Google and NASA launch AI model to track methane emissions globally
New deep learning tool detects 50% more methane plumes than human analysts with 60m precision.
1Breakthrough in methane detection
Google Research and NASA JPL have developed a deep learning model that identifies methane plumes with 60-meter precision from space. The AI detects 50% more plumes than human analysts, including 23,000 previously undetected emissions. This includes 24 of the world's 25 highest-emitting landfills, significantly improving upon existing NASA detection capabilities. The technology represents a major leap in environmental monitoring accuracy.
2Technical specifications
The vision transformer model was trained on 3.6 million simulated methane plumes. It processes raw spectral data from NASA's EMIT instrument aboard the International Space Station. At 60 meters per pixel resolution, it can pinpoint specific landfill cells - a dramatic improvement over previous 5.5km resolution systems. The model's code is available on GitHub, with datasets hosted on Earth Engine under Creative Commons licensing.
3European context
While Europe's Methane Regulation mandates satellite monitoring for oil, gas and coal sectors, it excludes waste management. The Google-NASA tool fills this gap by identifying landfill emissions. Prognos estimates European landfills will emit 1.5 billion tonnes of CO2 equivalent as methane by 2050. Current EU systems like Copernicus offer only 5.5km resolution, insufficient for facility-level monitoring required by new regulations.
4Comparison with existing systems
The new AI model outperforms NASA's previous methane detection product, which identified only 17 of the top 25 emitting landfills. Europe's Sentinel-5P satellite provides daily global coverage but at much lower resolution. UCLA's 2023 analysis using Carbon Mapper data revealed the limitations of existing systems, prompting this technological advancement for more precise emission tracking.
5Regulatory implications
Article 31 of Europe's Methane Regulation requires member states to monitor super-emitting events, but current tools lack precision. The Google-NASA model could become the de facto standard despite its unofficial status. However, its 51.6-degree orbit inclination means optimal coverage isn't over Europe, highlighting gaps in regional monitoring capabilities even with advanced technology.
6Upcoming European solutions
Europe is developing higher-resolution alternatives, with AIRMO securing €5M funding for a 2027 satellite launch. The Copernicus CO2M mission, also scheduled for late 2027, will complement these efforts. These projects aim to provide regulatory-grade data for compliance monitoring, but face years-long development timelines while methane emissions continue unchecked.
7Accessibility and limitations
The methane plume database is freely available on Earth Engine, with model code on Kaggle and GitHub. However, the EMIT instrument was originally designed for mineral dust mapping in arid regions, not optimized for methane detection. This serendipitous application demonstrates how space technology can be repurposed for climate action beyond original design parameters.
8Industry impact
This breakthrough highlights the growing role of AI in environmental monitoring. The technology's precision enables targeted regulatory action and corporate accountability. As climate reporting requirements tighten globally, such tools will become essential for compliance. The open-source approach also sets a precedent for collaborative climate tech development between tech giants and space agencies.
Original article: The Next Web — Google and NASA launch AI model to track methane emissions globally