Yuanlei Chen , Evan David Sherwin , Erin B. Wetherley, Petr V Yakovlev, Elena S.F. Berman, Brian B Jones, Benjamin Hmiel, David R Lyon, Riley Duren, Daniel H Cusworth, Adam R Brandt 

Reducing methane emissions from oil and gas operations is key to minimizing the climate impact of fossil fuels. Two comprehensive aerial studies in 2019 in the Permian Basin revealed excess emissions compared to official estimates. Although both studies suggested high emissions, the estimates from the two aerial surveys seemed to differ greatly: one study measured 153 (+12/-10, 95% CI) metric tons of methane per hour (t/h), or 7.5% (+0.6%/-0.5%) of gross gas production from aerially detectable point sources in the New Mexico Permian Basin, while the other estimated 246±96 t/h, or 2.7±0.9% of the gross gas production in the larger Texas and New Mexico portions of the Permian Basin. This paper explores causes of this apparent discrepancy by comparing observations of ultra-emitters (>500 kg/h) detected by each survey across a large, spatially overlapping survey region. We account for differences in sensor performance, study scope and design, and data processing practices of the two aerial studies. By aligning approaches, we reconcile the mean ultra-emitter emissions estimates in the applicable overlapping survey area with relative differences as low as 13%, down from 176% for the two full estimates before alignments. T-tests show a p-value increase from 1.2e-5 to 0.182, indicating that the differences between the two aerial-based estimates are not statistically significant after reconciliation. The apparent discrepancy between the studies as published is due to sub-basin level heterogeneous emissions, differing sensor minimum detection limits, and missed ultra-emitters over 1 t/h due to infrequent surveys. Temporal variability in emissions raises an estimation challenge, but this can be mitigated with repeated comprehensive surveys. This work points to methods to improve comparability and repeatability of future estimates, and offers methods to ensure that measured assets are representative of the full area of interest.