Real-time coastal flood risk assessment using IoT-integrated satellite data and machine learning models for predicting flooding events and informing resilient coastal planning for Durban coastal region, South Africa

Coastal regions around the world are facing growing threats from frequent and intense flooding events driven by climate change, rising sea levels, and unplanned coastal development. To address these challenges, this study aims to develop a real-time coastal flood risk assessment system in Durban, metropolitan area that integrates satellite data, Internet of Things (IoT) sensors, and machine learning techniques. The primary goal is to enhance flood prediction accuracy and support informed, resilient coastal planning and disaster preparedness. The proposed system combines high-resolution satellite imagery with real-time environmental data collected from strategically placed IoT sensors. Key variables include elevation, land use/land cover, slope, rainfall, and tidal fluctuations. A Random Forest machine learning model was used to process and classify these datasets into distinct flood risk categories. The model was trained using historical flood incidents and validated with ground-truth observations, ensuring strong predictive performance and reliability. The flood risk analysis revealed significant spatial variation across the coastal study area. Specifically, 33.15 % (847.35 sq.km) of the region was classified as “Very High” risk, followed by 28.26 % (722.63 sq.km) as “High,” 15.69 % (401.09 sq.km) as “Moderate,” 11.73 % (299.96 sq.km) as “Low,” and 11.14 % (284.97 sq.km) as “Very Low.” These findings emphasize the urgency of focusing risk mitigation strategies in the most vulnerable zones. By providing real-time, data-driven insights, this integrated framework offers a practical and scalable solution for coastal flood risk management. It empowers local authorities, planners, and communities to make proactive decisions that reduce risk and build long-term resilience.