A six-step diagnostic framework for understanding discrepancies between methane inventories and site-level measurements.
Background
Source-level inventories, often referred to as “bottom-up inventories” are based on an accounting of emissions attributable to equipment and activities. Scientific measurement studies provide a different perspective, using observations of methane in the atmosphere together with models and other inference methods to estimate emissions at source, site, regional, or larger scales. Historically, these approaches have sometimes produced substantially different estimates. The landmark study by Alvarez et al. from 2018 put the U.S. natural gas supply chain’s emissions roughly 60% above the Environmental Protection Agency inventory estimate. Such discrepancies helped motivate greater use of empirical observations to evaluate and improve emissions inventories.
Increasingly, empirical observations are being incorporated into emissions inventories through approaches often described as measurement-informed inventories (MIIs). The term encompasses a wide range of methods and does not, by itself, imply a particular level of accuracy, specificity or completeness. Measurements may provide information about individual sources, entire facilities, or larger spatial scales, and converting atmospheric observations into emission-rate estimates generally requires models, assumptions, and inference. Source-level measurements typically provide information about a discreet source in a particular operating mode that can be used in place of generic information that may be unrepresentative of the operator’s circumstances. Measurements at greater spatial scale (e.g. equipment-level, site-level) are usually used to provide complementary information to assess the completeness and representativeness of existing information. Measurement integration uses these measurements, together with operational and other contextual information, to build an inventory based on multiple observational scales. This inventory represents the best available estimate for methane emissions within the reported boundary. It accounts for applicable sources, operating modes, incidents and malfunctions observed within the entirety of the information. Each source of information carries its own uncertainties and limitations, making the basis on which different sources of information are evaluated together particularly important.
The UN Environment Programme’s Oil & Gas Methane Partnership 2.0 (OGMP) reporting framework requires member companies to build a source-level inventory for Levels 3 and 4, and then use site-level measurements to evaluate and reconcile the inventory at Level 5. OGMP frames reconciliation as an iterative process of investigation and not a one-off comparison of two independent numbers. The objective is to determine whether the inventory and observations are reasonably consistent given their respective temporal and spatial coverage and uncertainties, and to investigate discrepancies that may reveal incomplete or inaccurate characterization of emissions.
The term reconciliation doesn’t have a single, shared definition within the oil and gas industry. As used here, reconciliation is the structured evaluation of an emissions inventory against independent empirical observations to assess and improve consistency, completeness, and representativeness. It also includes investigating discrepancies among independent estimates. Theoretically, discrepancies should reduce in significance as the base emissions inventory improves. To establish a reasonably comparable basis, spatial boundaries, temporal coverage, operating conditions, detection limits, and uncertainty of the independent sets of information must be well understood. Understanding these parameters is critical for genuinely assessing areas of overlap in the data. Where inconsistency and discrepancy remain, the objective is to understand their cause and determine whether the underlying emissions characterization should be refined. Discrepancies may be identified that indicate both overestimates and underestimates of source-level inventories.
Diagnostic check-list:
Simply updating an emissions estimate retroactively to better represent historical operations misses much of the learning opportunity that is inherent to the reconciliation process. A discrepancy is a diagnostic signal, not a verdict that either estimate is wrong. It may reflect an incomplete or inaccurate characterization of emissions, but it may also arise because the two estimates represent different spatial boundaries, time periods, operating conditions, or portions of the emissions distribution.
When a source-level inventory and site-level measurements appear inconsistent, the first question is whether they represent the same emissions on a reasonably comparable basis. Once comparability has been established, remaining discrepancies tend to fall into a manageable set of causal categories that operators can investigate systematically and, hopefully, incorporate into mitigation planning.
The following checklist provides a step-by-step approach for diagnosing those discrepancies and determining when they indicate that the underlying inventory requires refinement to improve consistency, completeness, or representativeness.
1. Scope mismatch
An apparent discrepancy between an inventory and a site-level measurement-derived estimate may simply reflect that they do not represent the same sources and boundaries. The two numbers might include different lists of sites, so one includes facilities that the other has left out. Or perhaps a plume drifted over from a neighbouring operator’s well pad, adding emissions from equipment outside of the target inventory. Emissions from gathering equipment may get included as part of the estimate for production facilities, or the other way around.
These differences in scope can make two otherwise sound estimates appear inconsistent. Diligent reconciliation studies invest the resources upfront to align the two estimates before comparing them, ensuring they cover the same sites and equipment for the same period of time.
2. Temporal mismatch
Site-level measurements are usually snapshots of emissions at specific times. Bottom-up inventories, by contrast, commonly represent emissions over an annual reporting period. Trying to put the two together without accounting for the differences in spatial and temporal factors is like trying to compare apples and oranges.
