Article pubs.acs.org/est
Impact of Marcellus Shale Natural Gas Development in Southwest Pennsylvania on Volatile Organic Compound Emissions and Regional Air Quality Robert F. Swarthout,*,†,° Rachel S. Russo,‡,$ Yong Zhou,‡ Brandon M. Miller,‡ Brittney Mitchell,§ Emily Horsman,‡ Eric Lipsky,∥ David C. McCabe,⊥ Ellen Baum,⊥,¶ and Barkley C. Sive‡,#,@ †
Natural Resources and Earth System Science Program, University of New Hampshire, Durham, New Hampshire 03824, United States ‡ Department of Chemistry, Appalachian State University, Boone, North Carolina 28608, United States § Department of Chemistry, Physics and Geoscience, Meredith College, Raleigh, North Carolina 27607, United States ∥ Department of Mechanical Engineering, Penn State Greater Allegheny, McKeesport, Pennsylvania 15132, United States ⊥ Clean Air Task Force, Boston, Massachusetts 02108, United States # Environmental Science Program, Appalachian State University, Boone, North Carolina 28608, United States S Supporting Information *
ABSTRACT: The Marcellus Shale is the largest natural gas deposit in the U.S. and rapid development of this resource has raised concerns about regional air pollution. A field campaign was conducted in the southwestern Pennsylvania region of the Marcellus Shale to investigate the impact of unconventional natural gas (UNG) production operations on regional air quality. Whole air samples were collected throughout an 8050 km2 grid surrounding Pittsburgh and analyzed for methane, carbon dioxide, and C1−C10 volatile organic compounds (VOCs). Elevated mixing ratios of methane and C2−C8 alkanes were observed in areas with the highest density of UNG wells. Source apportionment was used to identify characteristic emission ratios for UNG sources, and results indicated that UNG emissions were responsible for the majority of mixing ratios of C2−C8 alkanes, but accounted for a small proportion of alkene and aromatic compounds. The VOC emissions from UNG operations accounted for 17 ± 19% of the regional kinetic hydroxyl radical reactivity of nonbiogenic VOCs suggesting that natural gas emissions may affect compliance with federal ozone standards. A first approximation of methane emissions from the study area of 10.0 ± 5.2 kg s−1 provides a baseline for determining the efficacy of regulatory emission control efforts.
■
methane emissions.6 Emission inventories have demonstrated that emissions from oil and natural gas sources can dominate VOC emissions in regions of intensive oil and natural gas production.7−9 Independent measurements and estimates have suggested that methane and VOC emissions from natural gas production may exceed industry estimates by a factor of 1.5− 10.6,10−15 However, these reports are geographically and temporally limited and there is a need for independent emission assessments from additional natural gas production regions. Here, we report the results of a regional air sampling effort in the Marcellus Shale region of southwest Pennsylvania. Our measurements demonstrate the magnitude and geographic extent of natural gas emissions. We use a source apportionment model to estimate the contribution of natural gas emissions to
INTRODUCTION Pennsylvania is one of the fastest growing states in natural gas production, with proved reserves and production increasing by factors of 10 and 11, respectively, from 2008 to 2012.1 This increase in reserves and production is a result of advances in directional drilling and hydraulic fracturing, which have led to an increase in technically recoverable natural gas in the Marcellus Shale formation. In 2011, there were over 54 000 active natural gas wells in Pennsylvania.1 Of these, approximately 4800 were unconventional natural gas (UNG) wells, meaning that they were drilled horizontally, hydraulically fractured, or both.2 This number increased by approximately 30% to over 5500 wells by the end of 2012, and by the end of 2013 over 6500 UNG wells were active.2 The rapid expansion of UNG production has raised concerns about greenhouse gas emissions and impacts on air quality. Emissions from UNG production can include toxic species,3 increase downwind ozone production,4,5 and increase the overall greenhouse gas footprint of natural gas use through © XXXX American Chemical Society
Received: September 13, 2014 Revised: December 8, 2014 Accepted: January 16, 2015
A
DOI: 10.1021/es504315f Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
ratory NAM 12 km potential temperature profiles for June 16− June 17. Wind direction and wind speed were also estimated from the NAM analysis. Wind direction was consistently from the southeast (152 ± 18°) at 2.3−4.8 m s−1 during the day and 1.2−3.8 m s−1 at night. A mobile atmospheric observatory was deployed at RCSP from 15 to 16 July and at a semirural, residential site in Hickory from 16 to 17 July. Selected VOCs and oxygenated VOCs (OVOCs) were monitored by a quadrupole proton-transfer reaction-mass spectrometer (PTR-MS; Ionicon Analytik). The PTR-MS monitored 14 m/z ratios on a 2 min cycle. Carbon dioxide (CO2) was measured every minute by infrared spectroscopy (LI-7000, LI-COR Environmental). Canister samples were also collected hourly at each site. Canister samples were analyzed on two multidetector gas chromatography (GC) systems. A five-detector GC system was used to quantify C2−C10 VOCs.16−18 Methane (CH4) and CO2 were analyzed using a GC system equipped with a flame ionization detector and a methanizer prior to the analytical column.19,20 All measurements were calibrated using NISTtraceable standards.
