ArticlePDF Available

Cancer risk assessment of selected hazardous air pollutants in Seattle


Abstract and Figures

The risk estimates calculated from the conventional risk assessment method usually are compound specific and provide limited information for source-specific air quality control. We used a risk apportionment approach, which is a combination of receptor modeling and risk assessment, to estimate source-specific lifetime excess cancer risks of selected hazardous air pollutants. We analyzed the speciated PM(2.5) and VOCs data collected at the Beacon Hill in Seattle, WA between 2000 and 2004 with the Multilinear Engine to first quantify source contributions to the mixture of hazardous air pollutants (HAPs) in terms of mass concentrations. The cancer risk from exposure to each source was then calculated as the sum of all available species' cancer risks in the source feature. We also adopted the bootstrapping technique for the uncertainty analysis. The results showed that the overall cancer risk was 6.09 x 10(-5), with the background (1.61 x 10(-5)), diesel (9.82 x 10(-6)) and wood burning (9.45 x 10(-6)) sources being the primary risk sources. The PM(2.5) mass concentration contributed 20% of the total risk. The 5th percentile of the risk estimates of all sources other than marine and soil were higher than 110(-6). It was also found that the diesel and wood burning sources presented similar cancer risks although the diesel exhaust contributed less to the PM(2.5) mass concentration than the wood burning. This highlights the additional value from such a risk apportionment approach that could be utilized for prioritizing control strategies to reduce the highest population health risks from exposure to HAPs.
Content may be subject to copyright.
Cancer risk assessment of selected hazardous air pollutants in Seattle
Chang-fu Wu
, Szu-ying Wu
, Yi-Hua Wu
, Alison C. Cullen
, Timothy V. Larson
John Williamson
, L.-J. Sally Liu
Department of Public Health, National Taiwan University, Taipei, Taiwan
Institute of Environmental Health, Institute of Occupational Medicine and Industrial Hygiene, National Taiwan University, Taipei, Taiwan
Evans School of Public Affairs, University of Washington, Seattle, WA, USA
Department of Civil and Environmental Engineering, University of Washington, Seattle, WA, USA
Air Quality Program, Washington State Department of Ecology, Bellevue, WA, USA
Department of Environmental and Occupational Health Sciences, University of Washington, Seattle, WA, USA
Institute of Social & Preventive Medicine, University of Basel, Switzerland
abstractarticle info
Article history:
Received 19 June 2008
Accepted 24 September 2008
Available online 8 November 2008
Risk assessment
Source apportionment
Diesel exhaust
Air toxics
The risk estimates calculated from the conventional risk assessment method usually are compound specic
and provide limited information for source-specic air quality control. We used a risk apportionment
approach, which is a combination of receptor modeling and risk assessment, to estimate source-specic
lifetime excess cancer risks of selected hazardous air pollutants. We analyzed the speciated PM
and VOCs
data collected at the Beacon Hill in Seattle, WA between 2000 and 2004 with the Multilinear Engine to rst
quantify source contributions to the mixture of hazardous air pollutants (HAPs) in terms of mass
concentrations. The cancer risk from exposure to each source was then calculated as the sum of all available
species' cancer risks in the source feature. We also adopted the bootstrapping technique for the uncertainty
analysis. The results showed that the overall cancer risk was 6.0910
, with the background (1.6110
diesel (9.8210
) and wood burning (9.45 10
) sources being the primary risk sources. The PM
concentration contributed 20% of the total risk. The 5th percentile of the risk estimates of all sources other
than marine and soil were higher than 1 10
. It was also found that the diesel and wood burning sources
presented similar cancer risks although the diesel exhaust contributed less to the PM
mass concentration
than the wood burning. This highlights the additional value from such a risk apportionment approach that
could be utilized for prioritizing control strategies to reduce the highest population health risks from
exposure to HAPs.
© 2008 Elsevier Ltd. All rights reserved.
1. Introduction
In conventional risk assessment studies of air pollutants, the
estimated cancer risk, i.e. the additional risk of developing cancer
due to continuous lifetime exposures to carcinogenic compounds,
is calculated as the product of unit risk and the exposed concen-
tration (USEPA, 2005a). The concentration estimates are usually
based on measurements of individual hazardous air pollutants
(HAPs) and the risk estimates are compound specic. For example,
using measurements at 25 air monitoring sites in Minnesota, Pratt
et al. (2000) estimated that the cancer risks of 16 pollutants ranged
between 4.710
and 11.010
. Similarly, Tam and Neumann,
(2004), utilizing measurements at ve air monitoring sites in Port-
land, OR, calculated the cancer risks for 43 HAPs. They showed
that 17 HAPs exceeded the cancer risk level of 1 10
at all sites,
with carbon tetrachloride, 1,3-butadiene, formaldehyde, and 1,1,2,2-
tetrachloroethane contributing 50% of the total lifetime cancer risks
From the risk reduction viewpoint, compound-specicrisk esti-
mates provide limited information for source-specic air quality
management, due to contributions of multiple sources to each
compound. For example, formaldehyde was identied as one of the
main risk contributors in many regions (e.g. Sax et al., 2006; Woodruff
et al., 2000). However, in an urban environment it comes from a
multitude of sources such as diesel exhaust, wood burning smoke,
industrial activities, photochemical products, etc. For developing
effective control strategies for sources imposing the highest health
risks, source-specicrisk estimates would provide more indicative
information. The US National Scale Air Toxics Assessment (NATA)
study adapted this approach with modeled ambient concentrations as
the main inputs for generating source-specic population exposure
and risk estimates with regard to major, area, and mobile sources. It
was estimated that the national-wide total risk was 4.1510
the greatest contribution from on-road mobile sources (USEPA, 2006).
The NATA approach depends largely on emission inventory data and
various models without actual measurements. Errors propagating
Environment International 35 (2009) 516522
Corresponding author. National Taiwan University, Room 717, No.17, Xu-Zhou Rd.,
Taipei 100, Taiwan. Tel./fax: +886 2 3322 8096.
E-mail address: (C.-f. Wu).
0160-4120/$ see front matter © 2008 Elsevier Ltd. All rights reserved.
Contents lists available at ScienceDirect
Environment International
journal homepage:
along the modeling processes could heavily inuence the risk
estimates. In addition, these risk estimates were limited to those
pre-specied source categories available from the emission inventory
(Cook et al., 2007; USEPA, 2006).
An alternative method less constrained by the aforementioned
limitations is the risk apportionment approach (Mukerjee and Biswas,
1992), which is a combination of risk assessment and receptor modeling
using actual measurements. Receptor models estimate contributions
from individual sources in terms of mass concentrations. Mukerjee and
Biswas (1992) made the initial demonstration by using the Chemical
Mass Balance model to apportion six samples of total suspended
particulate matters collected at a receptor site near several industrial
sources. The source-specic risks were calculated by summing the
species-specic risks for each source prole including paved road dust
), blast furnace (6.810
) and sinter plants (3.410
Our previous source apportionment study in Seattle, WA (Wu et al.,
2007) estimated that wood burning smoke and secondary sulfate
contributed equally to the ne particulate matters (PM
, aerody-
namic particle diameter b2.5 µm) mass concentrations (24%), higher
than that from diesel exhaust (10%). However, the source apportion-
ment results alone do not provide information on health hazards of
these sources. In this study, we expanded and improved the risk
apportionment approach to estimate source-specic lifetime exces-
sive cancer risks of selected hazardous air pollutants in Seattle, WA.
