Contributions of condensable particulate matter to ... - ACS Publications

May 30, 2018 - 1). In both emission surveys, only concentrations of organic carbon ..... combustion and marine shipping sources were relatively high (...
0 downloads 0 Views 2MB Size
Subscriber access provided by NEW MEXICO STATE UNIV

Environmental Modeling

Contributions of condensable particulate matter to atmospheric organic aerosol over Japan Yu Morino, Satoru Chatani, Kiyoshi Tanabe, Yuji Fujitani, Tazuko Morikawa, Katsuyuki Takahashi, Kei Sato, and Seiji Sugata Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b01285 • Publication Date (Web): 05 Jul 2018 Downloaded from http://pubs.acs.org on July 6, 2018

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 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 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.

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 33

Environmental Science & Technology

1

Contributions of condensable particulate matter to atmospheric organic aerosol

2

over Japan

3 4

Yu Morino,*,† Satoru Chatani,† Kiyoshi Tanabe,† Yuji Fujitani,† Tazuko Morikawa,‡

5

Katsuyuki Takahashi,|| Kei Sato,† and Seiji Sugata†

6 7



8

305-8506, Japan

9



10

||

11

210-0828, Japan

National Institute for Environmental Studies, 16-2, Onogawa, Tsukuba, Ibaraki,

Japan Automobile Research Institute, 2530 Karima, Tsukuba, Ibaraki 305-0822 Japan

Japan Environmental Sanitation Center, 10-6 Yotsuyakami-Cho, Kawasaki, Kanagawa,

12 13

To be submitted to Environmental Science and Technology

14

Last revised on May 30, 2018.

15

ACS Paragon Plus Environment

Environmental Science & Technology

16

ABSTRACT:

17

Because emission rates of particulate matter (PM) from stationary combustion

18

sources have been measured without dilution or cooling in Japan, condensable PM has

19

not been included in Japanese emission inventories. In this study, we modified an

20

emission inventory to include condensable PM from stationary combustion sources

21

based on the recent emission surveys using a dilution method. As a result, emission

22

rates of organic aerosol (OA) increased by a factor of seven over Japan. Stationary

23

combustion sources in the industrial and energy sectors became the largest contributors

24

to OA emissions over Japan in the revised estimates (filterable-plus-condensable PM),

25

while road transport and biomass burning were the dominant OA sources in the previous

26

estimate (filterable PM). These results indicate that condensable PM from large

27

combustion sources makes critical contributions to total PM2.5 emissions. Simulated

28

contributions of condensable PM from combustion sources to atmospheric OA

29

drastically increased around urban and industrial areas, including the Kanto region,

30

where OA concentrations increased by factors of 2.5–6.1. Consideration of condensable

31

PM from stationary combustion sources improved model estimates of OA in winter but

32

caused overestimation of OA concentrations in summer. Contributions of primary and

33

secondary OA should be further evaluated by comparing with organic tracer

34

measurements.

35

ACS Paragon Plus Environment

Page 2 of 33

Page 3 of 33

Environmental Science & Technology

36

TOC Art

37 38

ACS Paragon Plus Environment

Environmental Science & Technology

39

INTRODUCTION

40

Regulation of air pollutants has been successfully implemented in developed

41

countries over the last several decades, and emissions from combustion sources, such as

42

power plants, industrial factories, and motor vehicles, have decreased over the last two

43

decades1, 2. The relative contribution of non-combustion sources to air pollutants is

44

therefore thought to have increased during this period3. However, contributions of

45

combustion sources to air pollutants are still significant4, 5.

46

To evaluate the contributions of combustion sources to the ambient PM2.5 (particulate

47

matter with diameter less than 2.5 µm), understanding of PM2.5 produced by both

48

primary-emitted and secondary-produced PM2.5 is needed. Primary-emitted PM2.5 is

49

composed of compounds that range from non-volatile to semi-volatile6-9. Conditions,

50

such as temperature and dilution ratios, under which PM2.5 is sampled from combustion

51

sources should therefore be carefully selected to provide an accurate estimation of the

52

contributions of combustion sources to PM2.5. However, only filterable particulate

53

matter (PM), which consists of particles directly emitted under stack conditions, has

54

been measured from stationary combustion sources in many countries. Condensable PM,

55

which is in the gas phase under stack conditions and condenses into PM immediately

56

after discharge from the stack, should also be considered.

