Methane, Black Carbon, and Ethane Emissions from Natural Gas

Subscriber access provided by HACETTEPE UNIVERSITESI KUTUPHANESI

Article

Methane, black carbon, and ethane emissions from natural gas flares in the Bakken Shale, ND Alexander Gvakharia, Eric A. Kort, Adam R. Brandt, Jeff Peischl, Thomas B. Ryerson, Joshua P. Schwarz, Mackenzie L Smith, and Colm Sweeney Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b05183 • Publication Date (Web): 12 Apr 2017 Downloaded from http://pubs.acs.org on April 13, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 29

Environmental Science & Technology

TOC Art 84x47mm (300 x 300 DPI)

ACS Paragon Plus Environment


Page 2 of 29

Methane, black carbon, and ethane emissions from natural gas ares in the Bakken Shale, ND Alexander Gvakharia,

∗, †

Eric A. Kort, † Adam Brandt, ‡ Je Peischl, ¶ § Thomas B. ,

Ryerson, § Joshua P. Schwarz, § Mackenzie L. Smith, † and Colm Sweeney ¶ †Climate and Space Sciences and Engineering, University of Michigan, Ann Arbor,

Michigan, USA ‡Department of Energy Resources Engineering, Stanford University, Stanford, California,

USA ¶Cooperative Institute for Research in Environmental Science, University of Colorado

Boulder, Boulder, Colorado, USA §NOAA ESRL Chemical Sciences Division, Boulder, Colorado, USA E-mail: [email protected]

Phone: 757 232-3228 1

Abstract

2

Incomplete combustion during aring can lead to production of black carbon (BC)

3

and loss of methane and other pollutants to the atmosphere, impacting climate and

4

air quality. However, few studies have measured are eciency in a real-world setting.

5

We use airborne data of plume samples from 37 unique ares in the Bakken region of

6

North Dakota in May 2014 to calculate emission factors for BC, methane, ethane, and

7

combustion eciency for methane and ethane. We nd no clear relationship between 1


Page 3 of 29


8

emission factors and aircraft-level wind speed, nor between methane and BC emission

9

factors. Observed median combustion eciencies for methane and ethane are close

10

to expected values for typical ares according to the US EPA (98%). However, we

11

nd that the eciency distribution is skewed, exhibiting lognormal behavior. This

12

suggests incomplete combustion from ares contributes almost 1/5 of the total eld

13

emissions of methane and ethane measured in the Bakken shale, more than double the

14

expected value if 98% eciency was representative. BC emission factors also have a

15

skewed distribution, but we nd lower emission values than previous studies. The direct

16

observation for the rst time of a heavy-tail emissions distribution from ares suggests

17

the need to consider skewed distributions when assessing are impacts globally.

TOC Art

18

Introduction

19

Over 140 billion cubic meters (BCM) of gas is globally ared each year. 1 Flaring is used to

20

dispose of gas at production and processing facilities that lack infrastructure and means to

21

capture or use the gas. The United States ares about 8 BCM per year, with almost half of

22

that coming from North Dakota alone. 2 From 2004 to 2014, the amount of gas annually ared

23

in North Dakota increased from 0.08 BCM to 3.7 BCM, and in 2014 about 28% of North

24

Dakota's total produced natural gas was ared. 3 Flaring has implications for the atmosphere.

25

Although ideally gas would be captured instead of lost, it is preferable to are rather than 2



26

vent, since aring destroys methane (CH 4 ) and volatile organic compounds which aect air

27

quality, converting them to CO 2 . CH4 is a potent greenhouse gas, the 2nd most important

28

anthropogenic greenhouse gas behind CO 2 based o integrated radiative forcing. 4,5 Flaring

29

is not 100% ecient, and through incomplete combustion it can be a source for CH 4 and

30

VOCs. 6,7 Flaring can also create black carbon (BC) as a byproduct, an anthropogenic forcer

31

of climate with public health implications. 811 The World Bank recently introduced a "Zero

32

Routine Flaring" initiative to end aring worldwide by 2030 through government incentives

33

and institutional cooperation, hoping to mitigate economic losses due to aring and relieve

34

its burden on the atmosphere. 12

35

Inventories that account for aring often use a combustion eciency value of 98% of the

36

initial gas, citing an EPA technical report. 13,14 This eciency value assumes are stability,

37

and can decrease based on wind speed and other factors such as ow rate or aeration.

38

Studies have investigated are eciency in laboratories using scaled-down are simulations

39

in a controlled environment, reporting 98-99% are combustion eciency, 15,16 but there

40

have been few eld studies done to assess are eciency and directly measure emissions

41

in a real-world environment. Thus, scaled-up laboratory results may not be representative

42

of real-world aring. A study of two are sites in Canada calculated an average observed

43

combustion eciency of 68 ± 7%, much lower than the assumed eciency. 17 One remote

44

sensing study in the Netherlands found high eciencies of 99% but only analyzed three

45

ares, with up to 30% error in the measured gas concentrations, and noted the lack of in

46

situ data. 18 There was also a comprehensive study to observe industrial are emissions and

47

eciency but the tests were conducted at a are test facility, not directly at well sites. 19

48

To our knowledge the only extensive study of in situ are eciency for CH 4 sampled ten

49

ares in the Bakken Shale in North Dakota and one in western Pennsylvania. 20 This study

50

reported high are eciencies up to 99.9% but based on their identication techniques, they

51

acknowledged a possible bias towards larger, brighter burning, and thus more ecient, ares.

52

Black carbon emissions from gas aring have been investigated, but there are not many

3


Page 4 of 29

Page 5 of 29


53

studies that use direct observations of aring. Schwarz et al. (2015) 10 quantied total eld

54

emissions of BC and derived an upper-bound on BC emission factor for aring from the

55

Bakken using the same aircraft campaign data as used in this paper. Their emission factor

56

was obtained using BC ux calculated with a mass balance technique for the entire eld.

57

Hence, it did not target individual ares. It includes all BC sources in the region (e.g. diesel

58

trucks, generators, limited agriculture, etc.) and is expected to provide an upper bound.

