Application research of potential large-scale landslide model using sensitivity analysis

Under the climate change, extreme climate-induced disasters have increased the scale and the frequency of occurrence. The excessive rainfall of Morakot Typhoon triggered the serious landslides disaster in Taiwan. Previous study established the model for the potential of landslide risk/scale assessment using the integrated environmental indicator. However, the suitability of the factors considered in the model is worth exploring. In this study, the important of the landslide risk/scale factors were determined through the sensitivity analysis to understand and discuss on the influence of each factors in the model. Taiwan area with the landslides disaster as a result of Morakot Typhoon was selected as a case study. According to the potential large-scale landslide model, sensitivity analysis was examined the factor reduction to understand an amount contribution of each factor to the overall output of regression analysis. The potential landslide risk/scale was categorized into low, medium and high level of danger grade using K-means cluster analysis. Then the relationship between the danger grade and the protection target areas will be discussed as a reference for the further management of relative authorities. The results show the sensitivity of landslide risk factors which could be prioritized as follow: return period > hazard (Maximum daily rainfall of Typhoon Morakot) > green deterioration > road development. Besides, the sensitivity of landslide scale factors could be ranked as follow: headward erosion > river concave > dip slope. The maximum daily rainfall has a large contribution to the landslide risk model and can be used as a major factor to simplify the model. Also, the index of headward erosion shows the significant influence in potential landslide scale because of the site of large-scale landslide is generally located at the headwater areas with well-growth forest and depth soil.在氣候變遷下，極端氣候誘發的災害規模與頻率皆會增加。有鑑於莫拉克極端降雨事件對於台灣造成嚴重土砂災害，過去研究以巨量數據將各個環境指標進行崩塌發生機率及規模模式之建置。然而這些模式所考量參數之適宜性值得探討。本文以敏感度分析探討崩塌機率及規模模式之適宜性，瞭解模式內各因子對模式的貢獻程度並進行討論。 本文以台灣作為研究樣區並以莫拉克颱風作為災害事件。將崩塌和崩塌規模潛勢模式，以敏感度分析將模式內因子逐步剔除並進行迴歸分析，以了解各因子之貢獻程度。另將崩塌發生機率和規模之潛勢，透過K-means群集分析分為低、中、高三種危險等級再與保全對象探討其位向關係，可作為有關單位未來管理之相關依據。 結果顯示，崩塌發生機率模式之指標因子敏感度為：極端降雨>危害度(莫拉克一日最大降雨量) >植生綠劣>道路開發。極端降雨為對此模式貢獻最大，可作為崩塌潛勢之依據；而影響崩塌規模之指標因子，其敏感度分析結果為：向源侵蝕>順向坡>河道凹岸。在崩塌規模潛勢中，以向源侵蝕對於此模式貢獻最大，主因為大規模崩塌的區位多位於植生生長良好土砂料源較多之源頭集水區。摘要 I ABSTRACT II 目錄 IV 圖目錄 VI 表目錄 VIII 1 第一章 前言 1 1.1 研究動機 1 1.2 研究目的 1 1.3 內容架構與研究流程 2 2 第二章 文獻回顧 2 2.1 崩塌之定義及分類 2 2.2 崩塌影響因子 3 2.3 崩塌風險之研究 4 2.4 崩塌規模之研究 5 2.5 土砂災害對保全對象之影響 7 2.6 敏感度分析 9 3 第三章 研究材料與方法 11 3.1 研究樣區概述 11 3.2 研究材料 11 3.3 研究方法 13 3.3.1 集水區分析單元劃定 13 3.3.2 正規化法 15 3.3.3 崩塌風險分析 15 3.3.4 崩塌規模分析 27 3.3.5 崩塌潛勢因子敏感度分析 34 3.3.6 危險等級劃定 34 4 第四章 結果與討論 35 4.1 敏感度分析結果 35 4.2 崩塌潛勢分析結果 40 4.3 崩塌規模分析結果 42 4.4 危險等級劃定結果 44 4.5 現地勘查 46 4.5.1 崩塌潛勢重點區位現勘結果 46 4.5.2 崩塌規模重點區位現勘結果 48 4.5.3 危險度重點區位現勘結果 50 第五章 結論與建議 52 5 參考文獻 53 6 附表一 崩塌因子相關研究 58 7 附表二 水利署雨量資料 6