Atmospheric Pollution Research 6 (2015) 662 672

© Author(s) 2015. This work is distributed under the Creative Commons Attribution 3.0 License.

AAtm spheric PPollution RResearchwww.atmospolres.com

Rapid physical and chemical transformation of traffic–related atmospheric particles near a highway

Cheol–Heon Jeong 1, Greg J. Evans 1, Robert M. Healy 1, Parnian Jadidian 1, Jeremy Wentzell 2, John Liggio 2,Jeffrey R. Brook 2

1 Southern Ontario Centre for Atmospheric Aerosol Research, University of Toronto, Toronto, ON, Canada2 Air Quality Research Division, Environment Canada, Toronto, ON, Canada

ABSTRACTThe health of a substantial portion of urban populations is potentially being impacted by exposure to traffic–relatedatmospheric pollutants. To better understand the rapid physical and chemical transformation of these pollutants, thenumber size distributions of non–volatile traffic–related particles were investigated at different distances from a majorhighway. Particle volatility measurements were performed upwind and downwind of the highway using a fast mobilityparticle sizing spectrometer with a thermodenuder on a mobile laboratory. The number concentration of non–denudedultrafine particles decreased exponentially with distance from the highway, whereas a more gradual gradient wasobserved for non–volatile particles. The non–volatile number concentration at 27 m was higher than that at 280 m by afactor of approximately 3, and the concentration at 280 m was still higher than that upwind of the highway. Theproportion of non–volatile particles increased away from the highway, representing 36% of the total particle number at27 m, 62% at 280 m, and 81% at the upwind site. A slight decrease in the geometric mean diameter of the non–volatileparticle size distributions from approximately 35 nm to 30 nm was found between 27 m and 280 m, in contrast to thegrowth of non–denuded particles with increasing distance from the highway. Single particle analysis results show thatthe contribution of elemental carbon (EC)–rich particle types at 27 m was higher than the contribution at 280 m by afactor of approximately 2. The findings suggest that people living or spending time near major roadways could beexposed to elevated number concentrations of nucleation–mode volatile particles (<30 nm), Aitken–mode non–volatileparticles (30–100 nm), and EC–rich fine–mode particles (>100 nm). The impact of the highway emissions on air qualitywas observable up to 300 m.

Keywords: Near–road, non–volatile particles, traffic–related air pollutant, spatial variation, single particle analysis

Corresponding Author:Cheol–Heon Jeong

: +1 416 978 5932: +1 416 978 8605: ch.jeong@utoronto.ca

Article History:Received: 13 October 2014Revised: 15 January 2015Accepted: 26 January 2015

doi: 10.5094/APR.2015.075

1. Introduction

Traffic–related atmospheric pollutants (TRAPs), includingultrafine particles (UFP, <100 nm in diameter), black carbon (BC),polycyclic aromatic hydrocarbons (PAHs), and volatile organiccompounds (VOCs), are believed to adversely impact the health ofpopulations living and working near roadways. Proximity to majorroadways and exposure to TRAPs may be associated with a numberof respiratory and cardiovascular issues including asthma, reducedlung function, adverse birth outcomes, cardiac effects, respiratorysymptoms, premature mortality, and lung cancer (e.g., HEI, 2010).Most recently, diesel exhaust was officially classified as carcinogenic by the International Agency for Research on Cancer (IARC) ofthe World Health Organization (Straif et al., 2013). Gauderman etal. (2007) found that children who lived within 500 m of a freewayhad substantial deficits in lung function, when compared tochildren who lived 1 500 m or more from a freeway. A cohort studyin Toronto, Canada suggested that traffic proximity (residencewithin 50 m of a major road or 100 m of a highway) may beassociated with an increase in circulatory mortality of over 40%(Jerrett et al., 2009). In Canada approximately 4 million and10 million people live within 100 m and 250 m of a major road,respectively, based on 2006 Census data (Evans et al., 2011). Thusa high proportion of Canada’s urban populations are potentiallyexposed to TRAPs. For example, in Toronto, the most populatedcity in Canada, 13% and 56% of the population lives within 50 mand 250 m of a major road, respectively.

Assessment of exposure to TRAPs can be challenging due tothe intra–urban spatial variability of pollutants, such as UFP andBC. Near–road field studies have shown a large difference in theconcentrations of traffic–related pollutants between near–roadmeasurements and typical air quality monitoring stations (e.g.,Riediker et al., 2003). A number of on–road measurements haveshown that significantly higher concentrations of traffic–relatedpollutants can occur on roadways as compared to backgroundlevels measured 150–500 m away (Hitchins et al., 2000; Bukowieckiet al., 2002; Riediker et al., 2003; Kittelson et al., 2004; Pirjola etal., 2004; Kaur et al., 2005; Weimer et al., 2009; Hagler et al., 2010;Pirjola et al., 2012). Roadside field studies have also consistentlyfound decreasing concentrations of UFP and NO2 with increasingdistance from a road, while there were inconsistent trends in PM10and PM2.5 (Roorda–Knape et al., 1998; Zhu et al., 2002a; Zhu et al.,2002b; Westerdahl et al., 2005; Beckerman et al., 2008; Hagler etal., 2009; Durant et al., 2010; Gordon et al., 2012; Salimi et al.,2013).

Additionally, there is a growing interest in understanding thespatial variability of particle chemical properties near busyroadways. In particular, particle volatility is an important propertyof UFP in terms of the potential health effects of particles as wellas the spatial variation of particles near major roadways. Traffic–related UFP generally contain compounds that range in volatility,such as organics from unburned fuel, lubricating oil, and graphiticcarbon (e.g., Kittelson, 1998; Sakurai et al., 2003). To characterizethe volatility of traffic–related particles, different types of tandemdifferential mobility analyzers (TDMA) have been deployed near

Jeong et al. – Atmospheric Pollution Research (APR) 663

major roadways (Wehner et al., 2004; Biswas et al., 2007; Tiitta etal., 2010). The TDMA system provides detailed information aboutthe volatile fraction of an individual particle diameter, but noinformation about the volatile/non–volatile fractions of the entiresize distribution. Little work has been done to differentiatebetween the entire volatile and non–volatile fraction of UFP nearmajor roadways (Wehner et al., 2004; Pirjola et al., 2012). Thespatial variation of the particle volatility in the entire sizedistribution can provide useful information about exposure tonon–volatile particles as a function of distance from a roadway.

In this paper, the evolution of volatile and non–volatile UFP ina near–road environment was examined using a thermodenuder(TD) and a fast mobility particle sizing spectrometer (FMPS) atdistances up to 280 m from a major highway. In addition, singleparticle analysis using an aerosol time–of–flight mass spectrometer(ATOFMS) provided insights into the size and composition of theaccumulation mode particles (100–3 000 nm). The measurementsof aerosol volatility and chemical composition provide an improvedunderstanding of the evolution of traffic–related particles in thevicinity of major roadways. This information is potentially usefulfor understanding the associated exposure and health impacts.

2. Methods

2.1. Measurement design

The study was performed in the Greater Toronto Area,Ontario, Canada from August 17 to September 17, 2010 as a partof the Fast Evolution of Vehicle Emissions near Roadways (FEVER)campaign. The University of Toronto mobile laboratory (MAPLE)was deployed to measure the decay gradients of TRAPs nearhighway 400. This major highway has 6 lanes that, when combined,carry 90 000 vehicles/day (Liggio et al., 2012). The mobile lab,MAPLE, was moved back and forth among sites 27 m (Site A) and280 m (Site B) downwind of the highway, as well as at 90 m (Site C)upwind of the highway (Figure 1). During these mobile measurements, traffic–related pollutants were sampled at each site for atleast 1 hour before MAPLE moved to the next site. Only thosemorning and evening rush–hour periods on August 19, 26, and 27during which the wind was perpendicular to the highway (200–330) were chosen for UFP gradient analysis. Information on themobile measurement days and times is listed in Table 1.

