Leveraging Internet Radio for Sustainable Disaster Management: An Integrated IoT and Machine Learning Framework

Konstantinos Papatheodosiou, Ioannis Georgakopoulos, Stamatios Ntanos, Vasileios P. Rekkas, Panagiotis Sarigiannidis, Sotirios K. Goudos

Natural disasters represent a critical intersection of environmental degradation, climate change, and societal vulnerability, posing a severe threat to sustainable development. Building a resilient communication infrastructure is therefore paramount for environmental sustainability and community survival. This paper addresses the shortcomings of traditional systems—such as high latency, limited coverage, and unreliable infrastructure—by proposing a novel integrated disaster management system built on Internet Radio technology. The framework combines a robust early warning system with an efficient emergency information broadcaster, offering global reach, real-time capabilities, and significantly reduced resource requirements. Its low-power consumption and minimal physical infrastructure make it an environmentally sustainable and cost-effective solution, aligning with goals for reducing the ecological footprint of critical services. A comprehensive 6-month case study for the Dodecanese Islands, Greece—with focused implementation on Symi Island—was conducted to validate the system. IoT-based meteorological stations and machine learning models (Random Forest) achieved a temperature prediction RMSE of 1.5 °C (a 35% improvement over traditional models), a wind velocity RMSE of 3.1 km/h, and an F1-Score of 0.80 for rainfall prediction. The integrated system demonstrated end-to-end latency of 10–25 s (210× faster than traditional systems), 98% coverage, 94% user comprehension, and a 70% reduction in operational costs. System-wide testing confirmed an alert accuracy of 92%, a false alarm rate of 12%, and a missed event rate of 10%, all within acceptable thresholds. The system achieved 99.2% overall uptime with redundant components ensuring continuous operation. Comparative analysis shows the proposed system outperforms traditional Greek EWS by 210× in latency, improves coverage by 327%, and reduces costs by 70% while maintaining three UN SDG alignments. The research fills a critical gap by integrating sustainable communication technology with modern predictive analytics, offering a replicable model for island communities worldwide.