A decadal synthesis of atmospheric emissions, ambient air quality, and deposition in the oil sands region
This article is part of the special series “A Decade of Research and Monitoring in the Oil Sands Region of Alberta, Canada.” The series documents the history of monitoring in the region and critically reviews a synthesis of monitoring results published in key environmental theme areas to identify patterns of consistent responses or effects; significant gaps in knowledge; and recommendations for improved monitoring, assessment, and management of the region.
This article includes online-only Supporting Information.
Abstract
This review is part of a series synthesizing peer-reviewed literature from the past decade on environmental monitoring in the oil sands region (OSR) of northeastern Alberta. It focuses on atmospheric emissions, air quality, and deposition in and downwind of the OSR. Most published monitoring and research activities were concentrated in the surface-mineable region in the Athabasca OSR. Substantial progress has been made in understanding oil sands (OS)-related emission sources using multiple approaches: airborne measurements, satellite measurements, source emission testing, deterministic modeling, and source apportionment modeling. These approaches generally yield consistent results, indicating OS-related sources are regional contributors to nearly all air pollutants. Most pollutants exhibit enhanced air concentrations within ~20 km of surface-mining activities, with some enhanced >100 km downwind. Some pollutants (e.g., sulfur dioxide, nitrogen oxides) undergo transformations as they are transported through the atmosphere. Deposition rates of OS-related substances primarily emitted as fugitive dust are enhanced within ~30 km of surface-mining activities, whereas gaseous and fine particulate emissions have a more diffuse deposition enhancement pattern extending hundreds of kilometers downwind. In general, air quality guidelines are not exceeded, although these single-pollutant thresholds are not comprehensive indicators of air quality. Odor events have occurred in communities near OS industrial activities, although it can be difficult to attribute events to specific pollutants or sources. Nitrogen, sulfur, polycyclic aromatic compounds (PACs), and base cations from OS sources occur in the environment, but explicit and deleterious responses of organisms to these pollutants are not as apparent across all study environments; details of biological monitoring are discussed further in other papers in this special series. However, modeling of critical load exceedances suggests that, at continued emission levels, ecological change may occur in future. Knowledge gaps and recommendations for future work to address these gaps are also presented. Integr Environ Assess Manag 2022;18:333–360. © 2021 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).
INTRODUCTION
Canada's proven oil reserves make up approximately 10% of the total global reserves, which is the third largest of any country (Natural Resources Canada, 2020). Of these proven reserves, the oil sands (OS) deposits account for 97% (Government of Alberta, 2017) and are spread over an area of approximately 142 200 km2 in three distinct regions: Athabasca, Cold Lake, and Peace River (Figure 1). As of 2019, half of the OS bitumen production came from surface mines and the other half from in situ methods (Alberta Energy Regulator, 2021). Currently, 88% of OS bitumen production occurs in the Athabasca Oil Sands Region (AOSR; Alberta Energy Regulator, 2021). Additional detail on development in the oil sands region (OSR) is presented in the introductory paper in this series (Dubé et al., 2021).
Atmospheric emissions are important to characterize as they can affect human health and ecosystems because they are transported, transformed, and deposited onto the environment. Most ambient air pollutant and deposition measurements in the OSR are conducted outside industrial operating areas. Effects may occur beyond the OSR, across provincial and territorial boundaries, where monitoring is less intensive.
Regional air monitoring in the OSR is delivered primarily by three multistakeholder airshed organizations (Figure 1) responsible for long-term ambient air monitoring: Wood Buffalo Environmental Association (WBEA) in the AOSR (Foster et al., 2019), Peace River Area Monitoring Program in the Peace River OSR, and the Lakeland Industry and Community Association in the Cold Lake OSR. These organizations collect monitoring data in accordance with Alberta's Air Monitoring Directive (Government of Alberta, 2019a) and are publicly accessible online (http://airdata.alberta.ca). Continuous air-monitoring stations that measure and report on air pollutant concentrations in near-real time are the backbone of these networks (see Figure 1). Many stations are located in communities and/or linked to specific monitoring requirements in the Environmental Protection and Enhancement Act industrial approvals (Government of Alberta, 2020). Airshed organizations also perform integrated air monitoring, which involves collecting samples periodically (e.g., every sixth day, monthly) on a sampling medium for subsequent laboratory analysis.
Additional information has been derived from focused studies, wherein more complex measurement technologies and platforms (e.g., airborne measurements) are used to investigate air concentrations and impacts. Measurements are supplemented by modeling studies used to estimate concentrations and deposition within and downwind of the OSR. The introductory paper in this series provides additional details on environmental monitoring in the OSR, as well as the history and evolution of the Oil Sands Monitoring Program (Dubé et al., 2021).
Considerable emissions, ambient air quality, and deposition monitoring data have been generated in the OSR. Although there have been integrated assessments of ambient air concentrations and deposition impacts on terrestrial ecological indicators (e.g., Davidson et al., 2020 and references therein), similar integrated assessments have not been conducted across the three OSRs or across all environmental media, hence our motivation for conducting the present review. Recent literature reviews that synthesize published findings cover the topics of polycyclic aromatic compounds (PACs; Harner et al., 2018), atmospheric deposition (Wentworth & Zhang, 2018), measurements of trace elements (TEs) and nonmetals in air and deposition in the AOSR (Huang et al., 2016), forest health effects caused by deposition (Davidson et al., 2020; Foster et al., 2019), air quality and odor events in the area of Fort McKay (Alberta Energy Regulator and Alberta Health, 2016), and a critical review of the main findings of Environment and Climate Change Canada (ECCC) work in support of the Oil Sands Monitoring Program (OSMP) (Brook et al., 2019). These publications are valuable in their syntheses of large amounts of scientific work and draw important conclusions regarding information gaps, priority issues for future work, and impacts.
This review paper builds on these efforts by broadening the scope to include all air pollutants and deposition to all ecosystems, regardless of the organization that led the work. Another novel aspect of this review is that it is part of a series that synthesizes the current state of knowledge across all media (i.e., air, aquatic, and terrestrial). A summary paper integrates results across air, water, terrestrial biological, and community-based monitoring themes to identify crosscutting, system-wide linkages, and pathways that are not immediately apparent within a single theme review (Roberts et al., 2021a). In addition to the novelty of this paper, the need for an integrated monitoring system in the OSR has been noted previously (e.g., Hopke et al., 2016; Swanson, 2019a, 2019b), which requires a comprehensive cross-thematic understanding of the current state of knowledge of the environment in the OSR. The issues of fate and ecological effects of air contaminants are largely captured in the surface aquatics (Arciszewski et al., 2021) and terrestrial biological monitoring (Roberts et al., 2021b) reviews, although key findings of ecological effects resulting from deposition are discussed briefly. The introduction (Dubé et al., 2021) and synthesis (Roberts et al., 2021a) papers also summarize topics relevant to multiple themes (e.g., cumulative environmental effects).
The main objectives of this paper are to (i) summarize the current body of knowledge of OSR atmospheric emissions and the processes of transport, transformation, and deposition; (ii) support the review and synthesis of other theme areas in this series for which air emissions might result in impacts (i.e., Arciszewski et al., 2021; Roberts et al., 2021b). This paper meets these objectives by (i) summarizing the methodology; (ii) synthesizing the state of knowledge of air emissions, contaminant pathways (i.e., transport, transformation, and deposition), and impacts; (iii) reaching a conclusion that highlights key knowledge gaps and recommendations for future work. One limitation of this review paper is that it largely considers peer-reviewed literature, which may not be the preferred reporting source for some pertinent information (e.g., community-based monitoring, evaluation of compliance monitoring data). In addition, the magnitude and variety of topics considered are challenging to synthesize in a concise yet representative manner.
MATERIALS AND METHODS
This review series synthesizes the substantial collection of peer-reviewed articles published between 2009 and 2019 on environmental monitoring activities and research studies in the OSR. Some more recent literature was included for specific topics and issues. In addition to the peer-reviewed literature, there were relevant technical reports, although this information lacks the independent evaluation inherent in a peer-reviewed journal. As a result, these reports were not included in the publication count statistics.
Conceptual models adapted from Swanson (2019a, 2019b) for each theme area were used to achieve a coherent multithemed technical review series, and are explained in detail in the introductory paper (Dubé et al., 2021). Briefly, a conceptual model visualizes simplified elements of the complex OS-environmental system as a box-and-line diagram representing physical and chemical pathways. These models are organizational diagrams that quickly summarize if or how specific effects are linked to specific causes (Figure 2). There are four main categories in the conceptual model used here: emission source (pressure), air pollutant (stressor), transport-transformation-deposition (pathway), and impact (response; European Commission and Eurostat, 1999). Given that emission source, air pollutant, transport, transformation, deposition, and impact are more common terms in the atmospheric literature, we have opted to use these terms throughout the manuscript while recognizing they are analogous to the pressure, stressor, pathway, and response vernacular used elsewhere in this review series. Connections between pressures, stressors, pathways, and responses were determined within individual monitoring themes and then linked between themes for the system as a whole in the overall synthesis (Roberts et al., 2021a). Ultimately, this broad-scale view of the current state of knowledge can be used to guide future monitoring and research activities in the OSR (Roberts et al., 2021a).
Emission sources are divided according to whether they are OS-related sources or non-OS sources. The OS-related sources are further categorized as (i) stacks (inclusive of upgrader and flaring stacks), (ii) off-road vehicle exhaust (e.g., diesel heavy hauler and shovel fleet emissions), (iii) mine faces, (iv) tailings ponds, (v) fugitive dust (e.g., unpaved roads, land clearing, petroleum coke (petcoke) piles, exposed tailings, bare surfaces, overburden soil piles, limestone quarry and crushing operations), and (vi) fugitive plant or processing gaseous emissions (e.g., leaks, venting). Non-OS sources include (i) “other” anthropogenic sources in the OSR (e.g., on-road vehicles, aircraft, limestone quarries, and residential heating, particularly in the urban service area of Fort McMurray [pop. ~75 000]); (ii) wildfires and prescribed burning activities; (iii) long-range transport of pollutants that originate outside the OSR from provincial, national, or global sources and are transported into the OSR; (iv) natural landscapes (e.g., biogenic emissions, volatilization from lakes and soils).
The pollutants are grouped by contaminant: sulfur dioxide (SO2), nitrogen oxides (NOx = NO + NO2), and other oxidized forms of nitrogen (NOz; e.g., nitric acid [HNO3]), ammonia (NH3), particulate matter (PM), PACs, volatile organic compounds (VOCs), reduced sulfur compounds (RSCs), mercury, ozone (O3), and greenhouse gases (GHGs). There is some inherent overlap in the stressor categories (e.g., PM can contain PACs and mercury).
Due to the complexity of PM emissions, formation, and composition, as well as the differing impacts of PM components, PM is further divided into five subcategories: base cations (BCat); sulfate/nitrate/ammonium (SNA; comprised of particulate sulfate [pSO42−], nitrate [pNO3−], and ammonium [pNH4+]); TEs; primary and secondary organic aerosol (POA/SOA); and black carbon (BC). For example, BCat and SNA are linked to acidification and eutrophication; BC affects radiative forcing; TEs can elicit toxic effects.
The transport and transformation of pollutants include both gas-phase and aqueous-phase reactions. Impacts that are known or suspected are grouped into ambient air quality, odor events, and ecosystem responses. Although odor is usually included in the term “air quality,” it is treated separately here due to frequent odor events and concerns in communities surrounding OS facilities. The number of papers covering each topic in this review is listed on the conceptual model to indicate which components have received relatively more attention in the literature.