the observed VOC mixing ratios, to estimate the impact of natural gas emissions on exposure to hazardous air pollutants and potential ozone production, and to calculate regional methane emission rates.
■
EXPERIMENTAL SECTION A regional measurement campaign was conducted from 16 to 18 June 2012 in the Marcellus Shale region of southwest Pennsylvania (PA), USA (Supporting Information (SI) Figure S1), to assess the impact of natural gas production on air quality. The campaign involved two measurement components: whole air samples were collected throughout an 8050 km2 area surrounding Pittsburgh, PA (hereafter referred to as regional grid samples); and a mobile laboratory was deployed at a site remote from natural gas activities, with just one well within 10 km, Raccoon Creek State Park (RCSP), and a site in Hickory, PA with 294 UNG wells within 10 km (Figure 1). Regional grid
■
RESULTS AND DISCUSSION Elevated VOC Mixing Ratios near Unconventional Natural Gas Wells. As of July 2012, the 8050 km2 study area contained approximately 1000 active UNG wells, and 8000 active conventional wells (Figure 1). Summary statistics for selected VOCs observed throughout the study area and grouped based on their proximity to natural gas wells are given in Table 1; additional compounds are presented in Table S1. Higher nighttime (22:00−07:00 EDT) mixing ratios of CH4, the most abundant component of natural gas, were observed to the southwest and the northeast of Pittsburgh corresponding with a higher spatial density of UNG wells (Figure 1A). Mixing ratios of CH4 were not elevated near conventional natural gas wells. This may be because the conventional wells in this area produce lower volumes of gas compared to UNG wells, or because emissions are associated with gas development activities (i.e., drilling, hydraulic fracturing and other well completion activities).2 Similar spatial distributions were observed for many of the C2−C8 alkanes (Figure S2). An opposite spatial distribution was seen for combustion and urban emission associated VOCs including ethyne (Figure 1B), ethene, benzene, and C2Cl4 (Figure S2), which were elevated along the urban corridor extending from Pittsburgh to the northwest and lower near UNG wells, which tend to be in less urbanized areas. Alkanes in daytime grid samples (10:00−20:00 EDT) showed no consistent spatial pattern relative to well locations (Figure S2), likely due to dilution of emissions through increased vertical mixing and horizontal transport. The nighttime samples were more indicative of localized sources because emissions were likely constrained within a shallower boundary layer21 as indicated by lower wind speeds during the nighttime sampling period (see Experimental Section). To statistically evaluate the spatial relationship between UNG well locations and trace gas mixing ratios in the nighttime regional grid samples, two comparisons were made. First, the 72 samples were divided into quintiles based on the number of UNG wells within a 10 km radius (Figure 2A−E). Second, because high emissions of fugitive raw natural gas can occur during well construction and completion,22,23 samples were grouped into quintiles based on the distance to the nearest
