Near-Road Air Pollutant Measurements: Accounting for Inter-Site

Jul 19, 2018 - Parameterized Yields of Semivolatile Products from Isoprene Oxidation under Different NOx Levels: Impacts of Chemical Aging and Wall-Lo...
1 downloads 0 Views 572KB Size
Subscriber access provided by Kaohsiung Medical University

Energy and the Environment

Near-Road Air Pollutant Measurements: Accounting for Inter-Site Variability using Emission Factors Jonathan M Wang, Cheol-Heon Jeong, Nathan Hilker, Kerolyn K. Shairsingh, Robert Healy, Uwayemi Sofowote, Jerzy Debosz, Yushan Su, Michiyo McGaughey, Geoff Doerksen, Tony Munoz, Luc White, Dennis Herod, and Greg J. Evans Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b01914 • Publication Date (Web): 19 Jul 2018 Downloaded from http://pubs.acs.org on July 22, 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 29

Environmental Science & Technology

1

Near-Road Air Pollutant Measurements:

2

Accounting for Inter-Site Variability using

3

Emission Factors

4

Jonathan M. Wang*,†,‡, Cheol-Heon Jeong†, Nathan Hilker†, Kerolyn K. Shairsingh†, Robert M.

5

Healy‡, Uwayemi Sofowote‡, Jerzy Debosz‡, Yushan Su‡, Michiyo McGaughey⊥, Geoff

6

Doerksen⊥, Tony Munoz‡, Luc White§, Dennis Herod§, Greg J. Evans†

7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30



Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario M5S3E5 Canada ‡

Environmental Monitoring and Reporting Branch, Ontario Ministry of the Environment and Climate Change, Etobicoke, Ontario M9P3V6 Canada

§

Air Quality Research Division, Environment and Climate Change Canada, Ottawa, Ontario K1A0H3 Canada

⊥Air

Quality and Climate Change, Metro Vancouver, Burnaby, British Columbia V5H4G8 Canada

Corresponding author: Jonathan M. Wang Dept. Chemical Engineering and Applied Chemistry University of Toronto 200 College Street, Room 123, Toronto, Ontario, Canada M5S3E5 Tel. 416-978-5932 Fax. 416-978-8605 Email. [email protected]

ACS Paragon Plus Environment

Environmental Science & Technology

31

ABSTRACT

32

A daily-integrated emission factor (EF) method was applied to data from three near-road

33

monitoring sites to identify variables that impact traffic related pollutant concentrations in the

34

near-road environment. The sites were operated for twenty months in 2015-2017, with each site

35

differing in terms of design, local meteorology, and fleet compositions. Measurement distance

36

from the roadway and local meteorology were found to affect pollutant concentrations

37

irrespective of background subtraction. However, using emission factors mostly accounted for

38

the effects of dilution and dispersion, allowing inter-site differences in emissions to be resolved.

39

A multiple linear regression model that included predictor variables such as fraction of larger

40

vehicles (>7.6 m in length; i.e., heavy-duty vehicles), vehicle speed, and ambient temperature

41

accounted for inter-site variability of the fleet average NO, NOx, and particle number EFs

42

(R2:0.50–0.75), with lower model performance for CO and black carbon (BC) EFs (R2:0.28-

43

0.46). NOx and BC EFs were affected more than CO and particle number EFs, by the fraction of

44

larger vehicles, which also resulted in measurable weekday/weekend differences. Pollutant EFs

45

also varied with ambient temperature and because there was little seasonal changes in fleet

46

composition, this was attributed to changes in fuel composition and/or post-tailpipe

47

transformation of pollutants.

