Gen AI creates accurate satellite images of flooding

Known as the ‘Earth Intelligence Engine’, the tool was developed as a visualisation aid to inform the public of the potential effects of impending storms. The MIT team has made the tool available online so that people can see its results in action. The work is published in the journal IEEE Transactions on Geoscience and Remote Sensing. 

“The idea is: One day, we could use this before a hurricane, where it provides an additional visualisation layer for the public,” said research lead Björn Lütjens, a postdoc in MIT’s Department of Earth, Atmospheric and Planetary Sciences.

“One of the biggest challenges is encouraging people to evacuate when they are at risk. Maybe this could be another visualisation to help increase that readiness.”

Initially, the researchers used AI on its own to create the synthetic images. They applied a generative adversarial network (GAN), a type of machine learning method that can generate realistic images using two competing neural networks. While this model produced realistic images, it also generated ‘hallucination’ floods at locations where flooding was not currently possible.