The oil and gas industry is Canada’s largest contributor to national methane (CH4) emissions. To quantify the input of active and inactive (suspended and abandoned) oil and gas infrastructure to regional CH4 budgets, we conducted truck-based measurements (transect-based and OTM 33A) with a greenhouse gas analyzer, complimented with optical gas imaging at oil-producing sites of Saskatchewan, including understudied regions. We found that inactive sites regionally accounted for roughly 43% of total measured CH4 emissions in Lloydminster, 9% in Kindersley, and 0% in Swift Current. Thus, CH4 emissions from oil production in southwestern Saskatchewan are underestimated by almost 25% if emissions from inactive sites are ignored. Measured mean CH4 emissions of actively producing oil and gas infrastructure in Lloydminster were at least 50% lower (36 ± 7 m3/day) than found in previous studies potentially due to declines in production schemes, effective implementation of emission reduction approaches, or spatial differences between sampled sites. Unlike previous studies, measured emissions in Lloydminster were lower than reported values (147 ± 10 m3/day). In contrast, measured emissions in Kindersley (64 ± 17 m3/day) and Swift Current (23 ± 16 m3/day) were close to reported emissions despite observed tank vents and unlit flares. Unlit flares emitted at least 3 times more CH4 than other infrastructure types and were the “super-emitters” in this study. Currently, provincial and federal regulations target only active infrastructure, but regulators may consider extending regulations to inactive sites where data suggest significant emission reduction potential.
The oil and gas industry is Canada’s largest contributor to national methane (CH4) emissions (Environment and Climate Change Canada, 2019). In the upstream oil and gas sector, the greenhouse gas CH4 is mostly released to the atmosphere in the form of vented (intentional) or fugitive (unintentional, uncontrolled leaks) emissions. CH4 is rated a toxic and flammable gas, which also has a high potential to warm the Earth (up to 83 times greater than that of the greenhouse gas carbon dioxide over a 20-year period; Forster et al., 2021). Therefore, CH4 emissions from the oil and gas industry should be reduced to mitigate global warming.
Saskatchewan is the second largest oil producing province in Canada. In 2019, the Canadian province of Saskatchewan released Oil and Gas Emissions Management Regulations together with a Methane Action Plan (Ministry of Energy and Resources, Government of Saskatchewan, 2019a). With these initiatives, the Government of Saskatchewan plans to target CH4 emission reductions from the upstream oil and gas sector by 4.5 million tonnes CO2 equivalent annually by 2025 and will enforce a subsequent emissions cap until 2030. Some argue that the Saskatchewan measures fall short of the federal targets and measures introduced earlier (Gorski, 2019), but the federal government did accept the provincial measures as equivalent. Due to the impact of the COVID-19 pandemic on the oil and gas industry, the provincial government introduced temporary relief measures of CH4 emission regulations, including exemptions regarding leak detection and repair surveys (Ministry of Energy and Resources, Government of Saskatchewan, 2020a).
Federal and provincial regulations were crafted with only partial understanding of present-day emission levels and sources. Various oversight studies using trucks and aircraft have pegged emissions at higher levels than inventory estimates (Lan et al., 2015; Yacovitch et al., 2015; Atherton et al., 2017; Johnson et al., 2017; Zavala-Araiza et al., 2018; Baillie et al., 2019; MacKay et al., 2019; O’Connell et al., 2019; Chan et al., 2020; Li et al., 2020; MacKay et al., 2021). By now, most oil and gas regions in western Canada have been measured but oil-producing sites in western Saskatchewan have escaped oversight. Emission rates may vary significantly per region (MacKay et al., 2021), infrastructure type (Lavoie et al., 2022), and over time assuming adaptation to new CH4 emission reduction regulations. Therefore, it is important to frequently conduct measurements covering a wide geographical region with different infrastructure types to obtain an accurate picture of overall CH4 emissions.
