Logistic Regression with Hydraulic Factors for Landslide Susceptibility of Riverbank

• Main Authors: Chun-Tzu Chang, 張純慈
• Other Authors: Hsun-Chuan Chan
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/24e5f3

碩士 === 國立中興大學 === 水土保持學系所 === 106 === The sources of sediment in mountain watershed are soil erosion, bank scour and landslide. The exceed sediment deposited in the river and blocked the flow caused the variations of river bed and influences of riverbank instability. This study focused on the landslide susceptibility analysis of the riverbank which is considered as the main source of sediment in a river. Tai-An river passes through midst of Taiwan with high population density. Considering the deposition of sediment endangers lives and property, the upstream areas of Tai-An river was selected as the study area. The factors in the analysis included not only the traditional ones (topography and geological factors), but also the triggered factor of flow (called the hydraulic factors). In this study, the geological factors included lithology and dip slope index and topography factors included slope, slope high, aspect, and greenness index. The hydraulic factors included the channel gradient, sinuosity, radius of curvature, and slip-off or undercut slope. The Logistic regression method was used to establish the landslide susceptibility model and different combinations of factors were used in the model. These combinations included models with 1. hydraulic factors; 2. hydraulic & geological factors; 3. hydraulic & topography factors; and 4. hydraulic, geological & topography factors. The results showed that once the models with hydraulic factors and the overall accuracy were higher than 60%. Moreover, the overall accuracy of the model with hydraulic & topography factors and hydraulic, geological & topography factors were more than 70% and AUC were about 0.80. The results showed hydraulic factors were effective in predicting landslides of riverbank. The developed model is expected to provide as references of river regulations in the study area.