Volatile organic compound distributions during the NACHTT campaign at the Boulder Atmospheric Observatory: Influence of urban and natural gas sources



[1] A comprehensive suite of volatile organic compounds (VOCs) was measured at the semirural Boulder Atmospheric Observatory (BAO) in northeast Colorado during the Nitrogen, Aerosol Composition, and Halogens on a Tall Tower (NACHTT) campaign during the winter of 2011. A signature of elevated nonmethane hydrocarbon (NMHC) mixing ratios was observed throughout the campaign. The C2-C5 alkane mixing ratios were an order of magnitude greater than the regional background. Light alkane mixing ratios were similar to those at urban sites impacted by petrochemical industry emissions with ethane and propane reaching maximums of over 100 ppbv. The mean (± standard deviation) calculated total OH reactivity (7.0 ± 5.0 s−1) was also similar to urban sites. Analysis of VOC wind direction dependence, emission ratios with tracer compounds, and vertical profiles up to 250 m implicated regional natural gas production activities as the source of the elevated VOCs to the northeast of BAO and urban combustion emissions as the major VOC source to the south of BAO. Elevated acetonitrile and dimethyl sulfide mixing ratios were also associated with natural gas emissions. Fluxes of natural gas associated NMHCs were determined to estimate regional emission rates which ranged from 40 ± 14 Gg yr−1 for propane to 0.03 ± 0.01 Gg yr−1 for n-nonane. These emissions have the potential to impact downwind air quality as natural gas associated NMHCs comprised ≈24% of the calculated OH reactivity. The measurements described here provide a baseline for determining the efficacy of future policies designed to control emissions from natural gas production activities.

1 Introduction

[2] Novel fossil fuel extraction techniques, including horizontal drilling and hydraulic fracturing, have allowed for efficient recovery of natural gas from shale and tight sand deposits. Consequently, natural gas has become an increasingly important part of the U.S. energy supply [Energy Information Administration (EIA), 2011]. Rapid expansion in natural gas production has raised concerns regarding the impacts of the industry on human and environmental health through its potential effects on water quality [Osborn et al., 2011] and air quality [Schnell et al., 2009]. Volatile organic compound (VOC) emissions from natural gas production have the potential to be toxic [Colburn et al., 2012; McKenzie et al., 2012], to contribute to ozone (O3) [Schnell et al., 2009; Kemball-Cook et al., 2010] and secondary organic aerosol formation, and to impact climate [Howarth et al., 2011; Alvarez et al., 2012]. Despite public interest in this issue, few peer-reviewed VOC measurements near or within natural gas fields have been reported [see, e.g., Viswanath, 1994; Katzenstein et al., 2003; Lelieveld et al., 2005; Petron et al., 2012; McKenzie et al., 2012; Gilman et al., 2013]. Here we characterize atmospheric mixing ratios of a large suite of VOCs measured at a semirural site located within one of the top 10 natural gas fields in the U.S. [EIA, 2009].

[3] Natural gas production and exploration in the U.S. has increased dramatically over the past decade. The number of active production wells increased 50% between 2000 and 2011, while total natural gas production increased by 19% over the same time period [EIA, 2013]. Natural gas production is projected to continue to increase because it is a cleaner burning energy source than coal and there are abundant reserves in the U.S. [EIA, 2011, 2012, 2013]. As a result of these production increases, recent observational studies have reported deteriorating air quality in regions within or downwind of natural gas basins, including the Jonah Field and Pinedale Anticline in the Upper Green River Basin (UGRB) in Wyoming [Wyoming Department of Environmental Quality-Air Quality Division, 2012], the Uintah Basin in Utah [Lyman and Shorthill, 2013], the Piceance Basin in western Colorado [Garfield County Public Health Department, 2012], and the Denver-Julesburg Basin in northeast Colorado [Petron et al., 2012]. Schnell et al. [2009] suggested that VOC emissions from natural gas processing in the UGRB, combined with cold, stable atmospheric conditions, caused rapid photochemical O3 production with mixing ratios reaching over 140 ppbv. Modeling studies have indicated that emissions from natural gas production could cause 8 h average O3 to increase by ≈1–10 ppbv in the west-southwest U.S., which may impact the O3 attainment status of these regions and downwind areas [Kemball-Cook et al., 2010; Rodriguez et al., 2009; Environmental Protection Agency (EPA), 2008].

[4] In 2010, 7% of the nation's natural gas was produced in Colorado, making it the fifth largest contributor to natural gas production in the U.S. (following Texas, Louisiana, Wyoming, and Oklahoma) [EIA, 2013]. The Energy Information Administration (EIA) tracks the number of producing natural gas wells by state. According to the EIA, the number of active gas wells in Colorado increased by 34% between 2000 and 2011. As of 2011, Colorado had the sixth highest number of natural gas wells in the U.S. with 30,101 producing wells, which represented 6% of the nation's total. The number of wells in Colorado is continuing to increase. The majority of natural gas production and exploration in Colorado occurs in Weld County in northeastern Colorado. The Baker Hughes interactive rig count (http://gis.bakerhughesdirect.com/RigCounts/default2.aspx) showed that ≈40% of the active drill rigs in Colorado were located in Weld County in February–March 2011. The 2008 Environmental Protection Agency (EPA) National Emissions Inventory (http://www.epa.gov/ttnchie1/net/2008inventory.html) reported that oil and natural gas (O&NG) production activities were the largest VOC emission source in Weld County comprising over 60% of total VOC emissions. VOCs are emitted during various stages of the natural gas extraction process, including during well construction, drilling, hydraulic fracturing, flowback, completion, storage and transport of raw gas, flaring of vented gas, flashing of gas condensate, and dehydration [Bar-Ilan et al., 2008a]. The elevated levels of VOC emissions have contributed to the air quality issues documented in the Annual Air Quality Data Reports for the Northern Front Range of Colorado conducted by the Colorado Department of Public Health and Environment (CDPHE) (available at http://www.colorado.gov/airquality/tech_doc_repository.aspx). In November 2007, the EPA designated Denver and the Northern Front Range region as being in nonattainment for the 8 h O3 standard of 0.08 ppmv, and regional municipalities have developed plans to reduce emissions of ozone precursors.

[5] Until recently, natural gas production has been largely exempt from federal environmental regulation. In response to ongoing environmental concerns, several state agencies have adopted water and air pollution regulations to ensure the safety of natural gas production. In 2008, the CDPHE enacted new permitting requirements and more stringent regulations for VOC emissions from natural gas condensate tanks, glycol dehydrators, and engines (complete CDPHE rules available at http://www.colorado.gov/cs/Satellite/CDPHE-Main/CBON/1251601911433). The 2008 rules required implementation of VOC emission reduction technologies in the Colorado Front Range 8 h O3 nonattainment area, including low-bleed or no-bleed valves on pneumatic devices and auto igniters to flare emitted VOCs, to achieve a targeted 90% reduction in VOC emissions during the summer and 70% reduction throughout the year. Statewide regulations implemented in 2008 included mandatory VOC emission reductions of 95% from large condensate tanks and 90% of VOC emissions from glycol dehydrators compared to uncontrolled emissions. In addition, production wells were required to use green completion practices and flare vented gas where technically feasible. Similar regulations were adopted by the EPA in 2012 and will be phased in between 2012 and 2015 [EPA, 2012]. Ambient VOC mixing ratios, emission ratios, and emission rates provided in this work provide a useful measure of the effect of the 2008 emission regulations and a baseline for comparison to future measurements in this region to determine the efficacy of ongoing regulatory efforts.

[6] The measurements reported here were conducted as a component of the Nitrogen, Aerosol Composition, and Halogens on a Tall Tower (NACHTT) campaign conducted at the National Oceanic and Atmospheric Administration's (NOAA) Boulder Atmospheric Observatory (BAO) in Weld County, Colorado, during the time period of 18 February to 13 March 2011. The objectives of NACHTT included studying the processes associated with wintertime halogen activation and heterogeneous aerosol chemistry, characterizing wintertime radical budgets and the role of nitrogen oxides in producing tropospheric halogens, and developing an improved inventory of chlorine (see Brown et al., [2013], for a campaign overview). A unique VOC source signature was observed throughout the campaign, which prompted additional study beyond the original NACHTT objectives. This work presents the results from a large VOC data set collected during NACHTT and is focused on (1) identifying the sources of specific classes of VOCs, (2) quantifying the regional emission fluxes of VOCs associated with local sources, and (3) examining the impact of the VOC distribution observed at BAO on hydroxyl radical (OH) reactivity as a proxy for potential O3 production.

2 Experimental

[7] The NACHTT campaign was conducted at the National Oceanic and Atmospheric Administration's Boulder Atmospheric Observatory (BAO) from 18 February to 13 March 2011. Located in a semirural area of Weld County, Colorado (Figure 1), the BAO maintains a 300 m tower that facilitates convenient atmospheric profiling. BAO is located in a primarily agricultural region; however, the site is nearby major urban areas: ≈35 km north of Denver and ≈30 km east of Boulder. In addition, the site is located within the Denver-Julesburg Basin, which is an active O&NG exploration and production region. An overview of the NACHTT campaign and measurements is given in Brown et al. [2013].

Figure 1.

Map of study region showing the location of the Boulder Atmospheric Observatory (BAO) and urbanized areas, highways, and natural gas wells. The BAO is located in southwest Weld County, approximately 30 km east of Boulder and 35 km north of the Denver metropolitan area. Interstate 25 runs north from Denver and to the east of the sample site. The BAO is surrounded by natural gas wells in the Wattenberg Field, which extends primarily to the northeast of the site through much of southern Weld County.

[8] Two sets of whole air samples were collected in 2 L electropolished stainless steel canisters. The first set was collected at the top of each hour, beginning at noon (Mountain Standard Time; MST) on 18 February and continuing through noon on 13 March 2011 from a height of 22 m on the BAO tower, hereafter referred to as hourly samples (N = 550). The second set of samples, hereafter referred to as profile samples, were collected at heights of 50 m (N = 37), 100 m (N = 35), 150 m (N = 32), 200 m (N = 32), and 250 m (N = 31) using the service elevator on the tower. Profile samples were collected two to four times every 24 h from 24 February to 11 March 2011 with an effort to collect two profiles during the day (07:00–18:00 h MST based on JNO2 measurements) and two during the night (18:00–07:00 h MST) as weather conditions permitted; the service elevator on the tower could not be used when wind speeds were greater than 12 m s−1 or when there was ice on the elevator guide rails. A total of 37 individual profiles were collected, 14 during the day and 23 during the night. Collecting a complete profile took approximately 35 min, and several profile collections were interrupted by dangerous weather conditions resulting in an unequal number of samples being collected at each height.

[9] Prior to the field campaign, the canisters were manually flushed three times in the laboratory by evacuating the canister to 10−2 Torr and filling to 760 Torr with ultrahigh purity helium that had passed through an activated charcoal/molecular sieve (13X) trap immersed in liquid nitrogen followed by a final evacuation to 10−2 Torr. Ambient air was continuously flushed through the stainless steel sampling line to the trailer next to the BAO tower containing the canister sample collection manifold using a metal bellows pump (MB-302, Senior Aerospace, Sharon, MA). Before the sample was collected, the canister was manually flushed five times with ambient air by filling the evacuated canister to approximately 20 psig then allowing the canister to vent back down to ambient pressure. The canister was then pressurized to 30 psig with the metal bellows pump.

[10] The canisters were analyzed for a suite of VOCs immediately following the campaign in the laboratory at the University of New Hampshire. Details of the analytical system are given in Sive et al. [2005], Zhou et al. [2008], and Russo et al. [2010a, 2010b]. Briefly, the analytical system used three gas chromatographs equipped with two flame ionization detectors, two electron capture detectors, and a mass spectrometer. Each 1785 cm3 (STP) sample aliquot was analyzed for C2-C10 nonmethane hydrocarbons (NMHCs), C1-C2 halocarbons, C1-C5 alkyl nitrates, several oxygenated VOCS (OVOCs), and reduced sulfur compounds. The specific compounds discussed in this work along with the limits of detection are listed in Table 1. The OVOCs, carbonyl sulfide (COS), and dimethyl sulfide (DMS) were detected using electron impact mass spectrometry. Response factors for NMHCs, halocarbons, alkyl nitrates, and reduced sulfur compounds were determined using multiple standard cylinders that spanned a range of expected mixing ratios. The OVOC response factors were determined using multipoint dilutions of multiple independent, National Institute of Standards and Technology-traceable gravimetrically prepared standards (AiR Inc., Denver, CO). Analytical precision was 1–8% for the NMHCs, 3–8% for the alkyl nitrates, 3–5% for the reduced sulfur compounds, and 3–15% for the halocarbons. Mean precision for the OVOCs was 8% and ranged from 5% for acetone to 10% for acetaldehyde.

Table 1. Summary Statistics of VOC Mixing Ratios and OH Reactivities for All Hourly Samples Collected at 22 m
 Mixing Ratio (pptv)OH Reactivity (s−1)  
CompoundMean (SD)MedianRangeBackgroundaMean (SD)Background (SD)aLODN > LOD
  1. a

    Background is defined as the mean and standard deviation of mixing ratios observed during periods when clean air was sampled: portions of 20 and 26 February 2011 and 3 and 10–12 March 2011.

  2. b

    The formaldehyde mixing ratio used for OH reactivity calculations was estimated as the average of the wintertime 24 h median values reported by Eisele et al. [2009, Table 4.1] at two sites in the Colorado Front Range.

  3. c

    The methane mixing ratio used for OH reactivity calculations was estimated as the mean reported by Petron et al. [2012] at BAO.

  4. d

    Total OH reactivity was calculated as the sum of the reactivity of each individual VOC, and values below the detection limit were estimated as random values between 0 and the limit of detection.

