Landslide Susceptibility Modelling Using the Random Forest Machine Learning Algorithm in the Watershed of Rais-Ali Delvari Reservoir

The aim of this study was to model the landslide susceptibility using the Random Forest Machine learning technique and prioritization of effective factors on landslide occurrence in the watershed of the Rais-Ali Delvari Reservoir. The landslide inventory map was prepared using extensive field surveys and the Iranian Landslides Working Party Data Bank. Of the total of 279 identified landslide locations, 70% were used for the modelling processes and the remaining (30%) were applied for validation of the developed model. Different thematic layers including elevation, slope angle, plan curvature, profile curvature, topographic wetness index (TWI), distance from rivers, drainage density, distance from faults, distance from roads, lithological units, and the normalized difference vegetation index (NDVI) were selected. According to the relationship between the dependent (landslides) and the independent (effective factors) variables in the R statistical software, the random forest algorithm was run using the “Random Forest” package, and a landslide susceptibility map was prepared. Accuracy of the model was tested using the receiver operating characteristic (ROC) curve based on 30% of unused landslides in the modelling process. Accuracy results indicated that the Random Forest model with an AUC value of 0.983 had an excellent precision. Also, prioritization of the effective factors showed that the slope angle, elevation, plan curvature, distance from road, and lithological units had the highest effect on landslide occurrence. Therefore, it maybe suggested that the prepared landslide susceptibility map could be effective in decision making for land use planning, and in the managing of the Rais-Ali Delvari Reservoir Watershed.