Optimization study on the selection method for non-landslide points based on landslide susceptibility zoning

Landslide susceptibility assessment is crucial for geological disaster prevention, and the selection of non-landslide points significantly affects evaluation results. Taking Nyingchi City, Xizang, as an example, this study explores how selection of non-landslide points at various ratios from high- and low-susceptibility zones affects landslide susceptibility assessments. Based on factors such as terrain, geology, hydrology, and human activity, the study area is divided into high, medium, and low susceptibility zones. Different selection schemes for non-landslide points are designed, and ROC curves are used for evaluating the accuracy and stability of the assessment results. As indicated by the findings, the selection ratio of non-landslide points notably affects the results: a higher proportion from low-susceptibility areas better represents the spatial distribution of landslides, whereas a small number from high-susceptibility areas enhances model accuracy. At the same time, this research highlights the influence of non-landslide point selection on landslide susceptibility evaluations, suggests an optimized strategy based on zoning differences, and provides valuable insights for future assessments.