Characterization of trace gases measured over Alberta oil sands mining operations: 76 speciated C 2 –C 10 volatile organic compounds (VOCs), NO and

. Oil sands comprise 30% of the world’s oil reserves and the crude oil reserves in Canada’s oil sands deposits are second only to Saudi Arabia. The extraction and processing of oil sands is much more challenging than for light sweet crude oils because of the high viscosity of the bitumen contained within the oil sands and because the bitumen is mixed with sand and contains chemical impurities such as sulphur. Despite these challenges, the importance of oil sands is increasing in the energy market. To our best knowledge this is the ﬁrst peer-reviewed study to characterize volatile organic compounds (VOCs) emitted from Alberta’s oil sands mining sites. We present high-precision gas chromatography measurements of 76 speciated C 2 –C 10 VOCs (alkanes, alkenes, alkynes, cycloalkanes, aromatics, monoterpenes, oxygenated hydrocarbons, halocarbons and sulphur compounds) in 17 boundary layer air samples collected over surface mining operations in northeast Alberta on 10 July 2008, using the NASA DC-8 airborne laboratory as a research platform. In addition to the VOCs, we present simultaneous measurements of CO 2 , CH 4 , CO, NO, NO 2 , NO y , O 3 and SO 2 , which were measured in situ aboard the DC-8. In addition to the emission of many trace gases, the natural drawdown of OCS by vegetation was absent above the surface mining operations, presumably because of the widespread land disturbance. Unexpectedly, the mixing ratios of α -pinene and β -pinene were much greater over the oil sands (up to 217 pptv and 610 pptv, respectively) than over vegetation in the background boundary layer (20 ± 7 pptv and 84 ± 24 pptv, respectively), and the pinenes correlated well with several industrial tracers that were elevated in the oil sands plumes. Because so few independent measurements from the oil sands mining industry exist, this study provides an important initial characterization of trace gas emissions from oil sands surface mining operations. ozone (O 3 ) , nitric oxide (NO), nitrogen dioxide (NO 2 ) and total reactive nitrogen

ues and, with the exception of CO, were greater over the oil sands than at any other time during the flight. Twenty halocarbons (e.g., CFCs, HFCs, halons, brominated species) either were not enhanced or were minimally enhanced (<10%) over the oil sands. Ozone levels remained low because of titration by NO, and three VOCs (propyne, furan, MTBE) remained below their 3 pptv detection limit throughout the flight. Based on their correlations with one another, the compounds emitted by the oil sands industry fell into two groups: (1) evaporative emissions from the oil sands and its products and/or from the diluent used to lower the viscosity of the extracted bitumen (i.e., C 4 -C 9 alkanes, C 5 -C 6 cycloalkanes, C 6 -C 8 aromatics), together with CO; and (2) emissions associated with the mining effort, such as upgraders (i.e., CO 2 , CO, CH 4 , NO, NO 2 , NO y , SO 2 , C 2 -C 4 alkanes, C 2 -C 4 alkenes, C 9 aromatics, short-lived solvents such as C 2 Cl 4 and C 2 HCl 3 , and longer-lived species such as . Prominent in the second group, SO 2 and NO were remarkably enhanced over the oil sands, with maximum mixing ratios of 38.7 ppbv and 5.0 ppbv, or 383× and 319× the local background, respectively. These SO 2 levels are comparable to maximum values measured in heavily polluted megacities such as Mexico City and are attributed to coke combustion. By contrast, relatively poor correlations between CH 4 , ethane and propane suggest low levels of natural gas leakage despite its heavy use at the surface mining sites. Instead the elevated CH 4 levels are attributed to methanogenic tailings pond emissions. I. J. Simpson et al.: Trace gases measured over Alberta oil sands mining operations In addition to the emission of many trace gases, the natural drawdown of OCS by vegetation was absent above the surface mining operations, presumably because of the widespread land disturbance. Unexpectedly, the mixing ratios of α-pinene and β-pinene were much greater over the oil sands (up to 217 pptv and 610 pptv, respectively) than over vegetation in the background boundary layer (20 ± 7 pptv and 84 ± 24 pptv, respectively), and the pinenes correlated well with several industrial tracers that were elevated in the oil sands plumes. Because so few independent measurements from the oil sands mining industry exist, this study provides an important initial characterization of trace gas emissions from oil sands surface mining operations.

Introduction
Hydrocarbons are the basis of oil, natural gas and coal. Crude oil contains hydrocarbons with five or more carbon atoms (i.e., ≥C 5 ), with an average composition of alkanes (30%), cycloalkanes (49%), aromatics (15%) and asphaltics (6%) (Alboudwarej et al., 2006). Oil sands comprise 30% of total world oil reserves (Alboudwarej et al., 2006) and are a mixture of sand, water, clay and crude bitumen -a thick, sticky extra-heavy crude oil that is "unconventional", i.e., does not flow and cannot be pumped without heating or dilution. With 179 billion barrels in its deposits, most of which occurs as oil sands, Canada has the world's 2nd largest crude oil reserves after Saudi Arabia and is currently the world's 7th largest producer of crude oil, generating about 1 million barrels day −1 with almost 4 million barrels day −1 expected by 2020 (www.canadasoilsands.ca). Alberta is Canada's largest oil producer and has oil sands deposits in three relatively remote regions: Athabasca (which is serviced by Fort McMurray), Peace River and Cold Lake ( Fig. 1). Shallow oil sands deposits (<75 m deep) in the Athabasca region can be surface mined, which has disturbed about 600 km 2 of land and comprises ∼20% of Alberta's oil sands production (www.oilsands.alberta.ca). Each barrel of oil generated from surface mining requires about 2 tons of oil sands, 3 barrels of water -much of which is re-used -and 20 m 3 of natural gas (http://www.neb.gc.ca/ clf-nsi/rnrgynfmtn/nrgyrprt/lsnd/pprtntsndchllngs20152006/ qapprtntsndchllngs20152006-eng.html). Deeper deposits occur in all three regions and require other in situ recovery methods such as steam injection and have higher natural gas requirements, but they have the advantage of much less land disturbance and no need for tailings ponds (www.oilsands.alberta.ca).
Oil sands processing extracts the bitumen from the oil sands and upgrades it into synthetic crude oil. The surface mining operators (e.g., Syncrude Canada, Suncor Energy, Albian Sands Energy) extract the bitumen using a hot water process (e.g., www.syncrude.ca/users/folder.asp? FolderID=5918). Solvents/diluents are used to remove water and solids from the extracted bitumen and also decrease the bitumen's viscosity so that it can be piped to upgraders and refineries for processing. Diluent such as naphtha is used, which is a flammable liquid mixture of C 3 -C 14 hydrocarbons with major fractions of n-alkanes (e.g., heptanes, octane, nonane) and aromatics (e.g., benzene, toluene, ethylbenzene, xylenes) (Siddique et al., 2007). Paraffinic solvents/diluents may also be used, which consist of pentanes and hexanes (Siddique et al., 2006;D. Spink, personal communication, 2010). Upgraders crack and separate the bitumen into a number of fractions, e.g., fuel gas, synthetic crude oil, carbon (coke), and sulphur. There are upgraders at some of the surface mining sites north of Fort McMurray, at an in situ site south of Fort McMurray, and at downstream industrial centers such as the industrial heartland in Fort Saskatchewan, Alberta. Every 1 m 3 of oil sands also generates 4 m 3 of tailings waste, which includes both residual bitumen and diluent (Holowenko et al., 2000;Siddique, 2006). For example the Syncrude Mildred Lake tailings pond contains about 200 million m 3 of mature fine tailings and became methanogenic in the 1990s (Holowenko et al., 2000;Siddique et al., 2008). The methanogens are believed to anaerobically degrade certain components of naphtha (C 6 -C 10 n-alkanes and aromatics such as toluene and xylenes) into methane (CH 4 ), at rates of up to 10 g CH 4 m −2 d −1 or 40 million L CH 4 d −1 (Holowenko et al., 2000;Siddique et al., 2006Siddique et al., , 2007Siddique et al., , 2008Penner and Foght, 2010).
Despite the emerging importance of oil sands in the energy market, characterizations of the emissions of volatile organic compounds (VOCs) and other trace gases from the oil sands industry are extremely scarce in the peer-reviewed literature. Whereas downstream oil sands upgrading and refining facilities can be monitored by ground-based studies (e.g., Mintz and McWhinney, 2008), independent studies of oil sands mining emissions are particularly difficult because the mining operations are not accessible to the public (Timoney and Lee, 2009). Instead, the majority of oil sands studies are reported in the so-called "grey literature", consisting of discipline-specific reports, industrial monitoring reports, reports by industrially-controlled consortia, and reports commissioned by non-governmental organizations (Timoney and Lee, 2009). Air quality in the oil sands surface mining air shed is monitored locally by the Wood Buffalo Environmental Association (WBEA), which is a multi-stakeholder organization that represents industry, environmental groups, government, communities, and Aboriginal stakeholders. Longterm monitoring data and reports are available at the WBEA website (www.wbea.org) but these data have not been published in the peer-reviewed literature.
We present independent observations of 76 speciated C 2 -C 10 VOCs, CH 4 , sulphur dioxide (SO 2 ), carbon dioxide (CO 2 ), carbon monoxide (CO), ozone (O 3 ), nitric oxide (NO), nitrogen dioxide (NO 2 ) and total reactive nitrogen (NO y = NO + NO 2 + HNO 3 + PANs + other organic nitrates + HO 2 NO 2 + HONO + NO 3 + 2×N 2 O 5 + particulate NO − 3 + . . . ) near surface oil sands mining and upgrading operations in Alberta's Athabasca region. Although present in trace quantities in the Earth's atmosphere, these gases drive the atmosphere's chemistry and radiative balance (Forster et al., 2007 and references therein). For example hydrocarbons and nitrogen oxides (NO x = NO + NO 2 ) are key atmospheric constituents that can react together in the presence of sunlight to form tropospheric O 3 , itself a greenhouse gas and air pollutant. Many hydrocarbons (e.g., benzene, toluene) can also be toxic or carcinogenic. Sulphur dioxide is produced by industrial processes including petroleum combustion and can contribute to photochemical smog and acid rain. Our measurements were made on 10 July 2008 during a 17 min boundary layer (BL) flight leg over the Alberta oil sands using the NASA DC-8 aircraft as a research platform (http:// airbornescience.nasa.gov/platforms/aircraft/dc-8.html). The VOCs were collected in seventeen 45 s integrated whole air samples that were subsequently analyzed at the University of California, Irvine (UC-Irvine). The remaining trace gases were sampled continuously, with CO 2 , CH 4 , CO, NO, NO 2 , NO y and O 3 reported every 1 s and SO 2 reported every 30 s. Although clearly limited in temporal and spatial extent, to our best knowledge these measurements represent the first independent characterization of speciated VOCs and many other trace gases from oil sands mining in the peer-reviewed literature.

