Black Carbon Emissions from Associated Natural Gas Flaring

Jan 14, 2016 - School of Aviation & Transportation Technology, Purdue University, West Lafayette, Indiana 47907, United States. Environ. Sci. ... Appr...
1 downloads 13 Views 402KB Size
Subscriber access provided by UNIV OF CALIFORNIA SAN DIEGO LIBRARIES

Article

Black carbon emissions from associated natural gas flaring Cheryl L. Weyant, Paul B. Shepson, R Subramanian, Maria O. L. Cambaliza, Alexie Heimburger, David McCabe, Ellen Baum, Brian H. Stirm, and Tami C. Bond Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.5b04712 • Publication Date (Web): 14 Jan 2016 Downloaded from http://pubs.acs.org on January 25, 2016

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 24

Environmental Science & Technology

Black carbon emissions from associated natural gas flaring Cheryl L. Weyant,∗,† Paul B. Shepson,‡ R. Subramanian,¶ Maria O. L. Cambaliza,§ Alexie Heimburger,‡ David McCabe,k Ellen Baum,⊥ Brian H. Stirm,# and Tami C. Bond†

1

Civil and Environmental Engineering, University of Illinois Urbana-Champaign, IL, Department of Chemistry, Purdue University, West Lafayette, IN, Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, PA, Department of Physics, Ateneo de Manila University, Quezon City, Philippines, Clean Air Task Force, Boston, MA, Climate and Health Research Network, Bowdoinham, ME, and School of Aviation & Transportation Technology, Purdue University, West Lafayette, IN E-mail: [email protected] Phone: (217) 333-6967. Fax: (217) 333-0687

1

ACS Paragon Plus Environment

Environmental Science & Technology

Page 2 of 24

Abstract

2

3

Approximately 150 billion cubic meters (BCM) of natural gas is ared and vented

4

in the world annually, emitting greenhouse gases and other pollutants with no energy

5

benet. About 7 BCM per year is ared in the United States, and half is from North

6

Dakota alone. There are few emission measurements from associated gas ares and lim-

7

ited black carbon (BC) emission factors have been previously reported from the eld.

8

Emission plumes from 26 individual ares in the Bakken formation in North Dakota

9

were sampled. Methane, carbon dioxide, and BC were measured simultaneously, al-

10

lowing the calculation of BC mass emission factors using the carbon balance method.

11

Particle optical absorption was measured using a three-wavelength Particle Soot Ab-

12

sorption Photometer (PSAP) and BC particle number and mass concentrations were

13

measured with a Single Particle Soot Photometer. The BC emission factors varied over

14

two orders of magnitude, with an average and uncertainty range of 0.14 ± 0.12 g/kg hy-

15

drocarbons in associated gas and a median of 0.07 g/kg which represents a lower bound

16

on these measurements. An estimation of the BC emission factor derived from PSAP

17

absorption provides an upper bound at 3.1 g/kg. These results are lower than previous

18

estimations and laboratory measurements. The BC mass absorption cross section was

19

16 ± 12 m2 /g BC at 530 nm. The average absorption Ångström exponent was 1.2 ±

20

0.8, suggesting that most of the light absorbing aerosol measured was black carbon and

21

the contribution of light absorbing organic carbon was small.

22

Introduction

23

Petroleum deposits often have a gaseous component, commonly called associated gas, that

24

consists of methane and other short chain hydrocarbons.

25

by-product of oil production may be ared or vented if it is not collected or used on-site.

26

Associated gas extracted as a

One hundred and forty to one hundred and seventy billion standard cubic meters (BCM)

27

of associated gas is ared and vented in the world annually ( 1 ,

28

contributes about 7 BCM per year and about half of that is from North Dakota ( 3 ). The 2

ACS Paragon Plus Environment

2 ).

The United States

Page 3 of 24

Environmental Science & Technology

29

volume of natural gas ared in North Dakota increased from 0.09 BCM in 2004 to 3.7

30

BCM in 2014, a 40-fold increase in ten years ( 4 ).

31

hydraulic fracturing has led to increased oil production in the Bakken region, North Dakota.

32

However, production wells have preceded the development of gas transportation pipelines,

33

so a signicant fraction of the gas produced is ared. About 30 % of the gas produced in the

34

Bakken was ared in 2014, down from a peak of almost 50 % in 2008, although the magnitude

35

of aring has increased in that time period. In comparison, the rest of the USA ares less

36

than 1% of associated gas ( 3 ).

