Annual mean methane emissions from individual oil-gas point sources worldwide inferred from multi-satellite plume-imaging observations

James D. East , Daniel J. Jacob, Harshil M. Kamdar, Jack Bruno, Lucas A. Estrada, Itziar Irakulis-Loitxate, Manuel Montesino San Martin, Juan Emmanuel Johnson, Gonzalo Mateo Garcia, Dylan Jervis , Daniel J. Varon , Vassiliki Mancoridis, Daniel H Cusworth, Emily K. Reidy, Yuanlei Chen

Plume-imaging satellite instruments routinely reveal large methane point source emissions from oil and gas systems. But the observations are for instantaneous plumes, and the instruments have limited detection and revisit capabilities, making it challenging to infer time-averaged emission rates as needed for emission inventories and reporting. Here we develop a general method to infer the probability of detection (POD) of any plume-imaging instrument, over any surface, and to estimate the time-averaged point source emission rates on the basis of a limited collection of data. Applying this method to 12080 plume observations from the Sentinel-2, Landsat, EnMAP, PRISMA, EMIT, and Tanager instruments for each year in 2022-2025, we find that PODs are lower than commonly reported from controlled releases conducted under favorable conditions. We quantify annual mean emission rates for 1924 individual oil and gas point sources worldwide, estimate that point sources detected by these instruments emitted 7.6 Tg y-1 in 2025, identify 245 sources that emitted more than 1 ton h-1 on an annual basis, and show 2022-2025 trends for illustrative individual sources. 18% of sources emit for only one year and subsequently cease. Accounting for sub-detection-threshold emissions reveals higher source persistence (averaging 56%) than previously reported for oil-gas sources globally.