Satellite-Derived Estimate of City-Level Methane Emissions from Calgary, Alberta, Canada

Abstract

Cities are important sources of anthropogenic methane emissions. Municipal governments can play a role in reducing those emissions to support climate change mitigation, but they need information on the emission rate to contextualize mitigation actions and track progress. Herein, we examine the application of satellite data from the TROPOspheric Monitoring Instrument (TROPOMI) to estimate city-level methane emission rates in a case study of the City of Calgary, Alberta, Canada. Due to low and variable annual observational coverage, we integrated valid TROPOMI observations over three years (2020–2022) and used mass balance modeling to derive a long-term mean estimate of the emission rate. The resulting column-mean dry-air mole fraction (XCH₄) enhancement over Calgary was small (4.7 ppb), but within the city boundaries, we identified local hot spots in the vicinity of known emission sources (wastewater treatment facilities and landfills). The city-level emission estimate from mass balance was 215.4 ± 132.8 t CH₄/d. This estimate is approximately four times larger than estimates from Canada’s gridded National Inventory Report of anthropogenic CH₄ emissions and six times larger than the Emissions Database for Global Atmospheric Research (EDGAR v8.0). We note that valid TROPOMI observations are more common in warmer months and occur during a narrow daily overpass time slot over Calgary. The limited valid observations in combination with the constrained temporal observational coverage may bias the emission estimate. Overall, the findings from this case study highlight an approach to derive a screening-level estimate of city-level methane emission rates using TROPOMI data in settings with low observational coverage.

1. Introduction

Reduction in methane (CH₄) emissions is considered an effective near-term pathway to reduce anthropogenic greenhouse gas emissions that exacerbate climate change [1,2]. CH₄ emissions from the energy (e.g., oil and gas production and coal mining), agriculture (rice cultivation and feedlots), and waste (landfills) sectors are three major sources of anthropogenic CH₄ [3,4,5,6,7].

Urban areas, in particular, are very poorly understood sources of CH₄ emissions due to the large diversity of sources. In urban areas, wastewater treatment facilities and landfills are generally well known to be large emitters [8,9,10,11]. However, urban areas also contain many smaller, widely distributed sources such as leaks in natural gas (NG) distribution infrastructure, small-scale agriculture, sewer emissions, natural gas vehicles, and slip from NG combustion [8,12,13,14,15,16]. Most of these sources are poorly understood and may vary considerably in emission characteristics and rates.

Due to the diversity and number of sources across urban areas, conventional bottom-up (BU) CH₄ emission inventories are normally inaccurate. There are considerable challenges associated with maintaining and implementing BU inventories in urban areas as inventories must model both the number of discrete emission sources and predict appropriate emission rates. Granular BU inventory models in urban environments will always carry considerable uncertainty. And by extension, inventory-based mitigation policy will be inherently inefficient as this uncertainty carries through to policy decisions.

To reduce the uncertainty of BU inventories and understand if BU inventories are adequately simulating emissions, it is helpful to use larger-scale top-down (TD) measurements. Measurements of atmospheric concentrations of CH₄ in urban areas can be carried out with platforms such as ground-based stations and networks, vehicle- and aircraft-based systems, and satellites, e.g., [17,18,19,20,21,22,23]. The measurement resolution varies substantially from the equipment level to the basin level depending on the platform. The TD approach uses these measurements with atmospheric transport modeling to estimate emission rates and help understand the fidelity of BU estimates.