Experimental
Boundary layer air was sampled over the Athabasca surface mines as part of the 2008 Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) field mission (www-air.larc.nasa.gov/missions/arctas/arctas. html). The summer deployment of ARCTAS was based in Cold Lake, Alberta (54 • 25 N; 110 • 12 W) and included eight 8-h science flights (Flights 17-24) from 29 June-13 July 2008. Although the major focus of the summer phase of ARCTAS was boreal biomass burning emissions (Jacob et al., 2009), on 10 July 2008 (Flight 23) there was an opportunity for the DC-8 to make two descents into the BL over northeast Alberta as part of a return transit flight from Thule, Greenland to Cold Lake (Fig. 2). During the first BL excursion (Leg 7) the DC-8 circled over the Athabasca oil sands mining area between 11:27-11:44 local time, at altitudes between 720-850 m and within an area bounded by 56 • 34 -57 • 09 N and 111 • 01 -111 • 50 W. The aircraft flew over both boreal forest and cleared industrial land including tailings ponds, tailings sand and upgrader facilities (Fig. 2b). Ten-day backward trajectories show that during Leg 7 the air masses arrived at the aircraft's pressure level from the west (Fig. 3a). The second BL run was flown shortly after, from 12:00-12:15 local time at altitudes between 980-1410 m (Leg 9). Leg 9 was an intercomparison flight leg between the DC-8 and the NASA P-3B aircraft that occurred in generally clean air approximately 1 • further south (55 • 39 -56 • 16 N and 112 • 44 -113 • 47 W), i.e., not over the oil sands. The ten-day backward trajectories for Leg 9 show that the sampled air masses also arrived from the west and not from the oil sands mines to the north (Fig. 3b). Therefore, even though Leg 9 occurred at higher altitudes within the BL, we believe it provides a reasonable concurrent local background against which to compare the oil sands VOC enhancements (see additional discussion in Sect. 3.2.3).

Airborne whole air sampling
UC-Irvine has measured speciated VOCs from diverse environments for more than 30 years, using both ground-based and airborne platforms (e.g., Rowland, 1988, 1995;Colman et al., 2001;Blake et al., 2003Blake et al., , 2008Katzenstein et al., 2003;Barletta et al., 2005Barletta et al., , 2009Simpson et al., 2002Simpson et al., , 2006. Our sampling technique collects whole air samples (WAS) into 2-L electropolished, conditioned stainless ) Flight track of the NASA DC-8 aircraft during Flight 23, a transit flight from Thule, Cold Lake, Alberta on July 10, 2008. The DC-8 circled within the boundary layer over for 17 min (Leg 7) then descended back into background boundary layer air 16 min . (b) Detail of the flight path during Leg 7. The locations of selected samples (in which lues were measured) are highlighted in yellow. The prevailing wind direction was from t quadrant, and samples 4, 5 and 6 were collected directly downwind of the oil sands he grey patch just above the center of the figure). from Thule, Greenland to Cold Lake, Alberta on 10 July 2008. The DC-8 circled within the boundary layer over the oil sands for 17 min (Leg 7) then descended back into background boundary layer air 16 min later (Leg 9). (b) Detail of the flight path during Leg 7. The locations of selected samples (in which maximum values were measured) are highlighted in yellow. The prevailing wind direction was from the southwest quadrant, and samples 4, 5 and 6 were collected directly downwind of the oil sands operations (the grey patch just above the center of the figure). steel canisters each equipped with a Swagelok Nupro metal bellows valve (Solon, OH). The electropolishing minimizes the surface area and any surface abnormalities, and the canister conditioning (baking the cans in humidified air at ambient pressure and 225 • C for 12 h) forms an oxidative layer on the interior surface that further passivates the canister walls. The DC-8 payload included 168 of our 2-L air sampling canisters for each flight of the ARCTAS mission.
To prepare the canisters for field use, they are taken to the University of California Crooked Creek Station in the Sierra Nevada mountains (altitude = 10 200 ft) for a pumpand-flush procedure that is repeated ten times, in which each canister is pressurized to 40 psig with ambient air and then vented to ambient pressure. Next the canisters are returned to our UC-Irvine laboratory where they are evacuated to 10 −2 Torr and then pressurized to 1000 Torr with ultra-high purity helium before a final evacuation to 10 −2 Torr (E2M12 dual-stage vacuum pumps, Edwards Vacuum, Wilmington, MA). Lastly each canister is humidified by adding ∼17 Torr of purified water (the approximate vapour pressure of water at room temperature) to minimize surface adsorption and improve the reproducibility of our analytical split ratios during laboratory analysis (see Sect. 2.1.2). Rigorous sensitivity tests have shown that alkene growth in our passivated canisters is limited to a maximum of 0.2 pptv day −1 and all other compounds reported here are stable over the typically short period that the cans are stored before analysis (Sive, 1998). By analyzing the Alberta samples within 7 days of collection, we were able to limit the size of any alkene artefact to 1.4 pptv or less (i.e., to negligible values).
During each flight the whole air sampling is manually controlled and uses a stainless steel dual head metal bellows pump (MB-602, Senior Aerospace Metal Bellows, Sharon, MA) that is configured in series to draw outside air into a window-mounted 1/4 forward-facing inlet, through our air sampling manifold, and into one of the 168 canisters until it is filled to 40 psig. During Leg 7 we collected seventeen 45second air samples, or approximately one sample per minute. During Leg 9 we collected 1-min integrated air samples every 3 min, for a total of 6 samples.

