Journal of Exposure Analysis and Environmental Epidemiology (2003) 13, 364–371
r 2003 Nature Publishing Group All rights reserved 1053-4245/03/$25.00
www.nature.com/jea
Influence of traffic patterns on particulate matter and polycyclic aromatic
hydrocarbon concentrations in Roxbury, Massachusetts
JONATHAN I. LEVY, DEBORAH H. BENNETT, STEVEN J. MELLY, AND JOHN D. SPENGLER
Department of Environmental Health, Harvard School of Public Health, Boston, MA, USA
Vehicle emissions have been associated with adverse health effects in multiple epidemiological studies, but the sources or constituents responsible have not
been established. Characterization of vehicle-related exposures requires detailed information on spatial and temporal trends of various pollutants and the
ability to predict exposures in unmonitored settings. To address these issues, in the summer of 2001 we measured continuously particle-bound polycyclic
aromatic hydrocarbons (PAHs), ultrafine particles, and PM2.5 at a number of sites in Roxbury, a neighborhood of Boston, Massachusetts with significant
diesel and gasoline-fueled traffic. We took measurements at the side of the road and at varying distances from the road, with simultaneous collection of
traffic counts and meteorological conditions. Across all nine sites, median roadside concentrations were 8 ng/m3 of particle-bound PAHs (range: 4–57),
16,000 ultrafine particles/cm3 (range: 11,000–53,000), and 54 mg/m3 of PM2.5 as measured with a DustTrak (range: 12–86). Concentrations of all
pollutants were lower at greater distances from the road, upwind, and at higher wind speeds, with greater concentration gradients for PAHs and ultrafine
particles. In linear mixed effects regression models accounting for temporal autocorrelation, large diesel vehicle counts were significantly associated
with roadside concentrations of PAHs (P ¼ 0.02), with a moderate association with ultrafine particles and little relation with PM2.5. Although
more comprehensive information would be needed for epidemiological applications, these data demonstrate significant spatial and temporal heterogeneity for traffic-related pollutants during the summer in an urban center, with our monitoring and analytical methodology helping to inform source
attribution.
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13, 364–371. doi:10.1038/sj.jea.7500289
Keywords: Diesel, geographic information systems, polycyclic aromatic hydrocarbons, PM2.5, traffic, ultrafine particles.
Introduction
There is growing evidence of the effect of traffic-related air
pollution on respiratory and cardiovascular health, drawn
from two primary types of studies. One set of studies relies on
central site monitors for specific pollutants and uses statistical
techniques to infer a relation with traffic rather than other
pollution sources (Laden et al., 2000; Tolbert et al., 2000; Yu
et al., 2000; Janssen et al., 2002). Although these studies
yield interpretable estimates of population-wide effects, they
provide no ability to discern highly exposed subpopulations
or geographic clustering of impacts potentially associated
with pollution ‘‘hot spots’’.
Other studies have evaluated the relation between proximity to a major road or high traffic volume and a variety of
health outcomes, including increased respiratory symptoms
or reduced lung function (Wjst et al., 1993; Weiland et al.,
1994; Oosterlee et al., 1996; Brunekreef et al., 1997). Similar
investigations have used Geographic Information System
(GIS) methods to evaluate the relation between proximity to
1. Address all correspondence to: Dr. J.I. Levy, Department
of Environmental Health, Harvard School of Public Health, Landmark
Center Room 404K, PO Box 15677, Boston, MA 02215, USA. Tel:
+1-617-384-8808. Fax: +1-617-384-8859. E-mail: [email protected]
Received 5 November 2002; accepted 25 April 2003
a major road and medical visits for asthma (English et al.,
1999) or risk of wheeze (Venn et al., 2001). While these
studies are valuable, they provide limited insight about the
precise pollutants or sources influencing respiratory health,
making it difficult to determine the relative contributions of
pollution and other risk factors. Those studies that have
evaluated different types of traffic and have made simultaneous measurements of air pollution have found a stronger
relation with truck traffic and related pollutants (van Vliet
et al., 1997).
For traffic-related pollutants to influence health as a
function of proximity to the road, their concentrations must
have sufficient spatial heterogeneity. While outdoor concentrations of some pollutants that are principally related
to regional transport (such as PM2.5 or ozone) would be
unlikely to vary greatly within an urban center, concentrations of a subset of traffic-related primary pollutants
have been shown to vary substantially over small geographic
areas.
