Comprehensive aerial survey quantifies high methane emissions from the New Mexico

2 Limiting emissions of climate-warming methane from oil and gas (O&G) is a major 3 opportunity for short-term climate benefits. We deploy a basin-wide airborne sur4 vey of the New Mexico Permian Basin, spanning 35,923 km2, 26,292 active wells, and 5 over 15,000 km of natural gas pipelines using an independently-validated hyperspec6 tral methane point source detection and quantification system. The airborne survey 7 repeatedly visited over 90% of the active wells in the survey region throughout October 8 2018 to January 2020, totaling 117,658 well visits. We estimate total O&G methane 9 emissions in this area at 194 (+72/-68, 95% CI) metric tonnes per hour (t/h), or 10 9.4% (+3.5%/-3.3%) of gross gas production. 50% of observed emissions come from 11

Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy.  Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy. The middle bar breaks down extrapolated emissions into undetected emissions within the partial detection range (PDR), emissions from assets not measured in the survey area, and emissions that are below minimum detection limit (MDL). The right bar shows that the estimate of total methane emissions in the survey area from upstream and midstream O&G operations is 194 (+72/-68) t/h, 9.4% (+3.5%/-3.3%) of gross gas production. (b) The distribution of asset-typespecific persistence-averaged emission source sizes, which follow heavy-tailed distributions.
(c) Cumulative emission fraction as a function of persistence-averaged emission source sizes.
of all assets (including wells, gathering lines, storage tanks, and compressor stations) found 120 on a congruent gravel or concrete area containing at least one well. Midstream assets were 121 also a significant source, with 29±20 t/h emitted from pipelines (including underground gas 122 gathering pipelines) and 26±16 t/h emitted from compressor stations without a well on site.

123
The remainder was emitted from stand-alone storage tank sites (9±6 t/h), gas processing plants Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy.
persistence of the heavy tail for distributions of large emissions demonstrates the significant 132 potential for mitigating methane by detecting and fixing these high-consequence sources.

133
Sensitivity tests show robust support for a mean natural gas fractional loss rate of at least 134 8.1% of gas produced. As listed in Table 1, switching from a sublinear fit to a linear fit for the 135 calibration step, described in Section S5, brings the loss rate estimate up to 10.2% (+4.1%/-136 3.6%). A linear fit forced through the origin leads to an estimate of 11.0% (+5.0%/-4.6%). In 137 the calibration fitting process, leaving out large controlled releases improves the statistical va-138 lidity of the fit due to the underlying asymmetric error distribution at high emission rates, and 139 also increases the total emission estimate, as described in the SI, Section S1.5. Using an alter-  concentration of prior emissions at O&G production sites, as opposed to midstream assets). We 167 explore this comparison more in the SI, Section S8.

168
It is important to explore further a key strength of our method compared to prior studies:

169
Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy. very large study sample size. We explore this by simulating the impact of small sample sizes 170 on total emissions estimates (Fig. 3b).

171
Suppose that we only visited 100 sites, a typical sample size for ground-based campaigns. for 9% (+4%/-3%) of our study total, suggesting that higher sensitivity would lead to only a 198 modest increase in total estimated emissions relative to simulated levels.

199
In conclusion, we conducted a site-level, basin-wide field survey of methane emissions in 200 one of the most active oil-producing regions in the world. We estimate emissions to be 9.4%

201
Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy.
(+3.5%/-3.3%) of the gross gas production for the region, much higher than found in previous 202 studies with overlapping, although not identical, domains. The increase is partly because our 203 method allows us to inspect the entire O&G-producing population using an independently-

221
To compute basin-wide total emissions, we combine a statistical analysis of direct measure-222 ments with a literature-based estimate for emissions below the instrument's detection threshold. 223 We deploy an analysis workflow illustrated in Figure 1e. The SI, Section S3.1 details each step 224 in the workflow. Inputs into this workflow include wind-independent emission rate in kgh/mps 225 Non-peer reviewed preprint submitted to EarthArXiv. Submitted to Nature Energy.
for each plume and wind speed at imaging time and plume coordinates by HRRR's estimate.

226
For each plume, we multiply these two input terms to derive emission rates in kg/h.

227
In this study, we refer to the the single-blind test of the instrument by Sherwin,Chen et al. 228 to determine the instrument's detection limit and quantification accuracy and precision (see the 229 SI, Section S1). Data from the single-blind test shows the instrument's apparent overestimation 230 tendency for larger releases, possibly due to an underlying nonlinearity or a boundary bias for 231 calibration (detailed in the SI, Section S1.6).