48

ACS Paragon Plus Environment

Page 2 of 29

Page 3 of 29

49

Environmental Science & Technology

ABSTRACT ART

50 51 52

KEYWORDS

53

Real-world, vehicle, emission factor, near road, air quality monitoring, urban, heavy-duty

54

vehicles, ultrafine particles, black carbon

55

ACS Paragon Plus Environment

Environmental Science & Technology

56

INTRODUCTION

57

Traffic is one of the main contributors to urban air pollution where both traffic and

58

population density tend to be higher. Exposure to traffic emissions have been associated with

59

numerous negative health effects1-3 including increased risk of cardiovascular and respiratory

60

mortality, cancer,4, 5 adverse birth and developmental outcomes,6 respiratory diseases7 including

61

asthma8 and premature mortality.9, 10 Thus traffic represents a public health concern because a

62

large percentage of the population live near traffic sources and are potentially exposed to

63

elevated levels of air pollution. For example in 2013, it was estimated that approximately 32% of

64

the Canadian population live within 500 m of a highway and/or 100 m of a major roadway.2, 11 In

65

the United States, 19% of the population live within 500 m of a major roadway, however, in

66

more urbanized areas this percentage increases significantly for example 40% in California.12 A

67

previous modelling study showed that emitted pollutants dilute by a factor of ~1000 from

68

tailpipe-to-roadway within seconds and by a factor of 10 in the near road region in a matter of

69

minutes.13 Additionally, downwind measurements of particle number concentration showed that

70

at 27 m from a highway, levels were 11 times higher than upwind while at 280 m, concentrations

71

were double those upwind.11 Furthermore, downwind pollutant concentrations may not only

72

depend on vehicle emissions but also ambient meteorological conditions.14, 15 Thus, it is

73

increasingly important to continuously measure pollutant levels and trends in the near-road

74

environment, an area that is not typically captured by ambient air quality monitoring programs,16

75

and has only recently been introduced in the United States.17

76

Many factors affect near-road pollutant concentrations, which can complicate the

77

interpretation of the measurement data. In addition to the already dynamic emissions from the

78

vehicle fleet, the measurement distance (and height) from the roadway, obstructions, the

ACS Paragon Plus Environment

Page 4 of 29

Page 5 of 29

Environmental Science & Technology

79

presence of a street canyon, and meteorological conditions including wind direction and speed

80

can all affect the concentrations measured.18 In order to account for some of the variance arising

81

from dilution/dispersion, emission factors (EFs) can be used to isolate and normalize the local

82

traffic contribution. Past real-world studies have used various methods in determining EFs19

83

including integrated20, tunnel21, 22, remote sensing23, 24, upwind/downwind25, 26, and

84

chasing/tailpipe27, 28 measurements; however, the inter-site variability inherent across a network

85

of near-road stations, and unavoidable due to urban siting constraints, requires a more dynamic

86

EF determination method. A plume-based EF method was initially explored, because this had

87

been successfully used in an earlier study for evaluating vehicle and fleet emission

88

characteristics on the individual plume level.29 The application of the plume-based method was

89

more feasible in that case due to lower traffic density and a street canyon. However, at sites with

90

higher traffic density such as highways, it was found that this method could not be easily applied.

91

Thus, a broader, low-resolution method (daily timescale) that did not depend on the

92

identification of individual plumes was used to determine daily fleet mean EFs that could be

93

applied and compared at all sites investigated.

94

In this study, the core goal is to develop, evaluate, and apply a daily-integrated EF method

95

to 21 months of measurements from a near-road monitoring network in order to isolate vehicle

96

fleet emissions and better account for differences in site setup parameters and local meteorology.

97

The evaluation compares the daily-integrated EF method with a previously developed plume-

98

based EF method. Additionally, the calculation of a background spline for subtraction was

99

assessed through comparison with a local background site. Inter-site comparisons were made

100

between pollutant concentrations and EFs, where the local contribution was determined. The

101

value of this EF method and the resulting dataset is illustrated through the novel insights it

ACS Paragon Plus Environment

Environmental Science & Technology

102

provides into the parameters affecting weekday vs. weekend and seasonal trends of vehicle

103

emissions. These parameters were combined into a multiple linear regression model that was

104

used to explore the relative importance of the various factors that impact vehicle emissions.

