Journal of Exposure Analysis and Environmental Epidemiology (2003) 13, 134–143
r 2003 Nature Publishing Group All rights reserved 1053-4245/03/$25.00
www.nature.com/jea
Comparison between different traffic-related particle indicators: Elemental
carbon (EC), PM 2.5 mass, and absorbance
JOSEF CYRYS,a,b JOACHIM HEINRICH,a GERARD HOEK,c KEES MELIEFSTE,c MARIE LEWNÉ,d
ULRIKE GEHRING,a,b TOM BELLANDER,d PAUL FISCHER,e PATRICIA VAN VLIET,c
MICHAEL BRAUER,c,f H.-ERICH WICHMANN,a,b AND BERT BRUNEKREEFc
a
GSF, National Research Center for Environment and Health, Institute of Epidemiology, Neuherberg, Germany
Ludwig-Maximilians-University of Munich, Institute of Medical Data Management, Biometrics and Epidemiology, Chair of Epidemiology, Munich,
Germany
c
Environmental and Occupational Health Unit, Institute for Risk Assessment Sciences, Utrecht University, The Netherlands
d
Department of Environmental Health, Stockholm County Council, Stockholm, Sweden
e
Laboratory of Exposure Assessment and Environmental Epidemiology, National Institute of Public Health and the Environment, The Netherlands
f
School of Occupational and Environmental Hygiene, The University of British Columbia, Vancouver, Canada
b
Here we compare PM2.5 (particles with aerodynamic diameter less than 2.5 mm) mass and filter absorbance measurements with elemental carbon (EC)
concentrations measured in parallel at the same site as well as collocated PM2.5 and PM10 (particles with aerodynamic diameter less than 10 mm) mass and
absorbance measurements. The data were collected within the Traffic-Related Air Pollution on Childhood Asthma (TRAPCA) study in Germany, The
Netherlands and Sweden. The study was designed to assess the health impact of spatial contrasts in long-term average concentrations. The measurement
sites were distributed between background and traffic locations. Annual EC and PM2.5 absorbance measurements were at traffic sites on average 43–84%
and 26–76% higher, respectively, compared to urban background sites. The contrast for PM2.5 mass measurements was lower (8–35%). The smaller
contrast observed for PM2.5 mass in comparison with PM2.5 absorbance and EC documents that PM2.5 mass underestimates exposure contrasts related to
motorized traffic emissions. The correlation between PM10 and PM2.5 was high, documenting that most of the spatial variation of PM10 was because of
PM2.5. The measurement of PM2.5 absorbance was highly correlated with EC measurements and suggests that absorbance can be used as a simple,
inexpensive and non-destructive method to estimate motorized traffic-related particulate air pollution. The EC/absorbance relation differed between
countries and site type (background/traffic), supporting the need for site-specific calibrations of the simple absorbance method. While the ratio between
PM2.5 and PM10 mass ranged from 0.54 to 0.68, the ratio of PM2.5 absorbance and PM10 absorbance was 0.96–0.97, indicating that PM2.5 absorbance
captures nearly all of the particle absorbance.
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13, 134–143. doi:10.1038/sj.jea.7500262
Keywords: traffic-related pollutants, particles, PM10, PM2.5, absorbance, EC.
Introduction
Recent epidemiological studies have suggested that trafficrelated air pollution contributes to health effects associated
with long-term exposure to air pollution (Wjst et al., 1993;
Weiland et al., 1994; Oosterlee et al., 1996; Vliet van et al.,
1997; English et al., 1999; Nyberg et al., 2000). Exposure
assessment in these studies has included self-reported traffic
volumes on residential streets, quantitative data on traffic
counts collected by local agencies or the distance to major
roads. Only a few studies have incorporated actual measure-
1. Address all correspondence to: Dr. J. Cyrys, GSF-National Research
Center for Environment and Health, Institute of Epidemiology, Ingolstaedter Landstr. 1, D-85764 Neuherberg, Germany. Tel.: +49-89-31874156. Fax: +49-89-3187-3380. E-mail: [email protected]
Received 28 February 2002
ments of traffic-related air pollutants. Most of these studies
used NO2 as an indicator for traffic-related air pollution
(Nakai et al., 1995; Ciccone et al., 1998; Krämer et al.,
2000). Some other studies have included particle measurements, such as PM2.5, PM10, or absorbance of PM2.5 filters
(Vliet van et al., 1997; Janssen et al., 2001).
The health effects of particles probably differ between
particles from different sources, for example, resuspended
road dust or combustion-derived particles. In a study
conducted by Brunekreef et al. (1997) health effects were
associated with local truck traffic counts and with black
smoke concentrations in schools, whereas no associations
were observed for car traffic counts, suggesting a specific
effect of diesel exhaust particles. Duhme et al. (1996) found
a positive association between self-reported symptoms of
asthma and allergic rhinitis and self-reported frequency of
truck traffic. In addition, recent experimental studies (Diaz-
Traffic-related particle indicators
Sanchez, 1997, Bayram et al., 1998) highlight the role of
diesel exhaust particles in enhancing inflammatory and
allergic responses in the respiratory system. As several studies
reported strong correlations between elemental carbon (EC)
concentrations and diesel vehicle traffic (Delumyea et al.,
1980; Keeber, 1990) and documented that the majority of
EC emissions in urban environments originate from diesel
exhaust engines (Wolff et al., 1985; Hamilton and Mansfield,
1991; Fischer et al., 1997; Gray and Cass, 1998), EC
measurements were often used to estimate the contribution
of diesel exhaust engines to the particle concentrations.
More recently studies showed that nowadays in Western
Europe and the US, between 67 and 90% of the
atmosphere’s content of EC is estimated to be produced by
diesel-powered vehicles (Hamilton and Mansfield, 1991;
Fischer et al., 1997). Nevertheless, it must be still considered
that the contribution of other primary combustion sources
such as coal-burning, heating or industry can vary significantly between cities (Fischer et al., 1997; Hies et al., 2000).
