Methane emission rate estimates of offshore oil platforms in Newfoundland and Labrador, Canada

Oil and gas production is responsible for more than 50% of methane (CH₄) emissions from the energy sector, and oil production facilities are known to emit more CH₄ than gas production facilities (International Energy Agency [IEA], 2021). Following the Paris Agreement to keep global temperature increases below 2°C, regional and national regulators began to address the problem by implementing new regulations to reduce greenhouse gas (GHG) emissions. Canada has committed to reducing national CH₄ emissions by 45% before 2025 and has more recently committed to a 75% reduction from the oil and gas industry by 2030 (Environment and Climate Change Canada [ECCC], 2022) after releasing Canada’s Strengthened Climate Plan: A Healthy Environment and a Healthy Economy on December 2020. In 2021, Canada’s upstream oil and gas sector was responsible for more than 28% of GHG emissions from extraction, oil refining, and fugitive emissions, which is almost 38% of CH₄ emissions from fugitive sources only (National Inventory Report [NIR], 2021).

Regulations inspecting the release of CH₄ and certain volatile organic compounds from upstream oil and gas in Canada came into force in January 2020 along with equivalent agreements in some provinces, which corresponds to a stricter leak detection and repair program focused on fugitive emissions exclusively and putting limits on venting. Based on the regulatory scenario, Canada’s emission reduction commitments seem achievable. Offshore rigs contribute to 20%–30% of global oil production (Statista, 2024a) and offshore Canada accounts for 5.3% of total oil production from just a few offshore facilities, making it the third highest provincial producer following Alberta and Saskatchewan (Statista, 2024b). Notably, all this production originates from only 4 platforms, and the Canadian offshore hosts only very large installations including Hibernia, which is the fourth largest facility in the world as measured by production level (The 6 Biggest Offshore Oil Platforms in the World, n.d.). These huge offshore oil production facilities are exempt from most of the new Canadian CH₄ mitigation requirements, based on the low CH₄ intensity reported by offshore operators and due to strict safety protocols used for offshore facilities that necessitate careful handling of combustible gas (Legislative Services Branch, 2023). Since onshore oil and gas operations in Canada have been shown to emit 25%–60% more GHGs than reported in federal inventory reports (Baillie et al., 2019; O’Connell et al., 2019; Chan et al., 2020; MacKay et al., 2021; Vogt et al., 2022), it is also reasonable to question the reported CH₄ emission rates and CH₄ intensity of Canada’s offshore oil production, and few other offshore CH₄ studies have targeted such a small number of very large facilities.

Studies have reported different rates of CH₄ emissions for offshore platforms as a function of production level, facility type, and region. Zavala-Araiza et al. (2021) measured 10 times lower CH₄ emissions rates than reported values for Mexican offshore production, while Gorchov Negron et al. (2020) results were almost 2 times higher than values reported for U.S. offshore production in the Gulf of Mexico. Similar results were observed by Riddick et al. (2019) with 1.5 times higher emissions during gas production compared to official reports. This all emphasizes the importance of carrying out actual measurement to assess Canada’s offshore oil production as a tool to understand emissions and their possible sources and update the mitigation methods for GHG emissions.

Accessibility is the challenge in measuring CH₄ emissions from offshore oil production platforms, which can be addressed by using ship (Riddick et al., 2019; Yacovitch et al., 2020) or airborne (Gordon et al., 2015; Gorchov Negron et al., 2020; France et al., 2021; Zavala-Arazia et al., 2021) measurements. Previously used analytical methods varied by region and data type, including but not limited to Gaussian Dispersion (GD; Riddick et al., 2019; Yacovitch et al., 2020) and mass balance methods (Gordon et al., 2015; Conley et al., 2017; Gorchov Negron et al., 2020; France et al., 2021). Erland et al. (2022) compared some of the aforementioned mass balance approaches and showed that depending on assumptions and conditions, results could be comparable with each other (Gordon et al., 2015; Conley et al., 2017; Erland et al., 2022). Mass balance approaches require defining an enclosed volume with a rectangle (Kalthoff et al., 2002; Gordon et al., 2015; France et al., 2021), polygon (Kalthoff et al., 2002), or ellipse base (Conley et al., 2017; Gorchov Negron et al., 2020; Zavala-Arazia et al., 2021).

Here, we quantify CH₄ emissions and production-weighted intensities from 3 active oil production facilities in offshore Newfoundland and Labrador: Hibernia, Hebron, and SeaRose, during the autumn of 2021. We combined aircraft-based measurements with the Top-down Emission Rate Retrieval Algorithm (TERRA) mass balance approach, introduced by Gordon et al. (2015) and a GD method for emissions rate estimation and deriving production-weighted emission intensities for each platform. Two separate flight patterns were conducted for each platform during routine production. We compare our results to other offshore studies, reported values from the industry, and Canadian onshore environments, which are thought to have higher CH₄ intensities.