In addition to the mobile measurements, UFP was continuously measured at a stationary monitoring site (Site D) located

15 m from the highway. The site was surrounded by agriculturalareas with wide open level terrain and there were no nearbysignificant point sources of ultrafine particles or other mainroadways. In order to explore the influence of the highway trafficon particle number concentrations, the relationship between UFPconcentration and wind direction was depicted by using a windrose plot generated with 1 minute averaged data at Site D (see theSupporting Material, SM, Figure S1). The results show that highernumber concentrations (>40 000 cm–3) occurred when the windwas coming from the highway, further confirming that the highwaywas the only significant source of UFP present.

In order to evaluate the micro–scale gradient of UFP within20 m, UFP number concentrations were simultaneously measuredat 3 m (near the shoulder of the highway) and 18 m downwind ofthe highway on September 1. The FMPS was connected to a longpiece of conductive tubing ( 15 m) for sampling at the 3 mlocation, whereas a short piece ( 30 cm) was used for the secondFMPS at the 18 m sampling. The long sampling inlet was movedback to the 18 m location every hour to calculate the diffusionallosses through the long sampling line. The wind speed and winddirection were relatively constant over the 8 h measurementperiod. To investigate the spatial transformation of individualplumes from heavy emitters, 37 high particle number events wereselected by inspecting particle number spikes (>40 000 cm–3 at the3 m location) and video observations.

From September 2 to September 17, MAPLE was parked atSite A or B to characterize individual ambient particles. To ensurethat vehicle emissions were sampled, the analysis was limited tosingle particle mass spectral data measured during morning rushhours (5:30 am to 11:00 am) on the days (September 14 and 15)with winds between the southwest (200°) and the northwest(330°). It should be noted that data from low wind speed(<0.5 m s–1) conditions were also excluded from the analysis.

Table 1 summarizes the weather conditions and traffic densityon these sampling days. While wind direction was relativelyconstant for these selected days, average temperature and windspeed varied from 12 to 31 °C and from 2 to 8 m s–1, respectively.The number density of light duty vehicles (i.e., cars) was relativelyconstant, although on August 27 (Friday) the traffic density washigher than on other weekdays. Heavy–duty diesel vehicles (i.e.,heavy class) accounted for approximately 4% of the total vehicleson August 26 and 27. This traffic pattern has been described indetail elsewhere (Liggio et al., 2012).

Figure 1. Location of the measurement sites near Highway 400, Ontario, Canada: Site A (27 m from the highway), Site B(280 m from the highway), Site C (upwind 90 m from the highway), and Site D (stationary site 15 m from the highway)

(created in Google Maps, maps.google.com).

Jeong et al. – Atmospheric Pollution Research (APR) 664

Table 1.Meteorological parameters and traffic density during the particle volatility (August 19, 26, 27), the micro–scale gradient(September 1), and the single particle analysis study (September 14, 15)

Date MeasuringTime

Meteorological Parameters Traffic Density (vehicles h–1)Temp a

(°C)WD b

(Degree)WS c

(m s–1) Car Medium d Heavy e

08/19/2010 06:40–11:40 22 220–270 4.8 n/a f n/a f n/a f

08/26/2010 06:40–11:5018:10–20:10 18 280–330 6.2 5 300 270 270

08/27/2010 08:30–12:0015:00–18:10 20 250–310 2.7 7 500 280 240

09/01/2010 10:30–18:20 31 200–250 8.4 5 100 230 22009/14/2010 05:30–10:30 13 270–290 4.8 5 100 220 24009/15/2010 05:30–09:40 12 290–330 2.4 5 500 230 220

a Average temperature (Temp), b wind direction (WD), and c wind speed (WS) at 7 m height. dmedium (between passengercars and heavy–duty diesel trucks) sized vehicles, e heavy–duty diesel vehicles. f not available

2.2. Aerosol measurement and data quality assurance

The mobile lab, MAPLE, included measurements of particlesize distribution and chemical composition using a fast mobilityparticle sizing spectrometer (FMPS, TSI 3091), an aerosol time–of–flight mass spectrometer (ATOFMS, TSI 3800), an aethalometer forBC (Magee AE–21), a particle–bound PAH (p–PAH) monitor(EcoChem PAS 2000CE), two PM2.5 DustTraks (TSI 8520), a weathersensor, and a video camera. A second FMPS (TSI 3091) wasdeployed at Site D located 15 m from the highway to continuouslymeasure number and size distributions of UFP. The UFP data atSite D was used to normalize data at Sites A, B, and C to accountfor the temporal variations of ambient UFP; these variations occurnot only due to traffic sources, but also due to changes inbackground concentrations and atmospheric stability. In order toaccount for the temporal variations of UFP while MAPLE wasmoving between Sites A, B, and C, the number concentrations ateach location were adjusted according to the temporal variationsof UFP measured at the stationary Site D using the followingequation:

(1)

where, the subscript k denotes the site (i.e., Site A, Site B, Site C),Nk* is the adjusted particle number concentration at site k, Nk isthe measured number concentration at site k, is the averagenumber concentration at Site D during the entire measurementperiod, and Ns,k is the number concentration measured at Site Dduring the corresponding measurement period at site k. Theadjustment factor ( ) of the temporal variation ranged from0.84 to 1.38 over the measurement period on August 19, 26, and27.

The FMPS instruments provide particle size distributions from6 nm to 560 nm in electrical mobility diameter with 1 second timeresolution. Particle electrical mobility diameter is defined as thediameter of a spherical particle with the same mobility in anelectric field as the particle of interest (Baron and Willeke, 2001).The performance and correction of the FMPS has been described indetail elsewhere (Jeong and Evans, 2009). A comparison betweenthe two FMPSs and a scanning mobility particle sizer (SMPS) as wellas a CPC (TSI 3786) was conducted before and after the campaignin a laboratory, showing good correlations (r2>0.9) and agreementwithin 20%. Weekly intercomparison of the FMPSs and dailybackground checks with zero air were conducted at the field site.Particles smaller than 8 nm were excluded due to poor countingefficiency in this size range. The number size distribution ofparticles in the size range of 8 to 560 nm measured by the FMPSwere corrected as recommended by Jeong and Evans (2009) andZimmerman et al. (2015). In addition, the last three corrected sizebins (>540 nm in the corrected diameter) of the FMPS were

excluded due to poor counting efficiency. Zimmerman et al. (2014)recently reported that an overestimation of particle countingefficiency in the FMPS is observed for fresh diesel particles with amode diameter of 80 nm due to the high fractal dimension of sootagglomerates. Since these fresh agglomerates occur primarily inplumes from high emitting vehicles, Zimmerman et al. (2014)suggest that this FMPS error is negligible when using longeraveraging time for number concentrations. In this work, averageconcentrations were estimated for >30 min. In order to evaluatethe change of particle size distributions, the geometric meandiameter (GMD, dg) of the FMPS size distribution was calculated byEquation (2):

(2)

where, N is the total number concentration, m is the first size bin,n is the last size bin, di is the midpoint diameter of size bin, i, andNi is number concentration within the size bin i.