Laboratory analysis of VOCs
After Flight 23 the whole air samples were returned to our UC-Irvine laboratory for analysis using three gas chromatography (GC) ovens coupled with a suite of detectors that together are sensitive to the 76 C 2 -C 10 VOCs that we seek to measure. We use two flame ionization detectors (FIDs) to measure hydrocarbons, two electron capture detectors (ECDs) for halocarbons, and a quadrupole mass spectrometer detector (MSD) for sulphur compounds.
Complete analytical details are given in Colman et al. (2001). For each sample a 1520 cm 3 sample aliquot is introduced into the analytical system's manifold and passed over glass beads contained in a loop maintained at liquid nitrogen temperature. A mass flow controller (Brooks Instrument; Hatfield, PA; model 5850E) keeps the flow below 500 cm 3 min −1 to ensure complete trapping of the less volatile sample components (e.g., VOCs) while allowing more volatile components (e.g., N 2 , O 2 , Ar) to be pumped away. The less volatile species are re-volatilized by immersing the loop in hot water (80 • C) and are then flushed into a helium carrier flow. The sample flow is split into five streams, with each stream chromatographically separated on an individual column and sensed by a single detector, namely: (1)  and connected to an ECD; and (5) a DB-5ms fused silica capillary column (J&W Scientific; 60 m, I.D. 0.25 mm, film 0.5 mm) connected to an MSD. The split ratios are highly reproducible as long as the specific humidity of the injected air is above 2 g-H 2 O/kg-air, which we ensure by adding purified water into each canister (Sect. 2.1.1). The signal from each FID and ECD is output to a personal computer and digitally recorded using Chromeleon Software; the MSD output signal uses Chemstation software. To ensure that the measurements are of the highest calibre, each peak of interest on every chromatogram is individually inspected and its integration is manually modified. For Flight 23 more than 10 000 peaks were hand-modified.
Calibration is an ongoing process whereby new standards are referenced to older certified standards, with appropriate checks for stability and with regular inter-laboratory comparisons. The hydrocarbons standards are NIST-traceable and the halocarbon standards are either NIST-traceable or were made in-house and have been compared to standards from other groups such as NOAA/ESRL. Multiple standards are used during sample analysis, including working standards (analyzed every four samples) and absolute standards (analyzed twice daily). Here we used working standards collected in the Sierra Nevada mountain range. International intercomparison experiments have demonstrated that our analytical procedures consistently yield accurate identification of a wide range of blindly selected hydrocarbons and produce excellent quantitative results (e.g., Apel et al., 1999Apel et al., , 2003. Table 1 shows the limit of detection (LOD) and the measurement precision and accuracy for each VOC. Even though our LOD is conservative, the accuracy and precision of many species does deteriorate as we approach this limit.
2.2 SO 2 , CO 2 , CH 4 , CO, NO, NO 2 , NO y and O 3 In addition to VOCs measured by UC-Irvine, we also present SO 2 , CO 2 , CO, CH 4 , NO, NO 2 , NO y and O 3 mixing ratios, which were measured in situ aboard the DC-8 by four research teams each using fast-response, high precision, continuous real-time instruments. The measurement precision and accuracy of these compounds are given in Table 1. The time response is 1 s for all compounds except NO and NO 2 , which have a 3 s time response (SO 2 has a 1 s time response but was reported as a 30 s average). Briefly, SO 2 was measured using the Georgia Tech Chemical Ionization Mass Spectrometer (GT-CIMS) instrument, which uses SF − 6 ion chemistry to selectively ionize SO 2 . Carbon dioxide CO 2 was measured using the NASA Langley Atmospheric Vertical Observations of CO 2 in the Earth's Troposphere (AVOCET) instrument, which uses a modified Li-Cor model 6252 differential, non-dispersive infrared (NDIR) gas analyzer at the 4.26 µm CO 2 absorption band (Vay et al., 1999. Methane and CO were measured by the NASA Langley Differential Absorption CO Measurement (DACOM) instrument, which uses two tunable diode lasers in the infrared spectral region to simultaneously measure the absorption of light by CH 4 (3.3 µm) and CO (4.7 µm) (Fried et al., 2008). Nitric oxide, NO 2 , NO y and O 3 were measured using the 4-channel NCAR NO xy O 3 chemiluminescence instrument (Weinheimer et al., 1994).

Results and discussion
Mixing ratio time series for Flight 23 are shown for selected species in Figs. 4 and 10, and altitudinal profiles are shown for many measured species in Figs. 5-9. Values that are below their LOD have been given a value of "0" so that they are visually represented on the graphs. Measurement statistics for Flight 23 -including the BL excursions over the oil sands mining operations (Leg 7, n = 17) and in background air (Leg 9, n = 6) -are given in Table 1. For comparison, statistics for free troposphere (FT) measurements made earlier in the flight are also shown (60-76 • N; n = 66). Results for individual samples collected during Leg 7 (Table 2) show that not all of the Leg 7 samples were influenced by emissions from the oil sands because some samples were collected south of the mining operations as the plane manoeuvred (Fig. 2b). For many compounds their maximum enhancements were measured directly downwind of the Syncrude Mildred Lake Facility (i.e., samples 4, 5 and 6 in Table 2 and Fig. 2b). For other compounds such as isoprene, Table 1. Statistics of boundary layer measurements for 84 compounds measured near oil sands surface mining north of Fort McMurray, Alberta on 10 July 2008 (n = 17). Concurrent local background values in the boundary layer (Bkgd) are also included (n = 6), as are free tropospheric measurements (FT) collected between 60-76 • N (n = 66). Max. Enh. = Maximum enhancement of oil sands values over background values (oil sands max/bkgd avg) -when the background mixing ratio was below detection limit (LOD) a mixing ratio of 1.5 pptv was assumed in the enhancement calculations; Min = minimum; Max = maximum; Avg = average; StD = standard deviation; n/a = not applicable. Units: pptv unless otherwise stated. a Lifetimes of short-lived OH-controlled compounds are shorter (longer) during the summer (winter), when there are more (fewer) hydroxyl radicals (OH) available for oxidative reactions. b The VOC precision deteriorates as we approach our detection limit; at low values the precision is either the stated precision (%) or 3 pptv, whichever is larger. The NO, NO 2 , NO y and O 3 precision values are for high mixing ratios as were encountered over the oil sands; at low mixing ratios their precision is 20 pptv for the nitrogen species and 0.1 ppbv for O 3 . c Based on OH rate constants from Atkinson and Arey (2003) and assuming a 12-h daytime average OH radical concentration of 2.0 × 10 molec cm −3 . The lifetime estimates for furan and methylcyclopentane are from Atkinson et al. (2005) and also use a 12-h daytime OH value of 2.0 × 10 molec cm −3 . d The OH-lifetimes of m-xylene and p-xylene are 12.0 and 19.4 hours, respectively. e Total tropospheric lifetime based on Jacob et al. (2002).
f Total lifetimes based on Clerbaux et al. (2007). g CH 4 mixing ratios were not measured during Leg 9, and the background value was determined from the Cold Lake landing (see text).
h Ethanol data were not available for the first seven samples of Leg 7, and the statistics presented here are based on samples 8-17.
which has a dominant biogenic source, the maximum occurred further south over vegetation (i.e., sample 14). Because the Leg 7 samples ranged from near-background air to strongly polluted industrial plumes, the discussion below includes a "maximum enhancement" over the Leg 9 background average, based on each compound's maximum mixing ratio during Leg 7. For correlation purposes, the SO 2 , CO 2 , CH 4 , CO, NO, NO 2 , NO y and O 3 data are based on the average only of those 1 s measurements (30 s for SO 2 ) that overlapped the VOC sampling times (i.e., the so-called hydrocarbon data merge). The complete 1 s and 30 s data sets for these eight compounds, together with all the VOC data from the summer phase of ARCTAS, are available at ftp://ftp-air.larc.nasa.gov/pub/ARCTAS/DC8 AIRCRAFT/.

General features
Atmospheric trace gas mixing ratios are generally greater in the Earth's BL than above in the FT. Here we seek to quantify mixing ratio increases over the oil sands that exceed the increases that may occur as the aircraft descends into the boundary layer. The 84 trace gases presented here can be classified into three groups: (1) compounds that were strongly enhanced (>10%) over the oil sands relative to levels in the local background BL air; (2) compounds that showed minimal increases over the oil sands (<10% greater than the local background average); and (3) compounds that showed no statistical enhancements over the oil sands compared to the local background. The first group, easily the largest, includes SO 2 , NO, NO 2 , NO y , CO 2 , CH 4 , CO and 70% of the measured VOCs, namely alkanes, aromatics, cycloalkanes, alkenes, oxygenated hydrocarbons, ethyne, short-lived solvents (C 2 Cl 4 , C 2 HCl 3 , CHCl 3 , CH 2 Cl 2 ) and some HCFCs (HCFC-22 and HCFC-142b) (Sects. 3.2, 3.3.3 and 3.4-3.6). The maximum enhancements were 1.1-397× the background BL values, most notably nheptane (397×), SO 2 (382×), NO (319×), n-octane (242×), NO 2 (210×), m + p-xylene (181×) and methylcyclohexane (113×). Interestingly this group comprises α-pinene and βpinene, which are usually associated with biogenic emissions (Sect. 3.2.3). In the second group, halocarbons including HFC-134a, HCFC-141b, the halons and the methyl halides were minimally enhanced over the oil sands (Sect. 3.3.2).
In the third group, long-lived industrial halocarbons (e.g., CFCs, CCl 4 , CH 3 CCl 3 ), several brominated species (e.g., CHBr 3 , CH 2 Br 2 ) and ozone (O 3 ) were not enhanced over the oil sands (Sects. 3.3.1 and 3.5). In fact O 3 was relatively depleted because of titration by high levels of NO. Before discussing the results below, it is important to first recognize the limitations of this data set. Because of its short time-frame (17 min) and sample size (n = 17), the representativeness of the measured VOC enhancements is unclear. On the other hand, because emissions from oil sands mining are so poorly characterized in the peer-reviewed literature due to the inaccessibility of mining sites to independent observers, this study provides important close-range observations of the VOCs that are being released, i.e., which individual species are emitted and with what other species do they correlate. The large number of compounds and relatively small number of samples in this data set make it unsuitable for emission estimates or factor analysis such as Principal Component Analysis or Positive Matrix Factorization (e.g., Thurston and Spengler, 1985;Paatero, 1997;Choi et al., 2003), and we have instead performed linear correlations among the measured compounds using least squares fits to better understand their source influences.