In a global assessment of CH4 emissions from the oil and gas sector, suspended and abandoned infrastructure was assumed to be non-emitting (Scarpelli et al., 2020). However, several studies identified that abandoned infrastructure also emits measurable amounts of CH4 to the atmosphere (Kang et al., 2014; Boothroyd et al., 2016; Kang et al., 2016; Townsend-Small et al., 2016; Pekney et al., 2018; Riddick et al., 2019; Williams et al., 2019; Saint-Vincent et al., 2021). The Canadian federal inventory does account for emissions from suspended and abandoned infrastructure, but not in a clear way since emission factors are based on a single study restricted to a selection of abandoned wells, which may be plugged or unplugged in the United States (Townsend-Small et al., 2016; Environment and Climate Change Canada, 2020). However, in Canada jurisdictions differ among provinces, so that the plugging status of a well often remains unclear based on the information reported by operators when comparing nationwide. In addition, numbers of abandoned wells and their associated CH4 emissions generally constitute large uncertainties in CH4 emission inventories, particularly in Canada and the United States (Environment and Climate Change Canada, 2020; Williams et al., 2021). The contribution of suspended and abandoned infrastructure to CH4 emissions on a provincial and federal level remains unclear, as this infrastructure type is often ignored in measurement-based studies. If the overall contribution is significant, but remains overseen, Canada may fall short on achieving CH4 emission reduction targets in the near future.
To quantify the contribution of suspended and abandoned oil and gas infrastructure compared to emissions from active infrastructure in western Saskatchewan, we conducted mobile and stationary truck-based measurements in 3 different regions.
2. Materials and methods
2.1. Study area
For this study, 3 different oil and gas producing regions in Canada’s western province Saskatchewan were selected: Swift Current, Kindersley, and Lloydminster. In the surveyed areas, wells produce oil and gas from the Hogeland Basin in Swift Current, from the Bow Island Arch in Kindersley, and from the Athabasca Oil Sand Zone (Alberta Basin) in Lloydminster. Its economic distinction as the “Heavy Oil Capital of the World” correctly implies that Lloydminster produces mostly heavy oil. Production in Kindersley is dominated by gas and light oil, whereas Swift Current is characterized by gas and medium-dense oil production (Ministry of Energy and Resources, Government of Saskatchewan, 2020b; 2021).
2.2. Reported emissions
In Saskatchewan, solution gas emissions from the oil and gas industry are reported to the provincial government each month. Emissions are assessed by reporting facility connected with a certain number of wells via pipelines. The reports are made public each year, including the volume of solution gas emitted to the atmosphere through fuel processes, venting, and flaring (Ministry of Energy and Resources, Government of Saskatchewan, 2020c). The CH4 content of solution gas is roughly 80% in Swift Current, 85% in Kindersley, and 90% in Lloydminster. We assumed that for fuel emissions, CH4 was fully combusted and did not contribute to CH4 release to the atmosphere, whereas 5% of flared and 100% of vented emissions were associated with reported CH4 emissions. In addition to the publicly available annual report, we obtained the monthly report for September 2020 from the Government of Saskatchewan upon request.
2.3. Field measurements
All field measurements were conducted in September 2020 with a mobile laboratory. This laboratory consisted of a truck equipped with an Off-Axis Integrated Cavity Output Spectrometer measuring CH4, CO2, and H2O (Ultraportable Greenhouse Gas Analyzer, Los Gatos Research, San José, CA, USA), a heated 2D Ultrasonic anemometer (Model 86004, RM Young, Traverse City, MI, USA), a high-sensitivity GPS sensor (GPS 18 × 5 Hz, Garmin, Olathe, KS, USA), and an electronic compass (Model 32500, RM Young, Traverse City, MI, USA) to collect atmospheric measurements at a 2 Hz frequency (further details in the Supplement, Figure S1). During times when the GPS sensor was not working properly, we used auxiliary truck positions obtained with a cell phone (GPS Tracker Application). Real-time data were monitored on a laptop and simultaneously recorded on a datalogger (CR1000X, Campbell Scientific, Edmonton, AB, Canada). In addition, we used a hand-held optical gas imaging (OGI) camera (FLIR GF320, Wilsonville, OR, USA) to visually detect gas leaks and vents in situ.
2.3.1. Mobile surveys
To obtain CH4 emission rates attributed to oil and gas infrastructure located within the 3 studied regions (Figures S2 and S3), measured CH4 and carbon dioxide (CO2) concentrations, wind parameters, GPS locations, and compass directions were analyzed based on an established algorithm modeled after Zavala-Araiza et al. (2018) and further described elsewhere (Hurry et al., 2016; Atherton et al., 2017). In brief, we first filtered out inconclusive values, corrected and smoothed wind parameters with GPS and compass data, and linearly interpolated gaps. Subsequently, we identified excess levels of CH4 and CO2 (Figures S2 and S3) subtracting background levels determined with a running minimum over a moving window varying between 1 and 30 min depending on the best fit. Where the ratio of excess (e) concentrations eCO2:eCH4 < 100 indicated significant CH4 enhancement relative to the regular atmospheric concentration, we conducted back trajectory analysis in an effort to attribute the enhancement to known infrastructure upwind and within 400 m. The eCO2:eCH4 ratio was used to distinguish biological and combustion sources from oil- and gas-related emitters, which are subject to substantial CH4 relative to CO2 enhancement. Where attribution was possible, CH4 emission rates were estimated for each detected anomaly with an inverse Gaussian plume model considering peaks of enhanced CH4 concentrations, height of the attributed source, distance to the source, as well as atmospheric stability determined using meteorological data from nearest weather stations.
