本研究針對大漢溪流域之匹亞溪集水區實地量測土壤深度，建立土壤深度與坡度間的線性關係式，並以數值地形（DTM）計算坡度及推估全區之土壤深度分佈。利用艾利颱風誘發山崩資料，根據結合水文模式的無限邊坡理論搭配植生覆蓋指數（NDVI）對有效凝聚力的影響進行反算，獲得集水區內各地層的強度暨水文參數。反算過程是以高義雨量站的資料，讓研究區在災前常態下，大多為穩定，而在降雨引發崩塌的事件下，可能發生崩塌的地區與真正發生崩塌的位置大致相符。 研究中同時使用SHALSTAB和TRIGRS兩種模式進行降雨誘發淺層崩塌的模擬分析，再將模擬結果與實際崩塌資料進行比對，據以檢視這兩個模式的適用性。SHALSTAB和TRIGRS兩種模式的分析結果所得到的總體正確率分別為85.6%和87.2%，成功率曲線的曲線下面積（AUC）分別為0.811和 0.801。從總體正確率得知，TRIGRS 略高於SHALSTAB，而成功率曲的曲線下面積，SHALSTAB 略高於TRIGRS。總體而言兩個模式的分析結果大致相同，皆能有效解釋崩塌地分布，而具有模擬及預測之功能。本研究考慮了土壤深度、土壤凝聚力等參數的空間變異性，確能更有效地解釋山崩分布而可望獲得更佳的預測效果。 In this study, we measured soil depth in the Piya creek watershed of the Tahan River, and established the relationship between soil depth and slope angle. The functional relationship was used to estimate soil depth in the entire watershed with employing slope angles which are derived from a Digital Terrain Model (DTM). We used a hydrology model and an infinite slope model to perform back analysis for soil strength and hydrologic conductivity for each stratigraphic zone in the watershed from an event-based landslide inventory and the event rainfall data of typhoon Aere. In the back analysis, the effective cohesion of soil is assumed to be affected by roots and can be modeled by a function using Normalized Difference Vegetation Index (NDVI) as an independent variable. The ultimate result of the back analysis was satisfied by a criterion of maximum overall accuracy of interpreting landslides. In the analysis, the SHALSTAB and the TRIGRS models were used for slope stability analysis, and the results were compared by carefully validations of predicting landslides. The overall accuracy of SHALSTAB and TRIGRS are 85.6% and 87.2%, respectively. The area under the curve of success rate curve of SHALSTAB and TRIGRS are 0.811 and 0.801, respectively. The performances of these two models are similar; both of them can predict the landslide distributions effectively. To consider the spatial variation of soil depth and effective cohesion in this study do help increasing the accuracy of predicting landslides.