59

Weyant et al. (2016) 21 calculated BC emission factors from targeted ares in the same

60

region and found an average value well below the upper bound of Schwarz et al. (2015), and

61

to our knowledge is the only previously published peer-reviewed study of BC emissions from

62

aring that directly sampled ares. BC emission factors have been shown to vary based on

63

fuel chemistry and stability of the are, necessitating the use of specic emission factors or

64

a distribution rather than using a single average value as representative. 22

65

The lack of direct, in situ observations of aring eciency suggests that estimates of

66

emissions from incomplete combustion may be inaccurate. Also, using a single value for

67

aring emission factors or combustion eciency does not take into account the various pa-

68

rameters that may aect a are, 23 and a statistically robust sample of aring eciency

69

would help identify a representative distribution. Total fugitive emissions from oil and gas

70

production and leakage can be a substantial source of atmospheric CH 4 and are underrepre-

71

sented in inventories. 24 Studies have observed non-normal distribution of CH 4 emissions in

72

some elds, where less than 10% of sampled sources contributed up to 50% of the sampled

73

emissions. 2529 A study of are emissions using Greenhouse Gas Reporting Program and Gas

74

Emission Inventory data found that 100 ares out of 20,000 could be responsible for over

75

half the emissions in the US, but this conclusion results from the non-normal distribution

76

of gas volume ared and not from a skewed are combustion eciency (which is not repre-

77

sented). 30 In addition to the non-normal distribution of gas volume ared, there may be a

78

skewed distribution of emissions from incomplete combustion in ares based on eciency as

79

well.

4



80

We present an analysis of combustion eciency and emission factors of CH 4 , BC, and

81

C2 H6 for thirty-seven distinct ares in the Bakken Shale Formation in North Dakota using

82

data obtained during a May 2014 aircraft campaign, to our knowledge the largest study of

83

aring emissions in the eld based on number of ares and the rst to include C 2 H6 . This

84

gives us sucient statistics to obtain an eciency distribution and determine the implications

85

for total fugitive emissions from incomplete combustion in actual eld conditions.

86

Methods

87

Flights and Instrumentation

88

All observations used in this analysis were made as part of the TOPDOWN 2014 (Twin Otter

89

Projects Dening Oil/gas Well emissioNs) study, and were collected on-board a National

90

Oceanic and Atmospheric Administration (NOAA) DHC-6 Twin Otter aircraft. 10,31,32 This

91

campaign focused on understanding the atmospheric impact of fossil fuel extraction activities.

92

17 research ights were conducted on 11 separate days between May 12-26, 2014, totaling

93

40 hours. Flights were typically 3-3.5 hours in duration, and were primarily conducted at

94

low-altitudes (400-600 magl) within the planetary boundary layer at an average speed of 65

95

m/s. Vertical proles were performed in each ight to dene the mixed layer height. Flights

96

dedicated to mass balance conducted transects around the Bakken region, and although a

97

few ares were sampled during these transects, most of the ares were identied on "mowing-

98

the-lawn" ghts that swept across the region to target point sources as well as some ights

99

dedicated to point source identication. Flares were circled multiple times during these

100

ights between 400-600 magl although some were sampled higher up, around 1000 magl.

101

Flares were not specically targeted for any particular characteristic such as size, brightness,

102

or aring volume. Flares were sampled over the entire region rather than in a particular

103

cluster, giving low spatial sampling bias. However, due to the nature of the sampling,

104

brighter ares were more easily identiable from the plane and thus more likely to have been 5


Page 6 of 29

Page 7 of 29


105

targeted. Not all passes by a are produced a well-dened peak that could be used in the

106

eciency analysis. Many of the ares were sampled at a distance on the order of hundreds

107

of meters to kilometers downwind. This gave the are plume time to disperse and allowed

108

us to measure large plumes over a time period of 10-20 seconds, providing more data per

109

plume than if we sampled closer and lower.

110

CH4 , CO2 , carbon monoxide (CO), and water vapor (H 2 O) were measured with a Picarro

111

2401-m cavity ringdown spectrometer with a sampling rate of 0.5 Hz. CH 4 was measured

112

with an accuracy of ±1.4 ppb and precision of ±0.2 ppb; CO2 with an accuracy of ±0.15

113

ppm and precision of ±0.03 ppm. 33,34 An Aerodyne mini direct absorption spectrometer was

114

used to continuously measure C 2 H6 , deployed as described previously in literature 35,36 along

115

with hourly measurements of a standard gas to conrm stability. 32 Sampling was conducted

116

at 1 Hz with precision of < 0.1 ppb and average accuracy of ±0.5 ppb. 32 Due to the Aerodyne

117

ethane instrument having a response time of 1 second, compared to the Picarro's 2 second

118

response time, there were sharper, narrower peaks in C 2 H6 than CO2 and CH4 . To enable a

119

point-by-point comparison of C 2 H6 to CO2 , a weighted moving average (WMA) was applied

120

to the C2 H6 data. The total integrated value of the C 2 H6 peak did not signicantly change

121

with the WMA lter, indicating conservation of mass with the method.

122

All trace gases are reported as dry air mole fractions, converted from the measured

123

wet air mole fractions using water vapor observations from the Picarro. A single-particle

124

soot photometer (SP2 by Droplet Measurement Technology Inc., Boulder, CO) was used to

125

measure refractory black carbon (rBC) for particles containing rBC in the mass range of 0.7-

126

160 fg. The SP2 provided 1-second rBC mass-mixing ratios with systematic uncertainty of

127

25%. 10,37 Two dierential GPS antennae on the fuselage of the Twin Otter provided aircraft

128

heading, altitude, latitude, longitude, ground speed, and course over ground. Wind speed

129

was calculated as described in Conley et al. (2014), 38 with estimated uncertainties of ±1

130

m/s in magnitude and ±6◦ in direction. A Rosemount de-iced Total Temperature Sensor,

131

model number 102CP2AF, measured ambient temperature. Calibration before and after the

6



132

eld project indicate measurement performance with precision of ±0.2 ◦ C and accuracy of

133

±1.0 ◦ C.

134

Flare Identication

135

We identied ares in the following ways. During the science ights, all signicant events

136

were logged, including when the plane ew by a are. These ight notes thus provide times

137

when a are was visually conrmed, and these are plumes were identied in the data for

138

the corresponding ight and agged. After locating all the ares conrmed by the ight

139

notes, we searched through the remaining data to nd plumes that could be possible ares

140

but were not noted during the ight, such as smaller ares that might have been hard to

141

see on the ground. To identify the other possible are plumes, we looked for peaks in CO 2

142

where ∆CO2 , the peak enhancement, was greater at its maximum point than 4 σ of the CO2

143

background variability, indicating a statistically signicant elevation of CO 2 as a result of

144

combustion from a are. We also looked for a peak less than 20 seconds in time. At a mean

145

ground speed of 65 m/s this corresponds to a source about 3 km away using Gaussian plume

146

theory, 39 which is about the distance we tended to sample where the plume still presented

147

a robust signal above background. Figure 1 shows the research aircraft ight paths, known

148

are locations, 3 and where we sampled plumes.

