Extreme weather, fires and critical infrastructure failures can cause crises that pose complex challenges for municipalities and public authorities. Relevant data is often scattered and diffi-cult to combine in a current and consistent situational awareness picture. This is where an AI-driven dashboard for regional civil protection can help. Researchers at the Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institut, EMI are jointly developing this dashboard as part of the RegioGuard-PRO project with partners from industry and research.

The project builds on the results of two earlier projects funded by the German Federal Ministry of Research, Technology and Space (BMFTR): AKRIMA (AI-powered crisis monitoring for logistics) and HERAKLION (a data space for municipal resilience analysis). These results are being used to develop a practical, digital and extensible platform for civil protection. The platform will enable municipalities and other public-sector entities to respond rapidly and appropriately to hazards during crises, while identifying potential cascading effects in critical infrastructure at an early stage.

Earlier detection of risks from combining AI and geodata

“Municipalities often have access to vast amounts of data during crises and disasters, it is scattered across different sources and is not interconnected,” says Fraunhofer EMI research scientist Kai Fischer. “Our digital platform consolidates and analyzes large sets of data from different sources. This enables early detection of potential hazards and faster and more effective initiation of appropriate action.” For example, flood hazard maps no longer have to be analyzed entirely manually. Instead, the system determines which buildings or critical infrastructures are in a flood zone, identifies critical infrastructure and shows which areas could be cut off from the outside world.

Emergency responders can also use a language-based input to retrieve relevant information through an integrated AI model. “The AI model works much like a chatbot, giving users access to our data analyses,” Fischer explains. The goal is to bring all data and analysis results together on a single platform to create a real-time warning system. A ready to use demonstrator is planned within two years.

Individual resilience phases are described in precise detail

For a comprehensive analysis, the researchers integrate a range of data sources, including social, infrastructure, transportation, population, environmental, building, asset and weather data. This yields an integrated situational awareness picture, showing the extent of flooding, which buildings are affected and whether fire stations can be reached.

Using statistical and stochastic methods, researchers analyze the cause and the effect of extreme weather events. Their findings can be used to establish municipal performance indicators for disaster risk management and to quantify resilience. The analysis accounts for factors such as the availability and usability of critical infrastructure and the potential scale of damage. By consolidating and analyzing the data, this approach gives decision support before, during and after a disruptive event. This enables precise characterization of each phase of resilience, including preparedness, prevention, protection, response and reconstruction. This methodology helps identify potential vulnerabilities and strengthen municipalities’ capacity to manage disasters.

What-if scenarios help prepare for emergencies

RegioGuard-PRO integrates resource and process planning with resilience analyses to develop and optimize alarm and response plans. AI-powered what-if scenarios help assess potential courses of action and incorporate them in situational assessments.

For one use case, the project partners are studying the impact of a 100-year flood event. They combine flood hazard maps from the Federal Institute of Hydrology with road network data from OpenRouteService and OpenStreetMap. They also draw on data from the German Federal Agency for Cartography and Geodesy on key facilities, fire department locations, digital building models and landscape and land-use models. “The what-if analysis reveals a municipality’s vulnerabilities and calculates expected water quantities and many other parameters. This allows officials to take targeted precautions in response to future severe weather warnings,” Fischer explains.

Testing at demonstration centers under real-world operating conditions

Demonstration centers for AI-assisted civil protection are being established in the cities of Trier, Mainz and Wildau to test the dashboard under realistic field conditions, with the aim of eventually deploying it in other regions. The sites will also serve as knowledge-sharing hubs, passing on findings and lessons learned to other municipalities and civil protection organizations. In partnership with the Lörrach district, the researchers are also exploring applications related to extreme weather, including where people can cool off or seek shelter during heat waves and how they can reach these locations.