A Network-Enabled Intelligent Decision Support System for Disaster Response Using Spatiotemporal Data Fusion and Predictive Health Analytics

Abstract

Natural and man-made disasters kill hundreds of thousands of people every year, creating cascading health crises that overwhelm systems designed to respond in an emergency. Current DSS does not provide real-time spatiotemporal fusion or consider the key interactions between disaster dynamics and population health outcomes.We propose the Intelligent Disaster-Response Decision Support System (IDSDR), a novel end-to-end architecture integrating: (i) a Multi-Source Spatiotemporal Data Fusion module (MSTDF) combining satellite imagery, IoT sensors, social media, and meteorological feeds through a Cross-Modal Attention Fusion Network (CAFN); (ii) a Population Health Integration Module (PHIM) that co-models disaster severity with epidemiological risk and healthcare resource demand; (iii) a Resource Allocation Graph Neural Network (RA GNN) for dynamic, context-aware relief routing; and (iv) an Explainable AI (XAI-SHAP) interface enabling transparent decision support for incident commanders.Tested on six real-world disaster datasets (floods, earthquakes and cyclones) collected between2019–2024, IDSDR obtains an overall accuracy of 95.1%, a F1 score of 95.0% for disaster detection, 93.2% F1 score for the prediction of health risks and 94.0% F1 in resource allocation — outperforming state-of-the-art rule-based baselines by +23.3 percentage points. Average time in advance of alert was an improvement by 47 minutes over conventional systems.IDSDR is a key milestone towards integrated health competent disaster intelligence. The approach of jointly modeling spatiotemporal disaster dynamics and population health vulnerability addresses a critical gap in the literature while providing a scalable, explainable framework for deployment across both high- and low-resource emergency management settings.