7


Page 8 of 29

Page 9 of 29


Figure 1: Left panel shows ight paths (black lines), wells with known aring (grey triangles), 3 and are plume locations (red points) from the TOPDOWN 2014 campaign in the Bakken eld in northwest North Dakota. Times when the plane circled around an area multiple times to repeatedly sample can be seen in the middle of the region. Right panel is zoomed in on a single are plume, with ight path (black line), are plume (red points), and wells with reported aring (lled triangles with corresponding monthly aring amounts). The arrows indicate the wind direction. We used the wind direction, distance from well, and aring amount to verify that the plume was caused by aring.

149

To verify if these additional plumes identied in the data were indeed caused by aring, we

150

co-plotted the locations of these events with all nearby wells with reported aring and other

151

CO2 producers such as processing facilities and gas plants using the EPA GHG Reporting

152

Program as seen in Figure 1. Certain are locations were cross-checked with additional data

153

from the VIIRS Active Fire Map and North Dakota Oil and Gas ArcIMS Viewer. Then, using

154

Gaussian plume theory, we estimated how far away the source of a plume was based on the

155

plume width and wind conditions, matching the plume to a possible are source. 39 Although

156

the science ights were conducted on days with steady winds, leading to low variability

157

in wind speed and direction, we accepted plumes that were within 20% of the theoretical

158

distance to account for deviation in other factors such as not ying directly in the center

159

of the diused plume. If a plume was located downwind from a well with aring, was not

160

downwind of another CO 2 source, and had a width and distance consistent within 20% of

161

Gaussian plume theory, we considered it likely due to a are source. If a plume was not

8



162

downwind of a are at a distance consistent with Gaussian plume theory, or had interference

163

from another CO2 source, we omitted it from the analysis. 39 are plumes were identied

164

with the ight notes, and out of 17 additional plumes in the data, 13 were accepted using

165

our verication method and 4 rejected for a total of 52 are plumes from 37 unique ares.

166

Other sources for methane or black carbon closely co-located with ares (such as diesel

167

engines or fugitive losses from production wells) could contribute to the observations we are

168

attributing to aring, and we assess their potential impact on our analysis here. Using gas

169

composition data from over 550 samples, the average chemical plume from the Bakken is

170

0.7% CO2 , 3.7% N2 , 49% CH4 , 21% C2 H6 , and the rest in higher-order hydrocarbons. 40 This

171

results in a molecular weight of about 29 g/mol, close to that of air and nearly double the

172

weight of natural gas from other elds with higher CH 4 ratios. 21 An unburned source of gas

173

is therefore neutrally buoyant compared to a hot are exhaust plume which will rise in the

174

atmosphere. 41 However, the are plume can entrain these other sources, mixing them as the

175

buoyant plume rises in the atmosphere. If we assume a are converts 98% of its hydrocarbons

176

to CO2 , and that enhancements near a well pad due to other emissions are 50 ppm CH 4 and

177

415 ppm CO2 , then if the are plume entrained a volume equal to its own (50% dilution) the

178

resulting CH4 /CO2 slope measured by the aircraft (see Figure 2) would change by less than

179

1%, smaller than the uncertainty range in tting the slope. Considering typical values for

180

methane and CO2 enhancement (40 ppb and 5 ppm on average, respectively), we estimate

181

the slope error (and thus error on calculated emission factors) would be less than 1% with

182

10% as a upper bound. Adjusting the are eciency in this estimation does not signicantly

183

aect the result (using a 90% combustion eciency, all else equal, would also have an impact

184

of 1% on the slope). Although we cannot denitively rule out all potential contributions for

185

such sources to the plumes we are analyzing, these considerations of possible entrainment

186

suggest it is not signicant in this analysis, though the potential impact would suggest our

187

results may represent a lower bound for combustion eciency.

9


Page 10 of 29

Page 11 of 29


188

Combustion Eciency

189

Destruction eciency and emissions factors were calculated for each are sampled. Black car-

190

bon emission factors were determined following the methodology of Weyant et al. (2016), 21

191

using eq. 1

EFBC = 1000 × F

CCO2

CBC + CCH4 + CBC

(1)

192

Here CCO2 , CCH4 , and CBC are the mass concentrations of carbon in g/m 3 for each

193

product with the respective background removed and F is the ratio of carbon mass to total

194

hydrocarbon mass, calculated to be 0.79 from gas composition data for the Bakken. 40 CO2

195

and CH4 data were converted from molar ratios to g/m 3 using a molar volume at standard

196

temperature and pressure (273K, 1013 mb) to match the conditions of the BC mass concen-

197

trations. This EFBC value is given in grams of BC per kg of gas, and can be converted to

198

g/m3 using a gas density of 1.23 ± 0.14 kg/m3 for the composition. 40 For some of the ares

199

we did not detect a strong BC enhancement correlated with CO 2 , causing skewed or negative

200

emission factor values. To account for this, when the peak enhancement ∆BC was below

201

the detection limit of 4 σ of the background, we used a value of half the detection limit in

202

the EF calculation as in Weyant et al. (2016); 21 this was observed in a third of the plumes.

203

The measured BC concentrations were scaled up by 15% to account for accumulation-mode

204

mass outside of the SP2 detection range, as described in Schwarz et al. (2015); 10 rBC mass

205

in either the coarse mode or a sub-accumulation mode size range would not be accounted

206

for by this adjustment. Generally, as in Schwarz et al. (2015), 10 the accumulation mode

207

size distribution is well-t with a lognormal function, and any additional smaller or larger

208

populations of BC particles are revealed by deviations from the lognormal t at the smaller

209

or larger limits of the detection range respectively. Here, there was no evidence of additional

210

non-accumulation modes.

211

Emission factors for CH 4 and C2 H6 were obtained by rst calculating the peak enhance-

10



Page 12 of 29

212

ment of CH4 , C2 H6 , and CO2 . We calculated a mean background value for each plume

213

using the concentration data from 5-10 seconds before the start and after the end of the

214

plume, and then subtracted the background from the plume values to obtain ∆CH4 , ∆C2 H6 ,

215

and ∆CO2 . ∆CH4 and ∆C2 H6 were t with a Reduced Major Axis (RMA) regression to

216

∆CO2 for each peak to obtain the emission factor, in ppm CH 4 or C2 H6 per ppm CO2 . 20

217

Figure 2 shows an example plume from a are and its CH 4 regression. Regressions were

218

well-correlated with 10-20 data points in each are plume. Uncertainty in EF for CH 4 and

219

C2 H6 was given by 95% condence intervals from the regression. For all plumes, EF BC from

220

eq. 1 linearly correlated with the slope of BC vs CO 2 with an R2 of 0.97. This t was used

221

to derive uncertainty in EF BC from 95% condence intervals of the regression of BC and

222

CO2 .