Among TD methods, satellites are of particular interest. The high observational density and large-scale geographic coverage of the TROPOspheric Monitoring Instrument (TROPOMI) on board the Copernicus Sentinel-5 Precursor satellite provides one of the best options for satellite TD measurements. Multiple approaches have been proposed and used to estimate urban CH₄ emissions using TROPOMI observations, including transport inversion, two-dimensional Gaussian modeling, and tracer–tracer approaches [10,23,24,25]. Based on CH₄ measurements from a multi-tiered observing framework (including observations from TROPOMI, an aircraft-based system, and ground-based stations), Cusworth et al. [24] used geostatistical inverse modeling and prior emissions to estimate CH₄ emissions from the Los Angeles megacity. De Foy et al. [10] used a two-dimensional Gaussian model with TROPOMI observations and estimated CH₄ emissions from 61 urban areas globally. The tracer–tracer method, which quantifies emissions by scaling the ratios of prior estimates of CH₄:CO or CH₄:CO₂ with atmospheric measurements, has been used to estimate CH₄ emission rates from urban centers in North America [23,25].

The majority of the city-level CH₄ emission estimate methods require high observational coverage and fine spatial resolution [26,27]. For urban areas in Canada, it is more challenging to use TROPOMI observations to estimate the city-level CH₄ emissions because of the relatively low number of valid observations and the small sizes of the urban areas. In this case study of the City of Calgary, we explore the feasibility of TROPOMI to estimate the city-level CH₄ emission rate and present an approach using data integration to overcome low observational coverage. We integrated all valid observations during the 3-year period from 2020 and 2022 and applied the mass balance method of Buchwitz et al. [28] to derive a long-term screening-level estimate of Calgary’s CH₄ emission rate. We explore the context of these measurements relative to existing inventories and estimates from other cities. We also examine how the characteristics of the TROPOMI satellite in high-latitude settings could influence representative emission estimates.

2. Materials and Methods

2.1. Study Area

Calgary is located in southern Alberta in the transitional area between the foothills of the Rocky Mountains and the North American Great Plains. In 2021, Calgary covered a land area of 820.62 km² [29]. The city center coordinates are 51.04° (latitude) and −114.06° (longitude). Calgary’s population in 2022 was 1.4 million [30]. The urban area bordered by the City of Calgary’s municipal boundaries is shown in Figure 1. In our work, the entire urban area was predefined as the source region and further screened using the TROPOMI-observed CH₄ enhancements.

2.2. TROPOMI CH₄ Observation

As the only payload on Sentinel-5P (S5P), TROPOMI is a nadir-viewing push-broom grating hyperspectral spectrometer with a field-of-view of 108°. TROPOMI covers wavelengths of ultraviolet–visible (UV-VIS, 270 nm to 495 nm), near-infrared (NIR, 675 nm to 775 nm), and shortwave infrared (SWIR, 2305 nm–2385 nm) [31]. The operational S5P Level-2 processor employs the RemoTeC full-physics algorithm to retrieve the atmospheric CH₄ content from Earth radiance measurements in the NIR and SWIR spectral bands [32,33]. Independent validation by the S5P Mission Performance Centre and the S5P Validation Team concluded that the CH₄ total column data have good overall agreement with reference measurements collected from a global network of ground-based stations (e.g., Total Carbon Column Observing Network and Network for the Detection of Atmospheric Composition Change) and the data products from other satellite platforms (i.e., Japanese Greenhouse Gases Observing Satellite) [32,33]. The retrieved atmospheric CH₄ content is provided as the column-mean dry air mixing ratio of CH₄ (XCH₄, ppb) in the TROPOMI Level-2 data product. To correct the underestimation of the TROPOMI XCH₄ data for low albedo, bias-corrected XCH₄ is also provided in the data product.

The data used in this work included 3 years of the TROPOMI Level-2 data product, which spanned the period from 1 January 2020 to 31 December 2022. The spatial resolution of all observations in this work is 7 km × 5.5 km. For each observation, the bias-corrected XCH₄ value and corresponding retrieval quality indicator “qa_value” were extracted from the data for analysis. In this study, only high-quality, bias-corrected XCH₄ values were used (hereafter: valid observations), as indicated by qa_value > 0.5. The overpass time of TROPOMI for Calgary was between 12:00 p.m. and 14:00 p.m. local time. The 3-year mean XCH₄ observations from 2020–2022 at 0.05° × 0.05° resolution were obtained from all valid observations during the study period.