At the time of this study, the Canadian province of Newfoundland and Labrador had 4 sites of offshore oil production: SeaRose, Hebron, Hibernia, and Terra Nova. Terra Nova was not producing at the time of measurement, and therefore, it was excluded from our study. The platforms were 300–350 km from the province’s coastline in the Jeanne d’Arc Basin (Figure 1).

Hibernia is a gravity-based structure (GBS) producing oil since 1997, making it the oldest Canadian offshore facility. The production was reported to be 36.2 MMbbl yr⁻¹ (Million barrels per year) in 2021. Hebron is also a GBS and is the newest facility that started production in November 2017. It had the highest production level, totaling 50.6 MMbbl yr⁻¹ in 2021. SeaRose is a floating production, storage, and offloading (FPSO) vessel, active since 2005, with 5.80 MMbbl yr⁻¹ of oil produced in 2021. Terra Nova is another FPSO which started production in 2002. At the time of our study, Terra Nova was docked for maintenance (Canada-Newfoundland and Labrador Offshore Petroleum Board [C-NLOPB], 2024). Reported CH₄ emission estimates from each platform are published annually through the Greenhouse Gas Reporting Program (GHGRP) by the ECCC (GHGRP, 2021). Using values from the GHGRP (2021) and the production data (C-NLOPB, 2024), the offshore production’s total GHG emission intensities were calculated as 1.99 g CO₂ eq MJ⁻¹, 0.82 g CO₂ eq MJ⁻¹, and 1.91 g CO₂ eq MJ⁻¹, respectively, for Hibernia, SeaRose, and Hebron in 2019. At the time of this study, GHGRP 2004–2021 was available online, so we used 2019, 2020, and 2021 values to compare with the measurement results from this study.

In order to perform continuous measurement of CH₄ on Twin Otter aircraft, the plane was equipped with a high precision (0.10 ppb) Picarro G2210-i gas analyzer (Picarro, 2024) directly connected to the air inlets (0.25 in. Synflex tubing) that goes outside the copilot window using diagram pumps. The data were collected every 0.5 s for CH₄ mixing ratio. An Aventech wind and meteorological sensor (AIMMS-30, 2024) was used to measure the three-dimensional wind data and also was used to collect relative humidity, pressure, and temperature with the Air Data Probe mounted on the fuselage above the cockpit. The sensor had ±0.50 m s⁻¹ uncertainty for wind speed, ±0.30°C for temperature, and ±2.0% for relative humidity. The 2 GPS antennas came with the wind sensor were installed on top of the wings to collect aircraft location data. For real-time visualization, all instruments were connected through a central processing module. Accuracy of wind measurement was vital for good mass balance measurements. The Aventech wind measurement system incorporated a three-dimensional pitot tube sensor with 2 GPS units on the wings and a central processing unit. The raw measurements included barometric pressure, air speed, angle of attack, angle of sideslip, temperature, relative humidity, movement of the aircraft, and measured wind vectors. From these measurements, the three-dimensional winds in the atmosphere were output by the central processing unit at 1 Hz frequency with a precision of 0.50 m s⁻¹ in x, y, and z directions at a 150-knot air speed (77 m s⁻¹). A full aerodynamic calibration was completed at the beginning of the campaign involving various control maneuvers moving into the wind, and away from the wind, to relate the sensor input at its mounting point of the pitot tube, with the GPS and other information. The aerodynamic calibrations were checked by the sensor manufacturer to confirm goodness of fit. Although we did not have weather stations offshore against which we could check, we did compare each day’s data against the airport values when we were in vicinity of the airport, and they agreed well. This article’s methodology for measurement and analysis has been previously defined and reported (Risk et al., 2022).

Taking the tube length into account before each flight, Picarro’s response time was determined. Breath tests were performed before each flight to observe the lag time of the analyzer, which was usually between 49 and 56 s (Risk et al., 2022).

The gas analyzer was calibrated before the campaign using 2 gas standards of differing CH₄ mixing ratio prepared by Ameriflux and referenced to <1 ppb against the World Meteorological Organization scale for CH₄ using standards maintained by the U.S. National Oceanographic and Atmospheric Administration. We used Ameriflux FB03987 (4.52 ppm CH₄, 516.54 ppm CO₂, prepared August 2019) from which we found an average departure of 0.008 ppm CH₄, and Ameriflux FB04007 (1.81 ppm CH₄, 385.18 ppm CO₂, prepared August 2019) from which the average offset was 5.6 × 10⁻⁴ ppm CH₄. Before measuring the standards, the analyzer cavity was purged for several hours using zero gas water vapor measurements that were below 0.05%. The gas analyzer was checked immediately after the campaign using the same process to check for drift during the weeklong campaign.

