Comparison of Landslide Susceptibility Maps using Logistic Regression (LR) and Generalized Additive Model (GAM)

Landslide is one of the most common natural disasters that endanger the lives and properties of people in mountainous areas. Therefore, identification of risk exposure areas of landslide is essential to prevent and reduce damages by landslides. The purpose of this study is compared to logistic regression (LR) and generalized additive models (GAM) and the evaluation of their performance for landslide susceptibility mapping in the Chahardangeh Watershed, Mazandaran Province. At the first, landslide locations were identified by Google Earth images and extensive field survey. Then, the landslide inventory map was randomly divided as training data 70% for modeling and the remaining 30% was applied for the model validation. The landslide conditioning factors including topographic, hydrologic, geology and human factors were constructed in GIS. Finally, the receiver operating characteristic (ROC) Curve was used for the model validation. The validation of results showed that the area under the ROC curve for LR and GAM models were 81.2% and 82.4%, respectively. So, both of the models are suitable and efficient methods for landslide susceptibility mapping in the study area. Although, the obtained results showed that the GAM model performed is slightly better than the LR model for determining regions of susceptible to occurrence of landslide in the study area.