For the analysis of micro–scale variability of UFP within 20 m,the particle diffusion losses through the long sampling line used forthe measurement at 3 m from the highway were empiricallycorrected for by comparing paired measurements with andwithout the sampling line. As expected, particle losses increased asparticle size decreased. For example, there were approximately74% and 16% losses for 8 nm and 108 nm particles, respectively.The size resolved correction factors were applied to theconcentrations recorded by the FMPS with the long sampling line.

Black carbon concentrations were measured at 1 minuteresolution using a dual channel aethalometer (880 nm and370 nm). The concentrations of BC were calculated from lightattenuation measurements at 880 nm of the aethalometer using amass absorption coefficient of 16.6 m2 g 1. The flow rate of 5 Lpmand background concentrations with filtered air were checked on aweekly basis. The PAS is a portable photoelectric device whichprovides mass concentration of particle–bound PAHs absorbed oncarbon particles by UV–photoionization. No attempt was made tocalibrate and quantify the PAS since there is no absolutecalibration standard available. Instead, results were interpreted asrelative units. Semi–quantitative 1 minute resolution PM2.5 massconcentrations were measured by the 8520 DustTrak equippedwith a 2.5 μm inlet nozzle. Background checks of the DustTrakwere also performed on a daily basis. The second DustTrak (TSI8520, DT2) was used to monitor the performance of the firstDustTrak (DT1) by measuring ambient particles every week.Excellent correlations (r2>0.98) between these DustTraks werefound, while DT1 was consistently higher ( 24%) than the readingof DT2 over the measurement period. Mass concentrations ofparticles smaller than 0.6 μm (PM0.6) were estimated using theFMPS number concentration in the size range of 8 to 540 nm with

Jeong et al. – Atmospheric Pollution Research (APR) 665

the assumption of spherical particle shape and a density of1.2 g cm 3. The PM2.5 mass concentration of the DT1 was highlycorrelated with the PM0.6 mass concentration (r2=0.89), and theslope of the regression was 2.3 with the intercept of zero. It hasbeen reported by other studies that DustTrak measurements usinga light scattering technology could be overestimated (e.g., Evans etal., 2008). However, no correction factor was applied for theDustTrak measurement due to the absence of referencemeasurements for PM2.5 in our study.

2.3. Volatility measurement

In order to measure the number concentrations and sizedistributions of non–volatile particles, the FMPS was equippedwith an automated thermodenuder (TD, Dekati Ltd) allowing forrapid alternation between TD–treated and untreated samplesevery 5 min. In this study, particles that evaporate in thethermodenuder at a temperature of 250 °C are considered as“volatile”, while the particles remaining after heating are referredto as “non–volatile”. The non–volatile fraction by number (Nnv) isdefined as the ratio of the number concentration after heating at250 °C to the number concentration without heating, while thevolatile faction by number is defined as 1–Nnv.

Particle losses caused by diffusional and thermophoreticprocesses inside the TD were theoretically and experimentallyevaluated for various temperatures (30–250 °C) and particle sizes(15 nm–100 nm) at a flow rate of 10 L min–1. The residence time ofthe particles in the heating column of the TD was approximately0.3 second. To evaluate the transmission efficiency, monodispersed 40 nm, 60 nm, 80 nm and 100 nm gold particles weregenerated by an atomizer (TSI 3076) and differential mobilityanalyzer (DMA, TSI 3081) system. The particles alternately wentthrough either the TD or to a bypass line and were then measuredby a condensation particle counter (CPC, TSI 3786). Particle losseswith heating at 30 °C were negligible, while the transmissionefficiencies at 250 °C were approximately 76%. The experimentaltransmission efficiencies were in good agreement with the overalltheoretical transmission efficiency ( 77%) estimated by combiningthe diffusional and thermophoretic efficiencies in the 15–100 nmsize range (Baron and Willeke, 2001; Lin and Tsai, 2003). Thetheoretical diffusional and thermophoretic transmission efficiencies were approximately 99% and 78%, respectively, in the system.With decreasing particle size and increasing temperature, thetransmission efficiency decreased due to particle diffusion andthermophoretic losses. Longer residence time resulted in poorerparticle transmission efficiencies. The number concentration dataobtained using TD–FMPS were corrected for these particle losses.

2.4. Single particle analysis

The ATOFMS has been used to provide size resolved chemicalcomposition for fine particulate matter (PM) larger than 100 nm inreal time and to derive plausible sources of PM in many locations(e.g., Jeong et al., 2011; Healy et al., 2012; Rehbein et al., 2012). Tocharacterize individual ambient particles using the ATOFMS,ambient air is drawn through an aerodynamic focusing lens to theATOFMS sizing region. The particle size is determined by measuringthe transit time between two lasers in the sizing region. The size–resolved particles enter the mass spectrum region where a 266 nmNd:YAG laser desorbs and ionizes these particles. The positive andnegative ion mass spectra of each particle are obtained in the massspectrometer region. Metal standards and polystyrene latex (PSL)particles were used for mass spectra and particle size calibrationbefore and after the campaign.

During the field campaign, a total of 91 317 particles rangingfrom 0.1 to 2.3 μm were chemically analyzed using the ATOFMS atthe 27 m and 280 m locations. In order to concentrate on traffic–related particles, approximately 1 000 particle mass spectra werechosen and classified for both the 27 m and 280 m locations. All

individual raw particle mass spectra were peak–listed using the TSIMS–Analyze software. Single particle mass spectral data for eachparticle were classified into a number of particle types using aclustering toolkit, YAADA (www.yaada.org). The AdaptiveResonance Theory Artificial Neural Network (ART–2a) algorithm inthe Matlab–based toolkit was used for the clustering analysis witha vigilance factor of 0.85 and a learning factor of 0.05 (Song et al.,1999). Using the analysis, 35 particle clusters for the 27 m site and39 clusters for the 280 m site were initially obtained and re–grouped into 5 and 7 particle types, respectively, based on theirsimilarities in temporality, mass spectral signatures, and sizedistributions.

3. Results and Discussion

3.1. Distance decay gradient of ultrafine particles

The number concentrations of non–denuded particles(without TD) between 8 nm and 540 nm in diameter measuredupwind and downwind of the highway are summarized in Table 2.In this study the mean and the 95% confidence interval of UFPconcentrations were obtained for three weekdays. As discussedabove, the concentrations at the 27 m, 280 m and –90 m (upwind)locations were adjusted based on the concentrations measured atthe stationary site (Site D) located 15 m downwind of the highway.This adjustment is useful to account for any diurnal or day–to–dayvariations of particle concentrations while MAPLE was movingbetween these sites. The results were further corrected to isolatethe highway related particles, by subtracting the backgroundparticle concentrations measured 90 m upwind of the highway.The average UFP concentration upwind of the highway (Site C) wasabout 50% lower than the concentration at the 280 m downwindlocation (Site B).

As expected, a strong exponential decrease in the downwindUFP number concentrations with increasing distance from thehighway was observed (Figure S2). Based on these results, it wasestimated that particle numbers in the 8 to 540 nm size range canbe elevated by factors of 5 and 3 at distances of 15 m and 27 m,respectively, when compared to the average at a distance of 280 mfrom the highway. Background corrected particle numbers, whichcould be more representative of traffic–related particles, at 15 mand 27 m were higher than the concentrations at 280 m by factorsof 10 and 5, respectively. The decay rate of UFP number concentration near the highway was comparable to that measured at ahighway in Los Angeles, USA where UFP at 30 m downwind of thehighway was a factor of 4 higher than the average concentrationsat 300 m downwind (Zhu et al., 2002a).