Ethane22,100 (22,300)13,6001,280–101,7003,2000.13 (0.13)0.013 (<0.01)5550
Propane15,900 (19,400)8,700343–121,1001,5000.43 (0.53)0.022 (<0.01)5550
i-Butane2,890 (3,590)1,52048–23,3003000.15 (0.19)<0.015550
n-Butane7,060 (8,950)3,52086–59,4006200.4 (0.51)0.018 (<0.01)5550
i-Pentane1,980 (2,420)1,04023–15,4002100.18 (0.21)0.01 (<0.01)3550
n-Pentane2,010 (2,450)1,06022–15,4002000.19 (0.23)<0.013550
n-Hexane528 (620)2887–4,120630.068 (0.079)<0.013550
n-Heptane150 (169)864–1,360210.025 (0.028)<0.013543
n-Octane52 (54)332–4519<0.01<0.012511
n-Nonane19 (16)141–1245<0.01<0.011498
n-Decane10 (9)71–793<0.01<0.011530
Neopentane17 (17)113–994<0.01<0.013432
2,3-Dimethylbutane31 (30)193–1947<0.01<0.013489
2,2-Dimethylbutane31 (28)203–1518<0.01<0.013416
2-Methylpentane470 (501)27222–3,170940.059 (0.064)<0.0120542
3-Methylpentane245 (267)1426–1,700380.031 (0.034)<0.015542
2,4-Dimethylpentane24 (21)163–1589<0.01<0.013538
2,3-Dimethylpentane31 (29)224–20613<0.01<0.013441
2-Methylhexane50 (38)404–29428<0.01<0.014535
2,2,4-Trimethylpentane44 (43)293–30627<0.01<0.012503
2,3,4-Trimethylpentane9 (4)82–287<0.01<0.012108
2-Methylheptane34 (33)223–2639<0.01<0.012417
3-Methylheptane24 (22)142–1658<0.01<0.012398
Cyclopentane117 (135)637–87118<0.01<0.015550
Methylcyclopentane236 (275)1323–1,510270.044 (0.052)<0.013542
Cyclohexane143 (175)773–1,200150.024 (0.03)<0.013542
Methylcyclohexane143 (168)805–1,380200.033 (0.04)<0.015540
Total alkanes54,400 (61,800)30,9001,890–354,2006,4001.81 (2.15)0.11 (0.04) 550
Ethene434 (465)28413–4,1901040.096 (0.1)0.019 (0.023)5550
Propene104 (142)545–1,680280.077 (0.1)0.016 (0.02)4549
1-Butene17 (17)113–19280.011 (0.013)<0.013450
trans-2-Butene14 (22)93–2246<0.01<0.013146
cis-2-Butene13 (17)83–1697<0.01<0.013153
1-Pentene8 (7)73–806<0.01<0.013213
trans-2-Pentene8 (9)63–774<0.01<0.01386
cis-2-Pentene7 (7)53–583<0.01<0.01361
2-Methyl-1-butene10 (10)83–1006<0.01<0.013152
2-Methyl-2-butene10 (15)63–1414<0.01<0.013104
1-Hexene8 (5)73–17<LOD<0.01<0.0126
cis-3-Hexene19 (16)123–102240.011 (0.02)0.016 (0.03)3172
trans-2-Hexene9 (8)63–469<0.01<0.01249
cis-2-Hexene9 (7)64–23<LOD<0.01<0.0136
Total alkenes670 (749)42855–7,1002090.24 (0.28)0.083 (0.062) 550
Ethyne633 (382)523178–3,4203460.016 (<0.01)<0.016550
Benzene186 (126)14639–86578<0.01<0.015550
Toluene190 (190)12610–1,620460.028 (0.028)<0.015550
Ethylbenzene17 (19)101–2015<0.01<0.011549
m + p-Xylene74 (91)402–1,050180.035 (0.042)<0.011550
o-Xylene23 (26)131–2976<0.01<0.011545
Styrene5 (5)31–412<0.01<0.011346
i-Propylbenzene2 (1)21–162<0.01<0.011337
n-Propylbenzene4 (3)31–313<0.01<0.011375
m-Ethyltoluene10 (13)51–1545<0.01<0.011503
p-Ethyltoluene5 (6)31–643<0.01<0.011412
o-Ethyltoluene3 (3)21–343<0.01<0.011326
1,3,5-Trimethylbenzene6 (6)31–763<0.01<0.011446
1,2,4-Trimethylbenzene18 (23)91–29470.014 (0.019)<0.011541
1,2,3-Trimethylbenzene4 (3)31–373<0.01<0.011319
1,3-Diethylbenzene4 (4)31–454<0.01<0.011345
1,4-Diethylbenzene3 (3)31–293<0.01<0.011299
1,2-Diethylbenzene2 (2)21–142<0.01<0.011213
Total aromatics556 (525)37564–4,8702090.11 (0.13)0.025 (0.023) 550
Isoprene8 (4)84–307<0.01<0.01259
alpha-Pinene2 (3)11–421<0.01<0.011295
Camphene1 (2)11–131<0.01<0.01198
beta-Pinene1 (1)11–161<0.01<0.011191
3-Carene1 (2)11–131<0.01<0.01181
d-Limonene3 (6)21–621<0.01<0.011135
p-Cymene1 (1)11–81<0.01<0.011311
g-Terpinene1 (1)11–2<LOD<0.01<0.0114
Total biogenics19 (19)148–18520.013 (0.028)<0.01 311
MeONO24 (1)43–134  0.2550
EtONO24 (1)42–93  0.2550
2-PrONO216 (8)143–438  0.2550
1-PrONO22 (1)21–61  0.2544
2-BuONO227 (18)233–9210  0.2550
3-PenONO27 (6)61–262  0.2549
2-PenONO212 (10)101–473  0.2549
C1-C5 RONO273 (44)6413–23531   550
COS411 (32)411324–548418<0.01<0.015548
DMS12 (9)92–58<LOD<0.01<0.015263
Total S compounds423 (41)420326–606419<0.01<0.01 548
Acetonitrile139 (80)11336–503110<0.01<0.0120550
Formaldehydeb965NANANA0.2 (NA)0.2 (NA)NANA
Methanol4,660 (2,240)4,300875–13,9002,5300.11 (0.05)0.067 (0.045)100528
Ethanol2,620 (2,020)1,990105–15,7001,4200.2 (0.16)0.11 (0.079)80536
Acetaldehyde703 (584)52059–4,1603760.24 (0.22)0.09 (0.1)50498
Acetone1,300 (390)1,240736–6,0901,190<0.01<0.01100547
MEK331 (199)28964–2,900189<0.01<0.0150541
MBO23 (17)189–110340.02 (0.022)0.019 (0.029)8280
Total OVOCs10,700 (5,500)8,5001,880–43,4005,5000.78 (0.37)0.5 (0.16) 550
Total NMHCs + OVOCs123,200 (128,800)75,0008,950–702,10013,8002.98 (2.68)0.73 (0.24) 550
CH4 (ppbv)c1,867NANANA0.29 (NA)0.29 (NA)NANA
CO (ppbv)174 (51)156104–6211450.64 (0.18)0.51 (0.05)0.04511
NO2 (ppbv)8 (8)51–403.83.13 (2.82)1.53 (0.94)170352
NO (ppbv)3 (6)11–440.7  46197
Total OH reactivityd    7.04 (4.96)3.06 (0.97)  

[11] Additional measurements used in this analysis include carbon monoxide (CO), nitric oxide (NO), and nitrogen dioxide (NO2). Carbon monoxide was measured at 22 m every 15 or 20 min by infrared absorbance spectroscopy (modified Thermo Scientific 48C Trace Level) with a detection limit of 0.04 ppmv and 3 ppbv precision [Andrews et al., 2013]. This sample cycle allowed for measurements at other heights, twice hourly baseline checks, and twice daily calibration. We used the CO measurement reported closest to the top of each hour to correspond to the time when each hourly canister was being filled. Nitrogen oxides (NO and NO2) were measured every second by diode laser-based cavity ring down spectrometry [Wagner et al., 2013], located on a moveable carriage attached to the tower which conducted continuous vertical profiles from 0 to 250 m throughout the campaign. Detection limits for NO and NO2 measurements were 170 and 46 pptv, respectively, and measurement precision was 5% for both species. We used the average NO and NO2 mixing ratios measured when the movable carriage height was between 15 and 25 m during the first 10 min of each hour in order to correspond to the time when the canister sample was collected (or the 10 min prior to the start of the hour if data during the first 10 min of the hour were not available). Methane (CH4) mixing ratios were estimated as the mean mixing ratio reported by Petron et al. [2012] for the BAO site, and formaldehyde mixing ratios were estimated as the average of the wintertime 24 h median values at two sites in the Colorado Front Range reported by Eisele et al. [2009, Table 4.1] for the purpose of calculating OH reactivity.

3 Results and Discussion

3.1 Elevated Mixing Ratios of Selected VOCs

[12] Throughout the campaign, individual VOC mixing ratios varied by 1 to 4 orders of magnitude in the hourly samples (Table 1), indicating significant variability in the sampled air masses and local air quality conditions. This is demonstrated by the numerous periods of both relatively clean (e.g., portions of 20 and 26–27 February and 9–11 March) and polluted (e.g., portions of 19, 24–25, and 28 February and 2–3 and 5 March) air masses observed at BAO (Figure 2).

Figure 2.

Time series of selected VOC mixing ratios measured during NACHTT. Samples were collected at the top of each hour at 22 m on the BAO tower. Missing data points indicate that mixing ratios were below the limit of detection. Concurrent peaks in mixing ratios of (a) the natural gas associated alkanes ethane and propane, and (b) dimethylsulfide (DMS) and (c) acetonitrile occurred throughout the campaign. Mixing ratios of (d) the combustion associated compounds ethyne and ethene tracked those of (e) aromatic compounds benzene and toluene, and to a lesser extent, (f) the oxygenated VOCs methanol and ethanol. (g) Mixing ratios of the alkyl nitrates 2-butyl nitrate and 2-pentyl nitrate and total alkyl nitrates showed a different temporal distribution indicating their primarily photochemical source.

[13] Of the measured VOCs, the C2-C5 alkanes (excluding cyclopentane and 2,2-dimethyl propane), methanol, ethanol, acetone, and acetaldehyde were the 10 most abundant (Table 1). High mixing ratios of alkanes were observed throughout the entire monthlong campaign. At BAO, the average total measured VOC (C2-C10 alkane + C2-C6 alkene + C6-C10 aromatics + ethyne + biogenics + OVOCs + alkyl nitrates + DMS + COS) distribution was dominated by alkanes (mean ± standard deviation: 69 ± 17%). The C2-C5 alkanes represented approximately 66% of the total VOC mixing ratio, and the maximum total C2-C5 alkane mixing ratio was 324 ppbv. The prevalence of this high alkane signal over the monthlong campaign indicates the presence of a continuous source of alkane emissions.

[14] Alkane mixing ratios generally decreased with increasing carbon number. Of the measured VOCs, ethane (mean ± standard deviation: 22 ± 22 ppbv) and propane (16 ± 19 ppbv) were the two most abundant, with maximum mixing ratios of each exceeding 100 ppbv, representing 31% and 20% of total measured VOCs, respectively (Table 1). The C4 alkanes comprised 11% of the total measured VOCs with mean mixing ratios of 7.1 and 2.9 ppbv for n-butane and i-butane, respectively. Iso-pentane and n-pentane mixing ratios were nearly identical with means of 2.0 ppbv and medians of 1.05 ppbv.

[15] Relatively clean air masses were also sampled throughout the campaign (e.g., portions of 20 and 26 February and 10–12 March; see Figure 2). Mean mixing ratios from these periods were used to define the background VOC mixing ratios (see Table 1). Mean C2-C7 n-alkane mixing ratios for the entire campaign period (including the background observations) were between 7 times higher for ethane and 11 times higher for n-butane than the background levels. Mean C2-C6 n-alkane mixing ratios were approximately an order of magnitude higher than those observed at other semirural and remote sites, including Thompson Farm in Durham, New Hampshire, USA [Russo et al., 2010a]; Alert, Canada [Gautrois et al., 2003]; and Whiteface Mountain, New York [Gong and Demerjian, 1997]. Other alkanes exhibiting a similar enhancement above the background mixing ratio included n-heptane, n-octane, i-butane, i-pentane, 2-methylpentane, 3-methylpentane, cyclopentane, methylcyclopentane, cyclohexane, and methylcyclohexane.

[16] In contrast to the alkanes discussed above, mixing ratios of other VOCs were similar to measurements from other semirural sites. Mean mixing ratios of commonly used tracers of urban emissions, ethyne (633 ± 382 pptv) and tetrachloroethylene (C2Cl4) (10 ± 7 pptv), were only double the background mixing ratio and were similar to previously reported wintertime measurements at semirural sites [Russo et al., 2010a; Gautrois et al., 2003; Yokouchi et al., 1996; Gong and Demerjian, 1997]. Mean mixing ratios of the dominant alkenes, ethene (434 ± 465 pptv) and propene (104 ± 142 pptv), and aromatics, benzene (186 ± 126 pptv) and toluene (190 ± 190 pptv), were approximately 2 times higher than the background mixing ratio (Table 1); however, all were within one standard deviation of the wintertime mean reported at Thompson Farm [Russo et al., 2010a]. Mean OVOC mixing ratios exhibited smaller enhancements or no enhancement above background (factor of 1 to 2 times the background value). Mean mixing ratios of C3-C5 alkyl nitrates, excluding 1-propyl nitrate, were elevated above background by a factor of 2–4. Biogenic NMHCs were observed at low mixing ratios during the campaign as would be expected outside of the growing season.

