本研究透過整合跨尺度的崩塌潛勢，社會脆弱性和危害暴露量分析，來建立南投縣崩塌災害風險地圖。首先，崩塌潛勢機率係採用統計方法為研究取徑，透過網格基礎的崩塌潛勢模型進行評估。第二，社會脆弱性使用包含防災訓練次數、消防人力面積比例、避難所的距離、撫養比、淨綜合收入和低收入家庭數量等六個不同尺度的主題圖計算。第三，危害暴露量則是考慮人口與土地利用狀況，採用多層多類別衡量模型，重新分配人口，提高暴露量估算的準確性。在網格層級進行風險指標計算後，整合至坡單元和最小統計區進行災害風險評估。研究結果顯示，透過網格進行整合，能突破現有單元限制並進行跨尺度的評估，但缺點為無法符合現有救災的體制。坡單元能符合崩塌的地形特性；最小統計區是最基礎的統計單元，其劃分亦參考高程與坡度等地形特性和人口分佈狀況及密度，可做為風險評估的基本單元，不僅能夠突顯各村里之間的風險差異，也呈現村里內部風險的空間變異，符合現行防災人員所需之更詳盡且實用的風險空間資料。

This work adopts a multi-scale approach to integrating landslide hazards, social vulnerability and hazard exposure to develop a landslide disasters risk map in Nantou. First, landslide hazards are assessed by using a statistical and grid-based landslide susceptibility model. Second, social vulnerability is calculated by using six multi-scale maps, including frequency of disaster prevention training, ratio of firepower, distance to shelter, dependency ratio, net consolidated income and number of low-income households. Third, hazard exposure is determined based on demographics data and land-use map for considering both non-spatial and spatial factors. A multi-layer, multi-class dasymetric model (MLMCD) is applied to redistribute the human population for improving the spatial accuracy of exposure assessment. The natural and human factors are evaluated based on the slope-unit and minims administrative units, respectively. Analytical results show that use of grid-based units can break through the limitation of administrative districts, but conflicts with the organization of current rescue management. Conversely, the slope unit is relevant to the landslide distribution, and the minims administrative unit was originally developed to meet the distributions of topography and population. This approach presents not only the landslide risk in each village, but also the spatial pattern within a village. Additionally, the landslide risk map provides more precise and detailed information, making it useful for disaster workers.