Truck-based measurements occurred at a site level and cannot be used to discriminate between colocated infrastructure. For this study, we grouped colocated infrastructure at a radius of 45 m, where all infrastructure within the radius was associated with the same group. The groups consisted of 1–12 wells or facilities. The terms “group” and “site” are used interchangeably in this study. Infrastructure groups were considered sampled when the truck passed by downwind within a distance of 400 m at least twice. If anomalous CH4 concentrations were detected from sampled infrastructure on more than 50% of passes, the group was tagged as emitting. However, if several upwind groups fulfilled this criterion, only the closest group was tagged as emitting.
Infrastructure groups were classified according to production status. Active groups consisted of at least one piece of active infrastructure plus any number of inactive infrastructure. Inactive groups consisted solely of inactive infrastructure. Suspended and abandoned wells were defined as inactive. Following the provincial Directive PNG032 (Ministry of Energy and Resources, Government of Saskatchewan, 2018), suspended wells are those which are capable but temporarily not producing and usually unplugged. We assume here that abandoned wells must be plugged as per Saskatchewan’s provincial Directive (Ministry of Energy and Resources, Government of Saskatchewan, 2019b), which differs from jurisdictions in other provinces in that well infrastructure may still be present albeit plugged. All other wells were assumed to be actively producing or injecting fluid. Inactive facilities were either decommissioned or suspended, although facility data did not distinguish between suspended and abandoned status. This classification relies on compliance of oil and gas operators who report and update well and facility status. Therefore, the distinction between active and inactive groups may be inaccurate in rare cases, but we assume that databases are correct. Emissions from active infrastructure might consist of both vented and fugitive emission sources, but since inactive sites should not be emitting, all emissions would be considered as fugitive in nature.
For in-depth analysis of emissions from inactive sites, we subcategorized inactive infrastructure groups based on the well status (suspended or abandoned), disregarding facility status in case the group was assigned a facility. The subcategories for inactive infrastructure groups were thus “suspended,” “abandoned,” “suspended and abandoned” for mixed groups, or “unknown” for groups only consisting of facilities where the distinction between suspended and abandoned status was unclear based on our dataset.
Sampled infrastructure was measured several times, and we report the mean estimates for each emitting group. In addition to standard statistics on a regional level, bootstrapped confidence intervals (CIs) of the mean were also determined with 1,000 bootstrap samples generated with replacement using the bootstrap percentile method (Efron and Tibshirani, 1986). The emission estimation approach used in this study entails a standard error (SE) of roughly 63% for point sources and tends to underestimate the emission rate as described in O’Connell et al. (2019). The same approach was used in numerous studies investigating emissions from oil and gas sites from mobile surveys (Hurry et al., 2016; Atherton et al., 2017; Baillie et al., 2019; MacKay et al., 2019; O’Connell et al., 2019; MacKay et al., 2021; Lavoie et al., 2022).
Where emission rates resulted from a one-time emission only despite several passes with the truck, we verified whether the respective infrastructure group was misattributed through visual inspection. As a note, active sites were prioritized over inactive sites for emission attribution during this process. Emission rates for misattributed sites (23 total) were reset to zero.
The emission frequency for each region was assessed by dividing the number of emitting infrastructure groups by the number of sampled infrastructure groups (emitting and non-emitting). Regional per-site emission rates were determined as the mean of all emission rate estimates including zeros, which corresponds to the product of the emission frequency and the mean emission rate of all emitting infrastructure groups per surveyed area.
During mobile CH4 measurements, we followed predetermined routes within the 3 regions stretching over nearly 2,000 km in total (Figures S2 and S3). These routes were subdivided into transects, with each transect surveyed consecutively in triplicate. Whenever we identified potentially high oil- and gas-related CH4 sources during mobile surveys, we conducted stationary CH4 measurements.