105 106

METHODS

107

Measurement Sites

108

Continuous measurements were made 2015-2017 at the near-road sites described in Table

109

S1 (Supporting Information), where the near-road region is defined as being within 100 m of a

110

major urban roadway (average daily traffic ≥15,000 vehicles) or highway.2, 30 Measurements

111

were made at NR-Tor (43º39'32"N, 79º23'43"W) located in the Toronto downtown area (2015-

112

07-01 to 2017-03-28), 15 m north of a four lane arterial roadway (17 m wide; ~15,000-26,000

113

veh./day) within a street canyon where buildings on either side of the roadway range between

114

three and five stories (~12-22 m); at NR-H401 (43°42'40"N, 79º32'36"W) located in a suburban

115

Toronto area (2015-07-18 to 2017-03-28), 14 m south of a 17 lane highway (104 m wide;

116

~30,000 veh./day) within an open terrain environment; and at NR-Van (49°15'37"N,

117

123º4'40"W) located in a suburban Vancouver area (2015-07-11 to 2017-03-28), 6 m from a six

118

lane major roadway (24 m wide; ~365,000-410,000 veh./day) within a low street canyon where

119

buildings on either side are two stories high (~6 m). Additional information on the site and local

120

meteorology is included in the Supporting Information (Figure S1 & Table S1).

121

Mean (range) ambient temperature and relative humidity during this period were +11 °C (-

122

25 to +35 °C) and 60% (11 to 93%) for Toronto, and +11 °C (-8 to +33 °C) and 73% (17 to 94%)

123

for Vancouver. Instrumentation is described in Table S1 (Supporting Information) where regular

124

calibrations were applied for the gas-phase instrumentation every 3-4 months at all sites. No

ACS Paragon Plus Environment

Page 6 of 29

Page 7 of 29

Environmental Science & Technology

125

substantial seasonal differences in the relationship between the AE33880nm BC (Black Carbon

126

measured at 880 nm) and Sunset Elemental Carbon (EC).31 However, it was found that the

127

AE33880nm BC was consistently a factor of ~1.5 higher than the Sunset EC. Thus, for better

128

comparability to past studies, this factor was used to convert the AE33880nm BC measurements to

129

EC mass concentration. Meteorological conditions (wind direction and speed, relative humidity,

130

ambient temperature and pressure) were measured using a Vaisala WXT520 weather sensor at

131

each site set on a tower 10 m above ground level (a.g.l.), with the exception of NR-Tor, where it

132

was at roof level (~22 m a.g.l.) and for temperature/relative humidity at NR-Van (8 m a.g.l.). The

133

Wavetronix SmartSensor 125 HD, a dual radar traffic detection system described in more detail

134

in the Supporting Information, was utilized to count traffic based on detected vehicle lengths in

135

three bins (C2: 1 to 7.6 m; C3: 7.6 – 15 m; C4: >15 m) and to measure the vehicle speed at the

136

near-road sites averaged across all detectable lanes. Due to the limitation of the SmartSensor and

137

width of the highway, only traffic counting measurements for the eight closest lanes out of

138

seventeen at NR-H401 were useable. However, independent video analysis indicated that fleet

139

characteristics were similar for the eight lanes used here and for all lanes.

140 141 142

Daily-Integrated Emission Factors Measurement data were processed and analyzed using Igor Pro 6.34 and R. All pollutant

143

data were averaged to hourly resolution. A background calculation method was tested by

144

comparison with Bkg-Tor (South) site (43°36'44"N, 79º23'19"W) background measurements and

145

is described in further detail in the Supporting Information. In short, a linear spline interpolation

146

of minima determined in eight hour time windows of hourly averaged measurements was used to

147

calculate the background (Figures S1 & S2) necessary for background subtraction. The local