We calculated source-specic concentrations of PM
and volatile
organic compounds (VOCs) using the Multilinear Engine (ME) model
without a priori source proles (Paatero, 1999). Cancer risks and the
associated uncertainties were further calculated using bootstrapping,
with an emphasis on health risks of the diesel particulate matters
2. Materials and methods
2.1. Sample collection and source apportionment
Speciated PM
and VOCs samples were collected at the Beacon
Hill monitoring site in Seattle from 2000 to 20 04 (N=268). Beacon Hill
is an urban-scale semi-residential site, located within 2 km of two
major interstate freeways and arterial roads, as well as within 4 km of
a warehousing area and a major seaport. It represents average PM
concentrations in a typical Seattle residential neighborhood (Goswami
et al., 2002). The PM
samples were collected using the IMPROVE
method and analyzed for mass, trace elements, sulfate (SO
), nitrate
), organic carbon (OC), and elemental carbon (EC). The VOCs,
collected using the DNPH cartridge and the SUMA canister (Wu et al.,
2007), included formaldehyde, acetaldehyde, Benzene, 1,3-Butadiene,
Carbon Tetrachloride, Chloroform, Tetrachloroethylene, and Trichlor-
oethylene. These VOCs were on the list of the top 33 urban HAPs
(USEPA,1999) and were chosen for monitoring due to the expected high
concentrations and health risks in the Seattle area.
The source-specic concentrations of each species were calculated
in our previous study using the ME model (Wu et al., 2007). In the ME
Table 1
Modeled and measured mass concentrations (ng/m
) of each species
Element Base data Bootstrap Unit risk
Abbr Measured Modeled Diff (%)
Median p75
Organic carbon OC 2561.9 2475.3 3.4 1443.4 2312.8 3250.0
Elemental carbon EC 612.9 585.6 4.5 407.7 551.9 720.8
Ammonium NH
456.1 450.5 1.2 310.6 400.2 548.9
Arsenic As 1.0 1.3 33.0 0.7 1.1 1.8 4.3E06
Bromine Br 2.0 2.4 19.9 1.8 2.4 3.1
Calcium Ca 26.7 25.1 6.1 19.6 24.8 31.2
Chlorine Cl 60.2 59.3 1.5 51.1 58.9 65.6
Chromium Cr 1.6 2.3 42.5 1.3 1.9 2.7 2.4E06
Copper Cu 4.3 4.4 1.9 3.1 4.1 5.4
Iron Fe 51.6 49.9 3.3 30.8 45.1 63.7
Lead Pb 3.7 5.2 38.6 2.7 4.7 6.9 1.2E08
Manganese Mn 3.0 3.9 30.4 2.4 3.3 4.7
Nickel Ni 2.3 2.4 4.4 1.3 2.0 2.8 4.8E07
Nitrate nonvolatile NO
459.3 460.1 0.2 357.3 425.6 544.6
Potassium ion K 29.5 38.5 30.6 25.1 36.5 49.1
Silicon Si 42.1 40.6 3.4 28.0 39.3 55.2
Sodium ion Na 151.8 147.6 2.8 117.7 136.9 169.4
Sulfate SO
1191.2 1188.9 0.2 831.5 1090.8 1446.7
Tin Sn 5.1 7.7 51.8 4.7 7.5 10.2
Titanium Ti 2.6 3.6 36.1 2.5 3.5 4.5
Vanadium V 3.5 3.9 12.2 2.3 3.2 4.6
Zinc Zn 8.8 8.3 6.1 5.5 8.0 10.8
Sum 5681.2 5566.5 2.0 3651.0 5164.5 7002.6
Acetaldehyde Ace 1448.4 1297.0 10.5 711.4 1171.1 1652.1 2.2E09
Formaldehyde For 1505.8 1257.6 16.5 595.8 1116.3 1698.1 1.3E08
Benzene Ben 1329.7 1123.6 15.5 757.4 1071.7 1443.7 7.8E09
1,3-Butadiene But 119.1 147.0 23.4 101.1 139.4 183.5 3.0E08
Chloroform Chl 237.8 234.6 1.4 149.5 215.9 287.1 2.3E08
Carbon tetrachloride Car 641.2 634.9 1.0 462.4 602.4 752.8 1.5E08
Tetrachloroethylene Tet 182.4 179.2 1.8 124.3 172.0 227.6 5.6E09
Trichloroethylene Tri 169.2 208.0 23.0 146.8 197.2 248.2 2.0E09
Sum 5633.6 5081.7 9.8 3048.7 4685.9 64 93.1
Diff%=(modeled concentrationmeasured concentration) /measured concentration 100%.
p25 and p75 represent the 25th and 75th percentiles calculated from the bootstrapping process.
Reference from IRIS, website:
Reference from Cal EPA, website:
Reference from IRIS (assuming that 20% of all atmospheric chromium is hexavalent).
Reference from IRIS (unit risk for nickel subsulde).
517C.-f. Wu et al. / Environment International 35 (2009) 516522
Fig. 1. Source feat ures from the ME mod el with the error estimates. The y-axis is the mas s fraction (%) o f each species in the feature p role. The solid bar represents the point
estimates from the base factor. The upper and lower li ne and circle sy mbol represen ts the 75th percentile, 25th percentiles and median calculated fromthebootstrapping
518 C.-f. Wu et al. / Environment International 35 (2009) 516522
model, individual species measured in each air sample is expressed as
the sum of contributions from individual sources as below:
xij =
fjkgki +eij ð1Þ
where x
is the jth species concentration measured in the ith sample,
is the source contribution dened as the mass concentration from
the kth source contributing to the ith sample, f
is the source feature
dened as the jth PM species or VOCs mass fraction from the kth
source, e
is the error term associated with the jth species con-
centration measured in the ith sample, and pis the total number of
independent sources. The main model outputs contain the species
composition of each source (i.e. f
or source features) and the overall
contributions of each source (i.e. g
). We used the missing mass, i.e.,
the concentration of the measured total particle mass minus the sum
of all analyzed species, as an additional variable and implemented an
auxiliary equation to constrain the sum of all species mass fractions
to be 100% (Larson et al., 2006; Wu et al., 2007). This 100% constraint
prevents any serious under- or overestimation in the subsequent risk
Using both PM
and VOCs measurements in the ME model, we
identied ten features, including wood burning, secondary sulfate,
nitrate, gasoline, diesel, fuel oil, aged sea salt, soil, marine, and the
urban background concentration of VOCs (i.e., other)(Wu et al.,
2007). We found a better model t for the PM
= 0.88) than for the
=0.72) portion. The major PM
emission sources included
wood burning (24%), secondary sulfate (24%) and nitrate (20%). The
majority of the measured vapor phase HAPs came from the general
urban background (26%), wood burning (14%), and diesel exhaust
In this study, we further calculated the mean source specic
concentrations of individual HAPs (x
) as follows:
=fjk ×gkð2Þ
where x
is the concentration of the jth species from the kth source,
is the source feature in Eq. (1), and gkis the mean contribution from
the kth source.