Figure 2: Example of a aring plume with CO 2 , CH4 , C2 H6 time series and regression to nd CH4 EF.

223

We calculated the destruction removal eciency (DRE) following the methodology of

224

Caulton et al. (2014) 20 using eq. 2, with a small correction to report the value as the

225

fraction of gas destroyed rather than remaining.

DRE(%) = 1 − 226

µCH4 ∗ 100 ((X) ∗ µCO2 ) + µCH4

(2)

µCH4 and µCO2 are the gas concentrations in ppm and X is the carbon fraction of CH 4 11


Page 13 of 29


227

in the total fuel gas before combustion. From gas composition data for the eld, 40 the value

228

of X is 0.26 ± 0.05 for CH4 .

229

This DRE calculation was done two ways. First, by integrating over the entire peak

230

to obtain a DRE value from the total integrated amount of CH 4 and CO2 . Second, by

231

calculating the DRE value for each point in the peak individually to get an aggregate DRE

232

dataset as seen in Caulton et al. (2014). 20 The respective baseline values were removed

233

from each gas concentration in both methods. Since the integral method calculates DRE

234

using the average concentration over the sampling time of the gases in the plume, and

235

the point-by-point mean represents the average instantaneous DRE, a signicant divergence

236

between the results would be indicative of a potential problem with the approach. For all

237

ares the integrated DRE diered from the mean point-by-point DRE by 1% on average,

238

demonstrating robustness between the two methods. C 2 H6 DRE was also calculated using

239

both methods, with X = 0.23 ± 0.03 for C2 H6 . The eect of X's variability on the DRE is

240

small and within the calculated uncertainty for DRE.

241

Detection Threshold

242

We compared the standard deviation of CH 4 background and the maximum peak CO 2 en-

243

hancement to calculate a "noise DRE" using eq. 2 to assess the impact of a potential signal

244

produced by background variability on the DRE. The distribution suggests a sensitivity

245

threshold around 99%. We compared the sensitivity distribution to the measured DRE

246

distribution, and an analysis of variance between the two produced a p value of 9 ×10−7 ,

247

suggesting they are statistically signicantly dierent. Thus it would be dicult to dis-

248

tinguish measured DRE values of greater than 99% as signicant compared to background

249

variability, but values less than 99%, as we have observed, are robustly detectable with our

250

approach. There is a trade-o between our sampling approach and the one used by Caulton

251

et al. (2014), 20 where they ew lower and closer to the ares. With our ights, we obtained

252

more points in each plume, allowing us to calculate regression lines for emission factors. 12



253

However, we encountered a lower signal-to-noise ratio, making it more dicult to precisely

254

measure the DRE of very ecient ares. We used the dierence between 100% and the DRE

255

calculated using the sensitivity as a proxy for DRE uncertainty in each individual are.

256

Results

257

Emission Factors

258

Figure 3 shows the calculated CH 4 and C2 H6 emission factors plotted against mean aircraft-

259

level wind speed for all are plumes. Previous laboratory are studies have observed a strong

260

nonlinear dependence of ineciency on crosswind speed, 15,16 and Caulton et al. (2014) 20 ob-

261

served a weak relationship in the ares they sampled in the eld. Considering our observed

262

emission factors and crosswind speeds we nd similar results to Caulton et al. (2014). An

263

exponential t of our data suggests a weak dependence, with parts of the data possibly follow-

264

ing dierent distinct curves. A Pearson correlation analysis of the data and the exponential

265

t produced a weak correlation coecient (0.34). Gas exit velocity and are parameters like

266

the stack diameter can aect the ineciency curve and may be the reason for the apparent

267

presence of multiple curves, but unfortunately these values were not known for our sampled

268

ares. More specic knowledge of the gas composition and ow rate would potentially be

269

illuminating for the possible bimodal distribution in CH 4 EF of low-eciency emitters (>30

270

ppb/ppm) and high-eciency emitters (0-20 ppb/ppm) but we can only hypothesize without

271

detailed information on specic ares at time of our sampling.

272

Some ares were circled repeatedly or revisited on dierent days, and so we transected

273

multiple plumes from the same are. The calculated EF for the are was not consistent

274

between dierent plumes, suggesting uctuation in the eciency. Caulton et al. (2014)

275

found large overall variability in CH 4 EF and inconsistency between sampling on dierent

276

days, but attribute the variability to the small sample size of their plumes. 20 Weyant et

277

al. (2016) reported inconsistent emissions of BC for ares sampled on dierent days, and 13


Page 14 of 29

Page 15 of 29


278

observed large variability in BC EF for multiple passes of the same are, citing variability

279

in gas ow rate and gas composition as possible sources. 21 From our data alone we cannot

280

resolve the cause of same-are variability, but it is apparently a feature consistent across

281

studies.

282

We did not observe a clear relationship between EF and wind speed for plumes from the

283

same are, possibly due to factors like ow rate or exit velocity. For some ares that were

284

sampled multiple times, we did not get a sharp, identiable peak in CO 2 or CH4 on every

285

pass, and so we were not able to analyze all possible passes. The EF calculation included

286

background points in the regression, removing these points from the t and forcing the line

287

through zero did not signicantly aect the results. Comparing CH 4 and C2 H6 emission

288

factors for each plume, we found a linear relationship with a R 2 value of 0.57, as plumes

289

with higher emissions of CH 4 had corresponding higher emissions of C 2 H6 , suggesting that

290

combustion eciency is somewhat uniform across hydrocarbons.

14



Figure 3: CH4 and C2 H6 EF plotted against wind speed for all plumes, with an exponential t in red. Error bars represent 95% condence intervals in EF and 1 σ in wind speed.

291

Like Weyant et al. (2016), 21 we did not observe a dependence between BC emission factor

292

and CH4 EF for each plume. Elevated CH 4 emissions from a are do not necessarily indicate

293

higher or lower BC emission. Adding in an ethane term to eq. 1 did not signicantly change

294

the BC EF calculation, as the ppm-order enhancement of CO 2 dominates the ppb-order

295

enhancement of C2 H6 and CH4 .

