Geolabe's AI Satellite Tech: The Future of Accurate Methane Tracking

Claudia Hulbert (Geolabe, Cohort 2023)

When it comes to mitigating climate change, much of the discussion centers on reducing CO2 emissions or trapping CO2. But methane, a greenhouse gas responsible for about a third of global warming to date, is a lot more efficient than carbon dioxide at trapping heat in the atmosphere in the short term. As atmospheric methane levels reach an all-time high, governments and other organizations are realizing the urgency to address this issue and are looking to take effective remedial action. Curbing methane emissions is widely considered to be one of the fastest ways to slow global warming. The biggest challenge up to this point, however, is that no technology could detect methane emissions clearly and at scale.

Until now.

In a recent study published in Nature Communications, the U.S.-based remote sensing company Geolabe, in partnership with a research team from Kyoto University in Japan, developed an AI algorithm capable of sifting through massive amounts of data produced by a powerful European Space Agency satellite constellation. This algorithm is the first ever to automatically detect individual methane emissions at a global scale, anywhere on Earth, every few days. This offers a revolutionary, effective alternative for detecting methane in the future.

So, what are the downfalls of traditional satellite tech that make it difficult to identify methane leaks at scale? And how, specifically, is this new algorithm changing the satellite methane detection landscape?

“First off, traditional satellite methane detection isn't fully automated, so human verification is often required, which isn't scalable,” says Geolabe CEO Claudia Hulbert (Cohort 2023), one of the study’s co-authors. “There’s just no way that humans could manually check and verify all methane leaks. It's impossible.”

Secondly, Hulbert explains that detecting methane requires looking at images with many spectral bands, not just the typical red, green, and blue prevalent in current computer vision methods. With current satellite tech, the imagery is noisy (meaning it has unwanted information that distorts the image), which makes it hard to pinpoint emission sources.

“Traditional satellite methane detection methods observe around 15 percent of emissions from large oil and gas basins such as the Permian Basin,” Hulbert says. “Our technology can detect up to 85 percent by identifying smaller leaks through data denoising.”

Because no solution has existed previously that could detect methane at scale, governments and organizations have long relied on imperfect information to try to curb emissions. In fact, when organizations report emissions, the amount of methane produced is often underestimated. For example, studies have shown that the U.S. Environmental Protection Agency (EPA) underestimates the amount of methane leaking from U.S. oil and gas operations by as much as half.

“In order to track the presence of methane accurately and give organizations the right data to take quality remedial action, we need to have a clearer picture of how much methane is being released,” Hulbert says.

This is particularly critical as new stringent EPA regulations come into play, and the demand for accurate methane data is higher than ever. Another exciting prospect for this technology is that it could be used to analyze historical satellite data retroactively. This capability would help validate past emissions and identify trends over time—something we’ve never been able to do at that level of detail.

“It’s exciting to think about the variety of ways in which it can be used,” Hulbert says. “The energy sector could utilize it to fix leaks, the financial sector to make better investments, and agriculture could monitor emissions. There are so many possibilities. And it’s all just beginning.”