Visualizing the potential effects of a hurricane on people's homes before it hits can assist residents prepare and decide whether to leave.


MIT researchers have actually established a method that produces satellite images from the future to illustrate how a region would take care of a possible flooding occasion. The method integrates a generative expert system model with a physics-based flood model to develop practical, birds-eye-view images of a region, showing where flooding is most likely to take place provided the strength of an oncoming storm.


As a test case, the group applied the technique to Houston and generated satellite images portraying what certain places around the city would look like after a storm similar to Hurricane Harvey, which hit the region in 2017. The team compared these produced images with actual satellite images taken of the very same regions after Harvey hit. They also compared AI-generated images that did not consist of a physics-based flood design.


The team's physics-reinforced method generated satellite pictures of future flooding that were more realistic and accurate. The AI-only approach, on the other hand, created pictures of flooding in places where flooding is not physically possible.


The group's technique is a proof-of-concept, implied to demonstrate a case in which generative AI designs can generate realistic, credible material when matched with a physics-based design. In order to use the approach to other areas to portray flooding from future storms, it will need to be trained on a lot more satellite images to discover how flooding would search in other areas.


"The concept is: One day, we might utilize this before a cyclone, where it offers an additional visualization layer for the general public," says Björn Lütjens, a postdoc in MIT's Department of Earth, Atmospheric and Planetary Sciences, who led the research study while he was a doctoral student in MIT's Department of Aeronautics and Astronautics (AeroAstro). "Among the greatest challenges is motivating people to leave when they are at danger. Maybe this might be another visualization to help increase that readiness."


To show the potential of the brand-new method, which they have dubbed the "Earth Intelligence Engine," the group has made it available as an online resource for others to try.


The scientists report their results today in the journal IEEE Transactions on Geoscience and Remote Sensing. The study's MIT co-authors consist of Brandon Leshchinskiy; Aruna Sankaranarayanan; and Dava Newman, professor of AeroAstro and director of the MIT Media Lab; together with partners from numerous organizations.


Generative adversarial images


The new study is an extension of the group's efforts to use generative AI tools to envision future climate scenarios.


"Providing a hyper-local perspective of environment appears to be the most effective method to communicate our clinical outcomes," says Newman, the research study's senior author. "People relate to their own postal code, their regional environment where their friends and family live. Providing regional environment simulations ends up being user-friendly, individual, and relatable."


For this research study, the authors utilize a conditional generative adversarial network, or GAN, a kind of device learning approach that can generate practical images using 2 completing, or "adversarial," neural networks. The very first "generator" network is trained on pairs of genuine information, such as satellite images before and after a typhoon. The second "discriminator" network is then trained to identify between the real satellite images and the one manufactured by the very first network.


Each network instantly improves its efficiency based on feedback from the other network. The concept, then, is that such an adversarial push and pull need to ultimately produce artificial images that are indistinguishable from the genuine thing. Nevertheless, GANs can still produce "hallucinations," or factually incorrect functions in an otherwise realistic image that shouldn't exist.


"Hallucinations can deceive viewers," states Lütjens, who started to wonder whether such hallucinations could be prevented, such that generative AI tools can be trusted to assist notify people, particularly in risk-sensitive situations. "We were believing: How can we utilize these generative AI models in a climate-impact setting, where having trusted data sources is so essential?"


Flood hallucinations


In their new work, the researchers considered a risk-sensitive situation in which generative AI is charged with developing satellite pictures of future flooding that could be credible enough to notify decisions of how to prepare and possibly leave people out of damage's way.


Typically, policymakers can get an idea of where flooding might take place based upon visualizations in the form of color-coded maps. These maps are the end product of a pipeline of physical models that generally begins with a typhoon track model, which then feeds into a wind design that replicates the pattern and strength of winds over a local region. This is integrated with a flood or storm rise design that forecasts how wind may press any nearby body of water onto land. A hydraulic design then maps out where flooding will occur based on the local flood infrastructure and produces a visual, color-coded map of flood elevations over a specific region.


"The concern is: Can visualizations of satellite imagery add another level to this, that is a bit more tangible and mentally engaging than a color-coded map of reds, yellows, and blues, while still being trustworthy?" Lütjens states.


The group first checked how generative AI alone would produce satellite images of future flooding. They trained a GAN on actual satellite images taken by satellites as they passed over Houston before and after Hurricane Harvey. When they charged the generator to produce new flood pictures of the exact same areas, they found that the images resembled common satellite imagery, but a closer look exposed hallucinations in some images, in the form of floods where flooding must not be possible (for example, in places at greater elevation).


To reduce hallucinations and increase the trustworthiness of the AI-generated images, the group paired the GAN with a physics-based flood design that integrates genuine, physical parameters and phenomena, such as an approaching hurricane's trajectory, storm rise, and flood patterns. With this physics-reinforced method, the group created satellite images around Houston that portray the exact same flood extent, pixel by pixel, as forecasted by the flood design.