Methane Surveys May Need Far More Samples Than Campaigns Currently Collect

Methane has become the most closely watched greenhouse gas in the oil and gas industry, and for good reason. Pound for pound, methane traps far more heat in the atmosphere than carbon dioxide over the near term, which means that finding and fixing leaks from wells, compressors, pipelines, and processing facilities is one of the fastest available levers for slowing climate change. But a new statistical study published in Communications Earth & Environment by William S. Daniels and Dorit M. Hammerling, researchers at the Colorado School of Mines with ties to the Energy Emissions Modeling and Data Lab, delivers an uncomfortable message for anyone designing the measurement campaigns that underpin this effort: the way methane emissions are distributed across facilities makes them fundamentally hard to average, and the number of measurements needed to do it right is often much larger than campaigns have assumed.

The core of the problem lies in the shape of the distribution. When researchers measure emission rates from individual pieces of oil and gas infrastructure, they do not find a tidy bell curve centered on a typical value. Instead, they find what statisticians call a right-skewed, heavy-tailed distribution: most sources emit relatively small amounts, but a small fraction of sources, the so-called super-emitters, release enormous quantities of gas. The tail of this distribution is so heavy that a handful of facilities can dominate the total emissions of an entire basin. Daniels and Hammerling set out to probe exactly what this statistical structure implies for sampling, using six United States oil and gas basins as their test cases.

The technical issue is one that statisticians know well but that field campaigns have often underappreciated. When a distribution is heavy-tailed, the sample mean, the average computed from a finite set of measurements, is an unstable estimator of the true population mean. Each individual observation carries enormous potential influence. If a sampling campaign happens to catch one or two very large emitters, the estimated average emissions for the region will be pulled sharply upward. If, by chance, the campaign misses those facilities, the estimate will fall far too low. In a well-behaved distribution, adding more observations steadily narrows the uncertainty around the average. In a heavy-tailed one, the rare extreme values control everything, and convergence can be painfully slow.

Daniels and Hammerling translated this abstract statistical concern into concrete, actionable numbers. For each of the six basins they examined, they computed a minimum sample size, the number of measurements a campaign would need to collect in order to bound the error that sampling variability alone introduces into the estimated average emission rate. In other words, they asked how many facilities must be measured before the random luck of which facilities happen to be included stops dominating the result. The answer, they found, is that very large sample sizes can be necessary. The largest emissions in each basin drive the behavior of the sample, and by extension they drive the sample size requirements, meaning that campaigns hoping to characterize a basin’s average emissions cannot simply assume that a few dozen or a few hundred site visits will suffice.

The direction of the bias matters as much as its magnitude. The study shows that samples will underestimate average emissions if super-emitters are observed below their true frequency in the population, and overestimate them if super-emitters appear above their true frequency. This is a subtle but consequential point for policy. A campaign that happens to miss the biggest leakers will report that a basin is relatively clean, potentially delaying regulatory attention and allowing large volumes of methane to continue escaping unchecked. A campaign that happens to catch an unusual cluster of super-emitters will paint an alarmingly pessimistic picture that may not represent the basin as a whole. Either way, the number itself, however precise it looks on a report cover, may be wrong, and nobody can tell which direction the error runs without understanding the underlying distribution.

Perhaps the most consequential finding of the study is its rejection of the one-size-fits-all approach. Because the characteristics of super-emitters differ between basins, the authors find that a uniform sampling strategy applied across all regions is suboptimal. A basin whose emissions are dominated by a small number of extremely large sources demands a different statistical treatment than a basin where emissions are spread more evenly across many moderate leakers. The minimum sample size that guarantees acceptable error in one region may be badly inadequate in another. The practical implication is that future methane measurement campaigns should be designed basin by basin, with the sampling plan tailored to the statistical fingerprint of the emissions distribution in that specific region rather than copied from a template developed elsewhere.

This work arrives at a moment when the stakes for methane measurement have never been higher. Governments are building methane intensity standards and import rules around measured emission rates, satellite missions are scanning producing regions from orbit, and companies are competing to demonstrate low-emission supply chains to climate-conscious buyers. All of these mechanisms depend on numbers produced by measurement campaigns, and all of them inherit the statistical vulnerabilities that Daniels and Hammerling document. An emissions factor derived from an underpowered sample is not a neutral data point; it propagates into inventories, certification schemes, and regulatory thresholds, where its hidden bias can distort decisions worth billions of dollars and millions of tonnes of avoided emissions.

The study also carries a lesson about the relationship between measurement and modeling. Heavy-tailed distributions do not just complicate fieldwork; they complicate the interpretation of every dataset built from limited observations, including the aircraft surveys, ground-based mobile measurements, and continuous monitoring records that feed national inventories. The authors’ framework, which provides basin-specific minimum sample sizes tied explicitly to the error budget of the average estimate, offers campaign designers a principled way to decide in advance how much data they need, rather than discovering after the fact that their estimates are too noisy to be useful. That kind of prospective design, grounded in the statistics of the emissions themselves, represents a meaningful shift in how the field can approach the problem.

Funding for the research came in part through the Energy Emissions Modeling and Data Lab and through the Colorado Ongoing Basin Emissions project, supported by the Colorado Department of Public Health and Environment, reflecting the growing interest of state agencies in reliable basin-level emissions data. The work was published open access in Communications Earth & Environment, with Daniels now based at Johns Hopkins University. As satellite constellations multiply and methane regulations tighten worldwide, the study’s central message is likely to echo through campaign planning for years to come: the rare, enormous leaks that dominate a basin’s climate footprint also dominate its statistics, and only sampling strategies designed with that heavy tail in mind, tailored to each basin’s own character, can deliver the trustworthy average emission rates on which climate accountability depends.

Subject of Research: Statistical sampling requirements for measuring methane emissions from oil and gas basins

Article Title: Future methane measurement campaigns require basin-specific sampling strategies

Article References: Daniels, W. S., & Hammerling, D. M. (2026). Future methane measurement campaigns require basin-specific sampling strategies. Communications Earth & Environment. https://doi.org/10.1038/s43247-026-04089-4

Image Credits: AI Generated

DOI: 10.1038/s43247-026-04089-4

Keywords: methane, oil and gas, super-emitters, heavy-tailed distributions, sampling strategies, emissions measurement, basin-specific, greenhouse gas monitoring, sample size, Communications Earth & Environment, climate mitigation, emissions statistics