Google Research and NASA’s Jet Propulsion Laboratory have developed a deep learning model that identifies methane plumes globally at unprecedented resolution from space, detecting significantly more emission sources than traditional human analysis.
- AI maps methane plumes at ultra-fine 60-meter resolution
- Detects 50% more plumes than human analysts, 23,000 more sources
- Europe’s methane rules cover fossil sectors but exclude landfills
What happened
Google Research and NASA’s Jet Propulsion Laboratory have released a deep learning model capable of detecting methane plumes globally at a remarkable spatial resolution of 60 meters per pixel. The model leverages a vision transformer trained on over 3.6 million simulated plumes, analyzing raw spectral data captured by the EMIT instrument onboard the International Space Station.
Results demonstrate the model identifies roughly 50% more methane plumes than human analysts, adding around 23,000 new plumes overall. Notably, the tool detected emissions at 24 of the world’s 25 largest landfills, substantially improving the identification of landfill methane sources compared to previous NASA products.
Why it matters
Methane is a potent greenhouse gas, and accurate, high-resolution monitoring is critical to tackling emissions from major sources such as landfills, oil, gas, and coal facilities. Current European Union Methane Regulations mandate satellite monitoring and alert systems for fossil fuel sectors but exclude municipal waste landfills, which contribute significantly to methane emissions.
This advanced AI model fills a monitoring gap by providing precise, publicly available data on methane releases from waste sites worldwide. The openness of the data and model encourages transparency and supports regulatory and market efforts to address a broader set of emission sources, including municipal waste.
What to watch next
Europe’s methane monitoring infrastructure is evolving with upcoming missions like the Copernicus CO2M satellite launch scheduled for late 2027 and commercial entrants such as AIRMO aiming to deploy dedicated satellites soon. These initiatives will complement AI-driven tools by expanding coverage and sector inclusion.
Meanwhile, regulators and environmental agencies may incorporate datasets like Google and NASA’s methane plume maps to target super-emitters beyond fossil fuels, improving the comprehensiveness of emission controls and verification. How quickly and widely these tools integrate into official monitoring frameworks will be important to track.