Horizontal drilling and high-volume

From a global warming perspective, aring is preferable to venting.

37

Flaring oxidizes

38

carbon and creates CO 2 , but reduces the overall global warming potential of the emissions

39

by destroying methane and other hydrocarbons.

40

than CO2 on a 100 year time-scale ( 5 7 ).

41

warming gases, but also creates other pollutants, such as NO x , CO, and black carbon (BC).

Fossil methane is 36 times more potent

Flaring reduces emissions of ammable global

Black carbon is a combustion by-product that contributes to climate warming.

42

BC

43

absorbs solar radiation in the atmosphere, inuences cloud dynamics, and darkens ice surfaces

44

and accelerates melting. Including all known forcing mechanisms, the current atmospheric

45

radiative forcing of BC is estimated to be +1.1 (+1.0/-0.93) Wm

46

values (8 ). BC radiative forcing is second only to CO 2 , which contributes 1.68 ( ±0.35) Wm

47

(6 ).

−2

, relative to pre-industrial

−2

The GAINS global emission inventory estimates that aring emits 4 % of global BC

48

−1

49

emissions, at 230 Gg yr

50

yr

51

associated gas ares were modeled to be the source of 42 % of BC surface concentrations in

52

the Arctic atmosphere, where BC-induced ice melting is hypothesized to occur ( 9 ).

53

−1

, a factor of three greater than on-road gasoline vehicles (80 Gg

) (9 ), (8 ). The global emissions are relatively small, yet due to emissions at high latitude,

The BC emission factor, the mass of BC emitted relative to the quantity of gas ared,

54

is an important metric that facilitates estimations of emissions from the industry.

55

are no published eld measurements of BC emission factors from individualassociated gas

3

ACS Paragon Plus Environment

There

Environmental Science & Technology

Page 4 of 24

56

ares.

Field and laboratory emission testing of ares have reported combustion eciency

57

or destruction eciency of fuel components ( 10 13 ). Ground-based optical measurements

58

have been used to estimate BC emission rates from gas ares in the eld ( 14 16 ), but

59

emission factors were not reported in these studies. Schwarz et al. (2015) ( 17 ) conducted

60

aircraft sampling in the Bakken and determined an average BC emission factor from the

61

region that includes BC from ares and other sources, but they did not sample individual

62

ares. Laboratory studies that measure turbulent diusion ames have found that fuel gases

63

and ame turbulence aect emission factors ( 18 20 ), so variability in individual emission

64

factors is likely.

65

None of these studies estimate BC emissions from ares as operated, and eld studies

66

evaluate either one or two ares or an average over an entire region. As a complement to

67

these studies that describes the emission behavior of individual ares, this study reports

68

measurements of BC emission factors from individual

69

gas ares were sampled using both laser-induced incandescence and absorption-based BC

70

measurements to provide lower and upper bounds, respectively, on emission factors. Flare

71

emissions were measured in the Bakken region, North Dakota, in March 2014.

72

associated gas ares.

Individual

Methods

73

Measurements of BC, CO 2 , and CH4 were conducted for 26 gas ares in the Bakken forma-

74

tion, North Dakota using aircraft sampling. The sampling campaign included 85 ight passes

75

through are plumes in nine ights in March 2014. Individual ares were sampled with 1-6

76

ight passes, and two ares were sampled on multiple days. Each ight was limited by the

77

fuel and was about ve hours. The Purdue University's Airborne Laboratory for Atmospheric

78

Research (ALAR), a Beechcraft Duchess aircraft, ew into the plume from downwind. This

79

small aircraft was instrumented for wind and chemical species measurements, and can y

80

relatively slowly (45 m/s) to enable adequate sampling in a small plume.

4

ACS Paragon Plus Environment

Page 5 of 24

81

Environmental Science & Technology

Flares were sighted from the aircraft and then selected for sampling. Both ground ares

82

and elevated stack ares were sampled and

83

position, instantaneous gas ow rates, and details about are designs were not known for

84

the sites measured. However, the monthly volume of gas ared in March 2014 was known

85

for some of the well sites ( 4 ) and an estimation of average gas composition in the Bakken

86

formation was available ( 21 ).