RESULTS AND DISCUSSION
Of the ~400 papers reviewed for this report series, 136 peer-reviewed publications were identified within the air theme. As shown in Figure 2, each paper was classified into five principal categories: Emissions (72 papers), Transport and Chemical Transformations (83 papers), Ambient Air Quality (26 papers), Odor Events (1 paper), and Deposition (56 papers). Of the 136 papers in this theme, 130 papers focused on the AOSR; hence most of the discussion below pertains to the AOSR unless specifically stated otherwise. Only six and eight papers included the Peace River and Cold Lake OSRs, respectively; however, these papers typically used a broad spatial analysis that covered all three OSRs as opposed to specifically assessing the Peace River or Cold Lake OSRs.
Although the number of papers published indicates the amount of research taking place under a given topic, it does not necessarily reflect the extent to which a consensus has appeared in the scientific community. Furthermore, we recognize that categorizing papers according to these topics is subjective, although it still provides a rough indication of level of scientific “effort.” The categorization of specific papers is provided in the Supporting Information, along with a list of acronyms used throughout this paper. The results and discussion are organized into sections that align with Figure 2: emissions, transport-transformations-deposition, and impacts.
Emissions
Emissions related to bitumen extraction and processing are complex, due to the complexity of the bitumen resource and different processes used by the various facilities. Figure 2 summarizes major OS-related and non-OS emission source categories. It is challenging to quantitatively compare emissions between various sources, because it depends on the spatial and temporal scales that are chosen, and is affected by the uncertainty in many emission rates. However, Figure 2 attempts to synthesize the body of knowledge to date on emissions in the AOSR by indicating the relevant sources of each pollutant in the region. There is little evidence that non-OS anthropogenic sources within ~100 km of the surface-mining facilities are substantial, with the exception of fugitive dust emissions from the Hammerstone Quarry, as well as the contribution of emissions in the town of Fort McMurray itself (e.g., on-road vehicles) to certain pollutants (e.g., VOCs, NOx) monitored in Fort McMurray (Bari & Kindzierski, 2018). The Emissions section first synthesizes information on OS-related sources based on methodology (i.e., emission inventories, top–down emission estimates, ground-based source emission testing, and receptor-based source apportionment). There is a comparatively small number of studies utilizing isotopic measurements to investigate sources; hence these are briefly summarized in the Supporting Information. As noted above, from a regional perspective, emissions from non-OS sources are typically much less then OS-related emissions; therefore, non-OS emissions are also summarized in the Supporting Information.
Emission inventories
Emission inventories quantify the emission rates of air pollutants on varying timescales (typically annually) and with varying degrees of spatial resolution. Inventories usually provide emission rates at the facility or sector level for each pollutant. The Supporting Information details three common inventories developed for Canada: the National Pollutant Release Inventory (NPRI), Air Pollutant Emissions Inventory (APEI), and Greenhouse Gas Reporting Program (GHGRP).
A comprehensive and accurate emission inventory will improve deterministic chemical transport models (CTMs), which provide insight into pollutant transport, chemical transformation, and deposition. Chemical transport models provide the ability to evaluate the impact of new emissions or estimate the benefits of various emission reduction scenarios to air quality and environmental fate (Zhang et al., 2018). The reliability of CTM simulations is directly related to the reliability of model input parameters, such as temporally and spatially resolved emission inventories.
There has been considerable effort (72 papers) on improving our understanding of emissions in the OSR. Zhang et al. (2018) summarized a multiyear effort to improve emission data for NOx, VOCs, SO2, NH3, carbon monoxide (CO), PM2.5, PM <10 μm in mass median aerodynamic diameter (MMAD; PM10), and mercury in the AOSR. The 100 × 100 km study area included the major surface-mining facilities as well as two in situ operations. The analyses included multiple emission inventories (national, provincial, and subprovincial), as well as continuous emissions monitoring system (CEMS) data for SO2 and NOx. A novel approach using GIS shapefiles and satellite images improved the spatial allocation of emissions in a given facility. Aircraft observations were used to evaluate and revise pollutant emission rates for several facilities. The result was several detailed hybrid emission inventories that provided a better understanding of emission sources in the AOSR, including non-OS sources (e.g., biogenic, on-road mobile). These hybrid inventories were used in subsequent regional model simulations to estimate downwind concentrations (e.g., Akingunola et al., 2018; Russell et al., 2019; Stroud et al., 2018) and potential ecosystem impacts (Makar et al., 2018).
Mercury emissions from reported inventories were adjusted to also include off-road vehicles and were used in subsequent efforts to estimate the relative impact of human versus natural sources on mercury concentrations and deposition in Canada (Fraser et al., 2018). However, mercury emissions from OS facilities remain poorly characterized. Deposition patterns of mercury and methylmercury in snowpack samples confirm mercury emissions from OS facilities. Gopalapillai et al. (2019) reported local enhancement factors (i.e., ratio of snowpack loadings at near-field sites relative to reference sites) of 5 and 6, for total mercury and methylmercury, respectively. Lynam et al. (2015) reported mercury enrichment factors, relative to crustal material, in wet-deposition samples collected in Fort McMurray ranging from 130 to 245, implying noncrustal mercury sources in the region. A subsequent analysis (Lynam et al., 2018) found that rainfall collected during a wildfire smoke event resulted in a 5–24 times increase in mercury wet deposition. That being said, long-range transport of mercury is also a known contributor to ambient mercury levels and deposition in the region.
Figure 3 shows the total annual emissions for SO2, NOx, NH3, and PM10 at 2.5 × 2.5 km resolution for Northern Alberta, which includes the three OSRs and the area surrounding Edmonton (pop. ~1 000 000). Analogous figures for CO, total VOCs, and PM2.5, as well as a description of emissions by category, are in the Supporting Information. These emission inventories were adapted from Zhang et al. (2018) and represent 2015 reported emissions for Canada, as well as 2018 CEMS data for large stack sources in Alberta, and do not include wildfire emissions. Figure 3A shows the dominance of large point sources for SO2 in Northern Alberta, including sources that emit several hundred tons of SO2 every year in the southern AOSR and Cold Lake OSR. Figure 3B shows that NOx sources in the surface-mineable AOSR are confined to a relatively small geographical area, whereas NOx emissions in the Peace River and Cold Lake OSRs are more diffuse. The emission inventory for NH3 (Figure 3C) shows the dominance of agricultural emissions in central and northwestern Alberta, which are likely too far away to be relevant to the surface-mineable region, but likely affect NHx (sum of NH3 and NH4+) deposition in the Cold Lake and Peace River OSRs. Finally, Figure 3D shows the large impact that surface mining has on PM10 emissions relative to in situ facilities. This is important because fugitive dust deposition is a substantial vector for BCat and TEs to ecosystems, and BCat has been shown to mitigate the effects of acidic deposition within approximately 10–20 km of surface-mining facilities (see sections Atmospheric Deposition and Ecosystem Responses). Collectively, these figures can help inform monitoring network design by identifying areas of high emissions.
There are currently 32 PACs and four PAHs reported to the NPRI and APEI, respectively, although alkylated PAHs (alk-PAHs) and dibenzothiophenes (DBTs) are not included in either, with the exception of 1-nitropyrene, which is reported to the NPRI. Qiu et al. (2018) improved existing PACs emission inventories for the OSR by including tailings ponds, mine faces, mine fleets, and point sources for PAHs. However, only emissions from tailings ponds, mine fleets, and transportation sources were considered for alk-PAHs, DBTs, and alkylated DBTs (alk-DBTs). The updated PACs emission inventories, when used in a dispersion model, provided much better agreement with most ambient PACs air concentrations near the center of mining activities, although downwind concentrations continued to be underestimated (Qiu et al., 2018). Widespread underestimates in alk-PAHs and DBTs remained, implying that these emission inventories are still missing relevant alk-PAH and DBT sources and/or atmospheric processes.
Numerous studies have focused on improving GHG emission inventories by using life-cycle assessments to estimate emission intensities associated with various OS fuel-production pathways, such as mining or in situ; bitumen or synthetic crude oil (Charpentier et al., 2009; Englander et al., 2015; Nimana et al., 2015). Substantial differences between emission intensities were observed as a result of variability in the choice of activities included or excluded and limited reliability of emission data.
The Government of Alberta recently conducted an analysis of GHG emissions from OS facilities that release more than 50 000 t CO2e (CO2 equivalents) annually. This represents 95% of total reported OS emissions from the three OS regions from 2011 to 2018 (Government of Alberta, 2019b). This analysis excluded emissions released during activities occurring outside facility boundaries such as transportation, refining, or combustion. From 2011 to 2017, there was a 19% decrease in total reported OS GHG emission intensity. However, although emission intensity has decreased, expanded development and growth in OS production has resulted in a 42% increase in overall reported CO2e emissions from 45 Mt in 2011 to 64 Mt in 2017. In Canada, oil and gas activities from all industrial and nonindustrial sources contribute 26% of nationally reported GHG emissions, with OS activities comprising 11% of national GHG emissions as of 2018 (Environment and Climate Change Canada, 2020).
Top–down emission estimates
Measurements from airborne platforms allow a top–down emission quantification approach that can be used to evaluate and improve bottom–up emission inventories, particularly for emissions that are calculated using emission factors or engineering estimates. These approaches are complementary, because the former may identify unaccounted or unknown sources, whereas the latter can provide more complete temporal coverage for some of the approaches used (e.g., CEMS).
One intensive aircraft campaign occurred during a six-week period in August–September 2013. A second, the data from which are still under analysis, took place in April and June of 2018. These campaigns, led by ECCC, measured an extensive suite of airborne pollutants, including CO, CO2, SO2, NOx, CH4, PM size and composition, VOCs, and NH3. The 2013 study had the following goals: (i) measure and quantify air emissions from OS mining facilities, (ii) assess atmospheric transport and chemical transformation of emitted pollutants, (iii) provide measurements for satellite comparisons, and (iv) improve prediction capabilities of the Global Environmental Multiscale-Modelling Air quality and Chemistry (GEM-MACH) CTM (Gordon et al., 2015). The model was also used during both measurement campaigns to provide real-time air pollution forecasts to aid in flight planning.
A mass-balance algorithm, Top–down Emission Rate Retrieval Algorithm (TERRA), was used to calculate hourly emission rates for several surface-mining facilities using the 2013 data (Gordon et al., 2015). Reported uncertainties for hourly emission rates from TERRA are <30%, with extrapolation to the surface typically being the largest contributor for nonelevated sources. Results from TERRA were used to evaluate and improve emission inventories (e.g., Zhang et al., 2018). However, in order to compare the hourly TERRA-derived emission estimates with reported emission inventories, it is necessary to “upscale” these hourly emissions to annual emissions. This approach varies by pollutant as a result of different sources and available information.
Gordon et al. (2015) found TERRA-derived SO2 emissions were within 10% of industry reported hourly CEMS SO2 data, supporting the use of TERRA in estimates of emissions of chemicals for which direct observations are not available. Subsequent work comparing SO2 top–down emission estimates derived from satellite data with surface concentrations and reported emissions after 2013 differed up to a factor of 2 (McLinden et al., 2021). The satellite-derived emission estimates demonstrate slight increases (<20%) in SO2 emissions from 2013 through 2017, in agreement with SO2 concentrations measured by air-monitoring stations. However, reported emissions reveal a 50% decrease between 2013 and 2018 (McLinden et al., 2021). The authors, in consultation with industry, were unable to identify the cause(s) of the discrepancy.