The data processing procedure is summarized in Figure 2. Since the original TROPOMI XCH₄ data products are provided with a non-normalized grid, the first step was to develop a normalized 0.05° × 0.05° latitude-longitude grid (hereafter: pixel) for reallocation of multiple observations. For each individual observation, every valid XCH₄ data point (i.e., qa_value > 0.5) was reallocated onto the pixel that corresponded with its geolocation. This process was conducted using a TROPOMI daily screening toolkit developed by Gao et al. [34].

The second step involved screening the number of valid observations per pixel in 2020, 2021, and 2022 for both the source region (City of Calgary boundary) and the surrounding region used to determine the background XCH₄ (Figure 3). This step was used to assess the coverage of the valid TROPOMI observations for subsequent processing to derive the emission rate estimate. Coverage describes the sample size of valid observations per pixel and within the source and surrounding regions and the temporal distribution of valid observations throughout the year. Generally, confidence in the CH₄ emission hotspot screening and emission rate estimate increases with increasing coverage. Considering the diversity of CH₄ sources in urban areas (e.g., landfills, natural gas point-sources, and end-use incomplete combustion), coverage should ideally capture variations in the characteristics and rates of emissions from these sources across space and through time. Therefore, as coverage increases, the representativeness of CH₄ emission rate estimates for the study period should improve.

Figure 3 indicates variable year-to-year and month-to-month coverage within the source and surrounding regions for the study period. The lowest coverage was in 2020, and all the pixels of the source region contained <10 valid observations. Coverage improved in 2021 but varied spatially, particularly with less valid observations over the source region. Coverage was higher and least variable in 2022. Monthly variations were also noted (Figure 3e). Five months in 2020 had no valid observations, and in all three years, there were no valid observations in December and January. In general, summer and fall had the highest coverage.

Given the year-to-year and month-to-month variations, we integrated the data from 2020–2022 to increase the sample size for emission quantification. By doing so, all pixels within the source and surrounding region were covered by at least 17 valid observations. We averaged the XCH₄ in each pixel across the valid observations. The 3-year mean XCH₄ map was then denoised using Gaussian filters prior to emission quantification. Filtering was completed using a 2D Gaussian smoothing kernel with a standard deviation of 1 and a filter size of 3.

2.3. Emission Quantification: Mass Balance Model

We used the method proposed by Buchwitz et al. [28] to estimate the city-level CH₄ emission rate. The method uses mass balance theory to build an integrated emission rate estimate from multiple TROPOMI data scenes. From the underlying theory, an air parcel with background XCH₄ travels through the source region (characterized by the effective length, $L$, calculated as the square root of source region area) as directed by the wind speed, $V$. It is assumed that the XCH₄ of the air parcel is enhanced as determined by the accumulation time, τ (duration of the air parcel travelling through the source region, calculated as $L/V$), and the emission rate ($Q$, mass per time) in the source region. The CH₄ column mass enhancement (ΔmCH₄, CH₄ mass per area) is calculated as $\left(Q\times \tau \right)/L^2$. It is assumed that when travelling through the source region, ΔmCH₄ increases linearly. Therefore, the mean column mass enhancement over the source region can be calculated as 0.5 × ΔmCH₄, where ΔmCH₄ is calculated from the observed XCH₄ by introducing a factor, $M$, with the unit of kgCH₄/(km² × ppb). As a result, the modeled column mass enhancement and the observation-based column mass enhancement are connected. The basic equation is as follows:

$$Q=∆{XCH}_4\times CF$$

(1)

where the ΔXCH₄ is the CH₄ enhancement in ppb, derived from the 3-year mean XCH₄ map and $CF$ denotes the conversion factor to convert the CH₄ enhancement over the source region to the emission rate, $Q$. $CF$ is calculated from the following Equation (2):

$$CF=M\times M_{exp}\times L\times V\times C$$

(2)