The Picarro gas analyzer was benchmarked regularly using a single compressed air cylinder to check for drift throughout the campaign. The cylinder of air was compressed at the beginning of the campaign, and for consistency, the same cylinder was used for the duration of the campaign. In order to perform benchmarking, an inlet would connect Picarro and the compressed air cylinder. The variations in detected values from cylinder were observed for 5 min to ensure consistency in readings. This would also help finding any leaks in tube connections (Risk et al., 2022). Before or during the campaigns, we observed no drift in calibration or benchmarking checks that exceeded 2 ppb, primarily due to high-precision analyzer (Picarro, 2024).

To be sure that the AIMMS sensor was reporting data correctly, we conducted a comprehensive assessment of wind speed data to determine precision and variability and to ensure that it behaves similarly during the orbits as in straight-line flying. We plotted wind speeds during orbits, and we did not observe any systematic variability when flying in different directions that would exceed the manufacturer-specified 0.5 m s⁻¹ resolution. Observed differences were within normal wind expectations for variation with time and height. As expected, Allan Deviation plots for speed and direction, for both transects and loops, showed that precision always improved as averaging intervals increased, and we saw no worsening of precision at the looping frequency of roughly 240 s, which would have occurred if there had been a systematic bias in wind speeds, for example, when flying up versus downwind. We also assessed Random Walk noise K, which showed the same result. The only flight for which precision was worse than 0.2–0.3 m s⁻¹ during the loops was one flight we excluded from the dataset due to very strong (>20 m s⁻¹) wind speeds in which we saw precision values as low as approximately 1 m s⁻¹. Overall, variance was always less than 0.43 m s⁻¹ and both loops, and subsequent transects (straight line horizontal flight legs) did not show different characteristics, which makes us believe the AIMMS-30 sensor was reporting correctly and within manufacturer specifications during loops.

Our mass balance measurement method was adopted from similar onshore and offshore experiments (Gordon et al., 2015; Conley et al., 2017). Depending on the atmospheric transport model, the measurements could have been performed with loops around the facility (cylindrical or box method) or downwind transects at a different altitude, along with at least 1 upwind measurement to account for the background concentration (Gordon et al., 2015; Conley et al., 2017; Gorchov Negron et al., 2020). Our field campaign included loops for a mass balance technique and downwind transects for the GD method (Risk et al., 2022).

For mass balance technique, 6–10 loops were flown around each platform (approximately 600–1,000 m radius), ascending from the lowest possible flying altitude (approximately 100 m ASL), with an increasing altitude between loops by approximately 100 m. This enabled us to capture the entire vertical profile of the plume. For Gaussian method, 3–6 transects were flown between about 1.5 and 15 km downwind (Hensen et al., 2019) of the platform. The distance was selected based on capturing a plume at a close location and moving away until there is no dispersion from the plume. For the downwind transects, we flew at diverse altitudes (in approximately 150 m increments), permitting us to seize vertical and horizontal profiles of the plume. Our downwind transects, conducted at various altitudes (in 150 m increments), captured both vertical and horizontal plume profiles (Risk et al., 2022).

A total of 9 flights were performed around each of the 3 active offshore facilities during 2 separate campaigns in October and November 2021. The aircraft departed from St. John’s airport toward each active facility 3 times and made loops and transects around each platform as previously described. Figure 2 shows different sets of transect and loop measurements from each platform which were almost close to ideal conditions for analysis methods. Gordon et al. (2015) suggested performing flights only if the wind speed was greater than 5 m s⁻¹ for mass balance analysis. We attempted to operate within the 5–20 m s⁻¹ wind speed threshold, but offshore winds were highly variable.

Following the measurement campaigns, data from individual instruments were processed and merged into consistent time series. We then selected plume measurements suitable for GD considering the length and strength of the plume and background concentrations. Onshore (O’Connell et al., 2019) and offshore (Riddick et al., 2019; Yacovitch et al., 2020) oil and gas facilities have used the Gaussian plume model to estimate CH₄ emissions. In the Gaussian plume model, source flux rate is a function of model fraction of gas downwind of the point source, the relative height and distance between source and sensor, stability class of atmosphere and horizontal wind speed (Gordon et al., 2015). The Gaussian equation for a continuous elevated point source is as follows:

where Q is the emission rate of the point source, σ_y and σ_z are dispersion coefficients in the crosswind and vertical direction (m), u is the horizontal wind speed (m s⁻¹), C(x,y,z) is the downwind CH₄ concentration enhancement (ppmv), h_m is the measurement height (m), h_s is the source height (m), and x is the downwind distance from the measurement to the point source (m). While the assumptions of Gaussian distribution may not be met for all cases, with no information on sources, it is impossible to control the constant rate, and wind speed does not vary significantly for each detected plume of emission (Risk et al., 2022).