2.3.2. Stationary measurements
To complement our understanding of high emitters, we also applied a method similar to the stationary Other Test Method 33A (Edie et al., 2020; hereafter: OTM) developed by the U.S. Environmental Protection Agency. The most recent study suggests a 2σ error of ±70% for a single OTM 33A measurement and a low bias of roughly 5% for an ensemble of measurements (Edie et al., 2020). In addition, several studies agreed upon the fact that OTM-based emission rates tend to be underestimated (Brantley et al., 2014; Bell et al., 2017; Robertson et al., 2017; Edie et al., 2020). In this study, we conducted mobile surveys and stationary measurements consecutively. When a high-emitting source was identified during mobile surveys, we subsequently returned to the source location and confirmed the emitting source with the OGI camera if possible (Figure S4). Then, the truck was parked downwind of the emitter and free of major obstructions, while the gas analyzer and other instruments were recording measurements and were powered by the truck battery. For emission rate estimation purposes, the distance from the truck to the source was measured using a laser range finder and the source height was estimated visually. Stationary CH4 measurements lasted between 20 and 45 min at each site, at distances between 27 and 209 m from the source.
As for mobile surveys, the inverse Gaussian plume approach was applied to obtain emission rate estimates for OTMs. The stationary field measurements were conducted under the U.S. Environmental Protection Agency framework with slightly different data processing procedures. For rate estimation purposes, data obtained during OTMs were split into 20-min long overlapping intervals shifted by 1 min from each other. Therefore, at least one emission estimate was calculated per source (for a 20-min long OTM) and up to 25 emission estimates (for a 45-min long OTM). We assumed that the local ambient background concentration of CH4 was measured during these 20-min windows. The lowest 10% quantile of the measured CH4 values was subtracted to obtain eCH4. The initial azimuth of the source relative to the truck location was derived from the wind direction from which we observed highest eCH4 concentrations. The wind speed was assumed constant over the 20-min intervals and equal to the median wind speed. Both the final source azimuth and atmospheric stability class were chosen based on the best fit of the model. The vertical and horizontal spread of the emission plume was calculated using the Turner model (Turner, 1970). The emission rate was estimated for each 20-min window based on wind direction versus eCH4 profiles, and the median of these emission rate estimates for each location was used for further analysis. To adequately meet the Gaussian dispersion assumption, truck-to-source distances had to exceed a minimum of 3 times the source height. In this study, all stationary measurements satisfied this criterion.
In the field, we visually assigned each OTM to the emitting infrastructure, and using field notes, photos, and infrastructure databases (further details in the Supplement), we determined whether the infrastructure was active or inactive (suspended or abandoned).
3. Results and discussion
During mobile surveys, we sampled 1,220 infrastructure groups consisting of 1 to 12 single pieces of infrastructure each. This corresponded to 2,226 pieces of infrastructure, primarily wells and batteries, injection plants, and meter stations. In total, we attributed CH4 emission rates to 317 infrastructure groups, which resulted in an overall emission frequency of 26%. Fifty-seven percent of total sampled infrastructure groups were categorized active and the remaining 43% inactive. Among the groups tagged as emitting, 69% were active and 31% inactive.
For stationary measurements, we estimated CH4 emission rates at 19 different locations. Eighteen of them were conducted downwind of active sites, and one at an inactive site. The OTM sources of interest were mostly tanks but also flare stacks, wellheads, and pneumatic or engine sheds.
3.1. CH4 emissions from active infrastructure groups
CH4 emission rate estimates of active infrastructure in this study varied widely among surveyed areas (Figure 1), showing the characteristic heavy-tailed emission rate distributions as documented in other studies (Brandt et al., 2016; Zavala-Araiza et al., 2018). Only 7 emitting (of 120 sampled) active groups were identified in the Swift Current region during mobile surveys, however the emitters were individually large which meant that the estimated mean CH4 emissions in Swift Current exceeded those found in Kindersley and Lloydminster (Table 1). The highest number (142 of 405) of emitting active groups were found in Kindersley, followed by Lloydminster with 71 (of 167) emitting active groups but lower mean emissions than in the other 2 regions (Table 1).
|Stats. .||Unit .||Swift Current .||Kindersley .||Lloydminster .|
|Stats. .||Unit .||Swift Current .||Kindersley .||Lloydminster .|
Emission frequencies are given in percent, emission rates in m3/day per area for all emitting, active groups: Mean, median, minimum, maximum, 25th and 75th percentile, standard deviation (SD) and error (SE), and bootstrapped 95% confidence interval (CI) of the mean for 1,000 bootstrap resamples with replacement. Minimum and maximum were associated with group emission estimates per anomaly, the remainder with group-mean emission rates. Values were rounded.
