XAI-Based Disaster Management System for Satellite Imagery

Natural disasters such as floods, wildfires, and earthquakes pose significant challenges to urban resilience, requiring real-time monitoring, damage assessment, and rapid response mechanisms. Traditional AI-driven disaster management systems often function as black-box models, limiting trust and interpretability. This research proposes an Explainable AI (XAI)-based disaster management system leveraging satellite imagery and deep learning to enhance decision-making. The system integrates CNNs for disaster detection, U-Net for damage segmentation, LSTMs for disaster spread prediction, and GIS-based evacuation route planning. XAI techniques such as SHAP, LIME, and Grad-CAM provide interpretability to AI predictions, ensuring reliability in decision-making. The proposed approach improves real-time disaster response, misinformation filtering, and resource optimization, contributing to Sustainable Development Goal (SDG) 11 for sustainable cities and communities.