For example, in a pilot study in Harlem, New York,
elemental carbon levels ranged by a factor of four across sites
in close proximity to one another, while PM2.5 levels were
quite similar (Kinney et al., 2000). Similarly, polycyclic
aromatic hydrocarbon (PAH) concentrations in an urban
center differed by a factor of three between measurements on
Levy et al.
Traffic patterns and concentrations of PM and PAHs
a street and in a park, with traffic contributing an estimated
80% of ambient concentrations (Nielsen, 1996). Urban
PAH concentrations have also been linked with traffic
volume, with minimal concentrations during low traffic and
higher concentrations during rush hour (Levy et al., 2001).
Ultrafine particle concentrations have been strongly correlated with traffic patterns, with particle number concentrations a factor of five greater at peak traffic periods than
at night (Ruuskanen et al., 2001) and a factor of seven
greater within 15 m of the road compared with average
urban exposure levels (Hitchins et al., 2000). Particle number
concentrations have also been related to wind speed and
direction, with one study demonstrating decreasing concentrations at higher wind speeds (Zhu et al., 2002b) and
a second finding an inverted U-shaped curve with lower
concentrations at both high and low wind speeds (Zhu et al.,
2002a).
While these studies provide evidence of spatial heterogeneity for selected pollutants in urban areas, it is important
to understand which vehicle types and site characteristics
most influence concentrations, to help inform policy decisions
and to allow for the extrapolation of findings to other
settings. Studies that have developed models for trafficrelated pollutants have generally been able to predict
concentrations reasonably as a function of selected site
characteristics. For example, nitrogen dioxide concentrations
in three cities in Europe were estimated using traffic volume,
proximity to roadway, road type, land use type, and altitude
(R2 between 0.79 and 0.87), with the parameters used
depending on available information at the site (Briggs et al.,
1997). Similarly, long-term PM2.5 concentrations in three
European settings were predicted using traffic intensity
at various distances, distance to road, road configuration,
road type, or population density (R2 between 0.5 and 0.73)
(Brauer et al., 2003).
However, both of these studies relied on integrated
pollution samples and provided only coarse indicators of
traffic volume (with no temporal resolution or distinction
among vehicle types). To better understand the influence of
nearby traffic on ambient concentrations, as well as to
capture short-term peaks potentially relevant for some health
outcomes, there is a need to develop predictive regression
equations using continuous pollution and traffic data.
In this study, we took continuous measurements of traffic
counts and concentrations of multiple traffic-related air
pollutants (PM2.5, ultrafine particles, and particle-bound
PAHs) on a variety of roads in an urban section of Boston,
Massachusetts. We took measurements at the roadside and at
varying distances from the road and collected information on
meteorological conditions and traffic characteristics that
could influence pollutant patterns. We present regressions
to predict concentrations of all pollutants and conclude by
discussing the implications for future exposure assessment or
epidemiological studies.
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
Methods
In July and August of 2001, we measured PM2.5, ultrafine
particles, particle-bound PAHs, and traffic counts at a
number of sites in Roxbury, Massachusetts (a neighborhood
of Boston). Roxbury is a neighborhood that contains
multiple bus routes as well as bus and truck depots,
indicating the likelihood of significant diesel traffic and
related emissions, and spatial pollution patterns have been
analyzed there in the past (Levy et al., 2001).
To ensure that we captured a range of traffic densities, we
used GIS software (ArcGIS 8.1, ESRI, Redlands, CA, USA)
to estimate traffic density scores in the study area. Traffic
count and roadway classification data from the Massachusetts Highway Department (Massachusetts Highway Department, 2001) were used to estimate average daily traffic for
roadways. As actual traffic count data were not available
for most of the roads in the study area, many estimates
were based on average values for roads of the same functional class. To develop traffic indicators that more closely
corresponded to emissions, we adjusted traffic estimates to
account for Massachusetts Bay Transportation Authority
bus routes, assuming one bus was equivalent to 42 cars based
on total PM emissions from average US public buses versus
cars meeting 1990 emissions standards (US Environmental
Protection Agency, 1985, 1989).
The ArcGIS spatial analyst extension was used to create a
grid of 50 m cells with a traffic density score assigned to each
cell. A key parameter used by the software is the radius from
the center of each cell used to calculate the traffic density. In
order to investigate the impact of choosing different radii
(which would correspond to the radii of significant impact for
vehicles on a road), we calculated traffic densities with radii
of 50, 100, 200, and 300 m. These radii captured the range of
theoretically plausible behaviors of air pollution patterns
within this domain. The 100 m radius was selected for the
baseline traffic density scores, based on the qualitative
behaviors of the indicators with different radii (the 100 m
radius gave continuous values along major thoroughfares
while displaying differences between major and minor
streets). Quartiles of traffic density scores were calculated
and used to assign values 1–4 to each cell. Each proposed
sampling area corresponded to four cells and was assigned a
traffic density score corresponding to the sum of the four
cells. This allowed us to select sites with a range of traffic
densities, providing greater statistical power for our regression analyses.
