The performance of the maximum entropy algorithm and geographic information system in shallow landslide susceptibility assessment

Shallow landslide is one of the natural hazards that damage life and property of people in mountainous watershed. Due to the fact that a lot of landslides events have been occurred in this watershed, assessment the risk of shallow landslides by using appropriate methods and determine of effective factors in reduce the hazards of it is so effective. The potential of using maximum entropy modeling for landslide susceptibility mapping is investigated. In the case study of west of Ardabil province, 74 landslide occurrences were identified, 52 landslides (70%) used for training and the 22 landslides (30%) applied for validation purpose. environmental factors including continuous (altitude, slope, aspect, plan curvature, drainage density, and rainfall) and categorical (lithology and landuse) data were used as inputs for modeling. From the optimal setting test based on cross-validation, a continuous data and its combination with categorical data showed the best predictive performance. The results of validation showed that the ROC and AUC for success and prediction rate of model was 96.1 and 97.6%, respectively. Factor contribution analysis indicated that altitude and rainfall layers were the most influential factors. From interpretations on a response curve, steeply sloping areas that consisted of excessively covered with old alluvial terrace soils were very susceptible to landslides. Predictive performance of maximum entropy modeling was slightly better than that other models like of a logistic regression which has been used widely to assess landslide susceptibility. Therefore, Maximum entropy modeling is shown to be an effective prediction model for landslide susceptibility mapping.