ABSTRACT
Flooding remains a significant environmental hazard in northeastern India, with South Tripura severely impacted during the August 2024 event. This study presents a comparative flood susceptibility mapping using GIS-based Analytical Hierarchy Process (AHP) and Deep Neural Network (DNN) models, incorporating eight flood-conditioning parameters. The AHP model, a multi-criteria decision-making technique, was applied alongside a data-driven DNN model to evaluate flood-prone areas. Both models were validated using satellite-derived flood inundation data. Results indicate that while AHP offers rapid assessment, it exhibited lower predictive accuracy than the DNN model, which provided improved spatial representation of flood risk zones. The comparative findings highlight the enhanced potential of deep learning approaches in flood risk mapping and their scalability for regional planning. In hilly terrains like Tripura, where floods often co-occur with landslides, such integrated models can significantly contribute to geotechnical hazard zoning and infrastructure resilience.
