Near-road vehicle emissions air quality monitoring for exposure modeling

Journal Pre-proof Near-road vehicle emissions air quality monitoring for exposure modeling Jennifer L. Moutinho, Donghai Liang, Rachel Golan, Stefanie E. Sarnat, Rodney Weber, Jeremy A. Sarnat, Armistead G. Russell PII:

S1352-2310(20)30059-5

DOI:

https://doi.org/10.1016/j.atmosenv.2020.117318

Reference:

AEA 117318

To appear in:

Atmospheric Environment

Received Date: 25 September 2019 Revised Date:

24 January 2020

Accepted Date: 26 January 2020

Please cite this article as: Moutinho, J.L., Liang, D., Golan, R., Sarnat, S.E., Weber, R., Sarnat, J.A., Russell, A.G., Near-road vehicle emissions air quality monitoring for exposure modeling, Atmospheric Environment (2020), doi: https://doi.org/10.1016/j.atmosenv.2020.117318. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier Ltd.

Near-road Vehicle Emissions Air Quality Monitoring for Exposure Modeling

Jennifer L. Moutinho1, ¶, Donghai Liang2,# ¶, Rachel Golan3, Stefanie E. Sarnat2, Rodney Weber1, Jeremy A. Sarnat2, Armistead G. Russell1

1

School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, USA

2

Department of Environmental Health, Rollins School of Public Health, Emory University, Atlanta, USA

3

Department of Public Health, Ben-Gurion University of the Negev, Beer Sheva, Israel

# Address correspondence to: Donghai Liang, Ph.D. Department of Environmental Health, Rollins School of Public Health, Emory University, 1518 Clifton Rd NE, Atlanta, GA 30322, USA. Tel: 404-712-9583, Email: [email protected] ¶ These authors contributed equally to this work.

Credit Author Statements: Jennifer L. Moutinho: Data Collection, Data Analysis, Writing, Reviewing and Editing Donghai Liang: Investigation, Methodology, Statistical Analysis, Writing, Reviewing and Editing Rachel Golan: Data Collection, Investigation Stefanie E. Sarnat: Data Collection, Investigation, Methodology Rodney Weber: Data Collection, Investigation, Methodology, Reviewing and Editing Jeremy A. Sarnat: Conceptualization, Investigation, Methodology, Supervision, Reviewing and Editing Armistead G. Russell: Conceptualization, Investigation, Methodology, Supervision, Reviewing and Editing

1

Near-road Vehicle Emissions Air Quality Monitoring for Exposure Modeling

2 3

Jennifer L. Moutinho1, ¶, Donghai Liang2,# ¶, Rachel Golan3, Stefanie E. Sarnat2, Rodney Weber1, Jeremy

4

A. Sarnat2, Armistead G. Russell1

5 6

1

School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, USA

7

2

Department of Environmental Health, Rollins School of Public Health, Emory University, Atlanta, USA

8

3

Department of Public Health, Ben-Gurion University of the Negev, Beer Sheva, Israel

9 10

# Address correspondence to: Donghai Liang, Ph.D. Department of Environmental Health, Rollins School

11

of Public Health, Emory University, 1518 Clifton Rd NE, Atlanta, GA 30322, USA. Tel: 404-712-9583,

12

Email: [email protected]

13

¶ These authors contributed equally to this work.

14 15 16 17 18 19 20 21 22 23 24

Declaration of interests: none.

1

Abstract

2

Exposure to vehicular emissions has been linked to numerous adverse health effects. In response

3

to the arising concerns, near-road monitoring is conducted to better characterize the impact of mobile

4

source emissions on air quality and exposure in the near-road environment. An intensive measurement

5

campaign measured traffic-related air pollutants (TRAPs) and related data (e.g., meteorology, traffic,

6

regional air pollutant levels) in Atlanta, along one of the busiest highway corridors in the US. Given the

7

complexity of the near-road environment, the study aimed to compare two near-road monitors, located in

8

close proximity to each other, to assess how observed similarities and differences between measurements

9

at these two sites inform the siting of other near-road monitoring stations. TRAP measurements, including

10

carbon monoxide (CO) and nitrogen dioxide (NO2), are analyzed at two roadside monitors in Atlanta, GA

11

located within 325m of each other. Both meteorological and traffic conditions were monitored to assess

12

the temporal impact of these factors on traffic-related pollutant concentrations. The meteorological factors

13

drove the diurnal variability of primary pollutant concentration more than traffic count. In spite of their

14

proximity, while the CO and NO2 concentrations were correlated with similar diurnal variations, pollutant

15

concentrations at the two closely sited monitors differed, likely due to the differences in the siting

16

characteristics reducing the dispersion of the primary emissions out of the near-road environment.

17

Overall, the near-road TRAP concentrations at all sites were not as elevated as seen in prior studies,

18

supporting that decreased vehicle emissions have led to significant reductions in TRAP levels, even along

19

major interstates. Further, the differences in the observed levels show that use of single near-road

20

observations will not capture pollutant levels representative of the local near-road environment and that

21

additional approaches (e.g., air quality models) are needed to characterize exposures.

22 23

Key words: Traffic-related air pollutants, Near-road monitoring, Exposure assessment, Diurnal

24

profile of traffic emissions

25

1

Introduction

2

Measurement of traffic-related air pollutants (TRAPs) in near-road environments has been an

3

active area of research given the extensive evidence linking exposure to primary traffic pollution to a

4

range of adverse health effects (Brugge et al. 2007; Chen et al. 2017; Golan et al. 2018; HEI 2009a, b;

5

Liang et al. 2019; Smith et al. 2017; WHO 2005). High levels of air pollutants exist within near-road

6

microenvironments due to vehicle exhaust and mechanically-generated emissions (Baldauf et al. 2013;

7

Karner et al. 2010). Concentrations of TRAPs, including nitrogen oxides (NOx), volatile organic

8

compounds (VOCs), carbon monoxide (CO), primary fine particulate matter (PM2.5), black carbon (BC),

9

and organic carbon (OC), in particular, have been found to be elevated near heavily trafficked roads

10

(Baldauf et al. 2012; Beckerman et al. 2008; Boogaard et al. 2011). Such emissions can interact

11

chemically and physically with each other and other pre-existing pollutants in the roadway environment,

12

leading to a complex, multicomponent mixture (Saha et al. 2018).

13

Historically nitrogen dioxide (NO2) concentrations exceeded the hourly National Ambient Air

14

Quality Standard (NAAQS) in the near-road environment (Vijayaraghavan et al. 2014), though emissions

15

controls have led to reduced pollutant concentrations and exceedances of the national standards

16

(DeWinter et al. 2018). Where high NO2 measurements are still a concern, vehicle emissions are the

17

major contributing source (Lin et al. 2016). Elevated TRAP concentrations are affected by a number of

18

factors including traffic volume, vehicle types, local meteorological conditions (Karner et al. 2010), and

19

local topography such as natural (Baldauf 2017) and built highway features (Baldauf et al. 2016). These

20

conditions can result in a wide range of pollutant concentrations observed along different roadway

21

segments within the same urban environment (McAdam et al. 2011). A concern for the possibility of high

22

concentrations of NO2 initiated the U.S. Environmental Protection Agency (EPA) to implement a national

23

near-road monitoring network to specifically measure TRAPs (EPA 2010, 2011). The near-road

24

monitoring network focuses on locating monitors near the most heavily trafficked roads in urban cores

25

around the country. One objective for the network was to improve exposure assessments to primary traffic

26

emission for urban populations vulnerable to this pollution source. The implementation of the monitors

1

began January 1, 2014 and as of January 1, 2015, there were 61 active monitoring sites, two of which

2

were located within 10m of two different heavily trafficked highways in Atlanta, GA.