Hourly emissions of CH4 calculated using TERRA found that, on average, the largest sources were tailings ponds (45%) and open-pit surface mines (50%), with the remainder being fugitive emissions from facilities (Baray et al., 2018). The Syncrude Mildred Lake Settling Basin was responsible for most of the tailings pond CH4 emissions from the five facilities studied. When the TERRA-derived emissions were upscaled and compared with the 2013 reported facility emission rates, the former was 48% ± 8% larger, highlighting the uncertainty in the CH4 emission inventory and/or the assumptions used for upscaling. Liggio et al. (2019) used CO2 data from the same study to derive annual CO2 emissions and intensities for four surface-mining facilities in the AOSR. Aircraft-derived emission intensities were 13%, 36%, 38%, and 123% greater than reported for the facilities. Collectively, the TERRA-derived annual CO2 emissions were found to be 64% greater than publicly reported emission inventories for these four facilities.
Hourly TERRA-derived emission rates for PM2.5 from six different facilities were scaled to two-month emission rates by Zhang et al. (2018). For five of these facilities, the two-month emissions were between 1.5 and 5 times greater than annually reported emissions of PM2.5, suggesting a large underestimation of PM2.5 emissions in reported inventories. Most of the measured PM2.5 (65%–95%) were 1.28–2.56 μm, suggesting that most of the emitted PM2.5 mass originates from fugitive dust, because other sources (e.g., combustion) tend to produce particles with smaller diameters. A similar approach was used to calculate annual BC emission rates from OS surface mining to be 707 ± 117 t a−1, which is within 16% of reported BC emissions (Cheng et al., 2019). However, the aircraft-derived estimates suggest that 73% of BC is emitted from off-road vehicles, whereas the reported inventory indicates roughly equivalent contributions from off-road vehicles and stacks.
Similarly, Li et al. (2017) reported TERRA-derived emission rates of 69–89 nonbiogenic VOCs from four different OS surface-mining facilities and found that the ∑VOC emission rate was 2–5 times higher than reported in the NPRI. The chemical profile of VOC emissions varies between the facilities and reflects differences in plant processes (e.g., the use of naphtha versus paraffinic solvents in bitumen–sand–water separation) and subsequent tailings pond emissions. Understanding the different VOC chemical profiles may aid in air pollutant source apportionment efforts in nearby communities. TERRA-derived VOC emissions from Li et al. (2017) were incorporated into the emission inventory used by GEM-MACH, and generally resulted in improved model-measurement agreement for within plume VOC and organic aerosol (OA) concentrations (Stroud et al., 2018). This study, and others referenced, highlight the utility of using top–down estimated emissions to evaluate and bound bottom–up reported emission inventories.
Emissions of 18 gas-phase low-molecular-weight organic acids (LMWOA; ~C1–C10) were calculated with TERRA using the 2013 data. Approximately 12 t day−1 of LMWOA were emitted directly from OS surface-mining facilities, primarily from off-road diesel vehicles in open-pit mines (Liggio, Moussa, et al., 2017). However, this was a relatively small fraction of the LMWOA formed downwind of OS facilities caused by photochemical reactions (up to 184 t day−1). Similarly, isocyanic acid (HNCO) emissions from OS facilities were estimated to be 6.2 ± 1.1 kg h−1, again mostly from off-road diesel vehicles (Liggio, Stroud, et al., 2017). This was much smaller than the 116–186 kg h−1 estimated to be formed downwind from photochemical oxidation of diesel exhaust. The top–down estimates allow the determination of the relative importance of direct emission versus secondary formation for key pollutants.
Atmospheric measurements from a single flight around the Syncrude Mildred Lake and Suncor Millennium facilities in 2008 were used to calculate fluxes of NOx, NOy, SO2, pSO42−, OA, PM mass, BC, CO, and CO2 from both facilities (Howell et al., 2014). Measurements of 76 VOCs from the same flight were used to calculate enhancement factors in OS plumes above background levels (Simpson et al., 2010). These data were analyzed using linear least-squares fit to identify two source groups for the measured pollutants: (i) direct evaporative emissions from OS themselves and/or diluent, and (ii) emissions from the mining processes (e.g., petcoke combustion, bitumen upgrading).
These top–down emission studies often yield observation-based emission estimates that are higher than reported values, although aircraft-based emission estimates of SO2 and NOx were generally in good agreement with CEMS data during this period. Top–down emission estimates for chemicals without CEMS data (e.g., VOCs, CH4, PM, CO, CO2) were, in general, higher than the bottom–up reported values in the inventories. This may reflect the indirect means used to estimate these emissions in bottom–up inventory reporting.
Ground-based source emission testing
Many studies have also measured or calculated emissions from specific OS sources using on-the-ground approaches. Similar to the airborne top–down approaches, these studies provide limited temporal coverage. Although these measurements are usually more difficult to compare with reported emission inventories, they provide a detailed assessment of individual sources. For instance, Small et al. (2015) reviewed modeled and measured emission rates of VOCs, CO2, CH4, and RSCs from tailings ponds to derive emission factors. The authors identified the need to better understand the effects of microbial activity and ice melt on emissions, as well as more studies utilizing methods that better account for spatial and temporal variability. More recently, Zhang et al. (2019) measured CH4 and CO2 fluxes from a tailings pond in the AOSR using eddy covariance over several weeks during summer. Fluxes of CH4 and CO2 were larger during the nighttime and daytime, respectively. Furthermore, CH4 fluxes measured concurrently using a flux chamber approach yielded results that were an order of magnitude higher than the eddy covariance approach. This discrepancy could be attributed to the small footprint (<1 m2) of flux chamber measurements that can yield large variability between measurements (Zhang et al., 2019). This is in contrast to a five-week study that compared tailings pond CH4 emissions, calculated using three micrometeorological flux methods, with traditional flux chamber measurements (You, Staebler, et al., 2021). CH4 fluxes calculated using the eddy covariance approach agreed with inverse dispersion model fluxes (±30%) and gradient fluxes (±18%), whereas flux chamber measurements were 64% lower. Fluxes derived from a larger footprint of micrometeorological measurements result in more robust and representative emissions (You, Staebler, et al., 2021). During the same study, NH3 fluxes (0.05 g m−2 day−1) and total alkane fluxes (1.05 g m−2 day−1) were measured using inverse dispersion modeling (You, Moussa, et al., 2021). If these fluxes are upscaled assuming no variation across seasons (which is unrealistic given that ponds freeze during winter), then they equate to annual emissions of 42 and 881 t a−1 for NH3 and total alkanes, respectively. However, NH3 emissions have not been previously reported for this pond, and extrapolated annual NH3 emissions far exceed the 0.82 t a−1 NH3 emissions reported to the NPRI for the entire facility (You, Moussa, et al., 2021). Measurements during other seasons are needed to understand the seasonality of NH3 emissions from ponds. Additional results from the 2017 sampling campaign are still being analyzed to quantify fluxes of VOCs, TRS, and PACs.
Tailings ponds are a source of PAHs to the atmosphere (Galarneau et al., 2014; Moradi et al., 2021; Parajulee & Wania, 2014). Multimedia fate modeling demonstrated that PAH concentrations in several environmental compartments could not be reproduced using reported emissions, and it was suggested that unreported releases from tailings ponds were partially responsible (Parajulee & Wania, 2014). Fluxes based on two-film theory using concentrations of 16 PAHs measured in a tailings pond and air confirmed this finding (Galarneau et al., 2014). When extrapolated to the entire AOSR, these tailings pond emissions were found to be almost five times greater than all combined AOSR emissions reported to the NPRI (Galarneau et al., 2014). Most recently, co-located measurements of 6 PAH and 21 alk-PAH concentrations in air and water at a different pond were assessed (Moradi et al., 2021). Three flux calculation methods each revealed that PAHs were released to the atmosphere from this particular pond (Moradi et al., 2021). Extrapolating this to other ponds also receiving water from naphtha recovery units revealed that phenanthrene and benzo[a]pyrene from tailings ponds would contribute approximately 36% and 0.25%, respectively, of the total fugitive emissions (including mine fleet, fugitive dust, and stacks; Moradi et al., 2021).
The PM emissions were measured from three upgrader stacks in the AOSR using a dilution sampling system over several hours across 2–3 days for each stack (n = 6–7 samples per stack). Most PM was in the fine mode (PM2.5) and was between 54% and 91% fully neutralized ammonium sulfate in two of the stacks, whereas PM2.5 in the third stack was 48% of sulfuric acid. The TEs and PAHs were detectable but relatively low. In a separate stack study, the isotopic composition of pNH4+, pNO3−, and pSO42− was measured in the emission stream from two industrial stacks within the AOSR, in addition to the isotopic signatures of source materials and upgrading by-products (Proemse, Mayer, Chow, et al., 2012). These stack measurements are useful for the development of source profiles for source apportionment modeling (Landis et al., 2012; Landis, Studabaker, Pancras, Graney, White, et al., 2019; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019). However, for PM, the bulk of the emissions originate in surface-mining operations and other sources of fugitive dust release (Zhang et al., 2018).
Measurements of CO2, CH4, CO, nonmethane hydrocarbons (NMHC), NOx, PM2.5 mass, and BC from diesel heavy haulers used in the OS surface mines were made using on-board portable monitoring systems (Wang et al., 2016). Emission factors for each pollutant were derived through 16 different tests under a variety of operating conditions (e.g., idling, traveling with a load). Individual pollutants exhibited different behavior during the various operating cycles. For example, emissions of NOx and ultrafine particles were higher for engine idling than for travel, whereas emissions of PM2.5 and BC were higher for travel than for idling. The PM2.5 was 10%–40% organic carbon and 45%–60% BC. Emissions of SO2, hydrogen sulfide (H2S), and NH3 were reported to be negligible (Wang et al., 2016). Reviewing diesel emissions calculated using established emission factors revealed large variability between facilities, attributable to different calculation methods. Overall, the calculated emissions of CO, NMHC, and PM2.5 were higher than the measured values; calculated NOx values were both above and below the measured values. This work reveals that the characteristics of diesel emissions are affected by several factors and are not static, even for the same vehicle.
Core samples from undeveloped mine regions around the AOSR were used to derive fugitive emission factors for CH4 and CO2 from mine faces (Johnson et al., 2016). Using mined OS values in 2015, annual emissions were roughly 38–58 kt of CH4, and 0.2 Mt of CO2 (Johnson et al., 2016). For comparison, Baray et al. (2018) extrapolated TERRA results to estimate 2013 annual CH4 emissions of ~170 kt from all surface-mining sources (~85 kt of which are fugitive emissions from the mine faces). The values in Johnson et al. (2016) are from core samples in regions that have not yet been developed, and so the emission estimates may exclude the impact of physical and chemical processing of bitumen.
The estimates of fugitive PM emissions are affected by highly stochastic mechanical processes, surface conditions, variable composition and size distributions of source materials, and meteorological conditions. In particular, chemical composition and size distribution have serious consequences for transport, deposition rates, and environmental impact. One study characterized PM emissions from 64 sites, including OS mining facilities, quarry operations, and paved and unpaved roadways, using a portable in situ wind erosion laboratory (Wang, Chow, Kohl, Yatavelli, et al., 2015). Unpaved roads and bare land with large amounts of loose clay and silt were the highest emitting sites. Surface watering reduced dust emissions by 50%–99%, and surface disturbances (e.g., vehicular traffic) increased PM10 emission potential 9–160 times. These results reinforce the challenges for estimating PM10 emissions. Another study by Wang, Chow, Kohl, Percy, et al. (2015) characterized the chemical composition of 27 different samples from representative locations in the AOSR. Chemical profiles were grouped into six categories for use in emission inventories: paved road dust, unpaved road dust near and far from OS operations, overburden soil, tailings sands, and forest soils. These chemical profiles were incorporated into the updated emission inventories developed by Zhang et al. (2018). Using these updated emission inventories, Akingunola et al. (2018) found that modeling 12-bins of PM size instead of 2-bins reduced model bias in GEM-MACH by 32% (from −2.62 to −1.72 μg m−3) when compared with surface and airborne PM2.5 observations.