In Equation 1, C is the CH₄ concentration calculated by subtracting the background concentrations of CH₄ from the raw CH₄ measurements. Background CH₄ is the baseline CH₄ in ambient air plus any additional sources upwind the location of measurement. Background CH₄ was calculated using the average of 30 s of measurements from either side of the plume, as done in other offshore measurement studies such as France et al. (2021) and Gorchov Negron et al. (2020). We did this calculation separately for every observed plume, as the calculated background CH₄ can vary due to changes in atmospheric stability. The source height (i.e., flare or platform deck) was derived for each platform based on publicly available information. σ_y and σ_z were estimated using the Pasquill–Gifford stability class scheme and corresponding equations defined by Turner (1970) (Risk et al., 2022). There are different methods for calculating σ_y and σ_z, but there was not enough information based on our data for a reliable assumption. In the 1985 study, Hanna et al. described modifications to the stability class calculation for the Environmental Protection Agency’s Offshore and Coastal Dispersion model with more details. The calculation was based on roughness length and Obukhov length which required data on mixing height, surface temperature, air temperature, and wind speed at 10-m height. They also suggested that extrapolation could be used to derive these values but, in our case, the lowest flight altitude was approximately much higher at 160 m and extrapolation might introduce significant uncertainties. In applying the modifications proposed by Hanna et al. (1985), we did adopt the convention of using stability class C in place of A and B. Therefore, a range of stability classes (C–F) were considered instead, corresponding to predefined σ_y and σ_z values for each stability class (Turner, 1970). This was reflected in our uncertainty ranges for the Gaussian model estimates.

Based on the principle of conservation of mass, it is possible to apply the mass balance approach even without knowing the location of a source inside the reference volume. The amount of gas emitted from a facility can be roughly calculated by subtracting background values and mixing ratio that leaves the enclosed volume from the downwind side.

For this study, we adopted the TERRA method (Gordon et al., 2015) where the total emission rates as a sum of all possible sources coming in and going out of the reference volume:

TERRA considered the horizontal advection (E_C.H) and diffusion (E_C.HT) flux and vertical advection (E_C.V) and diffusion (E_C.VT) flux from the sides and top of the enclosed volume, increase of the mass within the volume caused by chemical reactions (E_C.X), changes in the density of the air (E_C.M) and deposition (E_C.VD) (Gordon et al., 2015). These terms account for the conservation of mass for horizontal and vertical advection of the enclosed volume minus the air mass changes within the volume.

One average fitted ellipse was required to represent each flight to create the enclosed volume for mass balance analysis. This ellipse was defined around the facility to project the three-dimensional location data into a two-dimensional cylinder screen surrounding the facility. The cylinder reached from sea level to the highest measured altitude and sat perpendicular to the sea level around the facility. Please refer to the Supplemental Information section SI.1, where additional details and figures were provided for fitting a loop and creating the enclosed volume (cylinder).

Water level variation during the survey was considered as the baseline for the cylinder. Water level varied by 0.30–1.70 m from the mean sea level during the campaign (Fisheries and Oceans Canada [DFO], 2021).

Since no measurements are available for below flight level, we extrapolated meteorological variables for the area between the sea surface and the lowest measured altitude (164 m for this campaign). Meteorological data below flight level were acquired using the ERA5 reanalysis data from the European Center for Medium-Range Weather Forecasts (Bell et al., 2021). Further information on the process of fitting a curve to the obtained forecast data for values below flight level was included in the Supplemental Information section SI.2. Once wind fits were available, we used the Kriging interpolation (Gordon et al., 2015) to make the mass balance enclosed volume (SI.3).

After projecting the wind speed data onto the two-dimensional screens, we used simple kriging with the range, nugget, and sill as 500, 0, and 1, respectively, to interpolate wind speed for the entire screen surface with a 20 m × 40 m resolution. We assumed a well-mixed boundary layer and used the fitted values from ERA5 data to extrapolate grid properties below flight level (Gordon et al., 2015).

The same procedure, called an air flux screen, was used to plot the variations in air parcel movements within the reference volume. The values for the air flux screen were interpolated using surface pressure and temperature (Gordon et al., 2015).