EC represents rather complex species with a number of
functional groups, e.g., alcoholic, phenolic, carbonylic, or
carboxylic. Besides ‘‘elemental carbon’’, other terms such
as ‘‘soot’’ and ‘‘black carbon’’ are used loosely and often
interchangeably by air quality, health and industrial
researchers. The terms define a similar fraction of the
carbonaceous aerosol and are supposed to be comparable,
but have a slightly different thermal, optical, and chemical
behavior in most cases. These properties depend on the
sources and atmospheric age of carbonaceous aerosols. The
analytical determination of EC is complex. It requires a
noncarbon filter, the measurements are time-consuming,
expensive, and filter destructing.
Furthermore, multiple methods exist for measuring the
elemental carbon content of particles, which may yield
substantially different concentrations (Schmid et al., 2001).
Since elemental or black carbon is the main light-absorbing
substance in ambient air (Horvarth, 1993), EC measurements should be highly correlated with reflectance measurements, as in the old Black Smoke method. Measuring the
reflectance of PM filters would be an alternative and cheaper
way to assess exposure to traffic-related particulate matter.
Several studies have documented that black smoke (BS),
derived from absorbance coefficients, is well correlated with
the concentration of EC or soot (Ulrich and Israel, 1992;
Janssen et al., 2001; Kinney et al., 2000). All those studies
refer to correlation of daily samples, whereas the correlation
between annual averages has not been studied till now.
However, this correlation is the relevant correlation for
studies on chronic effects of air pollution where spatial
contrasts are evaluated.
As part of an international collaborative study on the
impact of Traffic-Related Air Pollution on Childhood
Asthma (TRAPCA) exposure to ambient particles were
estimated for three separate birth cohorts in Munich
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Cyrys et al.
(Germany), The Netherlands, and Stockholm County
(Sweden). We conducted a 1-year monitoring program to
estimate the annual average concentrations of PM2.5, PM10,
absorbance of PM2.5 and PM10 filters, and EC as described
in detail by Hoek et al. (2002). The measurement sites were
selected in order to maximize the variation among the trafficrelated predictor variables (e.g., traffic intensity, population
density near the sites). Thus, the data set offers the unique
possibility to compare the different particle measurements
with regard to their suitability as indicators for traffic. Here
we test the assumption that the absorption coefficient is a
good marker for EC by comparison of the variability of
PM2.5 and absorbance of PM2.5 filters with the variability of
EC and by comparison of the contrast between background
and traffic sites for PM2.5 mass, absorbance of PM2.5, and
EC. Furthermore, we analyze the difference and correlation
between absorbance measured on PM2.5, and PM10 filters
and compare it to the difference and correlation between the
PM2.5 and PM10 mass concentrations.
Methods
Study period and study design
As part of the TRAPCA study, measurements of ambient
exposure to motorized traffic emissions were conducted
for three separate birth cohorts in three locations, The
Netherlands, Munich (Germany), and Stockholm County
(Sweden). Air pollution measurements were made between 1
March 1999 and 20 April 2000 in The Netherlands, 16
March 1999 and 21 July 2000 in Munich, and 9 February
1999 and 8 March 2000 in Stockholm.
In Munich and The Netherlands 40, in Stockholm County
42 measurement sites were selected. For budgetary reasons it
was not possible to obtain all particle measurements (PM2.5,
PM10, and EC) for each site. Most pump units were operated
with two inlets simultaneously, the first one for the regular
PM2.5 measurements and the second one for other purposes,
such as EC or PM10 measurements or quality control
purposes. Owing to this design we measured PM2.5
concentrations and the reflectance of PM2.5 filters at all 122
sites, EC at 36 sites (these sites are called in the following
‘‘EC sites’’), PM10 and the reflectance of PM10 filters at
additional 36 sites (‘‘PM10 sites’’). At each site, four 14-day
average samples were collected, spread over 1 year. For all
pollutants, annual averages were calculated as described by
Hoek et al. (2002). This strategy resulted in reasonably
precise estimates of the annual average, for example, the
standard error of the mean was less than 1 mg/m3 for PM2.5.
Site selection
The distribution of the sampling sites differed between the
three study areas because of the different characteristics and
distribution of the birth cohorts. In The Netherlands, a large
135
Traffic-related particle indicators
Cyrys et al.
part of the whole country was included. The spatial scale of
the Dutch area was approximately 200 km. The study areas
in Germany and in Sweden were rather small with a spatial
scale of 20 km in Germany and 25 km in Sweden. In Sweden,
the study area involved the city of Stockholm and three
communities near Stockholm (Solna, Sundbyberg, and
Järfälla); in Germany, the study area was restricted to the
city of Munich.
The measurement sites were selected in order to capture
sufficient variability with regard to the traffic intensity and/or
population density near the sites. In each country, we selected
both, background as well as traffic sites, according to
common criteria. Background sites were defined as sites that
were not influenced substantially by sources in the ‘‘direct
vicinity’’. Within a circular buffer of 50 m there were no
major roads and no other important sources of combustion
gases and particulate matter (construction works, small
industries, district heating plant, parking lot/garage). Traffic
sites were those where no sources except the street were
present near the site. The background sites were distributed
well between the inner city and the suburbs. The traffic sites
were adjacent to streets with different traffic intensities and
the distance to the street varied from site to site. The traffic
sites were located both at main roads and side roads. The
mean distance between samplers at the traffic sites and the
nearest street was 6 m in The Netherlands, 10 m in Munich,
and 19 m in Sweden. In addition to the urban background
and urban traffic sites, rural sites were also included in The
Netherlands and Sweden. The distribution of EC and PM10
measurement sites among the different sites environments is
given in Table 1.
Sampling and analysis methods
All air samples were collected with Harvard impactors
(Marple et al., 1987) according to a standard operating
procedure SOP TRAPCA2.0 (Hoek et al., 2002). The
sampling duration was 14 days. In order to avoid overloading of the filters, timers were used to sample for 15 min
Table 1. Distribution of EC and PM10 measurement sites in Germany
(Munich), the Netherlands and Sweden (Stockholm County). Note that
EC and PM10 sites are different sites.