The magnitude and persistence of high-emission events vary immensely. A 2022 study drawing on Barnett Shale data reports that 1% of production sites accounted for almost half of total methane emissions. The high emitting sites had an average rate of 39 kg/h. By contrast, the average emission rate for all production sites in the region was 1.76 kg/h (Schissel & Allen, 2022).
The study included an analysis of the impact of sampling frequency on the quantification of emissions from oil and gas production sites. Relatively stable emissions may require fewer observations to characterize than large, intermittent, or highly variable releases. It takes more planning to accurately capture the larger and more sporadic releases. The temporal dimension of emission events is a key element of designing effective measurement campaigns, along with a detailed understanding of operational conditions during the measurement campaign.
3. Activity data
Once scope and temporal comparability are established, activity data are one of the first elements to check. Source-level inventories depend on complete and accurate activity data, but equipment counts, event records, and operating information may sometimes be incomplete or inferred. In some cases, intermittent events like compressor blowdowns are not logged at all so the inventory assumes an event frequency instead of recording one. Emissions cannot be estimated accurately for equipment that’s never been counted.
A report for the U.S. Department of Energy on gathering compressor stations built a new national station count and arrived at a figure nearly 900 facilities above the Environmental Protection Agency’s official inventory (Zimmerle et al., 2019). The new estimate rested on a stronger foundation, using annual operator equipment reports rather than a 2015 analysis of air permits. The National Academies’ review of U.S. methane characterization identified that several source categories, such as pneumatic controllers, fugitives and engine exhaust, were still resting on activity and emission factor assumptions from a study published back in 1996 (NASEM, 2018).
4. Operating modes
Source-level inventories may represent equipment using normal or expected operating conditions, unless reporting criteria explicitly require abnormal states and failure modes, or an operator characterizes them voluntarily. Site-level measurements can provide evidence that equipment was operating differently from the conditions represented in the inventory, such as a tank with a failed control, an unlit flare, or a stuck dump valve.
Zavala-Araiza et al. (2017) demonstrated that super-emitting sites in the Barnett Shale region in Texas could not be explained by normal operations even at their high end of emission production. Abnormal process conditions, especially super-emitting events, explained most of the gap between site-measurements and component-based inventories. The study points to an aerial survey that found that emitting tank vents and hatches were the predominant high-rate sources of observed emissions, likely from operational mechanisms not characterized within source-level inventories.
Researchers at Colorado State’s METEC facility built a computer model that simulates the gas physically moving through a site’s equipment, including where it escapes when a part fails. Rather than applying a fixed average, or emission factor, to each piece of equipment, the model calculates emissions based on the conditions at the site. The model illustrated an important limitation of factor tuning when equipment failures contribute materially to emissions; an emission factor can be tuned to reproduce average emissions across a population that includes both normally operating and failed equipment, but doing so collapses fundamentally different operating states into a single average. In the model, a failure like a stuck dump valve sharply raised a site’s emission rate (Mollel et al., 2025).
Representing these failures through a higher population-average factor may improve the aggregate emissions estimate, but it obscures the distinction between normal and failed equipment. That makes measurement comparisons harder and sacrifices information about the frequency and duration of failure states, which is needed to develop targeted mitigation strategies. It has historically not been a requirement in various methane reporting criteria to report the duration and frequency of failure modes, particularly intermittent failures that may not materially affect facility operations over an extended period of time. Improving the characterization of these modes to match operational realities requires operational and other contextual informational, often along with site-level measurements at different observational scales for validation, depending on a facility’s remote real-time monitoring capabilities.
5. Measurement
Site-level measurement-derived estimates also carry uncertainty. Part of the issue can be attributed to the limits of the sensors used to detect methane, meaning that some plumes remain unidentified. And when a plume is detected, quantification is a notoriously challenging science due to atmospheric variability. Even the most sophisticated technologies can still miss the mark by many times the true emission rate.
Wind complicates site-level measurements. Some technologies measure methane concentrations in the plume and incorporate these estimates into atmospheric dispersion models to estimate emission rates. In a 2026 controlled-release test, where known amounts of methane were emitted so the results could be checked against the release rate, several teams underestimated emissions because their wind sensors were either malfunctioning or sitting in the pocket of still air behind nearby equipment (McManemin et al., 2026).
Sensor limitations and variable wind speeds and directions can push a single reading too high or too low. Adjusting an inventory to match a measured number that is carrying these errors results in revisions that are tuned to errant information. Controlled-release studies can characterize detection probability, bias, precision, and quantification uncertainty under defined test conditions. That information should be used to interpret the measurement-derived estimate and the weight placed on it during reconciliation.