296

Figure 4 shows the distribution and probability function of BC emission factor in g

297

BC/kg gas. The distribution is right-skewed, matching the results of Weyant et al. (2016), 21

298

and was t with a lognormal density using a maximum-likelihood method. The lognormal

15


Page 16 of 29

Page 17 of 29

299


distribution function is given by

f (x) = p

1

2 /(2σ 2 ))

(2π)σx

e−((logx−µ)

(3)

300

µ and σ are the mean and standard deviation of the logarithm. The Pearson correlation

301

coecient between the BC emission factor probability distribution and the lognormal dis-

302

tribution resulted in a correlation coecient of 0.96. We present the lognormal t as a way

303

to illustrate the skewed distribution and provide a quantitative representation. Results de-

304

rived from the combustion eciency distribution use the raw distribution, rather than an

305

approximation with the lognormal t.

Figure 4: On the left, a histogram of black carbon emission factor for all are plumes, with lognormal density (red line). On the right, distribution function of BC EF in black with lognormal distribution function in red.

306

We report BC EF from ares in g/kg, grams of BC produced per kilogram of hydrocarbons

307

in the fuel gas. The values ranged from 0.0004 to 0.287 g/kg. We can convert from g/kg

308

to g/m3 using a ared gas density of 1.23 ± 0.14 kg/m3 , 40 allowing us to express BC EF

309

in terms of gas ared volume and to compare the results with previous studies. Even with

310

the observation of a right-skewed distribution, our analysis nds lower BC emissions than

311

previously reported. Schwarz et al. (2015) 10 provide an estimate for all the BC sources in the

312

Bakken of 0.57 ± 0.14 g/m3 . This upper bound on aring is twice the highest emission value

313

we observed (Figure 4). Similarly, laboratory analysis by McEwen et al. (2012) 22 reported 16



314

emissions much larger than we observe (0.51 g/m 3 , o scale in Figure 4). The mean value of

315

0.13 ± 0.36 g/m3 measured with an SP2 by Weyant et al. (2016) 21 is within our observed

316

range, though it falls within the top 20% of emitters we observed. Our observed in-eld

317

ares thus appear to have produced less BC than would be predicted from previous studies.

318

The median, mean, and standard deviation of the mean BC emission factor we observed

319

were 0.021 g/m3 and 0.066 ± 0.009 g/m3 (or 0.017 and 0.053 ± 0.008 g/kg), respectively,

320

though given the skewed distribution care needs to be taken in interpreting these values.

321

Given that 3.7 BCM of gas was ared in the Bakken eld in 2014, 3 applying that to the

322

entire distribution of BC EF in g/m 3 suggests total BC emissions from aring of 0.24 Gg

323

BC/yr. However, the top quartile of ares contribute disproportionately, 0.17 Gg BC/yr,

324

which is 70% of the total emissions from ares. Overall, our emission rate of 0.24 Gg BC/yr

325

is two-thirds the rate of 0.36 Gg BC/yr calculated by Weyant et al. (2016) 21 for ares and

326

17% of the total Bakken emission rate (1.4 Gg BC/yr) reported by Schwarz et al. (2015). 10

327

Based on these results, using a single emission factor to estimate emissions from ares in

328

a region does not properly represent the wide variability in emissions that may be present.

329

Total emissions from aring could potentially be substantially reduced if the least ecient

330

ares alone are identied and addressed.

331

Combustion Eciency

332

For methane and ethane, the percent of gas remaining provides a useful metric for are

333

eciency - this is simply 100-DRE. In Figure 5 the distribution of percent remaining CH 4

334

and C2 H6 is illustrated and a lognormal relationship is apparent. As with emission factors,

335

we found a linear relationship between CH 4 and C2 H6 DRE for each plume, with a R 2 of

336

0.53.

17


Page 18 of 29

Page 19 of 29


Figure 5: Histogram of remaining CH 4 and C2 H6 (100-DRE) with density curve (dashed black) and lognormal t (red). These distributions were integrated to calculate the emissions due to incomplete combustion.

337

A Pearson correlation analysis of the DRE probability distributions and the lognormal

338

t distribution produced a correlation coecient of 0.99 for both CH 4 and C2 H6 . The

339

distribution of CH 4 and C2 H6 emission factors, which are theoretically consistent with the

340

DRE calculations, also exhibit a skewed distribution though a lognormal relationship is not

341

as apparent. The median DRE for CH 4 is 97.14 ± 0.37 using the integral method and 96.99

342

± 0.23 using the aggregate dataset. For C 2 H6 the median DRE is 97.33 ± 0.27 and 97.36

343

± 0.25 respectively. These median values are close to the expected eciency (98%), but the

344

right-skewed distribution indicates that 98% is not a representative destruction eciency

345

and would over-predict methane and ethane destruction.

346

We can assess the impact of this observed skewed distribution by considering the contri-

18



347

bution of incomplete are combustion to total eld methane and ethane emissions. Using

348

aircraft data and a mass balance technique, Peischl et al. (2016) 31 calculated a methane

349

ux for the Bakken region that extrapolates to an annual ux of 0.25 ± 0.05 Tg CH4 /yr.

350

As with black carbon, we can use reported aring gas volumes for North Dakota in 2014 3

351

and integrate the distribution of observed DRE values to produce an estimated emission of

352

methane from incomplete combustion of 0.052 Tg CH 4 /yr, or 21% ± 4% of the total emis-

353

sions reported by Peischl et al. (2016), using the uncertainty bound on the ux calculation.

354

This is more than double the contribution one would nd if the expected value of 98% was

355

assumed representative of the eld, which would predict emissions representing 8% ± 1.6% of

356

the total eld emissions. Caulton et al. (2014) reported much higher combustion eciencies,

357

and applying their median 99.98% value would suggest only a fraction of a percent (0.13%

358

± 0.03%) of the total eld emissions was from incomplete combustion in ares.

359

We performed the same analysis for ethane, and compareed with the total eld emissions

360

estimate of 0.23 ± 0.07 Tg C2 H6 /yr reported in Kort et al. (2016). 32 Again our observed

361

combustion eciencies suggest incomplete combustion from ares contributes substantially

362

to total eld emissions, 17% ± 5% of the total emissions (0.039 Tg/yr), more than double

363

that predicted by using 98% as a representative value.

364

The observed lognormal distribution results in a disproportionate impact from ares ex-

365

hibiting poor combustion eciencies. We nd the top quartile of methane emitters contribute

366

0.036 Tg CH4 /yr, which is 69% of total emissions from incomplete are combustion (and 14%

367

of the total eld emissions). Similarly, for ethane, the top quartile of emitters contributes

368

0.026 Tg C2 H6 /yr which is 66% of the total emissions from incomplete are combustion (and

369

12% of total eld emissions).