ACS Paragon Plus Environment

Environmental Science & Technology

Page 8 of 29

148

traffic contributions for each day, midnight-to-midnight, was then integrated. Daily-integrated

149

fuel-based EFs were calculated using a carbon-balance method described in Eq.1 as follows:

150

EF୔ = ቀ

∆ሾ୔ሿ

∆ሾେ୓మ ሿ

ቁ ‫ݓ‬େ

(1)

151

where EFP is the emission factor of pollutant P (in mass or particle number) per kg of fuel burned

152

assuming ambient conditions (i.e. 25 °C, 101.325 kPa) represented by the background-subtracted

153

integrated amount of carbon combustion products ∆CO2 (in kg of C m-3), and the carbon mass

154

fraction of a mixed-fuel fleet, wc = 0.86 kg C (kg-fuel-1).32-34

155

The daily-integrated EF method described above uses a similar approach as the plume-

156

based EF method described in Wang et al. (2015), with an important distinction that each capture

157

extends over an entire day, as opposed to a single vehicle exhaust plume,29 and is assumed (after

158

background subtraction) to represent predominantly the combined local traffic contribution. It is

159

not expected that the two different techniques would necessarily result in the same EF values for

160

two important reasons: 1) the daily-integrated EF method uses data of a much lower time

161

resolution and 2) the integration is applied over the entire day, which may not only affect the

162

baseline calculation between the two methods, but also potentially be affected by secondary

163

formation and non-traffic related sources depending on the pollutant. The “capture scale” of the

164

daily-integrated EF method may include a larger temporal period and hence geospatial area

165

depending on meteorological conditions. However, they should agree if EFs averaged over these

166

larger scales are similar to those in the single plumes. A comparison between the two methods is

167

described in EF method evaluation section.

168

ACS Paragon Plus Environment

Page 9 of 29

169 170

Environmental Science & Technology

Multiple Linear Regression Model Development Factors influencing fleet emission factors of different pollutants was investigated by

171

combining the EFs determined at all three sites (NR-Tor: ~28%, NR-Van: ~28%, NR-H401:

172

~44% of the dataset) for each respective pollutant and performing a multiple linear regression

173

(MLR) in R. The predictor variables examined include the percentage of C3 (%VehiclesC3) and

174

C4 (%VehiclesC4) vehicles, average vehicle fleet speed, ambient temperature, wind speed and

175

direction, and relative humidity. Firstly, predictor variables were ranked based on their

176

Spearman’s correlation coefficient with each pollutant EF. The highest-ranking variable was then

177

added in a supervised stepwise linear regression, where only variables that increased the adjusted

178

R2 by ≥1%, and reduced the residual error, were kept in the model. These criteria were adapted

179

from a previous method.35 In the case of collinearity, the variable that contributed to a higher

180

model R2 was kept. A final model was produced for each EF pollutant described in the MLR

181

section. Model performance was further tested using cross-validation where a training set

182

containing 70% of the data was randomly selected to create a new model. The remaining 30%

183

made up a test set to evaluate the model made and was compared to the actual EFs. This was

184

repeated 10 times with different randomizations.

185 186

RESULTS AND DISCUSSION

187

Evaluation of Emission Factor Methods

188

Overall, mean values between the daily-integrated and plume-based EF methods29

189

compared well. In Figure 1, mean PN and BC EFs compared the best, lying well within the

190

daily-integrated EF distribution. Matching NO and NOx EFs is complicated by the divergence in

191

the temporal scales of each method, leading to a bias towards fresher local emissions resulting in

ACS Paragon Plus Environment

Environmental Science & Technology

192

higher NO EFs using the plume-based EF method, and the opposite effect for the daily-

193

integrated EF method resulting in higher NO2 EFs (not shown, NOx – NO), whereas NOx EFs

194

were relatively close in value between the two methods (Figure 1). Mean CO EFs agreed in

195

magnitude, however, the plume-based mean EF was only 63% of the value of the daily-

196

integrated EF. The daily-integrated approach inherently assumes that the diurnal patterns are

197

only due to vehicle emissions. This assumption is less valid for pollutants such as CO where

198

traffic is not necessarily the only or even necessarily the dominant local source (i.e., cooking

199

sources, non-vehicle incomplete combustion).

