Research Information System
Environment, Development and Sustainability, 2026 (SCI-Expanded, Scopus)
This study assessed the landslide risk in the Bolaman micro-basin in Türkiye using a combination of expert-based and supervised machine learning methods. A landslide susceptibility map was produced using a random forest model. Validation of the results was conducted using a receiver operating characteristic curve (AUC), with an AUC value of 97%. A travel distance map was created using run-out values via inverse distance weighting interpolation. Vulnerability was defined in terms of the resistance exhibited by the element at risk (EaR), comprising buildings, roads, agricultural, and forest areas identified using satellite imagery. Information about these features was gathered through fieldwork. Lastly, a risk map was created by integrating susceptibility, run-out, and vulnerability maps through the application of a Mamdani Fuzzy Inference System. The risk map indicates that 34% of the area is at risk of landslides, with the highest risk areas situated along the Bolaman River. Additionally, 54.7% of the buildings, 62.1% of the provincial road, 62% of the local road, 36.5% of the agricultural lands, and 35.7% of the forest areas in the basin, were determined prone to high-very high landslide risk. The risk map was tested with twenty-three landslides that were not part of the training and testing datasets, employing general accuracy criteria. The accuracy of the risk model was found to be 95%. The results of this study provide that the proposed risk assessment methodology can be employed to identify high-risk areas and determine pre-countermeasures for disaster management.