The CH₄ mixing ratio flux approach followed a similar process, but this time, instead of using constant kriging values, a semivariogram was fitted to each round of simulation in TERRA. Five different methods for extrapolating below flight level values were included: zero, constant, zero to constant, linear fit, and exponential fit (Gordon et al., 2015). All 5 methods were applied to each dataset to account for the uncertainty of the extrapolation method.

The mixing ratio required a simple kriging interpolation method, as mentioned before. Gordon et al. (2015) applied a manual semivariogram to the residuals, modifying the base nugget, sill, and range for each set of simulations. We tested the use of an automated semivariogram fit method by modifying TERRA and adding a fitting engine. In some cases, the fitting engine did not work appropriately, and therefore, a case-by-case comparison between manual and automated fit was applied, and the fit was poor and needed manual improvements. It mainly occurred when the vertical motions were observable in the atmosphere. Flight HI2 and HI3 were examples of such conditions. Please refer to Supplemental Information section SI.3 for more examples of automated and manual fit comparison, complete details, and figures for wind and flux screen interpolation and semivariogram for the mixing ratio kriging step.

The 2 methods we used in this study have different sources of uncertainty for emission estimations. For the Gaussian plume model, the main sources of uncertainty are related to background concentration, source height assumption (i.e., flare or platform deck), and atmospheric stability class. The uncertainty on the background concentration was less than ±10% based on calculations, which is small compared to the uncertainty on stability classes and source height. For each measured plume selected for GD calculations, emission rates were estimated using different combinations of plausible stability classes and source heights, which were used to estimate the resulting uncertainty from these parameter assumptions.

Defining stability class showed to be the most significant source of uncertainty for the Gaussian method. Since we were unable to confidently determine values for σ_y and σ_z and as described earlier based on Hanna et al. (1985), we considered a range of plausible values (C–F) reflected in the uncertainty. Quantified plumes measured during the same flight were averaged to compute the facility’s overall mean emission rate, which was used for subsequent analysis and comparisons.

For the TERRA model, Gordon et al. (2015) introduced different parameters and assumptions as a source of uncertainty, including the extrapolation method for CH₄ concentration (below flight level), the extrapolation method for wind data (below flight level), distribution of wind speed time series, distribution of CH₄ concentration time series, mixing ratio rate leaving from the top of the cylinder, the cylinder’s height (or flight height), and background concentration for CH₄ (Gordon et al., 2015). We followed the same methods where possible, and we added the size of the fitted ellipse as a source of uncertainty in this study. The chosen method for fitting the loop resulted in changing the controlled volume and, consequently, the emission rates. The extrapolation method of the CH₄ mixing ratio above the ending flight height brought a significant source of uncertainty due to changes in boundary layer conditions for one of the flights and, therefore, was included in rate estimation.

Considering the boundary layer conditions and plume extension in height, our ending altitude was typically defined by other criteria, mainly cloud ceiling and visibility, which in most cases was up to approximately 1,000 m. Following the methodology outlined by Gordon et al. (2015), the boundary layer height for each flight was calculated based on dew point temperature. For some flights, this height ranged between 600 and 900 m, which was lower than the final flown altitude. For other flights, such as SR5, no distinct change in dew point temperature could be identified during the flight. The uncertainty associated with the boundary layer height was addressed in the TERRA approach by restricting the flight loop to the calculated boundary layer. Hence, we used interpolated surfaces or screens of 400, 500, 700, and 1,300 m heights.

Although other meteorological conditions like boundary layer height, vertical turbulence, change in air density, pressure, and temperature below flight level could also contribute to a level of uncertainty, there were no weather station or wind profiler data close enough to the platforms to permit a further investigation. Therefore, these variables did not contribute to the uncertainty analysis included in this study. We applied a bootstrapping method with 10,000 samples to account for the variation in range for TERRA analysis.

We calculated CH₄ intensities (megajoule of energy emitted per megajoule of energy produced) of Canadian offshore oil production using our estimated emission rates and reported yearly oil production for 2021 for each facility. Oil production was converted to megajoules using a conversion rate of 1 m³ = 38,510 MJ (light crude) and 40,900 MJ (heavy crude).

During the measurement campaigns, we encountered various meteorological conditions, including high winds and low cloud ceiling altitude. These conditions and other factors sometimes limited our ability to fly ascending loops evenly. To contextualize the quality of the measurements for applying atmospheric transport models like TERRA, we derived an equally weighted percentage (quality index) based on which of the following conditions were met in any given flight: wind speed between 5 and 20 m s⁻¹, absence of an outside source, reached upper boundary layer, plume center of mass captured, and sufficient number of loops.