57

Condensable PM from emission sources has been measured by the cooling method or

58

dilution method. In the United States, the cooling method with dry impingers (EPA

59

method 202/201a) is employed, although this method is known to have a positive bias

60

associated with gas adsorption. Efforts to reduce the bias have been made, but this bias

61

still remains10. The dilution method has been developed and used for PM measurements

62

at stationary combustion sources for a long time11-13, and International Organization for

ACS Paragon Plus Environment

Page 4 of 33

Page 5 of 33

Environmental Science & Technology

63

Standardization (ISO) has promulgated ISO 25597 for use of the dilution method to

64

measure condensable PM.

65

Some recent emission inventories in the United States14 and Europe15 have included

66

both filterable and condensable PM. However, filterable and condensable PM have not

67

been completely separated even in the National Emission Inventory (NEI) of the United

68

States14. Moreover, condensable PM is not measured in the emission surveys of many

69

countries, including Japan; and condensable PM is therefore not included in most

70

emission inventories4, 16, 17.

71

Previous studies have indicated that inorganic compounds (mostly sulfate) have been

72

the dominant contributors to condensable PM18-20, although recent studies based on

73

emission surveys of new power plant facilities have indicated that organic compounds

74

make the largest contributions to condensable PM21. In addition, van der Gon et al.22

75

have revised the emission inventory of Europe by considering condensable PM from

76

residential wood combustion23 and have shown that the revised simulation better

77

reproduces the observed OA concentrations. It has been shown that large amounts of

78

atmospheric organic materials are semi volatile (i.e., semi volatile organic compounds,

79

SVOC), and gas–particle partitioning of organic compounds is thus sensitive to

80

sampling conditions24, 25. Emission characteristics of SVOCs from motor vehicles and

81

biomass burning have been examined extensively8, 26, 27. However, to the best of our

82

knowledge, few studies have involved measurements of the volatility distributions of

83

SVOC from stationary combustion sources.

84

For atmospheric simulations of condensable PM, traditional models28,

29

are

85

unsuitable because they assume primary organic compounds to be non-volatile. The

86

volatility basis-set (VBS) framework is certainly one of the dominant methods for

ACS Paragon Plus Environment

Environmental Science & Technology

87

calculating gas–particle partitioning of organic compounds, including their responses to

88

changes of temperature or OA concentrations (dilution ratio)30, 31. However, the VBS

89

model includes large uncertainties, for example in multi-generational reaction rates and

90

emission profiles of SVOCs32, 33. Emission characteristics of SVOC are one of the key

91

uncertainties in current PM2.5 modeling.

92

In this study, we constructed an emission inventory of condensable PM in Japan, with

93

particular focus on OA. The standard methods in Japan for measuring PM from

94

combustion sources have been promulgated in Japan Industrial Standard (JIS) Z8808 or

95

Z7152, but these methods involve measurement of only filterable PM. Emissions of

96

condensable PM are not measured in Japan, and thus, condensable PM is not included

97

in the Japanese emission inventory4, 17. Recently, emission surveys of both filterable and

98

condensable PM from stationary combustion sources have been conducted in Japan34-36.

99

These data indicate that condensable PM makes important contributions to total PM2.5

100

and OA. The implication is that large amounts of SVOC are emitted from stationary

101

combustion sources. We modified the Japanese emission inventory by including

102

condensable PM from stationary combustion sources. We also briefly assessed

103

uncertainties of condensable PM emissions associated with uncertainties in our

104

estimation methods. With these emission data, we conducted a simulation with the VBS

105

model and estimated emission data to assess the contributions of SVOC (condensable

106

PM) to ambient OA concentrations.

107

108

METHODOLOGY

109

Estimation of emissions of condensable PM.

ACS Paragon Plus Environment

Page 6 of 33

Page 7 of 33

Environmental Science & Technology

110

As already noted, only filterable PM is considered in surveys of emissions from

111

stationary combustion sources in current emission inventories of Japan17. To estimate

112

rates of emission of condensable PM, we used emission survey data currently available

113

in Japan. From 2008 to 2013, the Tokyo metropolis34, 35 conducted emission surveys to

114

measure filterable and condensable PM at 53 stationary combustion sources, including

115

industrial gas boilers, heavy oil boilers, incinerators, marine ships, biomass burning,

116

ceramic furnaces, and electric furnaces. We should note that what is defined to be

117

filterable and condensable PM depends on the measurement methodology. Based on the

118

results of the emission surveys, we considered PM2.5 in the stacks to be filterable PM

119

and PM2.5 after dilution by a factor of 20 to be filterable PM plus condensable PM. The

120

Supporting Information (SI) provides details of the measurement methodology used in

121

the emission surveys. We used measurement data from 25 representative sources for the

122

analysis (Table 1). Emission rates of filterable and condensable PM from a

123

coal-combustion source were also measured with the same methodology by the Ministry

124

of Environment (MOE), Japan36, and these data were also used for this analysis.