Germany
The Netherlands
Sweden
EC PM10
sites sites
EC
sites
PM10
sites
EC PM10
sites sites
12
8
4
4
4
12
8
4
4
4
12
10
3
7
2
Total
12
Background
6
0
Background rurala
Background urbana 6
Traffic
6
a
12
7
0
7
5
12
9
3
6
3
In Munich, all sites were urban sites. In the Netherlands urbanization
degree 1 and 2 of the municipality is considered as urban, 3–5 as rural. In
Stockholm county, rural is Järfälla, other areas are considered as urban.
136
during each 2 h period. In this way, we collected effectively a
42-h sample representative for 14 days. All samples were
collected at a flow rate of 10 l/min. PM2.5 and PM10 were
collected on Anderson 37 mm 2 mm pore size Teflon filters
with a polyolefin support ring. For the EC measurements we
used Schleicher and Schuell GF20 quartz filters. Before and
after sampling, Teflon filters were conditioned during at least
24 h at a constant temperature and relative humidity and
then weighed using a microbalance by each center. In The
Netherlands and Munich, climate-controlled weighing rooms
were used (T ¼ 20721C and RH ¼ 3575% in Munich;
T ¼ 23711C and RH ¼ 3772% in The Netherlands). In
Stockholm, the temperature in the climate room was
22721C and the RH between 50 and 65%, so a desiccator
with a saturated solution of CaCl2 was used to maintain RH
at 30–40%.
After weighing, the reflectance of all Teflon filters was
measured according to a procedure described before (Fischer
et al., 2000). The exact measurement procedure is described
at the internet site for the EU multicenter study ULTRA
(http://www.ktl.fi/ultra). All reflectance measurements were
conducted in the Laboratory of the Environmental and
Occupational Health Unit, The Netherlands. The reflectance
of all Teflon filters was measured using the M43D Smoke
Stain Reflectometer (Diffusion Systems LTD), which
measures the reflection of the light incidence in percent.
After warming up for at least 15 min, the reflectometer is set
to zero. With the measuring head located over the white
standard tile, the reflectance reading is set to 100.0. Next, the
gray standard tile is measured, the reflectance should be
within the limits specified in the manufacturer’s manual. The
reflectance of five blank Teflon filters was measured before the
study period. To compensate for inhomogeneity of the filter,
the reflectance was measured on five different spots of each
filter and then averaged. The filter with the median average
reflectance was designated as the primary control filter, unless
the standard deviation of the five measurements was more
than 0.5 units (if it was higher a new Teflon filter was selected
and the procedure repeated). The primary control filter was
used in each analysis session to set the reflectometer to 100.0.
Sampled filters were measured at the five standard locations
as well. The averaged reflectance was transformed into an
absorption coefficient using the formula according to ISO
9835 (ISO, 1993):
a ¼ ððA=2Þ=V Þ lnðR0 =Rf Þ
where a is the absorption coefficient (m1 105), A the
loaded filter area (0.00078 m2), V the sampled volume (m3),
R0 the average reflectance of field blanks (%), and Rf the
reflectance of sampled filter (%). The average field blank was
used, but this is only valid if filters are from the same lot. The
absorbance coefficient was then multiplied by 105 to make
the readings more comprehensible. Reflectance applied in
the current work was not converted into units of mass
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Traffic-related particle indicators
concentrations since the transformation used to calculate
mass concentration is filter specific and no transformation
was available for the filters used in this study.
Analysis of EC filters was conducted at TÜV Süddeutschland Bau und Betrieb GmbH in Donzdorf, Germany,
according to the German reference method VDI 2465 (VDI,
1996). After a pretreatment of filter by liquid extraction of
the organic compounds with a nonpolar solvent for the
removal of extractable organic carbon, the filter is dried, and
the non-extractable organic carbon is removed by thermal
desorption under nitrogen at 5001C. The determination of
the remaining carbon (i.e., EC) is based on converting
carbon-containing compounds to CO2 and H2O in an
oxidizing atmosphere at 6501C, and the CO2 produced is
detected by coulometry.
We conducted an extended internal quality assurance/
quality control program as well as an external comparison
program which is described in detail by Hoek et al. (2002). In
general, the detection limits (DL) for all components were
low. The detection limit for PM2.5 was 3.4, 1.65, and 0.7 mg/m3
in Germany, The Netherlands, and Sweden, respectively. DL
for PM2.5 absorbance was 0.2 m1 105 in Germany and
0.1 m1 105 in The Netherlands and Sweden. All PM2.5
absorbance samples were above the detection limit. The
analytical DL for EC was 10 mg/filter, that is, 0.25 mg/m3. All
EC field blanks and one EC sample were lower as DL. The
precision of EC measurement calculated as the relative
standard deviation of the duplicates was determined in The
Netherlands. It was, with 12.9%, significantly higher than
the precision of PM2.5 (3.6%) or PM2.5 absorbance (1.9%).
Results
Levels and ranges
The comparison of EC levels with PM2.5 mass concentrations
and PM2.5 absorbance is possible only for sites where EC
measurements were also made (n ¼ 12 for each country).
Table 2 provides a description of the variability of EC annual
average concentrations compared with the PM2.5 and PM2.5
absorbance annual average concentrations. The levels of EC
and PM2.5 absorbance were comparable for Germany and
The Netherlands and lower in Sweden. The mean PM2.5
concentration was highest in The Netherlands, followed by
Germany and Sweden. The range (the difference between the
lowest and the highest annual average, expressed as a
percentage of the median) of EC ranged from 122% in
Sweden to 177% in The Netherlands. In contrast, the range
for PM2.5 mass concentrations was significantly smaller
(46% in Munich, 69% in The Netherlands, and 64% in
Sweden). The range for the corresponding absorption
coefficients of the PM2.5 filters was more comparable with
the EC range (105%, 139%, and 146% for Germany, The
Netherlands, and Sweden, respectively).
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Cyrys et al.
