Methane emissions from US low production oil and natural gas well sites

The study addresses the discrepancy between inventory-based and measurement-based estimates of methane (CH4) emissions from the US oil and natural gas (O&G) production sector. Prior work shows US O&G CH4 emissions are substantially higher than EPA greenhouse gas inventory estimates, with production sites a key contributor. US production is highly skewed: a small share of wells produces most O&G, while numerous older, low-productivity wells dominate site counts. CH4 emissions at production sites display heavy-tailed distributions, where a minority of sites contributes disproportionately to total emissions. The research question is how much low production well sites (≤15 boed) contribute to national O&G site-level CH4 emissions and how their production-normalized loss rates compare to the broader sector. The purpose is to integrate national production data with site-level measurements to estimate total CH4 emissions from low production sites and assess mitigation implications. The importance lies in informing effective CH4 reduction strategies and policy, given methane’s high near-term warming potential and the abundance of low production sites.

Measurement-based studies consistently report higher O&G CH4 emissions than EPA inventories, with Alvarez et al. (2018) estimating national emissions ~60–70% higher than inventory values for 2015. Discrepancies are attributed largely to production sector underestimation, including fugitive equipment leaks and storage tank emissions. Emissions at production sites exhibit heavy-tailed distributions where few sites dominate totals, a pattern documented across the O&G supply chain. Correlations between emissions and site attributes (production rate, age, water production) are generally weak, though production declines rapidly post-drilling, increasing the population of low-productivity wells. Recent basin-level and wellhead-focused studies show high CH4 loss rates at marginal wells, and aerial surveys reveal intermittent but very large emission events. These findings motivate a focused national assessment of low production sites’ contribution to total CH4 emissions.

Data sources and site definition: The authors use Enverus Prism proprietary monthly well-level production data for 2019, filtering for active onshore wells (n≈842,978). They aggregate wells to sites via geospatial clustering (buffer radius 25 m for vertical wells, 50 m for horizontal wells) to derive site-level O&G production (boed), estimating 565,000 low production well sites (≤15 boed) with +2/−5% uncertainty and an average 1.03 wells per site. Sites are categorized into four production cohorts: >0–2, 2–5.4, 5.4–9.7, and 9.7–15 boed. Regional groupings include Appalachian, Oklahoma/Kansas/Arkansas, Colorado/Utah/Wyoming, Permian, and Barnett. CH4 measurement dataset: The study compiles 240 site-level CH4 emission measurements for low production sites from six peer-reviewed studies across six basins (Appalachian, Delaware/Permian, Barnett, Uinta, Upper Green River, Denver-Julesburg). Measurements use ground-based downwind methods (tracer flux or Gaussian plume/OTM-33A) that estimate total site emissions without requiring site access. Reported attributes include mean site-level CH4 emission rate and contemporaneous O&G production; some sites are oil-only. Method detection limits are ~0.01–0.036 kg CH4/h. The dataset covers 0.01–15 boed, with noted oversampling of sites >5 boed accounted for in modeling. A Kolmogorov–Smirnov analysis confirms statistical similarity of emissions distributions among key basins with n>25, justifying data consolidation. Emissions modeling approach: The authors develop a hybrid framework combining (1) nonparametric bootstrap resampling and (2) a nonparametric Bayesian regression linked to production, plus Monte Carlo extrapolation: - Define high-emitters as the top 5% of sites by absolute emissions; empirical threshold >7.3 kg CH4/h. Bootstrap (10^4 samples) to estimate the frequency of high-emitters and their contribution to total emissions (median ~50%, range 20–75%), and to derive a distribution of high-emitter emission rates. Also bootstrap the frequency of below-detection-limit (BDL) sites. - For the bottom 95% of sites with detectable emissions, fit a nonparametric Bayesian regression (log-normal likelihood) of site-level CH4 emissions versus site-level production using a cubic B-spline with three knots (including 2 boed), weak priors, and NUTS sampling (PyMC3). Posterior draws (5,000 samples) generate production-dependent emission factors. This accounts for weak dependence above ~2–4 boed and declining emissions at ultralow production. - Extrapolation to national totals: Randomly sample the fraction of high-emitters and assign high-emitter rates (restricted to sites >2 boed). Assign bottom-95% sites production-binned posterior mean emissions, decrementing by sampled BDL frequency. Repeat 500 times to generate distributions for (i) high-emitters, (ii) bottom 95%, and (iii) all sites. Multiply site-level hourly emissions by reported production days to annualize. Compute 95% confidence intervals from the 2.5th–97.5th percentiles. Sensitivity/alternative model: A production-independent lognormal model yields an emissions factor of 3.2 kg CH4/h/site (95% CI: 0.8–18), higher but within the primary model’s uncertainty bounds. Uncertainty considerations include limited sample size (n=240), emissions variability, production data uncertainties, and representativeness across basins and production cohorts.