125

Emission rates of OA in filterable and condensable PM were estimated with the

126

following equations:

127 128

129



 FCPM =  . FPM ×  



. 

=  . FPM ×

  ∗ . 

130

  ! FCPM =  FCPM ×

131

=  FCPM ×

(1)

 ! 

 

 ! 

 

,

ACS Paragon Plus Environment

(2)

Environmental Science & Technology

132 133

where EX(FPM) is the emission rate of species X in filterable PM, EX(FCPM) is the

134

emission rate of species X in both filterable and condensable PM, CX(FPM) is the

135

concentration of species X measured without dilution (filterable PM), and CX(FCPM)* is

136

the concentration of species X measured after dilution (filterable-plus-condensable PM)

137

with a correction for the dilution ratio (i.e., CX(FCPM)* = CX(FCPM) × [dilution ratio]).

138

OMlsi is organic material that has low volatility, is semi-volatile, or has intermediate

139

volatility (C* ≤ 106 µg m–3) in both the gaseous and the particulate phases.

140

In this estimate, we assumed that condensable PM was omitted from the Japanese

141

emission inventory17 of PM2.5. This assumption is consistent with the fact that the

142

methodologies used in the emission surveys of stationary combustion sources were

143

based on JIS Z8808 or Z7152 (methodology to measure filterable PM). Thus rates of

144

PM2.5 emissions taken from the Japan Auto-Oil Program17 were equated to EPM2.5 (FPM)

145

in eq 1.

146

Data from the emission surveys conducted by the Tokyo metropolis34, 35 and the MOE

147

(2015) were used to estimate ratios of COA(FCPM)* to CPM2.5(FPM). Ratios of emission

148

rates (or emission factors) of compounds should be proportional to the ratios of the

149

concentrations of the compounds in the emission surveys. Considering that COA(FPM)

150

(i.e., OA concentrations in stacks) were sometimes very low, we chose to use

151

EPM2.5(FPM) rather than EOA(FPM) in eq 1 to reduce the uncertainty of the estimate. We

152

estimated the ratios of COA(FCPM) to CPM2.5(FPM) from eight emission sectors (Table

153

1). In both emission surveys, only concentrations of organic carbon (OC) were reported,

154

and we assumed the OA/OC ratio to be 1.237.

155

We should note that exhausts from on-road and off-road vehicles were measured after

ACS Paragon Plus Environment

Page 8 of 33

Page 9 of 33

Environmental Science & Technology

156

a dilution tunnel. However, because sampling temperatures for both on-road and

157

off-road vehicle exhausts were 47 °C38, some organic compounds that would have been

158

partitioned into the particulate phase at the reference temperature of 30 °C should have

159

evaporated during the sampling6, 39. To the best of our knowledge, the impact of the

160

sampling temperature on the estimates of total OA emissions from motor vehicles (in

161

emission inventories) has not been systematically evaluated. In this study, we simply

162

conducted sensitivity simulations to evaluate the contributions of condensable PM from

163

motor vehicles. May et al.39 conducted thermodenuder measurements of diesel vehicle

164

exhausts and found that about 30% of OA evaporated at around 47 °C. We therefore

165

conducted sensitivity simulations by enhancing OA emissions from motor vehicles by

166

30%, as detailed in a section of Results and Discussion.

167

Emissions of SVOCs were estimated with eq 2: &

 ! 

 

 ! 

(

 

) can be

, where Ci* is the saturation concentration (µg m-3) and fi,

168

calculated as 1)∑

%$169

the fraction of compounds with a saturation concentration equal to Ci*, is a metric of the

170

distribution of the volatility of the organic compounds. Information on fi was taken from

171

previous estimates24, 40.

172

The ratio

&'!∗ /

 ! 

 

depends strongly on fi and COA under stack conditions. Given  ! 