87

were distinguished by observation. Fuel com-

Aircraft and instrumentation

88

The aircraft was equipped with a Best Air Turbulence (BAT) probe for wind measurements,

89

a global positioning and inertial navigation system (GPS/INS), and a microbead thermistor

90

(22 ,

91

nose of the aircraft. It measured and recorded wind speed and wind direction at 50 Hz. The

92

measured pressure variations across the hemisphere of the probe were combined with 50 Hz

93

inertial data from GPS/INS to obtain three-dimensional wind vectors.

94

23 ).

The BAT probe is a nine-port pressure dierential probe that extended from the

Ambient air was drawn into the nose of the aircraft and through 5 cm diameter PTFE

−1

95

Teon tubing at a ow rate of 1840 L min

using a high-capacity blower located at the

96

rear of the aircraft. The sampled gases were drawn into three sampling instruments from

97

this central duct.

98

inlet probe, duct and sample line) as determined from ground measurements at sampling

99

ow rates. The particle loss was not a strong function of particle size for the particle sizes

100

measured (225-150nm). The sampling inlet was sub-isokinetic (duct velocity was about 18

101

m/s and the plane was traveling at about 45 m/s at the time of sampling). There would be

102

an enhancement of particle sampling eciency of 25 % for particles of 0.1 µm diameter.

There was a 20 % particle loss in the sampling system (including the

103

A cavity ring-down spectrometry system (CRDS, model G2301-m, Picarro Inc.) was used

104

for in-situ, real-time measurements of CO 2 , CH4 , and H2 O. The CRDS system measured gas

105

concentrations at 0.5 Hz at a ow rate of 850 ml min

106

in-ight calibration system for CH 4 and CO2 , using two NOAA/ESRL reference cylinders

−1

. The aircraft was equipped with an

5

ACS Paragon Plus Environment

Environmental Science & Technology

µmol/mol and cylinder 2:

Page 6 of 24

107

(cylinder 1: 1785.85 nmol/mol and 377.857

108

441.898

109

CH4 have routinely and consistently been found to be 0.05 % and 0.1%, respectively, and

110

comparisons to NOAA whole air samples support accuracies of 0.2 % and 0.1%, respectively,

111

corresponding to +/- 0.9 ppm for CO 2 and 3 ppb for CH 4 .

µmol/mol for CH4

2656.83 nmol/mol and

and CO2 , respectively). The measurement precision for CO 2 and

112

Two aerosol instruments were on-board the aircraft, a Single Particle Soot Photometer

113

(SP2, Droplet Measurement Technologies) ( 24 26 ) and a three wavelength Particle Soot

114

Absorption Photometer (PSAP, Radiance Research). The SP2 measures laser-induced in-

115

candescence of refractory BC-containing aerosols. The incandescence signal is proportional

116

to BC mass and this relationship was determined in laboratory calibrations using fullerene

117

soot. The SP2 was used to determine black carbon mass in individual particles, including

118

those internally mixed with other materials ( 27 ,

119

greater than 0.7 fg are detected with 100 % eciency (29 ). Although some particles were

120

below this threshold, less than 5

121

contribute more to the BC mass. In addition, coincident particle detection can cause un-

122

dersampling of 5-10 % of the BC mass. Primary SP2 data were analyzed as in Subramanian

123

et al. (2010) and then resolved at 1 Hz for further analysis ( 30 ). Leading edge scattering

124

signals have been used to predict scattering from BC particle coatings that are vaporized in

125

normal instrument use ( 31 ), but the SP2 in this study was not congured for this analysis.

126

The BC particle size ranged from 0.005 - 0.8

127

and is not signicantly dierent than the 0.188

128

Further details about the SP2 calibration, coincidence analysis, and BC coating thickness

129

estimation can be found in the Supplemental Information.

%

28 ).

Only particles with a mass of BC

of the mass was undetected because the larger particles

µm

with a mass median diameter of 0.22

µm

found in Schwarz et al.

µm

(2015) ( 17 ).

130

The PSAP continuously detects light absorption by particles at 467, 530, and 660 nm

131

on a lter, including absorption by non-BC particles. Absorption at each wavelength was

132

recorded every four seconds. Filter transmittance was maintained in the acceptable range of

133

0.5 to 1.0 throughout the entire sampling campaign. The ow through the instrument was

6

ACS Paragon Plus Environment

Page 7 of 24

Environmental Science & Technology

134

measured with a primary ow calibrator (mini-BUCK Calibrator M-5,A. P. BUCK, INC.)