Figure 1. Geospatial distribution of methane (A) and ethyne (B) in relation to urban areas and UNG wells. Each colored circle represents a single nighttime whole air canister sample and the size and color are proportional to the observed mixing ratio. Recently drilled UNG wells were those with a drilling start date within two months of the study date. Well locations and drilling start dates were from PADEP Oil and Gas Reporting Web site,2 and the processing plant information was obtained from U.S. EIA Natural Gas Annual Respondent Query System.1
air samples were collected into evacuated 2 L passivated stainless steel canisters.16 The full grid was sampled twice over a 24 h period: once during the day (10:00−20:00 EDT on 16 June), and once at night (22:00−07:00 EDT on 16−17 June). Refer to the Supporting Information for additional sampling details. A meteorological analysis of the study dates indicated a shallower, more stable boundary layer during nighttime sample collection than during the day. The boundary layer depth was estimated at 1800 ± 400 m during the day and 400 ± 100 m at night based on seasonal radiosonde data from Holzworth21 for Pittsburgh, PA, and an analysis of National Oceanic and Atmospheric Administration (NOAA) Air Resources LaboB
DOI: 10.1021/es504315f Environ. Sci. Technol. XXXX, XXX, XXX−XXX
a
5249 (6527) 1224−29368 3.5 0.0 74 0
3.0 5.4 × 10−1
1.7 5.5 × 101
2.36 (0.45) 1.8−3.92 8.3 × 10−5 1.7 × 10−3 12 11
2.04 (0.15) 1.82−2.34 1.9 × 10−4 3.6 × 10−4 7 4
2.28 (0.21) 1.88−2.63 1.1 × 10−5 5.0 × 10−3 6 16
2.04 (0.35) 1.81−3.86 2.3 × 10−4 2.7 × 10−4 9 3
1.2 × 10−4 5.8 × 10−4
1.1 × 10−5 5.3 × 10−3
mean (SD) range ERpropanea ERethynea UNG (%) combust (%)
mean (SD) range ERpropanea ERethynea UNG (%) combust (%)
mean (SD) range ERpropanea ERethynea UNG (%) combust (%)
C
mean (SD) range ERpropanea ERethynea UNG (%) combust (%)
ERpropanea ERethynea
ERpropanea ERethynea 1.0 0.0
1.0 0.0
1404 (1550) 431−9342 1.0 0.0 84 0
13140 (10900) 1980−50460 1.0 0.0 98 0
1893 (1152) 630−5115 1.0 0.0 74 0
3911 (5015) 437−23188 1.0 0.0 93 0
1665 (2789) 304−23188 1.0 0.0 82 0
propane (pptv)
n-butane (pptv) isopentane (pptv)
regional grid day and night samples (N = 144) 266 (512) 476 (765) 276 (317) 39−4268 86−6210 62−2233 1.8 × 10−1 2.7 × 10−1 9.5 × 10−2 0.0 1.2 × 10−1 4.8 × 10−1 85 76 40 0 4 22 night samples near wellsb (N = 29) 650 (910) 1123 (1404) 481 (468) 49−4268 106−6210 92−2233 1.8 × 10−1 2.8 × 10−1 8.4 × 10−2 0.0 0.0 8.6 × 10−1 94 91 61 0 0 25 night samples remote from wellsc (N = 32) 311 (205) 708 (467) 347 (261) 72−840 178−1726 81−1172 1.5 × 10−1 2.4 × 10−1 1.5 × 10−1 −1 0.0 2.1 × 10 3.7 × 10−1 73 57 41 0 14 29 Hickory (N = 24) 1778 (1291) 4348 (3438) 976 (618) 251−5898 646−15328 173−2838 1.2 × 10−1 3.1 × 10−1 5.5 × 10−2 3.2 2.9 2.7 84 93 73 13 5 20 Raccoon Creek State Park (N = 24) 221 (289) 408 (374) 306 (230) 45−1697 97−2057 88−1255 1.8 × 10−1 2.3 × 10−1 7.9 × 10−2 −1 0.0 3.0 × 10 6.1 × 10−1 73 58 24 13 26 51 isopentane/n-pentane ratio >1.20 (N = 134) 1.8 × 10−1 2.5 × 10−1 9.5 × 10−2 −1 0.0 4.1 × 10 5.7 × 10−1 isopentane/n-pentane ratio 45 wells within 10 km. c