A global satellite-based methane detection programme coordinated by the UN Environment Programme (UNEP) has contributed to more than 40 methane mitigation actions worldwide since it became operational in 2024. AI is allowing it to spot the leaks and notify those causing the emissions much faster.
The UNEP’s International Methane Emissions Observatory (IMEO) processes data from a network of more than 30 satellites that feed its Methane Alert and Response System (MARS) in order to notify governments and companies of opportunities to act on methane reductions.
It is a slow and laborious task for humans, but AI is dramatically speeding up the process, enabling IMEO analysts to process 12 to 15 times more data while maintaining scientific rigour, according to a new UNEP report.
In Spotlighting Opportunity: How Artificial Intelligence is Accelerating Methane Action, UNEP says AI-assisted workflows have identified 80–85% of confirmed methane detections before reaching the expert review stage, helping to mitigate sources estimated to have emitted 1.2m tonnes of methane.
Martin Krause, Director of UNEP’s Climate Change Division, said that, as new satellite missions increased the volume of methane data available globally, the challenge was no longer finding emissions but acting on them.
“UNEP’s experience shows how AI can help bridge that gap, enabling faster identification of major methane releases and helping convert data into measurable emissions reductions,” he said.
MARS integrates data from satellite instruments and uses AI models to distinguish methane emissions, such as leaks from oil and gas facilities, from environmental noise. Since 2023, the system has analysed over 1.3m satellite observations. The next step is to expand to other sectors, including coal and waste.
UNEP is keen to stress that humans remain firmly in the loop. Every AI-flagged detection is independently reviewed and verified by IMEO analysts before a notification is issued.
It has also been mindful of the potential for the programme to actually add to greenhouse gas emissions through energy-intensive AI processing in data centres. UNEP said the methane-monitoring models used were designed to be lightweight and energy efficient, so they would use relatively limited computing resources, while still delivering the required monitoring capacity.
Datasets and code used by the programme are also being made openly available to researchers, governments, and other organizations to broaden access to its methane-monitoring tools and experience.
Call to action
Despite the breakthroughs, turning the ability to detect leaks and notify emitters faster into targeted action on climate change remains a major challenge. UN Secretary-General, António Guterres, has called on countries to respond to 80% of methane alerts received through MARS by 2030. To date, only 10 countries have achieved that level of response
Raising the response rate is seen as crucial to tackling global warming. Methane is around 80 times more powerful as a greenhouse gas than CO₂ over a 20-year period, but it remains in the atmosphere for only about a decade. That means cutting methane emissions can make a significant impact on the rate of warming in the relatively near term, as well as providing benefits such as reduced air pollution and improved crop yields.
In the report, UNEP points to success in tackling a methane leak in Kazakhstan after satellite detection and AI-driven monitoring as an example of what can be achieved by the MARS system. As we reported in early 2024, a methane plume was detected by a satellite in the Mid Caspian Basin in 2023. This was then monitored by the MARS AI system, which detected seven plumes in mid-2024. The source was estimated to have had an average emission rate of 1.79 metric tons of methane per hour.
UNEP notified the government and the company involved, leading to an on-site inspection that traced the emissions to a malfunctioning valve. This was then replaced, and, by October 2024, emissions had ceased.
UNEP’s Krause said the techniques being developed by the programme had implications beyond methane detection.
“The real lesson from this work is that AI can help convert the explosion of environmental data into practical action. By combining scientific expertise with lightweight and energy-efficient AI tools, we can respond faster not only to methane emissions, but to a wide range of environmental challenges,” he said
(Photo: Shutterstock)