We used identical instrumentation as in our recent
exposure assessment studies in Boston (Levy et al., 2001,
2002), to ensure comparability of findings. We measured
PM2.5 concentrations using a DustTrak Model 8520 (TSI,
Minneapolis, MN, USA), a laser photometer fitted with an
impactor to exclude larger particles. It should be noted that
past studies have found that the DustTrak overstates ambient
365
Levy et al.
concentrations when not calibrated to the specific ambient
particles being measured, but that the bias is a consistent
multiplicative factor. For example, one study (Chang et al.,
2001) determined that PM2.5 concentrations from a DustTrak were a factor of 2.1 greater than concentrations from an
integrated personal exposure monitor during the summer,
with an R2 of 0.87. Similarly, another study (Levy et al.,
2002) found a factor of 2.8 difference between DustTrak and
TEOM measurements, with an R2 of 0.93. Although
humidity and other factors will influence this relation, these
variations are minimized by sampling in a defined geographic
region in a single season. Furthermore, a focus on PM2.5
rather than PM10 is likely to improve the relation between
our measures and gravimetric concentrations (Quintana et al.,
2000; Rea et al., 2001). Thus, our regression models and
evaluations of spatial heterogeneity based on DustTrak data
will be interpretable.
For ultrafine particle counts, we used a TSI P-Trak Model
8525, a condensation nuclei particle counter that condenses
an isopropanol vapor on smaller particles, allowing them to
grow to a size where they could be counted. Total particlebound PAH concentrations were measured with a PAS
2000CE (EcoChem Analytics, League City, TX, USA), a
photoelectric aerosol sensor that measures PAHs adsorbed
onto particles by using irradiation energy and lamp
wavelength specific to PAHs. We used a TSI Q-Trak Model
8551 to measure carbon monoxide (CO), carbon dioxide
(CO2), temperature, and relative humidity. Neither CO nor
CO2 is addressed in this analysis.
In addition, we measured traffic counts with a Trax I
(JAMAR Technologies, Horsham, PA, USA), an automatic
traffic data recorder that registers all vehicles passing over
pneumatic tubing and places them into one of 13 vehicle
classes. For our regression analysis, we only considered total
vehicle counts and counts of buses plus trucks (Class 4–13,
using the Federal Highway Administration Vehicle Classification Scheme (US Department of Transportation, 2002).
The latter term was considered as a proxy for diesel vehicles
and is hereafter referred to as ‘‘large diesel’’. We only
collected traffic counts at those sites at which traffic flow was
anticipated to be heavier, and our analysis focuses on this
subset of data. Finally, to capture wind speed and direction
for our analysis, we used a portable Weather Monitor II
(Davis Instruments, Hayward, CA, USA).
At each sampling site, one set of instruments was placed on
the sidewalk adjacent to the road, hereafter referred to as the
stationary site. The second set of instruments was moved
each hour in a predefined sequence, including measurements
25, 50, and 75 m from the stationary site in a downwind
direction, co-located measurements, and measurements
across the street and 25 m away from the stationary site in
an upwind direction. The upwind or downwind direction
was determined at the start of the day based on prevailing
wind direction, but measured meteorological data were used
366
Traffic patterns and concentrations of PM and PAHs
to recategorize each 10 min period as downwind (within 451
of the direction of the moving monitor), upwind (within
451of the direction away from the moving monitor),
or neutral (all other wind directions). All measurements
were taken on weekdays between 9:30 am and 4:30 pm,
given availability of field staff during these hours. We
collected 1-min average data for all parameters, but base
our primary regression analyses on 10-min averages to
ensure data stability.
For our stationary site regression analyses, we predicted
10-min average concentrations of all measured pollutants as a
function of traffic counts as well as temperature and relative
humidity. As traffic counts might not adequately represent
the source factor for short-term measurements in an urban
area (since traffic jams would yield low traffic counts but high
emissions), we also explored GIS-based traffic density scores
and the percentage of traffic that was diesel-fueled as
independent variables. For the moving measurements, we
evaluated distance from the road and wind speed as
predictors of the ratio between moving and stationary
measurements. Owing to the anticipated temporal autocorrelation between sequential 10-min average measurements,
we use linear mixed effects models with an AR-1 autoregressive correlation structure in all regressions, as implemented in SAS.