From the micro–scale gradient analysis between 3 and 18 mfrom the highway, the average UFP concentration (89 000±29 000 cm–3) measured at the 3 m location was found to be higherthan the concentration (42 000±7 000 cm–3) at the 18 m locationby a factor of approximately 2. This implies that the exposure riskof pedestrians and cyclists to UFP near major roadways can bevaried even within a near road microenvironment. More extensiveresearch on the micro–scale gradient study of TRAPs nearroadways is required to fully evaluate the exposure risk.

3.2. Differences in particle size distribution with distance fromthe highway

The size–resolved number concentrations observed at different distances are presented in Table 2 and Figure 2. A higherconcentration gradient between the 15 m and the 280 m locationswas observed for the smaller diameter particles. The gradient washighest for smaller UFP particles (<30 nm); the concentrations at15 m and 27 m were higher than those at 280 m by factors of 14and 7, respectively. The decay gradient for particles 30–200 nmbecame more gradual as particle size increased. No gradient wasdetectable for particles larger than 200 nm.

Jeong et al. – Atmospheric Pollution Research (APR) 666

Table 2. Average number concentrations with 95% confidence intervals of total particles (without TD) measured at the downwind andupwind sites during the particle volatility study on August 19, 26, and 27

Distance from Highway 15 m Downwind(Site D)

27 m Downwind(Site A)

280 m Downwind(Site B)

–90 m Upwind(Site C)

Total PN a (particle cm–3) 32 400±5 100 18 400±3 000 6 000±980 3 000±200PN8–20 nm (particle cm–3) 20 000±4 500 10 300±2 600 1 800±660 400±110PN20–30 nm (particle cm–3) 4 100±1 030 2 100±430 600±140 400±20PN30–50 nm (particle cm–3) 3 100±350 1 700±230 900±170 500±30PN50–100 nm (particle cm–3) 4 100±230 3 000±260 2 000±260 1 100±70PN100–540 nm (particle cm–3) 1 200±60 1 200±110 900±110 600±30PM0.6

b (μg m–3) 8.1±0.4 8.4±0.6 6.5±0.6 4.6±0.3BC (μg m–3) n/a 1.7±0.2 0.9±0.3 0.5±0.3p–PAH (relative unit) n/a 6.7 2.4 1PM2.5 (μg m–3) n/a 10±0.4 9±0.5 8±0.2

a Particle number concentrations (PN) at average temperatures of 18–22 °C and average wind speeds of 3–6 m s–1 withwind directions ranged from 220 to 330 °. The average traffic flow was approximately 7 000 vehicles h–1b Estimated particle mass concentrations (PM) based on the number concentrations of UFP measured by the FMPSassuming a density of 1.2 g cm–3

Figure 2. Average number size distributions of non–denuded particles(w/o TD) measured at the upwind and downwind sites on August 19, 26,

and 27. Error bars represent the 95% confidence intervals.

The size distributions at the downwind locations were bimodalwith one mode at about 11 nm and a second larger mode between30 nm and 100 nm. In contrast, the particle size distribution at theupwind location and at 280 m downwind was unimodal with apeak at approximately 70 nm (Figure 2). The size distributionscorrected by subtracting the upwind concentration to isolate thesize distributions of traffic–related particles only is also shown inFigure S3 (see the SM). Figure 3 depicts the average sizedistributions at the 3 m and 18 m locations during the 37 highemitter plume events. As shown in Figures 2 and 3, no change inmode between 3 m and 27 m was observed for the first peak at11 nm with a slight increase of the second mode. The geometricmean diameters (GMD) of the particle size distributions (8–540 nm) were 17 nm, 20 nm, 20 nm, 22 nm, 38 nm, and 48 nm atdistances of 3 m, 15 m, 18 m, 27 m, 280 m, and –90 m (upwind),respectively, indicating a shift to larger particle GMD as a functionof distance from the highway. The evolution of the sizedistributions as a function of distance from traffic sources wascomparable to near–highway field and modeling studies reportedby Zhu et al. (2002a) and Jacobson et al. (2005). Since dilution iseffective regardless of particle size, other processes such ascondensation or evaporation of smaller particles must have playeda role in the evolution of the size distributions during transportaway from the roadway. Stroud et al. (2014) suggested that

evaporation of semi–volatile organics from particles and production of less–volatile organics progresses as traffic–related airmasses move away from the highway. Jacobson and Seinfeld(2004) and Jacobson et al. (2005) described a simulation usingBrownian coagulation enhanced by Van der Waals forces, fractalgeometry, and evaporation that could account for the sizeevolution of UFP measured near roadways. Jacobson et al. (2005)also proposed that partial evaporation of volatile materials onparticles may enhance the size dependent decay due to increasedBrownian coagulation rates between larger particles and shrinkingparticles.

Figure 3. Average number size distributions of non–denuded particles(w/o TD) measured at distances of 3 m and 18 m from the highway on

September 1. Error bars represent the 95% confidence intervals.

3.3 Distance decay gradient of non–volatile ultrafine particles

The volatility of particles at different distances from thehighway was determined using a thermodenuder operating at250 °C. The size distributions of non–volatile particles measured atthe 27 m, 280 m, and the upwind locations are presented inTable 3 and Figure 4a. On a particle number basis, a lower non–volatile fraction (36%) in UFP was measured at the 27 m location ascompared to the non–volatile fraction (62%) at the 280 m location.The non–volatile fraction of background particles at the upwindhighway was found to be the highest (81%). This reduction inparticle number within the thermodenuder was presumably due to

Jeong et al. – Atmospheric Pollution Research (APR) 667

shrinkage of internally mixed particles, i.e., particles that all havethe same mixture of chemical species, below the minimumdetectable size and/or complete evaporation of externally mixedvolatile particles, i.e., particles that are each composed of onechemical species. The mixing state of these UFP could not bedetermined with the available instrumentation. It should be notedthat particle losses due to diffusion and thermophoresis in the TDwere corrected for before determining the size distributions ofnon–volatile particles.

Figure 4. Average number size distributions of non–volatile particles (a)and non–volatile particles corrected by the upwind concentration (b) onAugust 19, 26, and 27. TD operated at 250 °C. Error bars represent the

95% confidence intervals.

It is interesting to note that the average size of the non–volatile particles became smaller with increasing distance from thehighway, in contrast to the apparent growth observed for the non–denuded particles (Figure 4a). An evolution in the average sizedistributions of the non–volatile particles at 27 m, 280 m, and –90 m locations was observed with calculated geometric meandiameters of 35 nm, 30 nm, and 28 nm, respectively. The non–volatile particle size distributions were marked by modes at 20–30 nm and 40–60 nm at the 27 m location. As shown in Figure 4a, amode around 22 nm was observed both upwind and downwind ofthe roadway, suggesting the presence of background 22 nmparticles, possibly consisting of non–volatile cores of originallylarger particles. These non–volatile background particles couldhave originated from combustion sources and been transportedover large distances and time scales. Engler et al. (2007) foundnon–volatile particles smaller than 30 nm at a rural backgroundarea and proposed the presence of thermally stable organicresidues in this size range.