6. Emission rates
Emissions rates, often presented as averages in the form of emission factors, come last in the check-list because it’s tempting to reach for them first. Adjusting a factor doesn’t involve additional field work. If a meaningful residual discrepancy remains after the first five checks, the source-level emission rate or factor should be examined for whether it is accurate and representative of the equipment, process, and operating condition being characterized. The objective is to update the source characterization when the evidence supports doing so, rather than tuning a factor simply to force agreement with the site-level estimate.
Discussion
Different people within the industry, or even the same company, need different things from the reconciliation process. For an executive, the year-over-year corporate number may be the product, a trustworthy total that shows whether capital deployed on mitigation was money well spent. For an operations superintendent, a facilities engineering manager, or a capital project planner the value is often in the details that emerge, like the root cause, frequency, or duration of large sources. The same exercise serves both use cases when discrepancies are methodically addressed.
It’s important, however, not to conflate diagnosis with mitigation. Identifying where methane was emitted is not the same as understanding why it was emitted. Most large methane releases are process emissions, whether routine or abnormal, rather than leaks, which is why they survive whack-a-mole fixes and show up again in the recurrence data survey after survey. Methane emissions investigation should include causal analysis and a documented reason for the release, instead of just an attributed source or description of the condition of the equipment. This outcome yields far greater actionability when discrepancies can be linked with process knowledge and understood in the context of an operator’s circumstances.
Defensible methane accounting and reporting increasingly has tangible benefits. One of the pathways to market access under the EU’s methane import regulation requires measurement reconciliation as part of the inventory preparation with independent verification. Reconciliation, in other words, is becoming a condition for doing business, but it doesn’t have to be a black box. Most discrepancies can be traced to a handful of common causes and resolved with a step-by-step checklist.
Review: A better way
Reconciliation works best as a structured diagnostic process that first establishes whether the estimates are reasonably comparable, then investigates the causes of any meaningful residual discrepancy.
- Verify scope. Confirm the inventory and the measurement cover the same sites and sources.
- Align timing. Use operational information like operator logs and trend analysis of process data, event frequency and durations of known planned emissions, and repeated sampling, if necessary, to establish what the observation period represents relative to the reference inventory.
- Verify activity data and completeness. Develop a recurring process to internally verify the representativeness of activity data for significant emission sources. Verify that sources are not missing from the inventory.
- Establish operating modes. Determine the operating mode of each major process during measurement. Were tanks filling? Was the VRU running? Which compressors were operating or pressurized? Was the flare lit and what sources were routed to the flare?
- Characterize measurement and inference uncertainty. Consider detection limits, meteorology, quantification uncertainty, and technology-specific performance before treating the measured value as truth.
- Investigate source-level emission rates. If a residual gap survives the first five checks, examine whether the source-level rates or factors accurately represent the equipment, process, and operating conditions. Update them where the evidence supports doing so.
The goal is not to eliminate the numerical difference between two estimates. The goal is to understand whether the inventory is consistent with the available empirical evidence and to improve the underlying emissions characterization where warranted.
References:
Alvarez, R. A., et al. (2018). Assessment of methane emissions from the U.S. oil and gas supply chain. Science, 361(6398), 186–188. https://doi.org/10.1126/science.aar7204
McManemin, A., Juéry, C., Blandin, V., France, J. L., Burdeau, P., & Brandt, A. R. (2026). Controlled release testing of commercially available methane emission measurement technologies at the TADI facility. Atmospheric Measurement Techniques, 19, 923–934. https://doi.org/10.5194/amt-19-923-2026
Mollel, W., Zimmerle, D., Santos, A., & Hodshire, A. (2025). Using prototypical oil and gas sites to model methane emissions in Colorado’s Denver-Julesburg Basin using a mechanistic emission estimation tool. ACS ES&T Air, 2(5), 723–735. https://doi.org/10.1021/acsestair.4c00168
National Academies of Sciences, Engineering, and Medicine. (2018). Improving characterization of anthropogenic methane emissions in the United States. The National Academies Press. https://doi.org/10.17226/24987
Schissel, C., & Allen, D. T. (2022). Impact of the high-emission event duration and sampling frequency on the uncertainty in emission estimates. Environmental Science & Technology Letters, 9 (12), 1063–1067. https://doi.org/10.1021/acs.estlett.2c00731
Zavala-Araiza, D., et al. (2017). Super-emitters in natural gas infrastructure are caused by abnormal process conditions. Nature Communications, 8, 14012. https://doi.org/10.1038/ncomms14012
Zimmerle, D., et al. (2019). Characterization of methane emissions from gathering compressor stations [Final report, October 2019 revision]. Colorado State University. https://mountainscholar.org/items/ebf432f7-526a-4102-803f-0bf4a823ce8a