3

While the near-road remains a potential high exposure environment, the near-road environment

4

has changed with improved vehicle engine technologies, the associated emissions control systems, and

5

fuel regulations leading to reduced mobile source emissions affecting near-road concentrations (Ayala et

6

al. 2012; Henneman et al. 2015; Karner et al. 2010; McDonald et al. 2015; Vijayaraghavan et al. 2014). A

7

decrease in mobile emissions in Georgia over the last decade has contributed to an estimated 30%

8

reduction in PM from mobile sources (Zhai et al. 2017). New studies are needed to revisit potential

9

changes in the local spatial and temporal patterns.

10

The Dorm Room Inhalation to Vehicle Emissions (DRIVE) study was conducted in Atlanta, GA

11

in 2014 along one of the busiest highway corridors in the US to characterize factors leading to human

12

exposures in the near-road environment, both indoors and outside, and to assess statistical and integrated

13

traffic exposure metrics for applications in epidemiological studies (Liang et al. 2018a; Liang et al.

14

2018b; Sarnat et al. 2018) . Utilizing measurements from the DRIVE study, this paper aims to

15

characterize pollutant dynamics in the changing near-road environment and to examine how well the

16

near-road network monitors characterize TRAP concentrations in complex, dynamic urban environments.

17

This issue is addressed here by focusing on the dynamics of TRAP observations in relationship to

18

meteorology, traffic characteristics, and regional air pollution mixtures using two near-road monitoring

19

sites in close proximity to each other along the same highway segment located near downtown Atlanta.

20

Understanding how well a single monitor represents the concentrations across an urban area has

21

importance in both regulatory as well as health assessment frameworks. Therefore, as part of this analysis

22

we consider how the characterization of exposure and potential NAAQS exceedances might be different

23

at two near-road monitors along the same road segment.

24

1

Methods

2

Concentrations of primary tailpipe pollutants are used as tracers for the impact of on-road mobile

3

emissions to urban environments. Measurements were conducted continuously along a major highway in

4

Atlanta, GA at two near-road highway monitors located in relatively close proximity to each other (325

5

m). One of the near-road locations was part of the DRIVE study and the second was part of the national

6

EPA Near-road Monitoring Network operated by the Georgia Environmental Protection Division (GA

7

EPD) of the Department of Natural Resources since June 2014. This analysis assesses hourly

8

concentration measurements from two near-road monitoring locations along the same highway segment,

9

and examines the impact of key meteorological and traffic factors on corresponding measured pollutant

10

concentrations. Through comparing concentrations at the two sites within 325m of each other, the

11

analysis can further assess how representative the EPA near-road network monitors are for understanding

12

microscale exposure to populations within the near-road environment of an urban area.

13 14

Site descriptions

15

This study domain was centered around a segment of arterial interstate where Interstate 75 and

16

Interstate 85 (I-75/I-85) merge in the center of Atlanta, Georgia (Fig. 1). In 2014, this highway segment

17

along which the two monitoring locations are situated had an average annual daily traffic (AADT) of

18

330,000, composed primarily of light-duty gasoline passenger cars and trucks. Heavy-duty diesel trucks

19

made up approximately four percent of the average daily vehicles on this portion of the highway (GA

20

DOT 2012). Surface streets to the east of the highway follow a gridded pattern with an average block

21

length of 450 feet and an AADT of more than 15 times less than the AADT for the highway segment. The

22

land west of the highway segment for 1.5 km is the Georgia Institute of Technology (GIT) campus with

23

limited vehicle access and much lower AADT. Within the study domain area, the Southeastern Aerosol

24

Research and Characterization (SEARCH) network has maintained an urban background (UB)

25

monitoring site since 1998 (Hansen et al. 2003). The site is located 2.3km west of the highway and is a

26

long-term dataset that represents the historical Atlanta background concentration.

1

The near-road DRIVE sampling location (NR DRIVE) was located about a meter from the west

2

side of the fifteen-lane highway (eight southbound and seven northbound) to the south of 10th Street and

3

to the north of North Avenue (Fig. S1). The monitoring site was located in a parking lot with less than 85

4

passenger vehicle spots and the vertical height from the highway to the parking lot was 0.5 m. The nearby

5

EPA near-road monitoring network site (Fig. S2) was located on the Georgia Institute of Technology

6

campus (NR GIT) 325m north of the NR DRIVE site location and about a meter from the west side of the

7

highway. Trees were removed from the vegetation barrier to provide space for the site and a small, gated

8

parking lot for about 100 passenger vehicles is to the west of the site.

9 10

Air Quality Instrumentation, Meteorological Characterization, and Traffic Data

11

Air quality monitors at the NR DRIVE location collected continuous ambient air samples from

12

September 8, 2014 to January 5, 2015. All species concentrations, meteorological parameters, and traffic

13

data were measured at local standard time (LST); however, daylight savings time did end on November 2,

14

2014. Continuous measurements of black carbon (Magee Aethalometer AE31), carbon monoxide

15

(Thermo Model 48i), ozone (Thermo Model 49C), and nitrogen oxides (Teledyne API 200A) provided

16

concentration data for pollutants commonly associated with vehicle emissions. Real-time gas analyzers

17

collected measurements at 5-second averaging periods and the real-time black carbon monitor collected at

18

2-minute intervals. Data was collected using DAQFactory and WinWedge Pro software. Multipoint

19

calibrations, zero air, and span checks provided an assessment for accuracy throughout the study and were

20

used in time-weighted adjustments to the data. The sampling inlet height was approximately 3m and was

21

7m from the closest highway lane. All continuous data were averaged to hourly levels to assess temporal

22

variability differences between pollutants and possible indicators. Details of the instrumentation and

23

quality assurance can be found elsewhere (Sarnat et al. 2018). Continuous CO and NOx data from the NR

24

GIT site began on July 1, 2014. The sampling inlet height was approximately 3m and was 6m from the

25

closest lane. The hourly concentration data were downloaded from the EPA air quality system (AQS).

1

Traffic vehicle count and speed data were obtained from the Georgia Department of

2

Transportation Office of Transportation Data (GDOT 2014). The vehicle count data were collected at a

3

location on I-75/I-85 1.5 miles south of the measurement location using Automatic Traffic Records. No

4

major on or off ramps are located between the traffic count location and the near-road monitoring

5

locations. Meteorological data collected at the NR DRIVE site (HOBO U30, Onset Corp) included wind

6

speed, wind direction, temperature, and relative humidity. Wind speed and direction measurements used a

7

cup anemometer and wind vane sensor. The sensors were mounted to a pole 5m above the ground with

8

the temperature/relative humidity sensor placed in a protective solar radiation shield. The wind rose

9

(Figure 1) from the NR DRIVE location found winds from the east (between 45 to 135 degrees) 77% of

10

the time during the study period. With the NR DRIVE and NR GIT sampling locations west of the

11

highway, winds from the east lead to downwind concentrations measurements. Atmospheric mixing

12

height can be a critical factor for pollutant dispersion. For this reason, mixing height data was modeled

13

using the Weather Research and Forecast (WRF) model (NCAR) for the 4km grid that included the study

14

domain.