The decay gradient of non–volatile particles may be a goodindicator to measure how particle number concentrations areaffected by dilution. In order to estimate particle losses by dilution,a simple dilution equation was derived by using the non–volatileparticles as a dilution indicator as proposed by Zhang et al. (2004)who used carbon monoxide as a dilution indicator. The dilutionfunction between the two locations can be described as follows(N27–NB)/(N280–NB), where N27 and N280 are the non–volatileparticle number concentrations at the 27 m and the 280 mlocations, respectively, and NB is the background concentrationmeasured at the upwind location. Note that this equation wassimply derived based on the assumption that the numberconcentrations of non–volatile particles only decrease due todilution; dry deposition and coagulation are considered as minorremoval processes, and the number concentrations beyond theFMPS size limit are negligible. Figure 4b shows the background–corrected size distributions of non–volatile UFP. From the equationabove, the dilution ratio between 27 m and 280 m was found to beapproximately 3. For the non–denuded particles, the background–corrected decay ratio of the particle number at 27 m to theconcentration at 280 m was approximately 5 as discussed earlier,indicating that atmospheric dilution may be responsible forapproximately 60% of the reduction in UFP concentration awayfrom the highway. This implies that other removal processes suchas evaporation may also play a substantial role in decreasing thenumber concentration of UFP even beyond 27 m from a highway.

A direct comparison between the non–denuded and non–volatile UFP at the upwind location shown in Figure S4 (see theSM) presents a clear shrinkage of particles likely due to a partialremoval of organic coatings on internally mixed particles duringthe denuding process. Although the determination of the volatilityof an individual particle was not possible with the TD–FMPS systemin our study, the shift of the size distribution without substantialchange in number concentration could be explained by shrinkageof internally mixed particles. In contrast, the dramatic reduction inthe number of <30 nm particles at 27 m could be explained interms of either an external or internal mixture of particles.Regardless, it is clear that these <30 nm particles near the highwaywere composed primarily of volatile species. On the other hand,the non–volatile particle measurement near the highway (e.g., at27 m) suggests that UFP in the size range from 40 nm to 60 nmmay be almost entirely composed of non–volatile material, such asBC.

Only a few mobile studies have been published to estimatethe total non–volatile fraction of traffic–related particles (Wehneret al., 2004; Pirjola et al., 2012). Table 4 summarizes the pastroadside field measurements using a TD–particle sizer system. It ischallenging to compare the non–volatile fractions of particlesamong the studies due to variations in location and the lack ofsufficient data. As shown, the spatial gradient in the non–volatilefraction and background observation were only available in ourstudy. The average non–volatile fraction at 27 m in our study washigher than the fractions reported by the two other studies. Thisdifference can be explained by the large variations in the locationsof monitoring sites among the studies (i.e., 50 m from a highwayfor Wehner et al. 2004, averaged fraction between 8 and 60 mfrom a roadway for Pirjola et al. 2012). A higher number of heavy–duty vehicles was observed in our study, suggesting the higherfraction of non–volatile particles could be due to a strongerinfluence of diesel vehicles in our study. The mode of the non–volatile particle size distribution was comparable to that found byWehner et al. 2004, which showed a mode at 50 nm.

The TD–FMPS system used in our study alone cannotdistinguish the non–volatile fraction originating from differentengine types that may result in the variations of the non–volatilefraction near roads. The identification of individual exhaust plumesfrom known vehicles with the TD–FMPS is required to quantify theparticle volatility of diesel and gasoline exhausts or newly

Jeong et al. – Atmospheric Pollution Research (APR) 668

developed engines (i.e., gasoline direct injection engine). Inaddition, the real–world spatial variations of non–volatile particleswithin a micro–scale distance from a major road (<15 m) should beinvestigated to improve the accuracy of the dispersion andexposure of non–volatile particles. In order to investigate healtheffects of exposure to traffic–related UFP, the location ofmonitoring sites (i.e., distance from roadways) is one of the crucialfactors and the size dependent variation of particle volatilityshould be considered.

3.4. Decay gradient in the mass concentration of other pollutants

In addition to the UFP number and size measurements, BC, p–PAH, PM0.6 and PM2.5 concentrations were measured at the upwindand downwind locations with the exception of the 15 m location(Table 2). One minute measurements of BC at 27 m ranged from0.2 to 5.6 μg m–3 during the measurement period. The concentrations of many pollutants at 280 m remained higher than thoseupwind of the highway. The background–corrected concentrationsof BC and p–PAH at the 280 m location decreased by factors of 2.7and 4.2, respectively, when compared to those at 27 m. The decayratio of BC was comparable to that for the number concentrationof non–volatile particles. However, more gradual decay gradientswere observed for PM0.6 and PM2.5 mass concentrations betweenthe 27 m and 280 m locations by factors of 2.0 and 1.6,respectively.

3.5. Single particle analysis of traffic–related fine particles

The ATOFMS was used to provide insights into the evolution inthe near roadway environment of fine particles emitted fromtraffic sources to slightly aged particles. Figure 5 illustrates theaverage positive and negative mass spectra of EC–containingparticle types, which are more likely to be associated with traffic–related emissions from the highway. Four out of the five particletypes observed at the 27 m location were dominated by a series of

elemental carbon (EC) fragments (e.g., m/z ±12, ±24, ±36, ±48) andare thus believed to be associated with vehicular traffic exhaustsources. These EC based particle types, corresponding to fresh EC(EC–fresh), EC mixed with organic carbon (OC) (EC–OC), calcium–containing EC (EC–Ca), and potassium and sodium–containing EC(EC–K–Na), were observed at both Site A and B. Figure 6 shows thecontribution of the four traffic–related particle types at the 27 mand 280 m locations. Other particle types detected are summarizedin the SM.

At Site A, the mass spectrum of the EC–fresh type wasdominated by distinct dual–polarity EC fragments with smallerpeaks at m/z 37 [C3H+] and 40 [Ca+]. The peak at m/z +28 can beattributed to silicon (Spencer and Prather, 2006). The EC–freshparticle type may thus represent fresh soot agglomerates fromdiesel engine exhaust.

The EC–OC particles appeared to be unburned fuels and/orlubricating oil from gasoline combustion. These particles containedOC fragment signals in their positive mass spectrum (e.g., m/z 27[C2H3+/HCN+], 37 [C3H+]) with EC fragments as well as signals atm/z39 [K+] and m/z 40 [Ca+]. Lubricating oil contains calcium as anadditive, while potassium is present in gasoline fuel, whichdifferentiates it from diesel fuel (Spencer et al., 2006). However,the m/z 39 peak is not only associated with K, but could be due toan organic fragment [C3H3+] (Silva and Prather, 2000). Strongsignals in the negative spectrum were associated with ECfragments and organic nitrogen (m/z –26 [CN–], –42 [CNO–]) as wellas inorganic nitrate (–46 [NO2–], –62 [NO3–]), indicating incompletecombustion. The negative spectrum of cluster EC–OC alsocontained peaks at m/z –25, –37, –49, –61, and –73. These CnH–

peaks have previously been attributed to fragmentation of PAHsfrom fuel samples in laboratory measurements (Silva and Prather,2000; Spencer et al., 2006). The presence of more OC fragments(i.e., m/z 27, m/z 37–39) with nitrate has been previously observedfor gasoline vehicles (Toner et al., 2008).

