Probabilistic back analysis for landslide susceptibility assessment in data-scarce regions using a Bayesian approach

Incomplete knowledge about geotechnical conditions, owing to limited data availability, leads to uncertainties in slope stability analysis. Therefore, probabilistic back analysis has been adopted to obtain additional information and control the uncertainty, but the back analysis has previously been adopted only for individual slope problems, not for regional landslide analysis, in which the limited data causes serious inaccurate analysis results. In this study, the probabilistic back analysis was applied to a broad area where limited data is available for input parameters but numerous historical occurrence records are available. To address this issue, Bayesian back analysis was used to obtain supplementary data on input parameters based on observations of multiple landslide occurrence. These supplementary data were used to update the shear strength parameters, and landslide susceptibility was evaluated using the updated parameters. The results of the analysis conducted with the updated input parameters exhibited superior performance in comparison to those obtained with limited data. Moreover, the proposed method demonstrated robust performance regardless of the random selection and proportion of back analysis landslide locations. Consequently, the acquisition of supplementary data by applying probabilistic back analysis led to a reduction in uncertainty, thereby enhancing the accuracy of the landslide susceptibility analysis.