200 201 202 203 204 205

Figure 1: Comparison of mean EF values between mean plume-based EFs and daily-integrated EFs (n = 85 – 112, depending on the pollutant) over coinciding measurement time periods in 2014-2015 from NR-Tor. Box-and-whiskers represent the distribution. EF units for CO, NO, and NOx are in g kg-fuel-1, BC in mg kg-fuel-1, and PN in 1014 kg-fuel-1.

ACS Paragon Plus Environment

Page 10 of 29

Page 11 of 29

206

Environmental Science & Technology

The two methods were further tested by correlating the day-to-day variation in the daily-

207

integrated and plume-based EFs using the same dataset, so as to refine the data selection criteria,

208

as described in more detail in the Supporting Information. It was found that the selection criteria

209

that resulted in the best comparison between the two methods for all pollutants was a minimum

210

daily integrated amount of CO2 of 300 ppm-day. Not only is CO2 used in the calculation of EFs,

211

it acts as an indicator of the purity of the local traffic contribution because a very low CO2

212

integrated amount indicates that local contributions were not measurably different from the

213

regional air that was diluting it (i.e., the delta was not significantly different from zero). The

214

resulting R2 and p-values for each pollutant are shown in Figure S5 (Supporting Information),

215

with PN (R2: 0.61, p-value:

308

1). Relative to the other meteorological conditions and fleet factors, %VehiclesC4 had the highest

ACS Paragon Plus Environment

Environmental Science & Technology

309

WD/WE differences with a ratio of 2.2 at NR-H401 corresponding to higher weekday BC, NO,

310

NOx, and PN emissions expected from HDDVs (Figure 4a). At NR-H401, the largest mean

311

WD/WE EF ratios were for NO and BC, which were 2.4 and 2.0, respectively, while WD/WE

312

differences in PN and NOx EFs were also substantial with ratios of 1.8 and 1.6, respectively. At

313

NR-Van, the WD/WE ratios were both ~1.8 for BC and NO, while NOx and PN EFs were 1.4

314

and 1.6, respectively. Although WD/WE %VehiclesC4 ratio was only 1.7 at NR-Van, lower than

315

at NR-H401, WD/WE EF ratios with the exception of CO were much closer to the WD/WE

316

%VehiclesC4 ratio. Interestingly, PN and BC WD/WE EF ratios at NR-Van were close to those

317

observed at NR-H401. This implies that the weekend change in %VehiclesC4 at NR-Van may

318

have a larger relative impact than at NR-H401 or another factor is responsible for the higher

319

weekday EFs. A similar trend was observed when using the percentage of larger vehicles (%LV:

320

%VehiclesC3 + %VehiclesC4) and is discussed in the Supporting Information (Figure S7). At NR-

321

Tor, WD/WE EF ratios for BC and NO were similar to the ratio for %VehiclesC4, all being

322

around ~1.4 (Figure 4a), where there was a significant trend between %VehiclesC4 and NOx, PN,

323

and BC EFs (p-value>0.05).

324

ACS Paragon Plus Environment

Page 16 of 29

Page 17 of 29

325 326 327 328 329 330 331 332 333

Environmental Science & Technology

Figure 4: Box-and-whisker plots and mean values (diamonds) for pollutant EFs and percentage of C4 Vehicles (%VehiclesC4 = C4 Vehicles/Total Vehicles x 100%) (a), and EF distributions for each pollutant binned by %VehiclesC4 at each site (b-e: bins of 0.18 for NR-Tor, 1.5 for NR-Van, 2.0 for NR-HWY401; based on the range of the parameter) where left-panel box and whiskers, and right-panel light and dark shaded areas represent 75th and 90th percentiles, respectively. %VehicleC4 ranges were different for each site. CO was the only pollutant to exhibit WD/WE EF ratios below one at NR-H401, and was