Table 3. Average number concentrations with 95% confidence intervals of non–volatile particles (with TD) measured at the downwind and upwindsites during the particle volatility study on August 19, 26, and 27

Distance from Highway 27 m Downwind (Site A) 280 m Downwind (Site B) –90 m Upwind (Site C)

Total PN (particle cm–3) 6 700±540 3 700±580 2 400±110PN8–20 nm (particle cm–3) 1 700±190 1 100±110 800±50PN20–30 nm (particle cm–3) 1 300±80 900±50 600±30PN30–50 nm (particle cm–3) 1 400±80 700±90 400±20PN50–100 nm (particle cm–3) 1 900±140 800±230 400±10PN100–540 nm (particle cm–3) 600±60 200±100 100±10

Table 4. Comparison of total non–volatile fraction of traffic–related particles using a TD–Particle sizer system in this study and in previous research

This Study Wehner et al. (2004) Pijola et al. (2012)

Instrument TD–FMPS TD–SMPS TD–ELPI a

TD temperature (°C) 250 280 265Size range (nm) 8–540 b 10–400 7–10 000Distance from a major road (m) 27–280 50 8–60Upwind c Yes No NoHeavy–duty traffic ( vehicles h–1) 260 60–70 170 d

Mode of non–volatile size distribution (nm) 50 50 n/aNon–volatile fraction e 36% at 27 m, 62% at 280 m 14% 20–25%

a Electrical Low Pressure Impactor (ELPI, Dekati), b adjusted FMPS size, c non–volatile particle measurement at the upwind of a roadway,d daily average, e non–volatile fraction is calculated as follows: (number concentration using a TD system divided by number concentrationwithout TD system)×100%

Jeong et al. – Atmospheric Pollution Research (APR) 669

Figure 5. Average mass spectra of EC–related particle types observed at 27 m (a, b, c, and d) and 280 m (e, f, g, and h) from the highway.EC–fresh (a and e), EC–OC (b and f), EC–Ca (c and g), EC–K–Na (d and h).

The EC–Ca particles are believed to originate from dieselvehicles. These particles were characterized by the most intensepeaks at m/z 40 [Ca+] and 23 [Na+] in their positive ion spectra,while a series of EC fragment peaks and organic nitrogen weredominant in their negative spectra. Similar to other traffic–relatedparticle types, this class contained PAH fragment ions. Sodium andcalcium are present in conventional diesel fuels and lubricating oils(Spencer et al., 2006; Toner et al., 2006). Ca–rich EC clusters havebeen commonly identified as particles from traffic sources indynamometer and field experiments (e.g., Spencer et al., 2006;Toner et al., 2006; Dall’Osto and Harrison, 2012). EC–Ca particlesshowed a strong degree of similarity to particle types from bothdiesel lubricating oils and diesel exhaust (Spencer et al., 2006). It isworth noting that the traffic–related particle type contained a lowfraction of sulphate [HSO4–] compared to nitrate, likely due to theprevalence of low sulfur fuels in Canada.

EC–K–Na, the most abundant type (35%) at the 27 m location,likely originated from road dusts (road wear and dust suspension)and/or non–tailpipe emissions (i.e., brake pad wear) internally

mixed with soot from vehicular emissions. The particle type wascharacterized by EC fragments in the negative spectrum andintense dust–like signals (e.g., Al, K, Na, Fe, Mn, and Ba) in thepositive mass spectrum. Positive ion peaks at m/z 54 and 56 couldbe assigned to iron, and their relative intensities are consistentwith the expected isotopic distribution of iron. A peak at m/z –88was likely due to iron oxide [FeO–] (Figure 5d). The presence of ironcould also be due to brake wear and soil dust. Silva et al. (2000)also found iron oxide and oxygen ions, m/z –16 [O–], and intensepeaks for nitrate (m/z –46 and –62) in dust particle samples.

The similarity of the particle types observed at the 27 m andthe 280 m locations was investigated by examining the uncenteredPearson correlations (r) between the average mass spectral signalsobserved at each distance from the highway. EC–fresh was clearlyidentified by distinct EC fragments at both locations, whereas theEC–OC, EC–Ca and EC–K–Na particles contained relatively low ECsignals at the 280 m location. In fact, the EC–OC type had thelowest correlation between sites (r=0.68) of all four particle typesinvestigated. More oxidized OC fragments (m/z 43 [C2H3O+]) and

Jeong et al. – Atmospheric Pollution Research (APR) 670

aromatic hydrocarbons (m/z 39 [C3H3+], 51 [C4H3+], 63 [C5H3+]) wereobserved in the positive mass spectrum of the EC–OC particle typeat the 280 m location. This could be due to the influences ofhydrocarbons and background organics coated on the traffic–related particles with increased distance from the highway. A muchsmaller difference in the positive mass spectral signatures of EC–Cawas observed between the two locations (r=0.88), although lowersignals of EC fragments in the negative spectrum were observed atthe 280 location (Figure 5g). The major difference of EC–K–Nabetween 27 m and 280 m was the relative contribution of thisparticle type by number: 35% at 27 m and 6% at 280 m. In addition,the average mass spectra of the particle types were somewhatdifferent between sites, showing a very low iron signal (m/z 54 and56) and m/z 55 [Mn+] at the 280 m location. The decrease in theEC–K–Na relative number contribution may be caused bygravitational sedimentation of larger EC/road dust internalmixtures with increasing distance from the highway.

Figure 6. Contribution of particle types obtained using single particlemass spectrum data measured at the 27 m and 280 m locations.

Overall, traffic–related particle types (i.e., EC–fresh, EC–OC,EC–Ca, EC–K–Na) accounted for approximately 75% and 33% of theATOFMS particles detected at the 27 m and 280 m locations,respectively. It is worth noting that the decay ratio of traffic–related particle types was comparable to those of BC and PM0.6mass concentrations between the 27 m and 280 m location asdiscussed in Section 3.4. These findings imply that particlecomposition, and specifically the various mixing states of EC maybe important in context of exposure and health effects attributableto traffic–related particles as a function of distance from a majorroadway.

4. Summary and Conclusions

The number and size distributions of ultrafine particles weremeasured using an FMPS (with and without TD) at distances of 3 mto 280 m from a busy highway. In addition, single particle massspectra of traffic–related particles were measured to determinethe spatial evolution in the mixing state of submicron particles.Traffic–related (background corrected) particle number concentrations at 15 m and 27 m were 10 and 5 times higher than theconcentration at 280 m downwind of the highway, respectively. Aneven steeper spatial gradient was observed for nucleation modeparticles (<30 nm).

The non–volatile fraction of UFP after heating at 250 °Cincreased with distance from the highway, 36% at 27 m and 62% at280 m vs. 81% at –90 m (upwind) by number. The lower particlevolatility at the upwind site also indicated that the vehicleemissions were influencing particle composition, even at a distanceof 280 m. A more gradual decay gradient was observed for non–

volatile particles: the concentration at 27 m was higher than thatat 280 m by a factor of 3 as compared to the factor of 5 for non–denuded UFP. The more gradual gradient in number concentrationfor non–volatile traffic–related UFP indicates that removalprocesses such as evaporation can contribute to near roadgradients in UFP concentrations over distances of several hundredmeters. A shift in size distribution to smaller GMD with distancefrom the highway was observed for non–volatile particles incontrast to a growth in average diameter for non–denuded UFP.The background corrected number size distribution of non–volatileparticles had a mode peaking at around 50 nm, representingprimary traffic–related particles (i.e., soot). The findings in thisstudy indicate that people living or spending time near a majorroadway can be exposed to elevated number concentrations ofvolatile UFP (<30 nm) and non–volatile UFP (30–100 nm).

Elemental carbon containing particle types also predominatedin the larger sub–micron particles detected by single particle massspectrometry: EC–fresh, EC–OC, EC–Ca, and EC–K–Na particleswere identified. The contributions of these particle types variedwith distance from the highway; the EC–related particle typesaccounted for approximately 75% and 33% of the particlesdetected by the ATOFMS at the 27 m and 280 m locations,respectively.