Non-methane volatile organic compounds (NMVOCs)
Non-methane volatile organic compounds (NMVOCs) are mainly emitted from three major sources: vegetation, biomass burning, and anthropogenic activities such as industry and fossil fuel production, distribution and combustion (e.g., Guenther et al., 2000;Ehhalt and Prather, 2001;Folberth et al., 2006). Remarkably, of the 48 C 2 -C 10 hydrocarbons that we measured, all but three (propyne, furan and methyl tert-butyl ether or MTBE) showed very strong enhancements over the oil sands (maximum values at least 80% more than the local background average). Based on their correlations with one another (Table 3), the NMVOCs appear to have two distinct sources: (1) the oil sands and its products and/or the diluent used to lower the viscosity of the extracted bitumen (enhanced levels of C 4 -C 9 alkanes, C 5 -C 6 cycloalkanes, C 6 -C 8 aromatics), and (2)  The boundary layer run over oil sands mining operations (Leg 7, shaded) is followed by a second boundary run (Leg 9) in cleaner air 1⁰ further south. Values below detection limit have been given a value of -0‖ for plotting purposes. Note that the mixing ratios of n-heptane, the cycloalkanes, the butenes and isoprene were below their detection limits for most of the flight. The boundary layer run over oil sands mining operations (Leg 7, shaded) is followed by a second boundary run (Leg 9) in cleaner air 1 • further south. Values below detection limit have been given a value of "0" for plotting purposes. Note that the mixing ratios of n-heptane, the cycloalkanes, the butenes and isoprene were below their detection limits for most of the flight.
that support the mining effort, such as upgrading (elevated levels of C 2 -C 4 alkanes, C 2 -C 4 alkenes, C 9 aromatics). The major upgrading processes at this location are distillation (to separate different hydrocarbons), thermal conversion/coking (to convert the bitumen into lighter, refinable hydrocarbons), catalytic conversion (an enhanced form of thermal conversion), and hydrotreating (the addition of hydrogen to unsaturated molecules to stabilize them). For example the catalytic hydrocracking of Athabasca bitumen vacuum bottoms (AB-VBs) produces gaseous by-products including hydrogen sulphide (H 2 S), C 1 -C 7 alkanes and, to a lesser extent, C 2 -C 4 alkenes (Dehkissia et al., 2004).
( Table 1). The ≥C 7 n-alkanes are so short-lived (<2 days) that they were undetectable in the background BL (<3 pptv). By contrast, all 12 C 2 -C 9 alkanes were strongly enhanced in the oil sands plumes. Maximum ethane (1490 pptv) and propane (714 pptv) levels occurred in sample 4 downwind of the Syncrude Mildred Lake Facility (Fig. 2b) and were almost double and quadruple the background BL values, respectively (Table 1; Fig. 5a-b). Despite minimal urbanization in northern Alberta, the oil sand plume values were greater than average summertime values measured in rural New England from -2007. Instead they were comparable to average ethane and propane levels measured in urban areas of Baltimore and New York City during a summer study of 28 US cities from 1999-2005 . By comparison, the ethane and propane values in the oil sands plumes were smaller than median values measured during summer 2006 in the Houston and Galveston Bay area, which is both a large urban area and a major petrochemical manufacturing center (4407 pptv ethane; 2713 pptv propane) (Gilman et al., 2009). The primary ethane and propane sources are fossil fuel production (mainly unburned gas in the case of ethane) and biofuel and biomass burning, plus propane is also emitted in gasoline exhaust (e.g., Watson et al., 2001;Buzcu and Fraser, 2006;Xiao et al., 2008). Ethane and propane correlated well with a range of industrial and combustion species such as other light alkanes, alkenes, NO, NO 2 , NO y and ethyne (Table 3). By contrast, even though ethane is the second most abundant component of natural gas after CH 4 (Xiao et al., 2008;McTaggart-Cowan et al., 2010), it did not correlate well with CH 4 (r 2 = 0.52). This relatively poor correlation is surprising given the heavy use of natural gas during bitumen extraction (Sect. 1) and appears to indicate low natural gas leakage levels (see Sect. 3.6). Because ethane and propane are primarily associated with evaporative rather than combustive fossil fuel emissions, we expect a co-located evaporative source from the industries that process the oil sands, possibly from fuel gas and/or hydrocracking.
The butanes are associated with fossil fuel evaporation (e.g., liquefied petroleum gas or LPG) and biomass burning (Blake et al., 1996;Watson et al., 2001;Buzcu and Fraser, 2006). The i-butane (290 pptv) and n-butane (765 pptv) maxima in the oil sands plumes were 10-12× greater than in the background BL (Table 1; Fig. 5c) and were comparable to maximum levels measured in fresh biomass burning plumes during ARCTAS (not shown). They were also similar to the lower range of values measured in the 28 US city study , and they were much greater than average July values measured in rural New England (70 ± 39 pptv and 105 ± 59 pptv, respectively; Russo et al., 2010), even though there is widespread LPG leakage in the northeastern United States (White et al., 2008;Russo et al., 2010). The ratio of i-butane/n-butane varies according to source, for example 0.2-0.3 for vehicular exhaust, 0.46 for LPG, and 0.6-1.0 for natural gas (Russo et al., 2010 and references therein). Interestingly, the i-butane/n-butane ratio for the dominant oil sands plumes, i.e., samples 4, 5 and 6 (0.42 ± 0.03 pptv pptv −1 ) was very close to that expected for LPG. We have not been able to determine that LPG is used in the mining operations, and this signal could potentially represent emissions from hydrocrackers and/or the fuel gas that is produced and used at the major oil sands operations. Even though they only contain 4 carbon atoms, the butanes correlated better with the suite of heavier compounds that is associated with direct emissions from the oil sands and/or diluent (e.g., n-hexane, xylenes, cyclopentane; 0.93 ≤ r 2 ≤ 0.96) than with species linked to industrial activities associated with the mining operations (e.g., propane, CHCl 3 , ethene, ethyne, NO, NO y ; 0.77 ≤ r 2 ≤ 0.84) ( Table 3). Because the butanes are not generally associated with combustive fossil fuel emissions, these results indicate multiple evaporative butane sources at the mining sites.
Crude oil is composed of ≥C 5 alkanes, aromatics, cycloalkanes and asphaltics (Sect. 1), and even though the oil sands are extra-heavy and include molecules with 50 or more carbon atoms, they are still expected to contain a certain fraction of volatiles that would be captured by our measurements (D. Spink, personal communication, 2010). The C 5 -C 9 n-alkanes showed very strong enhancements, with higher values over the oil sands than at any other time during the summer phase of ARCTAS. It is noteworthy that the maximum abundances were not ordered according to chain length. n-Heptane (596 pptv) showed the highest maximum mixing ratio, followed by the pentanes (510-560 pptv), n-octane (363 pptv) and n-hexane (294 pptv) (Table 1; Figs. 4c, 5df). These values represent enhancements of 24-397× the background BL values (in the enhancement calculations a background value of 1.5 pptv was assumed for BL measurements that were below our detection limit of 3 pptv). Oil having a plant origin is known to show a predominance for n-alkanes with an odd chain number, in contrast to n-alkanes that are produced from marine organic matter (Rogers and Koons, 1971). Because oil sands are derived from marine phytoplankton, the relative abundance of pentanes and nheptane may suggest a stronger signal from the diluent than from the oil sands, assuming the diluent is made from conventional plant-based oil. Branched alkanes such as 2,3dimethylbutane and the methylpentanes (which we present as a single measurement, i.e., 2+3-methylpentane) were also enhanced by factors of 19-34 over the local background (Table 1).
The cycloalkanes were close to or below our 3 pptv detection limit during Leg 9, as they had been in the FT during the transit from Thule. By contrast, all the cycloalkanes were strongly enhanced over the oil sands, with maximum mixing ratios ranging from 41 pptv (cyclopentane) to 339 pptv (methylcyclohexane) ( Table 1; Figs. 4e, 6a). Cycloalkanes are a major component of crude oil (Sect. 1), and these results are consistent with their evaporative release from oil sands surface mining operations.
Aromatics are associated with combustion (fossil fuel and biomass burning), fuel evaporation, industry, and biogenic emissions (e.g., Karl et al., 2009a;White et al., 2009). With the exception of the longest-lived aromatics that we measured -i.e., benzene (1-2 weeks) and toluene (2-3 days)none of the aromatics were detectable in either the FT or in background BL air, whereas they were all strongly elevated over the oil sands with maximum enhancements of 7-181× over the background BL (Table 1; Figs. 4e-f, 6b-d). Toluene was the most abundant aromatic in the oil sands plumes (401 pptv maximum) followed by the xylenes (127-272 pptv maxima) and benzene (82 pptv maximum). For comparison these aromatic enhancements are much smaller than the maximum values measured near major petrochemical complexes in urban/industrial areas of Houston, Texas and southern Catalonia, Spain (e.g., toluene maxima of 16 000-77 000 pptv; Gilman et al., 2009;Ras et al., 2009). That is, the release of aromatics from the oil sands mining sites appears to be much smaller than releases further downstream during petrochemical refining.  With the exception of the C 9 aromatics, the ≥C 5 alkanes, cycloalkanes and aromatics showed excellent mutual correlations (Table 3). Interestingly, the best correlations were not limited to compounds within the same class but were mixed among the compound classes. For example, n-heptane, the most strongly enhanced VOC, showed top 10 correlations with n-nonane, n-octane, toluene, methylcyclopentane, nhexane, benzene, cyclohexane, 2,3-methylpentane, m + pxylene, and ethylbenzene (0.95 ≤ r 2 ≤ 0.99). These results show the fugitive co-emissions of a wide range of C 5 -C 9 VOCs from oil sands surface mining sites in Alberta, e.g. from oil sands and/or diluent. By contrast, the C 9 aromatics (i.e., ethyltoluenes, trimethylbenzenes) correlated best with each other and with combustion and industrial tracers such as NO, NO 2 , NO y , ethene, propene and cis-2-butene, suggesting their release from industries at the mining sites as opposed to direct evaporative release from the oil sands or diluent.