370

Why do we nd higher methane emissions and lower black carbon emissions than other

371

studies conducted in the Bakken shale? 10,20,21 We cannot denitively pinpoint the reason.

372

We sampled in the same subregion of the Bakken as Caulton et al. (2014), 20 though we

373

did not sample any of the same ares they did, and our campaign was 2 years after theirs.

19


Page 20 of 29

Page 21 of 29


374

Weyant et al. (2016) 21 did not report specic are locations but were likely in the same

375

subregion as well, 2 months before our campaign.

376

There is a dierence in sampling methods, which could contribute. Caulton et al. (2014)

377

ew low and close to the ares, though specic altitudes and distances are not reported. 20 As

378

we did not specically target larger (and so potentially more ecient) ares, our approach

379

makes it more likely to sample higher emitting ares. Weyant et al. (2016) also likely ew

380

closer to the ares than we did, though at a slower speed (45 m/s) than at which we typically

381

sampled (65 m/s). 21 We did not observe a clear correlation between sampling distance and

382

combustion eciency in our data, but it certainly aects variables such as plume entrainment,

383

other emissions sources, turbulence, and environmental factors.

384

The largest source of discrepancy in results is likely that relatively few ares have been

385

sampled: 26 (85 passes) by Weyant et al. (2016), 10 by Caulton et al. (2014), 20 and 37 (52

386

passes) in our study, and thus there is large representation error. In our study we attempt

387

a statistical sampling for greater representativeness, but given that there were over 5500

388

wells with reported aring in the Bakken in 2014, 3 37 independent ares only represents

389

0.6% of active ares. Thus, we think our results should be considered in concert with the

390

Weyant and Caulton analyses, and our data should be considered in aggregate. In doing so,

391

it would subtly change our total estimated contribution (lower for methane and higher for

392

black carbon), but the observed lognormal distribution result would not change.

393

Global Implications

394

Our sampling provides sucient statistics to observe a heavy-tail distribution of combus-

395

tion eciencies. This heavy-tail characteristic has been observed and reported for methane

396

emissions from the oil and gas sector, 2527,29,42 but this represents a rst observation of the

397

heavy-tail for aring emissions of methane and ethane. This has important implications

398

for current and future contributions from aring activities. To illustrate, let us consider if

20



399

our observed distribution were globally representative. Globally, 143 ± 13.6 BCM of gas

400

is ared annually. 43 If 98% destruction removal eciency were representative of every are,

401

that would correspond to a range in methane emissions of 1.14-1.90 Tg CH 4 /yr for a gas com-

402

position range of 60%-100% CH 4 . Applying our observed distribution, that range changes

403

to 2.78-4.64 Tg CH4 /yr, more than doubling the amount emitted. In assessing the climate

404

and air quality impacts of aring, it is critical that skewed distributions are accounted for in

405

the cases of methane, ethane, and black carbon. Although our specic observed emissions

406

factors and eciencies are likely only representative of the Bakken eld, the observations of

407

a skewed distribution is likely general.

21


Page 22 of 29

Page 23 of 29

408


Acknowledgement

409

Data from the aircraft campaign reported in the manuscript are archived and available

410

at http://www.esrl.noaa.gov/csd/groups/csd7/measurements/2014topdown/. This project

411

was supported by the NOAA AC4 program under grant NA14OAR0110139 and NASA grant

412

NNX14AI87G. J. Peischl and T. Ryerson were supported in part by the NOAA Climate Pro-

413

gram Oce and in part by the NOAA Atmospheric Chemistry, Carbon Cycle, and Climate

414

Program. We thank NOAA Aircraft Operations Center sta and ight crew for their eorts

415

in helping collect these data.

416

References

417

(1) Elvidge, C. D.; Ziskin, D.; Baugh, K. E.; Tuttle, B. T.; Ghosh, T.; Pack, D. W.;

418

Erwin, E. H.; Zhizhin, M. A Fifteen Year Record of Global Natural Gas Flaring Derived

419

from Satellite Data.

420 421

422 423

424

Energies 2009, 2, 595.

(2) U.S. Energy Information Administration, Natural Gas Gross Withdrawals and Production. https://www.eia.gov/dnav/ng/ng_prod_sum_a_EPG0_VGV_mmcf_a.htm . (3) North Dakota State Government, North Dakota Drilling and Production Statistics.

https://www.dmr.nd.gov/oilgas/stats/statisticsvw.asp . (4) IPCC, In

Climate Change 2013: The Physical Science Basis. Contribution of Working

425

Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate

426

Change ; Stocker, T., Qin, D., Plattner, G.-K., Tignor, M., Allen, S., Boschung, J.,

427

Nauels, A., Xia, Y., Bex, V., Midgley, P., Eds.; Cambridge University Press: Cambridge,

428

United Kingdom and New York, NY, USA, 2013; Chapter SPM, p 1â30.

429 430

(5) Shindell, D. T.; Faluvegi, G.; Koch, D. M.; Schmidt, G. A.; Unger, N.; Bauer, S. E. Improved Attribution of Climate Forcing to Emissions.

22


Science 2009, 326, 716718.


431 432

(6) Ismail, O.; Umukoro, G. Global Impact of Gas Flaring.

Page 24 of 29

Energy Power Eng. 2012, 4,

290 302.

433

(7) Simpson, I. J.; Andersen, M. P. S.; Meinardi, S.; Bruhwiler, L.; Blake, N. J.; Helmig, D.;

434

Rowland, F. S.; Blake, D. R. Long-term decline of global atmospheric ethane concen-

435

trations and implications for methane.

Nature 2012, 488, 4904.

436

(8) Bond, T. C.; Doherty, S. J.; Fahey, D. W.; Forster, P. M.; Berntsen, T.; DeAn-

437

gelo, B. J.; Flanner, M. G.; Ghan, S.; Kärcher, B.; Koch, D.; Kinne, S.; Kondo, Y.;

438

Quinn, P. K.; Sarom, M. C.; Schultz, M. G.; Schulz, M.; Venkataraman, C.; Zhang, H.;

439

Zhang, S.; Bellouin, N.; Guttikunda, S. K.; Hopke, P. K.; Jacobson, M. Z.; Kaiser, J. W.;

440

Klimont, Z.; Lohmann, U.; Schwarz, J. P.; Shindell, D.; Storelvmo, T.; Warren, S. G.;

441

Zender, C. S. Bounding the role of black carbon in the climate system: A scientic

442

assessment.

