Assessment of rainfall-induced landslide susceptibility at the regional scale using a physically based model and fuzzy-based Monte Carlo simulation

The occurrence of landslides is controlled by various interrelated spatial and climatic factors, some of which cannot be determined accurately in detail and others that can be determined only with large degrees of uncertainty. Therefore, uncertainties pervade the field of landslide susceptibility analysis, and so the recognition and assessment of uncertainties is of paramount importance in this analysis. In particular, in a physically based model that has been widely used in regional landslide susceptibility analysis, uncertainties are inevitably involved since reliable information required to estimate input parameters in physically based models is frequently limited in extent and has an imperfect quality. In addition, some uncertainties related to measured geotechnical parameters of slope materials may be nonstochastic, but rather cognitive, arising from incomplete information. Under such conditions of limited information, it is more appropriate to adopt the fuzzy set theory. Therefore, in this study, the fuzzy set theory, coupled with interval MC simulation and the vertex method, was adopted for physically based landslide susceptibility analysis to properly handle uncertainty propagation through a physical slope model. The proposed fuzzy-based approach was applied to the study area to evaluate landslide susceptibility for the regional area, and subsequently, to evaluate the performance of the proposed approach, the analysis results were compared with landslide inventory. In addition, a probabilistic and a deterministic analysis were also carried out to compare with the fuzzy-based analysis results. The fuzzy approach showed better performance than the probabilistic and deterministic analyses and was more robust against variation in input parameters than other approaches. Therefore, in a regional-scaled landslide susceptibility analysis using a physically based model, the fuzzy approach can control the uncertainties appropriately and is particularly advantageous when the amount of reliable input data is very limited.