- National emissions from low production sites: Estimated 4 Tg CH4/y (95% CI: 3–6 Tg) in 2019 from 565,000 active onshore low production sites. - Share of total O&G production site emissions: Low production sites account for 54% (95% CI: 37–75%) of total US O&G site CH4 emissions, relative to 7.6 Tg/y for all production sites (Alvarez et al.). - Loss rates: Production-normalized CH4 loss rate for low production sites is 13% (95% CI: 8–17%) assuming 80% CH4 content in gas, which is 6–12 times higher than the ~1.5% mean loss rate for all O&G sites. - Comparison to major basin: Emissions from low production sites exceed total Permian Basin CH4 emissions (2.7 Tg/y) by ~50%. - Emissions distribution: Top 5% of sites (>7.3 kg CH4/h) contribute ~50% (95% CI: 20–80%) of low production site emissions. 90% of sites emit <1 kg/h. About 50% of sites emit >10% of their CH4 production. - Average site-level rate: Mean site-level emission rate is 0.8 kg/h/site (95% CI: 0.5–1.2), ~50% lower in absolute terms than the mean for all US natural gas production sites (1.7 kg/h/site), yet with much higher loss rates due to lower production. - Ultralow production cohort (≤2 boed): Accounts for 25% (95% CI: 17–49%) of low production site emissions, representing ~10% of all US production-site CH4 emissions while contributing only 0.7% of national O&G production. - Regional results: Appalachian region emits ~1.2 Tg CH4/y (95% CI: 0.8–1.9) from low production sites, with an estimated regional loss rate of 26% (95% CI: 17–40%). Ultralow sites constitute ~90% of low production sites in Appalachia and ~50% of its low-production emissions. - High loss rates at low production: Data indicate increasing CH4 loss rates as production declines; wellhead-only studies in WV and OH report mean loss rates of ~8.8% (0–3 boed) and ~21% (<1 boed), respectively. - Intermittent large events: Rare, very large emissions (∼50–800 kg/h) identified at a small fraction (~0.05% in Permian low production sites) further underscore mitigation potential.

Findings demonstrate that despite contributing only ~6% of national O&G output, low production sites are responsible for a disproportionately large share (about half) of production-site CH4 emissions, driven by heavy-tailed emission distributions and elevated production-normalized loss rates, especially at ultralow production levels. The tendency for higher loss rates at lower production is consistent with a shift from production-dependent to production-independent sources (e.g., fugitive leaks, pneumatic venting) as wells age and output declines. These results resolve part of the inventory–measurement gap by highlighting the significant role of infrequent but large emission events and widespread moderate emissions among low production sites. The analysis underscores that effective national methane mitigation must target low production wells, particularly identifying and repairing high-emitting sites and addressing maintenance deficiencies. Regional patterns (notably Appalachia) and operator distributions indicate targeted strategies could yield substantial reductions.

The study integrates national production data with consolidated site-level CH4 measurements to produce a national estimate of methane emissions from low production O&G well sites. It shows these sites emit approximately 4 Tg CH4/y (2019), representing roughly half of all O&G production site emissions in the US, with loss rates far exceeding sectoral averages. Ultralow production sites contribute substantially to emissions despite minimal production, and the Appalachian region is a major contributor with very high loss rates. The work highlights that achieving significant CH4 reductions necessitates inclusion of low production sites in mitigation and regulatory frameworks. Practical mitigation options include routine LDAR programs, addressing storage tank and pneumatic emissions, and prioritizing maintenance. Future directions include expanding measurement-based datasets across regions and site types, improving detection of intermittent high emitters, refining emissions models, reassessing economic incentives that sustain marginal wells, and evaluating policies for plugging and abandoning uneconomic, high-emitting low production wells.

- Limited measurement sample size (n=240) relative to the national population increases uncertainty, especially in characterizing the tails of the emissions distribution. - Ground-based downwind methods quantify site-level totals but generally cannot resolve source-specific contributions; some studies report below-detection-limit values that require modeling adjustments. - Production data are proprietary and may include reporting uncertainties; well-to-site aggregation and buffer assumptions introduce uncertainty (+2/−5% in low production site count). - Measurement dataset oversampled sites with >5 boed relative to the national distribution, requiring model correction; basin coverage, while broad, is not fully comprehensive. - Assumptions such as 80% CH4 content in gas and the definition of high-emitters (top 5%) influence loss rate and contribution estimates; intermittency of large events may lead to under- or over-estimation if not fully captured during measurements.