where $M$ is the factor to convert the atmospheric total column mixing ratio to mass change per area for standard conditions (i.e., surface pressure at 1013.0 hPa), which is 5.345 kg CH₄/(km² × ppb); $L$ (km) is the effective length of the source region; $V$ (km/d) is the ventilation wind speed during the TROPOMI overpass derived from in situ measurements at multiple air monitoring stations in the Calgary urban area; $C$ is a unitless factor of 2.0, which is the assumed linear increase in CH₄ column mass enhancement of the air parcel when travelling over the source region [28]; and $M_{exp}$ is a dimensionless factor to correct for the actual mass, calculated using the following equation:

$$M_{exp}=\frac{〈M_i〉}{M}\approx \frac{〈p_i〉}{1013.0}$$

(3)

where $p_i$ is the surface pressure of the i-th pixel (in hPa), which is obtained from the 3dimensional monthly mean in the MERRA-2 reanalysis data product [35]; $\langle \rangle$ denotes averaging over all pixels of the source region; and 1013.0 refers to the surface pressure (hPa) under standard conditions.

Background XCH₄ is required to calculate ΔXCH₄ over the source region. Background XCH₄ can be estimated using XCH₄ upwind of the study region or a statistical approach [29]. The number of valid upwind observations may be insufficient to provide a robust estimate of background XCH₄ given impeding factors such as cloud cover, surface albedo, and aerosol optical depth [26]. Similarly, given that background XCH₄ is non-homogenous in large study regions, background estimates derived with the statistical approach can be biased. Lauvaux et al. [36] proposed an optimized statistical background XCH₄ estimation approach in their study of global CH₄ ultra-emitters using noisy satellite imagery. Briefly, for a region with ultra-emitting CH₄ plumes, the computation of the background XCH₄ value is sensitive to the skewness of the XCH₄ values, determined by $\left({\overline{x}}_{{XCH}_4}{-\mu }_{{XCH}_4}\right)/{\sigma }_{{XCH}_4}$/, where ${\overline{x}}_{{XCH}_4}$, ${\mu }_{{XCH}_4}$, and ${\sigma }_{{XCH}_4}$ are the mean, median, and standard deviation of the XCH₄ values in the region. Thus, as indicated by Equation (4), for regions with highly skewed XCH₄, the background is computed as ${\mu }_{{XCH}_4}$. Otherwise, the background is computed as $l\times {\mu }_{{XCH}_4}-\left(l-1\right)\times {\overline{x}}_{{XCH}_4}$, where l is an empirical parameter that has been used to segment emission hotspots. To maximize the probability of emission hotspot segmentation, the value of l is typically 2.5 [36].

In this work, we adopted the statistical approach of Lauvaux et al. [36] to estimate the background XCH₄. Here, the source region (i.e., City of Calgary) replaces the ultra-emitters present in Lauvaux et al.’s study [36]. A broader region surrounding the source region was then used to calculate the background XCH₄. The size of the surrounding region was defined based on a compromise between several conditions to: (i) limit contributions from other anthropogenic sources of CH₄ emissions outside the source region, (ii) ensure similar or better coverage than in the source region, and (iii) avoid areas with no valid observations. According to these conditions, the surrounding region was defined as a rectangular area centered on the source region with latitude range from 50.5° to 51.5° and longitude range from −114.5° to −113.5°, as indicated by the dashed outlines on the maps in Figure 3.

