森林集水區邊坡穩定性之評估：以蒙地卡羅模擬無限邊坡模式之參數

Landslides have been the major concern of forest management in Taiwan since most part of the island consists of precipitous slopelands, which are composed of highly fractured geological formations. Mapping or delineating areas susceptible to landslides is essential for effective watershed management. Thus, this study aims to develop an effective method for allocating unstable slopelands. The process-based model uses GIS as an integration platform, and uses DEM to calculate slope gradient and specific catchment area for each grid, deriving NDVJ value from SPOT imagery. The model then uses Monte Carlo simulation to derive parameter values for cohesion and friction angle needed in the infinite slope model. Two simulation strategies are adopted. One treats the entire watershed as uniformity. The other uses NDVI to
differentiate spatial variation of cohesion, and the friction angles of individual geological formation are simulated separately.
The result shows that the accuracy rates of the two strategies in differentiating instable slopes are 64.0% and 76.4%, respectively. And the average rates for individual landslides being correctiy evaluated are 67.0% and 76.8% for the two respective strategies. In view of spatial distribution, most of the landslides in the Chi-Jia-Wan watershed are located in the EO formation (composed of quartzite, sandstone, slate, graphitic shale), while topographically, they are located in converging steep slopes. Landslides in the Yo-Sheng-Chi watershed mostly occurred along the river valley, and they were very likely to be triggered by human activities. Thus, the model performance was not as successful as in Chi-Jia-Wan watei shed.