Results
In total, in the 12 days of sampling in which we measured
both air pollution and traffic counts adjacent to the sampling
site, we collected data at nine different sites in 307 10-min
periods. Temperatures were generally warm (mean 301C,
standard deviation 41C) with moderate and stable relative
humidity (mean 44%, standard deviation 11%). The wind
was generally from the south-southwest with low wind
velocity (median 10-min average less than 1 m/s, with 85%
of observations less than 2 m/s).
Across all measurements, median concentrations at the
stationary sites were 8 ng/m3 of particle-bound PAHs,
16,000 ultrafine particles/cm3, and 54 mg/m3 of PM2.5 as
measured by the DustTrak (Table 1). Both PAH and
ultrafine PM had skewed distributions, with mean concentrations of 18 ng/m3 and 21,000 particles/cm3, respectively.
There were substantial variations in pollutant concentrations
across sites, with spatial patterns differing by pollutant
(Figure 1). Median PAH concentrations ranged from 4 to
57 ng/m3. Median ultrafine particle count concentrations
ranged from 11,000 to 53,000 particles/cm3 across the nine
sites, with concentrations well correlated spatially with PAH
concentrations (r ¼ 0.70). In contrast, for PM2.5, the spatial
patterns were not as well defined and the concentration range
was less substantial (between 37 and 58 mg/m3 with the
exception of two extreme sites, with concentrations essentially
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
Levy et al.
Traffic patterns and concentrations of PM and PAHs
Table 1. Distribution of 10-min average concentrations of particle-bound PAHs, ultrafine particle counts, and PM2.5 from stationary roadside monitors,
along with traffic counts for those sites.
Median
Mean
Standard deviation
Range of medians across sites
PAH
(ng/m3)
Ultrafine
PM (#/cm3)
PM2.5
(mg/m3)
Traffic counts
(vehicles/10 min period)
8.4
18
21
4–57
16,000
21,000
15,000
11,000–53,000
54
52
22
12–86
6
6.9
5.1
0.3–16
Figure 1. Map of concentration patterns at nine monitoring sites in Roxbury, Massachusetts. The shaded streets were included in an earlier exposure
assessment study (Levy et al., 2001), and the dot represents a large bus terminal.
uncorrelated with PAH or ultrafine concentrations across
sites). As indicated in Figure 1, the relative concentrations of
the three pollutants differed somewhat by site.
The influence of traffic patterns can be seen when
stratifying the sample by predefined traffic density scores.
Sites with traffic density scores above the median (100 m
radius) had median concentrations of 36 ng/m3 of particlebound PAHs, 26,000 ultrafine particles/cm3, and 46 mg/m3
of PM2.5, versus median concentrations of 5 ng/m3 of
particle-bound PAHs, 13,000 ultrafine particles/cm3, and
59 mg/m3 of PM2.5 for sites with lower traffic density. In
general, traffic density scores (across radii ranging from 50 to
300 m) did not predict PM2.5 concentrations, but were
positive significant predictors of PAH concentrations for all
radii and of ultrafine particle concentrations for radii of 100
and 200 m (P o 0.05).
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
We used our automated continuous traffic counts in
regression models to look at the influence of traffic on
roadside concentrations in more detail. Across sampling sites,
the median traffic counts ranged from 96 to 1150 vehicles per
hour (2–97 large diesel vehicles per hour). These traffic
counts were well correlated with the predefined traffic density
scores (correlation coefficient of 0.61 for a 100 m radius).
Since temperature and relative humidity were insignificant in
our initial regression models (likely due in part to the limited
range of values), we omit them from further consideration.
Total traffic counts were not significant predictors of any
stationary site pollutant concentrations (Table 2). In contrast,
large diesel vehicle counts were significant predictors of
particle-bound PAH concentrations (P ¼ 0.02). The relation
between large diesel vehicle counts and ultrafine particles
was positive but insignificant (P ¼ 0.12), and there was no
367
Levy et al.
Table 2. Summary of predictive regression equations for stationary site
concentrations as a function of traffic counts (10-min average measurements; p-values correspond to slope terms, with equations presented in
bold when Po0.05).