2.2. Risk assessment
With x
, the cancer risk from exposure to the kth source (R
) can
be calculated as the sum of cancer risks of all available species in its
source feature:
jk ×Unit Riskj
 ð3Þ
Unit risk values for each species were taken rst from the
Integrated Risk Information System (IRIS) (USEPA, 2005b). When the
IRIS unit risk of a specic species is not available, the Unit Risk
Estimate (URE) provided by the California Environmental Protection
Agency (CalEPA) was used (CalEPA, 2005). Species without any unit
risk values were not included in the risk assessment process. The unit
risk of chromium was adjusted by multiplying a factor of 0.2, assuming
that approximately 20% of the ambient airborne chromium was in the
toxic hexavalent form (Bell and Hipfner, 1997). The unit risk for nickel
was adopted from the unit risk for nickel subsulde in IRIS, assuming
that all the monitored nickel was in the insoluble and carcinogenic
2.3. Uncertainty analysis
As the uncertainty estimates of source features and source
contributions were not provided by the standard ME model, we
used the bootstrapping technique (Eforn and Tibshirani, 1993)to
obtain the uncertainties. The bootstrapping involves creating multiple
sets of subsamples and requires no prior statistical assumptions about
the underlying distribution of the dataset. Each set of the subsamples
is generated from re-sampling the data with replacement (i.e. any data
point could be sampled multiple times or not at all). The boot-
strapping process was executed with the SAS statistical software
(Version 8.02, SAS Institute Inc., Cary, NC, USA) in this study. To obtain
reliable results, 300 sets of different input data were generated with
299 sets of resampled data and one basedata without any replace-
ment (Eberly, 2005). Each dataset with 268 samples was analyzed
with the ME model, resulting in 299 sets of bootstrapping solutions as
well as one set of basesolutions. While the physical meaning of each
source feature in the base solutions was interpreted individually in
Wu et al. (2007), it is not practical to do so for all 299 bootstrapping
solutions in this study. Thus, these bootstrapping features were
matched or mappedto the base features automatically through cor-
relating the time series of the source contributions in the boot-
strapping results to those in the base results. The pair with the highest
correlation coefcient was retained, requiring that coefcient to be
larger than 0.6 (Eberly, 2005).
3. Results and discussion
The measured and modeled concentrations of each species were shown in Table 1.
Results shown in the Modeledcolumn represent the reconstructed concentrations
estimated from the base data while results in the Bootstrapcolumns represent the
statistics from the 299 sets of reconstructed concentrations through the bootstrap
resampling processes. The sums of the PM
and VOCs mass concentrations were
underestimated by only 2% and 10%, respectively. For individual species, the absolute
percent difference between the measured and modeled values ranged between 0 and
52%, with the larger differences (i.e. arsenic, chromium, lead, manganese, potassium, tin,
and titanium) due mostly to the higher percentages of samples below detection limit
(N20%) for these species. These biases would propagate into the following risk
assessment process. However, the bootstrapping results indicated that the measured
concentrations were within the 25th to 75th percentile of the bootstrapping range
(Table 1), suggesting that the variation from the ME modeling process wasmostly taken
into account in this uncertainty analysis.
The source features from the base data and the associated uncertainties calculated
from the bootstrapping process are shown in Fig. 1. During the mapping process, most
Table 2
Point estimates of the cancer risk (per million) for each source by the risk apportionment approach
Sources As Cr Ni Pb Ace Ben But Car Chl For Tet Tri Sum PM risk Sum VOCs risk Sum risk
Other 0.74 1.60 0 0.01 0.76 1.76 1.43 4.19 2.49 2.63 0.25 0.17 2.36 13.7 16.1
Diesel 0.33 2.17 0.32 0.01 0.2 2.18 0.96 0.31 0 3.17 0.13 0.05 2.83 6.99 9.82
Wood 3.17 0 0 0.02 0.37 2.08 1.07 0.6 0.36 1.64 0.12 0.03 3.18 6.27 9.45
Aged 0.32 0.44 0.02 0.01 0.26 0.08 0.44 1.69 0.7 1.79 0.07 0.08 0.79 5.11 5.90
Fuel 0.4 0.45 0.61 0.004 0.68 0.27 0.15 0.24 0.27 2.64 0.006 0.02 1.47 4.26 5.73
Sulfate 0.31 0 0.14 0.0 04 0.18 0.26 0.2 1.24 0.78 2.10 0.11 0.04 0.45 4.91 5.36
Nitrate 0.24 0.01 0 0.002 0.09 1.08 0.17 0.09 0.18 1.01 0.006 0 0.25 2.63 2.87
Gasoline 0 0.1 0 0.001 0.27 0.65 0 0.69 0.1 0.73 0.26 0 0.1 2.70 2.80
Soil 0 0.44 0 0 0.04 0.26 0 0.06 0.45 0.64 0.017 0.009 0.44 1.47 1.91
Marine 0.03 0.19 0.04 0 0.004 0.15 0 0.41 0.07 0 0.03 0.02 0.27 0.68 0.94
Sum 5.55 5.40 1.13 0.062 2.85 8.77 4.42 9.52 5.4 16.35 1 0.42 12.14 48.72 60.88
519C.-f. Wu et al. / Environment International 35 (2009) 516522
sources were mapped successfully to one of the base features at least more than 280
times with an rN0.6, while soil and othersources were mapped only 220 and 267
times, respectively, implying higher uncertainties for these two source features. Ideally,
the range (25th to 75th percentile) of the error estimates should cover the results from
the base data and the median from bootstrapping factor should be close to that from the
base data. In this study, although the source features from the base dataset may not
always agree well with those from the bootstrapping dataset, good agreement was
found for most important marker species that had high contributions in each source
Fig. 2. Source contributions to a) PM
and b) VOCs mass concentrations (data from Wu et al., 2007). Risk estimates by source contributions for c) PM
, d) VOCs, and e) the sum of
and VOCs.
520 C.-f. Wu et al. / Environment International 35 (2009) 516522
feature (Fig. 1, e.g. the VOCs fraction in othersource; SO
and NH
in secondary sulfate
source; V and Ni in fuel source; Fe and Mn in diesel source; and the Si and Ca in soil
The unit risks were available for As, Cr, Pb andNi in the PM
fraction and for all the
monitored VOCs (Table 1). The point estimates of excessive cancer risks from the base
data were summarized in Table 2 by source and species. The source specic risk values
ranged from 1.610
to 9.410
. All sources except for marinegave a sum of cancer
risks higher than 1 10
.Other, diesel and wood burning sources were the primary risk
sources in both PM
and VOCs fractions (Fig. 2). The sum of cancer risks were 6.110
with the PM
portion contributing 20% (1.210
). Formaldehyde, carbon tetrachloride
and benzene posed the highest cancer risks as estimated in other studies (Morello-
Frosch et al., 2000; Tam and Neumann, 2004). The overall risk value (11.810
estimated in the 1999 NATA study at the census tract of the Beacon Hill site was 2 times
higher than the overall risk in this study. This is mainly because more species (85 species
vs. 12 in this study) were simulated in the NATA study. Compared to the risk estimates
based on the measured ambient concentrations of 19 species in other cities (Los Angeles,
CA: 1.010
; New York City, NY: 1.3 10
), the overall risk in this study was still lower
than theirs (Sax et al., 2006).
The main risk contributors to the othersource were t he VOCs fraction (13.7/16.110 0% =
85%) with carbon tetrachloride, formaldehydeandchloroformasthetopcontributors.The
othersource from the ME model was considered as the background concentrations of
VOCs since the contribution for eachVOC was high in the source feature and the reconstructed
time series did not show obvious temporal variation (Wu et al., 2007). In the NATA study, the
background concentration was dened as components attributable to long-range transport,
unidentied emission sources, resuspensionof historical emissions, and other natural sources.
