Landslide Susceptibility Analysis of Riverbank by Using Logistic Regression method

Taiwan, a island located in Eurasian plate and Philippine Sea Plate intersection, due to the plates squeeze each other and orogeny developed, resulting in geographical characteristics of mountain and hills. By the curst uplifting rapidly in Taiwan, the geology is weak and the flurosion is strong. The flow scours the river bank lead to loss of the hillslope and the slope instability, causing the landslide close to streamway and increase river sediment concentration. It not only make river turbid, affecting river ecosystems, but cause channel deposit . In the case of channel-fill deposit, it is easily to appear the phenomenon of blocking water flow, and increase risk of flood and debris hazard happened during a typhoon. And it may damage lives and properties of citizen around the downstream.

Lots of specialists and scholars have been researching the landslide susceptibility, but landslide causative factors about river characteristics are not taken seriously relatively. To realize the coupling between landslide and river characteristics factors, such as channel gradient, sinuosity , concave bank , convex bank, and path of river, the study area focuses on riverbank slope. The topography, geological and river characteristics factors were chosen as the landslide causative factors. To select the factors efficiently, the relationship between the landslide causative and the landslide data were assayed. Logistic regression method was used for establishing the landslide susceptibility model, and receiver operating characteristic was used to evaluate the accuracy of the landslide predicted model. The result representing good accuracy of the model can be referenced for relative improvement projects in the research area.