Table 2. Distribution of annual average concentrations of EC, PM2.5
mass, and absorbance for each country separately (EC sites).
Range (%)a
Country
N
Min
Mean
Med
Max
EC (mg/m3)
Germany
Netherlands
Sweden
12
12
12
1.7
1.4
1.1
2.6
2.7
1.5
2.4
2.3
1.4
5.4
5.4
2.8
154.5
176.6
122.0
PM2.5 (mg/m3)
Germany
Netherlands
Sweden
12
12
12
11.5
14.4
8.1
13.8
18.3
10.3
13.4
18.4
9.8
17.6
27.1
14.4
45.6
69.1
64.0
PM2.5 absorbance (105 m1)
Germany
12
1.4
Netherlands
12
0.9
Sweden
12
0.7
1.9
1.8
1.3
1.8
1.6
1.1
3.3
3.1
2.3
104.7
139.3
146.0
a
Expressed as the difference between the lowest and the highest annual
average divided by the median (in %).
Table 3 summarizes the annual average concentrations of
EC, PM2.5, and PM2.5 absorbance for the three different site
types (background urban, background rural, and traffic site)
for each country separately. The annual averages of all
pollutants were lower at the background rural sites than at
the background urban sites. The urban–rural contrast was
substantially smaller than the background-traffic contrast.
EC accounts for 11.8–15.8% of the measured PM2.5 mass at
background urban sites, and for 18.1–21.7% of PM2.5 mass
at the traffic sites. Furthermore, the table provides the ratios
of the means with the background urban sites as the reference
group. The highest ratios between traffic and urban background sites were found for EC measurements (1.43–1.84),
followed by the ratios for the PM2.5 absorbance (1.26–1.76).
The lowest ratios were observed for the PM2.5 mass
concentration (1.08–1.35). Note that in Sweden there were
only two traffic sites, therefore the comparison would be less
meaningful. In Germany and The Netherlands, the absolute
difference in PM2.5 concentration between traffic and background sites was essentially the same as the difference in EC
concentrations. This suggests that the PM2.5 difference is
because of EC entirely and may reflect the important
contribution of diesel exhaust to spatial variability in urban
PM2.5 concentrations.
Since EC and PM10 were measured at different sites, the
comparison of PM10 and PM10 absorbance with EC was not
possible. Therefore, we compared the ratio of mean PM2.5
mass to mean PM10 mass with the ratio of mean PM2.5
absorbance to mean PM10 absorbance (Table 4). The
measured PM2.5 mass concentration was between 54%
(Sweden) and 68% (Germany) of the PM10 concentration,
whereas almost identical levels of PM2.5 and PM10
absorbance were measured. The range in annual averages,
expressed as a percentage of the median, was only moderately
higher for PM10 compared to PM2.5: in The Netherlands 41
137
Traffic-related particle indicators
Cyrys et al.
Table 3. Annual average concentrations for different particle indicators measured at traffic and urban background sites (EC sites).
Germany
EC
(mg/m3)
PM2.5
(mg/m3)
PM2.5 Absorbance
(105 m1)
Background
Background
Traffic
Background
Background
Traffic
Background
Background
Traffic
urban
rural
urban
rural
urban
rural
The Netherlands
a
Sweden
a
n
Mean
Ratio
n
Mean
Ratio
n
Mean
Ratioa
6
0
6
6
0
6
6
0
6
2.1
F
3.1
13.3
F
14.3
1.7
F
2.1
1.00
F
1.43
1.00
F
1.08
1.00
F
1.26
4
4
4
4
4
4
4
4
4
2.1
1.9
3.9
17.8
17.3
19.9
1.5
1.3
2.6
1.00
0.89
1.84
1.00
0.97
1.12
1.00
0.88
1.76
7
3
2
7
3
2
7
3
2
1.4
1.3
2.5
10.2
8.4
13.8
1.2
0.9
2.1
1.00
0.92
1.79
1.00
0.83
1.35
1.00
0.70
1.75
a
Ratio between the annual average concentrations for the traffic sites to the background sites.
Table 4. Annual average concentrations for PM2.5 and PM10 mass and absorbance (PM10 sites).
Germany
PM2.5 mass
(mg/m3)
PM10 mass
(mg/m3)
PM2.5
absorbance
(105 m1)
PM10
absorbance
(105 m1)
The Netherlands
n
Mean
12
14.3
12
21.0
12
2.0
12
2.1
a
Ratio
n
Mean
0.68
12
17.1
12
25.5
12
1.6
12
1.6
0.96
Sweden
a
Ratio
n
Mean
0.67
12
10.7
12
20.6
12
1.4
12
1.4
0.97
Ratioa
0.54
0.96
a
Mean of the ratios between the annual average (mass or absorbance) concentrations of PM2.5–PM10 calculated for each site separately
versus 44%; in Munich 62 versus 71%, and in Sweden 142
versus 80% (PM10 range listed first).
Relation between the pollutants
Figure 1a and b shows the correlation between the annual
averages of EC and PM2.5 mass concentrations and between
EC and the absorption coefficients of PM2.5 filters. The
correlation for the absorption coefficient of PM2.5 and EC is
clearly higher than for PM2.5 mass and EC in all three centers
(R2: 0.93 versus 0.67, 0.94 versus 0.56 and 0.73 versus 0.61
in Germany, The Netherlands, and Sweden, respectively).
The lowest correlation between the absorption coefficients
of PM2.5 filters and EC was observed for the Swedish
measurements. The regression models presented in Figure 1a
and b for Germany, The Netherlands, and Sweden differed
significantly from each other (Po0.01, ANOVA). After
adjusting for country, the regression models for traffic and
background sites differed significantly as well (Po0.01,
ANOVA). For the 24 background sites, the regression model
was EC ¼ 0.34 (0.28)+0.89 (0.24)*PM2.5 absorption; for
the 12 traffic sites, EC ¼ 1.78 (0.44)+1.98 (0.18)*PM2.5
138
absorption (value in brackets are the standard errors of the
regression coefficients). This suggests that while the correlation between EC and absorbance was high, site-specific
calibration with EC is necessary.