NATA assigned background concentrations to 29 species (i.e. 27 gaseous HAPs including
those 8 VOCs in our study and mercury and diesel PM)on top of the modeled concentrations
(USEPA, 2006). The 1999 NATA study estimated a cancer risk of 2.3 10
for the back-
ground concentrations of HAPs in the Beacon Hill area, which is comparable to our estimate
The diesel and wood burning sources presented similar cancer risks (Fig. 2cd),
even though the diesel exhaust contributed less to the PM
mass concentration than
the wood burning (Fig. 2a). On the other hand, secondary sulfate, which was mainly
formed in southern Washington, along the Canadian border, and in southwestern
British Colombia, Canada (Kim et al., 2004), contributed equally to the PM
concentration as wood burning (Fig. 2a) but posed less health risks (Fig. 2c). The above
observations highlight the additional information from such a risk apportionment
approach that could be utilized for prioritizing control strategies to reduce the highest
population health risks from exposure to HAPs.
The uncertainties in risk assessment of each source were shown in Fig. 3, with the
point estimates displayed in lled circles. For other, sulfate, gasoline and soil sources,
the point estimates using the base dataset were close to the median of the distribution,
suggesting small uncertainties. Point estimates for fuel, aged sea salt, wood burning and
diesel sources were near the higher end of the bootstrapping results, while the point
estimates of marine and nitrate were near the lower end. Nevertheless, the 5th per-
centile of the risk estimates still showed that most sources gave the sum of risks higher
than 110
, except for marine and soil.
For risks of DPM, we also utilized the simulated concentrations of DPM from NATA's
study and the unit risk (310
)fromCalEPA (2008) to calculate the cancer risk of
DPM. The modeled exposure concentrations of DPM from the 1999 NATA study at the
BeaconHill census tract,and within the bufferzones of 1 km, 3 km and5 km were 2.17,1.83,
1.63 and 1.55 µg/m
, respectively. The apportioned DPM mass concentration in our study
was 0.7 µg/m
. Multiplyingthese modeled concentrations by the unitrisk of DPM resulted
in estimated cancer risks of 6.5210
, 5.4910
and 4.6610
, respectively.
The DPM risk (2.8310
,Table 2) from our risk apportionment approach was 2 orders of
magnitude lower. This is probably due to our inclusion of only four metal species and no
polycyclic aromatic hydrocarbons (PAHs) for DPM in risk calculations. Even after adding
risks fromthe DPM associated VOCs,risk from diesel exhaustwas still low (9.8210
). On
the other hand, the cancer risk estimate (2.0910
DPM sourcecontributions (0.7µg/m
) and the CalEPA's DPM unit risk wasclose to those in
the NATA study.
The discrepancy for the DPM risk estimates can be attributed to the following
reasons. First, the risk apportionment approach tends to underestimate because the
results depend on the number of species taken into account. Although 30 species have
been analyzed by the ME model in this study, the unit risks are available only for twelve
of them. Further, species that are not analyzed cannot be included in the risk
calculations. Second, apportioning the diesel exhaust involves considerable uncertain-
ties, as evidenced by the fact that the point estimate of the DPM cancer risks fell outside
the 25th and 75th percentile range (Fig. 3). Third, many toxicological studies have
showed that the overall chemical reactivity and toxicity of PM or DPM may be greater
than the sum of individual components (e.g. Pan et al., 2004). Thus it is likely that the
DPM risk estimate based on the sum of risks from individual species is lower than that
from using a single totalunit risk for DPM. Fourth, the validity of the CalEPA's DPM unit
risk itself is still being debated, even though that value is derived from epidemiological
data (CalEPA, 2008). The US EPA in its health assessment document concluded that “…
human exposure-response data [for diesel exhaust] are considered too uncertain to
derive a condent quantitative estimate of cancer unit risk(USEPA, 2002). Further-
more, many studies did not have adequate quantitative exposure data, did not control
for potential confounders, and might not be suitable for newly developed diesel engine
technology (Hesterberg et al., 2006).
One major limitation of the risk apportionment approach is that the estimated risk
values could be underestimated by the limited set of species considered and the
incomplete data of toxicity. The 12 species considered in our study were among the 33
most important urban HAPs based on emissions and toxicities in a 1995 ranking
analysis (USEPA, 1999). Other urban HAPs were not monitored because they are
considered less stable or lacked approved sampling and analytical methods (PSCAA,
2003). For example, a group of 7 PAHs, which are probable human carcinogens and are
used as a surrogate for the much more complex mixture of polycyclic organic matter in
the US EPA National Air Toxics Program (USEPA, 1999 ), are not monitored routinely at
the Beacon Hill site. In the 1999 NATA study, naphthalene was estimated to contribute
13.9% to the total risks, after subtracting the risks from the background concentrations
(USEPA, 2005b). Yu et al. (2008) also estimated that the mean cancer risk calculated
from 17 species of PAHs in an urban environment was 2.9 10
. It is possible to include
PAHs in the source apportionment models and then calculate their contributions to the
estimated cancer risks. One previous source apportionment study identied six sources
(i.e. oil, coal, gasoline, diesel, wood, and other) of PAHs in the urban atmosphere (Larsen
and Baker, 2003). Another limitation of this study is that the risk estimates pertained to
outdoor air at a stationary location. Indoor exposure sources might also have important
contributions (Payne-Sturges et al., 2004; Sax et al., 2006).
4. Conclusions
The cancer risks associated with the major sources at Beacon Hill
were calculated by the risk apportionment approach. These estimated
risks were calculated based on the assumption of lifetime exposures to
ambient pollutants. They are not actual risk values and are generally
regarded as for screening purposes and for preliminary estimation. We
further adopted the bootstrapping technique for the uncertainty
analysis. The results showed that most of the sources, other than marine
and soil, had cancer risks exceeding the acceptable level of 110
The sourcewith the greatest health concernwas the background source
which gave a cancer risk of 1.6110
(5th95th percentile: 1.2 10
), which requires concerted regional efforts to address. The
source-specic risks estimated in this study were likely to beunderesti-
mated mainly because only a limited number of species was monitored.
This is more apparent for the DPM, for which usinga single unit risk that
is scientically derived and well examined might be more appropriate
for the risk calculation. For the other sources, the risk apportionment
approach might be the most feasible method since no source-specic
unit risks were developed. Despite the potential underestimation, the
estimated risks still exceeded the acceptable level, suggesting that the
HAPs in this region are worth further investigation. Our risk apportion-
ment results can provide the guidance for future healthrisk managers to
design the risk reduction strategy more effectively.
This study was conducted as a collaborative effort among the U.S.
EPA Region X, the Washington State Department of Ecology (the
Fig. 3. The risk estimates for each source. The box plot represents the 25th percentile,
median, and 75th percentile of the risk estimates. The left and right line represents the
10th and 90th percentiles, and the square symbols represent the 5th and 95th
percentiles. The circle symbols represent the risk estimates from the base factors.
521C.-f. Wu et al. / Environment International 35 (2009) 516522
Ecology), the Puget Sound Clean Air Agency, the Washington State
University, and the University of Washington. This study was funded
by the Ecology under a cooperative agreement with the Washington
Cooperative Fish & Wildlife Research Unit. This study was also
partially funded by the U.S. Environmental Protection Agency through
its Ofce of Research and Development under EPA Grant R827355.