Alkenes and alkynes
Because they are primarily biogenic tracers, isoprene and the monoterpenes are considered separately below (Sect. 3.2.3). The 7 C 2 -C 4 alkenes considered here are highly reactive, short-lived compounds that are primarily associated with industrial emissions and incomplete combustion (e.g., Sprengnether et al., 2002;Buzcu and Fraser, 2006;. Whereas only ethene was detectable in the background BL during Flight 23 (20 ± 2 pptv), all of the alkenes were strongly enhanced over the oil sands by factors of up to 5-85 (Table 1; Figs. 4g, 6e). The maximum mixing ratio decreased with increasing chain length, from ethene (270 pptv) to propene (128 pptv) to i-butene (37 pptv), etc. Although these oil sands values are much greater than the rural background, they are smaller for example than in fresh biomass burning plumes encountered during ARCTAS (e.g., ethene up to 18 690 pptv, propene up to 5465 pptv). They are also in the low range of summertime values in US cities (e.g., 260-2430 pptv ethene; 68-500 pptv propene; 33-1550 pptv i-butene; Baker et al., 2008). In the Houston area -where the greatest concentration of petrochemical facilities in the US is located and ethene and propene play a major role in rapid O 3 formation (Ryerson et al., 2003) -average respective ethene and propene mixing ratios of 2690 pptv and 1540 pptv were measured in the summer of 2006 (Gilman et al., 2009), and poor correlation between ethene and CO suggested the dominance of industrial point sources . Over the oil sands, ethene showed excellent correlation with the combustion tracer ethyne (r 2 = 0.96), some correlation with CO (r 2 = 0.60), and good correlation with the industrial tracer C 2 Cl 4 (r 2 = 0.79) ( Table 3), suggesting the mixed influence of industrial and combustion sources on ethene and other alkenes at the oil sands mining sites.
Propyne was not detectable in any of the samples that we collected during Flight 23. Ethyne -a tracer of incomplete combustion by biomass burning and urban fossil fuel (e.g., Blake et al., 2003; -correlated most strongly with ethene, NO y , NO and the trimethylbenzenes (0.91 ≤ r 2 ≤ 0.96) and showed a maximum mixing ratio of 138 pptv over the oil sands, compared to 59 ± 5 pptv in the local background (Table 1; Fig. 6f). As with the alkenes, this enhancement is small compared to average levels measured in Houston (473 pptv; Gilman et al. 2009) and in the 28 US city study (260-2390 pptv;Baker et al., 2008), indicating the relatively low impact of industrial combustion on the measured ethyne levels.

Isoprene and monoterpenes
Isoprene (C 5 H 8 ) is a short-lived (∼2.8 h) biogenic volatile organic compound (BVOC) with a major terrestrial plant source (e.g., Sharkey et al., 2008;Warneke et al., 2010). Isoprene levels vary widely in North America, with greatest emission rates in the eastern US during summer (Guenther  , 2006). For example in the 28 US city study the lowest average isoprene mixing ratio was measured in El Paso, Texas (47 ± 75 pptv) and the highest in Providence, Rhode Island (2590 ± 1610 pptv) . In our study, isoprene mixing ratios exceeded 200 pptv in all 17 air samples collected over the oil sands as well as in the background BL air. Interestingly, the average isoprene mixing ratio was 50% greater during Leg 7 (468 ± 147 pptv) than in the local background air (311 ± 95 pptv), with a maximum value of 780 pptv (Tables 1 and 2; Fig. 4h). Anthropogenic sources of isoprene include traffic (e.g., Reimann et al., 2000;Borbon et al., 2001;Barletta et al., 2002) and industry (e.g., Ras et al., 2009). However the isoprene maximum was measured in sample 14, south of the main surface mining operations (Fig. 2b). Isoprene was poorly correlated with all of the trace gases that we measured (r 2 < 0.37) and its enhancements appear to be biogenic rather than industrial in origin. We believe that the stronger isoprene enhancements during Leg 7 than Leg 9 most likely reflect natural source strength variability rather than an altitude effect. Isoprene is highly reac- tive with OH and it is the shortest-lived species that we measure (Table 1). Although the Leg 7 samples were collected at a lower average altitude (804 ± 38 m) than the background BL measurements (1135 ± 180 m) (Sect. 2) -and in principle more isoprene oxidation by OH could have occurred during transport to higher altitudes during Leg 9 -the individual measurements did not show a clear declining trend with altitude. Instead we expect that samples with higher isoprene levels represent parts of Leg 7 that were downwind of more, or more recent, isoprene emissions. For example aspen are a major isoprene emitter (Sharkey et al., 2008) and the forests in the Fort McMurray area include a mix of trembling aspen and white spruce. Like isoprene, monoterpenes (C 10 H 16 ) are short-lived (∼3.7-5.3 h) BVOCs that are emitted by vegetation, though in this case they are more strongly emitted by coniferous ecosystems than by temperate deciduous forests (Fuentes et al., 2000). During this study the pinenes were enhanced during both BL runs compared to measurements in the FT (Fig. 7a-b). Interestingly, however, the maximum α-pinene and β-pinene mixing ratios over the oil sands (217 pptv and I. J. Simpson et al.: Trace gases measured over Alberta oil sands mining operations 610 pptv, respectively) were much greater (7-11×) than their respective average values in the background BL air (20 ± 7 and 84 ± 24 pptv; Table 1, Fig. 7a-b). At first glance this appears to be consistent with a number of studies that have shown increased emissions of monoterpenes and other VOCs in response to various plant stress factors (Schade and Goldstein, 2003;Räisänen et al., 2008;Holopainen and Gershenzon, 2010;Niinemets, 2010). For example, monoterpene mixing ratios and emissions from a California ponderosa pine forest showed a 10-30 fold increase during and after major forest thinning, with measured mixing ratios exceeding 3000 pptv (Schade and Goldstein, 2003). However, unlike isoprene, the pinenes were most strongly enhanced in sample 4, downwind of the Syncrude Mildred Lake Facility (Table 2; Fig. 2b), and they showed strong correlations with many species that were elevated in the oil sands plumes. For example α-pinene correlated best with β-pinene (r 2 = 0.92) and industrial compounds such as C 2 Cl 4 (r 2 = 0.81), SO 2 (r 2 = 0.76), 1,2,3-TMB (r 2 = 0.74), i-butene (r 2 = 0.73) and ethyne (r 2 = 0.73). The land surrounding the upgraders is already disturbed and we do not expect the industrial upgrader plumes to have had the opportunity to mix with plumes representing damaged vegetation. Instead these unexpected observations lead us to infer that there could be a source of pinenes associated with the oil sands mining industry, which requires further investigation.
Methanol has a major biogenic source and minor sources including biomass burning and anthropogenic emissions (e.g., vehicles, solvent use and manufacturing; Jacob et al., 2005 and references therein). Its maximum mixing ratio (3570 pptv) was measured downwind of the oil sands operations in sample 6, which was enhanced by a factor of 1.8 over the local background average (Table 1; Fig. 7c). Methanol correlated best with MEK and the butenes (0.71 ≤ r 2 ≤ 0.75) suggesting its industrial release from the oil sands facilities. However the oil sands industry appears to be a relatively minor methanol source. For example, the mixing ratios in the oil sands plumes were comparable to those measured at higher altitudes and latitudes during earlier portions of Flight 23, and they were an order of magnitude smaller than maximum methanol values that were measured in fresh biomass burning plumes during ARCTAS (32 740 pptv).
In addition to being produced by plants and used as a solvent and chemical feedstock, ethanol is increasingly being used as a motor fuel and fuel additive (e.g., Kesselmeier and Staudt, 1999;Russo et al., 2010). The maximum ethanol mixing ratio during Leg 7 (141 pptv) was 1.9× the background average, and ethanol correlated best with CH 3 CCl 3 and the halons (0.65 ≤ r 2 ≤ 0.83). However, unlike the other gases that we measured, ethanol data were not available for the first seven samples of Leg 7 and therefore we are unable to fully characterize its emissions from the oil sands industry.
Acetone is produced by vegetation, the oceans and the atmospheric oxidation of C 3 -C 5 isoalkanes, especially propane (Jacob et al., 2002). It is used industrially as a solvent and polymer precursor and has been observed in abundance downwind of petrochemical complexes (Cetin et al., 2003). The maximum acetone mixing ratio over the oil sands (941 pptv) was 1.8× its local background average. Like methanol, the acetone levels over the oil sands were comparable to other times during the flight and were smaller than maximum values measured in fresh biomass burning plumes during ARCTAS (4552 pptv). Acetone correlated best with MAC, 1,2-dichloroethane and CH 2 Cl 2 (0.64 ≤ r 2 ≤ 0.69) suggesting that it is emitted by the oil sands industry but in smaller amounts than other regional sources such as biomass burning.
Methyl ethyl ketone is produced industrially and used as a solvent, and it is known for its characteristic sweet odour (Kabir and Kim, 2010). The maximum MEK mixing ratio over the oil sands (214 pptv) occurred in sample 4 and was 11× greater than the local background average. However most of the MEK enhancements over the oil sands were comparable to those measured at other times during the flight (Fig. 7d). Methyl ethyl ketone showed strong correlations (r 2 > 0.8) with a dozen industrial and combustion tracers including C 2 -C 4 alkenes, C 2 Cl 4 , NO and 1,2,3-TMB, showing its release from the oil sands industry in what appear to be relatively minor quantities.
Methacrolein and MVK are major isoprene oxidation products (e.g., Montzka et al., 1993;Stroud et al., 2001;Karl et al., 2009b). The MAC and MVK mixing ratios were greater over the oil sands than at any other time during the flight (Fig. 7e-f), with maximum respective values of 266 and 379 pptv in sample 7, or 8× and 6× the local background average (Table 1). Methacrolein and MVK correlated most strongly with each other (r 2 = 0.87), followed by species such as DMS and 1,2-dichloroethane (r 2 ≤ 0.69). Because the highest MAC and MVK mixing ratios occurred downwind of the Syncrude Mildred Lake Facility in samples 3-7, enhancements associated with the oil sands industry cannot be ruled out. However because their strongest correlations were with each other, we suggest that their primary local source is likely isoprene oxidation. The average isoprene mixing ratio during Leg 7 was 468 ± 167 pptv. Assuming MAC and MVK formation yields from OH and O 3 reactions in the range of 0.16-0.4 (Tani et al., 2010 and references therein), the observed MAC and MVK values during Leg 7 (92 ± 63 and 141 ± 109 pptv, respectively) appear to be within the range that can be explained by isoprene chemistry.
Methyl tert-butyl ether is a solvent and gasoline additive, and furan is an intermediate in the synthesis of industrial chemicals. The mixing ratio of MTBE remained below its detection limit of 3 pptv throughout Flight 23, showing that it is not being released by the oil sands industry. This is consistent with its declining use in North America because of concerns over groundwater contamination (e.g., Backer, 2010;Lu and Rice, 2010). Like MTBE, furan also remained below its 3 pptv detection limit throughout the flight, in contrast to a maximum mixing ratio of 2344 pptv in very fresh biomass burning plumes that were sampled during ARCTAS.

