Source: Phys.org

 
Phys.org reports on a groundbreaking AI tool developed by MIT researchers that generates realistic satellite images of potential flooding in regions susceptible to hurricanes. The tool combines a generative adversarial network (GAN) with a physics-based flood model, offering accuracy in visualizing flood impacts. Tested on Houston during scenarios mirroring Hurricane Harvey, the AI model produced satellite images that matched real flood patterns while avoiding errors common in AI-only models. This innovation aims to improve disaster preparedness by offering intuitive, location-specific visualizations that motivate public evacuation and readiness. The team emphasizes the importance of pairing AI with physics models to reduce "hallucinations" or inaccuracies in generative outputs. The project, called the "Earth Intelligence Engine," showcases how technology can aid climate risk management.