The pixel-specific CH₄ enhancement, ΔXCH₄, was derived by subtracting background XCH₄ from the 3-year mean XCH₄ for each pixel in the source region. Pixels with ΔXCH₄ < ${\sigma }_{{XCH}_4}$ were not considered as source pixels and were therefore not included in the calculation of ΔXCH₄. This was performed to exclude pixels with ΔXCH₄ biased by the regional variation in XCH₄.

$${XCH}_{4,background}=\left\{\begin{array}{c}{\mu }_{{XCH}_4},if\frac{{\overline{x}}_{{XCH}_4}-{\mu }_{{XCH}_4}}{{\sigma }_{{XCH}_4}}>0.3\\ 2.5\times {\mu }_{{XCH}_4}-1.5\times {\overline{x}}_{{XCH}_4},if\frac{{\overline{x}}_{{XCH}_4}-{\mu }_{{XCH}_4}}{{\sigma }_{{XCH}_4}}\le 0.3\end{array}\right.$$

(4)

Uncertainty in the emission estimate (σ) in this work mainly comes from the uncertainty in the calculations of ΔXCH₄, ${\sigma }_{{∆XCH}_4}$, and the conversion factor $CF$ (${\sigma }_{CF}$). Therefore, we applied the law of propagation of uncertainty [36,37] and calculated σ as the combined uncertainty of ${\sigma }_{{∆XCH}_4}$ and ${\sigma }_{CF}$ in the source region (e.g., Equation (5)). The main contributor to ${\sigma }_{CF}$ is uncertainty in wind speed. In this work, ${\sigma }_{CF}$ was assumed equivalent to ${\sigma }_V$, and was estimated from the temporal variation in wind speed during averaging. The calculations of ${\sigma }_{{∆XCH}_4}$ and ${\sigma }_{CF}$ are indicated by Equations (6) and (7), respectively.

$$\sigma =\sqrt{{\sigma }_{{∆XCH}_4}^2+{\sigma }_{CF}^2}$$

(5)

$${\sigma }_{{∆XCH}_4}={\sigma }_{{XCH}_4}\times M\times M_{exp}\times L\times V\times C$$

(6)

$${\sigma }_{CF}={\mathsf{\Delta }\mathrm{X}\mathrm{C}\mathrm{H}}_4\times M\times M_{exp}\times L\times {\sigma }_V\times C$$

(7)

3. Results and Discussion

3.1. Methane Enhancements

Figure 4 shows maps of the mean XCH₄ and ΔXCH₄ over the source region. Figure 4a shows that the mean XCH₄ per pixel within the source region was higher than the surrounding region. Within the source region, higher XCH₄ areas were found in central and northeast Calgary (Figure 4b). The mean XCH₄ in the source region was 1879.1 ± 2.7 ppb, while the mean and median XCH₄ within the surrounding region were 1873.3 ppb and 1874.5 ppb. The background XCH₄ estimated using Equation (4) was 1876.3 ppb. It should be noted that not all pixels within the source region were found to have elevated XCH₄. Therefore, the calculated mean XCH₄ for the City of Calgary does not exhibit substantially higher XCH₄ compared to the background XCH₄.

Pixel-specific enhancements, ΔXCH₄, were derived by subtracting the background XCH₄ from the observed XCH₄ (Figure 4b). Only the pixels with enhancements greater than the standard deviation (2.7 ppb) were defined as ‘sources’ within the source region. The mean ΔXCH₄ from the ‘sources’ was 4.7 ppb. The highest enhancement (7.4 ppb) was over central Calgary near one of three active landfills. There are 23 facilities in Calgary that are required to report annual CH₄ emissions to Canada’s Greenhouse Gas Reporting Program (GHGRP). As indicated in Figure 4b, nearly all of the facilities (21/23) were located within the XCH₄ enhanced areas as determined in our analysis.

Compared to TROPOMI-based ΔXCH₄ estimates from other urban areas in North America [10], the mean ΔXCH₄ in Calgary from 2020–2022 is near the upper end of city-level enhancements derived from cities with similar populations (Figure 5). For the 18 cities with populations of less than 5 million, ΔXCH₄ does not noticeably correlate with the population. Although Calgary’s population is only half of Toronto (Canada), our TROPOMI-based ΔXCH₄ estimate for Calgary (4.7 ppb) is slightly higher than that estimated by de Foy et al. [10] for Toronto (4.5 ppb). It should be noted that for de Foy et al. [10], ΔXCH₄ for the urban areas studied was estimated based on a two-dimensional Gaussian fit method and not strictly constrained by the city boundaries. Thus, different ΔXCH₄ estimation methods may introduce bias.