Bell RW, Hipfner JC. Airborne hexavalent chromium in southwestern Ontario. J Air
Waste Manage Assoc 1997;47:90510.
CalEPA. Air toxics hot spots program risk assessment guidelines: part II technical
support document for describing available cancer potency factors. Ofce of
Environmental Health Hazard Assessment; 2005.
CalEPA. Findings of the scientic review panel on the report on diesel exhaust, http://; accessed June 18, 2008.
Cook R, Strum M, Touma JS, Palma T, Thurman J, Ensley D, et al. Inhalation exposure
and risk from mobile source air toxics in future years. J Expo Sci Environ Epidemiol
Eberly S. EPA PMF 1.1 user's guide, 27711. Research Triangle Park, NC: US Environmental
Protection Agency; 2005.
Eforn B, Tibshirani RJ. An introductionto the bootstrap. London: Chapmanand Hall; 1993.
Goswami E, Larson T, Lumley T, Liu LJS. Spatial characteristics of ne particulate matter:
identifying representative monitoring locations in Seattle, Washington. J Air Waste
Manage Assoc 2002;52:32433.
Hesterberg T, Bunn WB, Chase GR, Valberg PA, Slavin TJ, Lapin CA, et al. A critical
assessment of studies on the carcinogenic potential of diesel exhaust. Crit Rev
Toxicol 2006;36:72776.
Kim E, Hopke PK, Larson T V, Maykut NN, Lewtas J. Factor analysis of Seattle ne
particles. Aerosol Sci Technol 2004;38:72438.
Larsen RK, Baker JE. Source apportionment of polycyclic aromatic hydrocarbons in the urban
atmosphere: a comparison of three methods. Environ Sci Technol 2003;37:187381.
Larson TV, Covert DS, Kim E, Elleman R, Schreuder AB, Lumley T. Combining size
distribution and chemical species measurements into a multivariate receptor model
of PM
. J Geophys Res Atmos 2006:111.
Morello-Frosch RA, Woodruff TJ, Axelrad DA, Caldwell JC. Air toxics and health risks in
California: the public health implications of outdoor concentrations. Risk Anal
Mukerjee S, Biswas P. A concept of risk apportionment of air emission sources for risk
reduction considerations. Environ Technol 1992;13:63546.
Paatero P. The multilinear engine a table-driven, least squares program for solving
multilinear problems, including the n-way parallel factor analysis model. J Comput
Graph Stat 1999;8:85488.
Pan CJ, Schmitz DA, Cho AK, Froines J, Fukuto JM. Inherent redox properties of diesel
exhaust particles: catalysis of the generation of reactive oxygen species by
biological reductants. Toxicol Sci 2004;81:22532.
Payne-Sturges DC, Burke TA, Breysse P, Diener-West M, Buckley TJ. Personal exposure
meets risk assessment: a comparison of measured and modeled exposures and
risks in an urban community. Environ Health Perspect 2004;112:58998.
Pratt GC, Palmer K, Wu CY, Oliaei F, Hollerbach C, Fenske MJ. An assessment of air toxics
in Minnesota. Environ Health Perspect 2000;108:81525.
PSCAA.Final report: Puget soundair toxicsevaluation. Pudget SoundClear Air Agency; 2003.
Sax SN, Bennett DH, Chillrud SN, Ross J, Kinney PL, Spengler JD. A cancer risk assessment
of inner-city teenagers living in New York City and Los Angeles. Environ Health
Perspect 2006;114:155866.
Tam BN, Neumann CM. A human health assessment of hazardous air pollutants in
Portland, OR. J Environ Manag 2004;73:13145.
USEPA. National air toxics program: the integrated urban strategy. Fed Regist 1999;64
USEPA.Health assessment document for dieselengine exhaust. Washington,DC: National
Center for Environmental Assessment; 2002.
USEPA. Guidelines for carcinogen risk assessment EPA/630/P-03/001B. Risk assessment
forum,. Washington, DC: National Center for Environmental Assessment; 2005a.
USEPA. Integrated risk information system. Available from: b
USEPA. National-scale air toxics assessment for 1999 Available from: bhttp://www.epa.
Woodruff TJ, Caldwell J, Cogliano VJ, Axelrad DA. Estimating cancer risk from outdoor
concentrations of hazardous air pollutants in 1990. Environ Res 2000;82:194206.
Wu CF, Larson TV, Wu SY, Williamson J, Westberg HH, Liu LJS. Source apportionment of PM
and selected hazardous air pollutants in Seattle. Sci Total Environ 2007;386:4252.
Yu YJ, Guo HC, Liu Y, Huang K, WangZ, Zhan XY. Mixed uncertainty analysis of polycyclic
aromatic hydrocarbon inhalation and risk assessment in ambient air of Beijing.
J Environ Sci (China) 2008;20:50512.
522 C.-f. Wu et al. / Environment International 35 (2009) 516522
... If the concentration is greater than the MDL, then the uncertainty is calculated as [(error fraction × concentration) 2 + (0.5 × MDL) 2 ] 1/2 . According to Wu et al. (2009), the mean sourcespecific concentrations (x * jk ) of individual HAPs can be calculated as follows: ...
... In this study, potential health risks were evaluated only for chronic exposure to air toxics via inhalation. To assess these risks, we applied a point estimation approach for a screening-level health risk assessment that has been adopted in several studies Kindzierski 2017, 2018;Fox et al. 2004;Jia and Foran 2013;Xiong et al. 2020;Wu et al. 2009Wu et al. , 2011. The toxicity values applied in this study are presented in Tables S4-S6. ...
... Overestimation could occur because (i) more conservative values were applied to some calculations, (ii) mean values from one year measurements were used to calculate long-term (70-year lifespan) toxicity values, and (iii) 50% of the MDL value was substituted when measured values were less than the MDL. Many of these limitations have also been addressed in previous studies Kindzierski 2017, 2018;Jia and Foran 2013;Ramírez et al. 2012;Wu et al. 2009Wu et al. , 2011. In addition, there was a possibility of overestimation of health risks because we determined PAHs and HMs from TSP samples, which would be theoretically higher levels than those from inhalable PM samples (such as PM 10 or PM 2.5 ) (Cheruiyot et al. 2015;Chow 1995). ...
Full-text available
We conducted ambient monitoring of various hazardous air pollutants (HAPs) for 2 years (2013-2015) in two adjacent Korean cities in a megacity area: Seoul and Incheon. Measured HAPs included volatile organic compounds (VOCs), polycyclic aromatic hydrocarbons (PAHs), and heavy metals (HMs). The objectives of this study were to evaluate the spatiotemporal variations of HAPs, to prioritize HAPs based on health risks, to identify sources using a receptor-based model, and to estimate source-specific risks. Overall, the HAP levels in Incheon were higher than those in Seoul. The concentrations of combustion-origin HAPs, such as PAHs and some HMs, were significantly higher during the heating period than during the non-heating period. However, most VOCs exhibited an opposite trend. Benzo[a]pyrene showed the highest cancer risk in both cities, followed by formaldehyde, arsenic, and benzene; trichloroethylene was the only species that exceeded the hazard quotient of 1. Cumulative cancer risks were 2.0 × 10⁻⁴ in Seoul and 2.7 × 10⁻⁴ in Incheon. Major sources and their contributions to each HAP concentration were estimated by positive matrix factorization modeling. Based on source-specific risk assessments, we suggest that both cities should give high priority to the control of traffic pollution and the supply of cleaner fuels in non-residential sectors. Reducing carbonyl concentrations in Seoul and industrial emissions in Incheon is also necessary. Establishing new ambient standards for benzo[a]pyrene and formaldehyde is worth considering as a long-term measure. This study provides scientific information on the occurrence, health risks, and sources of various HAPs in large urban areas.