Receptor-based source apportionment studies
Source apportionment studies use statistical approaches (e.g., positive matrix factorization [PMF], chemical mass balance [CMB], and Unmix models) to determine source factors and quantitative source contribution estimates based on variations in chemical composition through time or space (US Environmental Protection Agency, 2014). Studies that have used PMF, Unmix, or the CMB model to quantitatively apportion PM2.5 mass, PM10 mass, PACs loadings, or TE loadings in samples collected in the OSR are summarized in Figures 4A, 4B, 5A, and 5B, respectively. Additional details of each study are provided in the Supporting Information. Source factors were grouped according to similarity in Figures 4 and 5 to synthesize findings between studies. However, source factors with similar or identical names reported in different studies are not necessarily directly comparable, as a result of different sampling periods, locations, measurement techniques, pollutants considered, and/or statistical techniques.
Figure 4A indicates bitumen upgrading emissions that form secondary PM2.5 constitute between 21% and 59% of PM2.5 at air-monitoring sites near Fort McKay and Fort McMurray (Bari & Kindzierski, 2017; Landis et al., 2017; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019; Phillips-Smith et al., 2017). Fugitive dust (reported range of 30%–46%) and biomass burning and combustion (0%–26%) also contribute to PM2.5, whereas regional or long-range transport consistently contributed a <10% fraction of total PM2.5 mass. There are fewer studies for coarse PM (PM10–2.5); however, coarse PM in Fort McKay is overwhelmingly dominated (83%–95% contribution) by fugitive dust (Landis et al., 2017; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019). These findings are consistent with fugitive dust emission measurements reported by Wang, Chow, Kohl, Yatavelli, et al. (2015) and lichen biomonitoring results (Landis et al., 2012; Landis, Studabaker, Pancras, Graney, White, et al., 2019).
Figure 5A shows PMF and CMB results for percent contribution to ∑PAHs or ∑PACs loadings calculated using lichen (Landis, Berryman, et al., 2019) and moss or peat samples (Zhang et al., 2016) collected throughout the AOSR. Contribution from fugitive dust ranged from 34% to 69% throughout the AOSR, and increased to 70% (∑PAHs) and 82% (∑PACs) at sites <25 km from OS facilities. Petcoke dust (26%–54%) represented the largest fugitive dust component, which is consistent with studies that found PACs in large (10–100 μm) petcoke and ore particles using scanning electron microscopy (Jariyasopit et al., 2018; Zhang et al., 2016) and two-dimensional gas chromatography coupled with mass spectrometry (Manzano, Muir, and Marvin, 2016; Manzano, et al., 2017) on snowpack and passive air samplers. Recent work has quantified nitrogen-containing PACs (NPACs) in snow, lake sediment, and passive air samplers, as well as petcoke, haul-road dust, and unprocessed oil sands (Chibwe et al., 2019). Concentrations of NPACs in snow, lake sediment, and air decreased with increasing distance from surface-mining facilities until ~50 km away. Profiles of NPACs in environmental samples were most similar to petcoke samples, also suggesting this is a source and that NPACs could be used as source indicators in future studies. As shown in Figure 5A, biomass combustion (6%–15%) and stack emissions (5%–22%) were also identified as source factors in these two studies.
Hsu, and Harner, Li, et al. (2015) reported 24-h averaged atmospheric concentration of 22 PAHs sampled at four community sites in the AOSR in 2012 and 2013. Three of the community sites had peak PAH values associated with NO, NO2, PM2.5, and SO2, which implies combustion sources. However, the fourth site (~30 km SE of Fort McMurray) exhibited strong seasonality, suggesting nearby temperature-dependent emissions were the main driver of PAH air concentrations at that site (e.g., volatilization from water bodies, re-emission of previously deposited PAHs).
These PMF and CMB results for PACs are inconsistent with the emission inventory developed by Qiu et al. (2018), who reported ~75% of regional PAHs originating from point sources. The apparent discrepancy is likely because fugitive dust was not a source category included in Qiu et al. (2018), and possibly caused by other inventory data uncertainties. Although analyses of PACs profiles have been undertaken for some source material (e.g., oil sand, diluted bitumen, mine overburden, unpaved roads, tailings pond dikes, coke storage piles, diesel exhaust, upgrader stacks, flue gas desulfurization stacks), future analyses of fugitive dust sources (e.g., limestone and quarry stockpiles, roads, and exposed surfaces in tailings facilities) can help bridge the gap between emission inventories and source apportionment studies by improving source profiles (Wang, Chow, Kohl, Yatavelli, et al., 2015; Watson et al., 2012, 2014; Watson, Chow, Wang, Kohl, et al., 2013; Watson, Chow, Wang, Lowenthal, et al., 2013; Watson, Chow, Wang, Zielinska, et al., 2013; Yang et al., 2011). In addition, a standard list of target analytes and a reference analytical methodology will help improve the comparability of various studies (Studabaker et al., 2017).
As shown in Figure 5B, Landis, Berryman, et al. (2019) reported TE mass from lichen samples using PMF. The contribution of fugitive dust from petcoke piles (11%), haul roads (36%), and raw oil sands (31%) were substantial at sites within 25 km of production facilities (n = 56), with contributions from other factors being smaller (<6% each). When all lichen sampling sites are considered (n = 127), the source factor contributions were petcoke dust (7%), haul-road dust (34%), and oil sands dust (20%), with remaining factors being <13% each.
A hybrid source analysis (i.e., multiple linear regression, aluminum enrichment factors, and PMF) of snowpack samples collected in 1978, 1981, 2008, and 2011–2016 identified source categories of TEs (Gopalapillai et al., 2019). The two largest bitumen producing facilities were found to be key contributors of many TEs (e.g., V, Al, Ti, W, Ga, Fe, Be, Cs, Co, Mo, Rb, Pb, As, U, Ba, Y, Ce, and La). Fugitive dust, particularly from petcoke piles and roads, was a major contributor of these elements, with the exception of V and W (Gopalapillai et al., 2019). Near-field snowpack loadings of V, Al, and Ti decreased from 1978 to 2016, although were still an order of magnitude larger than snowpack loadings at reference sites in the Peace Athabasca Delta. Positive matrix factorization was used to analyze snowpack samples collected in 2015 (n = 154) and resulted in four factors: two types of road dust, road salt, and raw oil sand dust, although the authors did not report source factor contributions to total TE loadings. Delayed petcoke was not identified as a source factor because petcoke contains a relatively low percentage of TEs (Gopalapillai et al., 2019; Landis, Berryman, et al., 2019).
Concentrations of TEs were measured in cranberries, lingonberries, and blueberries at sites both near and far from OS sources in the AOSR (Stachiw et al., 2019). Dust particles were observed on the surface of berries using scanning electron microscopy. The TEs that exhibited strong correlation (e.g., Al, Cr, Pb, U, and V) with Y, a lithophilic element, were inferred to originate from dust deposition, whereas TEs without a strong correlation were assumed to be from plant uptake via roots (e.g., Ba, Cd, Cu, Mn, Mo, Ni, Rb, Sr, and Zn). The TEs correlated with Y were 2–24 times more abundant in AOSR sites relative to sites hundreds of kilometers away. The concentrations of these TEs were reduced after washing the berries with water (Stachiw et al., 2019).
Pathways
Transport and transformation
Atmospheric transport of pollutants depends on emission height, plume buoyancy, meteorology, topography, and characteristics of the pollutant (e.g., reactivity, solubility, particle size). Transport can be inferred by on-the-ground measurements, although this requires a spatially extensive measurement network, which is challenging for remote areas like the OSR. Satellite or airborne measurements also provide information to evaluate transport for a number of air pollutants (e.g., NO2, SO2, CO). In addition, CTMs simulate the effects of topography, meteorology, and pollutant reactivity and solubility on transport, and so may be used to infer transport. Similar to the literature examining OSR emission sources, most transport, transformation, and deposition papers to date focus on the surface-minable AOSR. This section deals with measurements and modeling results that reveal the extent of pollutant transport as well as spatial patterns of air pollutants, which are, unsurprisingly, highly similar to those for deposition.
In the AOSR, the WBEA operates a network of several dozen passive air samplers throughout the region to measure monthly (or bimonthly) averaged concentrations of NH3, HNO3, NO2, SO2, and O3. Hsu et al. (2016) reported higher NO2 and SO2 in winter, presumably the result of decreased vertical mixing and photochemistry, as well as increased NOx emissions resulting from the heating of buildings. NH3 and HNO3 were higher in summer due to temperature-dependent emissions (NH3), wildfire emissions (NH3), and photochemistry (HNO3). Air concentrations of NO2, SO2, and NH3 were enhanced within 50 km of OS facilities, whereas HNO3 was enhanced at a larger spatial extent because it is a transformation product. Enhancements of NH3 were also observed in the southern AOSR, likely as a result of agricultural, urban, and/or wildfire emissions. Edgerton et al. (2020) reported long-term average concentrations ranging from 0.30 to 2.76 ppb (SO2), 0.17 to 5.78 ppb (NO2), 0.19 to 1.1 ppb (NH3), and 0.07 to 0.15 ppb (HNO3) across the different sites.
All air pollutants revealed a pattern that follows the Athabasca River Valley, with enhanced concentrations extending further in the north–south direction relative to east–west. Ambient air concentrations were factors of 8, 20, 4, and 3 higher adjacent to facilities than the edges of the monitoring network for SO2, NO2, HNO3, and NH3, respectively (Edgerton et al., 2020). The seasonal trends and spatial patterns reported by Hsu et al. (2016) were consistent with earlier studies analyzing data from the same passive sampler network (Bytnerowicz et al., 2010; Hsu, 2013). Passive SO2 measurements revealed statistically significant (p < 0.05) negative trends since 2000 at 18 of 30 sites (Edgerton et al., 2020). On the other hand, NO2 trends were only statistically significant at two sites (one negative and one positive). Concentrations of O3 were lower closer to the OS facilities as a result of titration by NO (Hsu, 2013). The same seasonal trends for SO2, HNO3, and NH3 were reported by Hsu and Clair (2015), who analyzed hourly measurements of water-soluble gases and ions in PM2.5 using an Ambient Ion Monitor-Ion Chromatograph system in Fort McMurray. Comparison of PM2.5 and PM10 filter data from four sites in the AOSR revealed that >80% of SO42− was in the PM2.5 fraction and >60% of the Ca2+, Mg2+, Na+, and Cl− were in the PM10–2.5 fraction (Edgerton et al., 2020). Monthly averages revealed that Ca2+ peaked in March/April to November, whereas pSO42−, pNO3−, and pNH4+ peaked from November to March (Edgerton et al., 2020).
Spatial patterns of NO2 and SO2 described above are consistent with satellite observations that demonstrated distinct enhancements in a ~30 × 50 km area encompassing the surface-mining operations in the AOSR (McLinden et al., 2012). Follow-up studies used improved satellite retrievals to generate higher resolution NO2 and SO2 vertical column density (VCD) maps (McLinden et al., 2014, 2016). Spatial patterns were consistent with the previous study, although substantial increases in peak VCDs over specific facilities were more distinguishable. More recent advances can identify individual NO2 plumes, which could allow for emission estimates to be derived for specific sources (Griffin et al., 2019). Satellite-based estimates of SO2 emissions have been compared with surface monitoring network SO2 concentrations over multiple years, revealing similar increases and decreases from 2009 to 2018 (McLinden et al., 2021).
