Satellite-detected oil-gas methane point sources emitted 7.6 Tg y-1 in 2025

Why it matters

Satellites see only brief snapshots of methane plumes, yet inventories need yearly averages. This method turns sparse, open-access plume data into annual rates for individual sources. The results suggest many large sources are underestimated or missing in a bottom-up inventory built from national reports.

What they did

The authors used plume observations and clear-sky non-detections from Sentinel-2, Landsat, EnMAP, PRISMA, EMIT and Tanager, compiled by UNEP’s IMEO and Carbon Mapper. They fit a probability of detection (POD) for each instrument with Bayesian inference. They adjusted it locally using background noise maps from EMIT and Sentinel-2. For each source and year they combined the emission rate distribution, local wind speed and persistence into an annual mean rate.

Key findings

• The fitted PODs are lower than those commonly reported from controlled releases under favorable conditions.
• Annual mean rates were estimated for 1924 individual sources. Detected point sources totaled 7.6 Tg y-1 in 2025, and 245 sources emitted more than 1 ton h-1 on an annual basis.
• 18% of sources detected in 2022-2024 were not detected in later years. Of 78 sources detected in 2022, 47 were detected in all four years.
• Counting emissions below the detection threshold gives a mean persistence of 51%, and 56% in 2023. Jervis et al. (2025) reported 16%.
• Estimates correlate weakly with the GFEI v3 inventory (r = 0.18) and are 2.5-fold larger on average in observed grid cells. Adding missing point sources would raise GFEI by 5.8 Tg y-1 (14%).

Limitations

• Error on annual rates averages 67%, mostly from the lognormal fit. Individual estimates are therefore uncertain.
• The POD is derived from detections only, so it overstates a satellite’s true global detection ability. Offshore sources and low-wind observations are excluded, and cloudy tropical basins are poorly observed.
• Agreement with the GHGSat estimates is modest (r = 0.36). The method assumes ergodicity and that local noise ratios from EMIT and Sentinel-2 apply to the other instruments.

Glossary

• Probability of detection (POD): The chance that an instrument detects a plume of a given emission size.
• Persistence: The fraction of time a source is actually emitting.
• Plume-imaging satellite: A satellite that images methane plumes from individual sources using sunlight reflected in shortwave infrared.
• GFEI: The Global Fuel Exploitation Inventory, a bottom-up map of fossil fuel methane emissions based on national reports.