Halocarbons
Long-lived halocarbons (e.g., CFCs, HCFCs) contribute to stratospheric ozone depletion, and short-lived halocarbons (e.g., CHCl 3 , C 2 Cl 4 ) can be toxic or carcinogenic. Of the 26 C 1 -C 2 halocarbons that we measured, 6 showed both strong enhancements (≥10%) over the local background average and larger mixing ratios over the oil sands than at any other time during the Flight, 9 showed minimal enhancements over the local background average, and 11 were not statistically enhanced over the oil sands.

Strongly enhanced halocarbons
The first group includes the CFC-substitutes HCFC-22 and HCFC-142b, and the relatively short-lived (2-5 mo) industrial solvents chloroform (CHCl 3 ), trichloroethene (C 2 HCl 3 ), tetrachloroethene (C 2 Cl 4 ) and dichloromethane (CH 2 Cl 2 ). HCFC-22 and HCFC-142b are long-lived refrigerants, foam-blowing agents and solvents whose atmospheric concentrations are rapidly increasing Derwent et al., 2007;Montzka et al., 2009). Their mixing ratios over the oil sands were greater than at any other time during the flight, with maximum values of 212 pptv HCFC-22 and 22 pptv HCFC-142b -or respective enhancements of 12% and 18% over the average background (Table 1; Fig. 8a) -showing their usage and release at the mining sites.
The short-lived solvents CHCl 3 , C 2 HCl 3 , C 2 Cl 4 and CH 2 Cl 2 also showed stronger enhancements over the oil sands than during any other portion of the flight (Figs. 4b,  8b). In fact, the maximum C 2 HCl 3 and C 2 Cl 4 levels (4.8 pptv and 5.9 pptv, respectively; Table 1) were higher in the oil sands plumes than at any other time during the summer phase of ARCTAS. For perspective, these plume values are smaller than average levels measured in urban centers such as Hong Kong, where C 2 HCl 3 and C 2 Cl 4 mixing ratios of 70 ± 31 and 29 ± 9 pptv were measured during autumn 2007 (Zhang et al., 2010). Maximum respective CH 2 Cl 2 and CHCl 3 mixing ratios were 35.1 pptv and 15.6 pptv, or 22% and 45% greater than the background average. Dichloromethane, C 2 Cl 4 , and to a lesser extent CHCl 3 , were also enhanced relative to background mixing ratios during the descent into Cold Lake, showing their use as solvents in local urban areas as well as by the mining industry.

Minimally enhanced halocarbons
The second group of compounds, which were minimally enhanced over the oil sands, includes 1,1,1,2-tetrafluoroethane (HFC-134a), some chlorinated species (1,2-dichloroethane, HCFC-141b), the halons (H-1211, H-1301, H-2402) and the methyl halides (CH 3 Br, CH 3 I, CH 3 Cl). HFC-134a is a CFC-12 replacement that is rapidly increasing in the atmosphere Clerbaux and Cunnold, 2007). HFC-134a levels of up to 60 pptv were measured in U.S. pollution plumes (Barletta et al., 2009) compared to a maximum of 48.9 pptv in the oil sands plumes, or 8% greater than the background average, suggesting that the oil sands mining industry is a relatively minor HFC-134a source (Table 1, Fig. 8c). 1,2-Dichloroethane is used as a chemical intermediate and solvent, and its maximum mixing ratio over the oil sands (9.4 pptv) was 16% greater than the background BL average but within the range of mixing ratios measured at other times during the flight, also suggesting relatively small emissions by the oil sands industry (Table 1). Likewise, the CFC replacement compound HCFC-141b showed fairly small (<7%) enhancements over the oil sands (Table 1). The halons are ozone-depleting substances that are still used in North America because of their critical role as fire extinguishing agents (Butler et al., 1998;Fraser et al., 1999;Montzka et al., 2003). Maximum H-1211, H-1301 and H-2402 levels over the oil sands were 5-10% greater than the local background average (Table 1), compared to average H-1211 enhancements of 75% in cities in China ( Barletta et al., 2006) where 90% of global H-1211 production occurs (Fraser et al., 1999). Therefore the halon enhancements over the oil sands also appear to be relatively minor, though we note that the H-1301 maximum over the oil sands was greater than at any other time during the summer phase of ARCTAS.
Methyl bromide, an ozone-depleting substance with natural and anthropogenic sources (e.g., oceans, salt marshes, fumigation and biomass burning; Yvon-Lewis et al., 2009 and references therein), had a maximum mixing ratio of 8.3 pptv both over the oil sands and upon final descent into Cold Lake. This 8% enhancement over the local background average was small compared to CH 3 Br mixing ratios of up to 15 pptv in fresh biomass burning plumes that were sampled during ARCTAS, suggesting a minor impact of the oil sands industry on CH 3 Br levels in the sampled plumes.
Methyl iodide is a very short-lived species (1-2 d) with a major oceanic source (Yokouchi et al., 2008) and minor terrestrial sources (e.g., Redeker et al., 2003;Sive et al., 2007). Its mixing ratio declined sharply with altitude, consistent with its short atmospheric lifetime (Fig. 8d). In fact, the shape of the profile and the magnitude of the mixing ratios were remarkably similar to those measured elsewhere over North America during the Intercontinental Chemical Transport Experiment-North America (INTEX-NA) campaign in July andAugust, 2004 (Sive et al., 2007). In addition, the mixing ratios of CH 3 I during Legs 7 and 9 (<0.45 pptv) were at the low end of global background values (0.5-2 pptv; Yokouchi et al., 2008 and references therein) and CH 3 I did not correlate with other species over the oil sands (r 2 ≤ 0.37). Therefore the oil sands do not appear to be a significant CH 3 I source. The maximum CH 3 I mixing ratio was measured in the final sample of Flight Leg 7 (sample 17) and the enhancements of CH 3 I during Leg 7 may be attributable to methyl iodide's terrestrial source.
Methyl chloride is the most abundant chlorine-containing compound in the atmosphere. It has a complex budget with multiple sources (e.g., tropical vegetation, biomass burning) and sinks (e.g., OH, soil) (Clerbaux et al., 2007). During the first half of the flight the average CH 3 Cl mixing ratio in the FT was 530 ± 13 pptv. It then decreased to 503 ± 6 pptv in the background BL, consistent with CH 3 Cl removal at low altitude by its soil sink. By contrast, the average CH 3 Cl mix-ing ratio over the oil sands (522 ± 12 pptv) was enhanced compared to the local background, with a maximum of 545 pptv and a range of values similar to the FT (Fig. 8e). Whereas CH 3 Cl showed poor correlation with most compounds over the oil sands (r 2 < 0.34), including combustion tracers such as CO and ethyne (r 2 < 0.01), it showed some correlation with OCS and HCFC-22 (r 2 = 0.58 for both) followed by HCFC-142b and HFC-134a (0.46 ≤ r 2 ≤ 0.47). Carbonyl sulphide has a strong vegetative sink (Sect. 3.4) and Montzka et al. (2007) found some correlation between OCS and other species with known surface sinks such as CH 3 Cl. Therefore the CH 3 Cl enhancement is expected to be related to the loss of its soil sink at the mining site, with the possibility of its co-emission from the oil sands industry together with species such as HCFC-22.