The variability of the decay gradients of TRAPs and thecontributions of traffic–related particle types indicated that notonly the number concentration, but the composition, volatility, andmixing state of the particles can change substantially with distancefrom the roadway. These gradients were not reflected by measurements of PM2.5 alone. Furthermore, the influence of the highwaywas still evident up to 280 m downwind. Thus a comprehensiveanalysis of physical and chemical properties is required for betterestimates of the exposure of traffic–related pollutants in thevicinity of roadways. This exposure may potentially impact thehealth of the 10 million Canadians who live within 250 m of amajor roadway.

Acknowledgements

This work was supported by Environment Canada, CanadaFoundation for Innovation, the Ontario Innovation Trust, and theOntario Research Fund.

Supporting Material Available

Clustering analysis of single particle data (Text), A wind roseplot of 1 minute average number concentration of ultrafineparticles measured on August 19, 26 and 27 (Figure S1), Ultrafineparticle number concentrations as a function of distance from thehighway over the period of August 19, 26, and 27 (Figure S2),Average number size distributions of non–denuded particles (w/oTD) corrected by the upwind concentration on August 19, 26, and27 (Figure S3), Average size distributions of non–denuded (w/o TD)particles (AA) and non–volatile particles (TD) at the downwind(27 m, 280 m) and upwind (–90 m) locations (Figure S4), Averagemass spectra of the K–rich Biomass particle type observed at the27 m (a) and at 280 m (b) locations (Figure S5), Average massspectra of particle types observed at the 280 m location. (a) Dust–Fe, (b) Dust–Mn, (c) Dust–Na, and (d) Phosphate (Figure S6). Thisinformation is available free of charge via the Internet athttp://www.atmospolres.com.

References

Baron, P.A., Willeke K., 2001. Aerosol Measurement; Principles, Techniques,and Application. 2nd Edition, John Wiley & Sons Inc., New York.

Beckerman, B., Jerrett, M., Brook, J.R., Verma, D.K., Arain, M.A., Finkelstein,M.M., 2008. Correlation of nitrogen dioxide with other trafficpollutants near a major expressway. Atmospheric Environment 42,275–290.

Jeong et al. – Atmospheric Pollution Research (APR) 671

Biswas, S., Ntziachristos, L., Moore, K.F., Sioutas, C., 2007. Particle volatilityin the vicinity of a freeway with heavy–duty diesel traffic. AtmosphericEnvironment 41, 3479–3493.

Bukowiecki, N., Dommen, J., Prevot, A.S.H., Richter, R., Weingartner, E.,Baltensperger, U., 2002. A mobile pollutant measurement laboratory–measuring gas phase and aerosol ambient concentrations with highspatial and temporal resolution. Atmospheric Environment 36, 5569–5579.

Dall'Osto, M., Harrison, R.M., 2012. Urban organic aerosols measured bysingle particle mass spectrometry in the megacity of London.Atmospheric Chemistry and Physics 12, 4127–4142.

Durant, J.L., Ash, C.A., Wood, E.C., Herndon, S.C., Jayne, J.T., Knighton,W.B., Canagaratna, M.R., Trull, J.B., Brugge, D., Zamore, W., Kolb, C.E.,2010. Short–term variation in near–highway air pollutant gradients ona winter morning. Atmospheric Chemistry and Physics 10, 8341–8352.

Engler, C., Rose, D., Wehner, B., Wiedensohler, A., Bruggemann, E., Gnauk,T., Spindler, G., Tuch, T., Birmili, W., 2007. Size distributions of non–volatile particle residuals (Dp<800 nm) at a rural site in Germany andrelation to air mass origin. Atmospheric Chemistry and Physics 7, 5785–5802.

Evans, G.J., Jeong, C.–H., Sabaliauskas, K., Jadidian, P., Aldersley, S.,Larocque, H., Herod, D., 2011. Design of a near–road MonitoringStrategy for Canada, Contract Report, Environment Canada.

Evans, G.J., Peers, A., Sabaliauskas, K., 2008. Particle dose estimation fromfrying in residential settings. Indoor Air 18, 499–510.

Gauderman, W.J., Vora, H., McConnell, R., Berhane, K., Gilliland, F.,Thomas, D., Lurmann, F., Avol, E., Kunzli, N., Jerrett, M., Peters, J.,2007. Effect of exposure to traffic on lung development from 10 to 18years of age: A cohort study. Lancet 369, 571–577.

Gordon, M., Staebler, R.M., Liggio, J., Li, S.M., Wentzell, J., Lu, G., Lee, P.,Brook, J.R., 2012. Measured and modeled variation in pollutantconcentration near roadways. Atmospheric Environment 57, 138–145.

Hagler, G.S.W., Thoma, E.D., Baldauf, R.W., 2010. High–resolution mobilemonitoring of carbon monoxide and ultrafine particle concentrations ina near–road environment. Journal of the Air & Waste ManagementAssociation 60, 328–336.

Hagler, G.S.W., Baldauf, R.W., Thoma, E.D., Long, T.R., Snow, R.F., Kinsey,J.S., Oudejans, L., Gullett, B.K., 2009. Ultrafine particles near a majorroadway in Raleigh, North Carolina: Downwind attenuation andcorrelation with traffic–related pollutants. Atmospheric Environment43, 1229–1234.

Healy, R.M., Sciare, J., Poulain, L., Kamili, K., Merkel, M., Muller, T.,Wiedensohler, A., Eckhardt, S., Stohl, A., Sarda–Esteve, R.,McGillicuddy, E., O'Connor, I.P., Sodeau, J.R., Wenger, J.C., 2012.Sources and mixing state of size–resolved elemental carbon particles ina European megacity: Paris. Atmospheric Chemistry and Physics 12,1681–1700.

HEI (Health Effects Institute), 2010. Panel on the Health Effects of Traffic–Related Air Pollution, Traffic–Related Air Pollution: A Critical Review ofthe Literature on Emissions, Exposure, and Health Effects, HEI SpecialReport 17, Health Effects Institute, Boston, MA.

Hitchins, J., Morawska, L., Wolff, R., Gilbert, D., 2000. Concentrations ofsubmicrometre particles from vehicle emissions near a major road.Atmospheric Environment 34, 51–59.

Jacobson, M.Z., Kittelson, D.B., Watts, W.F., 2005. Enhanced coagulationdue to evaporation and its effect on nanoparticle evolution.Environmental Science & Technology 39, 9486–9492.

Jacobson, M.Z., Seinfeld, J.H., 2004. Evolution of nanoparticle size andmixing state near the point of emission. Atmospheric Environment 38,1839–1850.

Jeong, C.H., Evans, G.J., 2009. Inter–comparison of a fast mobility particlesizer and a scanning mobility particle sizer incorporating an ultrafinewater–based condensation particle counter. Aerosol Science andTechnology 43, 364–373.

Jeong, C.H., McGuire, M.L., Godri, K.J., Slowik, J.G., Rehbein, P.J.G., Evans,G.J., 2011. Quantification of aerosol chemical composition usingcontinuous single particle measurements. Atmospheric Chemistry andPhysics 11, 7027–7044.

Jerrett, M., Finkelstein, M.M., Brook, J.R., Arain, M.A., Kanaroglou, P., Stieb,D.M., Gilbert, N.L., Verma, D., Finkelstein, N., Chapman, K.R., Sears,M.R., 2009. A cohort study of traffic–related air pollution and mortalityin Toronto, Ontario, Canada. Environmental Health Perspectives 117,772–777.

Kaur, S., Nieuwenhuijsen, M., Colvile, R., 2005. Personal exposure of streetcanyon intersection users to PM2.5, ultrafine particle counts and carbonmonoxide in Central. London, UK. Atmospheric Environment 39, 3629–3641.