334

approximately 1 for NR-Van and NR-Tor. CO emissions from the fleet between weekday and

335

weekend, assumed to be mostly from gasoline fueled passenger vehicles at these sites, is not

336

expected to change substantially. However, with fewer larger vehicles on the weekend, the

337

contribution of CO relative to CO2 would be higher resulting in higher average CO EFs on the

338

weekends. Interestingly, the PN EF WD/WE ratio at NR-Tor was also closer to 1 than the other

339

sites, implying that particle emitters in and around the NR-Tor site were similar regardless of the

340

day of week, which was not the case at NR-H401 and NR-Van. The influence of larger vehicles

341

on some pollutants is further demonstrated in Figure 4b-e, where pollutant EFs were binned by

342

%VehiclesC4. For all three sites, but more obviously at NR-H401 and NR-Van, NOx and BC EFs

343

increased as the %VehiclesC4 increased (Figure 4c,e), whereas CO EFs had the opposite or small

344

trends (Figure 4b). The comparison for PN EFs was more complicated, where both NR-Tor and

ACS Paragon Plus Environment

Environmental Science & Technology

345

NR-Van had smaller differences relative to %VehiclesC4 when compared to NR-H401, where

346

only the 75th and 90th percentiles of the data showed increases in PN EFs with higher

347

%VehiclesC4 (Figure 4d).

348 349 350

Seasonal Trends in Emission Factors Seasonal EF trends were similar for each pollutant at all three sites and were found to

351

coincide with changes in ambient temperature. When EFs are binned by temperature as shown in

352

Figure 5c-f, EFs had a clear relationship with ambient temperature. CO, NOx, and PN EFs

353

increased as ambient temperature decreased whereas BC EFs had the opposite trend increasing

354

with warmer temperatures. Because fleet composition was the major factor influencing WD/WE

355

differences, seasonal differences in the fleet were initially investigated to see whether differences

356

in the fleet influenced seasonal EF trends. %VehiclesC4 showed no significant correlation with

357

ambient temperature, except for NR-Tor (p-value: ~7x10-3). Additionally, the seasonal

358

differences in %VehiclesC4 were small relative to those for most pollutant EFs with

359

winter/summer ratios ranging from 0.9 at NR-H401 to 1.2 at NR-Van. The seasonal change in

360

PN EFs has been discussed in more detail in Wang et al. (2017) for the NR-Tor site.57 In short,

361

increased condensation and/or nucleation during colder temperatures, and increased evaporation

362

on warmer days combine to create strong seasonal effects on PN EFs post-tailpipe. This is

363

further supported by the small seasonal change (±10%) in %LV, whereas mean PN EFs were

364

~30-80% higher in the winter depending on the site. All sites showed relatively comparable PN

365

EFs above 0 °C, however, below 0 °C, PN EFs began to increase at the Toronto sites to much

366

higher values at -10 °C (Figure 5e). This increase in PN EFs at low temperatures was not

367

observed at NR-Van because few data were available below 0 °C. Considering that both sites are

ACS Paragon Plus Environment

Page 18 of 29

Page 19 of 29

Environmental Science & Technology

368

characterized by similar fleet composition, the effect of warmer temperatures may have been one

369

of the contributing factors towards lower PN EFs at NR-Van.

370 371 372 373 374 375 376 377

Figure 5: Box-and-whisker plots and mean values (diamonds) for ambient temperature (a) and %VehiclesC4 (b) (left), and pollutant EFs distributions in 5 °C temperature bins (c-f; right) where left-panel box and whiskers, and right-panel light and dark shaded areas represent 75th and 90th percentiles, respectively. EF trends in the extreme temperature bins at each site were limited by the number of data points (n