J. Geophys. Res., [Atmos] 2013, 118, 53805552.

443

(9) Stohl, A.; Klimont, Z.; Eckhardt, S.; Kupiainen, K.; Shevchenko, V. P.; Kopeikin, V. M.;

444

Novigatsky, A. N. Black carbon in the Arctic: the underestimated role of gas aring

445

and residential combustion emissions.

Atmos. Chem. Phys. 2013, 13, 8833.

446

(10) Schwarz, J. P.; Holloway, J. S.; Katich, J. M.; McKeen, S.; Kort, E. A.; Smith, M. L.;

447

Ryerson, T. B.; Sweeney, C.; Peischl, J. Black Carbon Emissions from the Bakken Oil

448

and Gas Development Region.

Environ. Sci. Technol. Lett. 2015, 2, 281285.

449

(11) Anenberg, S. C.; Schwarz, J.; Shindell, D.; Amann, M.; Faluvegi, G.; Klimont, Z.;

450

Janssens-Maenhout, G.; Pozzoli, L.; Van Dingenen, R.; Vignati, E.; Emberson, L.;

451

Muller, N. Z.; West, J. J.; Williams, M.; Demkine, V.; Hicks, W. K.; Kuylenstierna, J.;

452

Raes, F.; Ramanathan, V. Global air quality and health co-benets of mitigating near-

453

term climate change through methane and black carbon emission controls.

454

Health Perspect. 2012, 120, 831839.

23


Environ.

Page 25 of 29

455 456


(12) The World Bank, Zero Routine Flaring by 2030. http://www.worldbank.org/en/

programs/zero-routine-flaring-by-2030 .

457

(13) U.S. EPA Oce of Air Quality Planning and Standards (OAQPS), Parameters for

458

Properly Designed and Operated Flares. https://www3.epa.gov/airtoxics/flare/

459

2012flaretechreport.pdf .

460

(14) Emission Standards Division, U.S. Environmental Protection Agency, Oce of

461

Air Radiation, OAQPS, Basis and Purpose Document on Specications For

462

Hydrogen-Fueled Flares. 1998;

463

implementation/air/rules/Flare/Resource_5.pdf .

464 465

466 467

468 469

470 471

http://www.tceq.state.tx.us/assets/public/

(15) Johnson, M.; Kostiuk, L. Eciencies of low-momentum jet diusion ames in crosswinds.

Combust. Flame 2000, 123, 189 200.

(16) Johnson, M.; Kostiuk, L. A parametric model for the eciency of a are in crosswind.

Proc. Combust. Inst. 2002, 29, 1943 1950. (17) Leahey, D. M.; Preston, K.; ; Strosher, M. Theoretical and observational assessments of are eciencies. (Technical Paper).

J. Air Waste Manage. Assoc. 2001, 51, 1610.

(18) Haus, R.; Wilkinson, R.; Heland, J.; Schafer, K. Remote sensing of gas emissions on natural gas ares.

Pure Appl. Opt. 1998, 7, 853.

472

(19) Knighton, W. B.; Herndon, S. C.; Franklin, J. F.; Wood, E. C.; Wormhoudt, J.;

473

Brooks, W.; Fortner, E. C.; Allen, D. T. Direct measurement of volatile organic com-

474

pound emissions from industrial ares using real-time online techniques: Proton Trans-

475

fer Reaction Mass Spectrometry and Tunable Infrared Laser Dierential Absorption

476

Spectroscopy.

477

Ind. Eng. Chem. Res. 2012, 51, 1267412684.

(20) Caulton, D. R.; Shepson, P. B.; Cambaliza, M. O.; McCabe, D.; Baum, E.; Stirm, B. H.

24



478

Methane Destruction Eciency of Natural Gas Flares Associated with Shale Formation

479

Wells.

Environ. Sci. Technol. 2014, 48, 95489554.

480

(21) Weyant, C. L.; Shepson, P. B.; Subramanian, R.; Cambaliza, M. O. L.; Heimburger, A.;

481

McCabe, D.; Baum, E.; Stirm, B. H.; Bond, T. C. Black Carbon Emissions from As-

482

sociated Natural Gas Flaring.

483

26764563.

484 485

486

Environ. Sci. Technol. 2016, 50, 20752081, PMID:

(22) McEwen, J. D. N.; Johnson, M. R. Black carbon particulate matter emission factors for buoyancy-driven associated gas ares.

J. Air Waste Manage. Assoc. 2012, 62, 307321.

(23) Kahforoushan, D.; Fatehifar, E.; Soltan, J. The Estimation of CO2 Emission Factors for

487

Combustion Sources in Oil and Gas Processing Plants.

488

Util. Environ. E. 2010, 33, 202210.

Energy Sources Part A: Recov.

489

(24) Brandt, A. R.; Heath, G. A.; Kort, E. A.; O'Sullivan, F.; Pétron, G.; Jordaan, S. M.;

490

Tans, P.; Wilcox, J.; Gopstein, A. M.; Arent, D.; Wofsy, S.; Brown, N. J.; Bradley, R.;

491

Stucky, G. D.; Eardley, D.; Harriss, R. Methane Leaks from North American Natural

492

Gas Systems.

Science 2014, 343, 733735.

493

(25) Yuan, B.; Kaser, L.; Karl, T.; Graus, M.; Peischl, J.; Campos, T. L.; Shertz, S.;

494

Apel, E. C.; Hornbrook, R. S.; Hills, A.; Gilman, J. B.; Lerner, B. M.; Warneke, C.;

495

Flocke, F. M.; Ryerson, T. B.; Guenther, A. B.; de Gouw, J. A. Airborne ux mea-

496

surements of methane and volatile organic compounds over the Haynesville and Mar-

497

cellus shale gas production regions.

498

2015JD023242.

J. Geophys. Res., [Atmos.] 2015, 120, 62716289,

499

(26) Mitchell, A. L.; Tkacik, D. S.; Roscioli, J. R.; Herndon, S. C.; Yacovitch, T. I.;

500

Martinez, D. M.; Vaughn, T. L.; Williams, L. L.; Sullivan, M. R.; Floerchinger, C.;

501

Omara, M.; Subramanian, R.; Zimmerle, D.; Marchese, A. J.; Robinson, A. L. Mea-

502

surements of Methane Emissions from Natural Gas Gathering Facilities and Processing 25


Page 26 of 29

Page 27 of 29


503

Plants: Measurement Results.

504

25668106.

505 506

507


(27) Allen, D. T. Methane emissions from natural gas production and use: reconciling bottom-up and top-down measurements.

