GIS-Based Landslides Risk Assessment Applying Certainty Factor (CF) and Ensemble with Deep Learning Neural Network (DLNN): a Study of Cachar District of Assam, India

Due to its extensive network of rivers, natural hazards like landslides, floods and other disastrous events are commonly occurred in Assam to which have a detrimental consequence on the growth of this state. Each year during the monsoon season, over 50 tributes flow into the Brahmaputra and Barak Rivers and causing distressing landslide. It was a heavily affected district in the first wave of flooding in June, 2022, and several people died in flood induced landslides. Accordingly, the present study aimed to evaluate landslide risk in Cachar district, Assam, North-East India. So, different physico-geographical parameters (landslide causative factors) were considered and mapped, and landslide risk was examined using the certainty factor (CF) and its ensemble with deep learning model. Primarily, 16 landslide causative factors and inventory map were prepared, and then the dataset were split into 70% for training and 30% for validating. The finding identified that the northern and south eastern parts of Cachar district is highly prone to landslides, whereas the western and central part is least prone to landslides. Based on ROC curve validation results, DLNN-CF was identified as the most efficient model, with AUC = 0.98, compared to standalone CF model, with AUC = 0.958. The present study would be a source to the local planners for managing future landslides in the Cachar district.