X-variables
PAH
Ultrafine PM
PM2.5
368
Total traffic
Large diesel
vehicles
Fraction large
diesel
14+0.048x
P=0.25
27,00055x
P=0.08
540.002x
P=0.40
15+0.56x
P=0.02
20,000+230x
P=0.12
530.008x
P=0.37
17+17x
P=0.45
19,000+35,000x
P=0.01
520.19x
P=0.81
PAH
5
4
3
Upwind/neutral
2
Downwind
1
0
0
20
40
60
Distance (m)
80
Ratio
(Moving /Stationary)
Ultrafine PM
1.5
1
Upwind/neutral
Downwind
0.5
0
0
20
40
60
Distance (m)
80
PM2.5
Ratio
(Moving /Stationary)
relation with PM2.5 concentrations. To address the fact that
traffic counts may be an imperfect marker of source
characteristics, we also tested regressions with fraction of
traffic categorized as large diesel vehicles as the independent
variable. This term was a significant predictor of ultrafine
particle counts (P ¼ 0.01), but was not significant for
particle-bound PAHs or PM2.5.
When we considered the moving monitor measurements,
the ratios between moving monitor and stationary monitor
concentrations generally declined as a function of distance
from the road and were greater at sites downwind of the road
than upwind, as would be expected given the influence of
traffic on concentrations (Figure 2). The magnitude of the
distance gradient is most apparent for particle-bound PAHs
and ultrafine particles, with a slight gradient for PM2.5.
Furthermore, the wind speed would be expected to influence
the magnitude of the moving monitor concentrations (as well
as the stationary site concentrations), with lower concentrations at higher wind speeds. Given sample size limitations, we
combined all distances and simply looked at the relation
between wind speed and pollutant concentrations. As
indicated in Figure 3, concentrations of all three pollutants
decreased with increasing wind speed in this illustrative
comparison.
In our regression models, we predicted the ratio between
moving and stationary monitor concentrations as a function
of distance from the road and wind direction (as a categorical
variable representing downwind or upwind/neutral), including an interaction term to reflect potential differences in
concentration gradients as a function of prevailing wind
direction. Since wind speed would theoretically influence
concentrations at all sites, and given limited statistical power,
we omitted this variable from our moving/stationary ratio
regressions.
The PAH ratio decreased significantly as a function of
distance from the roadway (P ¼ 0.03), with ratios that were
higher downwind (although the wind direction term was not
statistically significant) (Table 3). In contrast, there was no
statistically significant distance gradient for either ultrafine
Ratio
(Moving/Stationary)
Traffic patterns and concentrations of PM and PAHs
1.2
1
0.8
Upwind/neutral
0.6
Downwind
0.4
0.2
0
0
20
40
60
80
Distance (m)
Figure 2. Mean ratio between moving monitor and roadside pollutant
concentrations as a function of distance from the roadway and wind
direction. Error bars represent standard errors of the mean ratios.
PM or PM2.5, although both coefficients were negative as
anticipated. However, downwind concentrations were significantly higher than upwind/neutral concentrations for both
ultrafine PM and PM2.5, and the interaction terms indicated
that concentrations declined as a function of distance for
downwind sites but not upwind/neutral sites.
To determine whether our findings were related in part to
the short averaging times used, we also evaluated all
regressions using 1-h average concentrations (reducing the
sample size from 307 to 35, given incomplete data for some
hours). The general conclusions were nearly identical, with
the only differences being that stationary PM2.5 concentrations were significantly predicted by total traffic counts and
the ratio of moving and stationary monitor PM2.5 concentrations was not significantly predicted by any covariates.
Discussion
Our investigation found that spatial variations of particlebound PAHs and ultrafine particle counts in a limited
geographic area were generally more substantial than
variations in PM2.5 concentrations. This finding agrees with
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
Levy et al.
Traffic patterns and concentrations of PM and PAHs
PAH
200
Concentration (ng/m3)
180
160
140
120
100
80
60
40
20
0
0
1
2
3
4
5
6
7
8
9
Wind speed (m/s)
Ultrafine PM
Concentration (#/cm3)
60000
50000
40000
30000
20000
10000
0
0
1
2
3
4
5
6
7
8
9
8
9
Wind speed (m/s)
PM 2.5
160
Concentration (µg/m3)
140
120
100
80
60
40
20
0
0
1
2
3
4
5
6
7
Wind speed (m/s)
Figure 3. Relation between moving monitor pollutant concentrations
and wind speed, aggregated across all measurement sites.
past investigations and supports the notion that local
transportation sources contribute a greater proportion of
particle-bound PAHs or ultrafine particles than PM2.5.
Furthermore, this conclusion is supported by the physical
properties of ultrafine particles, where coagulation processes
lead to significant changes in particle number and particle size
distributions over a matter of meters (Zhu et al., 2002b).
