Rainfall-Induced Landslide Susceptibility Using a Rainfall–Runoff Model and Logistic Regression

Abstract

Conventional landslide susceptibility analysis adopted rainfall depth or maximum rainfall intensity as the hydrological factor. However, using these factors cannot delineate temporal variations of landslide in a rainfall event. In the hydrological cycle, runoff quantity reflects rainfall characteristics and surface feature variations. In this study, a rainfall–runoff model was adopted to simulate the runoff produced by rainfall in various periods of a typhoon event. To simplify the number of factors in landslide susceptibility analysis, the runoff depth was used to replace rainfall factors and some topographical factors. The proposed model adopted the upstream area of the Alishan River in southern Taiwan as the study area. The landslide susceptibility analysis of the study area was conducted by using a logistic regression model. The results indicated that the overall accuracy of predicted events exceeded 80%, and the area under the receiver operating characteristic curve (AUC) closed to 0.8. The results revealed that the proposed landslide susceptibility simulation performed favorably in the study area. The proposed model could predict the evolution of landslide susceptibility in various periods of a typhoon and serve as a new reference for landslide hazard prevention.