135

before and after every day of sampling and was used to calibrate an internal real-time ow

136

sensor. The PSAP signal was corrected for ow rate, spot size, and adjusted to atmospheric

137

pressure (32 ) and a loading-dependent correction was applied.

138

The study design included both SP2 and PSAP measurements to provide constraints

139

on emission factors given uncertainties inherent in each technique.

140

to underestimate emission factors because it does not detect the smallest BC particles and

141

is subject to saturation in high concentrations.

142

to analyze particle-level details. The PSAP can only capture integral absorption over the

143

entire plume and is not exclusively selective for BC absorption. However, the PSAP does

144

not have the same size and saturation constraints as the SP2. Both instruments produced

145

good results; the SP2 shows little bias due to missing particles and the PSAP is not strongly

146

inuenced by non-BC absorption.

147

important characteristic of emissions: the absorption per mass, which is required in order to

148

infer the radiative eect of black carbon.

149

The SP2 is expected

On the other hand, the SP2 can be used

The paired data are thus useful to calculate another

Data analysis

150

Emission factors (in grams BC per kg associated gas) were calculated using the ratio of BC

151

mass to the total carbon mass measured in each are pass. Emissions were also reported in

152

grams BC per volume associated gas using gas composition data from the Bakken (see SI

153

Table 1). This carbon balance method is typical for emission factor calculations for combus-

154

tion emissions ( 33 35 ), especially for sources without stacks. Carbon, above background and

155

measured as CO 2 , CH4 , and BC, in the plume was assumed to be entirely from the combus-

156

tion of associated gas, so that the mass of excess carbon in the plume was equal to the mass

157

of carbon from the fuel. This calculation assumes that emissions of other carbon-containing

158

species are negligible. Carbon monoxide and other hydrocarbon emissions are assumed to

159

be zero in this analysis. The CO/CO 2 ratio in a laboratory are has been measured as less

7

ACS Paragon Plus Environment

Environmental Science & Technology

160

Page 8 of 24

than 0.002 (36 ), suggesting that CO forms a negligible fraction of carbon.

161

For each species, the carbon balance calculation requires the concentration of excess car-

162

bon in the plume, which is obtained by subtracting background from plume concentrations.

163

Background concentrations were determined by assigning the lowest 10th percentile of each

164

full ight data set as background. While this cuto value is arbitrary, use of the 70th per-

165

centile would decrease the BC emission factors by less then 5 % and the mass absorption

166

cross sections by less than 4 %.

167

Each signal peak was integrated to determine the amount of pollutant captured in a

168

single pass.

169

due to the dierent response times and sampling rates of the instruments, the peaks they

170

generated ranged from 3-10 seconds. Alignment of the signals for a point-by-point analysis,

171

as in the gas-phase measurements of Caulton et al. (2014) ( 11 ), was not feasible. For each

172

signal separately, twenty seconds of data around each peak was run through a bootstrapping

173

routine that minimized the residuals between a Gaussian t and the data, and identied

174

a Gaussian curve that best t the data ( 37 ) (Figure 1). The width of the Gaussian peak

175

was used to determine the integration period for each signal. An exception was CH 4 , which

176

was measured with the same instrument as CO 2 .

177

the well pad in addition to the are. An assumption was made that CH 4 within the CO 2

178

peak originated from the are plume itself, while CH 4 outside that peak came from leaks.

179

This assumption was implemented by integrating CH 4 over the same time period as the CO 2

180

peak.

181

The aircraft typically sampled the are plume for about two seconds, but

Methane can be emitted from leaks on

Passes for which the maximum CO 2 peak was greater than 10 times the standard devi-

σ)

182

ation of the background (10

above the average background were included in the dataset.

183

When BC mass, aerosol absorption, or CH 4 were below the detection limit (3

184

background), they were assigned a value of half the detection limit for the emission factor

185

calculation.

186

vehicle emissions, diesel engines, and small ares on storage tanks. The contributions from

σ

above

Other combustion emission sources may be present near the are, including

8

ACS Paragon Plus Environment

Page 9 of 24

Environmental Science & Technology

σ

188

cuto identies are plumes sampled that are clearly elevated above background. Only ve

189

passes were omitted due to this constraint. Two were small ares and for three, the ight

190

was too far from the source.

