| Summary: | 碩士 === 國立中興大學 === 水土保持學系所 === 107 === The sediment from landslides directly fell into the river was considered as near-bank landslides in the study. And the near-bank landslides on the upper catchment area of Qishan River was selected as a study case, using Logistic regression method with factors, including topography, geological and hydraulic, to establish the landslide susceptibility models of different combinations. The combinations were whether hydraulic factors included and adding different hydraulic factors, and compare their differences in near-bank landslides prediction. In addition, the near-bank landslides in the study were classified into four types according possible causes and different appearances. TypeⅠ caused by the rainfall erosion presented wide-upper appearance. TypeⅡ caused by the runoff collecting in the trench presented slender appearance. TypeⅢ caused by bank scour presented wide-lower appearance. TypeⅣ caused at the source of the tributary presented continuous appearance. The results showed that the AUC for the model established with the topography factors (slope, aspect, terrain roughness and NDVI) and geological factors (lithology and dis slope) was 0.754. And in the combinations with different hydraulic factors, the AUC for the model added hydraulic factors of riverbank type and riverbed slope is increased to 0.774. Especially for the wide-lower appearance landslides caused by the river scouring, the prediction is significant improvement in the model. Finally, based on the landslide susceptibility map obtained by this study, combined with the protected targets data for risk assessment. Identified the distribution of high-risk areas to provide government reference for riverbank maintenance, or regulations order.
|
|---|