Emission frequencies were 6% for Swift Current, 35% for Kindersley, and 43% for Lloydminster (Table 1). Mean per-site emission rates (±SE) for active infrastructure including non-emitting sites in Swift Current were 23 ± 16 m3/day (95% CI: 1–59 m3/day), 64 ± 17 m3/day in Kindersley (36–101 m3/day), and 36 ± 7 m3/day in Lloydminster (24–52 m3/day) based on our mobile surveys (Figure 2).
OTM-based emission rate estimates for active infrastructure ranged from 45 to 4,155 m3/day with highest emissions from unlit flares (Figure 3). Top-emitter rate estimates from active infrastructure during mobile surveys were in the same order of magnitude as the top OTM-based estimates. In the Kindersley area, highest group estimates from mobile surveys (mean ± SE; 4,719 ± 1,303 m3/day) slightly exceeded hotspot OTM-based estimates, whereas top estimates in Swift Current (1,700 ± 1,400 m3/day) and Lloydminster (562 ± 54 m3/day) were lower during mobile surveys. OTM-based emission estimates for active infrastructure were generally higher than in-motion estimates due to the fact that the truck was stationary for up to 45 min allowing for local concentration increases within the emission plume to be captured as opposed to quick passes with the moving truck. Although we observed differences between the two approaches, the differences were not systematic. Both types of measurements are shown in Figure 4.
According to our field observations during OTMs, which we carried out at large-emitting sites, we noticed a dominance of venting tanks in Kindersley, but unlit flares generally tended to emit highest amounts of CH4 (Figure 3). Overall, unlit flares emitted (mean ± SE) 3,508 ± 371 m3/day, venting tanks 1,091 ± 311 m3/day, venting pneumatic or engine sheds 1,082 ± 747 m3/day, and leaks from wellheads 301 m3/day. Therefore, unlit flares emitted on average at least 3 times more CH4 to the atmosphere than any of the other active oil and gas infrastructure. Based on our OTM measurements, 40% of total estimated CH4 emissions were attributed to emissions caused by (3) unlit flares and 50% to (12) venting tanks, while the remaining 10% were emissions from sheds and wellheads. While CH4 emissions from unlit flares can per se be easily avoided, tank vents are a known issue contributing to large CH4 emissions (Tyner and Johnson, 2021). However, regulations currently do not explicitly target the mitigation of tank vents (Environment and Climate Change Canada, 2019) and preventative measures are needed.
CH4 emission rates from the upstream oil and gas sector in western Canada have been determined in previous OGI, stationary, truck-based, and airborne studies (Anhalt, J., 2016; Atherton et al., 2017; Johnson et al., 2017; Roscioli et al., 2018; Zavala-Araiza et al., 2018; Baillie et al., 2019; MacKay et al., 2019; Chan et al., 2020; MacKay et al., 2021). Mean site-level CH4 emissions varied widely from <1 m3/day in the Weyburn CO2-Enhanced Oil Recovery field (MacKay et al., 2019) to 324 ± 79 m3/day (reported as 24.1 ± 5.9 t/h) in Lloydminster, Alberta (Johnson et al., 2017). Compared to other studies, mean per-site CH4 emission rates estimated for Swift Current were on the same order as Drayton Valley (26 m3/day, reported as 6 t/yr; Anhalt, 2016) and Red Deer (27 m3/day, reported as 7 t/yr; Anhalt, 2016 and 29 m3/day; MacKay et al., 2021) in Alberta, and Midale in southeastern Saskatchewan (20 m3/day; Baillie et al., 2019). Mean per-site emission rates in Kindersley were comparatively high, similar to emission estimates previously measured in Lloydminster (72 m3/day in 2016 and 74 m3/day in 2017; MacKay et al., 2021).
Most emission quantification studies in western Canada assessed that measured emissions exceed the amount reported to provincial regulatory bodies and in national CH4 emission inventories (Atherton et al., 2017; Johnson et al., 2017; Roscioli et al., 2018; Zavala-Araiza et al., 2018; Baillie et al., 2019; Chan et al., 2020; MacKay et al., 2021). We compared measured emission rates in each region with emission values reported to the Government of Saskatchewan. These reported emissions comprise both vented and flared emission estimates, which are typically intentional, whereas emission estimates obtained from our mobile surveys also include uncontrolled fugitive emissions where present. We compared measured to reported emissions in Figure 5, assuming that our momentary emission estimates were representative for emissions during the course of 2020.
