Repeated satellite looks estimate large methane emissions per Permian facility

Why it matters

Single plume detections do not give yearly figures for a facility. This method turns sparse satellite sightings into facility-level annual estimates with uncertainty. These could be set beside bottom-up inventories and aerial surveys. The totals cover only large, satellite-detectable emissions, so they are not full basin inventories.

What they did

The authors linked GHGSat plume detections and clear-sky non-detections from 2023 to 2025 to mapped facilities in the Permian Basin, about 100,000 in all. They built a Bayesian model for each facility-year. It accounts for the chance that each observation would miss a plume. It also borrows information from similar facilities in the same class: wells, compressor/gathering stations and gas processing plants. They compared estimates between years and summed them into basin totals.

Key findings

• Mean annual large-emitter emission rates were about 0.46 kg h⁻¹ for wells, 32–36 kg h⁻¹ for compressor/gathering stations and 139–144 kg h⁻¹ for gas processing plants.
• Large-emitter persistence averaged about 0.001 for wells, 0.089–0.091 for compressor/gathering stations and 0.238–0.249 for gas processing plants.
• About 30–35% of midstream facilities accounted for 50% of estimated annual large-emitter emissions.
• More than 98% of matched midstream facilities showed no year-to-year change larger than the combined uncertainty.
• Extrapolated basin totals were 1.70, 1.66 and 1.63 Mt y-1 for 2023, 2024 and 2025. That is roughly 32% to 59% of earlier basin-wide estimates.

Limitations

• Estimates cover only large emissions the satellite can see (about 50 kg h⁻¹ or more). They exclude smaller emissions and classes such as transmission pipelines, so they are not complete basin inventories.
• Well estimates are just the class-level average, because well detections were too rare for facility-specific estimates. Basin totals also scale observed facilities up to all mapped facilities.
• This is a preprint. Comparisons with aerial and satellite studies are limited by differences in study period, facility definitions and detection thresholds.

Glossary

• Large-emitter persistence: The fraction of time a facility is in an emission state within the large-emitter range, not the fraction of time it emits at any rate.
• Probability of detection: The chance that a satellite observation would spot a plume of a given emission rate under that day’s wind and retrieval noise.
• Hierarchical Bayesian method: A statistical approach that combines each facility’s own data with a prior drawn from similar facilities.
• Credible interval: A range that contains the true value with stated probability (here 95%), given the model and data.