GIS-based landslide susceptibility mapping of the Meghalaya-Shillong Plateau region using machine learning algorithms

Landslides are a common geological hazard causing impairment of public works and loss of lives worldwide and in India, especially in the Himalayan region. The present study aims to map the landslide susceptibility for the Shillong Plateau region of India using different machine learning algorithms, namely artificial neural network (ANN), extreme gradient boosting (XGBoost), K-nearest neighbors (KNN), random forest (RF), and support vector machine (SVM) and provides insights into influential factors, with a focus on disaster risk reduction. For this purpose, the geospatial database containing 15 landslide conditioning factors related to regional geo-environmental settings and a landslide inventory with 1330 locations are prepared. The landslide susceptibility maps (LSM) reveal that the south-southeastern portion of Meghalaya, mainly slopes along the southern escarpment, are more susceptible to landslides. The model robustness is demonstrated using the area under the receiver operating characteristic curve (AUC), F1-score, kappa, and other statistical metrics. The XGBoost and RF machine learning models with AUC = 0.971 have shown the best performance, followed by SVM (0.958), KNN (0.951), and ANN (0.945), which is consistent with other applied statistical parameters and higher than the traditional MCDA methods. However, the problem of overestimation is observed in the case of ANN and XGBoost models. The generated LSMs will assist decision-makers and planners in identifying high-risk areas, prioritizing mitigation measures, and guiding regional development.