15 16 17 18

Multivariate regression modeling We used multivariate linear mixed regression modelling to assess the factors that affected the temporal variability in the concentration of each TRAP:

19 20

=

+

+

(Eq. 2)

21 22

where

23

coefficient of interest that describes the influence of factor

24

assessed include time period of the day (categorical), temperature (continuous), wind speed (continuous),

25

relative humidity (continuous), wind direction (categorical), weekend (categorical, Saturday and Sunday),

26

and hourly traffic counts (continuous). The temporal factor was divided into four periods: morning rush

denotes the concentration of BC, CO, NO, NO2, NOx, or O3 measured during hour and

is the

on the hourly pollutant level. The factors

1

hour (6 – 9am), mid-day (10am – 3pm), evening rush hour (4 – 8pm, used as reference group), late

2

evening (9 – 24pm), and early morning (1 – 5am). The wind direction factor was divided into three

3

directions: north (315 – 45 degrees), east (45 – 135 degrees, which leads to the monitoring sites being

4

downwind of the highway), and south (135 – 225 degrees).

5

used to capture potential variations not explained by

6

The regression relationship between pollutant concentrations and driving factors developed a simplified

7

method compared to the use of chemical transport models or dispersion models. A positive regression

8

coefficient indicates an association between the pollutant concentration level at the measurement site and

9

the factor, while controlling for all other factors included in the model. The multivariate regressions

10

provide a direct relationship for health studies to better understand the driving factors for near-road

11

exposures.

and

represents time-specific random intercepts represents residual random normal error.

12 13

Results and Discussion

14

Observed near-road air pollutant concentrations

15

The NR DRIVE site and the NR GIT site measured CO, NO, NO2, NOx, and BC concentration

16

continuously from September 8, 2014 to January 5, 2015. The NR GIT site began sampling BC from

17

November 3, 2014. The NR DRIVE site also measured ozone (O3), wind speed, and temperature. CO and

18

NO2 mean (standard deviation) concentrations at the NR DRIVE site were 425 ppb (210 ppb) and 29 ppb

19

(15.5 ppb), respectively. At the NR GIT site, average (standard deviation) concentrations measured for

20

CO and NO2 were 624 ppb (338 ppb) and 19.5 ppb (8.6 ppb), respectively (Table 1 and Table 2).

21

During the sampling period, CO and NO2 hour maximums at both sites remained below the

22

hourly national standards of 35 ppm for CO and 100 ppb for NO2 despite the prominent wind direction

23

being from the east. The maximum hourly concentration at the NR DRIVE and NR GIT sites were 1.9

24

ppm (CO) and 93.8 ppb (NO2), and 2.2 ppm (CO) and 51.6 ppb (NO2), respectively (Table 1). Due to

25

high concentrations that skewed the distribution causing a non-normal distribution, sites were compared

1

using a Spearman’s rank correlation. The NR DRIVE site on average measured lower CO and NOx as

2

well as higher NO2 than those measured at the NR GIT location (Table 1). Temporal variability in CO

3

and NO2 hourly concentrations between the two sites lead to a Spearman’s correlation of 0.18 (CO) and

4

0.72 (NO2). Both sites captured the morning and evening increase in TRAP concentrations in the same

5

hour; however the average diurnal profiles show lower CO and NO2 concentrations measured at the NR

6

DRIVE site and the higher NO2 concentrations measured at the NR GIT site (Fig. 2). Less dispersion at

7

the NR GIT site due to vegetation drove different mixing rates with ozone, leading to differences between

8

the NO and NO2 concentrations at the NR DRIVE and NR GIT sites. This is also partly shown by the

9

closer NOx concentrations with the mean (standard deviation) concentration of 50 ppb (35 ppb) and 57

10

ppb (34 ppb) at the NR DRIVE and NR GIT sites, respectively, and the higher NOx Spearman’s

11

correlation between the two sites of 0.72. The lower NO2 concentrations at the NR GIT site suggests less

12

ozone titration occurs where reduced dispersion occurs.

13

At the NR DRIVE and NR GIT sites, the mean (standard deviation) BC concentrations were 1.6

14

(1.3) ug m-3 and 1.7 (1.2) ug m-3. While the BC measurement at the NR GIT site was only operational

15

during November and December, similar monthly averaged concentrations were observed at both sites

16

during these months. Diurnally, the BC concentration at both sites followed a similar trend with a

17

morning peak from 9am to 11am and an afternoon minimum at 5pm (Fig. 2). The BC diurnal trend

18

mimicked the NO trend with a bimodal distribution observing a maximum concentration in the morning

19

1.5 times greater than the evening peak. These pollutant distributions differed from the more balanced

20

bimodal distributions of the CO and NO2 diurnal profiles, which observed similar peak concentrations in

21

the morning and evening.

22

To understand the impact vehicle emissions have on local concentrations, the near-road

23

measurements were also compared to the urban background and rural measurements around Atlanta, GA.

24

Measurements for the urban background concentrations were collected at the highly instrumented, long-

25

term, Jefferson St. site, part of the Southeastern Aerosol Research and Characterization (SEARCH)

26

network (Blanchard et al. 2013a; Edgerton et al. 2005, 2006; Hansen et al. 2003; Hansen et al. 2006; Liu

1

et al. 2005; Solomon et al. 2003) located 2.3km west of the near-road sites. Average (maximum) hourly

2

concentrations for urban background CO and NO2 were 266 ppb (1732 ppb) and 12.6 ppb (94.4 ppb).

3

Also part of the SEARCH network, the Yorkville site located about 40 miles northwest of Atlanta

4

provides rural background pollutant concentrations. The average (maximum) hourly concentrations for

5

rural CO and NO2 were 175 ppb (524 ppb) and 2.2 ppb (26.9 ppb) respectively. Based on the difference in

6

the means for the measurements from September 8, 2014 to January 5, 2015, the regional background

7

contributed about 28% to the NR GIT site CO measurements, and the urban emissions contributed 15% to

8

the NR GIT site CO measurements. The highway vehicle emissions were a significant source of the CO

9

concentration measured in the near-road environment contributing the remaining 57% to the measured

10

CO concentration at the NR GIT site. In contrast, the regional background was a smaller percentage of the

11

NO2 concentration measured in the near-road environment (11%) and the urban background contributed

12

to about 53% of the NR GIT NO2 measurements. City scale regulations for NO2 would help overall

13

exposure since NO2 highway emissions contribute only 35% to the near-road measurements.

14

Single pollutant concentrations measured in the near-road environment are commonly used as key

15

indicators of the impact vehicle emissions have on local air quality within the near-road

16

microenvironment. In order to assess the impacts from vehicle emissions in the near-road environment,

17

CO and NO2 were commonly measured to represent primary vehicle emissions (HEI 2010). The near-

18

road measurements for both key TRAPs were elevated above the Atlanta urban and rural background

19

concentrations suggesting traffic emissions contribute to elevated concentrations in the near-road

20

environment. The NR DRIVE site concentrations for CO and NO2 were 35% and 57% higher than the

21

urban background concentrations. This is consistent with other EPA Near-road Monitoring Network sites

22

in 2014, which reported mean NO2 concentrations ranging from 9 to 24 ppb (DeWinter et al. 2018; EPA

23

2016). While traffic emissions contribute to the elevated levels of primary pollutants in the near-road

24

environment, other urban and region sources contribute a significant fraction affecting their use as tracers

25

for vehicle emissions. Overall, the pollutant levels measured at the DRIVE study location, an open site

26

with good dispersion characteristics, showed a relatively small roadside increment and low impact of the

1

16-lane interstate highway compared to historic near-road field data, especially those measured from

2

more enclosed sites (i.e. street canyon locations). Nevertheless, the pollutant levels we measured during

3

this extensive monitoring period were consistent with measurements from the US EPA’s near-road

4

monitoring network pollutant trends analysis and emissions estimates.