3.2. City-Level CH₄ Emission Rate and Comparisons with BU Inventories

Using the mass balance method, we calculated Calgary’s 3-year mean emission rate as 215.4 t CH₄/d with an uncertainty of 132.8 t CH₄/d over the period 2020–2022. The uncertainty is dominated by ${\sigma }_{{∆XCH}_4}$ (111.2 t CH₄/d). The large ${\sigma }_{{∆XCH}_4}$ is mainly attributed to spatial variations in ΔXCH₄ (Figure 4b). As discussed in Buchwitz et al. [28], spatial variations in emissions over the source region can bias the emission estimate. The uncertainty associated with wind observations was smaller (72.6 t CH₄/d) and may reflect the potential bias caused by the temporal variations in wind speed in the source region during the study period. Therefore, the uncertainty of 132.8 t CH₄/d in this work indicates the extent to which our emission estimate may be biased by the spatial and temporal variations in the emissions sources and wind speed in Calgary, respectively.

Figure 6 shows TROPOMI-based city-level CH₄ emission estimates for select North American cities [10,23]. In general, the emission rate increases with population. Compared to CH₄ emission rate estimates from other major cities, Calgary’s emission rate is relatively low but comparable with cities that have slightly lower and higher emissions and populations, respectively (e.g., Portland, Charlotte, Kansas City, Washington, DC, and Baltimore). Although the city-level emission estimates were all derived from TROPOMI observations, the methods used to quantify emissions varied and differed from the mass balance method in this work. Plant et al. [23] used the tracer–tracer ratio approach (ΔCH₄:ΔCO), while de Foy et al. [10] used a two-dimensional Gaussian fit method. The noticeable discrepancies between the emission estimates by Plant et al. [23] and de Foy et al. [10] for the same cities (e.g., New York City, Mexico City, Boston, and Baltimore) imply that measurement-based, city-level CH₄ emission estimates are methodologically dependent. For current and emerging satellite-based platforms capable of measuring CH₄, it is crucial to establish a suite of proven and well-tested emission quantification approaches to improve confidence and enable direct comparisons between estimates.

City-level CH₄ emission estimates from BU emission inventories are much lower than our measurement-based estimate.

Table 1. City-level CH₄ emission rate estimates for Calgary.

Data Source	Emission Rate (t CH4/d)	Year
EDGARv8.0	31.6	2020
EDGARv8.0	38.2	2021
EDGARv8.0	34.9	2022
EDGARv8.0	34.9	2020–2022
Gridded NIR	48.3	2018
GHGRP 1	20.3	2020–2021
This work	215.4 ± 132.8	2020–2022

¹ Only includes sources with annual emission rates > 10 kt CO₂-eq.

lists the city-level CH₄ emission rates in Calgary from several gridded emission inventories and emissions reported to Canada’s Greenhouse Gas Reporting Program (GHGRP). The most recent EDGAR inventory (i.e., v8.0) estimates annual city-level CH₄ emissions for Calgary at 31.6 t CH₄/d (2020), 38.2 t CH₄/d (2021), and 34.9 t CH₄/d (2022), with a mean of 34.9 t CH₄/d. Canada’s gridded national inventory report (NIR) of anthropogenic CH₄ emissions estimated 48.3 t CH₄/d for Calgary in 2018 [38]. This is higher than the EDGARv8.0 estimates but still only accounts for less than one quarter of the measurement-based estimate reported herein using TROPOMI. The mismatch between the BU emissions inventories and the TROPOMI-based estimate from this study is consistent with the discrepancies observed between inventory and measurement-based estimates of CH₄ emissions for other major cities in North America [10,23,25,39].