However, in spite of the substantial variations in
concentrations, our regression equations often displayed
limited predictive power. This is due in part to sample size
limitations, but also may highlight the fact that traffic counts
are not an ideal proxy of vehicle emissions for short
averaging periods in busy urban areas, due to potential low
counts during periods of high traffic density. Past studies that
found greater predictive power for traffic terms (Briggs et al.,
1997; Brauer et al., 2003) focused on longer-term concentrations, for which traffic volume might be a more reasonable proxy for emissions. Regardless, our regressions did
yield some compelling findings for particle-bound PAHs.
For example, interpreted literally, the ‘‘large diesel vehicle’’
regression in Table 2 indicates that a baseline PAH concentration of 15 ng/m3 would increase to 24 ng/m3 if one moved
from the lowest to highest traffic count street in our sampling
region.
It is instructive to compare the results of our investigation
with an earlier study that characterized PAH and PM2.5
concentrations in a neighborhood of Roxbury (Levy et al.,
2001). This study used a mobile monitoring protocol with
1-min averaging times to cover a substantial portion of
the neighborhood and used statistical techniques to account
for spatial and temporal autocorrelation. As indicated in
Figure 1, the spatial domain overlapped with the domain
in our analysis. The earlier study concluded that PAH
concentrations were higher with proximity to the bus
terminal or on bus routes, with PM2.5 levels slightly elevated
on high traffic streets (but not significantly so), agreeing with
our general findings. In addition, despite differences in the
monitoring protocol and time of day when measurements
Table 3. Linear mixed effects regression models of 10-min average ratios of moving monitor over stationary monitor concentrations on distance from the
road and wind direction.
PAH
Coefficient
(SE)
Intercept
Distance
Direction
Upwind/neutral
Downwind
Distance*direction
Upwind/neutral
Downwind
2.4 (0.54)
0.020 (0.011)
Ultrafine PM
P-value
(type 3 test of
fixed effects)
0.029
0.21
0.85 (0.61)
F
Coefficient
(SE)
1.0 (0.09)
0.0026 (0.0020)
P-value
(type 3 test of
fixed effects)
0.57
0.05
0.27 (0.11)
F
0.34
0.013 (0.013)
F
PM2.5
Coefficient
(SE)
0.96 (0.035)
0.0012 (0.00074)
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
0.63
0.0083
0.15 (0.041)
F
0.14
0.0038 (0.0026)
F
P-value
(type 3 test of
fixed effects)
0.035
0.0020 (0.00092)
F
369
Levy et al.
were taken, spatial patterns of PAHs appear similar to those
in our analysis. PAH levels were highest toward the northeast
(downwind of the bus terminal) in both studies, and if we
average all measurements from Levy et al. (2001) that were
taken at the same sites as in our current investigation,
the measurements are well correlated (r ¼ 0.64). Although
this is far from definitive, it does provide some support for
the possibility that the spatial trends of particle-bound PAHs
in our investigation might exist on a regular basis, at least for
working hour measurements during the summer.
We can also draw preliminary conclusions from the
magnitude of the concentrations measured in our study.
The mean particle-bound PAH concentration was slightly
below the mean from the earlier Roxbury study (Levy et al.,
2001) and concentrations at a semiurban location in Boston
(Dubowsky et al., 1999), although concentrations at some
sampling sites exceeded these levels, perhaps due to the
greater density of large diesel vehicles. Direct comparability is
impaired by the lack of traffic count data during the sampling
periods in earlier studies, as well as by differences in monitor
placement (e.g., a fourth-floor apartment in the semiurban
location in Boston, versus roadside measurements in Roxbury). Ultrafine particle count concentrations in our study
were somewhat higher than levels reported at sites in the Los
Angeles basin (Kim et al., 2002) or at background urban
sites in Australia (Hitchins et al., 2000), but were lower on
average than near-road measurements in Australia or
concentrations in other high-traffic settings in Boston (Levy
et al., 2002). It is difficult to compare PM2.5 concentrations
with levels in other studies due to differences in instrumentation, but the mean concentration was higher than concentrations measured in other outdoor settings in Boston using the
DustTrak (Levy et al., 2002). There is therefore some
indication that pollution levels in our study are within the
range of those measured in other urban settings, although
concentrations at some sites are at the high end of reported
ambient values.