5 6

Assessment of traffic volume and meteorological factors impacting roadside concentrations

7

Traffic count and meteorological conditions are key factors driving near-road TRAP

8

concentrations. Average weekday traffic data during the sampling period on the interstate demonstrated

9

three peaks with the morning and evening rush hour events as well as a mid-day peak at about 2pm

10

(Figure 3a). Instead of a common bimodal traffic count distribution (Baldauf et al. 2012; Batterman et al.

11

2015), this segment of highway had a consistently high traffic volume with vehicle counts rising quickly

12

from 5am to 7am and more slowly until reaching a maximum at 3pm. Vehicle count slowly decreased

13

from 3pm to 7pm and quickly dropped reaching a minimum at 3am. The daily trend was consistent across

14

all four months of the study with variability in the travel behavior based on weekday (Figure 3b, Figure

15

S3) and weekend (Figure 3c, Figure S3). Traffic volume on the weekend displayed no peak in the

16

morning or evening, further highlighting the significance of both work commuting trips and the use of the

17

highway for other daily trips.

18

The mean normalized diurnal profiles of the key TRAP species, vehicle counts, and

19

meteorological conditions driving dispersion showed the extent of daily variation for each pollutant and

20

factor for the NR DRIVE site and the NR GIT site (Figure 4). The normalized CO, NO2, and BC

21

concentrations had similar diurnal profiles with a morning concentration peak at 10am, an evening peak at

22

10pm, and minimum concentrations observed at 3am and 4pm. The normalized hourly O3 concentration

23

measured at the NR DRIVE site had a maximum concentration at 6pm and a minimum at 10am. Since the

24

minimum primary pollutant concentrations occurred during the hours with maximum vehicle counts, this

25

showed that highway traffic count alone is a poor indicator of diurnal pollutant levels observed at the

26

DRIVE study location.

1

The average vehicle speed and corresponding congestion patterns (Figure 4c) showed highway

2

traffic speeds remaining high throughout the night and reaching a minimum at 6pm during the evening

3

rush hour period. As traffic counts rose at 6am, corresponding reductions in mean traffic speeds were

4

observed from about 70 to 40mph by 8am. While traffic counts remained high throughout the day, mean

5

traffic speeds increased from 45mph at 10am before dropping to 15mph at 5pm. Since vehicle emissions

6

rates remain fairly constant above 20mph (Barth et al.) and traffic counts remain elevated throughout the

7

day, diurnal emissions trends would suggest the highest vehicle emissions occur between 2pm and 8pm

8

when measured species concentrations are lowest.

9

Mixing height data was generated by the WRF model for the 4km grid including the two central

10

near-road locations, and varied diurnally. The mixing height remained low (about 260m) until about 7am,

11

increased during the day until reaching a maximum height at about 3pm (typically about 1000m), and

12

decreased until approximately 10pm. Mixing height as well as ozone formation driven by photochemical

13

activity resulted in a peak for both between 1pm and 6pm. As the mixing height increased in the morning,

14

TRAP concentrations decreased reaching a minimum concentration at 3pm while the traffic count was

15

reaching a maximum. By multiplying the concentration by the mixing height (Figure 4d and 4e), the

16

increase during mid-day when the mixing height was greatest showed the emissions increased with traffic

17

count, but mixing height drove daily dynamics leading to minimum concentrations mid-day.

18

The two near-road sites observed trends in primary TRAPs explained by rapid pollutant

19

dispersion, particularly associated with the increase in convective mixing and increased wind speeds.

20

Even at near-road sites, the diurnal convective mixing and wind speed had the dominant impact of TRAP

21

concentrations (Fig. 4). The high impact of meteorological factors compared to traffic count suggests a

22

change in fate and transport properties affecting near-road concentration variability. This change

23

influences the applicability of traffic count as a mobile source tracer in quantifying exposure to traffic

24

emissions. Our results were consistent with recent findings from other near road study (Hilker et al. 2019;

25

Sofowote et al. 2018; Wang et al. 2018). Further, an implication of the changing near-road environment is

26

that future exposure studies aimed at characterizing health impacts of mobile emissions will need to

1

consider different approaches for determining the mobile source contribution to ambient concentrations of

2

single pollutants.

3

The pollutant diurnal patterns at the two near-road network sites were consistent with prior

4

studies showing elevated concentrations of primary traffic pollutants occurring during morning rush hours

5

when the atmospheric mixing is weak and emissions are high, then concentrations decreasing at the

6

boundary layer increases (Grosjean 1983; HEI 2010; Menut et al. 2012). During the evening rush hours,

7

the concentrations again increased and remained high throughout the night in spite of the greatly reduced

8

emissions due to the low boundary layer. Both sites observed the typical bi-modal diurnal profile for the

9

primary traffic-related air pollutants, but did not observe the same diurnal profile for vehicle count, which

10

is often used as a proxy of exposure to traffic pollutants in health effect studies (Batterman et al. 2015).

11

Traffic counts on I-75/I-85 through metro Atlanta exhibited distinctive patterns of rising sharply in the

12

early morning consistent with the beginning of the morning rush hour and reached a consistent peak

13

vehicle count of approximately 20,000 vehicles per hour from 10am to 4pm. The differences in the

14

diurnal profile patterns between the vehicle counts and primary traffic pollutant concentrations highlights

15

the predominant role meteorology and its influence on vertical dispersion have on the impact of traffic

16

hotspots on adjacent areas.

17 18

Assessment of combined traffic volume and meteorological factors leading to roadside concentrations

19

As part of the DRIVE study, exposures were modeled by integrating hourly concentrations.

20

Further, the NAAQS for both NO2 and CO have one-hour components. Thus, there is a need to link local

21

and regional factors to hourly concentrations. This can be done using more complex dispersion modeling

22

approaches or statistical methods. Statistical modeling can help identify if different factors influence the

23

observations at the three near-road locations differently. Here, linear mixed modeling evaluates

24

associations between pollutant concentrations and multiple possible contributing predictors to assess

25

factors that drive the temporal variability observed at the different near-road sites. The regression

26

coefficients for the models developed for the NR DRIVE site and the NR GIT site were compared to

1

assess whether site differences along the same road segment can lead to significant differences in

2

pollutant dynamics (Table 3). Significance of a factor was determined by a p-value less than 0.05.

3

During the 2014 study period, the NR DRIVE site and the NR GIT site regression coefficients for

4

BC and NO concentrations were positively associated with late evening to morning rush hour period

5

(9pm to 9am) when the traffic count and mixing height were low and beginning to increase (Table 3).

6

Also for both sites, wind direction and increasing wind speed were significantly associated with decreases

7

in all the primary pollutants (BC, CO, NO, NO2, and NOx), indicative of dispersion away from the

8

pollutant source (Table 3 and Figure S4). Additionally, weekend days showed an association with a

9

significant decreasing concentration for all pollutants (NO, NO2, and BC) except CO. While many of the

10

factors were significant for both sites, temperature was negatively significant for NO, NO2, and NOx only

11

at the NR GIT site. Temperature was also significantly negative at the NR GIT site for BC and CO, while

12

positively significant for the NR DRIVE site. The NR DRIVE site is located in an open parking lot

13

indicative of increased photochemical reactions compared to the NR GIT site, which was located within a

14

tree barrier along the highway. The regression models also highlight the diminishing predictive power of

15

traffic count on near-road pollutant levels. While the coefficient for traffic count was significant at both

16

near-road sites and for all pollutants, the magnitude of the coefficient was low and therefore was not a key

17

factor driving temporal variability of pollutant concentration at the study domain. Nevertheless, while

18

traffic counts alone has been found not to be an important predictor of TRAP concentrations, a recent near

19

road study completed in Canada has identified heavy-duty diesel traffic volume as an important predictor

20

(Sofowote et al. 2018; Wang et al. 2018).