... However, to the best of our knowledge, little work has been done to address and quantify the risk assessment of VOCs in China (Du et al., 2014;Tong et al., 2019;Zhang et al., 2017), let alone source-risk apportionment, a method coupling source apportionment techniques and health risk assessment (Mukerjee and Biswas, 1992). Source-risk apportionment has been applied in limited studies worldwide Kindzierski, 2017, 2018a, b;Wu et al., 2009) but is far less understood in China . ...
... This result might be attributed to the implementation of strict controls on coal burning, solvent usage, and industry in Beijing, which are important benzene sources. The overall cancer risk, including various PM 2.5 elements, was 6.09 × 10 − 5 in Seattle between 2000 and 2004 (Wu et al., 2009). Bari et al. assessed the health risks in several places in Canada, and the cancer risks ranged from 8.2 × 10 − 8 to 4.1 × 10 − 6 (Bari and Kindzierski, 2017, 2018a, b), far below the tolerable risk. ...
... To determine the health impacts of emission sources, the point estimate approach was used for risk apportionment based on PMF-derived factors. This approach has been used in several studies of risk apportionment, including VOCs and PM 2.5 Kindzierski, 2017, 2018a, b;Khan et al., 2016;Wu et al., 2009). The carcinogenic risk and non-carcinogenic risk of each source are presented in Fig. 7. ...
Full-text available
The long-term and high temporal resolution measurement of ambient volatile organic compounds (VOCs) was continuously performed in urban Beijing in 2016. Historical VOC data were compiled and compared with our results. A positive matrix factorization (PMF) model was used to determine potential emission sources. The backward trajectories and the concentration, ozone formation potential (OFP), and secondary organic aerosol potential (SOAP) of VOCs in the air masses from different regions were used to estimate the VOC chemical behaviour and transport-direction contributions. A health assessment, including the carcinogenic risk and non-carcinogenic risk of specific hazardous VOC species, was performed, while the point estimate approach was used for risk apportionment based on emission sources. In this study, the average (±standard deviation, Std) concentration of total observed VOCs (TVOCs) was 101.5 ± 65.2 μg/m³, ranging from 21.4 to 439.1 μg/m³. The VOC concentration in Beijing had decreased but was still higher than that in other Chinese cities according to compiled reported VOC data. Combustion-related sources were the primary sources of VOC pollution in urban Beijing. VOC contributions were increasingly coming from the northwestern regions instead of solely from regions to the south of Beijing according to the VOC concentration, OFP, and SOAP data. A potential health risk from VOCs should be considered for residents living in Beijing. The results of this study are expected to provide basic data for efficient VOC emission management and public health protection in Beijing.
... Windblown Soil, Soil (3.7%; 2.57%): Dominated by Al, Si, Ca, Ti and Fe, the five key crustal matter elements in their oxide form, commonly associated with windblown soils (Wu et al., 2009;Amato and Hopke, 2012). The [Al/Si] ratio was 0.34 at both sites, which was in the range common for alumina-silicate clays (0.25-0.35). ...
... Mixed Industry with Aged Sea Air, IndSaged (14.0%; 14.4%): Dominated by Na, S and BC -indicative of Na 2 SO 4 most likely formed by the chemical reaction of sea spray with sulfate (Qin et al., 1997;Wu et al., 2009), sometimes referred to as aged sea air. Of the total measured Na, 77% (at Liverpool and 73% at Mascot) was allocated to this fingerprint. ...
... Fresh Sea Spray, Sea (8.8%; 13.6%): Dominated by Na and Cl, with small amounts of Br (indicative of fresh sea spray; e.g. Qin et al., 1997;Wu et al., 2009). 23% of the total Na and 99% of the total Cl was allocated to this fingerprint at Liverpool and at Mascot 25% of the Na and 99% of the Cl was allocated to this fingerprint. ...
Sampling of PM2.5 has been undertaken twice per week at the Liverpool and Mascot sites (in Sydney, Australia) since 1998. Ion Beam Analysis (IBA) was applied to each sample to determine the concentrations of 21 elements from hydrogen to lead and the black carbon concentration was determined using photon transmission techniques. Sampling days that displayed high and low airborne soil concentrations were identified and three distinct sets of soil fingerprints were determined using Positive Matrix Factorisation (PMF) source apportionment techniques. A fingerprint for all sampling days (representing the average soil fingerprint for each site), a fingerprint corresponding to low soil days associated with local retrained road dust and a fingerprint for high soil days associated with agricultural activities. The ratios of key soil elements (i.e. Si, Al, Fe) displayed larger temporal variation for the high soil days, whereas lower variation was observed for low (local) soil days. Furthermore, it was found that the El Niño-Southern Oscillation (ENSO) affected the concentration of windblown soil dust in the atmosphere. The average soil fingerprint, for all data, was heavily influenced by sampling days containing higher concentrations of soil dust, thus representing the dominant soil type. However, we did observe differences in the K/Fe and Ca/Si ratios to be a distinguishing factor between the average soil fingerprint and the high soil day fingerprint. The Soil fingerprint for the low soil concentration days had a large fraction of black carbon associated with vehicle emissions, represented retrained road dust.
... Some previous studies have coupled source apportionment (chemical mass balance, modified PMF, and multilinear engine model) with health risk assessment to obtain sourcespecific risk estimates for the atmospheric particle matters (Mukerjee and Biswas et al. 1992;Wu et al. 2009;Liao et al. 2015), soils (Huang et al. 2018a;Yang et al. 2019), and road dust (Men et al. 2020). Wu et al. (2009) evaluate the source-specific risk of contaminants in the particulate matters using the multilinear engine and carcinogenic risk assessment. ...
... Some previous studies have coupled source apportionment (chemical mass balance, modified PMF, and multilinear engine model) with health risk assessment to obtain sourcespecific risk estimates for the atmospheric particle matters (Mukerjee and Biswas et al. 1992;Wu et al. 2009;Liao et al. 2015), soils (Huang et al. 2018a;Yang et al. 2019), and road dust (Men et al. 2020). Wu et al. (2009) evaluate the source-specific risk of contaminants in the particulate matters using the multilinear engine and carcinogenic risk assessment. Men et al. (2020) calculated the source-specific risks combining the risk assessment process with PMF model, in which the particle size was not taken into account in the source-specific risk evaluation. ...
... Similarly, Cao et al. (2016) showed that Cr had the highest contribution to the carcinogenic risk of children despite the lower concentration of Cr in the environmental media in the industrial city. These results indicated that a minor contributor to the accumulation of heavy metals does not necessarily posed a least health risk due to large-scaled risk factors for the individual heavy metals (Wu et al. 2009;Liu et al. 2018). ...