Table 5 shows the Spearman correlation coefficients
between EC, PM2.5 mass, and absorbance of PM2.5 filters.
The crude- and country-adjusted correlations were rather
similar. The highest correlation was found for the relation
between EC and PM2.5 filter absorbance (rcrude ¼ 0.94,
radj ¼ 0.93), followed by PM2.5 versus PM10 (rcrude ¼ 0.78,
radj ¼ 0.85). The correlations between EC and PM2.5 mass
and between PM2.5 mass and PM2.5 absorbance were
weaker, but still significant. The correlations between the
pollutants were generally stronger at traffic sites than at
background sites. The strongest correlation at traffic sites was
found for the relation between EC and PM2.5 absorbance
(r ¼ 0.97), followed by PM2.5 versus PM10 (r ¼ 0.87) and EC
versus PM2.5 absorbance (r ¼ 0.79). The correlations at the
background sites were weaker.
The high correlation between PM10 and PM2.5 (Table 5,
Figure 2a) documents that large portion of the spatial
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Traffic-related particle indicators
Germany
for the three countries and the two site types were
significantly different from each other (Po0.01), but the
differences were quantitatively small.
Sweden
The Netherlands
y = 0.1877x - 0.4188
R2 = 0.6049
y = 0.2552x - 2.0268
R2 = 0.5631
y = 0.4944x - 4.2051
R 2 = 0.672
a
Cyrys et al.
6
5
Discussion
EC (g/m3)
4
3
2
1
0
0
5
10
15
20
25
30
3
PM2.5 (g/m )
Germany
y = 2.0168x - 1.1875
R2 = 0.9338
b
Sweden
The Netherlands
y = 0.9048x + 0.3560
R2 = 0.7298
y = 1.6053x - 0.2620
R2 = 0.9371
6
5
3
EC (g/m )
4
3
2
1
0
0
1
2
3
-5
4
-1
Absorption coefficient (10 m )
Figure 1. Correlation between the annual averages of elemental
carbon (EC) and (a) PM2.5 mass concentrations and (b) absorption
coefficients of PM2.5 filters for Germany, The Netherlands, and
Sweden.
variation of PM10 was because of PM2.5. The regression
models for the three countries were significantly different
from each other (Po0.01). The respective regression models
predict for an annual average PM10 concentration of 20 mg/m3
(the current EU guideline for 2010), PM2.5 concentrations of
13.4, 13.7, and 10.5 mg/m3 for The Netherlands, Germany,
and Sweden, respectively. There was no significant difference
for the regression models of background and traffic sites
(calculated for all countries combined, but with adjustment
for country). The correlation between PM2.5 absorbance and
PM10 absorbance was almost 1.0 for background sites as well
for traffic sites (Table 5, Figure 2b). The regression models
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
EC, PM2.5 mass, and absorbance
The averaged concentrations of EC ranged from 1.5 to
2.7 mg/cm3 across the three sites studies, levels that are typical
of those reported in other urban areas, for example, average
EC concentration measured at four sites using a thermooptical method ranged from 1.5 mg/m3 in Research Triangle
Park, NC to 3.3 mg/m3 in Rubidoux, CA and Phoenix, AZ
(Tolocka et al., 2001). On the other hand, Countant and
Stetzer (2001) reported for 13 US sites lower EC concentrations (measured by thermal optical reflectance) as those
measured in our study. The EC concentrations ranged from
0.2 in Bismarck, ND to 1.4 mg/m3 in New York City (Bronx
Botanical Garden), NY. However, by the comparison of the
results from different studies one should keep in mind that
the EC amount is highly dependent upon the method used
for the EC determination (Schmid et al., 2001). The VDI
method, which we used in our study, was found to give
generally higher amounts of EC than the thermal–optical
methods. In our study, EC accounts for approximately 15–
19% of the measured PM2.5 mass. The values are higher than
the corresponding values reported by Countant and Stetzer
(2001), which ranged from 4 to 10%, but comparable with
the results obtained by Tolocka et al. (2001) (12–22%). The
highest EC/PM2.5 ratio was observed at traffic sites in
Germany (22%), followed by The Netherlands (20%) and
Sweden (18%). For comparison, in Germany about 20% of
the private cars and all buses use diesel. In The Netherlands,
about 13% of the private cars and all buses use diesel. In
Stockholm county, about 5% of private cars and 90% of the
buses (remainder ethanol) use diesel.
Filter absorbance measurements are definitely cheaper and
easier to conduct than EC measurements and offer superior
measurement precision (1.9% for absorbance versus 12.9%
for EC). The absorbance coefficients of PM2.5 filters were
clearly more highly correlated with simultaneously measured
EC concentrations than with PM2.5 mass concentrations
(radj ¼ 0.93 versus 0.74). The results support the assumption
that measuring the absorbance of PM filters is an alternative
way to estimate the relative concentration of incomplete
combustion-derived particulate matter. Several other studies
have documented that black smoke (BS), derived from
absorbance coefficients, is well correlated with the concentration of EC or soot. In a study in New York, the absorption
coefficients of PM2.5 filters was highly correlated (r ¼ 0.95)
with simultaneously measured EC concentrations (Kinney
et al., 2000). In a study in Berlin, outdoor BS concentrations
at a traffic-impacted site were highly correlated with EC
139
Traffic-related particle indicators
Cyrys et al.
Table 5. Pearson correlation coefficients and Pearson partial correlation coefficients between the particle indicators.
Site type
EC versus PM2.5a
EC versus PM2.5 absa
PM2.5 versus PM2.5 absa
PM2.5 versus PM10b
PM2.5 versus PM10 absb
Crude
All (n=36)
Partialc
All (n=36)
0.72
0.0001
0.74
0.0001
0.94
0.0001
0.93
0.0001
0.65
0.0001
0.72
0.0001
0.78
0.0001
0.85
0.0001
0.99
0.0001
0.99
0.0001
Partialc
Traffic (n=12)
Partialc
Background all (n=24)
0.79
0.0066
0.55
0.074
0.97
0.0001
0.65
0.0012
0.69
0.0286
0.64
0.0015
0.87
0.0011
0.82
0.0001
0.99
0.0001
0.98
0.0001
a
EC sites.