21 22

Changing Near-road Environment

23

Given the relatively short duration that we analyzed for the current study, caution should be taken

24

regarding the generalizability of the findings for making long-term inferences regarding how the near-

25

road environment has changed over the last decade. A state operated, routine monitoring site (SDK),

26

while not associated with the EPA Near-Road Monitoring Network, provides a long-term concentration

1

data set located within 650m of the I-285 bypass around metro Atlanta with an AADT of about 140,000.

2

The hourly average (maximum) concentrations from September 8, 2014 to January 5, 2015 for CO and

3

NO2 were 323 ppb (1259 ppb) and 9.7 ppb (57.5 ppb), respectively. From 2000 to 2010, the observed CO

4

and NO2 concentration at the SDK site decreased 33% (496 to 345 ppb) and 42% (18.1 to 13.7 ppb),

5

respectively. For the same period from 2000 to 2010, the CO concentration decrease was 50% (561 to 282

6

ppb) and 9% (180 to 163 ppb) at the urban (JST) and rural (YRK) background sites. Similarly, the NO2

7

concentration decrease was 30% (21.9 to 15.3 ppb) and 60% (5.4 to 2.2 ppb), respectively. Decreased

8

mobile emissions have contributed to a decrease in near-road concentrations for primary traffic-related air

9

pollutants; however, a decrease in overall concentrations locally and regionally have also contributed to

10

the decrease in the near-road environment. Further, the difference in the rate of decrease from sources

11

contributes to the changing near-road environment and how to characterize near-road exposure.

12

As observed here and elsewhere, near-road TRAP concentrations are less elevated above

13

background levels than prior near-road studies around the country (Beckerman et al. 2008; Sarnat et al.

14

2008), suggesting that future studies will need to consider different approaches for characterizing the

15

spatial gradients and exposures to mobile sources. Initial results from the EPA Near-road Monitoring

16

Network support these results. Of the 61 near-road sites active in 2015, only five hourly concentrations

17

exceeded the hourly NO2 standard of 100 ppb and at all the sites the 98th percentile of the daily 1-hour

18

maximum was below the standard (DeWinter et al. 2018). The NO2 concentration average across the

19

country ranged from about 9 ppb to 30 ppb.

20

Nationally, estimates of on-road mobile source emissions of NOx have decreased about 50% since

21

2004 and emissions of CO show that 2014 levels are about 49% of those in 2004 and 25% of those in

22

1994 (Dallmann and Harley 2010). In the metro Atlanta area, long term analysis data from the urban

23

background site part of the SEARCH network shows that CO, NOx, and BC levels have decreased by 350

24

ppb (7.2% per year), 35 ppb (7.3% per year), and 1.25 µg m-3(6.8% per year), respectively, from 1999 to

25

2011. Source apportionment analysis at this site indicated mobile source-related PM2.5 decreased by about

26

half during the same period (Blanchard et al. 2013b). These declining trends are expected to continue in

1

the future as fleet turnover to newer vehicles continues, new mobile source emissions controls are

2

introduced and additional policy interventions are implemented.

3 4

Conclusion

5

Elevated traffic-related air pollutants are linked to adverse health effects and often

6

epidemiological health analysis rely on only a few monitoring locations to quantify exposure of

7

individuals in the near-road environment. Our comparison of two near-road monitors along the same

8

highway segment showed site-to-site differences influencing pollutant concentrations. Site characteristics

9

can contribute to localized concentrations and this affect was observed when comparing the two near-road

10

sites in this study. The two sites were 325m apart and both within a meter of a heavily trafficked highway.

11

While the two sites observed similar diurnal and temporal variability, site differences including increased

12

vegetation around one location reduced dispersion in comparison to the location by an open, asphalt

13

parking lot. Reduced dispersion lead to higher CO and NO concentrations and lower NO2 concentrations

14

over the study period. These site differences contributed to different pollutant dynamics along the same

15

highway segment highlighting the importance of site location. The near-road pollutant concentrations

16

measured show a reduced impact from the highway sources with levels less elevated above background

17

concentrations than in prior studies. The regression models also highlighted the diminishing predictive

18

power of traffic count on near-road pollutant levels. These decreased concentrations indicate the

19

effectiveness of mobile source emission controls leading to a decreased relative contribution from

20

vehicles to urban air pollution. This finding indicates a changing near-road environment that will affect

21

future approaches to characterizing vehicle emission hotspots and their impacts on exposures.

22 23

Acknowledgments

24

Support for this project were provided through a contract with the Health Effects Institute (RFA

25

#4942-RFA13-1/14-3). The field study conducted as part of this study benefitted greatly from the

1

assistance of many students, staff, and faculty at both Georgia Tech and Emory. Specific thanks go to C.

2

Cornwell, K. Parada, S. Shim, Dr. K. Johnson and E. Yang for their tremendous help in conducting the

3

field study. We want to thank Dr. R. Weber, Dr. V. Verma, and Dr. D. Gao for their measurements of

4

oxidative potential of ambient fine particles via DTT assay. We are indebted to Dr. J. Schauer (U.

5

Wisconsin) for loaning us several instruments to supplement our sampling network. We would also like

6

to thank Dr. Seung-Hyun Cho from RTI, Inc. for her collaboration on this project. The Georgia EPD

7

allowed us access to their roadside monitoring site and helped provide data from those monitors, and we

8

particularly thank Ken Buckley for his assistance with this.

9

assembled for field studies conducted as part of the Southeastern Center for Air Pollution and

10

Epidemiology (SCAPE), which was funded by a US Environmental Protection Agency STAR grant

11

R834799. This publication was also supported by the HERCULES Center P30ES019776. The

12

information in this document may not necessarily reflect the views of the Agency and no official

13

endorsement should be inferred. Dr. R Golan gratefully acknowledges support by a post-doctoral

14

fellowship from the Environment and Health Fund, Jerusalem, Israel. We acknowledge NSF for providing

15

a fellowship to Dr. J Moutinho, and Dr. A Russell made use of funds provided by a generous gift from

16

Howard T. Tellepson. We owe a debt of gratitude to the numerous administrators at Georgia Tech for

17

allowing us to conduct this study on campus and in their residence hall facilities.

The study used the instrumentation

18

This material is based upon work supported by the National Science Foundation Graduate

19

Research Fellowship Program under Grant No. (NSF DGE-1650044). Any opinions, findings, and

20

conclusions or recommendations expressed in this material are those of the author(s) and do not

21

necessarily reflect the views of the National Science Foundation.

1

Figures

2

a b

3 4 5 6 7

Figure 1 (a) Map of sampling area. NR GIT site part of the EPA Near-road Monitoring Network in metro Atlanta, UB site part of the SEARCH Network. (b) Wind rose of hourly observations at the NR DRIVE site from September 8, 2014 to January 5, 2015

1 2 3 4 5

Figure 2 Diurnal profiles for CO, NO2, NO, NOx, BC, and O3 for the NR GIT and NR DRIVE sites from September 8, 2014 to January 5, 2015.

1 2 3 4

Figure 3 Hourly average traffic count of (a) total, (b) weekday, and (c) weekend variability from September 1, 2014 to December 31, 2014

1

2 3 4 5 6 7 8

Figure 4 Hourly mean (a) NR DRIVE concentration data, (b) NR GIT concentration data, and (c) traffic parameter data normalized by mean. (d) NR DRIVE concentration normalized by mean and multiplied by mixing height. (e) NR GIT concentration normalized by mean and multiplied by mixing height. Data from September 8, 2014 to January 5, 2015. TCNT: Traffic count, MH: Mixing height, SPD: Traffic highway speed.

