Imagine being in Venice, not on just any day, but during one of the increasingly frequent exceptional high waters. The sirens sound, the wind howls, the incessant rain hits the city. You know that the tide is rising, that Piazza San Marco will soon be submerged, but no one can predict exactly the extent of the phenomenon. What if we could see the future instead? Observe from above, with precision, the water that invades the streets and squares, the historic buildings surrounded by the waves, the islands of the lagoon that disappear under the fury of the sea? It’s not science fiction, but the promise of the “Earth Intelligence Engine”, an innovative artificial intelligence tool developed by MIT scientists.
This innovative technology promises to significantly improve the ability to predict and manage natural disastershelping communities prepare for and protect themselves from extreme weather events.
The “Earth Intelligence Engine” is based on a generative AI model trained on a large dataset of satellite images, topographic maps and meteorological data. Unlike traditional forecasting models, which simply provide color-coded maps, this tool generates real photographs of the futureshowing with millimeter precision the extent of flooding during extreme weather events.
“Our AI learns to recognize patterns and correlations between the different factors that influence floods, such as rainfall intensity, wind strength, terrain conformation and the presence of infrastructure,” he explains Björn Lütjensthe MIT researcher who led the project. “This way, we can more accurately predict where water will accumulate and which areas will be most affected.”
To validate the effectiveness of their model, the researchers compared the AI-generated images with real satellite images taken after extreme weather events, like Hurricane Harvey that hit Houston in 2017 and the recent floods that hit Spain, and in particular the province of Valencia. The results show that the “Earth Intelligence Engine” is able to accurately predict the extent of floods, exceeding traditional models in accuracy.
The possible applications of this technology are many. Competent authorities will be able to use AI-generated images to:
“Our goal is to create a tool accessible to all, which can help communities around the world prepare for and protect themselves from the effects of climate change,” he says Dava Newmanprofessor of aeronautics and astronautics at MIT and co-author of the study.
The “Earth Intelligence Engine” is still a prototype, but MIT researchers have already made a version of their tool available online, so that the scientific community can contribute to its development. This technology represents an important step in the application of AI to manage the effects of climate change.