[17] A comparison of NMHC mixing ratios observed at BAO with those reported in two major North American polluted areas (Galveston Bay near Houston, TX and Mexico City, Mexico), several U.S. cities, and a natural gas-producing region in the SW U.S. (Figure 3) demonstrates the magnitude of the observed mixing ratios. It should be noted that the results presented here were measured in late winter and include samples collected throughout the entire day. The mean and median C2-C4 alkane mixing ratios at BAO were higher than those observed in 28 different U.S. cities [Baker et al., 2008], while the range of pentane, ethene, ethyne, benzene, and toluene mixing ratios were comparable to major U.S. cities (Figure 3). The mean and median ethane and propane mixing ratios and the mean n-butane mixing ratio were higher than those observed in Galveston Bay and the Houston ship channel. The 75th percentile i-butane, n-pentane, and i-pentane mixing ratios at BAO were similar to the mean levels reported for Galveston Bay and the Houston ship channel [Gilman et al., 2009]. The C2-C4 alkane levels were higher at BAO than over the Anadarko Basin (Texas, Oklahoma, Kansas, eastern New Mexico) in 2001–2002 where Katzenstein et al. [2003] attributed the observed high C1-C4 alkane mixing ratios to emissions from O&NG production. The mean ethane mixing ratio at BAO was higher than in Mexico City, and the mean C3-C5 alkane mixing ratios were within the range of the 24 h averages observed in and downwind of Mexico City [Apel et al., 2010]. The ethene, ethyne, benzene, and toluene levels at BAO were lower than in Mexico City. This comparison illustrates that the observed alkane mixing ratios at BAO represent a relatively large, unrecognized source of VOCs.

Figure 3.

Box and whisker plot of several NMHCs measured at BAO during 18 February to 13 March 2011 in comparison to measurements from other polluted sites. In the box and whisker plot, the solid black line = median, dashed black line = mean, top and bottom of box = 75th and 25th percentiles, respectively, top and bottom whisker = 90th and 10th percentiles, respectively, and top and bottom black circle = 95th and 5th percentiles, respectively. Measurements from other polluted sites are represented by different symbols: red triangle = range of mean mixing ratios reported for 28 U.S. cities measured during summer between 1999 and 2005 [Baker et al., 2008]; blue square = mean mixing ratio in Houston ship channel/Galveston Bay, August–September 2006 [Gilman et al., 2009]; pink diamond = 24 h mean mixing ratio in Mexico City (TO, top pink diamond) and at site T1 30 km northeast of Mexico City (bottom pink diamond) during March 2006 [Apel et al., 2010]; green triangle = mean mixing ratio in SW U.S. during September 2001 (top green triangle) and April 2002 (bottom green triangle) [Katzenstein et al., 2003].

3.2 Source Identification

3.2.1 Vertical Profiles Suggest Local Ground Level VOC Sources

[18] Vertical profiles conducted throughout the campaign indicated the influence of local anthropogenic emission sources on the air sampled at BAO. This analysis includes all of the profile samples collected between 50 and 250 m and the hourly samples collected at 22 m on the top of the hour both before and after the profile samples were collected. Mixing ratios of C2-C7 n-alkanes, as well as i-butane, i-pentane, 2-methylpentane, 3-methylpentane, cyclopentane, methylcyclopentane, cyclohexane, and methylcyclohexane, decreased rapidly with height (Figure 4 and supporting information Table S1), suggesting a strong local ground level emission source. Alkane mixing ratios were elevated at 250 m relative to rural areas [e.g., Russo et al., 2010a; Gautrois et al., 2003; Gong and Demerjian, 1997] and background values (Figure 4 and supporting information Table S1), suggesting that the impact of surface emissions extended to the 250 m level and most likely beyond into the daytime mixing layer. Petron et al. [2012] reported midday measurements of a small set of NMHCs made at 300 m on the BAO tower throughout 2007–2010 and showed that this alkane enhancement was regional and was not a rare event. The propane, n-butane, i-pentane, and n-pentane mixing ratios measured at 250 m during the NACHTT campaign were comparable to the wintertime C3-C5 alkane mixing ratios in the north and east sectors reported by Petron et al. [2012]. The similarity in mixing ratios measured in different years suggests that anthropogenic emissions did not decrease between 2007 and 2011 and that the relative contributions of various emission sources of C3-C5 alkanes remained consistent.

Figure 4.

Vertical profiles of (a) ethane, (b) propane, (c) ethyne, (d) benzene, and (e) 2-butyl nitrate (2-BuONO2). Symbols in Figures 4a–4e are the mean mixing ratios for all profile samples ± standard error of the mean at 22, 50, 100, 150, 200, and 250 m. Mixing ratios at 22 m represent the hourly samples at 22 m at the top of the hour immediately prior to and following the collection of each profile. Grey dashed lines in Figures 4a–4e are the mean background mixing ratio observed in the hourly cans at 22 m.

[19] Mean profiles of many other compounds showed evidence of local emissions of a lesser magnitude than the light alkanes. For instance, ethyne (Figure 4c), propene, benzene (Figure 4d), toluene, methanol, and ethanol all decreased with altitude but only by a factor of 1.5–2 between 22 and 250 m. Individual C4 and C5 alkyl nitrate mixing ratios decreased by a factor of 1.3 between 22 and 250 m (Figure 4e). The C1-C3 alkyl nitrates, acetone, and acetaldehyde showed neutral profiles suggesting that they did not have strong local sources at the BAO tower.

3.2.2 Wind Direction Dependencies Distinguish Between Urban and Natural Gas Emissions

[20] VOC mixing ratios were dependent on wind direction (Figure 5). For this analysis, only the hourly samples collected at 22 m are used to define three wind sectors: (1) the northeast sector (0°–90°), (2) the south sector (135°–225°), and (3) the west sector (225°–315°). In addition, this analysis included only samples collected when the wind speed was above 3 m s−1 to minimize the influence of proximal emission sources.

Figure 5.

Wind roses showing wind dependence of mixing ratios for selected (a) alkanes, (b) alkenes and ethyne, (c) dimethyl sulfide (DMS), (d) aromatics, (e) alkyl nitrates, and (f) OVOCs. Symbols are the mean mixing ratio in 45° wind direction sectors when wind speeds were greater than 3 m s−1.

[21] Mean mixing ratios of C2-C5 alkanes, DMS, and alkyl nitrates were highest in the northeast wind sector and showed a second, lesser peak in the south sector (Figures 5a, 5c, and 5e). Numerous branched and cyclic alkanes showed a similar wind direction dependence, including 2- and 3-methylpentane, methylcyclopentane, cyclohexane, and methylcyclohexane. Combustion tracers, such as C2-C5 alkenes and ethyne, were highest in the south sector with a second peak in the northeast sector (Figure 5b). A similar pattern was observed for C7-C9 aromatics and the major OVOCs, including methanol, ethanol, and acetaldehyde (Figures 5d and 5f).

[22] BAO is located in the southwest of the Wattenberg Field, which was the ninth most productive natural gas field in the U.S. as of 2009 [EIA, 2009]. While BAO is surrounded by natural gas wells, the Wattenberg Field extends northeast along the Denver-Julesburg Basin, and the area to the northeast of BAO contains the greatest number of natural gas wells (Figure 1). This active natural gas exploration region contained over 14,000 active wells in early 2011 according to the Colorado Oil and Gas Conservation Commission (COGCC) geographic information system database (http://cogcc.state.co.us/Home/gismain.cfm), implicating natural gas production activities as the major source of the elevated alkanes and DMS.

[23] The Denver metropolitan area to the south of BAO is the major urban emission source in the region. In addition to the elevated C2-C5 alkenes and ethyne observed in the south sector, the mixing ratios of CO, NO, and NO2, additional combustion tracers, were highest in the south sector [Brown et al., 2013]. Mixing ratios of the industrial solvent C2Cl4 were also highest in the south wind sector corroborating the stronger relative influence of an urban emission source to the south of BAO. While mixing ratios of alkenes and aromatics were also elevated in the northeast sector, suggesting that these compounds were emitted from natural gas production activities, the higher mixing ratios observed in the south sector indicate that mixing ratios of these compounds at BAO were more strongly influenced by urban emission sources.

[24] A combination of emission sources to the northeast, east, and southeast of BAO contributed to increased benzene mixing ratios relative to the west sector (Figure 5d). Benzene is a component of automotive and urban emissions [Baker et al., 2008] and is also a minor component of natural gas [EPA, 1989]. In the northeast sector, both emissions from natural gas and from automobile traffic were possible sources of benzene. Automotive emissions from the Denver metropolitan area to the south and from Interstate 25 to the east likely contributed to the elevated benzene mixing ratios observed at BAO (Figure 1). Similar to the findings of Gilman et al. [2013], these results suggest that a mixture of urban, automotive, and natural gas emissions influenced the measured benzene.

[25] Mean mixing ratios of all NMHCs were lowest in the west wind sector (Figure 5). The mean wind speeds (5.4 ± 3.2 m s−1) and O3 mixing ratios [Brown et al., 2013], were highest in this sector, suggesting that observations from the west sector likely represent transport of processed air masses from over the Rocky Mountains.

[26] Ratios of VOCs also indicate the influence of natural gas production in the northeast sector and urban air masses in the south sector. Iso-pentane is an abundant component of gasoline and is elevated relative to n-pentane [McGaughey et al., 2004], but mixing ratios of these gases in natural gas from the Wattenberg Field are approximately equal [COGCC, 2007]. The mean i-pentane to n-pentane ratio was lowest in the northeast sector (0.94 ± 0.04) and highest in the south sector (1.11 ± 0.21). The ratio of toluene to benzene is also useful as an indicator of automotive and urban emissions as both gases are emitted by similar sources and toluene reacts more rapidly with OH than benzene [Warneke et al., 2007]. The mean toluene to benzene ratio was highest in the south wind sector (1.09 ± 0.40), lowest in the west sector (0.57 ± 0.35), and at an intermediate value in the northeast sector (0.76 ± 0.25). These VOC ratios suggest that air masses from the south were more frequently impacted by fresh automotive emissions from the Denver metropolitan area and that air masses from the northeast were impacted by emissions from natural gas production.

[27] Cluster analysis of hourly air mass 72 h back trajectories modeled using the NOAA Air Resources Laboratory Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT; http://www.arl.noaa.gov/ss/models/hysplit.html) for all 22 m hourly samples indicated that approximately 85% of all air masses sampled during NACHTT originated from the west of BAO with the remaining 15% coming from the northeast (supporting information Figure S1a). Approximately 35% of the sampled air masses traveled through the Wattenberg Field area prior to their arrival at BAO. Furthermore, a comparison of back trajectories for the samples with the top tenth percentile propane mixing ratios with those in the bottom tenth percentile showed that over 56% of the back trajectories for high propane samples passed through the area to the northeast of BAO, while none of the back trajectories for low propane samples passed through the northeast, with the majority coming directly from the west (supporting information Figures S1b and S1c). While wind direction at the BAO tower was not always representative of long-range air mass transport, these back trajectories indicate that wind direction at the tower was indicative of local and regional influences. The wind direction dependence of the VOC mixing ratios observed at BAO suggests that the site is impacted by anthropogenic emissions from both urban and combustion sources and natural gas production. In the following section, we examine mixing ratio correlations and emission ratios to further categorize the regional sources of the NMHCs observed at BAO.

3.2.3 Mixing Ratio Correlations and Emission Ratios for Natural Gas and Urban Sources

[28] Mixing ratios of C2-C6 alkanes were highly correlated (r2 > 0.9) with propane throughout the campaign, suggesting they shared a common source throughout the study region (Figure 6a). Light alkanes are prominent components of natural gas extracted from the Wattenberg formation [COGCC, 2007], and the alkane signature observed at BAO was similar to the pattern observed in Wattenberg natural gas (Figure 6b). The lower percentage of ethane observed in ambient air compared to natural gas indicates that a greater proportion of emissions come from flashing (evaporation caused by a rapid decrease in pressure or increase in temperature) of natural gas condensates, which causes an enrichment of higher hydrocarbons relative to methane and ethane [Petron et al., 2012]. This finding supports the estimates of bottom-up emission inventories for the Denver-Julesburg Basin, which found that the major source of flashing emissions, condensate tanks, accounts for ≈66% of total VOC emissions in the area and venting of natural gas accounts for ≈29% [Bar-Ilan et al., 2008a]. The similarity between the C2-C5 alkane ratios in Wattenberg natural gas and those observed at BAO suggests that these samples were influenced by regional natural gas production operations.

Figure 6.

Natural gas source identification. (a) Light alkanes were highly correlated with propane at BAO throughout the campaign, and (b) mole fractions of light alkanes observed at Boulder Atmospheric Observatory (BAO) were similar to the mean (error bars represent standard deviation) mole fractions in Weld County natural gas previously reported by the Colorado Oil and Gas Conservation Commission [COGCC, 2007]. (c) The distinct 1 to 1 relationship between i-pentane and n-pentane observed at BAO is a unique tracer for natural gas emissions and was lower than the ratio previously reported by Baker et al. [2008] for the Denver metropolitan area.

[29] The relationship between the pentane isomers was unique to the natural gas source signature at BAO. The mean ratio of i-pentane to n-pentane was 1.0 (±0.1) throughout the campaign (Figure 6c), matching that previously reported in Wattenberg natural gas [COGCC, 2007]. In many other regions i-pentane is approximately a factor of 1.5–4 larger than n-pentane [e.g., Russo et al., 2010a; Parrish et al., 1998; McLaren et al., 1996; Baker et al., 2008]. The higher i-pentane is often attributed to enhanced fuel evaporation emissions because i-pentane is a more prevalent component of gasoline than n-pentane [Harley et al., 1992; McGaughey et al., 2004]. Gilman et al. [2013] also reported a similar but lower i-pentane to n-pentane ratio (i-pentane/n-pentane = 0.885) at BAO. Because n-pentane has a slightly faster rate of reaction with OH than i-pentane, the i-pentane/n-pentane ratio is expected to increase with photochemical aging, and the i-pentane to n-pentane ratio could be a useful indicator of the impact of natural gas emissions in downwind regions.