Full-text available
Source-specific health risk apportionment for heavy metals is critical for pollution prevention and risk management in mining and smelting areas. An integrated method combining health risk assessments with the positive matrix factorization model was proposed to evaluate source-specific health risks for adults and children. A typical mining and smelting area was taken as an example in the present study to apportion the source-specific health risks to humans. A total of 37 road dust samples collected from the industrial (IA) and residential areas (RA) of Gejiu (China) were analyzed for heavy metals (Cd, Cr, Cu, Ni, Pb, and Zn). The results indicated that road dust in the study area was mainly contaminated with Cd, Cu, Pb, and Zn. Three potential sources, including atmospheric deposition, industrial waste, and natural sources, were identified and quantified, with contributions of 43.32%, 30.83%, and 25.85%, respectively. For non-carcinogenic risks, a similar trend of the source contribution was found for adults and children under the same land use; atmospheric deposition made the greatest contribution to the non-carcinogenic risk in both IA and RA. However, for carcinogenic risk, natural sources were the greatest contributor to human health risks in both IA and RA, followed by atmospheric deposition and industrial waste. The investigation in this study allowed the evaluation of health risks from potential contamination sources and the results provide valuable information on health risk mitigation strategies for environmental managers.
Full-text available
Traditionally, environmental authorities make regulatory policies for controlling volatile organic compounds (VOCs) pollution based upon the mitigation of dominant VOC sources. However, the emission from each VOC source has a unique combination of VOC species of different toxicities. Without quantitatively assessing the health risk associated with each source, the effectiveness of the mitigation policy could be undermined. To address this shortcoming, we developed a new health risk-oriented source apportionment method that can provide quantitative health risk assessment and source-specific mitigation strategies for hazardous VOCs. We estimated the integrated inhalation cancer risk (ICR) of hazardous VOCs was 7.7×10-5 in western Canada, indicating a 100% likelihood of exceeding Health Canada’s acceptable risk level (1.0×10-5). Anthropogenic sources were responsible for 56.3-73.8% of cancer risks across eight Canadian cities except for regional background island, where natural sources contributed over 77% to the integrated ICR. Thus, substantial environmental and health co-benefits could be achieved via reducing ambient levels of benzene and 1,3-butadiene by 39.3-75.7% and 14-69.3%, respectively, and mitigating emissions from fuel combustion (by 31.3-54.1%), traffic source (3.0-36.8%), and other anthropogenic sources (5.3-20.1%) in western Canada. Our study has significant implications for prioritizing air pollution mitigation policies, especially for quantitative reductions of hazardous air pollutants.
Exposure to potentially toxic trace elements (PTTEs) in inhalable particulate matter (PM10) is associated with an increased risk of developing cardiorespiratory diseases. Therefore, in multi-source polluted urban contexts, a spatially-resolved evaluation of health risks associated with exposure to PTTEs in PM is essential to identify critical risk areas. In this study, a very-low volume device for high spatial resolution sampling and analysis of PM10 was employed in Terni (Central Italy) in a wide and dense network (23 sampling sites, about 1 km between each other) during a 15-month monitoring campaign. The soluble and insoluble fraction of 33 elements in PM10 was analysed through a chemical fractionation procedure that increased the selectivity of the elements as source tracers. Total carcinogenic risk (CR) and non-carcinogenic risk (NCR) for adults and children due to concentrations of PTTEs in PM10 were calculated and quantitative source-specific risk apportionment was carried out by applying Positive Matrix Factorization (PMF) to the spatially-resolved concentrations of the chemically fractionated elements. PMF analysis identified 5 factors: steel plant, biomass burning, brake dust, soil dust and road dust. Steel plant showed the greatest risk contribution. Total CR and NCR, and source-specific risk contributions at the 23 sites were interpolated using the ordinary kriging (OK) method and mapped to geo-reference the health risks of the identified sources in the whole study area. This also allowed risk estimation in areas not directly measured and the assessment of the risk contribution of individual sources at each point of the study area. This innovative experimental approach is an effective tool to localize the health risks of spatially disaggregated sources of PTTEs and it may allow for better planning of control strategies and mitigation measures to reduce airborne pollutant concentrations in urban settings polluted by multiple sources.
Oil and natural gas (O&G) extraction operations emit hazardous volatile organic compounds (VOCs) in quantities that have adverse effects on human health. Our current understanding of the exposure risks associated with upstream O&G exploitations remains limited, and very few quantitative on-site remediation strategies have been proposed. To this end, we assessed the health risks associated with the emission of hazardous VOCs and presented a set of remediation goals for the city of Calgary, which is a major center of the Canadian oil industry. Results from probabilistic risk assessment (PRA) suggested that although VOCs had a negligible impact on chronic non-cancer-associated risk, inhalation-associated cancer risk remained a significant concern. Carbon tetrachloride, benzene, and 1,3-butadiene were the dominant VOCs, representing 88% of the integrated inhalation cancer risk (= 7.8 x 10-5); background, solid fuel combustion, and O&G extraction were among the primary sources that posed the greatest threat to human health. Results of a Monte Carlo simulation revealed that the probability of developing cancer due to inhalation of hazardous VOCs was ∼13.1% on clean air days and 45.9% on days with significant levels of air pollution. Preliminary remediation goals (PRGs) included reductions of 24.2–65.1% and 11.4–50.9% targeting priority VOCs and their sources, respectively. Taken together, our findings suggest that stringent control of the sources of VOCs, particularly fossil fuel combustion, is an urgent priority. PRA coupled with PRGs provides informative risk assessments and suggests quantitative remediation strategies that can be applied toward improved management of hazardous pollutants.
Full-text available
Chemical compositions of atmospheric fine particles like PM2.5 prove harmful to human health, particularly to cardiopulmonary functions. Multifaceted health effects of PM2.5 have raised broader, stronger concerns in recent years, calling for comprehensive environmental health-risk assessments to offer new insights into air-pollution control. However, there have been few studies adopting local air-quality-monitoring datasets or local coefficients related to PM2.5 health-risk assessment. This study aims to assess health effects caused by PM2.5 concentrations and metal toxicity using epidemiological and toxicological methods based on long-term (2007–2017) hourly monitoring datasets of PM2.5 concentrations in four cities of Taiwan. The results indicated that (1) PM2.5 concentrations and hazardous substances varied substantially from region to region, (2) PM2.5 concentrations significantly decreased after 2013, which benefited mainly from two actions against air pollution, i.e., implementing air-pollution-control strategies and raising air-quality standards for certain emission sources, and (3) under the condition of low PM2.5 concentrations, high health risks occurred in eastern Taiwan on account of toxic substances adsorbed on PM2.5 surface. It appears that under the condition of low PM2.5 concentrations, the results of epidemiological and toxicological health-risk assessments may not agree with each other. This raises a warning that air-pollution control needs to consider toxic substances adsorbed in PM2.5 and region-oriented control strategies are desirable. We hope that our findings and the proposed transferable methodology can call on domestic and foreign authorities to review current air-pollution-control policies with an outlook on the toxicity of PM2.5.
