Bayesian network based probabilistic approach for uncertainty analysis of earthquake induced landslides

In this study, a unique approach is used to calculate the likelihood of the safety factor and permanent displacement of natural slopes under earthquake shaking. The proposed approach was constructed using probabilistic modeling of landslide instability based on the Bayesian Network technique. First, the pseudo-static factor of safety was computed, considering it an uncertain parameter. Then, the permanent displacement of failure mass was estimated through probabilistic analysis considering the effect of critical and peak horizontal acceleration. In the process of probabilistic analysis, soil and slope properties (cohesion, friction angle, unit weight, slope angle, and failure depth) and peak horizontal acceleration were considered as random variables distributed as normal and exponential functions, respectively. To illustrate the applicability of the proposed approach, a hypothetical infinite slope was adopted from past literature. The results showed that due to the event of an earthquake, the slope might experience permanent displacement. Finally, based on the variation of permanent displacement, the likelihood of landslide occurrences was estimated. Validation of the study was established by comparing the outcomes with the results obtained from the multivariate probabilistic approach, first-order reliability method, and Monte Carlo simulation. To demonstrate the practical applicability of the proposed framework, a case study of earthquake-induced landslides was taken to estimate the factors of safety and permanent displacement probabilistically. The methodology presented in this study would lead to an estimation of landslide failures by taking uncertainties into account, which would increase the safety of city dwellers.