PM10 sites.
c
Adjusted for country.
b
concentrations (r ¼ 0.98) (Ulrich and Israel, 1992). The
correlation found by Janssen et al. (2001) was also high
(r ¼ 0.85); after excluding one outlier the correlation
coefficient increased to 0.93. However, all those studies
measured either EC or BS differently in comparison with our
study. In addition, they refer to correlation of daily samples,
whereas in our study the correlation between annual averages
was analyzed. The two types of correlations may differ, for
example, because common meteorology does affect the daily
correlation but not the averages.
The correlation between EC and PM2.5 absorbance was
higher at traffic sites (radj ¼ 0.97) compared to background
sites (radj ¼ 0.65). Even after removing of 2 traffic sites with
EC concentrations higher than 5 the correlation remains
high (radj ¼ 0.93). There was a significant difference in the
regression slopes relating absorbance and EC for background and traffic sites. There also was a significant
difference in the regression slopes relating absorbance
and EC, especially between Sweden and the other two
countries. Since both EC and absorption coefficient were
measured in the same laboratories for all study areas, this
is unlikely caused by methodological differences. This suggests that validation of the simpler absorbance method is
necessary with EC measurements at sites representative for
the study.
The spatial contrast in the absorption coefficient of
PM2.5 was more comparable with the spatial contrast for
EC than that for PM2.5. Also the contrast between the
traffic sites and the background sites for PM2.5
absorbance was similar to that of EC.
The high correlation between EC and PM2.5 absorbance is
also reflected by the variation of the particulate pollutants.
Since we selected sites in order to capture maximum variability
in traffic intensity among the streets, we observed a substantial
range in annual average concentrations of the particulate
pollutants in each study area. As expected, EC has the highest
variation among all pollutants, except in Stockholm County,
140
followed by the corresponding ranges for the absorption
coefficients of PM2.5 filters. The smallest ranges were observed
for PM2.5 mass concentrations in all three centers. A small
spatial variability in PM10 concentrations within an urban and
between urban and non-urban areas was observed also in
other studies (Hoek et al., 1997a; Monn et al., 1997; Cyrys
et al., 1998). A major fraction of PM2.5 consists of secondary
pollutants (sulfates, nitrates), which are formed within the
atmosphere from gas-to-particle conversion processes. Those
compounds are mainly products of long-range transport and
thus have relatively low spatial variability. Since EC and
absorbance of PM2.5 filters relate more to the primary
pollutants (e.g., combustion particles of local origin) they
have a larger spatial variability even in small areas.
Another possibility to compare the particulate pollutants
with regard to their suitability as indicators for traffic is to
compare the contrast between traffic sites and background
sites. PM2.5 mass concentrations at the traffic sites were 8–
35% higher than at the urban background sites. The contrast
for PM2.5 absorbance was considerably higher (26–76%),
which is more comparable to the contrast for EC (43–84%).
This is in line with several other studies that have documented
a stronger influence of traffic on filter ‘‘blackness’’ or EC
compared to PM2.5 or PM10 mass concentrations, with
street/background ratios from 1.8 to 4.1 for ‘‘soot’’ or EC
and from 0.9 to 1.3 for PM2.5 or PM10 mass concentrations
(Janssen et al., 1997; Fischer et al., 2000; Kinney et al., 2000;
Janssen et al., 2001). Although the reflectance measurements
have been widely conducted in the ‘‘old’’ BS measurement
method (OECD, 1964), the method based on reflectance
measurement and the results can be compared to ours.
Janssen et al. (1997) found 2.6 times higher BS concentrations measured at a street site compared to urban
background site. The lower contrast for the reflectance
found in our study can be explained by differences in
study designs. In the Janssen study, the measurements
were conducted only during the day at 0.5 m from the
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Traffic-related particle indicators
Germany
y = 0.5505x + 2.7055
2
R = 0.9026
a
Cyrys et al.
The Netherlands
y = 0.6739x - 0.1038
2
R = 0.7643
Sweden
y = 0.2816x + 4.8959
2
R = 0.8261
3
PM2.5 (g/m )
20
15
10
5
0
0
5
10
15
20
25
30
35
40
PM10 (g/m3)
Germany
The Netherlands
y = 0.8949x + 0.1138
R2 = 0.9795
Sweden
y = 0.8430x + 0.1437
R2 = 0.9888
4
-5
-1
Absorption coefficient of PM2.5 filters (10 m )
b
y = 1.0010x - 0.0854
R2 = 0.9944
3
2
1
0
0
1
2
3
4
Absorption coefficient of PM10 filters (10-5m-1)
Figure 2. Correlation between the annual averages of (a) PM2.5 and
PM10 mass concentrations and (b) absorption coefficients of PM2.5
and PM10 filters for Germany, The Netherlands, and Sweden.
edge of the road, the sampling height was 3 m. In the
present study, the measurement sites were located on average
11 m away from the nearest road (24 m from the nearest
major road), the average sampling height was 4 m and the
measurements were spread over 24 h. BS measured by
Nielsen (1996) in Copenhagen were also higher than in our
study (2.4 times higher in the main street), possibly because
of the larger traffic intensities in the Danish study
(60,000 vehicles/day) instead of the average of 10,000
vehicles/day/km in a circular buffer of 50 m in the present
study. In Amsterdam (Fischer et al., 2000) the contrast was
1.82 for PM2.5 absorbance and 1.84 for PM10 absorbance
(the average traffic count in the main street was 16,082
vehicles/day).
