POWER OF AI: Methane satellites proliferate, turning to AI to handle data deluge

Highlights

Supported by ESG, energy transition

Companies spending on machine learning

In early 2024, a new satellite known as MethaneSAT will beam its first megabyte of data to Earth where scientists stand ready with advanced modeling and machine learning -- artificial intelligence -- to pinpoint and measure emissions from leaking oil and gas production sites worldwide.

Around the same time, a coalition of partners led by the non-profit Carbon Mapper project will launch the first of two satellites with the urgent mission to capture high-resolution imagery of methane escaping from oil and gas facilities, as well as strong point sources such as landfills or agricultural operations. Carbon Mapper's data will be used to inform actions to reduce these potent greenhouse gases as well as support initiatives like Climate TRACE, a collaboration of academic and private-sector players who are training artificial intelligence to analyze and calculate greenhouse gas emission sources.

High-tech methane detection is gathering pace against the backdrop of ESG imperatives and energy transition, which have bolstered renewable energy and drawn increasing scrutiny to fossil fuels. Declining costs to launch monitoring satellites, as well as artificial intelligence (AI), which makes parsing terabytes of emissions data feasible, have given the oil and gas industry an emerging tool for environmental stewardship.

"Over the past five years, and I definitely anticipate that over just the next 12 months alone, there are a lot of satellites that we're kind of keeping our tabs on that are expected to go into orbit," Emmanuel Corral, senior emissions analyst for the Center of Emissions Excellence at S&P Global Commodity Insights said in an interview. "There are a few more within the next couple of months, so the landscape is really just constantly changing and evolving as more and more technology goes up in space."

Investors are watching

Increasingly, timely satellite images of methane plumes visible on dashboards and publicly available emissions data are also giving the finance sector critical information, proponents of the technology say.

"Knowing who's a better operator, who's the worst operator, who's leaking a lot -- I think that becomes very material to [investors]," said Deborah Gordon, a senior principal with RMI, a nonprofit that leads the modeling of emissions from the oil and gas sector for Climate TRACE. "The other party that's been very interested in this is the insurance and reinsurance industry."

Companies are spending at different rates on machine learning for emissions management, but there is immense interest in utilizing AI to cut emissions and maximize efficiency, according to interviews and public remarks of various industry players.

At the same time, some question the wisdom of spending billions on satellites and cloud computing to try to solve a problem that has been known and largely neglected for decades.

"Intelligence by itself, whether human or artificial, doesn't create a commitment to fix a problem," Clark Williams-Derry, an analyst with the Institute for Energy Economics and Financial Analysis, said in an email. "Knowledge isn't the same thing as a will to act. There's a risk that we'll pretend that AI will 'fix' a problem, when all it will do is give us yet another source of information that we'll collectively choose to neglect."

Leak detection is a necessary component of methane mitigation, but measurement alone does not equal emissions reductions, agreed Erin Tullos, a senior advisor to the United Nations' Oil and Gas Methane Partnership.

"What you want to do is ensure that your measurements are sufficiently accurate to inform your mitigation strategies," Tullos said in a May 2023 report about methane mitigation published by the Industrial Decarbonization Network. "You could try and accurately measure everything. The problem with that scenario is that it might just create a lot of cost and measurement for the sake of measurement."

Some also question the wisdom of spending potentially billions to track methane leaks from space. How far the satellite-AI field goes in pursuit of methane reductions also depends on public funding and private investment, observers say.

GHGSat, a company with nine satellites in orbit serving the oil and gas industry is the first of its kind in the US. The company in 2022 received a $7 million, five-year federal grant to support NASA with methane data, but federal funding for such projects has been largely absent.

Data supports mitigation

With AI, powerful computers can process and analyze huge amounts of data in a matter of a few days, work that used to take weeks or months, said Ritesh Gautam, a lead senior scientist with the team working to launch MethaneSAT. The automation also reduces the risk of human error, he said.

The satellite is a project of the Environmental Defense Fund, which has long pushed for reining in methane gas. The group is planning an advocacy campaign around MethaneSAT and expects to use the data the satellite collects to spur action among industry operators and regulators.

"That's why very, very high-quality data has to be produced that can withstand scrutiny of really anyone who wants to track mitigation... and the commitments made by the oil and gas industry and governments around the world," Gautam said.

More robust data on emissions -- produced from more accurate measurement -- could lend credibility to the argument that fugitive methane emissions need to be mitigated, for the sake of the environment as well as for improved production economics.

The raw satellite data is transmitted to an earth station and can then be found on cloud servers where Gautam and the rest of the MethaneSAT mission operations team compute and calibrate the data with the help of AI.

The team had already built a database of existing oil and gas infrastructure down to individual facilities on which the AI technology is trained and validated. The plan is to be able to link emissions shown on a satellite image to a particular site.

Importantly, the team will be the first in the industry to convert the satellite methane data, which is described as parts per billion, to emissions expressed as kilos of methane released every hour from a larger facility or a cluster of small sites, Gautam said.

It takes 'a constellation'

Using AI to ferret out methane pollution is a nascent field that may be on the cusp of taking off.

Duke Energy in 2022 pioneered a methane monitoring platform that uses satellites, ground-level sensors, AI and cloud computing to detect leaks and measure methane emissions from its natural gas distribution system in "near real-time."

The project, a collaboration with Accenture and Microsoft, recently received $1 million from the US Department of Energy to extend the platform to the Transco pipeline owned by Duke Energy supplier Williams.

In January, the United Nations Environment Programme partnered with French satellite imaging company Kayrros to detect and track methane leaks in the US and other nations that signed the Global Methane Pledge.

At the same time, satellites tasked with other jobs are finding they can also transmit images of methane sources from oil and gas fields, further broadening the field. A NASA mission launched in July 2022 to measure earth dust from the International Space Station, for example, found that it could also detect dozens of methane "super-emitter" events in a matter of a few months.

"It turns out that different satellites are designed to do different things well, so it really takes a constellation of these satellites to tell the whole story," Gordon said. "The more satellites you put up the more things you can do."