[30] Emission ratios of VOCs with source-specific tracer compounds were examined in order to determine which VOCs were associated with the natural gas, urban, and combustion sources suggested in the wind direction dependence analysis discussed above. Propane was used as a tracer for natural gas emissions (ERC3H8), ethyne (ERC2H2) and CO (ERCO) as tracers for combustion emissions, and C2Cl4 (ERC2Cl4) as a tracer for urban industrial emissions. Emission ratios were determined for each of the three wind sectors (northeast, south, and west) using only hourly samples collected when the wind speed at BAO at 10 m was greater than 3.0 m s−1. Compounds detected in less than half of the samples in any wind sector were excluded from the analysis including 2, 3, 4-trimethylpentane, the C4-C6 alkenes, styrene, i-propylbenzene, o-ethyltoluene, 1, 2, 3-trimethylbenzene, the diethylbenzenes, and all biogenic compounds.

[31] Emission ratios were determined from the slope of the linear regression with tracer compounds for all VOCs having a correlation coefficient with a tracer compound greater than 0.50 (Table 2a-d). Two distinct VOC sources were identified in the south wind sector based on ethyne to propane ratios in individual samples. The first of these sources was characterized by higher ERC2H2 for alkanes (Figure 7) and strong correlations between the majority of measured VOCs and propane (Table 2b) and is hereafter referred to as the southern natural gas source. All hourly samples from the south wind sector where wind speeds were greater than 3 m s−1 having a propane mixing ratio greater than 5 ppbv and an ethyne to propane ratio of less than 0.04 were included in the analysis of the southern natural gas source (N = 13). The second source was characterized by higher ERC2H2 of combustion-related compounds, which were similar to the ERC2H2 reported for these compounds in the Denver metropolitan area in the summer of 2004 [Baker et al., 2008] (Figure 7), and strong correlations between the measured VOCs and the tracers ethyne and C2Cl4 (Table 2c) and is hereafter referred to as the southern urban source. All hourly samples from the south wind sector that were not included in the southern natural gas source analysis were included in the southern urban source analysis (N = 40).

Table 2a. Correlation Matrix and Emission Ratios for VOCs With Tracer Compounds for the Northeast Wind Sector (0–90°)a
 Northeast Wind Sector
 Propane (pptv pptv−1)Ethyne (pptv pptv−1)CO (pptv ppbv−1)C2Cl4 (pptv pptv−1)
Compoundr2ERC3H8 (Error)r2ERC2H2 (Error)r2ERCO (Error)r2ERC2Cl4 (Error)
  1. a

    Emission ratios (ER) are the slope of the linear regression with the listed tracer compound for all samples from the given wind sector with wind speeds greater than 3 m s−1. The error is the residual error of the linear regression. Values are listed only for compounds which were correlated to tracer compounds with an r2 greater than 0.50.

Ethane0.951 (0.028)0.7954 (3.3)0.70479 (37)0.583150 (317)
Propane  0.7750 (3.3)0.68445 (36)0.663156 (266)
i-Butane0.970.18 (0.0035)0.819.4 (0.55)0.7586 (5.9)0.67586 (49)
n-Butane0.960.46 (0.01)0.824 (1.4)0.74215 (15)0.671464 (123)
i-Pentane0.910.13 (0.0046)0.847 (0.37)0.8065 (3.8)0.67426 (35)
n-Pentane0.920.13 (0.0046)0.837 (0.38)0.7865 (4)0.66427 (37)
n-Hexane0.870.035 (0.0016)0.872 (0.092)0.8018 (1.1)0.59112 (11)
n-Heptane0.790.0099 (0.0006)0.860.58 (0.028)0.785.3 (0.33)0.5030 (3.6)
n-Octane0.760.0031 (0.0002)0.860.19 (0.009)0.771.7 (0.11)  
n-Nonane0.660.0008 (0.0001)0.790.052 (0.0032)0.700.46 (0.036)  
n-Decane0.620.0004 (0.0001)0.790.024 (0.0015)0.740.22 (0.016)  
Neopentane0.900.0008 (0.0001)0.830.046 (0.0025)0.780.42 (0.027)0.602.6 (0.26)
2,3-Dimethylbutane0.820.0017 (0.0001)0.850.096 (0.0049)0.810.89 (0.052)0.565.3 (0.57)
2,2-Dimethylbutane0.760.0013 (0.0001)0.840.08 (0.0045)0.840.76 (0.043)0.554.4 (0.52)
2-Methylpentane0.870.028 (0.0013)0.851.6 (0.076)0.8114 (0.83)0.6290 (8.4)
3-Methylpentane0.870.015 (0.0007)0.870.83 (0.038)0.827.7 (0.43)0.6148 (4.5)
2,4-Dimethylpentane0.800.0012 (0.0001)0.880.071 (0.0032)0.840.66 (0.035)0.563.9 (0.41)
2,3-Dimethylpentane0.610.0012 (0.0001)0.650.072 (0.0067)0.590.65 (0.068)  
2-Methylhexane0.760.0019 (0.0001)0.860.12 (0.0055)0.791 (0.063)0.566.4 (0.66)
2,2,4-Trimethylpentane0.800.0022 (0.0001)0.880.13 (0.0057)0.801.2 (0.068)0.597.1 (0.71)
2-Methylheptane0.750.0019 (0.0001)0.880.11 (0.0053)0.791 (0.066)  
3-Methylheptane0.740.0012 (0.0001)0.870.075 (0.0037)0.790.67 (0.044)  
Cyclopentane0.890.0075 (0.0003)0.840.41 (0.022)0.763.7 (0.25)0.6324 (2.2)
Methylcyclopentane0.660.0094 (0.0008)    0.5533 (3.5)
Cyclohexane0.820.0094 (0.0005)0.810.53 (0.03)0.744.8 (0.33)0.5530 (3.2)
Methylcyclohexane0.770.0095 (0.0006)0.830.56 (0.03)0.735 (0.36)  
Ethene0.740.017 (0.0012)0.921.1 (0.037)0.889.8 (0.44)0.658 (5.7)
Propene0.670.0042 (0.0004)0.870.27 (0.013)0.852.5 (0.13)0.6316 (1.4)
1-Butene0.690.0005 (0.0001)0.750.03 (0.0023)0.760.28 (0.021)0.661.9 (0.18)
Ethyne0.770.015 (0.001)  0.909 (0.36)0.6153 (5.1)
Benzene0.790.0063 (0.0004)0.940.39 (0.012)0.863.6 (0.17)0.5721 (2.2)
Toluene0.730.0085 (0.0006)0.900.54 (0.021)0.875 (0.23)0.5529 (3.1)
Ethylbenzene0.630.0007 (0.0001)0.840.042 (0.0022)0.830.4 (0.021)0.552.3 (0.25)
m + p-Xylene0.670.0035 (0.0003)0.840.23 (0.012)0.82.1 (0.12)0.5012 (1.4)
o-Xylene0.600.0010 (0.0001)0.810.062 (0.0036)0.780.58 (0.037)  
n-Propylbenzene  0.650.006 (0.0007)0.660.056 (0.0061)  
m-Ethyltoluene0.560.0004 (0.0001)0.740.023 (0.0016)0.730.21 (0.016)  
p-Ethyltoluene  0.730.0099 (0.0009)0.70.092 (0.0088)  
o-Ethyltoluene  0.720.0042 (0.0005)0.670.039 (0.0054)  
1,3,5-Trimethylbenzene0.530.0002 (0.0001)0.680.01 (0.0009)0.650.093 (0.0089)  
1,2,4-Trimethylbenzene0.540.0006 (0.0001)0.680.037 (0.003)0.680.35 (0.029)  
DMS0.770.0003 (0.0001)0.710.017 (0.0018)0.590.14 (0.02)  
Methanol  0.564.5 (0.47)0.5743 (4.4)  
Ethanol0.550.058 (0.0062)0.683.6 (0.3)0.6433 (3)  
Acetonitrile0.740.0039 (0.0003)0.750.22 (0.015)0.742.1 (0.15)0.5413 (1.4)
C2Cl40.660.0002 (0.0001)0.610.011 (0.0011)0.710.12 (0.0089)  
Table 2b. Correlation Matrix and Emission Ratios for VOCs With Tracer Compounds for the South Wind Sector (135–225°) Natural Gas Sourcea
 South Wind Sector-Natural Gas Source
 Propane (pptv pptv−1)Ethyne (pptv pptv−1)CO (pptv ppbv−1)C2Cl4 (pptv pptv−1)
Compoundr2ERC3H8 (Error)r2ERC2H2 (Error)r2ERCO (Error)r2ERC2Cl4 (Error)
  1. a

    Emission ratios (ER) are the slope of the linear regression with the listed tracer compound for all samples from the given wind sector with wind speeds greater than 3 m s−1. The error is the residual error of the linear regression. Values are listed only for compounds which were correlated to tracer compounds with an r2 greater than 0.50.

Ethane0.961.5 (0.091)0.8577 (9.9)0.59511 (129)  
Propane  0.8051 (7.6)0.51320 (94)  
i-Butane1.000.19 (0.0028)0.809.7 (1.5)0.5061 (18)  
n-Butane0.990.44 (0.01)0.7622 (3.6)    
i-Pentane0.990.13 (0.0032)0.786.5 (1.1)    
n-Pentane0.980.13 (0.005)0.746.2 (1.1)    
n-Hexane0.990.039 (0.0013)0.731.9 (0.34)    
n-Heptane0.990.012 (0.0004)0.830.61 (0.084)0.533.9 (1.1)  
n-Octane0.980.0038 (0.0001)0.830.2 (0.027)0.551.3 (0.35)  
n-Nonane0.940.0011 (0.0001)0.830.06 (0.0083)0.530.38 (0.11)  
n-Decane0.880.0005 (0.0001)0.690.023 (0.0047)    
Neopentane0.980.001 (0.0001)0.830.052 (0.007)0.510.32 (0.095)  
2,3-Dimethylbutane0.920.0016 (0.0001)0.650.075 (0.016)    
2,2-Dimethylbutane0.960.0019 (0.0001)0.910.11 (0.011)0.600.67 (0.17)0.535.5 (1.6)
2-Methylpentane0.990.031 (0.0007)0.781.6 (0.25)    
3-Methylpentane1.000.017 (0.0003)0.780.84 (0.13)    
2,4-Dimethylpentane0.950.0016 (0.0001)0.840.084 (0.011)0.530.53 (0.15)  
2,3-Dimethylpentane0.850.001 (0.0001)0.790.055 (0.0085)0.640.39 (0.088)  
2-Methylhexane0.830.0021 (0.0003)0.620.1 (0.025)    
2,2,4-Trimethylpentane0.980.0021 (0.0001)0.830.11 (0.015)0.550.71 (0.19)  
2-Methylheptane0.980.0024 (0.0001)0.820.12 (0.017)0.530.79 (0.22)  
3-Methylheptane0.960.0017 (0.0001)0.890.095 (0.01)0.580.61 (0.16)  
Cyclopentane0.970.0062 (0.0004)0.720.3 (0.057)    
Methylcyclopentane0.980.015 (0.0006)0.720.74 (0.14)    
Cyclohexane0.980.011 (0.0005)0.800.56 (0.086)    
Methylcyclohexane0.890.0092 (0.0010)0.720.47 (0.087)    
Ethene0.790.018 (0.0029)0.981.2 (0.054)0.768.1 (1.4)0.7164 (12)
Propene0.710.0038 (0.00072)0.920.24 (0.021)0.801.8 (0.27)0.7414 (2.5)
1-Butene0.610.0005 (0.0001)0.850.036 (0.0046)0.800.27 (0.041)0.732.1 (0.39)
Ethyne0.800.016 (0.0024)  0.746.8 (1.2)0.6753 (11)
Benzene0.970.0085 (0.0005)0.900.47 (0.046)0.643.1 (0.71)0.5123 (6.7)
Toluene0.940.012 (0.0009)0.890.63 (0.066)0.654.3 (0.95)0.5231 (9)
Ethylbenzene0.750.00074 (0.0001)0.700.041 (0.008)0.540.28 (0.078)  
m + p-Xylene0.840.0048 (0.0006)0.750.26 (0.044)0.541.7 (0.48)  
o-Xylene0.720.0011 (0.0002)0.600.056 (0.014)    
DMS  0.800.032 (0.0073)0.540.22 (0.089)0.852.4 (0.43)
Ethanol0.780.1 (0.017)0.836.1 (0.83)0.5740 (10)0.73367 (68)
Acetonitrile0.900.0042 (0.0004)0.650.2 (0.045)    
C2Cl4  0.670.013 (0.0027)0.560.092 (0.025)  
Table 2c. Correlation Matrix and Emission Ratios for VOCs With Tracer Compounds for the South Wind Sector (135–225°) Urban Sourcea
 South Wind Sector-Urban Source
 Propane (pptv pptv−1)Ethyne (pptv pptv−1)CO (pptv ppbv−1)C2Cl4 (pptv pptv−1)
Compoundr2ERC3H8 (Error)r2ERC2H2 (Error)r2ERCO (Error)r2ERC2Cl4 (Error)
  1. a

    Emission ratios (ER) are the slope of the linear regression with the listed tracer compound for all samples from the given wind sector with wind speeds greater than 3 m s−1. The error is the residual error of the linear regression. Values are listed only for compounds which were correlated to tracer compounds with an r2 greater than 0.50.

  2. b

    A plume of enhanced propene was filtered out in the south wind sector-urban source samples to give the shown emission ratios.