173

the fi of Grieshop et al.40, we estimated

174

concentrations of 8800, 3000, 220, and 30 µg m–3, respectively. OA concentrations

175

under stack conditions during the emission survey used for the Japanese emission

176

inventory32 were not reported, and we therefore could not accurately calculate

177

 ! 

 

 

to be 2.5, 3.0, 5.0, and 7.5 with OA

from emission survey data. In previous studies24, 40,

ACS Paragon Plus Environment

 ! 

 

ratios

Environmental Science & Technology

 ! 

Page 10 of 33

178

were set to 7.5 or 2.5. Based on these values, we chose

179

the standard simulation and sensitivity simulation, respectively. The sensitivity to these

180

assumptions is assessed in the Results and Discussion section.

 

= 7.5 and 3.0 for

181 182

Chemical transport model.

183

We simulated the distributions of gaseous and particulate species by using a

184

three-dimensional chemical transport model, the Models-3 Community Multiscale Air

185

Quality (CMAQ, v5.0.2) modeling system developed by the United States

186

Environmental Protection Agency41. The model setups were the same as those used in

187

our previous studies:5,

188

Mechanism 05 (CB05) model of Yarwood et al.43, and the aerosol module was a

189

sixth-generation aerosol module of CMAQ (AERO6) coupled with a VBS module

190

(AERO6VBS). Originally, the AERO6VBS considered five classes of organic

191

compounds: one class for non-volatile compounds and four classes for semi-volatile

192

compounds with C* = 1, 10, 100, and 1000 µg m–3. For the simulation of this study, we

193

changed the framework of the AERO6VBS model so that nine classes of organic

194

compounds with C* ranging from 10–2 to 106 µg m–3 could be taken into consideration.

195

As already noted in our previous study42, we considered aging reactions for both

196

primary and secondary SVOC. We should note that we did not consider fragmentation

197

in this simulation.

42

the chemical mechanism was based on the Carbon Bond

198

Vaporization enthalpy (∆Hvap) is one of the key uncertainty factors in the atmospheric

199

simulations of OA44, 45, particularly in the representation of the temperature dependence

200

of the volatility distributions of OA. In general, measurement of the ∆Hvap of

ACS Paragon Plus Environment

Page 11 of 33

Environmental Science & Technology

201

low-volatility compounds is difficult because their concentrations in the gas phase are

202

low. The ∆Hvap of OA has been estimated from measurements of the temperature

203

dependence of the yields of secondary organic aerosol (SOA)

204

rates after heating or introduction into a vacuum49,

205

variability of the measurement data is the fact that the ∆Hvap values used in chemical

206

transport models are highly variable: Schell et al.28 set ∆Hvap equal to 156 kJ/mol,

207

whereas Carlton et al.29 used a ∆Hvap of about 40 kJ/mol. Epstein et al.49 have recently

208

derived an equation to calculate ∆Hvap as a function of C*: ∆Hvap = – 11 log10 C* + 129.

209

In contrast, Tsimpidi et al.51 set ∆Hvap = – 6 log10 C* + 100. Because ∆Hvap cannot be

210

uniquely determined from heating experiments52, its value is highly uncertain.

211

Uncertainties associated with ∆Hvap are further discussed in the next section.

46-48

or OA evaporation

50

. An indication of the large

212

The two simulation domains of this study are shown in Fig. 1. Domain 1 covered East

213

Asia with a horizontal resolution of 60 km, and Domain 2 covered Japan with a

214

horizontal resolution of 15 km. We conducted simulations for January 1 to February 29

215

2012 (winter), April 1 to May 31 2012 (spring), and July 1 to August 31 2012 (summer).

216

The spin-up time for the calculations was 10 days.

217

We used the observed OC concentrations at 14 sites (Fig. 1) to evaluate the OA

218

simulations. Filter samples with a sampling duration of 6 or 12 hours were analyzed

219

with a thermal/optical carbon analyzer (DRI model 2001A; Atmoslytic Inc., CA, USA

220

or Sunset Laboratory, Inc., OR, USA) on the basis of the IMPROVE protocol.53 Details

221

of this measurement are described in our previous study.42

222 223

RESULTS AND DISCUSSION

ACS Paragon Plus Environment

Environmental Science & Technology

224 225

Page 12 of 33

Emission estimates.

Figure 2 summarizes the emission rates of OA in filterable PM and condensable PM.