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
PM10 and PM2.5 mass and absorbance
The correlation between annual averages of PM2.5 and PM10
mass was high, documenting that a large portion of the
spatial variation of PM10 was because of PM2.5. PM2.5–
PM10 ratios are very similar in Germany and The Netherlands. In Sweden, PM2.5 is a lower fraction of PM10. We do
not have an explanation for this difference. Apparently, the
impact of local sources was stronger in Sweden, which is
consistent with the larger spatial variability of PM10 found in
Sweden. The difference is not because of outliers, since
median and mean ratios were similar. Neither differences in
sites nor distances to major roads explain the difference. The
correlation found in our study between PM10 and PM2.5
absorbance was very high, with r nearly 1 for the background
as well as for the traffic sites. Also, the absorption coefficient
values measured from PM2.5 filters were only slightly lower
than the values of absorption coefficients from PM10 filters
(4% lower in Germany and Sweden, 3% in The Netherlands). Similar results were found by Fischer et al. (2000) for
the urban area of Amsterdam. This is consistent with the size
distribution of elemental carbon (Chow, 1995) and it
supports the conclusion that PM2.5 absorbance is capturing
almost all of the absorbance of PM. As the health effects of
resuspended dust may differ from the health effects of
particulates derived from combustion processes (Schwartz
et al., 1996; Laden et al., 2000), it is necessary to differentiate
between the two components of particulate matter. Measurements of the blackness of the filters would be a potential way
to extract quantitative information on the composition of the
particles. As a consequence of our study, in studies using
absorption coefficient as a proxy for diesel soot, it is
unnecessary to measure both PM10 and PM2.5 filters.
Furthermore, it has been shown that measuring the
reflectance of PM10 or TSP (total suspended mater) filters
to estimate the relative concentration of combustion-derived
particles could be useful for retrospective exposure assessment (Penttinen et al., 2000). The quantitative relation
between EC and absorption coefficient would, however not
hold for other time periods, sites, and filter types. Large
differences between the reflectance of simultaneously measured particle concentrations on a membrane and paper filter
have been described before (Hoek et al., 1997b).
In conclusion, we have documented quantitatively that
measuring the reflectance of PM2.5 filters is a better proxy for
capturing the spatial variation in long-term average concentrations of combustion-derived particles than is a measurement of PM2.5 mass. The annual average absorbance
calculated from the reflectance method correlated well with
the annual average EC measurements sampled at the same
site. The estimation of reflectance is simpler, cheaper, and
more precise than the determination of EC and also has the
advantage of being non-destructive. There were however
differences in the EC/absorbance relation between countries
and site type (background/traffic), documenting the need for
141
Cyrys et al.
site-specific calibrations. The high correlation found between
PM10 and PM2.5 absorbance and negligible difference in
absolute levels of the absorption coefficient values measured
from PM2.5 filters and PM10 filters documents that for
reflectance measurements the (upper) size cut is not critical.
References
Bayram H., Devalia J.L., Sapsford R.J., Ohtoshi T., Miyabara Y.,
Sagai M., and Davies R.J. The effect of diesel exhaust particles on
cell function and release of inflammatory mediators from human
bronchial epithelial cells in vitro. J Respir Cell Mol Biol 1998: 18:
441–448.
Brunekreef B., Janssen N.A.H., Hartog de J., Harssema H., Knape M.,
and Vliet van P. Air pollution from truck traffic and lung
function in children living near motorways. Epidemiology 1997: 8:
298–303.
Chow J. Measurement methods to determine compliance with ambient
air quality standards for suspended particles. J Air Waste Manage
Assoc 1995: 45: 320–382.
Ciccone G., Forastiere F., Agabiti N., Biggeri A., Bisanti L., Chelline
E., Corbo G., Dell’Orco V., and Dalmasso P. Road traffic and
adverse respiratory effects in children. OccupEnviron Med 1998: 55:
771–778.
Countant B., and Stetzer S. Evaluation of PM2.5 speciation sampler
performance and related sample collection and stability issues: final
report. US Environmental Protection Agency, Research Triangle
Park, NC, 2001 Office of Air Quality Planning and Standards:
(Report No. EPA-454/R-01-008) (http://www.epa.gov/ttn/amtic/
pmspec.html).
Cyrys J., Heinrich J., Brauer M., and Wichmann H.E. Spatial
variability of acidic aerosols, sulfate and PM10 in Erfurt,
Eastern Germany. J Expos Anal Environ Epidemiol 1998: 8 ( 4 ):
447–464.
Delumyea R.G., Chu L.C., and Macias E.S. Determination of elemental
carbon component of soot in ambient samples. Atmos Environ 1980:
1: 647–652.
Diaz-Sanchez D. The role of diesel exhaust particles and their associated
polyaromatic hydrocarbons in the induction of allergic airway
disease. Allergy 1997: 52 ( 38 Suppl ): 52–56.
Duhme H., Weiland S.K., Keil U., Kraemer B., Schmid M., Stender
M., and Chambless L. The association between self-reported
symptoms of asthma and allergic rhinitis and self-reported traffic
density on street of residence in Adolescents. Epidemiology 1996: 7
( 6 ): 578–582.
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.
Fischer A., Gehrig R., and Hofer P. 1997. Russmesungen in der
Auenluft, Methodik und Resultate. Schriftenreihe Umwelt-Materialien Nr. 80 Luft. Bundesamt für Umwelt, Wald und Landschaft
(BUWAL), Bern.
Fischer P.H., Hoek G., van Reeuwiijk H., Briggs D.J., Lebret E.,
van Wijnen J.H., Kingham S., and Elliott P.E. Traffic-related
differences in outdoor and indoor concentrations of particles and
volatile organic compounds in Amsterdam. Atmos Environ 2000: 34:
3713–3722.
Gray H.A., and Cass G.R. Source contributions to atmospheric fine
carbon particle concentrations. Atmos Environ 1998: 32 ( 22 ):
3805–3825.
Hamilton R., and Mansfield T. Atmos Environ 1991: 25A: 715–723.
142
Traffic-related particle indicators
Hies T., Treffeisen R., Sebald L., and Reimer E. Spectral analysis of air
pollutants. Part 1: elemental carbon time series. Atmos Environ 2000:
34: 3495–3502.