Kittelson, D.B., 1998. Engines and nanoparticles: A review. Journal ofAerosol Science 29, 575–588.

Kittelson, D.B., Watts, W.F., Johnson, J.P., 2004. Nanoparticle emissions onMinnesota Highways. Atmospheric Environment 38, 9–19.

Liggio, J., Gordon, M., Smallwood, G., Li, S.M., Stroud, C., Staebler, R., Lu,G., Lee, P., Taylor, B., Brook, J.R., 2012. Are emissions of black carbonfrom gasoline vehicles underestimated? Insights from near and on–road measurements. Environmental Science & Technology 46, 4819–4828.

Lin, J.S., Tsai, C.J., 2003. Thermophoretic deposition efficiency in acylindrical tube taking into account developing flow at the entranceregion. Journal of Aerosol Science 34, 569–583.

Pirjola, L., Lahde, T., Niemi, J.V., Kousa, A., Ronkko, T., Karjalainen, P.,Keskinen, J., Frey, A., Hillamo, R., 2012. Spatial and temporalcharacterization of traffic emissions in urban microenvironments witha mobile laboratory. Atmospheric Environment 63, 156–167.

Pirjola, L., Parviainen, H., Hussein, T., Valli, A., Hameri, K., Aaalto, P.,Virtanen, A., Keskinen, J., Pakkanen, T.A., Makela, T., Hillamo, R.E.,2004. "Sniffer"–A novel tool for chasing vehicles and measuring trafficpollutants. Atmospheric Environment 38, 3625–3635.

Rehbein, P.J.G., Jeong, C.H., McGuire, M.L., Evans, G.J., 2012. Strategies toenhance the interpretation of single–particle ambient aerosol data.Aerosol Science and Technology 46, 584–595.

Riediker, M., Williams, R., Devlin, R., Griggs, T., Bromberg, P., 2003.Exposure to particulate matter, volatile organic compounds, and otherair pollutants inside patrol cars. Environmental Science & Technology37, 2084–2093.

Roorda–Knape, M.C., Janssen, N.A.H., De Hartog, J.J., Van Vliet, P.H.N.,Harssema, H., Brunekreef, B., 1998. Air pollution from traffic in citydistricts near major motorways. Atmospheric Environment 32, 1921–1930.

Sakurai, H., Tobias, H.J., Park, K., Zarling, D., Docherty, K.S., Kittelson, D.B.,McMurry, P.H., Ziemann, P.J., 2003. On–line measurements of dieselnanoparticle composition and volatility. Atmospheric Environment 37,1199–1210.

Salimi, F., Mazaheri, M., Clifford, S., Crilley, L.R., Laiman, R., Morawska, L.,2013. Spatial variation of particle number concentration in schoolmicroscale environments and its impact on exposure assessment.Environmental Science & Technology 47, 5251–5258.

Silva, P.J., Prather, K.A., 2000. Interpretation of mass spectra from organiccompounds in aerosol time–of–flight mass spectrometry. AnalyticalChemistry 72, 3553–3562.

Silva, P.J., Carlin, R.A., Prather, K.A., 2000. Single particle analysis ofsuspended soil dust from Southern California. AtmosphericEnvironment 34, 1811–1820.

Song, X.H., Hopke, P.K., Fergenson, D.P., Prather, K.A., 1999. Classificationof single particles analyzed by ATOFMS using an artificial neuralnetwork, ART–2A. Analytical Chemistry 71, 860–865.

Spencer, M.T., Prather, K.A., 2006. Using ATOFMS to determine OC/ECmass fractions in particles. Aerosol Science and Technology 40, 585–594.

Jeong et al. – Atmospheric Pollution Research (APR) 672

Spencer, M.T., Shields, L.G., Sodeman, D.A., Toner, S.M., Prather, K.A.,2006. Comparison of oil and fuel particle chemical signatures withparticle emissions from heavy and light duty vehicles. AtmosphericEnvironment 40, 5224–5235.

Straif, K., Cohen, A., Samet, J., 2013. Air Pollution and Cancer. IARCScientific Publication No. 162, IARC Press.

Stroud, C.A., Liggio, J., Zhang, J., Gordon, M., Staebler, R.M., Makar, P.A.,Zhang, J.H., Li, S.M., Mihele, C., Lu, G., Wang, D.K., Wentzell, J., Brook,J.R., Evans, G.J., 2014. Rapid organic aerosol formation downwind of ahighway: Measured and model results from the FEVER study. Journal ofGeophysical Research–Atmospheres 119, 1663–1679.

Tiitta, P., Miettinen, P., Vaattovaara, P., Joutsensaari, J., Petaja, T., Virtanen,A., Raatikainen, T., Aalto, P., Portin, H., Romakkaniemi, S., Kokkola, H.,Lehtinen, K.E.J., Kulmala, M., Laaksonen, A., 2010. Roadside aerosolstudy using hygroscopic, organic and volatility TDMAs: Characterizationand mixing state. Atmospheric Environment 44, 976–986.

Toner, S.M., Shields, L.G., Sodeman, D.A., Prather, K.A., 2008. Using massspectral source signatures to apportion exhaust particles from gasolineand diesel powered vehicles in a freeway study using UF–ATOFMS.Atmospheric Environment 42, 568–581.

Toner, S.M., Sodeman, D.A., Prather, K.A., 2006. Single particlecharacterization of ultra ne and accumulation mode particles fromheavy duty diesel vehicles using aerosol time–of–flight massspectrometry. Environmental Science & Technology 40, 3912–3921.

Wehner, B., Philippin, S., Wiedensohler, A., Scheer, V., Vogt, R., 2004.Variability of non–volatile fractions of atmospheric aerosol particleswith traffic influence. Atmospheric Environment 38, 6081–6090.

Weimer, S., Mohr, C., Richter, R., Keller, J., Mohr, M., Prevot, A.S.H.,Baltensperger, U., 2009. Mobile measurements of aerosol number andvolume size distributions in an Alpine Valley: Influence of traffic versuswood burning. Atmospheric Environment 43, 624–630.

Westerdahl, D., Fruin, S., Sax, T., Fine, P.M., Sioutas, C., 2005. Mobileplatform measurements of ultrafine particles and associated pollutantconcentrations on freeways and residential streets in Los Angeles.Atmospheric Environment 39, 3597–3610.

Zhang, K.M., Wexler, A.S., Zhu, Y.F., Hinds, W.C., Sioutas, C., 2004. Evolutionof particle number distribution near roadways. Part II: The 'road–to–ambient' process. Atmospheric Environment 38, 6655–6665.

Zhu, Y.F., Hinds, W.C., Kim, S., Sioutas, C., 2002a. Concentration and sizedistributions of ultra ne particles near a major highway. Journal of theAir & Waste Management Association 52, 1032–1042.

Zhu, Y.F., Hinds, W.C., Kim, S., Shen, S., Sioutas, C., 2002b. Study of ultrafineparticles near a major highway with heavy duty diesel traffic.Atmospheric Environment 36, 4323–4335.

Zimmerman, N., Jeong, C.H., Wang, J.M., Ramos, M., Wallace, J.S., Evans,G.J., 2015. A source–independent empirical correction procedure forthe fast mobility and engine exhaust particle sizers. AtmosphericEnvironment 100, 178–184.

Zimmerman, N., Pollitt, K.J.G., Jeong, C.H., Wang, J.M., Jung, T., Cooper,J.M., Wallace, J.S., Evans, G.J., 2014. Comparison of three nanoparticlesizing instruments: The influence of particle morphology. AtmosphericEnvironment 86, 140–147.