Aside from sample size considerations, the primary
limitation of our study arises from the focus on urban
settings in close proximity to one another. While this
addressed the question of small-scale spatial variations in
concentrations, it likely reduced the concentration gradients
and therefore impaired statistical power. To develop more
robust regression equations, future studies should attempt to
characterize a wider range of settings with differing source
characteristics and should include measurements at different
times of day and in different seasons to capture general
pollution trends. In addition, our study only captured a
relatively small proportion of moving monitor measurements
that were downwind of the hypothesized source of emissions,
given shifting wind patterns and Boston roads that do not
follow a standard grid (Figure 1). An investigation at a site
with roads parallel to one another and with more consistent
wind direction would strengthen the power of the analysis.
370
Traffic patterns and concentrations of PM and PAHs
Despite these limitations, our study provides useful
information for future transportation exposure assessments
or epidemiological investigations. The strong relation between PAH and ultrafine particle count concentrations and
predefined traffic density scores (along with the correlation
between these scores and measured traffic counts) indicates
that GIS methods can potentially yield exposure estimates
even within proximate sites in an urban area. Similarly, the
evidence of a relation between concentrations and distance
from roadways as well as wind speed and direction provides a
basis for more comprehensive regression-based approaches
for predicting concentrations. Finally, the stronger relations
associated with large diesel vehicles when compared with
total traffic imply that monitoring studies in areas with high
diesel traffic would be warranted, including observational
studies that can better capture the characteristics of the
vehicles linked to elevated concentrations.
Acknowledgments
Funding for this investigation was provided by the US EPA
(Grant #R825267) and by the Kresge Center for Environmental Health (NIEHS ES000002). We thank Maneesh
Anand, Kevin Banahan, Prashant Dilwali, Thy Do, Thomas
Dumyahn, Kevin Han, and Nicholas Lange for their
assistance with field measurements.
References
Brauer M., Hoek G., van Vliet P., Meliefste K., Fischer P., Gehring U.,
Heinrich J., Cyrys J., Bellander T., Lewne M., and Brunekreef B.
Estimating long-term average particulate air pollution concentrations: application of traffic indicators and geographic information
systems. Epidemiology 2003: 14: 228–239.
Briggs D.J., Collins S., Elliott P., Fischer P., Kingham S., Lebret E.,
Pryl K., VanReeuwijk H., Smallbone K., and VanderVeen A.
Mapping urban air pollution using GIS: a regression-based
approach. Int J Geograph Inform Sci 1997: 11: 699–718.
Brunekreef B., Janssen N.A., de Hartog J., Harssema H., Knape M.,
and van Vliet P. Air pollution from truck traffic and lung
function in children living near motorways. Epidemiology 1997: 8:
298–303.
Chang L.T., Suh H.H., Wolfson J.M., Misra K., Allen G.A.,
Catalano P.J., and Koutrakis P. Laboratory and field evaluation
of measurement methods for one-hour exposures to O3, PM2.5, and
CO. J Air Waste Manag Assoc 2001: 51: 1414–1422.
Dubowsky S.D., Wallace L.A., and Buckley T.J. The contribution of
traffic to indoor concentrations of polycyclic aromatic hydrocarbons.
J Expo Anal Environ Epidemiol 1999: 9: 312–321.
English P., Neutra R., Scalf R., Sullivan M., Waller L., and Zhu L.
Examining associations between childhood asthma and traffic flow
using a geographic information system. Environ Health Perspect
1999: 107: 761–767.
Hitchins J., Morawska L., Wolff R., and Gilbert D. Concentrations of
submicrometre particles from vehicle emissions near a major road.
Atmos Environ 2000: 34: 51–59.
Janssen N.A., Schwartz J., Zanobetti A., and Suh H.H. Air
conditioning and source-specific particles as modifiers of the effect
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
Traffic patterns and concentrations of PM and PAHs
of PM10 on hospital admissions for heart and lung disease. Environ
Health Perspect 2002: 110: 43–49.
Kim S., Shen S., and Sioutas C. Size distribution and diurnal and
seasonal trends of ultrafine particles in source and receptor sites of the
Los Angeles basin. J Air Waste Manag Assoc 2002: 52: 297–307.
Kinney P.L., Aggarwal M., Northridge M.E., Janssen N.A., and
Shepard P. Airborne concentrations of PM2.5 and diesel exhaust
particles on Harlem sidewalks: a community-based pilot study.
Environ Health Perspect 2000: 108: 213–218.
Laden F., Neas L.M., Dockery D.W., and Schwartz J. Association of
fine particulate matter from different sources with daily mortality in
six U.S. cities. Environ Health Perspect 2000: 108: 941–947.
