A sprawling, aggressive effort to measure the climate footprint of natural gas production has yielded striking results: methane emissions from the Barnett Shale in North Texas are at least 90 percent higher than government estimates.

That conclusion comes from a peer-reviewed study published Monday in Proceedings of the National Academy of Sciences. The paper is the most sweeping study to emerge from the Environmental Defense Fund’s $18-million project to quantify methane leaks from the natural gas industry. It was written by 20 co-authors from 13 institutions, including universities, government labs, EDF and private research firms.

Overall, the two-year study found that methane emissions from the Barnett Shale are nearly twice as much as estimated by the Environmental Protection Agency’s Greenhouse Gas Inventory, and 5.5 times the number from a separate global database.

Co-author Amy Townsend-Small, an environmental studies professor at the University of Cincinnati, said peer-reviewed papers often find larger emissions than EPA estimates.

The EPA’s databases are often based on decades-old methodology, Townsend-Small said, adding that the federal agency knows it has “a long way to go.”

Understanding the scope of methane leaks is crucial, because the answer will determine whether the ongoing shift from coal to natural gas-fired electricity creates a net benefit for climate change. Although gas power plants emit much less carbon dioxide than coal plants do, even small leaks of methane—the main component of natural gas—could undermine that advantage.

Methane is 86 times more potent than CO₂ as a greenhouse gas on 20-year timescales, and 34 times more powerful on 100-year timescales.

The PNAS study synthesized data from 11 previous papers. Those studies were part of EDF’s Barnett Coordinated Campaign, where a dozen research teams measured methane emissions from wells, compressor stations and other equipment.

The Barnett produces 7 percent of the nation’s natural gas. It was also the first oil and gas basin to use high-volume hydraulic fracturing on a large scale, starting in the 1990s.

First author Daniel Zavala-Araiza, an EDF scientist, said the study’s main goal was to reconcile the data obtained through different methods. Typically, “top-down” measurements, where scientists use airplanes to gauge emissions from an entire region, show higher leak rates than data collected from “bottom-up” methods, which involve analyzing individual facilities and adding up the results.

Zavala-Araiza said the scientists found greater agreement between their top-down and bottom-up emission rates than previous studies did.

The key, said Townsend-Small, is that the researchers created a statistical approach that more accurately accounts for “superemitters”—the small fraction of facilities that emit a disproportionate amount of methane.

For instance, just 2 percent of the oil and gas facilities in the Barnett released 50 percent of the methane. Zavala-Araiza said the team reached that conclusion after integrating its datasets and extrapolating the results over the entire shale field.

It required detailed calculations that took months: when the 11 papers behind the Barnett campaign were released in July, the scientists initially estimated methane leaks were 50 percent higher than EPA numbers. After completing the statistical analysis, they upgraded that figure to 90 percent.

The sheer scale of the campaign also enabled researchers to take more measurements than usual. While most top-down studies include one or two flights, the scientists on this team collected data on seven different flights. Townsend-Small said it allowed them to get a more representative sample of emissions, which can vary widely over time and by location.

“The study seems very well done in general, and the authors include several very capable scientists,” said Robert Howarth, a Cornell University professor and methane expert, via email. Howarth, who did not participate in any of the EDF studies, has been critical of EDF’s methane project, in part because some of the research was done in collaboration with industry partners.

EDF chief scientist Steven Hamburg said the vast majority of the Barnett data were collected on public property. The only exception was a series of measurements taken directly at compressor stations, and that data comprised “far less than five percent” of the total measurements, he said. The rest of the time, the researchers used fenceline measurements by driving downwind of the facilities and using mobile labs to track methane plumes. Industry did not provide any direct funding to the Barnett campaign, Hamburg said.

Townsend-Small said the scientists tried to minimize the measurements taken on company property because the first study from EDF’s methane project was criticized for potential “cherry-picking.” That study, published in 2013 and led by University of Texas professor David Allen, found lower-than-expected emissions from production sites. The scientists had received permission from nine oil and gas operators to take bottom-up measurements on company property, and the companies also helped to fund that study.

The Barnett paper references Allen’s study, which concerns Howarth.

Earlier this year, a study in the journal Energy Science & Engineering alleged that Allen and his co-authors had used a flawed methane detector that underestimated emissions.

In light of those accusations, said Howarth, the Barnett paper should not have cited the 2013 study uncritically. “EDF scientists have been aware of the problem with the Allen et al. work for at least the last 18 months,” Howarth said.

EDF and Allen have repeatedly denied that the 2013 results reflect a technical problem.

Howarth said he also worries the Barnett results may underestimate contributions from drilling and fracking, since the measurements were made after the period of peak production in the Barnett. It’s a “critical point [that’s] not adequately addressed in the paper,” he said.

Townsend-Small said the results represent what the natural gas fields of today will look like in 20 to 30 years, so scientists will know what to expect once a field progresses beyond the initial burst of activity.

Frank Flocke, an atmospheric scientist who wasn’t involved in the PNAS study, said the approach presented in the paper “appears pretty sound.”

Flocke, who works for the University Corporation for Atmospheric Research (UCAR) in Boulder, Colo., said he knows many of the study authors and respects their research record.

Because the data came from a single shale basin, Flocke cautioned against making sweeping conclusions. “Other areas might have different emissions rates, specifically given that the process plants [in the Barnett] seem to be the main emitters. Depending on how much processing the raw gas needs, total methane ‘leak rates’ from different natural gas extraction fields…may be very different,” he said in an email. “Much work remains to be done in order to assess the entire” natural gas industry.