Landslide Susceptibility Modeling and Mapping: A Comparison of Frequency Ratio and Decision Tree Models

Abstract

Landslides constitute one of the most damaging natural hazards in mountainous regions worldwide, and the Himalayas are particularly vulnerable due to their steep topography, complex geology, and intense monsoonal rainfall. This study develops a landslide susceptibility map for a part of Uttarakhand, India, using two comparatively simple yet robust modelling approaches—the bivariate Frequency Ratio (FR) method and the Decision Tree (DT) machine learning algorithm. A total of ten conditioning factors representing the key topographical and geo-environmental characteristics of the region were used, and a landslide inventory of 104 events was compiled for model training and validation. Model performance was evaluated using multiple statistical measures, including PPV, NPV, Sensitivity, Specificity, MAE, RMSE, and AUC. The results indicate that the DT model significantly outperforms the FR method, achieving an AUC of 0.848 compared to 0.578 for FR. The novelty of the work lies in the systematic comparison of these two widely used yet contrasting approaches in a Himalayan setting, demonstrating that the DT model provides a more reliable and transferable framework for susceptibility assessment in regions dominated by shallow landslides. These findings highlight the potential of simple machine-learning models to offer accurate and practical tools for landslide management in data-scarce mountainous terrains.