Observation-derived 2010-2019 trends in methane emissions and intensities from US oil and gas fields tied to activity metrics

Significance The United States accounts for a large share of global methane emissions from the oil/gas industry. Analysis of satellite and surface observations of atmospheric methane reveals larger-than-reported year-to-year variability of 2010 to 2019 US oil/gas methane emissions. This variability reflects trends in oil/gas production rates, number of active wells, and drilling of new wells. Emissions surged after 2017 as production increased. The methane intensity from the US oil/gas industry (methane emitted per unit methane gas produced) decreased steadily after 2010. Extension of this decreasing trend to 2030 (target date of the Global Methane Pledge) would result in a 32% decrease in US oil/gas methane emissions and 15% decrease in total anthropogenic emissions relative to 2019 despite an increase in production.


Figures S1 to S9
Tables S1 to S4          1.0 1.0 Normal 50% 5 a Settings different to the base inversion are underlined.For each inversion, we apply two methods to allocate the posterior correction factor of total methane emissions to the oil/gas sector, as introduced in Methods, so the 12-member inversion ensemble yields 24 estimates of oil/gas emissions.c Prior estimate of emissions from the oil/gas emissions in Permian was increased by a factor of 4 (-3.2Tg a -1 ) from the EPA inventory, reflecting previous evidence that the EPA inventory is too low.Table S2.Field and inversion estimates of oil and gas emissions (Gg a -1 ) in different US production regions.

Figure S1 .
Figure S1.Prior emissions used in the inversion.Prior methane emissions are shown for (A) all sectors, (B) oil, (C) gas, (D) wetlands, and (E) other emissions.Panel (F) shows the fraction of oil/gas emissions to total methane emissions.Anthropogenic emissions are from spatially gridded versions of the US, Canada, and Mexico official national inventories.Wetland emissions are from the mean of the high-performance subset of the WetCHARTs inventory ensemble.

Figure S2 .
Figure S2.Optimization of mean 2010-2019 methane emissions over North America.Results are from the base inversion using both GOSAT and GLOBALVIEWplus in situ observations, the GOSAT-only inversion, and the in-situ-only inversion.The left panels show the mean averaging kernel sensitivities (diagonal elements of the averaging kernel matrix).The degrees of freedom for signal (DOFS, defined as the trace of the averaging kernel matrix) are shown in the inset.The right panels show the posterior correction factors, i.e., the multiplicative factors applied to the total prior emissions in Fig. S1A.

Fig
Fig. S3 2010-2019 linear trends in methane emissions.The linear trends are fitted by linear regression to the inversion results for individual years.

Fig
Fig. S4 2010-2019 trends in oil/gas methane emissions in the US, Permian, Anadarko, and Marcellus from inversion ensemble.Each line represents an inversion result using different parameters.The red circles highlight the results from the base inversion (Inversion #1).

Fig. S5
Fig. S5 Same as Fig.3 but for different oil/gas production regions.

Fig. S6
Fig. S6Same as Fig.4but for methane intensity defined as oil/gas methane emissions normalized by the combined oil and gas production based on energy content, assuming one barrel of oil has the same amount of energy content as 6,000 cubic feet of natural gas.

Fig
Fig. S7 Evaluation of the posterior simulation to fit surface and tower, GOSAT, and TCCON observations for 2010-2019.Panels A and B show the mean differences between GEOS-Chem simulations and the observations using either prior or posterior methane emissions.Panels C-E show the model bias relative to surface and tower, GOSAT, and TCCON observations averaged for each year.The shadings represent the standard deviation of the bias.

Fig
Fig. S8 Ability of the inversion to separate posterior methane emission sectors.The figure shows posterior error correlation coefficients (r) between sectoral methane emissions in the US (CONUS) and three major oil/gas production regions, using the sector-aggregated error covariance matrix as described in Method.Error correlation coefficients indicate the ability of the inversion to separate emissions between sectors (0:perfectly, ±1: not at all).Grey shadings indicate that there is no emission from this sector in the region.Results are from the base inversion for the year 2015.Results for other years show similar patterns.

Figure S9 .
Figure S9.Comparison of uncertainty estimates in the prior and posterior oil/gas emissions over the US and individual production regions from derived from the posterior error covariance  ̂ and the inversion ensemble.

b
Adding the errors from individual sectors in quadrature following Maasakkers et al. (2021).

Table S1 .
Settings for generation of the 12-member inversion ensemble yielding 24 estimates of oil/gas emissions a .

Table S3 .
Summary of the multiple linear regression model used for prediction of oil/gas methane emissions in the US 1 .

Table S4 .
Summary of 2010-2019 mean oil/gas methane intensity (emission per unit methane gas production) and trends.