1

Tables

2

Table 1. Hourly averages of NR DRIVE and NR GIT near-road continuous instrumentation. September 8, 2014 to January 5, 2015.

Total 9/8-1/5

September 9/1-9/30

October 10/1-10/31

November 11/1-11/30

December 12/1 12/31

3 4 5 6

N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean

BC (ug m-3) DRIVE GIT 2282 1115 1.6 1.7 1.3 1.2 0.7 - 2.2 0.86 - 2.2 0.06 - 13 0.13 - 9.4 469 1.9 1.2 0.9 - 2.5 0.2 - 8.8 529 1.7 1.4 0.7 - 2.3 0.2 - 12 552 251 1.4 1.4 1.3 1.1 0.6 - 1.9 0.7 - 1.8 0.2 - 13 0.2 - 5.8 612 744 1.6 1.9

CO (ppb) DRIVE GIT 2178 2816 425 624 210 338 278 - 515 400 - 800 133 - 1860 0 - 2200 526 543 414 689 157 264 291 - 502 500 - 900 136 - 1306 0 - 1500 672 713 404 584 185 242 275 - 498 400 - 700 133 - 1304 0 - 1700 662 707 430 665 227 356 270 - 524 400 - 800 138 - 1594 100 - 2100 318 733 477 677

NO (ppb) DRIVE GIT 2666 2798 21 38 24 29 5.6 - 29 17 - 51 0 - 201 1.0 - 233 419 540 18 34 15 20 6.9 - 28 18 - 45 0 - 95 1.3 - 118 663 720 20 31 21 23 5.9 - 26 14 - 42 0.2 - 155 1.0 - 125 720 700 20 38 25 33 4.9 - 24 14 - 51 0 - 180 1.0 - 186 744 722 24 49

NO2 (ppb) DRIVE GIT 2666 2798 29 20 16 8.6 17 - 38 13 - 25 2.4 - 94 2.3 - 52 419 540 32 19 13 6.4 22 - 39 14 - 23 4.4 - 82 4.8 - 42 663 720 33 20 17 8.1 19 - 44 13 - 25 2.8 - 94 2.3 - 49 720 700 29 21 17 11 15 - 39 12 - 28 4.2 - 82 2.8 - 52 744 722 26 19

NOx (ppb) DRIVE GIT 2666 2798 50 57 35 34 25 - 67 32 - 74 2.7 - 252 4.9 - 263 419 540 50 52 26 24 30 - 66 34 - 68 3.7 - 174 11 - 139 663 720 53 50 33 27 27 - 71 30 - 66 3.7 - 202 5.8 - 150 720 700 49 59 39 40 21 - 62 28 - 77 4.0 - 225 4.9 - 213 744 722 50 68

SD

1.4

1.3

280

367

28

35

13

8

37

40

IQR Min-Max

0.7 - 2.0 0.06 - 12

1.0 - 2.4 0.1 - 9.4

285 - 593 168 - 1860

500 - 900 0 - 2200

5.5 - 34 0.05 - 201

23 - 65 1 - 233

16 - 34 2.4 - 75

12 - 24 3.7 - 47

24 - 67 2.7 - 252

39 - 89 5.1 - 263

BC: Black carbon, CO: Carbon monoxide, NO: Nitric oxide, NO2: Nitrogen dioxide, NOx: Nitrogen oxides, O3: Ozone, T: Temperature, N: Number of hours with observations, SD: Standard deviation, IQR: Inter Quartile Range, Min: Minimum observation, Max: Maximum observation

1

Table 2 Hourly averages of wind speed, temperature and traffic counts measured at the near-road continuous instrumentation.

2

September 8, 2014 to January 5, 2015.

Total 9/8-1/5

September 9/1-9/30

October 10/1-10/31

November 11/1-11/30

December 12/1 - 12/31

3

N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean SD IQR Min-Max N Mean SD IQR Min-Max

Wind Speed (mph) 2343 2.4 1.4 1.4 - 3.2 0 - 7.0 226 2.3 1.0 1.6 - 3.0 0 - 5.2 744 2.0 1.3 0.9 - 3.0 0 - 6.0 676 2.6 1.6 1.4 - 3.6 0 - 7.0 591 2.5 1.2 1.7 - 3.1 0.04 - 6.9

Temp (C)

Traffic Count*

2343 13 7 8 – 19 -5 - 30 226 22 3 19 - 23 14 - 29 744 19 6 15 - 23 4 - 30 676 9 6 5 - 14 -5 - 24 591 9 5 6 - 13 -1 - 22

1920 12300 5780 2180 - 19000 1400 - 20700 288 12600 5800 7300 - 17300 1560 - 20700 504 12500 5800 7500 - 17100 1500 - 20200 456 12100 5780 6870 - 17080 1400 - 20000 672 12100 5780 6680 - 17000 1490 - 20600

*Traffic count data were collected at a location on I-75/I-85 1.5 miles south of the measurement location using Automatic Traffic Records

1 2

Table 3 Regression coefficients from multivariate models examining associations between multiple factors and hourly pollutant concentrations at the NR DRIVE and NR GIT sites from September 8, 2014 to January 5, 2015. NR DRIVE BC Late Evening (9pm-12am) Early Morning (1-5am) Morning Rush Hour (6-9am) Mid Day (10am-3pm) Temperature (C) Relative Humidity (%) Wind Speed (mph) Northerly Wind Easterly Wind Southerly Wind Weekend Traffic Count (per 1,000) Mixing Heights (per 100 meters)

Est. 0.23* 0.24* 0.39* 0.14 0.01* 0.00 -0.20* -0.27* -0.73* -0.30 -0.61* 0.06* -0.03*

NR GIT BC

95% CI (0.04, 0.42) (0.03, 0.44) (0.13, 0.64) (-0.04, 0.32) (0.00, 0.02) (0.00, 0.01) (-0.26, -0.14) (-0.47, -0.08) (-0.95, -0.51) (-0.76, 0.17) (-0.97, -0.26) (0.04, 0.07) (-0.05, -0.01)

Est. 0.54* 0.52* 0.70* 0.49* -0.02* 0.00 -0.19* -0.74* -1.25* -0.81* 0.00* 0.00

NR DRIVE CO

95% CI (0.29, 0.79) (0.24, 0.79) (0.37, 1.03) (0.25, 0.72) (-0.04, -0.00) (-0.01, 0.01) (-0.27, -0.12) (-1.28, -0.20) (-1.79, -0.70) (-1.40, -0.22) (0.06, 0.09) (-0.04, 0.02)

Est. 6.08 -8.52 -11.98 -27.31 2.35* -0.37 -40.75* -36.78* -59.43* -7.17 -11.84 8.60* -3.41

NR GIT CO

95% CI (-27.25, 39.41) (-44.92, 27.88) (-55.88, 31.92) (-58.23, 3.61) (0.74, 3.97) (-1.31, 0.58) (-50.09, -31.40) (-70.63, -2.92) (-97.25, -21.61) (-83.98, 69.64) (-74.45, 50.77) (6.56, 10.64) (-7.12, 0.30)

Est. 47.63* -22.45 -1.59 25.87 -4.00* -0.33 -45.35* -90.25* -191.16* -222.118 -9.13 0.02* -6.39*

NR DRIVE O3

95% CI (6.97, 88.29) (-66.23, 21.34) (-54.11, 50.94) (-11.83, 63.57) (-6.13, -1.86) (-1.59, 0.93) (-57.35, -33.34) (-134.28, -46.23) (-240.46, -141.87) (-324.75, -119.47) (-92.74, 74.49) (19.11, 24.09) (-11.14, -1.64)

