| Summary: | 碩士 === 國立成功大學 === 資源工程學系 === 106 === Landsliding is one of the most common natural disasters in Taiwan. To predict mountain disasters, the original method is for the experts to be at the scene to determine the landslide factors based on their own experiences and to score according to the results. This kind of method requires large amount of manpower and material resources and it is also more subjective. Thus, this study uses the Geographic Information System (GIS) combined with the Instability Index Method to establish the prediction model of the potential landslide of Zengwen Watershed with classify of dip slope and undip slope.
In this study, ArcGIS creates and superimposes each layers for the final penitential landslide map with key factor of dip slope and undip slope. Eight key factors, such as slope, aspect, elevation, distance to the river, distance to structure, bedding, distance to the road and rainfall, are respectively used to examine the instability index as well as giving the score value and weight, and to superimpose the images of the sensitizing zones of Zengwen Watershed.
To increase the accuracy of the landslide susceptibility map, this study has filtered out the flat area, which includes the areas below 20 degree slope. Moreover, rivers are not affected in spite of landslide within the river areas when formulating the landslide catalog, therefore, those areas have been filtered out. This study has analyzed and discussed the susceptibility landslide of area with dip slope and undip slope.
In the study conducted by Li (2016) at the Zengwen area, when dip slope is not taken into account, and the average AUC is 0.755. This study has discussed the forecast accuracy of dip slope and undip slope respectively. According to the results, dip slope AUC average is 0.845, and undip slope AUC average is 0.774, which indicates that analyzing dip slope and undip slope individually has increased the preciseness of the model prediction.
Regardless of filtration, dip slope is more effective in the result than undip slope. The AUC value of dip slope without filtration is up to 84.8%, which verifies the 2016 landslide catalog AUC average, which is 80.7%. When both internal and external factors are consistent, dip slope has a very high weight value for collapse, which must be taken into account for disaster prevention.
Furthermore, among the bedding factor, safflower seed layer is most likely to collapse; Stumbling layer and alluvial layer are relatively less likely. Most landslides occur at 30 degrees or more in slope factors.
|
|---|