Ethane0.961.8 (0.049)0.8612 (1.4)0.5047 (12)0.74613 (83)
Propane  0.786.2 (0.95)  0.68326 (54)
i-Butane0.960.2 (0.0058)0.841.3 (0.17)0.585.4 (1.4)0.7971 (9.3)
n-Butane0.980.44 (0.011)0.802.8 (0.41)0.5310 (3.3)0.76153 (23)
i-Pentane0.900.18 (0.0094)0.881.3 (0.13)0.715.8 (1.2)0.8570 (7)
n-Pentane0.950.16 (0.0055)0.831.0 (0.13)0.624.2 (1)0.7856 (7.3)
n-Hexane0.920.049 (0.002)0.830.32 (0.038)0.641.3 (0.26)0.8418 (2)
n-Heptane0.920.017 (0.0008)0.870.11 (0.011)0.750.48 (0.079)0.856.4 (0.59)
n-Octane0.840.0078 (0.0005)0.840.055 (0.005)0.710.22 (0.036)0.802.9 (0.29)
n-Nonane0.700.0026 (0.0003)0.850.02 (0.0017)0.650.11 (0.018)0.841.1 (0.092)
n-Decane0.630.002 (0.0002)0.830.016 (0.0012)0.650.095 (0.014)0.910.95 (0.054)
Neopentane0.870.0012 (0.0001)0.740.0087 (0.0014)  0.610.43 (0.087)
2,3-Dimethylbutane0.760.0041 (0.0004)0.820.03 (0.0028)0.640.14 (0.025)0.911.8 (0.12)
2,2-Dimethylbutane0.790.0038 (0.0005)0.840.029 (0.0033)0.600.13 (0.03)0.851.6 (0.18)
2-Methylpentane0.830.043 (0.0028)0.830.30 (0.032)0.641.3 (0.27)0.8517 (1.7)
3-Methylpentane0.890.028 (0.0016)0.850.19 (0.017)0.740.86 (0.13)0.8811 (0.87)
2,4-Dimethylpentane0.580.0027 (0.0003)0.720.021 (0.0023)  0.651.1 (0.15)
2,3-Dimethylpentane0.630.0048 (0.0006)0.820.04 (0.0039)0.670.23 (0.04)0.862.2 (0.2)
2,2,4-Trimethylpentane0.570.0056 (0.0007)0.650.042 (0.0053)  0.722.5 (0.27)
2-Methylheptane0.810.0042 (0.0005)0.860.036 (0.0043)0.690.14 (0.033)0.861.9 (0.22)
3-Methylheptane0.870.0043 (0.0006)0.790.035 (0.0042)0.800.1 (0.019)0.801.9 (0.24)
Cyclopentane0.910.0083 (0.0004)0.840.056 (0.006)0.570.24 (0.055)0.823.1 (0.32)
Methylcyclopentane0.920.024 (0.0012)0.830.16 (0.015)0.620.65 (0.13)0.808.8 (0.85)
Cyclohexane0.930.011 (0.0004)0.810.072 (0.0091)0.530.32 (0.08)0.753.9 (0.5)
Methylcyclohexane0.930.012 (0.0005)0.760.073 (0.011)  0.713.9 (0.59)
Ethene0.740.15 (0.015)0.931.2 (0.055)0.866.9 (0.53)0.8866 (4.1)
Propeneb0.770.11 (0.012)0.680.7 (0.08)0.642.3 (0.32)0.7040 (4.6)
1-Butene0.740.0061 (0.0007)0.870.047 (0.0033)0.770.26 (0.029)0.842.6 (0.21)
Ethyne0.780.13 (0.012)  0.835.7 (0.48)0.8652 (3.8)
Benzene0.840.031 (0.0024)0.950.23 (0.0099)0.811.2 (0.11)0.8912 (0.78)
Toluene0.750.062 (0.006)0.900.48 (0.026)0.742.5 (0.27)0.9127 (1.5)
Ethylbenzene0.670.0069 (0.0008)0.840.054 (0.0039)0.610.28 (0.041)0.863.1 (0.23)
m + p-Xylene0.670.03 (0.0035)0.820.23 (0.018)0.561.1 (0.18)0.8614 (1)
o-Xylene0.720.0099 (0.0011)0.860.075 (0.0052)0.650.36 (0.047)0.864.2 (0.32)
n-Propylbenzene0.640.0015 (0.0002)0.680.011 (0.0014)  0.730.61 (0.077)
m-Ethyltoluene0.600.0053 (0.0007)0.690.04 (0.0045)  0.712.3 (0.26)
p-Ethyltoluene0.530.002 (0.0003)0.610.015 (0.0021)  0.650.86 (0.12)
o-Ethyltoluene0.530.0009 (0.0001)0.590.0066 (0.001)  0.620.38 (0.058)
1,3,5-Trimethylbenzene0.620.0022 (0.0003)0.700.016 (0.0019)  0.700.93 (0.11)
1,2,4-Trimethylbenzene0.520.0081 (0.0012)0.610.061 (0.0079)  0.623.5 (0.46)
Ethanol  0.705.7 (0.66)0.6543 (6.1)0.71324 (39)
C2Cl40.680.0021 (0.0002)0.860.016 (0.0012)0.570.079 (0.013)  
Table 2d. Correlation Matrix and Emission Ratios for VOCs With Tracer Compounds for the West Wind Sector (225–315°)a
 West Wind Sector
 Propane (pptv pptv−1)Ethyne (pptv pptv−1)CO (pptv ppbv−1)C2Cl4 (pptv pptv−1)
Compoundr2ERC3H8 (Error)r2ERC2H2 (Error)r2ERCO (Error)r2ERC2Cl4 (Error)
  1. a

    Emission ratios (ER) are the slope of the linear regression with the listed tracer compound for all samples from the given wind sector with wind speeds greater than 3 m s−1. The error is the error of the residuals of the linear regression. Values are listed only for compounds which were correlated to tracer compounds with an r2 greater than 0.50.

Ethane0.991.3 (0.012)0.7343 (2.7)  0.671977 (143)
Propane  0.6933 (2.3)  0.641522 (119)
i-Butane1.000.19 (0.0011)0.686.1 (0.43)  0.64286 (22)
n-Butane1.000.46 (0.003)0.6715 (1.1)  0.63704 (55)
i-Pentane0.980.13 (0.0017)0.694.2 (0.29)  0.66198 (15)
n-Pentane0.980.13 (0.0016)0.674.1 (0.3)  0.64195 (15)
n-Hexane0.980.032 (0.0005)0.701.1 (0.072)  0.6650 (3.7)
n-Heptane0.950.0084 (0.0002)0.710.29 (0.02)  0.6713 (1)
n-Octane0.940.0028 (0.0001)0.760.1 (0.0072)0.510.58 (0.071)0.704.7 (0.38)
n-Nonane0.800.0009 (0.0001)0.800.035 (0.0024)0.580.21 (0.025)0.761.7 (0.13)
n-Decane0.740.0005 (0.0001)0.810.02 (0.0011)0.670.13 (0.01)0.750.95 (0.061)
Neopentane0.980.0008 (0.0001)0.720.027 (0.0031)  0.641.3 (0.17)
2,3-Dimethylbutane0.950.0018 (0.0001)0.760.066 (0.0051)  0.693 (0.28)
2,2-Dimethylbutane0.920.0013 (0.0001)0.750.049 (0.0053)0.520.29 (0.053)0.742.4 (0.28)
2-Methylpentane0.950.026 (0.0006)0.660.86 (0.065)  0.6541 (3.2)
3-Methylpentane0.970.014 (0.0003)0.710.48 (0.033)  0.6722 (1.7)
2,4-Dimethylpentane0.870.001 (0.0001)0.760.037 (0.0022)0.600.22 (0.02)0.761.8 (0.1)
2,3-Dimethylpentane0.890.0018 (0.0001)0.780.068 (0.0059)0.550.4 (0.061)0.733.2 (0.33)
2-Methylhexane0.610.0017 (0.0002)0.500.061 (0.0066)    
2-Methylheptane0.900.0015 (0.0001)0.750.055 (0.0062)  0.692.6 (0.34)
3-Methylheptane0.820.0011 (0.0001)0.710.039 (0.0052)  0.671.9 (0.28)
Cyclopentane0.980.0073 (0.0001)0.700.24 (0.016)  0.6611 (0.84)
Methylcyclopentane0.950.014 (0.0004)0.710.49 (0.033)  0.6924 (1.7)
Cyclohexane0.940.0081 (0.0002)0.690.28 (0.02)  0.6313 (1)
Methylcyclohexane0.960.0088 (0.0002)0.670.3 (0.022)  0.6314 (1.1)
Ethene0.740.026 (0.0016)0.921.2 (0.035)0.847.4 (0.33)0.8454 (2.4)
Propene0.740.0058 (0.0004)0.880.25 (0.0097)0.831.6 (0.079)0.8612 (0.51)
1-Butene0.680.00083 (0.0001)0.880.04 (0.0024)0.810.28 (0.022)0.791.8 (0.15)
Ethyne0.690.021 (0.0015)  0.866.2 (0.26)0.7742 (2.4)
Benzene0.860.007 (0.0003)0.930.29 (0.0084)0.781.8 (0.099)0.8113 (0.65)
Toluene0.810.011 (0.0005)0.870.44 (0.018)0.772.9 (0.16)0.8221 (0.98)
Ethylbenzene0.740.001 (0.0001)0.860.045 (0.0019)0.790.3 (0.016)0.812.1 (0.1)
m + p-Xylene0.760.0051 (0.0003)0.830.21 (0.0099)0.721.4 (0.089)0.759.7 (0.58)
o-Xylene0.740.0014 (0.0001)0.850.061 (0.0027)0.770.4 (0.024)0.782.8 (0.16)
n-Propylbenzene0.580.0001 (0.0001)0.760.0051 (0.0005)0.620.033 (0.0047)0.640.24 (0.032)
m-Ethyltoluene0.580.0006 (0.0001)0.730.026 (0.002)0.640.17 (0.017)0.651.2 (0.11)
p-Ethyltoluene0.530.0002 (0.0001)0.640.0082 (0.001)0.530.053 (0.0082)0.510.36 (0.058)
o-Ethyltoluene0.560.0001 (0.0001)0.730.0044 (0.0006)0.610.029 (0.005)0.600.2 (0.035)
1,3,5-Trimethylbenzene0.610.0003 (0.0001)0.720.012 (0.0011)0.580.072 (0.0092)0.600.52 (0.063)
1,2,4-Trimethylbenzene0.570.0011 (0.0001)0.650.046 (0.0036)0.560.3 (0.029)0.602.1 (0.19)
DMS0.670.0004 (0.0001)      
Methanol0.520.12 (0.013)    0.55238 (23)
Ethanol  0.523.6 (0.38)    
Acetonitrile0.570.0035 (0.0003)      
C2Cl40.640.00042 (0.0001)0.770.018 (0.001)0.740.13 (0.0077)  
Figure 7.

Emission ratios relative to (a and b) ethyne and (c and d) propane for the northeast (NE), south (S), and west (W) sectors. Two source profiles were observed in the southern wind sector and are labeled as the southern natural gas source (S-NG) and the southern urban source (S-U). Bars represent the mean and error bars represent one standard deviation. The series labeled Petron are measurements made at BAO from Petron et al. [2012], and the series labeled Denver are measurements from Denver, CO in 2006 from Baker et al. [2008].

[32] All C2-C8 alkanes (except for methycyclopentane and 2,3-dimethylpentane), benzene, toluene, acetonitrile, and DMS were well correlated with propane (0.73 < r2 < 1.00) in the northeast wind sector, suggesting that natural gas-associated emissions were a source of these compounds (Table 2a; acetonitrile and DMS are discussed separately in section 3.4 below). The ERC2H2 of propane, n-butane, n-pentane, and i-pentane were similar to those previously reported for this site between 2007 and 2010, which were attributed to natural gas-related activities [Petron et al., 2012]. However, most of these VOCs were also correlated with ethyne (0.71 < r2 < 0.94) and CO (0.59 < r2 < 0.87) in the northeast sector, making it difficult to separate combustion emissions from direct natural gas emissions. The correlation between propane and the combustion tracers ethyne and CO in the northeast sector is not unexpected given the general collocation of direct natural gas emission sources (e.g., venting of raw natural gas during drilling, hydraulic fracturing, extraction, transport, storage, flashing and dehydration of natural gas condensates, and post extraction processing at regional processing plants) with combustion sources, (e.g., automobile traffic and diesel generators, truck engines, and drilling rigs used in natural gas production).

[33] The emission ratio profile of the southern natural gas source was similar to that of the northeast sector with the exception of a higher ERC2H2 for ethane (Figure 7). This could indicate that venting emissions are more prominent than flashing emissions to the south of BAO or could be an artifact of the smaller number of samples included in this calculation (N = 13) relative to the northeast sector (N = 73). Natural gas emissions could be expected to influence all wind sectors, including the south wind sector, because natural gas operations surround the sample site (Figure 1), making it difficult to separate out emission ratios for specific sources. However, samples representing the southern natural gas source comprised only 25% of the observations from the south wind sector, and VOC emissions from the southern urban source were the more prevalent (≈75%) in this wind sector allowing for the two sources to be examined separately.

[34] The ERC3H8 for C2-C8 alkanes in the southern urban wind sector was similar to that of other wind sectors, indicating that natural gas emissions were still the primary source of these alkanes (Table 2a), but the lower natural gas production activity in this wind sector resulted in lower C2-C8 alkane mixing ratios relative to urban and combustion tracers. Alkene, aromatic, and C9-C10 alkane ERC3H8 were higher in southern urban samples than in any other wind sector and were similar to, but generally lower than, values reported in Denver in 2004 [Baker et al., 2008]. These lower emission ratios indicate either a decrease in emissions of these compounds or an increase in urban propane emissions since 2004. They could also be indicative of seasonal differences as Baker et al. [2008] conducted their study in the summer of 2004. The general similarities between the emission ratios in southern urban samples to values measured in Denver, and the correlation with the urban tracer C2Cl4 for all NMHCs, except 2-methylhexane, suggest that the emission ratios for the southern urban source reported in Table 2c were characteristic of regional urban emissions during NACHTT.