Facility-level emissions based on measurements in September 2020 were not significantly different from annual or monthly reported emissions in Swift Current (95% CI: 38–58 m3/day in 2020, 49–76 m3/day in September 2020) and Kindersley (65–79 m3/day in 2020, 77–97 m3/day in September 2020). However, measured emissions in Lloydminster (24–52 m3/day) were significantly lower than emissions reported for the whole year of 2020 (86–106 m3/day), and even lower than those reported for September 2020 only (128–166 m3/day) as shown in Figure 5. The discrepancy between Lloydminster’s emissions in the annual and monthly reports can be explained by the fact that production peaked in September 2020 after recuperating from pandemic-driven drawbacks.
Among all Canadian oil and gas CH4 emission quantification studies, the Lloydminster region has always shown the highest mean site-level emissions, varying slightly between years and measurement methods: around 70 m3/day in 2016 and 2017 were found by MacKay et al. (2021), roughly 100 m3/day in 2016 according to O’Connell et al. (2019) and Zavala-Araiza et al. (2018), 170 m3/day in 2018 (MacKay et al., 2021), all from truck-based measurements, and 324 m3/day in 2016 for airborne measurements (Johnson et al., 2017). In this study, we measured only 36 m3/day on average in Lloydminster, which is a fraction of previously established rates for this development.
Why were measured values in Lloydminster so much lower in this study than past measurement estimates? There are a few possibilities. Differences in sample population could be the first explanation. Although we identified that most sampled and emitting wells produced oil (Figure S6), and that facilities were mainly the same heavy crude oil single-well batteries (Figure S7) deemed responsible for high CH4 emissions in previous studies (Sentio Engineering, 2015; Johnson et al., 2017; Roscioli et al., 2018; Tyner and Johnson, 2018), spatial differences in sampled infrastructure could have explained some of the departure from results shown in previous studies (further detailed in the Supplement, Figure S8). We also sampled fewer sites involving cold heavy oil production with sand (Petroleum Technology Research Centre, 2020) which was previously found to contribute largely to vented emissions (Sentio Engineering, 2015; Johnson et al., 2017; Roscioli et al., 2018; Tyner and Johnson, 2018). Declines in production could have explained the drop in observed emissions. Although Lloydminster’s oil and gas production significantly dropped at the beginning of the COVID-19 pandemic, prepandemic production levels were reattained by September 2020 when this study was carried out (Figure S5). But while production levels had recovered, fewer high-yield wells were contributing to production as the number of producing wells in Lloydminster has decreased by roughly half in past years (Figure S5). This decrease of active wells was verified and we found that numbers of suspended well counts increased by 3% between 2017 and 2020 in the studied area of Lloydminster. Therefore, some of the inactive sites in Lloydminster may have undergone recent suspension or abandonment, and possibly shut-in work and respective status reports may not have been transmitted into databases. The change in the character of production may have driven down CH4 emissions if marginal low production–high emission intensity wells were preferentially removed from service. Lastly, new regulation could have resulted in reduced emissions as compared with prior years. Overall, this study was carried out at an interesting moment in time, in a pandemic and during an oil price surge, in an oil and gas development in transition, where a slightly different population of oil and gas infrastructure was sampled than in past studies. We conclude that spatial and temporal variability are important determinants of CH4 emissions, and that neither can be considered as constants.
3.2. CH4 emissions from inactive infrastructure
In this study, 35% of all sampled sites were inactive in Swift Current, 31% in Kindersley, and 63% in Lloydminster (Figure 4). The inactive site emission frequencies in Swift Current, Kindersley, and Lloydminster were 0%, 19%, and 22% respectively (Table 2), which was lower than for active sites (Table 1). Roughly 9% of total (active plus inactive) site emissions in Kindersley were attributable to inactive groups of infrastructure, given that we sampled only about half as many inactive groups as active. In Lloydminster, 43% of total emissions came from inactive sites, but we sampled about 50% more inactive sites than active. We did not detect any inactive site emissions in the Swift Current region despite having sampled almost 70 groups composed entirely of inactive infrastructure. Per-site mean emission estimates for inactive infrastructure (Table 2, Figure 6) were 16 ± 5 m3/day in Lloydminster, 15 ± 4 m3/day in Kindersley, and 0 m3/day in Swift Current. For this study, we only conducted one OTM at an inactive site, which was located in Swift Current. The respective infrastructure was a leaking suspended water injection well according to our infrastructure database. The corresponding emission rate estimate was 12 m3/day, which was the lowest emission estimate among all OTMs.
|Stats. .||Unit .||Swift Current .||Kindersley .||Lloydminster .|
|Stats. .||Unit .||Swift Current .||Kindersley .||Lloydminster .|
CH4 emission frequency in percent, and rates in m3/day per area for all emitting, inactive groups: mean, median, minimum, maximum, 25th and 75th percentile, standard deviation (SD) and error (SE), and bootstrapped 95% confidence interval (CI) of the mean for 1,000 bootstrap resamples with replacement. Minimum and maximum were associated with group emission estimates per anomaly, the remainder with group-mean emission rates. No emissions from inactive sites were detected in Swift Current.