For example, for flight HE4, with zero loops, it is impossible to perform a TERRA analysis. Therefore, the quality index will be zero, no matter how other conditions were met during the flight. Table 1 summarizes flight details for this field campaign and the quality index for TERRA analysis. It is noteworthy to mention that for flights with a quality index below 60%, the uncertainty will be pretty high, and, in some cases, TERRA is incapable of performing an analysis for specific flights where the value is 40% or lower.

Based on our measurements, mean GD emission rate estimates for SeaRose, Hibernia, and Hebron are 3,400 m³ CH₄ day⁻¹ (30,300 m³ CO₂ eq day⁻¹), 4,200 m³ CH₄ day⁻¹ (37,200 m³ CO₂ eq day⁻¹), and 9,500 m³ CH₄ day⁻¹ (83,800 m³ CO₂ eq day⁻¹), respectively. The SR1 flight happened when the SeaRose platform was not producing, and therefore we did not observe a distinguishable plume for the Gaussian method. Since no enhancements were detected, we assume the platform produced little to no CH₄ emissions during the shutdown, as expected.

After filtering data to select measured plumes based on the criteria of a well-mixed boundary layer (temperature changes in height), consistent wind speeds (low to moderate), and clear CH₄ enhancements (nonlinear change by height) from the background, 15 out of 19 unique plumes detected during 41 downwind transects were selected for GD analysis. The remaining 4 plumes were deemed unsuitable for the Gaussian method because they either overlapped with elevation changes, which could mean that the source is not from the platform, or the plume shape had non-Gaussian characteristics (e.g., 2 peaks). Of the 15 plumes selected for GD, 10 were from Hebron, 4 were from Hibernia, and 1 was from SeaRose. Please refer to the Supplemental Information section SI.4 for a detailed explanation and figures of plumes that could not be used for this study.

As previously mentioned, several plausible emission rates were calculated for each plume, using different combinations of stability class and source height (platform base height and flare height) to incorporate the uncertainty related to these parameters. With the Gaussian method, average plume emission rates ranged from 3,060 m³ day⁻¹ (Hibernia) to 10,600 m³ day⁻¹ (Hebron). Individual plumes measured on the same flight showed good agreement.

Although we could not always observe a clear plume during all the selected loops like in Figure 2, we established that 6 out of 9 flights were suitable for the mass balance technique. Overall, TERRA mass balance-derived emission rates from our measurements were 2,900 ± 2,800 m³ CH₄ day⁻¹ (25,500 m³ CO₂ eq day⁻¹), 860 ± 360 m³ CH₄ day⁻¹ (7,600 m³ CO₂ eq day⁻¹), and 8,500 ± 7,600 m³ CH₄ day⁻¹ (74,600 m³ CO₂ eq day⁻¹), respectively, for SeaRose, Hibernia, and Hebron. The uncertainty range was wider for some flights, mainly due to the low-quality index factor, as mentioned in Table 1.

Flight HE6 loop is illustrated in the top right plot of Figure 2. High concentrations were measured while flying around the Hebron platform on the day of measurement. The mixing ratio values were close to the platform deck height, and no extra enhancement was observed at higher elevations. Considering the dominant wind was coming from the west, the source was most probably from Hebron. The same went for the flight HI2 loop on the top left plot of Figure 2. During flight HI2, the wind was mainly from the north, and a source was detected close to the ocean’s surface. Although there were other CH₄ enhancements observed at higher elevations, the detected emissions were not observed at higher elevations.

Flight height and loop size had the most significant contribution among all factors considered for sources of uncertainty. The estimated rate for changing the loop size based on 1 standard deviation was sometimes double or half of the bootstrapped average. The same result was observed when the same cylinder’s height was decreased without changing the loop’s size. However, the stability class of the atmosphere, high wind speeds, and low humidity were significantly more challenging, as observed in HI8 and SR9. Flights HI2 and HI3 had the highest uncertainty, mostly related to the height of the cylinder. The TERRA method encountered measurement constraints during the HI2 and HI3, where the center of the plume was below the lowest altitude, and the mixing layer height was lower than the final loop. This resulted in underestimations for mass balance approach. The SeaRose facility was shut down during the SR1 flight, which can be seen in the uncertainty range.

Unfavorable atmospheric conditions and cloud cover made it impossible for us to detect a plume during the orbit measurements for HE4, HI8, and SR9. As indicated in Table 1, the quality index for all these flights was 40%, and the requirement for the TERRA mass balance method was not met for these flights. Please refer to section SI.5 of Supplemental Information for further explanation and figures for HI8.