226

Before the correction for condensable PM, the total emission rate of OA equaled about

227

20% of the emission rate of PM2.5. However, after correction for condensable PM, the

228

emission rates of OA increased by a factor of seven and were even higher than those of

229

total PM2.5 in filterable PM. These results suggest that use of emission factors and

230

speciation of filterable PM has led to serious underestimation of OA emission rates from

231

stationary combustion sources in the current Japanese emission inventory. OA is a major

232

component of PM2.5 in the revised emission inventory.

233

Major sources of OA emissions in filterable PM were a transport sector and biomass

234

burning. However, after correction for condensable PM, large stationary combustion

235

sources in the industrial and energy sectors and incinerators accounted for the majority

236

of OA emissions. OA emission rates from stationary combustion sources increased by a

237

factor of 24. The percentages of OA among filterable PM2.5 (i.e.,

238

50%, whereas EOA(FCPM) values were comparable to or even larger than EPM2.5(FPM)

239

values for most combustion sources (Table 1). Because of these large differences,

240

consideration of condensable PM increased OA emissions substantially.

241

 

) were 1–

. 

It should be noted that this estimate includes uncertainties associated with (1) the

242

representativeness of tested emission sources and (2) the thermodynamic properties of

243

OA emitted from stationary combustion sources.

244

(1) As noted above, only a limited number of emission surveys have been conducted,  

. 

estimated from the survey of the Tokyo metropolis34, 35 does

245

and thus, the

246

not necessarily represent the emission sectors estimated by the Japanese emission

ACS Paragon Plus Environment

Page 13 of 33

Environmental Science & Technology

247

inventory17. To our knowledge, we have used the best available datasets from the

248

emission surveys of condensable PM in Japan, although a comprehensive analysis of the

249

uncertainty associated with their representativeness would be difficult. Instead, we

250

compared the emission rates of elemental carbon (EC) in filterable PM and

251

filterable-plus-condensable PM (EEC(FPM) and EEC(FCPM), respectively) to roughly

252

understand the ranges of uncertainty associated with the representativeness of the

253

emission sources. EC is practically non-volatile, and thus, EEC(FPM) and EEC(FCPM)

254

should be very similar if the PM2.5 speciation datasets are similar between the Japanese

255

emission inventory of the Japan Auto-Oil Program17 and the emission survey of the

256

Tokyo metropolis34, 35. Overall, EEC(FPM) and EEC(FCPM) agreed to within 10% (Fig.

257

2). This difference could have been caused by differences in the EC and PM2.5

258

speciation, and these differences could be a rough measure of the uncertainty associated

259

with the representativeness of the tested emission sources. The differences between

260

EEC(FPM) and EEC(FCPM) were relatively large in some sectors. The PM2.5 speciation

261

for each sector could therefore be one of key uncertainties in this estimate.

262 263

Uncertainties of total EOA(FCPM) and EEC(FCPM) associated with variabilities in  

. 

and

 

. 

(Table 1) were ±15% and ±7%, respectively. However,

264

these percentages do not necessarily indicate the overall uncertainties of these emission

265

estimates. Clearly, the

266

these sources contributed 40% of the estimated total EOA(FCPM). Thus, accurate

267

estimates for gas combustion sources should be a priority goal of future studies.

268 269

 

. 

ratio from gas combustion sources was very large;

(2) Choice of ∆Hvap and volatility distribution is another source of uncertainty. As shown in eq 2, EOMlsi(FCPM) depends on fi (volatility distribution). Air masses emitted

ACS Paragon Plus Environment

Environmental Science & Technology

270

from combustion sources are diluted by several orders of magnitude from stacks to the

271

ambient atmosphere, and during the course of this dilution, variations of OA

272

concentrations depend on two factors that have opposite effects: the OA concentration is

273

expected to increase because of condensation caused by cooling and then to decrease

274

because of evaporation due to dilution. The behavior of OA during dilution is

275

determined by ∆Hvap and volatility distribution. We evaluated whether the simulations

276

with the selected ∆Hvap and volatility distributions could reproduce the observed ratios

277

of OA concentrations before and after dilution from the stacks (i.e.,

278

COA(FPM)/COA(FCPM)*)34, 35 (Fig. 3).

279

The COA(FPM)/COA(FCPM)* ratios were high (0.5–1.8) from wood- and

280

field-burning sources with low stack temperatures (