[35] Emission ratio profiles in the west wind sector represented a combination of natural gas and urban sources. Alkene and aromatic ERC2H2 in the west sector were similar to those of the southern urban source. Mixing ratios of most VOCs were also correlated with C2Cl4 mixing ratios (Table 2d), suggesting that emissions from urban areas to the west, including Boulder, Longmont, and Fort Collins, impacted BAO. Distinct urban and natural gas sources were not able to be separated in the west wind sector as in the south sector because of the fewer number of samples for the west sector containing elevated levels of alkanes.

[36] Characterizing the relative importance of the urban/combustion source and the natural gas production source for specific classes of VOCs is useful in determining where to focus emission reduction efforts. Comparing the northeast and southern urban wind sector emission ratios allows for such an analysis (Figure 7). Compounds whose emission ratios with ethyne were roughly equal in all wind sectors but whose emission ratios with propane were lower in the northeast than in the southern urban wind sector were mostly influenced by urban and combustion emissions; natural gas emissions contributed less to the observed mixing ratios of these compounds. VOCs falling into this category included the alkenes, the aromatics, and the C9-C10 alkanes. In addition, as stated above, the emission ratios for this group of VOCs were similar to those measured in Denver in 2004 [Baker et al., 2008], supporting the conclusion that urban and combustion emissions were the major source of these compounds during NACHTT. Conversely, compounds whose emission ratios with propane were roughly equal in all wind sectors but whose emission ratios with ethyne were lower in southern urban samples than in the northeast wind sector were mostly influenced by natural gas emissions, and urban and combustion emissions made lesser contributions to the observed mixing ratios of these compounds. All of the C2-C8 alkanes were included in this category. Gilman et al. [2013] reported similar results for a subset of VOCs measured during NACHTT based on a multivariate regression source apportionment analysis. These results suggest that the most effective way to control emissions of aromatic compounds of human health concern in this area, such as benzene, toluene and the xylenes, would be to focus on reducing emissions from urban and regional combustion sources. On the other hand, tighter controls on the emissions of alkanes from natural gas production sources would be an effective way to reduce the ozone production potential of VOCs in the region (see section 3.5 for further details on the OH reactivity and associated ozone production potential of VOC emissions).

[37] The emission ratios with tracer compounds discussed above clearly distinguished the relative contribution of natural gas and urban/combustion sources on many of the VOCs observed at BAO. The strong correlation between light alkanes, branched alkanes and cyclic alkanes and the similarity of the C2-C5 alkane mole fractions to those observed in Wattenberg natural gas provide evidence for a prominent natural gas emission source influencing the air quality of northeastern Colorado.

3.3 Dimethyl Sulfide and Acetonitrile Associated With Natural Gas Emissions

3.3.1 Dimethyl Sulfide

[38] Dimethyl sulfide is the largest single reduced sulfur compound in the atmosphere [Bates et al., 1992]. The oxidation of DMS in the atmosphere produces sulfur dioxide (SO2) and sulfuric acid, which are sources of cloud condensation nuclei. Thus, the biogeochemical cycling of DMS influences cloud formation and residence times, precipitation rates, cloud and surface albedo, the radiative balance of the earth, and aerosol production, growth, and acidity [e.g., Charlson et al., 1987; Pandis et al., 1994].

[39] Elevated DMS mixing ratios were observed concurrently with peaks in the alkanes and ethyne in the hourly samples (i.e., 25–26 February, 28 February, 5 March, and 13 March) (Figure 2). DMS mixing ratios ranged from the detection limit (5 pptv) to 58 pptv with a mean (± standard deviation) of 12 ± 9 pptv. The highest percentage (50%) of samples with detectable DMS and the strongest DMS-propane correlation (r2 = 0.77) were observed in the northeast wind sector, followed by the west wind sector (r2 = 0.67) (Table 2a). In contrast, detectable mixing ratios of DMS were only observed in 33% of the samples from the south wind sector and 8% of the samples from the west wind sector. No tracer compounds were correlated with DMS in the southern urban samples. The correlation between DMS and propane was weak (r2 = 0.46) and was stronger with ethyne (r2 = 0.80) and C2Cl4 (r2 = 0.85) in the southern natural gas samples. Correlations with urban and combustion tracers could indicate an urban source of DMS but may also result from the small number of samples with detectable DMS in the southern natural gas sector (N = 7). DMS was below the limit of detection in the majority of the profile samples; nonetheless, the percentage of samples with detectable DMS decreased with altitude from 66% at 22 m to only 13% at 250 m. These observations suggest that DMS and propane had common or colocated, ground level sources located primarily in the direction of active natural gas production to the northeast of BAO.

[40] Dimethyl sulfide may be emitted at several points during natural gas extraction and processing. The main reduced sulfur compound in natural gas is hydrogen sulfide, but pptv levels of other compounds, including DMS, may be present [Cui et al., 2009]. Therefore, DMS may be present in fugitive emissions from natural gas wells, storage tanks, and pipelines. Federal law requires that odorants be added to natural gas to aid in leak detection for public safety purposes (Code of Federal Regulations, 49CFR 192.625, http://www.ecfr.gov/). Organosulfur compounds, including DMS, are common odorants, and DMS may be emitted during natural gas processing prior to transmission to consumers. Simpson et al. [2010] observed elevated mixing ratios of DMS in aircraft samples above the Alberta Oil Sands and attributed these emissions to industrial processes associated with gas production in the region. Given the combination of a strong coincidence of DMS and alkane peaks and the relatively strong correlation with propane in the northeast wind sector, it is likely that DMS is emitted at one or more stages during natural gas extraction and processing.

[41] Dimethyl sulfide is produced by marine phytoplankton and is the dominant biogenic sulfur compound emitted from the ocean [Bates et al., 1992], making transport from the Pacific Ocean another possible source. DMS was not correlated with the measured marine-derived halocarbons (e.g., CHBr3, CH2Br2, data not shown), suggesting that they did not share a common source. Moreover, Zhou et al. [2011] showed that, during the NACHTT campaign, the sampled air masses had little or no marine influence, strengthening the argument for a local DMS source. While biogenic sources of DMS are far greater than industrial and combustion sources on a global scale [Watts, 2000], anthropogenic sources may play an important role in regional continental chemistry. For example, sulfur dioxide produced from the oxidation of DMS from noncombustion anthropogenic sources may become more important as SO2 emissions from fossil fuel combustion (i.e., coal fired power plants) decrease and the use of low-sulfur diesel (SO2 emission controls) increases. The connection between natural gas production and reduced sulfur compound emissions should be a focus of future studies, and these emissions should be included in atmospheric chemistry and climate models.

3.3.2 Acetonitrile

[42] Similar to DMS, acetonitrile was correlated with propane in hourly samples in the northeast wind sector (r2 = 0.74) and weakly in the west sector (r2 = 0.57), but it was also strongly correlated with propane in the southern natural gas wind sector (r2 = 0.90) (Table 2a). Acetonitrile was also correlated with combustion tracers ethyne and CO in the northeast sector (r2 ≈ 0.75) and with ethyne in the southern natural gas samples (r2 = 0.65). These results suggest both natural gas associated emissions and combustion emissions contributed to the observed acetonitrile mixing ratios.

[43] As acetonitrile is produced during incomplete combustion of plant matter [Holzinger et al., 1999], it is a common tracer for biomass burning emissions. Several range fires were observed during NACHTT, including the Lefthand Canyon fire, which was visible to the NW of the BAO tower on 11–13 March 2011. While the relatively low mixing ratios (range = 36–503 pptv) indicate that no biomass burning plumes were sampled, the observed fires may have still contributed to regional acetonitrile mixing ratios at BAO. The ERCO for acetonitrile in the northeast wind sector of 2.1 ± 0.002 pptv ppbv−1 was within the range of values (0.4–2.6) reported for fresh smoke from biomass burning [e.g., Holzinger et al., 1999; Andreae et al., 2001] and similar to the values reported by de Gouw et al. [2003] in plumes sampled over Utah (2.0 ± 0.3) and the Yucatan Peninsula (2.7 ± 0.4). The weak correlations between acetonitrile and combustion tracers ethyne and CO (r2 < 0.5) to the west and south of BAO are likely the result of fresh inputs from urban combustion sources.

[44] Anthropogenic sources of acetonitrile include industrial solvent use [EPA, 1985] and, to a lesser extent, vehicle exhaust [Holzinger et al., 2001], but the lack of correlation between acetonitrile and the urban tracer C2Cl4 in any wind sector (Table 2a) indicates that industrial emissions were not a major source of acetonitrile at BAO. The strong correlation with propane and the consistent ERC3H8 in the northeast, southern natural gas, and west wind sectors (≈0.004 pptv pptv−1) suggests that natural gas production may result in acetonitrile emissions. A structurally similar compound, dibromo-acetonitrile, has been reported in hydraulic fracturing fluid as an antimicrobial additive [New York State Department of Environmental Conservation, 2009], but the proprietary nature of hydraulic fracturing fluid used in Colorado during NACHTT makes it difficult to determine if acetonitrile is used in a similar capacity and to what extent it may be used. The relationship between natural gas emissions and acetonitrile should be examined in future studies, particularly in the context of its use in hydraulic fracturing.

3.4 Alkyl Nitrates

[45] Alkyl nitrates (RONO2) are produced from VOC precursors through reactions involving OH radical and NO [Roberts, 1990; Bertman et al., 1995]. Observed total C1-C5 alkyl nitrates (total RONO2 = sum of individual C1-C5 alkyl nitrates) were dominated by 2-butyl nitrate (27 ± 18 pptv), 2-propyl nitrate (16 ± 8 pptv), and 2-pentyl nitrate (12 ± 10 pptv) (Table 1 and Figure 2g). The mean total RONO2 mixing ratio (73 ± 44 pptv) was more than double that observed at a semirural site during winter in New Hampshire [Russo et al., 2010b] and similar to summer values reported for the polluted Pearl River Delta in China [Simpson et al., 2006]. However, comparisons with previous studies are complicated because the mixing ratios of individual alkyl nitrates in benchmark standards used by multiple research groups have changed [Simpson et al., 2011]. Mixing ratios of C3-C5 alkyl nitrates were highly correlated with 2-butyl nitrate (r2 > 0.95) throughout the campaign. Ethyl nitrate showed a weaker correlation with 2-butyl nitrate (r2 = 0.83), and methyl nitrate was completely uncorrelated (r2 ≈ 0), suggesting the importance of other primary or secondary sources.

[46] Alkyl nitrate mixing ratios were not correlated with alkane mixing ratios (Table 2a); however wind direction dependence, vertical profiles, diurnal variations in mixing ratios, and photochemical age estimations indicate strong local photochemical production from natural gas-associated precursors as the major source at BAO. Similar to the C2-C5 alkanes, the mean total RONO2 was also highest in the northeast wind sector (Figure 5e). The decrease in mixing ratio with height in vertical profiles of individual C2-C5 RONO2 (see 2-butyl nitrate in Figure 4e) and total RONO2 suggests production from a local ground level source. Mixing ratios of total RONO2 were also highest during the afternoon (12:00–16:00 MST) supporting a primarily local photochemical source.

[47] The photochemical age of the alkyl nitrates in sampled air masses was estimated by comparing observed ratios of individual RONO2 to their parent hydrocarbons to values predicted using a simplified sequential reaction scheme [Bertman et al., 1995] (reaction (1):

display math(R1)

This reaction scheme assumes only OH-initiated photochemical production of a given alkyl nitrate (RONO2) from its parent hydrocarbon (RH) at the rate kA and photochemical removal at the rate kB as the only removal mechanism. The evolution of alkyl nitrates in a given air mass is then described by equation (1):

display math(1)

Where the alkyl nitrate to parent hydrocarbon ratio inline image, at a given time t, is calculated assuming a starting ratio inline image of 0. Predicted and observed alkyl nitrate to parent hydrocarbon ratios are plotted against the ratio for the most abundant alkyl nitrate, 2-butyl nitrate, and its precursor, n-butane, in Figure 8. The analysis could not be performed for methyl nitrate because methane was not measured. For C3-C5 alkyl nitrates, the observed values were similar to the expected values and indicated that most observations represented photochemical production of RONO2 from alkane emissions less than 1 day old (Figure 8). Observed ethyl nitrate to ethane ratios still indicated photochemical production from fresh emissions but were enhanced in ethyl nitrate relative to the expected values (Figure 8). This deviation suggests another source of ethyl nitrate at BAO, potentially production from oxidation of longer chain alkanes [Russo et al., 2010b; Bertman et al., 1995]. The high alkyl nitrate mixing ratios observed at BAO suggest that transport of alkyl nitrates produced in this region, and subsequent photolysis to yield NO, could be an important source of NOx in downwind locations with limited primary NOx emissions.

Figure 8.

Photochemical age of alkyl nitrates. Comparison of observed alkyl nitrate:parent hydrocarbon ratios (green circles) to predicted values (gray squares) provide an estimate of photochemical age. Most observed alkyl nitrate:parent hydrocarbon ratios indicated that alkyl nitrates were less than 1 day old (red dashed line) and all were less than 3 days old (blue dashed line).

3.5 OH Reactivity

[48] Tropospheric O3 is a criteria pollutant which can have detrimental effects on human health and vegetation [Lippmann, 1989] and is an important greenhouse gas with an estimated positive radiative forcing of 0.35 W/m2 [Solomon et al., 2007]. Ozone is produced in the troposphere from the OH-initiated oxidation of hydrocarbons in the presence of nitrogen oxides. Thus, the mixing ratios of O3 depend on the specific composition of hydrocarbons present in a region or an air mass. The OH reactivity is a measure of the initial rate of peroxy radical formation and can be interpreted as the potential of a specific compound to ultimately produce O3. Here we calculate the OH reactivity of the trace gases observed at BAO as a measure of their O3 production potential and estimate the OH reactivity attributable to natural gas associated emissions.

