JACKSONVILLE, Fla. – Scientists at MIT are developing an artificial intelligence (AI) tool that generates realistic satellite images to illustrate potential flooding scenarios.
RELATED: Flooding took away the love of his life after 35 years without a day spent apart
Recommended Videos
According to Space.com, the tool combines a generative AI model with a physics-based flood model to identify areas at risk of flooding.
It then creates detailed bird’s-eye views of what the region might look like after a flood, based on the intensity of an approaching storm.
However, GANs (Generative Adversarial Networks) sometimes produce “hallucinations,” which are features in images that may appear realistic but are inaccurate or shouldn’t be present.
“Hallucinations can mislead viewers. We were considering how we could use these generative AI models in a climate-impact context, where having reliable data sources is essential. This is where the physics model comes into play,” said Lütjens.
“The idea is, one day, we could use this before a hurricane, where it provides an additional visualization layer for the public," Björn Lütjens, a postdoc in the Department of Earth, Atmospheric, and Planetary Sciences at the Massachusetts Institute of Technology (MIT), said in a statement.
“One of the biggest challenges is encouraging people to evacuate when they are at risk,” Lütjens said. “Maybe this could be another visualization to help increase that readiness.”
To demonstrate the model, the researchers applied it to a scenario in Houston, generating satellite images of flooding in the city after a storm similar in strength to Hurricane Harvey.
They compared the AI-generated images to actual satellite images and to images created without the assistance of the physics-based flood model.
The AI images generated without the physics model were highly inaccurate, featuring numerous “hallucinations,” mainly showing flooding in areas where it wouldn’t be possible.
In contrast, the images produced using the physics-reinforced method closely matched the real-world scenario.
The scientists envision that this technology will be useful for predicting future flooding scenarios.
It will provide reliable visuals that will help policymakers prepare for and make informed decisions about flood planning, evacuation, and mitigation efforts.
In their press release, the scientists noted that policymakers typically assess potential flood areas using visualizations, often color-coded maps.
“The question is: Can visualizations of satellite imagery add another level to this that is a bit more tangible and emotionally engaging than a color-coded map of reds, yellows, and blues while still being trustworthy?” Lütjens said.
Currently, the team’s method is in the proof-of-concept stage and requires additional time to analyze other regions to predict the outcomes of various storms more accurately.
“We show a tangible way to combine machine learning with physics for a use case that’s risk-sensitive, which requires us to analyze the complexity of Earth’s systems and project future actions and possible scenarios to keep people out of harm’s way.” Dava Newman, professor of AeroAstro and director of the MIT Media Lab said. ”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."