Riverbank landslides are a major source of sediment in any given watershed. They occur because the flow of the river erodes the foot of the riverbank, destabilizing it. This study analyzed the landslide susceptibility of the
riverbank in the Chenyulan watershed. Logistic regression was used to construct a landslide susceptibility model based
on not only topographical, geological, and anthropic factors but also hydraulic factors, including the hydraulic sinuosity index, channel gradient, and bank curvature (concave or convex). The study area was divided into four regions according to the riverbed slope and the confluence of rivers. The results indicated that landslides were predominantly influenced by topographical and geological factors. Notably, hydraulic factors, which represent external forces, enhanced the predictive ability of the model. The most significant enhancement was attributed to the hydraulic sinuosity index,followed by bank curvature and riverbed slope. Consequently, lateral erosion of the river was considered to contribute to riverbank landslides in the study area, and control engineering of lateral erosion was suggested as a mitigation measure.Hydraulic factors improved the predictive performance of the model by a considerably larger extent in the downstream area compared with the upstream area, with the most substantial improvement being approximately 10% of the area under the curve. Using a hydrodynamic model for accurately estimating hydraulic factors may further enhance the predictive performance of the proposed landside susceptibility model.
Key Words: Riverbank, Landslide, Hydraulic factor, Logistic regression.