Hoek G., Forsberg B., Borowska M., Hlawiczka S., Vaskövi E.,
Welinder H., Braniss M., Benes I., Kotesevec F., Hagen L.O., Cyrys J.,
Jantunen M.J., Roemer W., Brunekreef B. Wintertime PM10 and
Black Smoke concentrations across Europe: results from the PEACE
study. Atmos Environ 1997a: 31: 3609–3622.
Hoek G., Meliefste K., Cyrys J., Lewné M., Bellander T., Brauer B.,
Fischer P., Gehring G., Heinrich J., Vliet vad P., and Brunekreef B.
Spatial variability of fine particle concentrations in three European
countries. Atmos Environ 2002: 36: 4077–4088.
Hoek G., Welinder H., Vaskovi E., Ciacchini G., Manalis N., Royset
O., Reponen A., Cyrys J., and Brunekreef B. Interlaboratory
comparison of PM10 and Black Smoke measurements in the PEACE
study. Atmos Environ 1997b: 31: 3341–3349.
Horvarth H. Atmospheric light absorption-a review. Atmos Environ
1993: 27A: 293–317.
ISO. Ambient air-determination of a black smoke index. International
Organization for Standarization. International Standard 9835-1993
(E), 1993.
Janssen N.A.H., van Mansom D.F.M., van Jagt K., Harssema H., and
Hoek G. Mass concentration and elemental composition of airborne
particulate matter at street and background locations. Atmos Environ
1997: 31: 1185–1193.
Janssen N.A.H., van Vliet P., van Aarts F., Harssema H., and
Brunekreef B. Assessment of exposure to traffic-related air pollution
of children attending schools near motorways. Atmos Environ 2001:
35: 3875–3884.
Keeber G. The source of aerosols elemental carbon at Allerghery
mountain. Atmos Environ 1990: 24A: 2795–2805.
Kinney P.L., Aggarwal M., Northridge M.A., Janssen N.A.H., and
Shepard P. Personal exposure to PM2.5 and diesel exhaust particles
on Harlem sidewalks. Environ Health Perspect 2000: 108:
213–218.
Krämer U., Koch T., Ranft U., Ring J., and Behrendt H. Trafficrelated air pollution is associated with atopy in children living in
urban areas. Epidemiology 2000: 11: 64–70.
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 ( 10 ):
941–947.
Marple V.A., Rubow K.L., Turner W., Spengler J.D. Low flow rate
sharp cut impactors for indoor air sampling: design and calibration.
J Air Pollut Control Assoc 1987: 37: 1303–1307.
Monn Ch., Carabias V., Junker R., Waeber R., Karrer M., and Wanner
H.U. Small-scale spatial variability of particulate matter
o10 mm (PM10) and nitrogen dioxide. Atmos Environ 1997: 31:
2243–2247.
Nakai S., Nitta H. and Maeda K. Respiratory health associated with
exposure to automobile exhaust. II. Personal NO2 Exposure Levels
According to Distance from the Roadside. J Expos Anal Environ
Epidemiol 1995: 5 ( 2 ): 125–136.
Nielsen T. Traffic contribution of polycyclic aromatic hydrocarbons in
the center of a large city. Atmos Environ 1996: 20: 3481–3490
Nyberg F., Gustavsson P., Järup L., Bellander T., Berglind N.,
Jakobsson R., and Pershagen G. Urban air pollution and lung
cancer in Stockholm. Epidemiology 2000: 11 ( 5 ): 487–495.
OECD. Method of Measuring Air Pollution. Organisation for Economic
Co-operation and Development, Paris, 1964.
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.
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Traffic-related particle indicators
Penttinen P., Alm S., Ruuskanen J., and Pekkanen J. Measuring
reflectance of TSP-filters for retrospective health studies. Atmos
Environ 2000: 34: 2581–2586.
Schmid H., Laskus L., Abraham H.J., Baltensperger U., Levenchy V.,
Bizjak M., Burba P., Cachier H., Crow D., Chow J., Gnauk V.,
Even V., ten Brink H.M., Giesen K.P., Hitzenberger R., Hueglin C.,
Maenhaut W., Pio C., Carvalho V., Putaud J.P., Toom-Sauntry D.,
and Puxbaum H. Results of the ‘‘carbon conference’’ international
carbon round robin test stage I. Atmos Environ 2001: 35: 2111–2121.
Schwartz J., Dockery D.W., Neas L.M. Is daily mortality associated
specifically with fine particles? J. Air Waste Manage Assoc 1996: 46:
927–939.
Tolocka M.P., Solomon P.A., Mitchell W., Norris G.A., Gemmill D.B.,
Wiener R.W., Vanderpool R.W., Homolya J.B., and Rice J. East
versus West in the US: chemical characteristics of PM2.5 during
Winter of 1999. Aerosol Sci Technol 2001: 34: 88–96.
Ulrich E. and Israel G.W. Diesel soot measurement under traffic
conditions. J Aerosol Sci 1992: 1 ( Suppl. 23 ): S925–928.
Journal of Exposure Analysis and Environmental Epidemiology (2003) 13(2)
Cyrys et al.
VDI. Verein Deutscher Ingenieure, 2465, Part 1: Measurements of Soot
(Immission) F Chemical Analysis of Elemental Carbon by
Extraction and Thermal Desorption of Organic Carbon. Bueth,
Berlin, 1996.
Vliet van P., Knape M., Hartog de J., Janssen N., Harssema H., and
Brunekreef B. Motor vehicle exhaust and chronic respiratory
symptoms in children living near motorways. Environ Res 1997:
74: 122–132.
Weiland S.K., Mundt K.A., Rueckmann 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., Nicola T., von LoeffelholzColberg E., von Mutius E. Road traffic and adverse effects on
respiratory health in children. Br Med J 1993: 307: 596–600.
Wolff P., Korsog P., Stroup D., Ruthkorky M., and Morrissey M.
The influence of local and regional sources on the concentrations of
inhalable particulate matter in South-eastern Michigan. Atmos
Environ 1985: 19: 305–313.
143