During flight SR5 to SeaRose, no major CH₄ peaks were detected while ascending until the last loop, and the quality index for this flight was 40%, suggesting that a particular atmospheric condition was causing the plume to rise higher than expected. However, we were able to perform upward extrapolation and still estimate an emission rate using TERRA. For further details on SR5 flight conditions, please refer to section SI.5 of Supplemental information.

Estimated emission rates from this study were compared to industry-reported CH₄ emissions from the most recent GHGRP report (2021). Since GHGRP reports rates in tonnes per year (1 ton = 1.47 × 10³ m³), to compare emission values from this study estimated in m³ day⁻¹, we divided the reported values by the number of production days for each facility in 2019 to compute a daily reported value. While Hibernia and Hebron were in regular operation in 2019, SeaRose was only active for about 180 days due to safety-related shutdowns (C-NLOPB, 2022). Figure 3 shows that the estimates from this study are comparable to those reported in the GHGRP (2021). Based on a comparison between GHGRP-reported values, after a change in regulations and shifting from estimates to the measurement for reporting, a significant decrease can be observed in values after 2019. It seems that before the regulations came into force in 2020, operators may have reported higher emissions, while in the following years, they shifted to estimating lower emissions. As shown in Figure 3, based on information from C-NLOPB, before 2020, the Canadian Association of Petroleum Producers Best Management practice was a reference for offshore production to report their emissions based on estimates. Still, in 2020, they shifted to using optical gas imaging (OGI) to locate the sources of fugitive emissions and using emission factors to estimate the emission rate from located sources. This aligns with Johnson et al. (2021), which suggests OGI measurement underestimates emissions by a scale of 1.6–2. Other studies also showed that traditional methods that use emission factors show up to 60% of underestimation for federal inventory (Baillie et al., 2019; O’Connell et al., 2019; Chan et al., 2020; MacKay et al., 2021; Vogt et al., 2022).

Flaring, power generation, venting, and fugitive equipment leaks are the primary sources of CH₄ emissions for Newfoundland and Labrador’s offshore production. Even though wind speed is shown to have a minor effect on flare slip CH₄ emissions at low/moderate wind speeds (<10 m s⁻¹), the combustion efficiency of simple open flares drops to 90% at 11 m s⁻¹ (40 km h⁻¹) (Castiñeira and Edgar, 2008). However, all 3 offshore facilities use an enclosed flare design, which should minimize the impact of wind speed on combustion efficiency (Johnson and Kostiuk, 2002). For the measured platforms, the proportion of CH₄ in flared gas varies, but assuming 85.9% CH₄ (ExxonMobil Canada Properties, 2011), we can use the total CH₄ emissions downwind to calculate an apparent efficiency and compare with realistic efficiency values to indicate whether the flare is the main contributor to downwind CH₄. To perform the calculation, we use a simplified version of the combustion efficiency formula as follows:

Here, CE_Methane refers to the percentage of CH₄ downwind, over the total flared CH₄. Apparent CE_Methane was 93.8 ± 2.80% and 94.6 ± 4.00% for SeaRose and Hibernia, respectively. In the case of Hebron, apparent combustion efficiency values were persistently below 60%. While some studies showed that the flare efficiency could be higher than 97% (ExxonMobil Canada Properties, 2011; Gvakharia et al., 2017), others suggested values as low as 60% (Plant et al., 2022). For the sake of this study, we assumed that these enclosed flares at these facilities should generally be meeting a 97%–98% efficiency targets and that apparently lower values suggested fugitive additions. For 2 of the platforms, the apparent combustion efficiency suggested that most of the downwind CH₄ could be explained by flare slip and that fugitive contributions were small. Low apparent flare efficiencies at Hebron suggested a greater contribution from fugitive sources.

Production-weighted CH₄ emission intensities for each offshore platform were calculated using the average estimated emission rate from measured data and reported oil production for the same year (2021). Emission and production rates were converted to energy units (MJ) to compute the amount of energy lost per energy produced, allowing for an equal comparison across different production types.

Figure 4 compares the CH₄ intensities of Canadian offshore and onshore production using all available data sources (MacKay et al., 2019; MacKay et al., 2021; NIR, 2021; Wang et al., 2021; C-NLOPB, 2022; MacKay et al., 2022; Vogt et al., 2022; Petrinex, 2024; Methane Reduction, n.d.). It should be mentioned that oil sands are excluded from this comparison. Our comparison indicated offshore oil production has the lowest emission intensity across Canada, which could partly result from safety-related emissions limitations for offshore production.