Spatial patterns of PACs in ambient air were reported by Harner et al. (2018) in the AOSR using a network of 17 passive sites that have measured a wide range of PACs since 2010. As expected, the PAC air concentrations declined with increasing distance from mining activities. A much more rapid decline was observed for alk-PAHs and DBT derivatives, likely caused by their abundance in coarse particles, which have a relatively high deposition velocity. PAHs differed between near-facility and distal sites by a factor of 2–3, consistent with the contribution from multiple non-OS sources (e.g., biomass burning).
Studies reporting air measurements of mercury in the AOSR are limited. Total gaseous mercury (TGM) was measured at Fort McMurray between 2010 and 2013 (Parsons et al., 2013). Average ambient concentrations (1.45 ± 0.18 ng m−3) were comparable with measurements at other sites in Alberta (1.36–1.65 ng m−3) and are consistent with Northern Hemisphere background concentrations (Sprovieri et al., 2016). No correlation between TGM and NOx or SO2 was reported, and the highest TGM levels measured over the study period were caused by wildfire smoke. In general, the measured TGM concentrations appeared to be consistent with background values and were predictably driven by diel and seasonal trends superimposed on a combination of long-range transport and regional surface-air flux of gaseous mercury (Parsons et al., 2013).
Shephard et al. (2015) compared satellite observations of NH3, methanol, formic acid, and CO with aircraft measurements and GEM-MACH simulations to evaluate and constrain the satellite measurements. Results demonstrated small positive biases for satellite-derived CO (10%), formic acid (20%), and NH3 (7%) relative to airborne measurements, but a large negative bias for methanol (−54%). This indicates that satellites are a promising tool to understand emissions, transport, and fate of these pollutants in the AOSR. A similar approach was taken to evaluate the potential of satellite aerosol optical depth measurements to understand PM patterns (Sioris et al., 2017); however, spatial heterogeneity of the land surface resulted in poor correlation with ground-based measurements.
Fioletov et al. (2016) used a ground-based spectral sun photometer to measure SO2 VCDs over a two-year period in Fort McKay. The VCDs correlated well with co-located surface SO2 measurements during plume impingement, but correlated poorly when the SO2 plume remained aloft. This underscores the importance of understanding plume dynamics for determining the transport, transformation, and deposition of pollutants emitted from stacks. Recent studies have improved the agreement between modeled and measured plume heights by better accounting for horizontal heterogeneity in meteorological conditions (Akingunola et al., 2018; Gordon et al., 2018). These studies found that using meteorological data from surface stations several kilometers away can result in a negative bias of modeled plume height. Whenever possible, meteorological input variables that drive plume dispersion models should be taken from the immediate vicinity of the stack.
As emissions are transported through the atmosphere, many will undergo transformations that influence their fate and subsequent impacts. Given the complexity of emissions and the remoteness of the downwind regions, it is challenging to directly investigate these transformations. However, four flights during the 2013 ECCC aircraft campaign sampled downwind transects to quantify formation rates and/or total quantities of SOA, particulate organic nitrate (pON), LMWOA, and HNCO. These flights were conducted during summer and may not be representative of other times of the year when photochemistry is less active and emission rates may be different. Further analysis of ECCC 2018 winter and summer flight data is currently underway.
Liggio et al. (2016) observed a roughly sixfold increase in OA mass within a 4-h transport time during the August 2013 flights, equivalent to between 55 and 101 t day−1, which is comparable with formation rates downwind of Mexico City and Paris. Lab experiments revealed that SOA composition measured aboard the aircraft was similar to SOA composition from the oxidation of compounds that off-gas from bitumen (Li, Liggio, Lee, et al., 2019; Li, Liggio, Han, et al., 2019). Modeling calculations confirmed that most SOA produced downwind of the surface-mineable AOSR is a result of semi- and intermediate-VOC (SVOC and IVOC) oxidation. Ground measurements have confirmed the presence of SVOCs and IVOCs near AOSR facilities (Tokarek et al., 2018). A component of SOA is pON, which was observed to form at a rate of 15.5 t day−1, and contributes up to half of the SOA mass (Lee et al., 2019). Laboratory photo-oxidation experiments confirmed that formation occurs primarily through IVOC oxidation in the presence of high NOx concentrations. Li, Liggio, Lee, et al. (2019) performed additional photo-oxidation experiments on vapors that volatilized from OS ore, naphtha, tailings pond water, bitumen, and diluted bitumen. The SOA yields were much higher for OS ore and bitumen vapors, suggesting that most SOA formed downwind may be from the volatilization of precursors related to open-pit mining operations.
Stroud et al. (2018) used VOC and PM speciation profiles derived from aircraft measurements to model VOC and OA concentrations in the AOSR using GEM-MACH. The revised VOC speciation profiles resulted in larger SOA production and a slightly better agreement with observations. However, there was still a negative bias in modeled OA compared with aircraft observations (−2.37 ppbv improved from −2.79 ppbv), suggesting the need for additional data on emissions and OA formation processes to correctly simulate downwind SOA and hence PM2.5.
Secondary formation rates of LMWOAs were estimated to be 180 t day−1 (Liggio, Moussa, et al., 2017), although 54%–77% of the precursors were unaccounted for with the currently known VOC emissions. The authors suggest that IVOC oxidation may also be contributing to the relatively large LMWOA formation rates downwind. The observed formation rate was more than an order of magnitude higher than estimated primary LMWOA emissions. Similarly, the aircraft-derived secondary HNCO formation rate (~150 kg h−1) was observed to be between 2 and 20 times higher than the estimated direct emissions (Liggio, Stroud, et al., 2017).
Cheng, Li, et al. (2018) used BC measurements from three of the downwind transformation aircraft flights to demonstrate that mass median and number median diameters of refractory BC particles were similar to fresh urban emissions and remained relatively unchanged during 3 h of transport. However, the coating thickness around the refractory BC cores increased from ~20 to 40 nm within the 3 h as a result of SOA formation. The size and coating thickness of BC particles are important input parameters in aerosol-climate models.
Because PACs are a highly varied class of compounds, the transformation products are difficult to quantify. Nonetheless, quinones and NPACs are known to be major photochemical transformation products of PACs, and the latter have been observed in air at sites around the AOSR in the low ng m−3 range (Harner et al., 2018; Wnorowski, 2017; Wnorowski & Charland, 2017). Quinones were found primarily in the particle-phase during the daytime.
Results from PMF analyses on ambient PM2.5 samples discussed previously indicated that secondary sulfate is a substantial contributor to PM2.5 concentrations (~30%–40%) in Fort McKay and Fort McMurray (Bari & Kindzierski, 2017; Landis et al., 2017). Under normal facility operations, upgrader stacks are the largest regional source of SO2, which will undergo oxidation via several pathways to form pSO42−. The chemistry of SO2 oxidation to form pSO42− has been extensively studied in the atmospheric chemistry literature (Seinfeld & Panadis, 2016).
Atmospheric deposition
Atmospheric emissions and transformation products deposit to the surface via wet (precipitation) and dry processes (gases and particles directly deposit to surfaces). A brief description of deposition terminology and methods is given in the Supporting Information, whereas this section focuses on results specific to the OSR. Edgerton et al. (2020) analyzed deposition data collected from 2008 to 2017 to evaluate bulk, dry, and total deposition patterns and trends in jack pine stands for inorganic N (NO2, NH3, NH4+, HNO3, and NO3−), S (SO2 and SO42−), and BCat (Ca2+, Mg2+, K+, and Na+). Total, N, S, and BCat deposition were calculated using a combination of inferentially modeled dry deposition from passive gas sampler, denuder, and filter pack data, and bulk wet-deposition and vegetative throughfall ion exchange resin (IER) data collected by the WBEA's Terrestrial Environmental Effects Monitoring (TEEM) Program. Total S deposition was estimated using throughfall observations, assuming that dry deposited SO2 and pSO42− would be completely removed from vegetated surfaces by precipitation. In contrast, bulk NO3− and NH4+ along with modeled dry deposition were used to estimate total N deposition. Throughfall observations were used to estimate Ca2+ and Na+, and throughfall K+ and Mg2+ were adjusted downwards (decreased), assuming that the levels of these cations would increase due to vegetation leaching.
Total deposition rapidly decreased with increasing distance from OS facilities, with deposition rates approaching regional background around ~50 km (Edgerton et al., 2020). Total N and S deposition values at 20 stations averaged over 3–8 years varied from ~2.0 to 5.7 kg N ha−1 a−1 and 2.0 to 14 kg S ha−1 a−1, respectively. Bulk deposition of NH4+ and NO3− were substantial contributors (40%–62%) to total N deposition, whereas dry NO2 deposition was dominant at sites close to OS facilities. Reduced N contributed between 28% and 58% of total N deposition, with this percentage typically increasing with distance from OS facilities. The relative contribution towards total N deposition across the entire network was wet NH4+ ≥ wet NO3− > NO2 ~ HNO3 > NH3 > pNO3− > pNH4+. Potential acid input (PAI) was calculated for the 2011–2012 data and was typically in the range of 0.1–0.2 keq ha−1 a−1, with higher values (0.3–0.7 keq ha−1 a−1) closer to OS facilities. Year-to-year variability in total N, S, and BCat deposition was ±50%, which suggests the necessity for long-term monitoring for trend detection.
Various periods and deposition sites have also been analyzed to calculate dry (Hsu et al., 2016) and bulk or total deposition rates (Fenn et al., 2015; Proemse, Mayer, & Fenn, 2012; Proemse et al., 2013; Watmough et al., 2014; Wieder, Vile, Scott, et al., 2016). These studies yielded results similar to those of Edgerton et al. (2020) in terms of spatial deposition patterns. Lynam et al. (2015) also reported event-based, wet-only deposition measurements for a 21-month period in Fort McMurray and found SO42−, NO3−, and NH4+ wet deposition to be 1.96, 1.60, and 1.03 kg ha−1 during the study period, respectively.
A new methodology based on aircraft observations was developed to determine lifetimes against dry deposition, as well as average deposition velocities, for total oxidized sulfur and total reactive oxidized nitrogen caused by dry deposition over a spatial scale of (3–6) × 103 km2 in the AOSR (Hayden et al., 2021). Airborne measurements during three ECCC flights in 2013 tracked a pollutant air mass at different intervals up to 107–135 km downwind of emission sources (equivalent to 4–5 h transport time). Direct comparisons were made with modeled dry deposition estimates to assess uncertainties and the spatial–temporal scales of air pollutant impacts. Dry deposition fluxes decreased exponentially with distance from the oil sands sources resulting in lifetimes of 2.2–26 h. Fluxes were 2–14 and 1–18 times higher than model estimates for oxidized sulfur and nitrogen, respectively, indicating dry deposition velocities that were 1.2–5.4 times higher than those computed for models. A Monte Carlo analysis with five common inferential dry deposition algorithms indicates that such model underestimates of dry deposition velocity are typical. These findings indicate that deposition to vegetation surfaces are likely underestimated in regional and global CTMs regardless of the deposition algorithm used. The model-observation gaps may be reduced if surface pH and quasi-laminar and aerodynamic resistances in algorithms are optimized.
The spatial patterns of N, S, and BCat deposition measurements are consistent with CTM results (see Figure 6), which reveal enhanced deposition around the surface-mineable OS facilities (Cho et al., 2017; Makar et al., 2018). Speciated and total N, S, and BCat deposition were simulated on a 2.5 × 2.5 km grid covering Alberta and Saskatchewan from August 2013 to July 2014 (Makar et al., 2018). Modeled annual wet deposition correlated well with wet-deposition observations (R2 between 0.72 and 0.90), although were biased low and high for BCat (factor of 0.4) and S (factor of 2.2), respectively (Makar et al., 2018). These modeled output fields were then corrected using model-measurement fusion with wet-deposition measurements, and further evaluated against aircraft measurements and snowpack deposition measurements to improve simulated deposition fields.
