AI helps identify mobile connectivity ‘cold spots’ for inclusive MHEWS

While mobile networks are increasingly recognised as a critical channel for disseminating early warnings, a significant proportion of the world's most vulnerable populations still live in areas with limited or no mobile connectivity. The problem worsens in the wake of disasters when infrastructure is damaged or destroyed. Without accurate, up-to-date information on where these connectivity "cold spots" are located, governments are unable to design equitable and effective alerting strategies. 

The International Telecommunications Union (ITU), as part of the Early Warnings for All (EW4All) initiative, developed the Early Warning Connectivity Map (EWCM) together with Microsoft AI for Good Lab, GSMA, Planet Labs and the Institute of Health Metrics and Evaluation at the University of Washington. The core challenge to solving the problem is threefold: determining which populations can be reached through available messaging channels; quantifying how many people are offline and therefore unable to receive alerts; and understanding how connectivity levels - and the capability to deliver warnings - change during and after disaster events. To address this challenge, ITU and partners developed the EWCM which uses an AI-powered methodology that integrates satellite imagery and advanced analytics to generate high-resolution, time-enabled population density maps. Overlaying this information with mobile network coverage data could then further identify areas in need of connectivity and inform decisions around rural network expansion. This AI-powered approach allows EWCM to provide a detailed and continuously updated view of connectivity gaps - crucial for planning EWS that truly leave no one behind. 

The EWCM was first piloted in Fiji, Tonga and Vanuatu. Following the success of these initial deployments, the approach has since been scaled to 30 more countries, the majority being supported as part of the EW4All initiative, where it has helped identify major coverage gaps where mobile early warning messages would fail to reach communities. This has strengthened national strategies for last-mile coverage and inclusion. 

Data variability and patchy network coverage from mobile operators posed challenges in ensuring consistency. To mitigate this, EWCM aggregates nine distinct connectivity datasets and supplements them with real-time, crowd-sourced data. Ongoing validation with ITU Member States and continuous improvement of AI tools helps address data quality concerns. Supporting technical innovation with capacity-building efforts is important, ensuring that national authorities not only have access to high-quality data, but also possess the skills and tools necessary to interpret and apply it effectively in decision-making.