Compared to CH4 emissions from active infrastructure, inactive infrastructure tended to emit less CH4 in the surveyed areas (Figure 6) on a per-site basis. Respective proportions of emissions from active and inactive infrastructure to emitting sites are shown in Figure 7. The further the lines are from the baseline, the higher the percentage of sites that emit most CH4. Accordingly, Kindersley showed the highest percentage of sites emitting most CH4 for active infrastructure groups, while Lloydminster outweighed Kindersley for inactive infrastructure groups (Table 3). Additionally, the emission distribution in Lloydminster was slightly more heavy-tailed for inactive than for active infrastructure groups. A more heavy-tailed distribution implies that a small number of heavier emitting sites skewed the distribution and the mean emission rates. Therefore, this study shows that CH4 emissions from inactive infrastructure can be substantial and should be included in CH4 emission inventories. In addition, this study emphasizes the importance of limiting the period of well suspension, or to extend CH4 regulation to these sites. Age, reservoir potential to emit, gas migration, and many other factors could drive CH4 emissions from suspended sites. The fact that inactive site emissions are very different across geographies suggests that reservoir or subsurface properties or age could be important. Work is needed to determine what causes fugitive emissions at inactive sites so that the issues can be addressed with sensible regulation.
|Area .||Status .||Gini Coeff. (%) .|
|Area .||Status .||Gini Coeff. (%) .|
Gini coefficients in percent for active and inactive infrastructure groups in the surveyed areas. The Gini coefficient represents a measure of the distribution of emissions across emitting sites, derived by the Lorenz curve in Figure 7.
A breakdown of emissions from inactive sites into suspended and abandoned is shown in Figure 6. A reasonable distinction of suspended and abandoned groups was possible for Kindersley where only 10 (of 178) groups consisted of either both suspended and abandoned infrastructure or of facilities with unknown inactivity status. However, for Lloydminster a large fraction of inactive groups (104 of 286) could not be categorized as either suspended or abandoned, so that conclusions on emission rate quantification from these sites could not confidently be drawn for this region. In Kindersley, we found that site-level mean emission rates amounted to 16 m3/day for both only suspended and only abandoned groups, whereas emissions did not exceed 6 m3/day for other inactive sites. However, we emphasize that for reliable component-level emission quantification, particularly at inactive oil and gas sites, methods other than transect-based mobile measurements are more adequate.
Previous studies focusing on CH4 fluxes from abandoned wells in the United States (Kang et al., 2014; Kang et al., 2016; Townsend-Small et al., 2016; Pekney et al., 2018; Riddick et al., 2019; Saint-Vincent et al., 2021; Townsend-Small and Hoschouer, 2021) showed a wide range of emissions from below detection limit up to roughly 2 m3/day per well (reported as 6.0 · 104 mg h–1 well–1; Kang et al., 2016), with an average estimate of 0.21 m3/day per well for the United States and Canada (reported as 6.0 g/h per well; Williams et al., 2021). The values in this study were appreciably higher in some regions which suggest that either inactive well-related emissions are higher than expected, or that associated aboveground infrastructure is also a root cause. Overall, it is hard to draw precise parallels between inactive infrastructure in Canada and the United States because definitions and jurisdictional expectations differ. In Canada, a suspended well or facility (or site) is temporarily out of production (without a time limit) but theoretically capable of production, and where production infrastructure may be largely intact, CH4 leakage is possible. Generally, a well that has been plugged can be considered abandoned. Some U.S. studies focus specifically on abandoned wells, which would have less potential to emit than suspended sites comprised of wells and associated intact infrastructure including flow lines and tanks. In contrast to Canada, many U.S. states do not allow for long-term suspension of sites, and Canadian regulators also differ in their guidance. As a result of these variations, and the regional differences we observed, it is difficult to estimate the potential inventory of CH4 attributable to inactive sources nationally. More work is needed to understand inactive site emissions and their root cause.