Figure 6 shows the total S, N, and BCat deposition from the GEM-MACH model-measurement fusion described above and as described in Makar et al. (2018). A similar rapid decrease in deposition with increasing distance from the facilities was seen in model simulations, with >10 kg N ha−1 a−1 within 5 km of the sources (Figure 6B), dropping to 1 to 3 kg N ha−1 a−1 at >30 km from the sources. Total S deposition was >10 kg ha−1 a−1 within 7–30 km of the main facilities, and values of 1–3 kg S ha−1 a−1 from 40 to 80 km downwind (Figure 6A). Makar et al. (2018) found that modeled S deposition was dominated by dry SO2 and wet HSO3− deposition close to sources, whereas wet SO42− deposition dominated downwind. For N deposition, dry NO2 and dry NH3 were the largest contributors near OS facilities. The relative impact of oxidized versus reduced nitrogen was consistent with Edgerton et al. (2020). The simulated deposition of base cations is shown in Figure 6C and has a similar pattern to total S and N deposition. CTMs make use of vegetation-type weighting within each model grid cell to estimate average deposition for an entire grid cell, while the data collected by Edgerton et al. (2020) were specific to jack pine forest canopies. The GEM-MACH also includes forms of N not captured by the measurement network (e.g., HONO, PAN, organic-N) reported in Edgerton et al. (2020). Further model simulations and a more formal comparison between modeled and observed deposition are underway.
Studies in the AOSR have demonstrated that acidifying deposition is, at least partially, neutralized by BCat deposition in areas to the northeast of surface-mining operations (Edgerton et al., 2020; Fenn et al., 2015; Makar et al., 2018; Watmough et al., 2014). Figure 6D shows the ratio of total BCat to total anion (S plus N) deposition, normalized to equivalents. Areas with a ratio below one mean that acidifying deposition exceeds BCat deposition. The model-measurement fusion results reveal areas within ~10–20 km of surface-mining facilities generally have BCat deposition in excess of acidifying deposition. However, this is in contrast to Edgerton et al. (2020), who reported an excess of acidifying deposition closer to surface-mining facilities (i.e., positive PAI). The cause(s) of this discrepancy is not clear, but could be related to the differences in methodology for the model-measurement fusion and model and observations noted above.
Based on the above measurements and modeling studies, as well as emission inventories shown in Figure 3, an expansion of N, S, and BCat deposition measurements to the Cold Lake OSR is warranted, because there is elevated N and S deposition with less mitigating BCat deposition relative to the surface-mineable region. However, because N and BCat deposition is likely influenced from spatially diffuse anthropogenic sources (e.g., agriculture, upwind urban emissions), it would be challenging to set up a monitoring network along a gradient of deposition.
Measurements of PACs deposition are complicated by the hundreds of different compounds classified as PACs. Each compound has unique properties (e.g., solubility, vapor pressure, reactivity) that affect both wet and dry deposition rates. In the OSR, PAC deposition has been assessed using snowpack, precipitation, lichens, moss, and novel dry deposition passive air samplers (Harner et al., 2018). Multiple studies have quantified wintertime PACs deposition at hundreds of sites across the AOSR and reported loadings ranging from 2.2 to 26 000 μg m−2 (Bari et al., 2014; Cho et al., 2014; Kelly et al., 2009; Manzano, Muir, Kirk, et al., 2016). The highest deposition fluxes were found closest to surface-mineable facilities, and rapidly declined with increasing distance until plateauing at ~20–40 km away. Average wintertime deposition at sites in the Peace Athabasca Delta (PAD; >100 km from current surface mines) were 25 μg m−2 in 2012 and 13 μg m−2 in 2013. The major ∑PACs component by mass was alk-PAHs (52%–66%), followed by DBTs and alk-DBTs (24%–39%; Harner et al., 2018). An assessment of temporal changes in snowpack data between 2008, 2012, 2013, and 2014 resulted in both a decreasing and relatively steady amount of ∑PACs deposition over time, depending on the sites chosen and interpolation techniques used.
Muir et al. (2012) reported wet-deposition measurements of PACs at three sites near surface-mineable facilities in 2011. PACs wet deposition was highest from December to March, ranging from 45% to 64% of annual wet deposition (830–3380 μg m−2 a−1). On average, most PACs in the precipitation samples were C1-C4 alk-PAHs (60%), followed by unsubstituted PAHs (23%) and DBTs (16%). Relative to the snowpack studies, the relatively smaller annual fluxes reported by Muir et al. (2012) are likely the result of the contribution of fugitive dust sources (e.g., petcoke and haul-road dust) to total deposition, which are not captured by wet-only samplers during dry periods (Harner et al., 2018).
To determine dry deposition of PACs, a passive dry deposition sampler (PAS-DD) was developed by Eng et al. (2014). In the AOSR, PAS-DDs were co-deployed with traditional PACs passive air samplers at five sites to evaluate and understand the contribution of gas and particle dry deposition. This technique also shows promise for measuring TE dry deposition (Gaga et al., 2019).
Through the WBEA TEEM Program, epiphytic lichen samples were collected in the AOSR in 2002, 2008, 2011, 2014, and 2017. Although lichen measurements cannot be used to calculate deposition fluxes (mass per area), they are an effective biomonitor of spatial and temporal deposition patterns and are useful for conducting source apportionment studies (Graney et al., 2017; Landis et al., 2012; Landis, Berryman, et al., 2019; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019). Spatial patterns of PACs concentrations in lichen are consistent with snowpack data, revealing a rapid decline with increasing distance from OS facilities to ~40 km (Landis, Berryman, et al., 2019; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019; Studabaker et al., 2012). Similar spatial patterns in moss concentrations were also reported for 13 PAHs and four alk-PAHs by Zhang et al. (2016).
Samples of lake sediment cores, peat cores, and tree cores have been used to assess PAC trends over the past ~100 years in the AOSR. Between 2011 and 2015, sediment cores were collected from 33 lakes in northeastern Alberta and northwestern Saskatchewan to quantify PACs sediment fluxes (Ahad et al., 2015; Jautzy et al., 2013; Kurek et al., 2013; Summers et al., 2016). Harner et al. (2018) reported on 28 of these lakes and found sediment fluxes of PAHs, alk-PAHs, and DBTs in near-field lakes (<25 km from upgraders and petcoke piles) increased by ~3 times between ca. 1980 and 2015, whereas far-field lakes (>25 km) exhibited increases by less than a factor of 2. Increases started ca. 1980 as surface-mining activities in the region were ramping up. Fluxes of alk-PAHs were approximately 10 times larger than fluxes of DBTs and PAHs. These results are consistent with peat cores collected by Zhang et al. (2016), which revealed increasing ∑PAHs and ∑alk-PAHs in peat cores since ca. 1960 at sites within 25 km of OS surface-mining facilities. No temporal trend was reported in the peat core collected ~40 km south of these facilities. Preliminary work has revealed that tree cores might be a promising archive to assess PACs historical trends in air, and preliminary results are consistent with the spatial and temporal patterns of PACs deposition described above (Rauert & Harner, 2016; Rauert et al., 2017, 2020).
A modeling framework to calculate dry deposition of 17 PAHs, 21 alk-PAHs, and 5 DBTs in the AOSR was developed by Zhang, Cheng, Wu, et al. (2015). Atmospheric PAC concentrations from three sites close to OS facilities were used to determine annual dry deposition of 3920–5380 μg m−2 a−1 (forest canopies) and 2850–4920 μg m−2 a−1 (grass and shrubs) for alk-PAHs. Contributions were smaller from PAHs (330–560 μg m−2 a−1 for forests; 270–490 μg m−2 a−1 for grass and shrubs) and DBTs (230–1120 μg m−2 a−1 for forests; 450–930 μg m−2 a−1 for grass and shrubs). Measurements of PACs in precipitation collected from 2011 to 2012 in Fort McKay and Fort McMurray were used to calculate scavenging ratios in rain and snow (Zhang, Cheng, Muir, et al., 2015). Results indicated that snow scavenging was more efficient than equivalent amounts of rain, which is consistent with fugitive dust being the primary source of PACs because large, irregularly shaped particles are efficient cloud condensation nuclei under cold conditions. A study by I. Cheng, Wen, et al. (2018) produced gridded deposition estimates of 17 PAHs, 21 alk-PAHs, and 5 DBTs for 2011 using the dry deposition framework, scavenging ratios, and gridded air concentrations developed by a dispersion model corrected to passive air sampler measurements. Total PACs deposition values ranged from 55 to 175 000 μg m−2 a−1 in grid cells across the surface-mineable AOSR (median fluxes of 760 μg m−2 a−1). The total contributions towards dry and wet deposition, respectively, were 19% and 42% (PAHs), 74% and 49% (alk-PAHs), and 7% and 9% (DBTs). However, Tevlin et al. (2021) reported that these modeled wet-deposition fluxes were approximately a factor of f4 higher than measured annual wet-deposition fluxes. In addition, modeled total deposition during the cold season was more than an order of magnitude higher than accumulated wintertime fluxes calculated with snowpack samples. Tevlin et al. (2021) recommended further comparison between modeled and measured deposition estimates at additional sites to understand and rectify the model-measurement discrepancies.
Deposition patterns of TEs and mercury have been assessed with wintertime bulk samplers (Bari et al., 2014), snowpack samples (Gopalapillai et al., 2019; Kelly et al., 2010; Kirk et al., 2014; McNaughton et al., 2019), lichen samples (Blum et al., 2012; Graney et al., 2012, 2017; Landis et al., 2012; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019), lake sediment cores (Cooke et al., 2017), and event-based, wet-only samplers (Lynam et al., 2015). Similar to PACs measurements, there is typically a “bullseye” pattern within ~25 km of surface-mining activities, although this varies among TEs and is affected by topography, meteorology, and PM size. For instance, Kelly et al. (2010) found that, in 2008, snowpack deposition was enhanced for all measured TEs, except selenium, at sites nearer to surface mines. Similarly, Kirk et al. (2014) reported increased snowpack mercury and methylmercury within ~25 km, and noted that ~80% of the total mercury was particulate-bound. A reanalysis of this snowpack data revealed similar spatial patterns, albeit at lower total loadings due to a revised interpolation method (McNaughton et al., 2019). An analysis of snowpack samples collected in 1978, 1981, 2008, and 2011–2016 revealed a general decrease in TE loadings over time, although more recent data for near-field sites are still an order of magnitude greater than reference sites in the PAD (Gopalapillai et al., 2019). Spatial patterns of loadings of 45 different TEs in >120 lichen sampling locations across the AOSR in 2008 and 2014 confirm that the spatial extent of increased deposition caused by surface-mining activities is ~20–30 km (Landis et al., 2012; Landis, Studabaker, Pancras, Graney, Puckett, et al., 2019). The lichen and snowpack data both suggest that coarse PM emissions dominate near-field TE deposition. On the other hand, Cooke et al. (2017) found little evidence of enhanced TE and mercury levels in lake sediment cores. However, only two of the 20 lake sites in that study were within 20 km of the “midpoint” of surface-mining operations.