Canada’s GHGRP collects information on greenhouse gas emissions annually from facilities across Canada. As a mandatory program, facilities that emit 10 kilotonnes (kt) or more in carbon dioxide equivalent (CO₂-eq) units per year, must report their emissions to Environment and Climate Change Canada [40]. A total of 23 facilities reporting to the GHGRP within Calgary’s city limits averaged 20.3 CH₄/d between 2020 and 2021, which is 9.4% of the emission estimate using TROPOMI data and 42.1% and 58.2% of total inventoried emissions in the gridded NIR and EDGARv8.0, respectively. The large gap between the GHGRP-reported emissions, BU emission inventories, and our measurement-based estimate suggests that Calgary’s emissions profile is dominated by CH₄ sources below the GHGRP reporting threshold of 10 kt CO₂e-eq/year or emissions from those sources are under-reported. Overall, our results highlight the need for a re-examination of the emissions factors, activity factors, and methods used to develop BU inventories and expansion of the facilities covered by GHGRP requirements.

3.3. Observational Bias

As suggested by EDGARv8.0, there is modest seasonal variation in Calgary’s CH₄ emissions, with slightly lower emissions in the summer than in the winter (Figure 7a). This is consistent with measurement-based estimates from the Baltimore and Washington, DC, metropolitan region in the U.S. [41]. Urban CH₄ sources (e.g., slip from incomplete combustion of natural gas at end-use, landfills, and wastewater treatment) are characterized by noteworthy temporal variations related to natural gas consumption, barometric pressure, temperature, and moisture [42,43,44,45]. It is possible that different emissions processes compete and dominate the city-level emissions profile at different times of the year. The valid TROPOMI observations for Calgary were mainly constrained to the period from June to October in each year. The higher observational coverage during summer and fall may fail to capture the CH₄ emission characteristics of other sources in other months of the year and thus bias the TROPOMI-based emission estimate.

As a supplemental indicator, the monthly natural gas consumption of City of Calgary-owned facilities was included in Figure 7a to examine potential seasonality- and temperature-induced fluctuation in CH₄ emissions related to natural gas end-use. Generally, EDGARv8.0 CH₄ emissions are higher in colder months and appear to correlate to some degree with the natural gas consumption data. For context, the main fuel used for space heating in the City of Calgary is natural gas. We used the EDGARv8.0 emission variability to seasonally correct our Calgary estimate from 215.4 ± 132.8 t CH₄/d to 220.1 t CH₄/d. Compared to the large uncertainty surrounding our mean estimate, this correction has little impact on the magnitude of the city-level emission rate estimate. The limitation of using the seasonality from EDGARv8.0 to temporally correct our estimate is that inventories have known issues with underestimating CH₄ emissions compared to measurement-based estimates. Therefore, it is challenging to draw more definitive conclusions on annual emissions using TROPOMI data because the observations are temporally constrained.

Another noteworthy point of using TROPOMI observations is the fixed sensing time slot on each day for Calgary. TROPOMI’s overpass time slot for Calgary is between 12:00–14:00 local time. For sources with strong diurnal variations, the fixed observation time may result in marked uncertainty when scaling up emission rate estimates from hourly to daily/monthly/annual. Measurement-based emission rate estimates conducted in London, UK, show the highest emission flux around 12:00 p.m. with a maximum-to-minimum ratio of 1.9 [46]. Figure 7b shows the average hourly variations in CH₄ concentrations from three ground-based air monitoring stations in Calgary. The data show that CH₄ concentrations are lower during the TROPOMI overpass time slot. Furthermore, the TROPOMI overpass time for Calgary spans typical work hours during the week. CH₄ emission measurements from a sensor mounted on a light-rail transit platform in Salt Lake Valley, Utah, found temporal variations in emissions from a manufacturing facility [47]. Plumes from this facility were only detectable during work hours. Similar sources of emissions likely exist in Calgary. This introduces more uncertainty into scaling the hourly emission rate up to an annual rate, as emission patterns from different sources may fluctuate between days/evenings, weekdays/weekends, and holidays.