[49] A compound's OH reactivity is calculated as the product of its concentration [X] in molecules cm−3 and its OH rate constant kOH, X in cm3 molecule−1 s−1 [Atkinson, 2003; Atkinson et al., 1997, 2006; Atkinson and Arey, 2003] (equation (2)):

display math(2)

Reactivity was calculated for CO, CH4, NO2, and the VOCs (C2-C10 alkanes, C2-C6 alkenes, C6-C10 aromatics, ethyne, biogenic hydrocarbons, acetonitrile, COS, DMS, and OVOCs) in all of the hourly samples (Figure 9). In order to better estimate total OH reactivity of VOCs at BAO, constant values for CH4 and formaldehyde were used because these compounds were not measured during the NACHTT campaign. The mean CH4 mixing ratio (1867 ppbv) reported by Petron et al. [2012] was used. For formaldehyde, the average winter 24 h median values reported by Eisele et al. [2009, Table 4.1] at two Colorado Front Range sites was used. Lastly, to avoid biasing the analysis towards samples with a greater number of compounds above the limit of detection (LOD), we assumed that the mixing ratio for any observation below the LOD was a random value between 0 and the LOD. This assumption adds a small amount of random noise to the calculation (<0.01% of the calculated total OH reactivity) but results in more representative total OH reactivity values. The propagated uncertainty of the total OH reactivity estimates presented here was 55% and included the uncertainty of each compound's rate constant for reaction with OH and the uncertainty in the mixing ratio of the compound.

Figure 9.

Percentage contributions of VOCs to OH reactivity for the entire campaign and during background periods for all measured VOCs plus NO2, CO and using constant values (a and b) for CH4 and formaldehyde which were not measured, (c and d) for NMHCs and OVOCs using a constant value for formaldehyde, and (e and f) for C2-C10 alkanes only. Numbers above each pie chart represent the total reactivity, and numbers within each slice are the percentage contribution of the given compound or class of compounds.

[50] The overall mean calculated OH reactivity was 7.0 ± 5.0 s−1 with a maximum of 31.3 s−1 (Table 1 and Figure 9a). Nitrogen dioxide made the largest contribution to OH reactivity (over 40%) followed by the alkanes (21%), the OVOCs (13%), CO (12%), CH4 (6%), alkenes (3%), and the aromatics (1%) (Figure 9). The mean NO2, CO, and CH4 contributions to OH reactivity were 3.1, 0.64, and 0.29 s−1, respectively. The mean calculated OH reactivity at BAO was on the lower end of total OH reactivity values reported in the literature which typically range from between 5 s−1 and over 100 s−1 [see Lou et al., 2010]. Similar to the mixing ratios compared in Figure 3, the mean calculated OH reactivity at BAO was within the range of total summertime OH reactivities measured in moderately sized cities including Houston, TX (range of 7–25 s−1 [Mao et al., 2010]), Nashville, TN (mean ± standard deviation of 11.3 ± 4.8 s−1 [Kovacs et al., 2003]), and Mainze, Germany (range of 6–18 s−1 [Sinha et al., 2008]). Calculated wintertime OH reactivity at BAO was on the low end of ranges reported for megacities including Paris, France (range of 10–120 s−1 [Dolgorouky et al., 2012]), Mexico City, Mexico (range of 10–200 s−1 [Shirley et al., 2006]), New York City, NY (range of 10–100 s−1 [Ren et al., 2003, 2006]), and Tokyo, Japan (range of 10–100 s−1 [Yoshino et al., 2006]). However, these comparisons are complicated by the fact that the direct total OH reactivity measurements from the studies mentioned above include contributions from species not measured during NACHTT, and total OH reactivity measurements at the BAO site would likely be higher than the results of the summation method discussed here. However, the similar magnitudes of OH reactivity at BAO and in these cities again suggests that regional VOC emissions were more representative of urban air quality and that there is the potential for enhanced O3 formation downwind of this site.

[51] The mean total VOC contribution to OH reactivity, which excludes NO2, CH4, and CO, was 3.0 ± 2.7 s−1, which is effectively equivalent to the OH reactivity reported during this campaign by Gilman et al. [2013], with a maximum of 14.6 s−1. The OH reactivity of the VOCs was dominated by the alkanes (1.8 ± 2.2 s−1, 49%), OVOCs (0.78 ± 0.37 s−1, 37%), and alkenes (0.24 ± 0.28 s−1, 9%) (Table 1 and Figure 9c). This result is consistent with the dominance of alkanes and OVOCs in the VOC mixing ratio signature discussed above. Propane and n-butane constituted the largest percentage of the alkane reactivity at 23% and 20%, respectively (Table 1 and Figure 9e). Ethane, i-butane, i-pentane, and n-pentane made similar contributions (8–10%) to the alkane reactivity. The dominance of propane in the alkane contribution to OH reactivity is caused by its greater rate constant for reaction with OH compared to ethane which had a higher mean mixing ratio than propane (section 3.1). The C6 alkanes (n-hexane, 2-methylpentane, 3-methylpentane, methylcyclopentane) each contributed 2–4% to the total OH reactivity of the VOCs (Table 1).

[52] Similar signatures dominated by alkane reactivity were observed in the megacities of Tokyo, Japan [Yoshino et al., 2006] and Mexico City, Mexico [Shirley et al., 2006]. Alkane emissions from petrochemical processing facilities in Houston, TX also resulted in a similar contribution from alkanes to total OH reactivity [Mao et al., 2010]. Comparisons to other rural or semirural sites are difficult because measurements in these areas are typically conducted during the summer to quantify the contribution of biogenic VOCs to OH reactivity. For example, the mean VOC reactivity at BAO was within one standard deviation of the mean reactivity (4.15 ± 2.64 s−1) reported during the summer at Thompson Farm, NH during the 2004 International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) campaign [White et al., 2008]. However, biogenic hydrocarbons accounted for the bulk of summertime VOC OH reactivity during ICARTT but accounted for less than 1% of the OH reactivity at BAO.

[53] Following the alkanes, OH reactivity of the OVOCs was the second most important contributor to total OH reactivity of the VOCs during NACHTT. The estimated formaldehyde (0.20 s−1) and the measured acetaldehyde (0.24 ± 0.22 s−1) and ethanol (0.20 ± 0.16 s−1) accounted for the bulk of OVOC reactivity accounting for 31%, 26%, and 24% of the OVOC reactivity, respectively. While methanol was the most abundant OVOC, its slower rate of reaction with OH resulted in it contributing only 14% to the total OVOC reactivity.

[54] Other classes of compounds made minor contributions to the VOC reactivity at BAO. Alkenes accounted for 9% of the calculated reactivity (Figure 9c), with ethene (0.10 ± 0.10 s−1) and propene (0.08 ± 0.10 s−1) contributing 38% and 28%, respectively, to the total alkene reactivity (Table 1). The mean aromatic OH reactivity at BAO was only 4% of the total OH reactivity from VOCs (Figure 9c) and was dominated by the xylenes (0.04 ± 0.04 s−1), toluene (0.03 ± 0.03 s−1), and the trimethylbenzenes (0.02 ± 0.03 s−1) (Table 1).

[55] To estimate the effect of natural gas emissions on OH reactivity, we compare the mean alkane contribution to OH reactivity observed for the whole campaign to the mean alkane contribution to OH reactivity calculated during the background periods defined in section 3.1 above. Only the OH reactivity of alkanes was considered as they were the dominant VOCs associated with natural gas production as described above in section 3.2. Methane was also excluded from this analysis as it was not measured during NACHTT. The mean alkane contribution to OH reactivity under background conditions was 0.11 ± 0.04 s−1, 94% less than the mean alkane contribution to OH reactivity for the whole campaign. Percent contributions of individual C2-C10 alkanes to the total OH reactivity of the alkanes were similar under background conditions as compared to the whole campaign period underscoring the strong correlation between these compounds and their common source. The difference in mean OH reactivity from alkanes between the entire campaign and the background periods was 1.70 s−1, amounting to 24% of the mean total OH reactivity or 57% of the mean OH reactivity of the VOCs (excluding CH4, CO, and NO2) observed during the campaign. This value is similar to that reported by Gilman et al. [2013] calculated using the results of their multivariate regression source apportionment. This additional OH reactivity attributable to natural gas emissions could have an important impact on local to regional tropospheric O3 formation. The longer photochemical lifetimes of alkanes during the winter result in transport of these compounds and subsequent increased O3 formation in downwind regions [e.g., Kemball-Cook et al., 2010]. Any additional input of reactive VOCs could be important for air quality in the Colorado Front Range as regional municipalities continue efforts to comply with federal air quality standards.

3.6 Fluxes

[56] The major VOCs emitted from natural gas processing, specifically the alkanes, have lifetimes ranging from days to months during the winter. Consequently, year-round natural gas VOC emissions will likely impact the air quality of downwind areas [i.e., Kemball-Cook et al., 2010]. In order to forecast and predict air quality, accurate local/regional emission rate estimates of photochemical smog precursors (i.e., VOCs) are necessary. Here we use the hourly samples from 22 m to calculate fluxes of alkanes and benzene at BAO and estimate regional emissions associated with natural gas production.

[57] Boundary layer development throughout the day and night was monitored using the meteorological parameters measured on the moving carriage on the BAO tower between the surface and 250 m. Vertical profiles of potential temperature were examined for each night to determine when the nocturnal boundary layer (NBL) was fully developed and to estimate the height of the NBL. A shallow nocturnal inversion layer with a depth of 10 to 80 m developed each night. The 22 m inlet height of the canister samples was below the top of the inversion layer on five of the 23 nights (18–19 February, 22–23 February, 27–28 February, 2–3 March, and 3–4 March). The variation of trace gas mixing ratios throughout the day can be calculated using the mass balance equation (equation (3)) [Talbot et al., 2005; Zhou et al., 2005; Sive et al., 2007; White et al., 2008]:

display math(3)

Where [XBL] is the concentration of compound X in the boundary layer, t is time, ER is the emission rate of compound X, H is the boundary layer depth, P is the chemical production rate of compound X, kOH is the rate constant for the reaction of compound X with OH, [OH] is the concentration of OH, Ve is the vertical transfer coefficient, X is the concentration of compound X in the mixed layer above the boundary layer, and advection is a term representing horizontal transport of compound X. The NMHCs investigated here are not produced chemically in the atmosphere; thus, chemical production P reduces to 0. The only important removal mechanism for alkanes is reaction with OH. If only measurements made at night are used, then chemical removal by OH, kOH[OH][XBL], can be neglected. The nitrate radical (NO3) is a potential nocturnal oxidant; however, given the low reaction rate constant for the reaction between NO3 and the alkanes considered here, this chemical removal term can be neglected. Under the stable NBL with low wind speeds and relatively homogeneous regional emissions, it can be assumed that advection and vertical mixing inline image are negligible, and changes in VOC mixing ratios can be attributed to local emissions. Based on our assumptions, equation (3) reduces to the following (equation (4)):

display math(4)

Thus, the flux or emission rate ER can be calculated when the change in trace gas concentration per unit time under the nocturnal boundary layer inline image and the boundary layer height H are known. We used a nocturnal boundary layer height of 40 m for these calculations based on observed potential temperature profiles (not shown) for the five nights used in the calculation.

[58] Hourly average mixing ratios of natural gas associated VOCs varied diurnally. Mixing ratios of these compounds (ethane, propane, i-butane, n-butane, i-pentane, n-pentane, n-hexane, 2-methylpentane, 3-methylpentane, cyclopentane, methylcyclopentane, cyclohexane, methylcyclohexane, and benzene) generally increased under the stable NBL and reached their maximum hourly average mixing ratios at 08:00–09:00 h MST (Figure 10). This nocturnal trend was followed by a steady decrease after sunrise as wind speeds, vertical mixing, and photochemical processing increased until mixing ratios reached a minimum at 14:00 h MST and remained at minimum levels through approximately 22:00 h MST (Figure 10). Emission fluxes were calculated for the compounds which exhibited a strong correlation (r2 > 0.7) between the change in mixing ratio and time between 01:00 and 05:00 h (Table 3). It is worth noting that only alkane and benzene mixing ratios increased (r2 = 0.71–0.98) consistently under the NBL suggesting that a strong local emission source, which was persistent throughout the night, was responsible for the increases observed.

Figure 10.

Hourly mean mixing ratios of selected VOCs. Symbols represent the mean mixing ratio for all hourly samples collected at 22 m on the BAO tower. Error bars represent the standard error of the mean. (a–c) Most VOCs displayed a diurnal cycle with higher mixing ratios during the night and lower mixing ratios during the late afternoon. (d) The alkyl nitrates showed an opposite diurnal pattern indicating their photochemical production during the day. The gray box highlights the 01:00 to 05:00 MST time period used for calculating VOC fluxes.

Table 3. Emission Rates of Selected VOCs at BAO and Extrapolated Regional Emission Rates
 BAO Emission RateaWattenberg FieldbWeld CountycPrevious Estimates for Weld Countyd
Compound109 molecules cm−2 s−1r2Gg yr−1Gg yr−1Gg yr−1
  1. a

    Emission rates were calculated as the change in average concentrations (δC; molecules cm−3) over a 5 h time period (δt; 01:00–05:00 MST) multiplied by the boundary layer height. Only data from nights when the sample inlet was within the NBL were included. Emission rates were calculated for only those compounds whose linear regression between the change in mixing ratio per unit time had an r2 > 0.70. Stated uncertainties are given in parentheses and are the propagated standard error of the linear regression between the change in NMHC concentration per unit time (01:00–05:00 MST) and the assumed variation in nocturnal boundary layer height (40 ± 20 m).

  2. b

    Extrapolated emission rates for the Wattenberg Field were calculated using an area of 2530 km2 representing the extent of the heavily drilled region.