Non-enhanced halocarbons
Carbon tetrachloride (CCl 4 ), methyl chloroform (CH 3 CCl 3 ) and the chlorofluorocarbons  were not statistically enhanced over the oil sands (Table 1; Fig. 4a). These long-lived compounds have been banned in developed countries since 1996 under the Montreal Protocol, and even though evidence exists for their continued emissions in North America (Millet and Goldstein, 2004;Hurst et al., 2006) their absence from the oil sands mining industry is not surprising. Ethyl chloride (C 2 H 5 Cl), another chlorinated hydrocarbon whose industrial demand is declining, also was not enhanced over the oil sands. Likewise, with the exception of methyl bromide (CH 3 Br), none of the measured brominated species (CHBr 3 , CH 2 Br 2 , CHBrCl 2 , CHBr 2 Cl) were enhanced over the oil sands (Table 1). These short-lived species are predominantly emitted from the ocean (e.g., Butler et al., 2007) and their lack of enhancement over the oil sands is not unexpected.

Sulphur species
Sulphur dioxide (SO 2 ) is strongly associated with fossil fuel combustion, especially coal and residential oil, and the sulphur content of crude petroleum can be up to 2% (Benkovitz et al., 1996). In the Athabasca oil sands developments, the major SO 2 emission sources are associated with the upgrading and energy production operations at the Suncor and Syncrude sites (Kindzierski and Ranganathan, 2006;D. Spink, personal communication, 2010). For example Suncor produces high sulphur fuel grade petroleum coke from oil sands at its Fort McMurray operations that is 5.7-6.8% sulphur on a dry basis (http://www.suncor.com/en/about/3408.aspx). Some of this coke is used to power its upgrading operations, emitting SO 2 into the atmosphere as a by-product, but most of the coke is buried as a waste product. Whereas Burstyn et al. (2007) measured monthly average SO 2 mixing ratios of 300-1300 pptv in rural locations of western Canada, and Kindzierski and Ranganathan (2006) measured a median outdoor SO 2 mixing ratio of 650 pptv in residential areas of Fort McKay (a small community 64 km north of Fort Mc-Murray, in proximity to the mining operations), SO 2 levels in this study were even smaller during most of the flight, averaging 17 ± 5 pptv in the FT during the first half of the flight and increasing to 102 ± 27 pptv upon descent into the background BL air (Table 1; Fig. 4d). However SO 2 showed remarkable enhancements over the oil sands, with a maximum mixing ratio of 38 730 pptv (38.73 ppbv) in sample 4, or 383× the background BL average. This value is similar to those measured in the urban/industrial environments of heavily polluted cities. For example, Talbot et al. (2008)  Here we attribute the elevated SO 2 levels to coke combustion. The very strong correlations between SO 2 and more than a dozen compounds (r 2 ≥ 0.9) -most notably the trimethylbenzenes (0.98 ≤ r 2 ≤ 0.99), NO, NO 2 and NO y (0.96 ≤ r 2 ≤ 0.97), the ethyltoluenes (0.90 ≤ r 2 ≤ 0.97) and ethene (r 2 = 0.92) -suggest that coke combustion may be associated with their enhancements as well.
Carbonyl sulphide (OCS) is the most abundant sulphurcontaining compound in the remote atmosphere (Ko et al., 2003 and references therein). Its sources include CS 2 oxidation, the oceans, biomass burning, coal burning and aluminum production (e.g., Watts, 2000;Kettle et al., 2002). Like CH 3 Cl, the average OCS mixing ratio was fairly constant in the FT during the first half of the flight (445 ± 19 pptv), then diminished with decreasing altitude to 413 ± 13 pptv in background BL air, consistent with its wellknown uptake by vegetation (e.g., Montzka et al., 2007;Blake et al., 2008;Campbell et al., 2008). By contrast OCS was not depleted over the oil sands, where its average mixing ratio (437 ± 26 pptv) was not significantly different than during the first half the flight (Table 1; Fig. 8f). Carbonyl sulphide did not correlate with most compounds over the oil sands (r 2 ≤ 0.38) but showed some correlation with the same suite of compounds as CH 3 Cl, including HCFC-22, HCFC-142b and HFC-134a (0.47 ≤ r 2 ≤ 0.58). Therefore the relative OCS enhancement over the oil sands compared to the background BL is most likely due to a lack of drawdown from the cleared land in the oil sands area, with the possibility of an OCS source associated with the oil sands industry. Note that even though previous work has shown similarities between OCS and CO 2 because of simultaneous uptake by photosynthetically active vegetation during the daytime Campbell et al., 2008;White et al., 2010), here OCS and CO 2 showed poor correlation (r 2 = 0.16) because CO 2 has a clear additional fossil fuel source at the oil sands developments (Sect. 3.6).
Dimethyl sulphide (DMS) is a short-lived sulphur species (1-2 days) with a major oceanic source and minor sources including vegetation and biomass burning (e.g., Watts et al., 2000;Gondwe et al., 2003;Meinardi et al., 2003). Like CH 3 I, DMS levels strongly increased with decreasing altitude. Dimethyl sulphide levels were below detection (<1 pptv) in the FT, increasing to 4.7 ± 0.8 pptv in the background BL. Its average mixing ratio doubled to 10.7 ± 3.9 pptv over the oil sands with a maximum enhancement of 18 pptv (Table 1). For comparison, DMS levels over productive oceanic regions are on the order of 100-250 pptv (e.g., Nowak et al., 2001). Dimethyl sulphide was most strongly enhanced in samples 4, 5 and 6 and correlated with a range of compounds, most strongly CO 2 (r 2 = 0.82) as well as H-1211, 1,2-dichloroethane and HCFC-142b (0.71 ≤ r 2 ≤ 0.74), indicating that its oil sands source is industrial.

NO, NO 2 , NO y and O 3
Like SO 2 , the major source of NO x is fossil fuel combustion (Benkovitz et al., 1996). Nitric oxide, NO 2 and NO y were all very strongly enhanced over the oil sands compared to their respective background values of 16 ± 6 pptv, 24 ± 11 pptv and 194 ± 33 pptv ( Table 1). The maximum NO, NO 2 and NO y mixing ratios occurred in sample 4 and were 4980, 4995 and 10 555 pptv, respectively, representing enhancements of 319×, 210× and 54× the local background (Table 1; Fig. 9a-b). Recall that these numbers are based on the average of those NO, NO 2 and NO y measurements that overlapped the VOC sampling times (Sect. 3). The maximum NO, NO 2 and NO y values over the oil sands based on 1 s measurements were even higher: 9545, 9205 and 21 800 pptv, respectively. The NO maximum was the highest recorded throughout the summer phase of ARCTAS, showing the very strong emissions of nitrogen oxides from the mining industry. The NO y levels in the oil sands plumes lie within the lower range of values measured in megacities such as Tokyo, Mexico City and Beijing, which can vary from 2000-200 000 pptv . The major NO x sources at the mining sites are (1) the upgraders, gaseous fuel fired boilers, heaters, and co-generation units for heat and power production, and (2) the heavy hauler mine fleets (D. Spink, pers. comm., 2010). Because NO, NO 2 and NO y correlated perfectly with each other (r 2 = 1.00) and very strongly with SO 2 , most C 2 -C 4 alkenes, and the C 9 aromatics (0.90 ≤ r 2 ≤ 0.99) (Table 3), we conclude that the observed emissions were from the upgraders, etc. rather than the heavy hauler fleet.
Ozone was not enhanced over the oil sands, and its maximum mixing ratio (31 ppbv) was the same as the average background BL value (31 ± 1 ppbv) (Table 1, Fig. 9c). In fact O 3 was anti-correlated with NO over the oil sands (Fig. 10) because the very strong NO emissions led to a titration of O 3 , a loss process which dominated over O 3 production on this short time-scale. Note that in Fig. 10 virtually all (∼95%) of the NO y is in the form of NO x .

