MIT scientists have developed an AI flooding tool that generates future satellite images to illustrate an area’s appearance after potential flooding.
They have tested the method on Houston, Texas, and generated satellite imagery depicting its appearance after a hurricane.
READ: Google AI predicts floods 7 days in advance
As a result, the AI tool generated more realistic and accurate satellite images than those from Hurricane Harvey in 2017.
How does the AI flooding tool work?
The MIT News website says the tool combines generative artificial intelligence with a physics-based flood model.
Specifically, they used a generative adversarial network (GAN). It’s a machine learning method that generates realistic images using two competing or “adversarial” neural networks.
On the other hand, the physics-based flood model incorporates real, physical parameters such as flood patterns.
Consequently, the AI flooding tool creates realistic, birds-eye-view images of a region, showing where flooding is likely to occur according to a storm’s strength.
More importantly, the physics-based flood model prevents hallucinations or false data, such as showing floods in places where it’s impossible.
However, the scientists emphasized that the AI flooding tool is a proof-of-concept. In other words, it only demonstrates how generative AI models can generate trustworthy content when paired with a physics-based flood model.
Björn Lütjens, a postdoc in MIT’s Department of Earth, Atmospheric, and Planetary Sciences, stated:
“One day, we could use this before a hurricane, where it provides an additional visualization layer for the public.”
“One of the biggest challenges is encouraging people to evacuate when they are at risk. Maybe this could be another visualization to help increase that readiness.”
Dava Newman, professor of AeroAstro and director of the MIT Media Lab, is excited for community leaders to use their AI flooding tool:
“We can’t wait to get our generative AI tools into the hands of decision-makers at the local community level, which could make a significant difference and perhaps save lives.”
The research team published their findings in the journal IEEE Transactions on Geoscience and Remote Sensing.