Compared with other offshore producing regions, Canada’s offshore facilities have relatively low absolute CH₄ emissions with moderate to low CH₄ emission intensity—and much lower intensity than most onshore sources in Canada. Gorchov Negron et al. (2020) used a mean regional estimate of 10 platforms in the Gulf of Mexico, while Yacovitch et al. (2020) performed a regional estimate for an area consisting of 103 offshore platforms. Zavala-Araiza et al. (2021) covered even more extensive areas in the Gulf of Mexico consisting of more than half of the offshore oil and gas production facilities. Riddick et al. (2019) reported the highest and lowest emission estimates for gas or oil and gas platforms. Hensen et al. (2019) consisted of different campaigns for several platforms in the North Sea. Figure 5 compares CH₄ intensities based on these other offshore studies. Although offshore oil production facilities seem to have high emission rates, considering the production level, the intensity is low (Hensen et al., 2019; Riddick et al., 2019; Gorchov et al., 2020; Yacovitch et al., 2020; Zavala-Araiza et al., 2021). We could not calculate the intensities for some of the mentioned studies due to a lack of accurate product type and rate information.

Considering all possible sources of CH₄, flaring and fugitive leaks could be 2 of the main contributors to offshore emissions. Flare combustion efficiency is expected to be 98%, but recent studies have shown that actual combustion efficiency is likely lower (Plant et al., 2022). Depending on the facility and the production levels, slight variations from that figure would represent substantial increases or decreases in CH₄ emissions. Aside from reducing fugitives on the facility from sources like compressors, flare optimization is essential, but more accurate measurement methods are needed to support that optimization effort, considering the offshore environment is challenging with high winds and frequent fog.

Fugitive equipment leaks and flaring seem to be the most probable emission source despite offshore platforms’ high safety and risk controls. The stationary measurement from the platform and bag sampling for the isotopic analysis could help understand the exact sources of emissions and the percentage of contribution for each. From flare combustion calculation for CH₄, fugitives can be one of the main contributors to the total emissions.

Venting is another possible emission source, but it is limited by federal regulations, with each venting event requiring a permit explaining the reason (such as in case of an emergency). Operators must also plan a safety management system that avoids venting as much as possible to reduce the environmental impact. There was also no venting during the days of measurement, so it is not possible to attribute these emissions to venting.

This research conducted the first measurement campaign of CH₄ emissions from Canadian offshore oil platforms. Our results showed that reported values agree (within uncertainty) with estimated emissions derived from direct measurements and 2 emission rate estimation methods (mass balance and GD). Although more frequent measurements during different production phases and times of the year would improve the results, our estimates suggest that offshore production has much lower emission intensities than onshore oil and gas production in Canada. We suspect fugitive emissions are the main contributor to the total CH₄ emissions from each platform. Our results also suggest that OGI measurement for reporting could be underestimating the total fugitive emissions for offshore oil production; however, more measurement is needed to confirm this.

Future work should focus on scheduled campaigns during different operational conditions and include measurements of mixing ratios and atmospheric conditions below flight level, CO₂ data, and staying within limits for wind speed and boundary layer conditions to reduce uncertainties and help determine the primary source of emission.

The datasets generated and analyzed in the current study, along with sources of other datasets used for external comparisons, are available (in xlsx format) for public download via https://doi.org/10.5683/SP3/AOOQHE.

The supplemental files for this article can be found as follows:

Khaleghi et al. 2024_Supplemental Infromation.docx

This project was supported by Natural Resources Canada’s Emissions Reduction Fund (Offshore Research, Development and Demonstration Program), which is managed by Energy Research & Innovation Newfoundland & Labrador.

This work was supported by Natural Resources Canada’s Emissions Reduction Fund, Offshore RD&D program, which was administered by Energy Research and Innovation Newfoundland and Labrador [project numbers E065, 2021].

The authors have declared that no competing interests exist.

Contributed to conception and design: AK, KM, AD, LAJ, DR.

Contributed to acquisition of data: KM, DR.

Contributed to analysis and interpretation of data: AK, KM, AD.

Drafted and/or revised the article: AK, KM, AD, LAJ, DR.

Approved the submitted version for publication: AK, KM, AD, LAJ, DR.

How to cite this article: Khaleghi, A, MacKay, K, Darlington, A, James, LA, Risk, D. 2024. Methane emission rate estimates of offshore oil platforms in Newfoundland and Labrador, Canada. Elementa: Science of the Anthropocene 12(1). DOI: https://doi.org/10.1525/elementa.2024.00025

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Guest Editor: Brian Lamb, Environmental Engineering Laboratory for Atmospheric Research, Washington State University, Pullman, WA, USA

Knowledge Domain: Atmospheric Science

Part of an Elementa Special Forum: Oil and Natural Gas Development: Air Quality, Climate Science, and Policy