Evaluation of a combined spatial multi-criteria evaluation model and deterministic model for landslide susceptibility mapping

• We compared predictive capability of two different models (SMCE, SHALSTAB).
• A total mismatch of 53.27% resulted for three susceptibility classes.
• Two results were combined to improve reliability of susceptibility map using NFR.
• Combination of SMCE with SHALSTAB can add rainfall triggering factor in semi-quantitative model.
• Combined model was more accurate than either individual model at delineating landslide-prone areas.

This study evaluated the application of a combined spatial multi-criteria evaluation model and deterministic model for landslide susceptibility mapping in Deokjeok-ri Creek, located in the northeastern part of Korea. This region has frequent shallow landslides often caused by intense rainfall on weathered granite soil slopes. This study compared the predictive capability of two different models: a spatial multi-criteria evaluation (SMCE) model, which is a semi-quantitative model, and a shallow landslide stability (SHALSTAB) model, which is a deterministic model used to produce shallow landslide susceptibility maps. For the SMCE model, input layers of landslide causative factors (i.e., topographic, hydrological, soil, forest, and geological factors) were prepared for pairwise comparison to obtain susceptibility weightage. For SHALSTAB, a digital elevation model was used to calculate slope and wetness indices. Field inventories were used to validate and combine the two models. A comparison of the susceptibility map obtained from the SMCE method with that obtained with the SHALSTAB method revealed that the total mismatch area between the two maps for all three susceptibility classes was about 53%. Therefore, the two results were combined to improve the reliability of the susceptibility map. The performance of the combined map was determined using the receiver operator curve (ROC). The area under curve (AUC) revealed a success accuracy of 79.56%, and the predictive accuracy was 83.6%. These results demonstrate that the combined model was more accurate than either individual model at delineating landslide-prone areas of weathered granite soil slopes.