Deposition of total mercury was simulated using GEM-MACH-Hg by Emmerton et al. (2018) and revealed that <2% of direct total mercury deposition at 50 lakes throughout the AOSR was a result of OS-related emissions. However, these 50 lakes were >30 km away from surface-mining activities. Modeled deposition within 30 km of the surface mines had a higher contribution of total mercury deposition from OS-related emissions (~10%), with most of the deposition originating from long-range transport as opposed to wildfires (<5%). More recently, Dastoor et al. (2021) found that GEM-MACH-Hg simulated mercury air concentrations and accumulated snowpack mercury were comparable with independent measurements in the AOSR. The authors found that OS sources had negligible impact on gaseous elemental mercury air concentrations, but average air concentrations of total oxidized mercury were elevated ~60% above background within 50 km of OS sources. Furthermore, although global sources dominated annual mercury deposition in the AOSR, deposition enhancements were observed within 100 km (winter) and 30 km (summer) of sources as a result of OS emissions.
The spatial patterns of PM air concentrations and deposition are not as well constrained as other pollutants due to (i) limitations in the available bottom–up emission inventory data, (ii) limitations for using satellites to infer spatial gradients of PM, (iii) the lack of spatially extensive PM monitoring surrounding the surface-mining facilities, and (iv) a paucity of PM size distribution measurements, especially in the range of 2.5–25 μm. There is also the added complexity of secondary PM2.5 formation as well as reversible gas-to-particle partitioning that will affect spatial patterns and transport of some PM2.5 components. The addition of surrogate surface samplers to the existing deposition network could help fill this gap and allow for improved deposition estimates of PM components such as TEs, PACs, mercury, and BCat (Hall et al., 2017).
Impacts
Ambient air quality
Air quality is commonly assessed by comparing air-monitoring data with established benchmarks, such as the national Canadian Ambient Air Quality Standards (CAAQS), provincial Alberta Ambient Air Quality Objectives and Guidelines, or those specified in the Lower Athabasca Region Air Quality Management Framework. This approach is straightforward, but typically applies only to a limited number of air pollutants (e.g., SO2, NO2, O3, PM2.5 mass). Furthermore, benchmarks such as the AAAQOs consider factors such as technical feasibility and socioeconomics, and do not necessarily protect against health risk for highly sensitive individuals or ecological systems. Nonetheless, these limits were typically developed based on integrated health impact assessments. Regulated air pollutants also have established federal reference methods and federal equivalent methods to ensure consistent and comparable measurement reporting standards. It is much more challenging to assess the extent to which “other” nonregulatory pollutants (e.g., most PACs, VOCs, and TEs) affect human or ecosystem health. The lack of benchmarks for some air pollutants is not an OSR-specific issue, so it will remain challenging to evaluate the impacts of some known OS-related stressors (e.g., alk-PAHs, DBTs, many VOCs). Alternative approaches to evaluating air quality impacts from OS-related stressors without benchmarks should continue to be explored and applied (e.g., oxidative health assays, co-located ecological-effects monitoring).
The benchmarks commonly applied in the OSR are listed in Supporting Information (Table S2). Various peer-reviewed papers and reports that have compared air concentrations of NO2, SO2, O3, and PM2.5 with benchmarks are summarized in the Supporting Information (i.e., Alberta Airsheds Council, 2019; Bari & Kindzierski, 2016; Brown, 2019a, 2019b; Bytnerowicz et al., 2016; Cho et al., 2012; Landis et al., 2018; Percy et al., 2012; Percy, 2013; Vijayaraghavan et al., 2016; Wood Buffalo Environmental Association, 2020). Typically these benchmarks are not exceeded, except during wildfire smoke events.
Assessing air quality impacts of most VOCs, PACs, and PM components (e.g., TEs) is challenging due to a lack of appropriate benchmarks. Nonetheless, Bari and Kindzierski (2018) analyzed 24-h averaged VOC data collected at Fort McKay and Fort McMurray between January 2010 and March 2015. Annually averaged benzene ranged from 0.6 to 1.3 μg m−3, below the annual AAAQO of 3.0 μg m−3. However, all other VOCs with AAAQOs are reported as 1-h averages. As a result, Bari and Kindzierski (2018) estimated carcinogenic and noncarcinogenic risks from VOC exposure by comparing ambient levels of eight measured VOCs to inhalation toxicity screening criteria developed by the US EPA's Office of Air Quality Planning and Standards. VOC concentrations were below chronic and acute toxicity screening criteria, except for acetaldehyde. Wentworth et al. (2018) reported that 24-h averaged acetaldehyde exceeded the 1-h AAAQO of 50 ppb on two days in August 2016 in the AOSR; however, the reasons for these exceedances could not be determined. Ambient concentrations of HNCO above 1000 pptv are associated with adverse health effects, which is above the average HNCO levels simulated in Fort McMurray (~25 pptv) in August and September (Liggio, Stroud, et al., 2017). About half of the simulated HNCO in Fort McMurray was attributed to OS industry related emissions. However, when Fort McMurray is downwind of OS surface-mining facilities, the predicted concentrations increased to 250–600 pptv, where >80% of the HNCO is from OS-related emissions.
With respect to PACs, AAAQOs only exist for naphthalene (3 μg m−3) and benzo[a]pyrene (0.3 μg m−3), both of which are annual averages. There is evidence that alk-PAHs are more toxic than their unsubstituted PAH congeners (Kelly et al., 2009; Ott et al., 1978); however, no benchmarks for alk-PAHs or DBTs in air currently exist. Jariyasopit et al. (2016) used PACs passive air samples from sites in the AOSR to conduct mutagenicity and cytotoxicity assays, which are considered a proxy for toxicity. This approach captures multiple analytes and synergistic effects; however, it is difficult to relate results to specific health effects. Air samples revealed weak correlation between mutagenicity and PAC concentrations, with no correlation for cytotoxicity, implying that the measured PACs were not driving mutagenicity or cytotoxicity. Irvine et al. (2014) studied excess cancer risks in Indigenous communities in the Cold Lake OSR resulting from PAH exposure from inadvertent ingestion of soils and inhalation of PAHs in air. The results did not indicate any statistically significant increased cancer risk from these exposure pathways.
Very few applicable benchmarks exist for components of PM. A few AAAQOs exist for ambient air concentrations of select TEs (i.e., arsenic, lead, manganese, nickel); however, these are either 1-h or annual averages, whereas ambient air concentrations are typically measured as 24-h averages every sixth day. Similar to their approach with PACs, Bari and Kindzierski (2017) evaluated health risks associated with inhalation of TEs in PM2.5, by comparing 24-h averaged PM2.5 composition data from 2010 to 2013 at sites in Fort McKay and Fort McMurray with toxicity reference values for six TEs. Ambient concentrations were considered to be within tolerable levels for carcinogenic risks and below a safe level of concern for noncarcinogenic risks.
To address limitations of comparing ambient air concentrations with single-pollutant benchmarks, Wren et al. (2020) developed an “event-based” approach to characterize complex air pollutant mixtures in Fort McKay from August 2013 to December 2016. Air concentration data from multiple instruments were analyzed using principal component analysis to categorize air pollutant mixtures related to hydrocarbon emissions, fossil fuel combustion, dust, and oxidized and reduced sulfur compounds. Data were used to isolate “events” of sustained air pollutant increases, defined as a ≥2-h period when the event indicator concentration exceeded the 75th percentile, with at least one 30-minute measurement exceeding the 95th percentile. Based on this approach, Fort McKay was affected by one or more events 47% of the time, with the median duration of each event between 5.5 and 9.5 h depending on the mixture type. Diesel exhaust from within OS facilities was likely the most important NOx source during events. A CMB model using source profiles from Li et al. (2017) revealed that 86% of total VOC mass during VOC events was from four OS facilities near Fort McKay. The framework developed by Wren et al. (2020) provides an important perspective on air quality in the community and complements the more “routine” approach of using a small number of single-pollutant air quality standards. This example highlights the utility of many co-located high time resolution measurements for identifying air pollution events and quantifying source contributions from specific OS facilities.
Studies that have investigated temporal trends in air pollutant concentrations in the OSR are summarized in the Supporting Information. The stepwise nature of OS development and considerable variability in emissions appears to limit the usefulness of traditional linear regression for air quality trending. More nuanced trending and analysis approaches appear necessary to understand the relationship between OS emissions and air quality impacts (e.g., Edgerton et al., 2020; Landis, Berryman, et al., 2019; Wren et al., 2020). Air quality models are another means by which the nonlinearity inherent in the chemical system may be incorporated and linked to OS-related emissions.
Odor events
Odor is a complex issue because of the number of odorous compounds, the wide difference in odor thresholds of compounds, differences in odor perception among individuals, and a lack of odor-related objectives (Davidson & Spink, 2018). Adding to the complexity is the transient and episodic nature of odorous events, which makes it challenging to link an odor event back to the specific emission source(s), meteorological conditions, or location (Gosselin et al., 2010). There have been relatively few peer-reviewed publications focused on odors in the OSR.
Odors are a recurrent issue for some OSR communities and can negatively affect quality of life (Clean Air Strategic Alliance, 2015). Odor issues have been documented via complaints submitted to the regulator, government-led investigative reports, public hearings, and research work conducted by communities and airshed organizations (e.g., Alberta Energy Regulator and Alberta Health, 2016; Beausoleil et al., 2021; Stantec, 2014). Recurring odor issues have been identified near surface-mining activities (AOSR) and near in situ production facilities (Peace River OSR).
There were 165 odor-based complaints received by the AER from Fort McKay residents between 2010 and 2014 (Alberta Energy Regulator and Alberta Health, 2016). To investigate the underlying cause, the AER and Alberta Health initiated an in-depth review of air quality and odor monitoring around Fort McKay (Alberta Energy Regulator and Alberta Health, 2016). The review considered air-monitoring data in the context of several odor-, air quality-, and health-based thresholds and analyzed air-monitoring and facility data collected at the time of the complaints to identify any possible operational upsets. A link was identified between odors in Fort McKay and OS mining operations, but the specific industry sources contributing to odors could not be clearly identified (Alberta Energy Regulator and Alberta Health, 2016).
Two publications examined odor events and AAAQO exceedances for odor perception in the AOSR. Both reported a substantial decrease in odor events and/or exceedances following modifications made ca. 2010 (not described in the literature) to reduce odorous emissions from OS operations (O'Brien et al., 2012; Percy, 2013). This is consistent with clear step decreases in ambient TRS concentrations measured in Fort McKay after 2010. The decrease in TRS was attributed to several factors, including diluent treatment to minimize off-gassing of RSCs from tailings ponds (Bari & Kindzierski, 2015). However, despite these improvements, odors are still apparent and remain a concern in communities close to OS developments (Beausoleil et al., 2021; Davidson & Spink, 2018).
A report by O'Brien (2014) indicated that thiophenes, cyclohexyl-isothiocyanate, benzothiozole, H2S, and/or carbon disulfide were likely the compounds responsible for most of the odors in the AOSR, which coincided with elevated TRS readings. In an ongoing effort to document and investigate odors in the AOSR, the WBEA has developed a Community Odour Monitoring Program (COMP), in close collaboration with communities. The COMP collects community odor complaints and compares them with concurrent ambient air quality monitoring data to identify any linkages or patterns between odor and ambient air pollutants (see https://wbea.org/odours/community-odour-monitoring-program/and; Beausoleil et al., 2021).
Numerous odor complaints were also submitted by residents in the Peace River OSR beginning in 2010, following a period of rapid industrial development in the Three Creeks, Reno, and Seal Lake areas. In January 2014, a public hearing panel heard from residents and independent subject matter experts on odors and emissions from regional heavy oil extraction processes (Alberta Energy Regulator, 2014). One important outcome was the 2018 release of AER Directive 084: Requirements for Hydrocarbon Emission Controls and Gas Conservation in the Peace River Area, which includes requirements for operators to eliminate routine flaring, prevent (reduce) nonroutine venting (flaring), reduce fugitive emissions, and minimize odors from transportation and maintenance activities. The AER maintains an online Peace River Performance Dashboard to track and communicate the industry's progress on reducing flaring and venting, and efforts to conserve gas in the Peace River OSR.
