Machine Learning and Sensitivity Analysis Approach to Quantify Uncertainty in LandslideSusceptibility Mapping

Mitigating the impacts of landslides requires quantifying the susceptibility of differentinfrastructures to this hazard through landslide susceptibility mapping. The mapping requires overlaying thespatial effects of multiple factors that contribute to the occurrence of landslide events (rainfall, land cover,distance to roads, lithology, and slope) and this process requires assigning weights to the different factorscontributing to landslides. This study introduces a new statistical approach for quantifying the weights used inlandslide susceptibility mapping and their associated uncertainty. The proposed approach combines machine learning(random forest classification) with large-scale sensitivity analysis to derive the uncertainty ranges of weights used inlandslide susceptibility mapping. The study demonstrates the approach for a case study of the Chittagong Hill Tracts andSylhet divisions of Bangladesh to understand the implications of weight uncertainty for road susceptibilityto landslides. The case study results show that distance to roads is the most influential factor to determine thelikelihood of the occurrence of landslide events, followed by the land cover type. Given weight uncertainty, thepercentage of road lengths in the study area under extremely high susceptibility to landslides ranges from around 20 to38 percent. The tolerance level to weight uncertainty is a crucial determinant of investment costs and is ultimately acritical element for decision making to relevant institutions and affected stakeholders. A conservativeselection of weights from within the uncertainty range (a weight combination that results in the highestsusceptibility) means that the risk is minimized but with a high investment cost.