Est. 1.31* 2.05* 1.85* 1.54* 0.40* -0.22* 2.54* 1.83* 5.06* 3.44* 2.34* -0.31* 0.47*

95% CI (0.24, 2.38) (0.89, 3.20) (0.42, 3.27) (0.56, 2.53) (0.35, 0.45) (-0.26, -0.19) (2.24, 2.84) (0.73, 2.93) (3.83, 6.29) (0.83, 6.04) (0.40, 4.28) (-0.37, -0.24) (0.06, 0.34)

3 NR DRIVE NO Late Evening (9pm-12am) Early Morning (1-5am) Morning Rush Hour (6-9am) Mid Day (10am-3pm) Temperature (C) Relative Humidity (%) Wind Speed (mph) Northerly Wind Easterly Wind Southerly Wind Weekend Traffic Count (per 1,000) Mixing Heights (per 100 meters)

4 5 6 7

Est. 4.73* 5.42* 7.53* 3.25* -0.10 0.21* -1.83* -5.64* -19.00* -20.66* -9.06* 1.09* -0.35

95% CI (1.89, 8.65) (3.69, 12.46) (-0.30, 5.97) (2.13, 9.47) (-0.28, 0.04) (0.13, 0.32) (-2.93, -1.02) (-29.36, -12.68) (-23.16, -15.18) (-9.29, -2.14) (-14.80, -3.15) (0.88, 1.30) (-0.75, 0.05)

NR GIT NO Est. 11.13* 11.92* 14.63* 7.81* -0.83* 0.12* -2.97* -9.16* -19.72* -26.59* -13.09* 0.00* 0.00

95% CI (6.97, 15.28) (7.35, 16.48) (9.22, 20.03) (3.97, 11.65) (-1.02, -0.63) (0.01, 0.24) (-4.15, -1.79) (-13.56, -4.76) (-24.62, -14.82) (-36.83, -16.35) (-20.32, -5.86) (2.10, 2.60) (-0.80, 0.16)

NR DRIVE NO2 Est. 0.27 -2.22* -1.85 -1.95* 0.04 -0.15* -5.06* -4.39* -10.21* -3.23* -4.97* 0.50* -0.64*

95% CI (-1.63, 2.17) (-4.26, -0.17) (-4.30, 0.60) (-3.71, -0.20) (-0.05, 0.13) (-0.20, -0.09) (-5.60, -4.53) (-6.36, -2.41) (-12.41, -8.00) (-7.84, 1.39) (-8.13, -1.82) (0.39, 0.61) (-0.86, -0.42)

NR GIT NO2 Est. 0.34 -1.00 -0.35 0.18 -0.05* -0.13* -2.39* -3.33* -5.89* -3.01* -2.44* 0.00* 0.00*

95% CI (-0.75, 1.43) (-2.20, 0.19) (-1.76, 1.05) (-0.82, 1.19) (-0.11, -0.00) (-0.16, -0.19) (-2.70, -2.07) (-4.50, -2.16) (-7.19, -4.59) (-5.72, -0.30) (-4.44, -0.45) (0.36, 0.50) (-0.45, -0.20)

NR DIRVE NOx Est. 5.00* 3.18 5.65 1.28 -0.06 0.06 -6.88* -10.04* -29.25* -23.95* -14.03* 1.58* -0.99*

95% CI (0.35, 9.65) (-1.82, 8.18) (-0.34, 11.64) (-3.00, 5.57) (-0.28, 0.16) (-0.07, 0.20) (-8.19, -5.56) (-14.88, -5.20) (-34.65, -23.85) (-35.24, -12.65) (-21.95, -6.10) (1.30, 1.86) (-1.53, -0.45)

NR GIT NOx Est. 11.39* 10.85* 14.18* 7.94* -0.89* 0.00 -5.33* -12.45* -25.53* -29.55* -15.40* 2.76* -0.64*

95% CI (6.59, 16.19) (5.58, 16.13) (7.94, 20.41) (3.50, 12.37) (-1.11, -0.66) (-0.14, 0.13) (-6.70, -3.96) (-17.54, -7.36) (-31.20, -19.85) (-41.41, -17.70) (-24.01, -6.80) (2.47, 3.05) (-1.19, -0.08)

*All covariates were included simultaneously in the model from each pollutant of interest. Est. Estimate of Coefficient; 95% CI- 95% Confidence Interval; Unit for BC: µg m-3; Unit for CO, NO, NO2, NOx, and O3: ppb; *Significant with a P-value of 0.05