Our study showed that emissions in the much-studied Lloydminster area were lower than expected, and lower than in previous years, which might be attributable to changes in well count or other spatiotemporal factors. This finding also shows the need to repeatedly conduct emission assessments. We hypothesize that marginal wells have been taken offline in favor of fewer higher yield wells from which emissions are more tightly controlled because these larger sites would have greater potential to emit and therefore would be equipped with emissions control equipment. If true, this would be a positive change because previous studies seem to show it is a standout region for emissions whereas other highly productive heavy oil regions in Canada like Peace River have been much less emissions intensive owing to hydrogen sulfide levels and the need for emissions control (Lavoie et al., 2022). More analysis is needed to understand the observed changes. Other oil producing regions we measured here emitted moderate amounts of CH4 on a per-site basis, and values were generally in line with regulatory reports. But in these areas, tank venting and unlit flares were major contributors to emissions and to a heavy tail distribution, and could be readily avoided. CH4 emission reductions by preventing unlit flares should be targeted, for example by igniting unlit flares to guarantee combustion of CH4 or by incorporating flare-gas-to-power projects. For future emission reduction efforts, we also suggest applying existing and developing new capture and commercialization opportunities.
Inactive infrastructure emitted less often and with less severity than active infrastructure, but emission levels were in some cases significant. In developments with high counts of suspended wells, and predisposing factors (which are unknown at this time but seem geography-specific), inactive sites could have a significant impact on overall CH4 emissions. Inactive site emissions may help reconcile some top-down mass balance studies, which sample bulk emissions from all sources, against other studies and regulatory reports that focus solely on emissions from active infrastructure. If CH4 control regulations are not extended soon to inactive sites, we would propose that such sites at least be measured occasionally, to help define the scope of the issue. In the 3 western provinces, there are roughly 200,000 suspended wells and facilities, and more than 300,000 abandoned wells (numbers from AccuMap, 31.03a Currency Update, by IHS Markit). If we scale up our mean per-site CH4 emission rate estimates with numbers of active (na) and inactive (ni) infrastructure groups in the broader regions of Swift Current (na = 883, ni = 521), Kindersley (na = 3,664, ni = 2,120), and Lloydminster (na = 1,176, ni = 2,108), almost 300,000 m3/day are emitted by active and 66,000 m3/day by inactive infrastructure in southwestern Saskatchewan. This means around 22% of overall emissions were attributed to inactive sites. Consequently, current provincial and federal CH4 emission estimates may be underestimated significantly if suspended and abandoned infrastructure is not included. As emissions from regulated active sources fall over the coming years, emissions from inactive legacy assets may become important in the inventory. We encourage further emission quantification studies of inactive infrastructure in different areas to manifest their importance and potential with regard to achieving Canada’s emission reduction target.
Data accessibility statement
Data was made publicly available: Vogt, Judith; Laforest, Justin; Argento, Mark; Kennedy, Sarah; Bourlon, Evelise; Lavoie, Martin; Risk, David, 2022, “Active and inactive oil and gas sites contribute to methane emissions in western Saskatchewan, Canada,” https://doi.org/10.5683/SP3/C9WLYG, Scholars Portal Dataverse.
The supplemental files for this article can be found as follows:
Figure S1. The mobile laboratory.
Figure S2. Surveyed routes in the studied areas.
Figure S3. Surveyed routes in Lloydminster, Saskatchewan.
Figure S4. Optical gas images taken with a forward-looking infrared camera.
Figure S5. Production history of wells.
Figure S6. Well characteristics.
Figure S7. Facility characteristics.
Figure S8. Emission factors of facility types.
We thank Environment and Climate Change Canada, Mitacs, and Geoverra for their support of this work. We also appreciate the Ministry of Energy and Resources at the Government of Saskatchewan for facilitating access to various datasets.
Funding was provided by Environment and Climate Change Canada (R0207021), Mitacs (R0196027), and Geoverra (R0176018).
The authors declare no competing interests.
Contributed to conception and design: JV, ML, DR.
Contributed to acquisition of data: JV, JL, MA, SK.
Contributed to analysis and interpretation of data: JV, EB, ML, DR.
Drafted and/or revised the article: All authors.
Approved the submitted version for publication: All authors.
How to cite this article: Vogt, J, Laforest, J, Argento, M, Kennedy, S, Bourlon, E, Lavoie, M, Risk, D. 2022. Active and inactive oil and gas sites contribute to methane emissions in western Saskatchewan, Canada. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2022.00014
Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA
Guest Editor: Stefan Schwietzke, Environmental Defense Fund, Boulder, CO, USA
Knowledge Domain: Atmospheric Science
Part of an Elementa Forum: Oil and Natural Gas Development: Air Quality, Climate Science, and Policy