Ecosystem responses
A thorough review of ecosystem responses related to the deposition is covered in other review papers in this series (Arciszewski et al., 2021; Roberts et al., 2021a; 2021b); however, major findings are introduced here. Acidification of soils and surface water can occur through excessive deposition of acidifying substances (e.g., SOx, NOx, NHx). Substantially lower soil or water pH can affect mobilization and solubility of metals. Different methodologies have been used to calculate the ecosystem acidification impacts of OSR emissions. The critical load concept is based on determining the local charge balance between anions and cations in surface waters and groundwater, and has been recommended and used in international agreements for N and S emissions controls in Europe (CLRTAP, see Makar et al., 2018 for details). The critical load methodology described here is a more explicit assessment relative to PAI, which is usually calculated as the difference between the sum of N and S deposition and BCat deposition (Cho et al., 2017; Edgerton et al., 2020).
Several studies in the AOSR have demonstrated that the current potential for widespread acidification in soils is limited, based on comparison of S deposition rates with critical loads of acidification derived from soil data (Government of Alberta, 2008; Jung et al., 2013; Whitfield et al., 2009). This is likely the result, at least in part, of the concurrent deposition of BCat (Edgerton et al., 2020; Fenn et al., 2015; Makar et al., 2018; Watmough et al., 2014). However, Makar et al. (2018) suggested that the neutralizing impact of BCat deposition decreases with distance (see Figure 6D), and exceedances of critical loads for forest ecosystems may occur further downwind. The WBEA's TEEM Program has collected 20 years of deposition, soil, and vegetation data in jack pine stands within ~150 km of the surface-mineable sources. A recent analysis revealed a lack of correlations between acidifying deposition and numerous ecological indicators of soil acidification, implying that widespread soil acidification is not occurring (Davidson et al., 2020). Similarly, studies that have used lake sediment cores to infer pH changes over time around and downwind of the AOSR have not detected region-wide lake acidification or changes in phytoplankton communities (Curtis et al., 2010; Hazewinkel et al., 2008; Laird et al., 2013).
Makar et al. (2018) used GEM-MACH simulated deposition of acidifying N and S compounds, as well as BCat, across Alberta and Saskatchewan to compare deposition rates against critical loads of acidification. The extent of critical load exceedances varied between 1 × 104 and 3.3 × 105 km2, for observation-corrected model fields. The extent of exceedances depended on the critical load calculation methodology and the ecosystem type (forest ecosystems were less sensitive than aquatic ecosystems). These results are not inherently inconsistent with previous studies that did not observe acidification, because the critical load exceedance metric does not indicate time-to-effect. Cho et al. (2019) analyzed 36 years of soil chemical properties at six long-term forest sampling sites in central and northern Alberta, and found decreasing trends of pH and base saturation, indicative of soil acidification, at five of the sites.
Terrestrial, aquatic, and wetland ecosystems can undergo eutrophication, a process by which excessive nutrient deposition (e.g., N, P) alters ecosystem structure and function. The TEEM Program reported positive correlations between N and S deposition and understory plant cover and richness, suggesting eutrophication may be occurring at jack pine sites within tens of kilometers of emission sources (Bartels et al., 2019; Davidson et al., 2020). Similar correlations between N deposition and ecological indicators have been observed in bogs in the AOSR (Wieder, Vile, Albright, et al., 2016; Wieder, Vile, Scott, et al., 2016). Artificial N-addition experiments conducted over five years in bogs and poor fens confirmed these findings, and found that N addition down-regulated N2-fixation, as well as increased N-leaching and vascular plant biomass (Vile et al., 2014; Vitt, 2016). However, accounting for confounding factors such as climate and N2-fixation rates is complicated (Vile et al., 2014; Wieder, Vile, Albright, et al., 2016). As a result of the N-addition experiments, a critical load of 3 kg N ha−1 a−1 was suggested for both poor fens and bogs in the region (Wieder et al., 2019, 2020). As discussed previously, observed total N deposition amounts exceed 3 kg N ha−1 a−1 at many monitoring sites extending 75 km to the north and south, and 50 km east and west of surface-mining facilities (Makar et al., 2018). Smaller regions with >3 kg N ha−1 a−1 level were also predicted south of Fort McMurray. This differs from Edgerton et al. (2020), which demonstrated N deposition >3 kg N ha−1 a−1 at only a handful of sites close to operations. Most sites exhibited N deposition between 2 and 3 kg N ha−1 a−1, with remote sites close to 2 kg N ha−1 a−1.
The plethora of PACs and TEs, as well as multiple exposure pathways (e.g., ingestion, inhalation, and absorption), make it challenging to conclusively identify stressor-pathway-response linkages. Literature evaluating contamination exposure in mammals, birds, and aquatic environments is presented in other review papers in this series (Arciszewski et al., 2021; Roberts et al., 2021b).
CONCLUSIONS
This review paper synthesizes more than 10 years of peer-reviewed literature on air and deposition research in the OSR. Conceptual model diagrams were developed for each paper in this series to illustrate how stressors (i.e., air pollutants) move along chemical and physical pathways from pressure (i.e., emission source) to elicit a response. Most (96%) peer-reviewed literature focused on the surface-mineable area of the AOSR. Future efforts should focus on improving our understanding of pathways and pressures for stressors that are known, or suspected, to cause a response, as described below and in Roberts et al. (2021a).
The conceptual model (Figure 2) displays which air pollutants have been linked to specific emission sources (72 papers). Air pollutants that have relatively better characterized emission sources are generally more straightforward to measure, and have several studies that use multiple lines of evidence (e.g., top–down aircraft measurements, direct source emission testing, and source attribution techniques) that yield consistent results. Emission inventories are used to understand spatial patterns of emission sources (Figure 3), which can inform monitoring site selection. However, substantial uncertainties remain for the quantification of emissions of NH3, VOCs, PACs, RSCs, and some PM components (e.g., >2.5 μm), as demonstrated by continued discrepancies between top–down and bottom–up approaches.
The transport of atmospheric emissions is generally well understood as a result of dense monitoring networks, satellite measurements, aircraft observations, and CTMs. In general, studies yield similar results: Air concentrations and deposition experience exponential decay with increasing distance from sources. Levels are generally elevated above regional background within ~20–30 km of surface mines for pollutants contained within PM >2.5 μm, whereas gaseous and PM2.5 pollutants are elevated across a larger spatial extent (>40 km). In addition, there is an “elongation” of the air concentration and deposition patterns in the north–south direction caused by airflow along the Athabasca River Valley. Many pollutants undergo transformations as they are transported; some of these transformations have been well studied (e.g., Seinfeld and Panadis, 2016). However, some transformations, particularly those for PACs and VOCs, are more complex and are an active area of research. Aircraft and laboratory studies have been conducted in the AOSR to understand the oxidation of IVOCs that arise from bitumen extraction (Lee et al., 2019; Li, Liggio, Han, et al., 2019; Li, Liggio, Lee, et al., 2019; Liggio et al., 2016), which result in the substantial downwind formation of secondary pollutants (e.g., SOA).
Air quality is commonly assessed against various benchmarks, which are single-pollutant standards typically designed for human health (i.e., not ecosystem protection). Although air concentrations in the OSR are often below applicable metrics, this is not a comprehensive lens through which to assess local or regional air quality. Many nonregulatory pollutants do not have benchmarks, or could act in synergistic ways, necessitating the development of multi-pollutant indicators for air quality (e.g., Han et al., 2020).
Co-located deposition and ecological monitoring has successfully identified ecological changes in jack pine, bog, and poor fen ecosystems. These responses are likely linked to elevated N deposition at these sites. Similar evidence of widespread regional acidification of lakes or soils has not been observed, although some modeling has indicated a risk of acidification in future. Long-term co-located monitoring at these ecological sites should continue because eutrophication and acidification of ecosystems are typically slower processes, but can take decades to reverse if they occur. Although ecological responses as a result of PACs, TEs, and mercury deposition are less clear, there is merit in better quantifying the deposition of fugitive dust, because it a key vector for these pollutants as well as BCat. Ongoing collection and analysis of snowpack and lichen samples allow for estimation of spatial and temporal patterns of all these parameters. A similar approach to Mehaffey et al. (2009) could be taken to quantify the relative ecological risk of exposure to hazardous air pollutants.
There is a dearth of air and deposition studies in regions outside the surface-mineable area (<6% of papers), despite significant development of in situ facilities in these regions—this is a key spatial gap that spans the conceptual model. In particular, the mitigating effect of BCat on acidification is likely not occurring to an appreciable extent in the Cold Lake OSR despite emissions of NOx (Figure 3B) and NH3 (Figure 3C). Key findings are discussed further in the final review paper in this series, particularly for issues that span multiple themes (Roberts et al., 2021a).
-
Reconcile discrepancies between top–down and bottom–up emission estimates for NH3, VOCs, PACs, RSCs, GHGs, and PM >2.5 μm.
-
Develop suitable mechanisms (e.g., emission factors) to translate tailings pond disposal data into estimates of fugitive releases.
-
Studies that examine spatial patterns of concentrations and/or deposition should adopt a “distance to nearest source” approach (e.g., Landis, Berryman et al., 2019) as opposed to choosing a “midpoint” of the surface-mineable AOSR.
-
Determine specific sources of odor-causing RSCs and VOCs through targeted studies that directly measure emission rates from suspected sources (i.e., tailings ponds, mine faces, fugitive plant emissions related to bitumen upgrading), and by conducting high-resolution, back-trajectory and/or source apportionment analyses (e.g., Wren et al., 2020) during odor events reported through the COMP.
-
Monitor PM using surrogate surface samplers (e.g., Hall et al., 2017) and/or size-resolved PM speciation measurements to support deposition estimates of fugitive dust, source attribution efforts, and investigations assessing linkages between fugitive dust deposition and ecological responses.
-
Refine and improve calculations for critical loads of acidification and eutrophication, including estimates of time to effects.
-
Determine and assess critical levels to measure ecosystem responses to short-term, high concentration exposures of vegetation to pollutants.
-
Quantify the relative inputs of specific sources to N deposition (especially NHx) at ecological sites through the use of CTMs (e.g., Whaley et al., 2018) and source apportionment techniques.
-
Increase the extent of co-located measurements of deposition and ecological indicators, especially at surface water sites, to help assess pathway-response linkages.
-
Set region-specific exposure criteria and clarify the ecological responses to increased N deposition by performing more N-addition (exposure) experiments, similar to those summarized in Wieder et al. (2019, 2020).
-
Utilize air quality models to inform site selection in regions without deposition monitoring (Cold Lake and Peace River OSRs), when necessary, to ensure that sites are located along a deposition gradient.
-
Expand deposition and effects measurements in the Cold Lake OSR, and in areas where deposition mapping reveals the greatest risk of acidification (e.g., aquatic sites in the AOSR).
ACKNOWLEDGMENT
This work was funded under the Oil Sands Monitoring (OSM) Program and is a contribution to the program but does not necessarily reflect the position of the program. The OSM Program is a joint federal–provincial initiative. The authors thank Sandro Leonardelli, Elisabeth Galarneau, Alicia Berthiaume, Leiming Zhang, Junhua Zhang, Matthew Landis, Eric Edgerton, Kenneth Foster, Tim Arciszewski, and David Roberts for reviewing the manuscript and providing constructive feedback, as well as Mina Nasr for helping to generate figures.
CONFLICT OF INTEREST
The authors declare that there are no conflicts of interest.
Open Research
DATA AVAILABILITY STATEMENT
Not applicable. No new data were generated by this literature review.
