。由於降雨預測結果是降雨-逕流演算之主要輸入項與土石流預警參考指標,所以雨量預測之準確性甚為重要。
本研究針對該系統之灰色理論預測模式、類似降雨查詢模式作探討;原始的GM(1,1)外差模式是以四點為理論基礎,於此文中加入三點預測模式,且配合移動平均及累積方式處理前期資料作預測;類似降雨查詢模式修正以臨前降雨時段、包含/不包含颱風資料、不同累積資料時間及次集水分區等方式進行預測。兩種模式預測結果與實際雨量值進行比較,探討不同方式預測值的準確度,希能得到更準確之預測結果,以提供防災預警資訊之應用,減少因暴雨所造成災害之損失。
研究結果發現,於灰色理論預測中,GM(1,1)三點外差模式較四點外差模式之預測效能為佳,且於第1小時後的預測,以累積方式預測效果較好,建議採三點平均累積方式取代原有的四點預測模式。至於第2小時後的預測效能,以修正類似降雨查詢模式較灰色三點平均累積方式為佳,由於兩種模式並無降雨物理作基礎,以致預測時段以降雨延時3小時為宜,且預測結果於實際應用時,須有輔助決策機制,以增加決策者信賴水準。

In the past few years, debris flow had caused heavy losses of lives and properties. Therefore, it becomes important to predict rainfalls which may cause mudflows. In 2002, the Water Resources Agency, Ministry of Economic Affairs set up the “runoff prediction system for the Jhuoshuei river basin” to predict short-term rainfalls using an analogical precipitation prediction model and a Grey theory prediction model. The accuracy of rainfall predictions is very important because the predicted rainfall amount is one of the most important input and indices for the rainfall-runoff calculations and debris flow warning systems.
This study only focuses on the Grey theory and analogical precipitation prediction models. The original GM(1,1) prediction model requires four items of data. In this study, we add the conjectured GM(1,1) three data model which combines the moving average and data accumulation techniques to make predictions. For the analogical precipitation prediction model, we use different periods of time, data contenting typhoons (if any), accumulative prediction results (if any), and different sizes of area to make predictions. The predicted results were compared with the real rainfall amount. Helpfully the rainfall related disasters can be reduced with these revised models.
In this study, we found that the GM(1,1) three data model had better prediction results than the four data model. Predictions using the accumulative data also gave better results for the first hour. Here, we suggest the GM(1,1) three data model with the moving average and data accumulation techniques to replace the GM(1,1) four data model. As for the predictions after the second hour, the analogical precipitation prediction model gave better results than the GM(1,1) model. These two prediction methods were not based on the hydrological physics, so, we suggest not making any predictions for more than three hours. When using the prediction results, we also suggest using other strategic decisions to assist in making a final, accurate decision.
URI: http://hdl.handle.net/11455/13266
Appears in Collections:土木工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.