Using AI to map disaster damage for faster rescue and recovery

After a hurricane, drones can capture detailed images of damaged homes and roads. However, the challenge lies in getting those images to emergency responders, as the files are large and connectivity is often unreliable in the aftermath of a storm. To address this issue, CLARKE, an AI system developed at Texas A&M University with support from the National Science Foundation’s AI Institute for Societal Decision Making, has been created to speed up the process.

After a drone lands, the CLARKE system can analyze the images without relying on an internet connection, allowing teams to subsequently send compact reports highlighting building damage and road conditions instead of hefty image files. This innovation is part of a larger initiative aimed at leveraging AI in emergency response efforts.

Following a significant earthquake in Myanmar in March 2025, Microsoft's AI for Good Lab utilized satellite imagery analysis to aid organizations in identifying damaged buildings. By minimizing delays in transferring and interpreting images, researchers predict that CLARKE could potentially enable emergency managers to make decisions approximately a day sooner, thus allowing them to prioritize the areas most severely affected and potentially save lives.

According to Tom Manzini, a researcher at Texas A&M, CLARKE can evaluate 2,000 homes in just seven minutes. Emergency responders have already utilized the system following hurricanes Debby and Helene in 2024, and it has since become a standard procedure for Florida’s disaster drone team, FL-UAS1. In summary, Manzini and his team have trained 125 emergency responders on the proper use of the CLARKE system.