CO2 (ppm)

these unquantied sources may appear in both the background and in the plume. The 10

Absorption m2 /m 3

187

30 20 10

150 100 50

0 05:25

05:30

05:35

05:40

05:45

05:50

05:25

05:30

4000

2000

05:40

05:45

05:50

0.05 0.04 0.03 0.02 0.01

0 05:25

05:30

05:35

05:40

05:45

05:50

05:25

05:30

Time (MM:SS)

05:35

05:40

05:45

05:50

Time (MM:SS) Latitude (deg)

850

Height (m)

05:35

Time (MM:SS) CH4 (ppm)

BC Mass (ng/m3 )

Time (MM:SS)

800 750

48.02 48 47.98 47.96 47.94

700 04:30

05:00

05:30

06:00

06:30

-102.9

-102.85

-102.8

-102.75

-102.7

Longitude (deg)

Time (MM:SS)

Figure 1: Measurements taken during a single are pass.

The real-time data is shown in

blue for the four signals: particle absorption from the PSAP (top left), CO 2 (top right), CH 4 (middle right), and the BC mass from the SP2 (middle left). Gaussian peak ts are shown as dashed red lines. The vertical black dashed lines show the width of the integration periods. The bottom left and right plots show the height and location of the of the aircraft at the time of sampling. The red portion identies the time period shown in the real-time plots.

191

The emission factor for BC was determined using Equation 1.

EFBC

192

193

C BC , C CO2 ,

and

C CH4

= 1000F

CBC CCO2 + CCH4 + CBC

are mass concentrations of carbon in each species in g/m

(1)

3

and

F

is the carbon mass fraction of the gas in grams carbon per grams of hydrocarbons in the fuel

9

ACS Paragon Plus Environment

Environmental Science & Technology

Page 10 of 24

194

gas. The carbon content of the fuel was not measured, but has little variation (0.77 - 0.75)

195

over a wide range of fuel mixtures (see SI Table 1). A value of 0.77 was used, which is based

196

on reported gas composition in the Bakken ( 21 ).

197

amounts of inert gases such as N 2 , CO2 or He. The emission factors given here are in grams

198

BC per kg of hydrocarbons in associated gas, which connects emission measurements to a

199

quantity that is easily measured and closely connected to energy content of the ared gas.

200

Any CO2 present in the fuel gas would bias BC emission factors low. Associated gas in the

201

Bakken typically contains less than 1 % CO2 (21 )(see SI Table 1).

202

Associated gas can contain signicant

2 The mass absorption cross-section of BC (MAC BC in m /g BC) is the ratio of the ab−1

2 3 (m /m ) to

C BC

203

sorption coecient ( b) in m

204

2 per mass of BC in m /g BC. The absorption Ångström exponent (AAE) is a measure of the

205

wavelength dependence of the absorption. AAE values of about 1 are typical for graphitic

206

carbon, while values of 3.5 - 7 are found for light absorbing organic aerosols ( 8 ). Here, AAE

207

is determined between the 467 and 660 nm wavelengths using Equation 2.

AAE

208

209

210

211

=−

and indicates the strength of the absorption

ln(b660 /b467 ) ln(660/467)

(2)

Reported mass absorption cross-section and AAE do not include values in the denominator below the detection limit.

Results and discussion Black carbon absorption

212

The magnitude of BC mass peaks was strongly correlated with the magnitude of the ab-

213

sorption peaks for all passes (R

214

BC emitted from this source. The average and standard deviation of the mass absorption

215

cross-section of BC for all the passes was 22

2

= 0.89), suggesting consistency in the absorption of the

±

16, 16

±

12, and 14

10

ACS Paragon Plus Environment

±

2 10 m /g BC at 467,

Page 11 of 24

Environmental Science & Technology

216

530, and 660 nm wavelengths, respectively. A histogram of MAC BC (530 nm) showing all

217

are passes is shown in Figure 2. MAC BC at 530 nm for individual passes ranged from 3.1 -

218

2 2 41 m /g BC, and the median was 16 m /g BC at 530 nm.

219

A typical value for MAC BC for pure BC is 7.5

±

2 1.2 m /g BC at 550 nm ( 8 ,

38 ),

220

about half the median value found here. This high value of MAC BC is not fully explained.