Levy J.I., Dumyahn T., and Spengler J.D. Particulate matter and
polycyclic aromatic hydrocarbon concentrations in indoor and
outdoor microenvironments in Boston, Massachusetts. J Expo Anal
Environ Epidemiol 2002: 12: 104–114.
Levy J.I., Houseman E.A., Spengler J.D., Loh P., and Ryan L. Fine
particulate matter and polycyclic aromatic hydrocarbon concentration patterns in Roxbury, Massachusetts: a community-based GIS
analysis. Environ Health Perspect 2001: 109: 341–347.
Massachusetts Highway Department. Traffic Volume Counts. Available
at: http://www.state.ma.us/mhd/trafficc/traffic.htm, May 2001.
Nielsen T. Traffic contribution of polycyclic aromatic hydrocarbons in
the center of a large city. Atmos Environ 1996: 30: 3481–3490.
Oosterlee A., Drijver M., Lebret E., and Brunekreef B. Chronic
respiratory symptoms in children and adults living along streets with
high traffic density. Occup Environ Med 1996: 53: 241–247.
Quintana P.J., Samimi B.S., Kleinman M.T., Liu L.J., Soto K.,
Warner G.Y., Bufalino C., Valencia J., Francis D., Hovell M.H.,
and Delfino R.J. Evaluation of a real-time passive personal particle
monitor in fixed site residential indoor and ambient measurements.
J Expo Anal Environ Epidemiol 2000: 10: 437–445.
Rea A.W., Zufall M.J., Williams R.W., Sheldon L., and Howard-Reed C.
The influence of human activity patterns on personal PM exposure: a
comparative analysis of filter-based and continuous particle measurements. J Air Waste Manage Assoc 2001: 51: 1271–1279.
Ruuskanen J., Tuch T., Ten Brink H., Peters A., Khlystov A.,
Mirme A., Kos G.P.A., Brunekreef B., Wichmann H.E.,
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(5)
Levy et al.
Buzorius G., Vallius M., Kreyling W.G., and Pekkanen J.
Concentrations of ultrafine, fine, and PM2.5 particles in three
European cities. Atmos Environ 2001: 35: 3729–3738.
Tolbert P.E., Klein M., Metzger K.B., Peel J., Flanders W.D., Todd K.,
Mulholland J.A., Ryan P.B., and Frumkin H. Interim results of
the study of particulates and health in Atlanta (SOPHIA). J Expo
Anal Environ Epidemiol 2000: 10: 446–460.
US Department of Transportation, Federal Highway Administration.
Traffic Monitoring Guide. Available at: http://www.fhwa.dot.gov/
ohim/tmguide/, August 16 2002.
US Environmental Protection Agency. Compilation of Air Pollutant
Emission Factors, Volume II: Mobile Source, 4th edn. US Government Printing Office, Washington, DC, 1985.
US Environmental Protection Agency. NEDS Source Classification
Codes and Emission Factor File. US EPA Office of Air Quality
Planning and Standards, Research Triangle Park, NC, 1989.
van Vliet P., Knape M., de Hartog J., Janssen N., Harssema H., and
Brunekreef B. Motor vehicle exhaust and chronic respiratory symptoms
in children living near freeways. Environ Res 1997: 74: 122–132.
Venn A.J., Lewis S.A., Cooper M., Hubbard R., and Britton J. Living
near a main road and the risk of wheezing illness in children. Am J
Respir Crit Care Med 2001: 164: 2177–2180.
Weiland S.K., Mundt K.A., Ruckmann A., and Keil U. Self-reported
wheezing and allergic rhinitis in children and traffic density on street
of residence. Ann Epidemiol 1994: 4: 243–247.
Wjst M., Reitmeir P., Dold S., Wulff A., Nicolai T., von LoeffelholzColberg E.F., and von Mutius E. Road traffic and adverse effects on
respiratory health in children. Br Med J 1993: 307: 596–600.
Yu O., Sheppard L., Lumley T., Koenig J.Q., and Shapiro G.G. Effects
of ambient air pollution on symptoms of asthma in Seattle-area
children enrolled in the CAMP study. Environ Health Perspect 2000:
108: 1209–1214.
Zhu Y., Hinds W.C., Kim S., Shen S., and Sioutas C. Study of ultrafine
particles near a major highway with heavy-duty diesel traffic. Atmos
Environ 2002a: 36: 4323–4335.
Zhu Y., Hinds W.C., Kim S., and Sioutas C. Concentration and size
distribution of ultrafine particles near a major highway. J Air Waste
Manage Assoc 2002b: 52: 1032–1042.
371