Performance Comparison of Landslide Susceptibility Maps Derived from Logistic Regression and Random Forest Models in the Bolaman Basin, Türkiye

Landslides often cause significant economic and human losses, and therefore landslide susceptibility mapping (LSM) has become increasingly important. Accurate assessment of LSM is important for appropriate land use management and risk assessment. The aim of this study is to define and compare the results of applying the random forest (RF) and logistic regression (LR) models for estimating landslide susceptibility, and also to confirm the accuracy of the resulting susceptibility maps in the Ordu-Bolaman River micro-basin. The study area was selected because it is one of the most landslide-prone areas in Türkiye. First, a total of 231 landslide locations were identified. Then 12 landslide-influencing factors were selected to generate landslide susceptibility maps. These maps were produced using the landslide influencing factors based on the RF and LR models in a geographical information system (GIS) environment. Finally, area under the curve (AUC) analysis, sensitivity, specificity, and accuracy were considered to assess and compare the performance of the two models. In addition, the maps were retested with large landslides not included in the training and test data sets, using general accuracy criteria. The results of the present study will be helpful for future landslide risk mitigation efforts in the research area.