221

However, many sources, including ares, were not included in the assessment of MAC BC in

222

2 Bond and Bergstrom (2006), and 7.5 m /g BC does not represent a conclusive value from

223

all sources. Absorption of BC can be enhanced when BC particles are coated with non-BC

224

aerosol components, such as water or organic material ( 28 ). However, the relative humidity

225

in the plume was low in these experiments (48

226

is unlikely. Some condensation of heavy hydrocarbons onto BC particles is possible. If the

227

associated gas contained 0.08 grams condensable hydrocarbons per gram methane ( 21 ), at

228

environmental temperatures, sucient heavy hydrocarbons (3 grams condensable hydrocar-

229

bons per gram BC) could condense on the BC to increase the absorption by a factor of 1.4

230

(see SI) (39 ). The SP2 did not detect the mass in very small particles, an underestimation

231

that would make the measured MAC BC higher than the true value. Additional absorption

232

by non-BC material can occur but is not consistent with these measurements as shown in

233

the next section or the coating thickness analysis in the SP2 (see SI).

±

25

%)

so the presence of condensed water

234

2 If the absorption measurements, along with a typical MAC BC of 7.5 m /g BC are used

235

to approximate the BC mass, the resulting BC mass emission factors would be 2.2 times

236

higher than measured by the SP2. This possible increase represents an upper bound on the

237

emission factors in this study.

11

ACS Paragon Plus Environment

Frequency

0 2 4 6 8 10

Flare ID Number

Environmental Science & Technology

8 9 10 11 12 13 14 15 16 17 18 22 23 24 25 27 30

0

10

20

Page 12 of 24

30

40

2

BC mass absorption cross−section (m /gBC) (530 nm)

Figure 2: Distribution of MAC BC at 530 nm. Top: histogram of all passes through a are 2 plume. The average and uncertainty range for all passes was 16 ± 12 m /g BC and are shown in the gure as the red and blue dashed, vertical lines. shows the range of MAC BC for each are.

The box and whisker plot

The centerline of each box is the median, the

box represents the 25th through 75th percentile, the whiskers represent the most extreme value within 1.5 times the interquartile range, and the points beyond the whiskers are values outside this range.

238

Absorption Ångström exponent ±

239

The average AAE measured for all the are passes was 1.2

240

greater variability in passes with low absorption (below 100 m

241

low absorption measurements, AAE ranged from -0.6 to 3.3, and for high absorption the

242

AAE was 0.9 to 2.3. However, this variability does not aect the average, which was 1.4

243

0.33 when low absorption peaks are excluded. This result suggests that most of the light

244

absorbing aerosol measured is black carbon and the contribution of organic carbon is small.

245

Using the method suggested in Lack et al. (2013), and assuming the AAE of pure BC is

12

ACS Paragon Plus Environment

0.8 (Figure 3).

−1

There was

s integrated peak).

For

±

Page 13 of 24

Environmental Science & Technology

1, an estimated eight percent of the absorption is predicted to be from non-BC absorbing

247

aerosols. However, the contribution of non-BC aerosol is not statistically dierent from zero

248

(40 ). AAE is uncorrelated with MAC BC , suggesting that organic carbon does not aect the

249

MACBC .

15 10 0

5

Frequency

20

25

246

-1

0

1

2

3

AAE (467 - 660 nm)

Figure 3: Histogram of the absorption Ångström exponent (AAE) between the 467 and 660 −1 nm. The gray bars show are passes with low levels of absorption (below 100 m s integrated absorption) and the maroon bars are passes above this threshold. The vertical lines represent the mean and uncertainty range (1.2

250

±

0.8).

Black carbon emission factor

251

Figure 4 shows the distribution of BC emission factors for all the passes, including a separa-

252

tion by individual are. Black carbon emission factors ranged over two orders of magnitude.

253

The average BC emission factor was 140 mg/kg hydrocarbons in associated gas with an

254

uncertainty range of of 120 mg/kg and a median of 68 mg/kg (median absolute deviation:

255

53 mg/kg).

256

emission factors is right-skewed.

Several observations were above 500 mg/kg and the histogram of all the BC

257

Average BC emission factors for individual ares varied from 2.3 to 330 mg/kg associ-

258

ated gas. Individual ares can be grouped into three statistically dierent categories: low,

259

medium, and high emitters, with averages of 5.0, 53, and 270 mg/kg (p 0.1).

stack ares (200

±

Page 14 of 24

±

46 mg/kg gas, n = 7)

280, n = 64), but the dierence was not statistically

264

Explanations for the large variability in the BC emission factor were sought. Environ-

265

mental factors measured were not predictive; the BC emission factor was not correlated (r