  3. c

    Extrapolated emission rates for Weld County were calculated using an area of 10,341 km2.

  4. d

    Range of estimates from Petron et al. [2012].

Ethane181 (81)0.799.4 (2.1)29 (9) 
Propane168 (80)0.7413 (3)40 (14)15–65
i-Butane39 (17)0.823.9 (0.8)12 (3) 
n-Butane97 (41)0.849.8 (1.9)31 (8)5–42
i-Pentane30 (12)0.863.7 (0.6)12 (3)1–20
n-Pentane28 (11)0.883.4 (0.6)11 (2)1–20
n-Hexane6 (2)0.860.89 (0.16)2.8 (0.6) 
n-Heptane1.4 (0.6)0.860.24 (0.04)0.74 (0.17) 
n-Octane0.4 (0.2)0.860.08 (0.01)0.25 (0.06) 
n-Nonane0.2 (0.1)0.780.03 (0.01)0.11 (0.03) 
Neopentane0.2 (0.1)0.940.03 (0.003)0.09 (0.01) 
2,3-Dimethylbutane0.3 (0.1)0.750.04 (0.01)0.14 (0.05) 
2,2-Dimethylbutane0.4 (0.1)0.860.05 (0.01)0.17 (0.04) 
2-Methylpentane5.3 (2.2)0.860.79 (0.14)2.5 (0.6) 
3-Methylpentane2.8 (1.2)0.830.41 (0.08)1.3 (0.3) 
2,4-Dimethylpentane0.2 (0.1)0.800.04 (0.01)0.12 (0.04) 
2,3-Dimethylpentane0.4 (0.2)0.860.07 (0.01)0.21 (0.05) 
2-Methylhexane0.4 (0.2)0.740.07 (0.02)0.22 (0.07) 
2,2,4-Trimethylpentane0.5 (0.2)0.830.09 (0.02)0.29 (0.08) 
2-Methylheptane0.3 (0.1)0.910.06 (0.01)0.19 (0.03) 
3-Methylheptane0.2 (0.1)0.940.05 (0.01)0.15 (0.02) 
Cyclopentane1.5 (0.6)0.900.18 (0.03)0.57 (0.11) 
Methylcyclopentane3.0 (1.0)0.980.44 (0.03)1.4 (0.1) 
Cyclohexane1.7 (0.6)0.970.25 (0.02)0.8 (0.08) 
Methylcyclohexane1.5 (0.5)0.980.25 (0.01)0.77 (0.06) 
Benzene1.4 (0.7)0.710.18 (0.05)0.57 (0.21)0.05–2

[59] Calculated emission fluxes generally followed the trend in mixing ratios at BAO with the highest fluxes calculated for ethane (181 ± 81 × 109 molecules cm−2 s−1) and propane (168 ± 80 × 109 molecules cm−2 s−1) followed by n-butane (97 ± 41 × 109 molecules cm−2 s−1) and fluxes of other compounds following a decreasing trend with increasing number of carbon atoms (Table 3). The total flux of the NMHCs listed in Table 3 (570 × 109 molecules cm−2 s−1) was higher than the O&NG associated total VOC flux of 370 × 109 molecules cm−2 s−1 reported in 2007 for Garfield County, CO, the county with the second highest number of natural gas wells in Colorado [Colorado Department of Public Health and Environment (CDPHE), 2009]. This value was calculated from total VOC annual emissions reported in CDPHE [2009], using a VOC mixing ratio-weighted molecular mass of 53.5 g mole−1 and an estimated area of natural gas operations in Garfield County of 2000 km2. In addition, the ethane emission rate for a portion of the Anadarko Basin in the southwest U.S. (study area of 590,400 km2) has been estimated at 0.3–0.5 Tg yr−1 [Katzenstein et al., 2003]. This emission rate corresponds to a flux of 32–54 ×  109 molecules cm−2 s−1, which is a factor of 3–5 lower than that estimated during NACHTT. While there is considerable uncertainty associated with the calculation of emission fluxes in all of these studies and the comparison is complicated by differences in the year of the measurement, it is likely that differences in the spatial density of wells or natural gas production volume between the Anadarko Basin, Garfield County, and Weld County contributed to the differences in estimated emission rates, and future studies should examine quantitative relationships between these factors and trace gas emissions.

[60] In order to compare emissions observed during NACHTT to previous emission rate estimates for this region, emission fluxes were extrapolated to regional emission rates for the Wattenberg Field and for Weld County, CO (Table 3). Regional emissions for the Wattenberg Field were calculated using an area of 2530 km2 corresponding to the area of highest spatial density of natural gas wells (see Figure 1), while the regional emission rates estimated using the area of Weld County, 10,341 km2, includes regions outside of the Wattenberg Field where the spatial density of natural gas wells is lower. Our estimates assume that NMHC emission rates at BAO were representative of regional emissions regardless of the number of wells in a given area; therefore, the regional emission rate estimates for the Wattenberg Field and Weld County represent lower bound and upper bound estimates, respectively.

[61] Petron et al. [2012] reported both bottom-up and top-down estimates of emissions of methane, propane, n-butane, i-pentane, n-pentane, and benzene from venting and flaring of natural gas in all of Weld County. Their three top-down emission scenarios were constrained by the observed methane to propane ratio but utilized different inventory-based molar ratio estimates for vented natural gas resulting in the range of emission rates shown in Table 3. A comparison of the top-down and bottom-up estimates by Petron et al. [2012] suggests that O&NG industry emissions estimates of the measured compounds are underreported by a factor of approximately 1.5–2, with the exception of benzene, which was lower by a factor of approximately 5. Propane, n-butane, pentane, and benzene emission rates extrapolated to the area of the Wattenberg Field during NACHTT were on the low end of the range reported by Petron et al. [2012] for Weld County (Table 3), while those extrapolated to the area of Weld County were in the middle of the range of previous estimates. The average of the two emission rate estimates for the Wattenberg Field and for Weld County calculated for NACHTT agrees most closely with the Petron et al. [2012] top-down emission estimate scenario that used the median methane to propane molar ratio from a survey of 77 Wattenberg natural gas samples (median = 15.43; [COGCC, 2007]) and resulted in their lowest top-down emission estimate (28.5 Gg yr−1 propane, 13 Gg yr−1 n-butane, 6 Gg yr−1 i-pentane, 6 Gg yr−1 n-pentane, and 400 Mg yr−1 benzene). Our results support the findings of Petron et al. [2012] but suggest that the discrepancy between emission inventories and actual emissions of these compounds is closer to a factor of 1.5 for the alkanes and a factor of 3 for benzene.

[62] Several factors could have contributed to the lower estimated emissions reported in this study compared to those of Petron et al. [2012]. BAO is located within the Wattenberg Field, but the highest spatial density of natural gas wells was northeast of BAO, and emission rates may be higher in areas with a greater well density. Similarly, emission rates are likely lower throughout the remainder of the county where the spatial density of natural gas wells is lower. Thus, there is considerable uncertainty in the assumption that the emission rate at BAO is representative of all of Weld County or of the Wattenberg Field. Also, NACHTT was conducted during the winter when emissions of these volatile compounds are likely different than during the warmer summer months included in the Petron et al. [2012] study. Our estimates are based on observations of emission rates on five nights; longer-term observations through multiple seasons may prove more representative. Lastly, differences in well drilling and completion practices and natural gas storage, transport, and processing practices may also contribute to the differences in emission rates discussed above.

[63] Annual emissions estimates for Weld County from NACHTT can be compared to previous estimates to examine the effect of the regulations instituted by the CDPHE in 2008. Table 4 compares our measurement-based emissions estimates for the combined total of the natural gas associated VOCs listed in Table 3, excluding ethane, to inventory-based estimates from 2006 (Western Regional Air Partnership Phase III inventory [Bar-Ilan et al., 2008b]). Ethane was excluded because we assumed that the 2006 inventory employed the EPA definition of VOCs, which excludes both methane and ethane, as Bar-Ilan et al. [2008a, 2008b] described their VOC estimates as the total of all state or EPA permitted sources. While the number of active wells increased by 61% and gas production increased by 27% from 2006 to 2011, the change in natural gas associated VOC emissions ranged from a 40% decrease to an 86% increase based on our lower and upper bound estimates, respectively. Using the average of our lower and upper bound estimates of VOC emissions yields a 23% increase relative to 2006 estimates, which is similar to the percent increase in natural gas production from Weld County. Normalizing VOC emissions to the number of gas wells and to the amount of natural gas produced shows that VOC emissions from each individual well likely decreased between 2006 and 2011 but that the large increase in the number of active wells and the decrease in the amount of gas produced per well over this time period resulted in nearly equivalent VOC emissions per unit of gas produced. These comparisons indicate that the 2008 regulations enacted by the CDPHE were most likely effective in controlling VOC emissions but that the continued development of natural gas resources in the region has offset the gains achieved and resulted in greater overall VOC emissions from the region.

Table 4. Comparison of VOC Emissions Estimates From Weld County in 2006 and 2011
 Natural Gas ProductionVOC Emissions
YearActive WellsaGas Produced (Bcf)bTons yr−1cTons yr−1 well−1Tons yr−1 Bcf gas−1
  • a

    The active well count for 2006 was obtained from Bar-Ilan et al. [2008a, 2008b], and data for 2011 were determined from data downloaded from the Colorado Oil and Gas Conservation Commission's data library (http://cogcc.state.co.us/).

  • b

    Production data were obtained from the Colorado Oil and Gas Conservation Commission's Colorado Oil and Gas Information System database (http://cogcc.state.co.us/cogis/).

  • c

    VOC emissions for 2006 are from Bar-Ilan et al. [2008a, 2008b] and VOC emission estimates for 2011 are the sum of the annual emission estimates of all compounds reported in Table 3, excluding ethane, for the upper and lower bound estimates and the average of the two.


[64] The EPA recently instituted new requirements for emission control and green completion of natural gas wells [EPA, 2012]. State and municipal governments across the country are also reevaluating their regulations in response to public concerns about human health and environmental quality. The baseline values for the wide suite of compounds reported here will be useful for determining the efficacy of future regulatory and policy changes on controlling emissions.

4 Summary

[65] Elevated mixing ratios of alkanes, especially C2-C6 alkanes, were observed at BAO during the monthlong NACHTT campaign (18 February to 13 March 2011). Mixing ratios of C2-C5 alkanes reached levels commonly observed in regions with large petrochemical production industries. The corresponding OH reactivity of the VOCs also reached levels typical of urban centers. Alkanes represented over 20% of the total OH reactivity, providing an abundant source of VOCs for downwind O3 production. Vertical profiles conducted on the BAO tall tower showed that mixing ratios of alkanes and other VOCs decreased with height and that elevated VOC mixing ratios extended to at least 250 m above ground level. These observations suggested the presence of a large source of alkanes in northeastern Colorado.

[66] The wind direction dependence of VOC mixing ratios demonstrated that the strongest source of alkanes was to the northeast of BAO, while an urban combustion source was prevalent to the south of the site. The C2-C5 alkane pattern observed at BAO matched that of natural gas extracted from the Wattenberg Field, an extensive natural gas production region to the northeast of BAO. Emission ratios of measured VOCs with tracer compounds showed that two major sources impacted BAO during NACHTT: a regional natural gas-related source and an urban and combustion-related source from the city of Denver to the south and the cities of Boulder, Longmont, and Fort Collins to the west. Elevated emission ratios of C2-C8 alkanes were associated primarily with the natural gas-related source, while C6-C8 aromatics and combustion-related VOCs, including C2-C4 alkenes and ethyne, were primarily associated with urban emissions. The observed alkane pattern, including the unique i-pentane:n-pentane ratio of 1.0, and the observed emission ratios for the natural gas-associated source provide useful tracers for natural gas emissions in future studies. The correlations between DMS and propane and acetonitrile and propane warrant further study of the use of these compounds in natural gas extraction and processing as they may also prove to be useful tracers for emissions from different stages of natural gas production. Emission rates of C3-C5 alkanes were within the range of previous estimates for this region, and this study provides emission rate estimates for a much larger suite of VOCs than previously available. Discrepancies between the emission rates at BAO and those from other natural gas production regions, including the Anadarko Basin, suggest that BAO may not necessarily be representative of emissions in other natural gas production regions. This may be a result of varying natural gas well densities, compositions, production volumes, or production practices. Future investigations should quantify the effects of natural gas well density and production volumes on VOC emission rates to facilitate estimation of more accurate regional, national, and global impacts of natural gas extraction on air quality and atmospheric chemistry.

[67] Annual total VOC emission estimates for Weld County in 2011 during NACHTT were lower than inventory-based estimates from 2006 when normalized to the number of active natural gas wells, indicating that 2008 regulatory efforts aimed at reducing VOC emissions were effective. However, the large increase in the number of active gas wells between 2006 and 2011 likely resulted in greater overall VOC emissions from Weld County in 2011. The emission ratios and compound-specific emissions estimates reported here may be useful in assessing the effects of future regulatory actions on emissions from different stages of natural gas production (e.g., venting and flashing).


[68] This work was supported by the National Science Foundation through a RAPID award to Appalachian State University (ANT-1127774). We acknowledge the use of the Boulder Atmospheric Observatory (BAO), and Daniel Wolfe and Bruce Bartram of the NOAA/ESRL Physical Sciences Division, and William Dubé and Nicholas Wagner of NOAA/ESRL Chemical Sciences Division and CIRES, for their help in conducting the measurements at the BAO. Nicholas Wagner and Steve Brown of NOAA/ESRL also provided the nitric oxide and nitrogen dioxide data. Carbon monoxide data from the BAO were provided by the Global Monitoring Division of NOAA's ESRL and were collected under the auspices of the North American Carbon Program. We also thank Dr. Huiting Mao, Chemistry Department, SUNY College of Environmental Science and Forestry, for helpful comments on early versions of the manuscript.