Assessment of Landslide Potential and Its Derived Disaster in a Reservoir Watershed

Due to the climate change, the frequency of severe rainfall has been increased in recent years. In view of the landslides of the upper catchment area, deposition of sediment, and high water turbidity in the Techi Reservoir which have seriously affected the ability of power generation and water storage, and decreased the life span of the reservoir caused by the typhoon Mindulle, how to delineate and manage the hot spot of landslides of sediment disaster potential on the slope of catchment would be a subject urged to discuss. This study is based on the risk-based models for potential large-scale landslide which was proposed by Lin et al. （2017） and discussed the concept of risk and scale of the primary disaster of landslides; and the risk models for assessing the derived disasters proposed by Lin et al. （2018） to analyze the derived disasters of landslides. The grid-based risk assessment models were used in this study and validated by the disaster ratio of the watershed subdivision under the typhoon Mindulle. Considering large-scale landslides of sediment disasters will exceed the even slope unit, Getis-Ord Gi* method is adopted to map the risk of derived disasters of landslides in the catchment area, and verified by graded the value into five level with the Jenks natural breaks classification method. Setting the highest risk level as a hot spot location, and the weights of the main factors influencing the derived disasters of landslides in the catchment area are discussed. The results of primary disaster of landslides assessment in the risk-based models for potential large-scale landslide under the typhoon Mindulle showed that, the risk of landslides was positively correlated with the landslide ratio （R2 =0.71**）, and the scale of landslides was negatively correlated with the landslide ratio （R2 =0.68**）. In terms of derivative disasters, there is a significant positive correlation between the risk of derivative disasters and the disaster ratio. In this study, the polynomial regression analysis of the Getis-Ord Gi* method and the Jenks natural breaks classification method showed that the relationship between the risk of derived disasters of landslides and disaster ratio can be raised to R2=0.99**. In general, the derived disasters of landslides in the catchment are mainly compound disasters. The main types are the different situations of sediment disasters caused by headward erosion in the catchment area happen on the occurrence section, transport section, and the accumulation section. The results of disaster ratio analysis showed that the transportation section is higher than accumulation section and is lower than occurrence section due to the lowest disaster ratio caused by relatively small projected area of landslides on the transport section. In terms of derived disasters caused of landslides in catchment areas, the aspect of management and governance lies mainly in the protected objects. Value-added uses for distribution of the risk of derived disaster can be carried out as a subsequent research by mapping with the spatial distribution of protected objects, and planning a complete system of disaster relief and prevention.近年因氣候變遷導致強降雨事件頻仍。有鑑於敏督利颱風造成德基水庫上游集水區多處崩塌、庫區土砂淤積、水質濁度飆高，嚴重影響水庫發電、蓄水等功能，大幅降低水庫壽命。如何劃定集水區坡面崩塌土砂災害潛勢之熱點區位並加以治理或管理為亟需探究之課題。 本研究以Lin et al.（2017）提出之大規模崩塌劃定模式為基礎，該模式在崩塌原生災害面相之論述兼具風險與規模概念，而衍生災害之評估亦利用Lin et al. 於2018所提出之衍生災害模式進行分析。模式係以網格為基礎之風險評估，並以敏督利事件各地形分區之致災率與衍生災害風險度作為模式驗證，考量大規模崩塌土砂災害會超出均勻區單元，特利用Getis-Ord Gi*法繪製集水區衍生災害風險圖，佐以Jenks自然斷點法將值域分五級進行驗證，將風險最高等級設為熱點區位，探討集水區衍生災害主要影響因子之權重。 Lin et al.模式在敏督利事件崩塌原生災害之推估結果，顯示集水區崩塌風險度與崩塌率呈顯著正相關（R2 =0.71**），而崩塌規模與崩塌率呈顯著負相關（R2 =0.68**）。而在衍生災害方面，集水區衍生災害風險度與致災率呈顯著正相關（R2 =0.72**），本研究採Getis-Ord Gi*法和Jenks自然斷點法經多項式迴歸分析，顯示集水區衍生災害之風險度與致災率關係可提升至R2=0.99**，一般而言，集水區衍生災害主要為複合型災害，其主要類型為集水區向源侵蝕造成土砂災害發生段、輸送段、以及堆積段之不同災情，其致災率分析結果顯示，發生段>堆積段>輸送段，因輸送段地形較陡峭造成崩塌致災投影面積較小故致災率最小，其主要受地形影響所致。以集水區衍生災害而論，在管理與治理之面相，主要在於保全對象，後續研究可將衍生災害風險分布做加值利用，套疊保全對象之空間分布，規劃完整之防救災體系。摘要i Abstractiii 目錄v 圖目錄vii 表目錄x 第一章 前言1 第一節 研究動機1 第二節 研究目的2 第三節 研究架構與研究流程4 第二章 文獻回顧6 第一節 土砂災害之定義與分類6 第二節 衍生災害與其發生成因9 第三節 崩塌因子之相關研究11 第四節 最適分析單元之劃分13 第五節 風險度之相關研究與沿革15 第六節 危害度之來源18 第七節 脆弱度相關方法之探討19 第八節 崩塌規模之相關研究21 第九節 土砂災害鏈之相關探討22 第十節 影像處理精確度之相關探討23 第三章 材料與方法25 第一節 樣區概述25 3.1.1 地形25 3.1.2 地質25 3.1.3 土地利用27 3.1.4 交通28 第二節 研究材料32 第三節 研究方法34 3.3.1 正規化34 3.3.2 危害度因子34 3.3.3 脆弱度因子36 3.3.4 崩塌規模因子44 3.3.5 泥砂遞移率49 3.3.6 地形濕度指標51 3.3.7 崩塌與致災區位萃取52 3.3.8 熱點分析（Getis-Ord Gi*法）54 3.3.9 Jenks自然斷點法58 第四章 結果與討論59 第一節 致災區位萃取成果59 第二節 原生災害之驗證61 第二節 衍生災害之驗證67 2.2.1 衍生災害風險度67 2.2.2 驗證結果69 第三節 重點管理區位71 第五章 結論與建議75 參考資料76 附錄8