Curr. Opin. Chem. Eng. 2014, 5, 78 83.

(28) Brandt, A. R.; Heath, G. A.; Cooley, D. Methane Leaks from Natural Gas Systems

508

Follow Extreme Distributions.

509

27740745.


510

(29) Frankenberg, C.; Thorpe, A. K.; Thompson, D. R.; Hulley, G.; Kort, E. A.; Vance, N.;

511

Borchardt, J.; Krings, T.; Gerilowski, K.; Sweeney, C.; Conley, S.; Bue, B. D.;

512

Aubrey, A. D.; Hook, S.; Green, R. O. Airborne methane remote measurements re-

513

veal heavy-tail ux distribution in Four Corners region.

514

2016, 113, 97349739.

515

Proc. Natl. Acad. Sci. USA

(30) Allen, D. T.; Smith, D.; Torres, V. M.; Saldaña, F. C. Carbon dioxide, methane and

Curr.

516

black carbon emissions from upstream oil and gas aring in the United States.

517

Opin. Chem. Eng. 2016, 13, 119 123, SI:13 Energy and Environmental Engineering

518

/ Reaction engineering and catalysis 2016.

519

(31) Peischl, J.; Karion, A.; Sweeney, C.; Kort, E. A.; Smith, M. L.; Brandt, A. R.;

520

Yeskoo, T.; Aikin, K. C.; Conley, S. A.; Gvakharia, A.; Trainer, M.; Wolter, S.; Ryer-

521

son, T. B. Quantifying atmospheric methane emissions from oil and natural gas pro-

522

duction in the Bakken shale region of North Dakota.

523

121, 61016111, 2015JD024631.

J. Geophys. Res., [Atmos.] 2016,

524

(32) Kort, E. A.; Smith, M. L.; Murray, L. T.; Gvakharia, A.; Brandt, A. R.; Peischl, J.;

525

Ryerson, T. B.; Sweeney, C.; Travis, K. Fugitive emissions from the Bakken shale

526

illustrate role of shale production in global ethane shift.

527

46174623, 2016GL068703. 26


Geophys. Res. Lett. 2016, 43,


Page 28 of 29

528

(33) Karion, A.; Sweeney, C.; Kort, E. A.; Shepson, P. B.; Brewer, A.; Cambaliza, M.; Con-

529

ley, S. A.; Davis, K.; Deng, A.; Hardesty, M.; Herndon, S. C.; Lauvaux, T.; Lavoie, T.;

530

Lyon, D.; Newberger, T.; Pétron, G.; Rella, C.; Smith, M.; Wolter, S.; Yacovitch, T. I.;

531

Tans, P. Aircraft-Based Estimate of Total Methane Emissions from the Barnett Shale

532

Region.

533

Environ. Sci. Technol. 2015, 49, 81248131, PMID: 26148550.

(34) Karion, A.; Sweeney, C.; Wolter, S.; Newberger, T.; Chen, H.; Andrews, A.; Koer, J.;

534

Ne, D.; Tans, P. Long-term greenhouse gas measurements from aircraft.

535

Techniques 2013, 6, 511.

Atmos. Meas.

536

(35) Yacovitch, T. I.; Herndon, S. C.; Roscioli, J. R.; Floerchinger, C.; McGovern, R. M.;

537

Agnese, M.; Pétron, G.; Koer, J.; Sweeney, C.; Karion, A.; Conley, S. A.; Kort, E. A.;

538

Nähle, L.; Fischer, M.; Hildebrandt, L.; Koeth, J.; McManus, J. B.; Nelson, D. D.;

539

Zahniser, M. S.; Kolb, C. E. Demonstration of an Ethane Spectrometer for Methane

540

Source Identication.

Environ. Sci. Technol. 2014, 48, 80288034, PMID: 24945706.

541

(36) Smith, M. L.; Kort, E. A.; Karion, A.; Sweeney, C.; Herndon, S. C.; Yacovitch, T. I.

542

Airborne ethane observations in the Barnett Shale: Quantication of ethane ux and

543

attribution of methane emissions.

Environ. Sci. Technol. 2015, 49, 81588166.

544

(37) Schwarz, J. P.; Spackman, J. R.; Gao, R. S.; Perring, A. E.; Cross, E.; Onasch, T. B.;

545

Ahern, A.; Wrobel, W.; Davidovits, P.; Olfert, J.; Dubey, M. K.; Mazzoleni, C.; Fa-

546

hey, D. W. The Detection Eciency of the Single Particle Soot Photometer.

547

Technol. 2010, 44, 612628.

548

(38) Conley, S. A.; Faloona, I. C.; Lenschow, D. H.; Karion, A.; Sweeney, C. A Low-Cost

549

System for Measuring Horizontal Winds from Single-Engine Aircraft.

550

Technol. 2014, 31, 13121320.

551

Aeros. Sci.

(39) Zannetti, P.

Air Pollution Modeling , 1st ed.; Spring US, 1990.

27


J. Atmos. Oceanic

Page 29 of 29


552

(40) Brandt, A. R.; Yeskoo, T.; McNally, M. S.; Va, K.; Yeh, S.; Cai, H.; Wang, M. Q.

553

Energy Intensity and Greenhouse Gas Emissions from Tight Oil Production in the

554

Bakken Formation.

555 556

(41) Beychok, M. R.

Energy Fuels 2016, 30, 96139621.

Fundamentals Of Stack Gas Dispersion , 4th ed.; Milton R. Beychok:

10 Marquette, Apt. 121, 2005; pp 1193.

557

(42) Subramanian, R.; Williams, L. L.; Vaughn, T. L.; Zimmerle, D.; Roscioli, J. R.; Hern-

558

don, S. C.; Yacovitch, T. I.; Floerchinger, C.; Tkacik, D. S.; Mitchell, A. L.; Sulli-

559

van, M. R.; Dallmann, T. R.; Robinson, A. L. Methane Emissions from Natural Gas

560

Compressor Stations in the Transmission and Storage Sector: Measurements and Com-

561

parisons with the EPA Greenhouse Gas Reporting Program Protocol.

562

Technol. 2015, 49, 32523261, PMID: 25668051.

Environ. Sci.

563

(43) Elvidge, C. D.; Zhizhin, M.; Baugh, K.; Hsu, F.-C.; Ghosh, T. Methods for Global

564

Survey of Natural Gas Flaring from Visible Infrared Imaging Radiometer Suite Data.

565

Energies 2016, 9, 14.

28


Methane, Black Carbon, and Ethane Emissions from Natural Gas

Recommend Documents