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

References Ayala A, Brauer M, Mauderly JL, Samet JM. 2012. Air pollutants and sources associated with health effects. Air Quality Atmosphere and Health 5:151-167. Baldauf R, Thoma E, Hays M, Shores R, Kinsey J, Gullett B, et al. 2012. Traffic and meteorological impacts on near-road air quality: Summary of methods and trends from the raleigh near-road study. Journal of the Air & Waste Management Association 58:865-878. Baldauf R. 2017. Roadside vegetation design characteristics that can improve local, near-road air quality. Transportation Research Part D-Transport and Environment 52:354-361. Baldauf RW, Heist D, Isakov V, Perry S, Hagler GSW, Kimbrough S, et al. 2013. Air quality variability near a highway in a complex urban environment. Atmospheric Environment 64:169178. Baldauf RW, Isakov V, Deshmukh P, Venkatram A, Yang B, Zhang KM. 2016. Influence of solid noise barriers on near-road and on-road air quality. Atmospheric Environment 129:265276. Barth M, Scora G, Younglove T. Estimating emissions and fuel consumption for different levels of freeway congestion. Transportation Research Record. Batterman S, Cook R, Justin T. 2015. Temporal variation of traffic on highways and the development of accurate temporal allocation factors for air pollution analyses. Atmospheric Environment 107:351-363. Beckerman B, Jerrett M, Brook JR, Verma DK, Arain MA, Finkelstein MM. 2008. Correlation of nitrogen dioxide with other traffic pollutants near a major expressway. Atmospheric Environment 42:275-290. Blanchard CL, Hidy GM, Tanenbaum S, Edgerton ES, Hartsell BE. 2013a. The southeastern aerosol research and characterization (search) study: Temporal trends in gas and pm concentrations and composition, 1999-2010. Journal of the Air & Waste Management Association 63:247-259. Blanchard CL, Tanenbaum S, Hidy GM. 2013b. Source attribution of air pollutant concentrations and trends in the southeastern aerosol research and characterization (search) network. Environmental Science & Technology 47:13536-13545. Boogaard H, Kos GPA, Weijers EP, Janssen NAH, Fischer PH, van der Zee SC, et al. 2011. Contrast in air pollution components between major streets and background locations: Particulate matter mass, black carbon, elemental composition, nitrogen oxide and ultrafine particle number. Atmospheric Environment 45:650-658. Brugge D, Durant JL, Rioux C. 2007. Near-highway pollutants in motor vehicle exhaust: A review of epidemiologic evidence of cardiac and pulmonary health risks. Environ Health-Glob 6:12. Chen H, Kwong JC, Copes R, Tu K, Villeneuve PJ, Van Donkelaar A, et al. 2017. Living near major roads and the incidence of dementia, parkinson's disease, and multiple sclerosis: A population-based cohort study. The Lancet 389:718-726. Dallmann TR, Harley RA. 2010. Evaluation of mobile source emission trends in the united states. Journal of Geophysical Research-Atmospheres 115:12. DeWinter JL, Brown SG, Seagram AF, Landsberg K, Eisinger DS. 2018. A national-scale review of air pollutant concentrations measured in the us near-road monitoring network during 2014 and 2015. Atmospheric Environment 183:94-105.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Edgerton ES, Hartsell BE, Saylor RD, Jansen JJ, Hansen DA, Hidy GM. 2005. The southeastern aerosol research and characterization study: Part ii. Filter-based measurements of fine and coarse particulate matter mass and composition. Journal of the Air & Waste Management Association 55:1527-1542. Edgerton ES, Hartsell BE, Saylor RD, Jansen JJ, Hansen DA, Hidy GM. 2006. The southeastern aerosol research and characterization study, part 3: Continuous measurements of fine particulate matter mass and composition. Journal of the Air & Waste Management Association 56:13251341. EPA U. 2010. Primary national ambient air quality standards for nitrogen dioxide. In: Final rule. 40 cfr parts 50 and 58, vol. 75. Federal register. No. 26. EPA U. 2011. Review of national ambient air quality standards for carbon monoxide. In: Final rule. 40 cfr parts 50, 53 and 58, vol. 76. Federal register. No. 169. EPA U. 2016. Integrated science assessment for oxides of nitrogen - health criteria. GDOT. 2014. Traffic counts in georgia Available: http://geocounts.com/gdot/. Golan R, Ladva C, Greenwald R, Krall JR, Raysoni AU, Kewada P, et al. 2018. Acute pulmonary and inflammatory response in young adults following a scripted car commute. Air Quality, Atmosphere & Health 11:123-136. Grosjean D. 1983. Distribution of atmospheric nitrogenous pollutants at a los angeles area smog receptor site. Environmental science & technology 17:13-19. Hansen DA, Edgerton ES, Hartsell BE, Jansen JJ, Kandasamy N, Hidy GM, et al. 2003. The southeastern aerosol research and characterization study: Part 1-overview. Journal of the Air & Waste Management Association 53:1460-1471. Hansen DA, Edgerton E, Hartsell B, Jansen J, Burge H, Koutrakis P, et al. 2006. Air quality measurements for the aerosol research and inhalation epidemiology study. Journal of the Air & Waste Management Association 56:1445-1458. HEI. 2009a. Traffic-related air pollution: A critical review of the literature on emissions, exposure, and health effects. HEI. 2009b. Air pollution and health: A european and north american approach (aphena). HEI. 2010. Traffic-related air pollution: A critical review of the literature on emissions, exposure, and health effects. (HEI Panel on the Health Effects of Traffic-Related Air Pollution). Boston, MA:Health Effect Institute. Henneman LRF, Holmes HA, Mulholland JA, Russell AG. 2015. Meteorological detrending of primary and secondary pollutant concentrations: Method application and evaluation using longterm (2000-2012) data in atlanta. Atmospheric Environment 119:201-210. Hilker N, Wang JM, Jeong C-H, Healy RM, Sofowote U, Debosz J, et al. 2019. Traffic-related air pollution near roadways: Discerning local impacts from background. Atmospheric Measurement Techniques 12:5247-5261. Karner AA, Eisinger DS, Niemeier DA. 2010. Near-roadway air quality: Synthesizing the findings from real-world data. Environmental Science & Technology 44:5334-5344. Liang D, Golan R, Moutinho JL, Chang HH, Greenwald R, Sarnat SE, et al. 2018a. Errors associated with the use of roadside monitoring in the estimation of acute traffic pollutant-related health effects. Environmental research 165:210-219. Liang D, Moutinho JL, Golan R, Yu T, Ladva CN, Niedzwiecki M, et al. 2018b. Use of highresolution metabolomics for the identification of metabolic signals associated with traffic-related air pollution. Environment international 120:145-154.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

Liang D, Ladva CN, Golan R, Yu T, Walker DI, Sarnat SE, et al. 2019. Perturbations of the arginine metabolome following exposures to traffic-related air pollution in a panel of commuters with and without asthma. Environment international 127:503-513. Lin C, Feng XF, Heal MR. 2016. Temporal persistence of intra-urban spatial contrasts in ambient no2, o-3 and ox in edinburgh, uk. Atmos Pollut Res 7:734-741. Liu W, Wang YH, Russell A, Edgerton ES. 2005. Atmospheric aerosol over two urban-rural pairs in the southeastern united states: Chemical composition and possible sources. Atmospheric Environment 39:4453-4470. McAdam K, Steer P, Perrotta K. 2011. Using continuous sampling to examine the distribution of traffic related air pollution in proximity to a major road. Atmospheric Environment 45:20802086. McDonald BC, Goldstein AH, Harley RA. 2015. Long-term trends in california mobile source emissions and ambient concentrations of black carbon and organic aerosol. Environmental Science & Technology 49:5178-5188. Menut L, Goussebaile A, Bessagnet B, Khvorostiyanov D, Ung A. 2012. Impact of realistic hourly emissions profiles on air pollutants concentrations modelled with chimere. Atmospheric environment 49:233-244. NCAR. Weather research and forecasting model. Available: https://www.mmm.ucar.edu/weather-research-and-forecasting-model. Saha PK, Reece SM, Grieshop AP. 2018. Seasonally varying secondary organic aerosol formation from in-situ oxidation of near-highway air. Environmental Science & Technology 52:7192-7202. Sarnat JA, Russell AG, Liang D, Moutinho JL, Golan R, Weber R, et al. 2018. Developing multipollutant exposure indicators of traffic pollution: The dorm room inhalation to vehicle emissions (drive) study:Health Effects Institute. Sarnat SE, Sarnat JA, Klein M, Goldman G, Mulholland J, Russell AG, et al. 2008. Applying alternative approaches to characterizing air pollution exposure in an epidemiologic study in atlanta. Epidemiology 19:S38-S38. Smith RB, Fecht D, Gulliver J, Beevers SD, Dajnak D, Blangiardo M, et al. 2017. Impact of london's road traffic air and noise pollution on birth weight: Retrospective population based cohort study. bmj 359:j5299. Sofowote U, Healy R, Su Y, Debosz J, Noble M, Munoz A, et al. 2018. Understanding the pm2. 5 imbalance between a far and near-road location: Results of high temporal frequency source apportionment and parameterization of black carbon. Atmospheric Environment 173:277-288. Solomon PA, Chameides W, Weber R, Middlebrook A, Kiang CS, Russell AG, et al. 2003. Overview of the 1999 atlanta supersites project. J Geophys Res 108:10.1029/2001JD001458. Vijayaraghavan K, DenBleyker A, Ma L, Lindhjem C, Yarwood G. 2014. Trends in on-road vehicle emissions and ambient air quality in atlanta, georgia, USA, from the late 1990s through 2009. Journal of the Air & Waste Management Association 64:808-816. Wang JM, Jeong C-H, Hilker N, Shairsingh KK, Healy RM, Sofowote U, et al. 2018. Near-road air pollutant measurements: Accounting for inter-site variability using emission factors. Environmental science & technology 52:9495-9504. WHO. 2005. Health effects of transportion-related air pollution. Zhai XX, Mulholland JA, Russell AG, Holmes HA. 2017. Spatial and temporal source apportionment of pm2.5 in georgia, 2002 to 2013. Atmospheric Environment 161:112-121.

Highlights Near-road Vehicle Emissions Air Quality Monitoring for Exposure Modeling Jennifer L. Moutinho, Donghai Liang, Rachel Golan, Stefanie Ebelt Sarnat, Rodney Weber, Jeremy A. Sarnat, Armistead G. Russell •

Lesser impact from highway source on pollutant levels compared to prior studies

•

Low predictive power of traffic count on near-road pollutant levels

•

Results indicative of the effectiveness of mobile source emission controls

•

Use of observations from a single near-road site can lead to biases in assessing exposures to mobile emissions

Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: