Methane emissions decline from reduced oil, natural gas, and refinery production during COVID-19

In the summer of 2020, the AVIRIS-NG airborne imaging spectrometer surveyed California’s Southern San Joaquin Valley and the South Bay (Los Angeles County) to identify anthropogenic methane (CH4) point source plumes, estimate emission rates, and attribute sources to both facilities and emission sectors. These flights were designed to revisit regions previously surveyed by the 2016–2017 California Methane Survey and to assess the socioeconomic responses of COVID-19 on emissions across multiple sectors. For regions flown by both the California Methane Survey and the California COVID campaigns, total CH4 point source emissions from the energy and oil & natural gas sectors were 34.8% lower during the summer 2020 flights, however, emission trends varied across sector. For the energy sector, there was a 28.2% decrease driven by reductions in refinery emissions consistent with a drop in production, which was offset in part with increases from powerplants. For the oil & natural gas sector, CH4 emissions declined 34.2% and significant variability was observed at the oilfield scale. Emissions declined for all but the Buena Vista and Cymric fields with an observed positive relationship between production and emissions. In addition to characterizing the short-term impact of COVID-19 on CH4 emissions, this study demonstrates the broader potential of remote sensing with sufficient sensitivity, spatial resolution, and spatio-temporal completeness to quantify changes in CH4 emissions at the scale of key sectors and facilities.


Introduction
The COVID-19 pandemic and associated lockdowns have resulted in an unexpected opportunity to quantify rapid changes in emissions of criteria air pollutants and greenhouse gases and potentially shed new light on some of the underlying processes. For example, Shi et al [1] documented reductions in particulate matter less than 2.5 microns in aerodynamic diameter (PM 2.5 ), nitrogen dioxide (NO 2 ), carbon monoxide (CO), and sulfur dioxide (SO 2 ) concentrations in northern China, Sharma et al [2] saw similar declines of PM 2.5 in India, and in the United Kingdom Jephcote et al [3] observed reductions in PM 2.5 and nitrogen dioxide (NO 2 ) but increases in ozone (O 3 ) concentrations due to reductions in local nitrous oxide (NO) emissions. Using activity data, Liu et al [4] estimated an 8.8% decrease in global CO 2 emissions in the first half of 2020 while Le Quéré et al [5] showed a 17.0% decline by early April 2020 compared to mean 2019 levels, but emphasized that these reductions are likely temporary. Regional studies by Yadav et al [6] reported decreases in CO 2  Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. observed declines in emissions, the global atmospheric growth rates of CO 2 and methane (CH 4 ) as measured by OCO2 and TCOON were not slowed in 2020 [7]. In fact, the latest observations by NOAA indicate that the global CH 4 atmospheric growth rate actually accelerated following 2020, underscoring the critical need to reduce emissions.
In the United States' Permian Basin, Lyon et al [8] used aircraft mass balance data and in situ tower network observations combined with tracer-transport modeling to show that CH 4 emissions were strongly correlated to the average daily oil price. Just prior to COVID-19 lockdowns, oil prices dropped precipitously in January 2020 [9]. For a 10,000 km 2 study area in the Permian basin, Lyon et al [8] showed a threefold reduction in CH 4 emissions from January through April 2020 associated with falling oil prices and reduced production, which reached a low on 20 April. As oil prices recovered from May through September 2020, CH 4 emissions were observed to increase and approach January 2020 levels. The Lyon et al [8] study demonstrates that at an oilfield scale, disruptive events like COVID-19 have the potential to impact oil prices, which can ultimately result in reductions to both production and associated CH 4 emissions.
The COVID-19 pandemic impacted many socio-economic sectors, including those that emit CH 4 to the atmosphere. Though previous studies generally show COVID-19 to result in temporary emission reductions for CH 4 and other trace gases, it is not clear if this persists across the full oil and gas supply chain, all basins, and all sectors. For example, reduced production was shown to result in lower emissions in the Permian [8]. However, in other regions, reductions to staffing could result in a larger proportion of unattended leaks, malfunctions, or other CH 4 emitting processes. Further, operators likely dealt with the drop in oil prices and staffing shortgages differently, some choosing to shut-in wells while others vented unsold natural gas to the atmosphere. To further test how the COVID-19 pandemic affected emissions across multiple sectors and geographic areas, the Next Generation Airborne Visible/Infrared Imaging Spectrometer (AVIRIS-NG) was flown in California to identify anthropogenic CH 4 point source plumes, estimate emission rates, and attribute sources to both facilities and emission sectors. During a period of COVID restrictions in the summer of 2020, AVIRIS-NG revisited energy sector and oil and gas regions previously surveyed by the 2016-2017 California Methane Survey [10]. By comparing results for the same as flown regions, we are able to assess changes in CH 4 emissions at the oil field scale as well as for specific emission sectors like refinery and power plant emissions.

Methane retrievals and emission estimates
The AVIRIS-NG and Global Airborne Observatory (GAO) airborne imaging spectrometers measure reflected solar radiation between 380 to 2,500 nm at approximately 5 nm spectral resolution and sampling [11]. These push broom instruments have a 34°field of view which permits covering large regions quickly with an image spatial resolution that scales with flight altitude. CH 4 retrievals utilize absorption spectroscopy in the shortwave infrared Thorpe et al [12][13][14][15][16][17]. For this study CH 4 emissions are calculated using the methodology described in Duren et al [10], including estimating CH 4 emissions for each observed plume, calculating an average emission rate for all plumes detected for a given source, and adjusting the average emission rate based on source persistence, which is derived from the number of times a source is detected and the number of times it is observed. This is important because many industrial CH 4 sources are highly intermittent.
This class of imaging spectrometer has a detection threshold for condensed CH 4 point sources as low as 2 kg CH 4 h −1 across multiple flight altitudes and wind conditions [13] and the CH 4 retrieval algorithm and emission quantification approach has been validated using controlled release experiments [18]. These instruments have been used in a number of previous CH 4 studies spanning different emission sectors and locations. In the United States, studies have surveyed the oil and gas sector for the Four Corners Region [19] and the Permian Basin [20], additional basins including the Uintah, Denver-Julesberg, and Marcellus [21], manure management and waste management sectors in California [10,22], and CH 4 hotspots associated with thermokarst lakes in Alaska [23,24]. Additional international flight campaigns have characterized CH 4 emissions from the oil and gas sectors in Canada [25] and the oil, gas, and waste management sectors in India [17].

Survey design
For the California COVID flights, AVIRIS-NG targeted the Southern San Joaquin Valley (Kern County) and the South Bay (Los Angeles County), which were previously surveyed as part of the 2016-2017 California Methane Survey (figure 1) [10]. Figure 2 shows the regions of overlap between the California Methane and California COVID flights for the Southern San Joaquin Valley (1,128 km 2 ) as well as the locations of the observed CH 4 plumes from either campaign. In 2020 California produced 143,114 Mbbl (thousand barrels of oil) [26] and 170,579 MCF (million cubic feet natural gas) [27], with the majority of production located in Kern County. While portions of 23 oil and gas fields were covered with AVIRIS-NG, the flights primarily targeted the largest fields including Lost Hills, Belridge, Cymric, McKittrick, Elk Hills, Midway-Sunset, Buena Vista, Poso Creek, Kern Front, and Kern River. Figure 3 shows the South Bay coverage (387.8 km 2 ) with AVIRIS-NG focused on refineries in El Segundo, Torrance, Carson, and Wilmington [28]. In 2020 California had a monthly production of around 9.5E6 barrels, of which over half came from Southern California [29]. While these regions included a number of power plants,  flight coverage was extended to Seal Beach and Huntington Beach to include additional facilities [30]. The Inglewood oilfield was also mapped during both flight campaigns.
For both the California Methane and California COVID flights, AVIRIS-NG was flown at 3 km above ground, providing image swaths of 1.8 km with 3 m image pixels and flights spanned the summer and fall seasons. For the overlapping region shown in figures 2 and 3, repeat mapping occurred on average 23 times (11.5 times for both the California Methane Survey and the California COVID flights).

Overview of observed CH 4 trends
As shown in figure 2, there were reductions in the number of CH 4 sources for most of the regions flown in the Southern San Joaquin Valley during the California COVID campaign. Particularly dramatic reductions were observed for the Poso Creek and Kern Front fields (north of Bakersfield), while the number of sources for the Buena Vista and Cymric fields increased during the California COVID campaign. Reductions in the number of sources are also visible for the South Bay (figure 3). In total, 222 CH 4 sources were observed during the California Methane Survey compared to only 120 sources for the COVID campaign. Figure 4 shows the cumulative distribution of CH 4 emissions from those sources observed during the California Methane Survey and California COVID campaigns. The histogram indicates a greater number of sources with emissions less than 75 kg hr −1 for the California Methane Survey. Total emissions from the energy (refineries and powerplants) and oil & natural gas sectors were 33.4% lower during the 2020 COVID flights (6,815 ± 2,173 kg hr −1 ) compared to the 2016-2017 California Methane Survey (10,232 ± 3,141 kg hr −1 ) (figure 5). Total emissions for the energy sector were 1,371 ± 526 kg hr −1 for the California Methane Survey, which dropped to 984 ± 291 kg hr −1 for the California COVID campaign (28.2% decrease). This was driven by reductions in refinery emissions, which was offset in part with increases from powerplants (figure 5, right). A 34.2% decrease is shown in figure 5 for the oil & natural gas sector (8,861 ± 2,615 kg hr −1 for the California Methane Survey, 5,830 ± 1,882 kg hr −1 for the California COVID). Figure 6 shows that significant variability was observed at the oilfield scale, with emissions declining for all but the Buena Vista and Cymric fields. Figure 7 shows examples of CH 4 plumes observed with AVIRIS-NG for the energy sector (refinery and powerplant) as well as well as from the oil & natural Gas sector (well and gathering line).

Refinery emissions
Petroleum refinery emissions are typically estimated using emission factors based on the type and use of equipment for a given facility. Refineries contain a number of potential CH 4 sources and have been documented to be significant emitters. For example, Chambers et al [31] used ground-based differential absorption light detection and ranging (DIAL) to estimate a CH 4 emission rate of 300 kg hr −1 for a Canadian refinery. Lavoie et al [32] used an aircraft Picarro cavity ring-down spectroscopy analyzer to quantify CH 4 emissions from three refineries that ranged between 360 ± 200 and 830 ± 240 kg hr −1 , reflecting values between 11 and 90 times higher than GHGRP emission estimates. Mehrotra et al [33] used a similar approach to measure CH 4 emissions ranging between 23 and 700 kg hr −1 for 3 California refineries sampled a total of 15 times, at rates significantly higher than values reported to the USEPA and CARB. More recently, Guha et al [34] surveyed 5 California refineries a total of 41 times using an airborne mass-balance technique and observing average emissions ranging between 300 and 600 kg hr −1 .  The CH 4 plumes observed in this study typically fall within the range of previous reported values. Of the 48 plumes observed during the California Methane Survey, the average emission was 246 ± 93 kg hr −1 and was 289 ± 113 kg hr −1 during California COVID for 11 plumes. For the California Methane Survey, the 48 plumes were assigned to 33 sources and average emissions were scaled by the frequency of observation as described in Duren et al [10] resulting in a total CH 4 emission rate of 1,197 ± 457 kg hr −1 ( figure 5). For the 2020 COVID campaign, 11 plumes were assigned to 5 sources with a total CH 4 emission rate of 323 ± 138 kg hr −1 ( figure 5). The dramatic decrease in the number of refinery sources in the South Bay can be clearly seen in figure 3.   with an average monthly production over this time period of 2.55E7 barrels (bbl). The significant drop in production in the first quarter of 2020 was associated with a crash in oil prices. The California COVID flights occurred 7/2020-9/2020 (blue) with an average monthly production that is 17.7% lower (2.10E7 bbl). While declining production is likely to lead to a reduction to CH 4 emissions, the observed 73.0% decline in refinery emissions during the COVID flights is significantly larger and suggests that a modest decrease in production has a large impact on CH 4 emissions (figure 5).

Power plant emissions
For the overlapping region between the two flight campaigns (figures 2 and 3), there were four observed CH 4 plumes from powerplants for the California Methane Survey (87 ± 31 kg hr −1 average emission rate). This is broadly consistent with previous studies, including Lavoie et al [32] that estimated CH 4 emissions from power plants between 75 ± 30 kg hr −1 and 240 ± 70 kg hr −1 and Hajny et al [35] that reported typical values between 17 ± 10 kg hr 1 and 117 ± 31 kg hr 1 .
Nine power plant CH 4 plumes for the California COVID flights had a 484 ± 113 kg hr −1 average emission rate and the plumes were typically significantly larger (see figure 7) with larger emission rates than plumes observed during the California Methane Survey. However, the power plant plumes were highly intermittent with a 14% average frequency for the California Methane Survey and 22% average frequency for the California COVID flights. After adjusting for frequency of observation, the total CH 4 emissions from the four power plant sources for the California Methane Survey was 174 ± 69 kg hr −1 and 661 ± 152 kg hr −1 from the six sources for the California Methane Survey ( figure 5).
Hajny et al [35] reported one anomalously high value (around 510 kg hr −1 ) for a power plant that was within 4 h of starting up and suggested that increased emissions might be associated with startup conditions. Continuous emissions monitoring systems (CEMS) hourly reported activity data was compared with all CH 4 plumes associated with the California Methane and California COVID power plant sources [36]. These results are shown in table 1 and indicate that for the California Methane survey 100% of the observed plumes occurred during times when any unit was active and 50% occurred within 2.5 h of a unit startup. For the California COVID survey 89% of the observed plumes occurred during times when any unit was active and 78% occurred within 2.5 h of a unit startup. This indicates the possibility that startup conditions could explain some of the observed emissions, with an increase in emissions for the California COVID associated with an increased prevalence of start ups. with an average monthly production over this time period of 2.55E7 bbl (gas and diesel production). The significant drop in production in the first quarter of 2020 was associated with a crash in oil prices. The California COVID flights occurred 7/2020-9/2020 (blue) with an average monthly production that is 17.7% lower (2.10E7 bbl).

Emissions by oil & gas field
It is clear from figure 6 that CH 4 emissions are highly variable at the oil & gas field scale. For those regions flown during both the California Methane and California COVID campaigns, most fields showed declining emissions with a notable exception for Buena Vista and Cymric. For the portions of each field covered by AVIRIS-NG, monthly oil and gas production are shown in figure 9 [37]. The California COVID flights occurred 7/2020-9/ 2020 (blue horizontal line) with an average monthly oil production that is 24.2% lower than the average monthly production for California Methane flights (9/2016-10/2016 and 6/2017-10/2017, orange horizontal lines). For the same periods, natural gas production declined by 20.0%. The observed 34.2% decline in CH 4 emissions for the oil & natural gas sector during the COVID flights (figure 5) is consistent with declining oil and gas production in those regions surveyed by AVIRIS-NG. This relationship is consistent but not linear, which suggests that a shock to an oil & gas supply chain (e.g., COVID, oil market collapse) may not have easily predictable results on emissions. Emission changes at the basin scale likely are the result to heterogeneous responses by individual operators and companies. Depending on circumstances, operators may respond differently to the same economic shock (e.g., shutting in wells, reduced maintenance, increased or decreased flaring). These responses likely have different outcomes on CH4 releases, which in aggregate may not linearly correlate with changes in production.
For the California COVID campaign, observed emissions for the 10 oil & gas fields with the highest emissions are plotted against estimated total BTUs for oil & gas production using a monthly average from the 7/ 2020-9/2020 reported data (figure 9). The linear regression with R 2 of 0.65 indicates a relationship between total oil & gas production (BTU) and observed emissions. While this seems intuitive, this is the first study using imaging spectrometer results to demonstrate this relationship, and corroborates the trend that Lyon et al (2020) observed in the Permian Basin. It is interesting to note that larger observed emissions were seen both for some fields dominated by gas production like Elk Hills and Buena Vista (66% and 72% of the total BTUs from gas) and for fields producing mostly oil like Midway Sunset and Cymric (96% and 97% from oil). Figure 9. Monthly oil and gas production for the portion of each field covered by AVIRIS-NG with totals from all fields shown in black [37]. The California COVID flights occurred 7/2020-9/2020 (blue horizontal line) with an average monthly oil production that is 24.2% lower than the average monthly production during the California Methane flights (9/2016-10/2016 and 6/2017-10/2017, orange horizontal lines). For the same periods, gas production declined by 20.0%. In the lower panel, observed emissions for the 10 oil & gas fields with the highest emissions during the COVID campaign are plotted against estimated BTU for oil & gas production (using monthly average from 7/2020-9/2020 reported data). The linear regression and R 2 indicates a robust relationship between total production (BTU) and observed CH 4 emissions. The average monthly BTU was calculated from oil and natural gas, while the legend indicates the percentage of BTU associated with natural gas.

Discussion
For the Lyon et al [8] Permian Basin study, oil & natural gas production was increasing prior to the COVID−19 pandemic. As a result, the observed relationship between declining oil prices, reduced production, and reduced CH 4 emissions was clearly visible. For the Southern San Joaquin Valley and South Bay, oil & natural gas (figure 9) and refinery production (figure 8) was declining prior to COVID-19. During COVID-19, declining oil prices would similarly result in declines in production at oil & natural gas fields and refineries, resulting in reductions to observed CH 4 emissions. However, declines associated solely with COVID-19 impacts are obscured by the prior trend of declining production for both emission sectors.
The total 34.8% decline in CH 4 emissions observed with AVIRIS-NG during the 2020 flights reflected a dramatic 73.0% reduction to refinery emissions and a 34.2% decrease for the oil & natural gas sector. Fields associated with the largest observed CH 4 emissions included those dominated by gas production (Elk Hills and Buena Vista), but other fields producing mostly oil (Midway Sunset and Cymric). There was significant variability in observed CH 4 emissions at the field scale, with emissions declining for all but the Buena Vista and Cymric fields. It remains unclear why emissions at these two fields increased. Like the other fields, oil and natural gas production at Buena Vista and Cymric was declining and there was no evidence of increased drilling or well stimulation, however, Cymric did report a number of surface expression issues in the 2019-2020 timeframe [38]. The observed increase in power plant emissions appears associated with powerplants being both active and within 2.5 h of a unit startup, however, it is unclear if these emissions resulted from a change in operation associated with the COVID-era or a different underlying cause.
As shown in the cumulative distribution of emissions ( figure 4, left), the COVID emissions are shifted the right relative to the California Methane results, indicating that the average emission rate per source actually increased. This is clearly shown in (figure 4, right) and is consistent with the total number of sources dropping 45.9% while total emissions dropped only 33.4%. During the COVID campaign, there were fewer low emission sources, suggesting that changes in production and activities in response to COVID-19 is an inefficient way to reduce emissions.
While questions remain regarding some of the underlying mechanisms for the observed declines, this study demonstrates the potential of remote sensing to document reductions in CH 4 emissions associated with long term trends like declining oil & natural gas production as well as disruptive economic events like COVID-19 and the associated crash in oil prices. This emphasizes the great potential for using these types of measurements to quantify CH 4 emissions, attribute them to specific emission sectors, and characterize emissions over time to improve our understanding of CH 4 budgets. In addition to characterizing the short-term impact of COVID-19 on methane emissions, this study demonstrates the broader potential of remote sensing with sufficient sensitivity, spatial resolution and spatio-temporal completeness to quantify changes in CH 4 emissions at the scale of key sectors and facilities. Given the dramatic rebound and acceleration in global CH 4 atmospheric growth rates following 2020, there is a more urgent need for emissions mitigation enabled by advanced observational data.

Conclusion
This study characterized the short-term impact of COVID-19 on CH 4 emissions. A dramatic 34.8% reduction in total CH 4 point source emissions from the energy and oil & natural gas sectors for the California COVID campaign (summer 2020) was observed when compared to the California Methane Survey (2016-2017). The 28.2% decrease in emissions for the energy sector was driven by reductions in refinery emissions, consistent with a drop in production, offset in part with increases from powerplants. For the oil & natural gas sector, observed emissions declined 34.2% and significant variability existed at the oilfield scale with a positive relationship between production and emissions. This study demonstrates the broader potential of remote sensing with sufficient sensitivity, spatial resolution, and spatio-temporal completeness to quantify changes in CH 4 emissions at the scale of key sectors and facilities.
Chapman, and Mark Helmlinger. RMD acknowledges additional support from High Tide Foundation. This work was undertaken in part at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004).

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.

Conflict of interest
The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.