Figure 7. Temporal variations in CH₄ emissions and sources in the City of Calgary: (a) monthly natural gas consumption data for City of Calgary facilities (bar) and BU inventory estimates from EDGAR v8.0 (line) and (b) average hourly (local time) CH₄ concentration levels from three air monitoring stations for the period from June to October in 2020, 2021, and 2022. Natural gas consumption data were obtained from the City of Calgary open data [48]. Continuous hourly ground-level CH₄ concentration measurements were retrieved from the Alberta Air Data Warehouse [49]. The shaded time slot in (b) is the TROPOMI overpass time for Calgary.

Figure 7. Temporal variations in CH₄ emissions and sources in the City of Calgary: (a) monthly natural gas consumption data for City of Calgary facilities (bar) and BU inventory estimates from EDGAR v8.0 (line) and (b) average hourly (local time) CH₄ concentration levels from three air monitoring stations for the period from June to October in 2020, 2021, and 2022. Natural gas consumption data were obtained from the City of Calgary open data [48]. Continuous hourly ground-level CH₄ concentration measurements were retrieved from the Alberta Air Data Warehouse [49]. The shaded time slot in (b) is the TROPOMI overpass time for Calgary.

The long-term mean CH₄ emission rate estimated from the mass balance method is relatively robust to poor satellite coverage but is well acknowledged to be less accurate than more complex methods because of the simplified transport algorithm. It is important to emphasize that our estimate for Calgary carries considerable uncertainty (see Buchwitz et al. [28] for an extensive discussion of the uncertainties inherent to the mass balance method). The observational bias discussed in this work likely introduces additional uncertainties in the CH₄ emission estimate. It is possible to correct this observational bias with support from BU emissions inventories, but inventories suffer from known issues, and thus, such corrections may be unreliable. Despite the observational bias inherent in our estimate, our results suggest that inventories may not be accurately estimating Calgary’s total CH₄ emissions and that emissions may be considerably higher.

4. Conclusions

In this work, we explored the feasibility of using TROPOMI observations and the mass balance method of Buchwitz et al. [28] to estimate a city-level CH₄ emission rate for Calgary, Alberta, Canada—a city with low annual satellite observational coverage. To overcome the low observational coverage from TROPOMI in Calgary, we integrated all valid observations from 2020–2022. The 3-year mean TROPOMI XCH₄ observations revealed elevated XCH₄ within the study area. Pixels of elevated XCH₄ overlapped with the distribution of larger, well-known CH₄ sources in Calgary. The mean XCH₄ enhancement from the pixels within the study area was 4.7 ppb.

The mean emission rate estimate for Calgary between 2020–2022 was 215.4 ± 132.8 t CH₄/d. This estimate falls on the lower end of other TROPOMI-based CH₄ emission rate estimates of other major North American cities and is comparable with the magnitude of emissions in cities with similar populations. Consistent with the findings of previous studies, the gaps between our TROPOMI-based city-level CH₄ emission rate estimate and BU inventories suggest that the latter are underestimating CH₄ emissions from the City of Calgary. Evidence shows that city-level CH₄ emissions in general (and Calgary specifically) are higher than initially predicted, but the exact source of the discrepancy cannot be reliably determined from these data. Future research to better understand the emissions profile of Calgary (and by extension, similar urban areas) with more granular measurements is warranted. BU inventories may not be accounting for all sources of CH₄ emissions, activity factors could be inaccurate, or some combination of the two. Thus, research to catalogue all potential sources of CH₄ emissions in cities may lead to better modeling and more accurate inventory estimates.

Our analysis shows that TROPOMI observations of CH₄ emissions from Calgary are biased to warmer months and are performed over a fixed daily time slot. As such, this bias likely introduces uncertainty into our emission estimate. Future work will assess whether more complex models or other data and methods can be used to correct this bias for cities like Calgary with limited satellite coverage.