碩士 / 逢甲大學 / 水利工程與資源保育研究所 / 96 / The purpose of this study is to integrate Geographical Information System and Logistic Regression for the potential assessment of landslide. As the case study, Shihmen Reservoir watershed was divided into 100 sub-watersheds for further analysis. With the aero photos, river system map, geological map, and Digital Elevation Model from the Soil and Water Conservation Bureau (SWCB), several factors were extracted with help of Geographic Information System, including watershed area, length of stream, mean watershed width, average slope, geological parameter, and road length. All values of these terrain factors for the 100 sub-watersheds were stored in a database for analysis of the logistic regression model. A logistic regression model for landslide potential was calibrated by 60 sub-watersheds and verified by the other 40 sub-watersheds. The model performed the landslide prediction with accuracy rate of 70% from the calibration sub-watersheds. This paper also defined the landslide potential as five-level scale, i.e., low potential, medium low potential, medium potential, medium high potential, and high potential. With landslide location map conducted by SWCB for Shihmen Reservoir, the verification results indicated that the higher potential this model predicts, the higher percentage of sub-watersheds did have landslide within them. In this study, the ratio of calibration to verification sub-watersheds, i.e., 60 to 40, 70 to 30, and 80 to 20, and the sampling methods, random sampling and stratified sampling, were discussed. However, the results showed that two sampling methods has no significant difference.
| Identifer | oai:union.ndltd.org:TW/096FCU05398018 |
| Date | January 2008 |
| Creators | Jing-bo Huang, 黃靖柏 |
| Contributors | none, 葉昭憲 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 64 |
Page generated in 0.1612 seconds