CO 2 , CH 4 and CO
Carbon dioxide (CO 2 ) is the leading contributor to the enhanced radiative forcing of the atmosphere, followed by methane (CH 4 ) (Forster et al., 2007). Compared to the FT, CO 2 was relatively depleted in the background BL air during Flight 23 (Table 1, Fig. 9d), consistent with its summertime uptake by terrestrial vegetation (Erickson et al., 1996;Randerson et al., 1997Randerson et al., , 1999. By contrast CO 2 showed a clear enhancement over the oil sands, with a maximum mixing ratio of 389 ppmv that is outside the range of values measured during the rest of the flight. 14 C in CO 2 was depleted in these enhanced samples, indicating a fossil fuel CO 2 influence (not shown). Carbon dioxide correlated with a wide range of compounds associated with the mining industry, including DMS, combustion tracers such as i-butene, ethene, ethyne and NO y , and industrial tracers such as MEK, 2,3-dimethylbutane, CHCl 3 and 3-ethyltoluene (0.72 ≤ r 2 ≤ Methane is long-lived and well-mixed in the remote atmosphere, and its sources include biogenic sources (e.g., wetlands, ruminants, rice agriculture), biomass burning, and fossil fuel mining and burning (Denman et al., 2007). Its mixing ratio varied little in the FT during Flight 23, averaging 1836 ± 10 ppbv (Table 1; Fig. 9e). Unfortunately CH 4 measurements were not available during the background BL run (Leg 9). However CH 4 shows a characteristic north-south gradient (e.g., Simpson et al., 2002), and because Cold Lake is at a similar latitude to Leg 9 and does not appear to have been appreciably impacted by local CH 4 sources (Fig. 9e), the average CH 4 mixing ratio in the BL during the final descent into Cold Lake (1843 ± 5 ppbv) is expected to be a reasonable proxy for the local background at this time of year. Methane mixing ratios were clearly enhanced over the oil sands, reaching 1983 ppbv in sample 6. By comparison the maximum CH 4 mixing ratio during the summer phase of ARCTAS (2000 ppbv) was measured in a fresh biomass burning plume. Natural gas is heavily used by the oil sands mining industry (Sect. 1), but the CH 4 results are surprising in that CH 4 did not correlate particularly well with ethane and propane (0.52 ≤ r 2 ≤ 0.57), which, like CH 4 , are components of natural gas. The composition of commercialgrade natural gas is variable, ranging from 70-95% CH 4 , with the remainder primarily ethane (e.g., Xiao et al., 2008;McTaggart-Cowan et al., 2010). Although we are not aware of the composition of the natural gas that is typically used by the oil sands industry, a specs sheet for natural gas from western Canada suggests the following typical composition on a per mole percentage basis: 94.6% CH 4 , 2.5% ethane, 0.2% propane, 0.03% n-butane, etc. (http://www.naesb.org/pdf2/ wgq bps100605w2.pdf). This suggests a relative release rate of 38 moles of CH 4 to 1 mole of ethane, which is more ethane than we observed. The maximum CH 4 mixing ratio over the oil sands (sample 6) represented an excess of 140 ppbv over its local background value of 1843 ppbv. The ethane mixing ratio in sample 6 was 1.106 ppbv, representing an excess of 0.325 ppbv. If the CH 4 enhancement were entirely due to natural gas emissions, we would have expected to measure a corresponding ethane excess of 3.7 ppbv. Therefore the measured ethane enhancement was about 10× smaller than would be expected from natural gas emissions, which suggests low natural gas leakage levels from the mining operations. Similarly, hydrocracking operations do not appear likely to explain the observed CH 4 enhancements. The catalytic hydrocracking of Athabasca bitumen vacuum bottoms has been found to give a weight per cent yield of 9.3% CH 4 , 10.5% ethane, 15.0% propane and 9.6% C 4 H 10 (Dehkissia et al., 2004), or relative molar yields of roughly 4:2:2:1 for CH 4 :ethane:propane:butanes. Again, this represents much more ethane than we observed, and assuming these hydrocracking yields apply here and are equivalent to emission rates, it does not appear that hydrocracking can be responsible for the high CH 4 excesses that were observed. Instead we suggest that the CH 4 enhancements are likely related to tailings pond emissions. As stated in Sect. 1, the Syncrude Mildred Lake tailings pond became methanogenic in the 1990s and releases up to 10 g CH 4 m −2 d −1 . Ethane is not produced in anaerobic environments, and tailings pond emissions of CH 4 explain the relatively large CH 4 enhancements compared to ethane as well as the relatively poor correlation between CH 4 and ethane.
Carbon monoxide (CO) is a potentially toxic gas that is also a precursor to photochemical smog. The maximum CO mixing ratio over the oil sands (144 ppbv in sample 5, or 48% greater than the local background of 97 ± 1 ppbv) was comparable to mixing ratios measured at other times during the flight (Table 1; Fig. 9f) and is not considered a large enhancement. For example Baker et al. (2008) found typical summertime CO mixing ratios of 300 ± 66 ppbv in US cities, and CO mixing ratios measured near fresh biomass burning plumes during ARCTAS exceeded 1900 ppbv (not shown). Interestingly, CO correlated most strongly with the suite of C 4 -C 9 alkanes, C 6 -C 8 aromatics and C 5 -C 6 cycloalkanes (0.85 ≤ r 2 ≤ 0.97) that were associated with direct emissions from the oil sands and/or diluent (Sect. 3.2). It also showed good correlations with CHCl 3 (r 2 = 0.84) and the C 9 aromatics (0.66 < r 2 < 0.83). Carbon monoxide is an urban/industrial combustion tracer and these results suggest that relatively low levels of CO are emitted throughout the mining operations.

Conclusions
Mixing ratios of 84 trace gases were measured in boundary layer air over oil sands surface mining operations in northern Alberta on 10 July 2008 aboard the NASA DC-8 research aircraft as part of the ARCTAS mission. Compared to local background air, 15 of these compounds showed no statisti-cal enhancements over the oil sands (propyne, MTBE, furan, CCl 4 , CH 3 CCl 3 , ethyl chloride, CHBr 3 , CH 2 Br 2 , CHBrCl 2 , CHBr 2 Cl and O 3 ). As is to be expected with high NO levels, O 3 was anticorrelated with NO and it appears that depletion of O 3 by NO dominated over O 3 production on this short time-scale. Another 9 compounds showed minimal enhancements over the oil sands (HFC-134a, HCFC-141b, 1,2-dichloroethane, H-1211, H-1301, H-2402, CH 3 Br, CH 3 I, CH 3 Cl).
Carbon monoxide showed a clear enhancement (up to 48%) and strong correlations with many other compounds, but its maximum mixing ratio of the flight was not over the oil sands. By contrast the remaining 59 compounds showed greater mixing ratios over the oil sands than at any other time during the flight, and in some cases than at any other time during the entire mission. The maximum enhancements were 1.1-397× the background BL values, most notably nheptane (397×), SO 2 (382×), NO (319×), n-octane (242×) and methylcyclohexane (113×). Based on their mutual correlations, the elevated trace gases fell into two groups. The first group includes CO and species associated with direct evaporative emissions from the oil sands themselves and/or from the diluent used to lower the viscosity of the recovered bitumen (C 4 -C 9 alkanes, C 5 -C 6 cycloalkanes, C 6 -C 8 aromatics). By contrast to the n-alkanes, the enhancements of aromatics such as benzene were relatively small, especially compared to values that have been measured downstream at petrochemical refineries. The second group also includes CO in addition to a wide variety of species associated with emissions from the mining effort, for example coke combustion (e.g., SO 2 ) and upgrading (e.g., NO, NO 2 , NO y ). The maximum SO 2 and NO levels over the oil sands were greater than at any other time during the summer phase of ARCTAS, and even though northern Alberta is a rural environment, the SO 2 and NO y levels were comparable to those measured in the world's megacities. The CO enhancements were generally small suggesting that the upgraders emit low levels of CO, which is to be expected for high temperature combustion where the carbon is converted mostly to CO 2 (i.e., efficient combustion). Although the oil sands industry is a major user of natural gas, a strong natural gas signal was not evident in the data suggesting low natural gas leakage levels and high efficiency combustion associated with the upgraders. Instead, the elevated CH 4 levels are consistent with methanogenic tailings pond emissions. Like CO, the butanes also fell into both groups and appear to have multiple sources from oil sands mining.
Isoprene, a biogenic tracer, was enhanced during the oil sands boundary layer run but with a maximum over a vegetated area south of the major mining operations. By contrast, the monoterpenes α-pinene and β-pinene (also biogenic tracers) were most enhanced in the oil sands plumes and showed good correlations with many industrial species associated with the mining effort including C 2 Cl 4 and SO 2 . The possibility of pinene emissions directly associated with the mining operations requires further study. Carbonyl sulphide and CH 3 Cl were also notable in that they failed to be drawn down over the surface mining sites, most likely because of the removal of their vegetation and soil sinks, respectively.
These measurements represent the only independent characterization of trace gas emissions from oil sands mining operations of which we are aware. Although the absolute mixing ratios of many VOCs were relatively modest compared to major petrochemical facilities that have been studied, they are significantly enhanced above background levels. The high reactivity of most of these gases, combined with significant emissions of NO x and SO 2 in what would otherwise be a relatively pristine area, mean that they do have the potential to form O 3 and acid conditions downwind of this activity. Further study of such potential effects is required. For example modeling and a multi-day day ground-based grid study near the mining sites would help to more completely characterize the trace gases that are emitted from and impacted by the Alberta oil sands industry.