In order to identify the contribution to health risk derived from various emission sources, this study investigated monsoon variations in PM2.5 mass and concentrations of the associated trace elements in a region with complex pollution sources in central Taiwan. This study applied the Chemical Mass Balance model to analyze the source contribution of PM2.5. The source apportionment to obtain the risk contribution of different sources were conducted for different monsoon periods according to the monsoon patterns. In this way, the contributions of individual sources and chemicals to health risk under different monsoon types can be understood to support development of effective control strategies. Among the top contributors of PM2.5 during the north-east monsoon were Secondary Aerosol 28.93% >Coal Boiler 19.82% >Crustal Dust 15.99%; in south-west monsoon were Coal Boiler 37.29% >Traffic Emission 21.19% >Secondary Aerosol 17.84%. The total risk of cancer was above the acceptable risk (3.07 × 10⁻⁶), while the non-carcinogenic risk was within the acceptable range (0.262). The variation in the concentration and composition of PM2.5 was related to the change of monsoon type. During the north-east monsoon, the air mass had a long transmission distance and the PM2.5 concentration was relatively high. During the south-west monsoon, the air mass had a short transmission distance and the composition was mainly influenced by nearby emission sources, which resulted in higher risk due to chemical characteristics. To provide sound air quality management, attention should be paid to the composition of PM2.5 in addition to its concentration.
Comprehensive air monitoring of hazardous air pollutants (HAPs) was conducted at four residential sites (three in the town and one in a suburb) in Pohang, where Korea's largest iron-steel industrial complex is located. The objectives of this study were to evaluate the occurrence and spatiotemporal distributions of HAPs and to identify important HAPs based on a health risk assessment. The 130 HAPs that were measured simultaneously included volatile organic compounds (VOCs), carbonyls, polycyclic aromatic hydrocarbons (PAHs), phthalates, and heavy metals (HMs). The impact of industrial emissions on the ambient levels of HAPs in Pohang appeared to be significant, because the concentrations of multiple HAPs at residential sites near the industrial complex were considerably higher than those at the suburban site. The concentrations of VOCs (except formaldehyde) and HMs did not exhibit a specific seasonal pattern, but PAH levels were generally 4–5 times higher in winter than in summer. The cumulative cancer risks posed by 28 HAPs averaged over the three residential sites and the suburban site were 2.3 × 10⁻⁴ and 1.6 × 10⁻⁴, respectively, both of which exceeded the tolerable risk criterion of 1 × 10⁻⁴. Benzo[a]pyrene posed the highest risk, followed by As, formaldehyde, benzene, and dibenz[a,h]pyrene. However, no single HAP exceeded the non-cancer risk criterion of 1. These results support the need for stricter controls on emissions of PAHs, VOCs, and HMs in Pohang, particularly for sources in the iron-steel industrial complex.
Full-text available
We introduce an extended receptor model, implemented with the multilinear engine ME2, which combines simultaneous but separate filter-based species information with size-resolved particle volume information. Our chemical data set consisted of 24-hour filter measurements reported by the EPA Speciation Trends Network at Beacon Hill in Seattle, Washington, from February 2000 to June 2003. We measured the particle size distribution at this site from December 2000 to April 2002 using a differential mobility particle sizer (DMPS) and an aerodynamic particle sizer (APS). The combined model extends the traditional chemical mass balance approach by including a simultaneous set of conservation equations for both particle mass and volume, linked by a unique value of apparent particle density for each source. The model distinguished three mobile source features, two consistent with previous identifications of ``gasoline'' and ``diesel'' sources, and an additional minor feature enriched in EC, Fe and Mn and ultrafine particle mass that would have been difficult to interpret in the absence of particle size information. This study has also demonstrated the feasibility of defining missing mass as an additional variable, and thereby providing additional useful model constraints and eliminating the posthoc regression step that is traditionally used to rescale the results.
Full-text available
This study reports the first field measurements of airborne hexavalent chromium (Cr(vi)) in southwestern Ontario. Hexavalent chromium was identified as an inhalation carcinogen and an air toxic of concern during the 1991-93 Windsor Air Quality Study. The results of that study indicated that approximately 20% of the routinely monitored ambient airborne chromium (Cr) was in the hexavalent form. In addition, the range of carcinogenic health risks attributable to airborne Cr(vi) was determined to be between 1.4 x 10(-5) and 3.0 x 10(-4) for people living in the Windsor area. During the summer of 1993, analyses of concurrent indoor and outdoor 24-hour air quality samples taken at 33 residences in Hamilton resulted in geometric mean Cr(vi) concentrations of 0.20 ng/m3 and 0.55 ng/m3, respectively, and little or no relationship between the indoor and outdoor sample sets. During the summer of 1994, an airborne Cr(vi) size-fraction study was conducted in Hamilton, the results of which suggested that the majority of the Cr(vi) was in the inhalable fraction.
Full-text available
We used monitoring and modeling to assess the concentrations of air toxics in the state of Minnesota. Model-predicted concentrations for 148 hazardous air pollutants were from the U.S. Environmental Protection Agency Cumulative Exposure Project (1990 data). Monitoring data consisted of samples of volatile organic compounds, carbonyls, and particulate matter [Less than and equal to] 10 microm in aerodynamic diameter collected at 25 sites throughout the state for varying periods of time (up to 8 years; 1991-1998). Ten pollutants exceeded health benchmark values at one or more sites by modeling, monitoring, or both (including acrolein, arsenic, benzene, 1,3-butadiene, carbon tetrachloride, chromium, chloroform, ethylene dibromide, formaldehyde, and nickel). Polycyclic organic matter also exceeded the benzo[a]pyrene health benchmark value assumed to represent this class of pollutants. The highest modeled and monitored concentrations of most pollutants were near the center of the Minneapolis-St. Paul metropolitan area; however, many smaller cities throughout the state also had elevated concentrations. Where direct comparisons were possible, monitored values often tended to exceed model estimates. Upper-bound excess lifetime inhalation cancer risks were estimated to range from 2.7 [times] 10(-5) to 140. 9 [times] 10(-5) (modeling) and 4.7 [times] 10(-5) to 11.0 [times] 10(-5) (using a smaller set of monitored carcinogens). Screening noncancer hazard indices summed over all end points ranged from 0.2 to 58.1 (modeling) and 0.6 to 2.0 (with a smaller set of monitored pollutants). For common sets of pollutants, the concentrations, cancer risks, and noncancer hazard indices were comparable between model-based estimates and monitored values. The inhalation cancer risk was apportioned to mobile sources (54%), area sources (22%), point sources (12%), and background (12%). This study provides evidence that air toxics are a public health concern in Minnesota.
A technique for fitting multilinear and quasi-multilinear mathematical expressions or models to two-, three-, and many-dimensional data arrays is described. Principal component analysis and three-way PARAFAC factor analysis are examples of bilinear and trilinear least squares fit. This work presents a technique for specifying the problem in a structured way so that one program (the Multilinear Engine) may be used for solving widely different multilinear problems. The multilinear equations to be solved are specified as a large table of integer code values. The end user creates this table by using a small preprocessing program. For each different case, an individual structure table is needed. The solution is computed by using the conjugate gradient algorithm. Non-negativity constraints are implemented by using the well-known technique of preconditioning in opposite way for slowing down changes of variables that are about to become negative. The iteration converges to a minimum that may be local or global. Local uniqueness of the solution may be determined by inspecting the singular values of the Jacobian matrix. A global solution may be searched for by starting the iteration from different pseudorandom starting points. Application examples are discussed—for example, n-way PARAFAC, PARAFAC2, Linked mode PARAFAC, blind deconvolution, and nonstandard variants of these.
Ambient particulate matter less than or equal to2.5 mum in aerodynamic diameter (PM2.5) samples were collected at a centrally located urban monitoring site in Seattle, WA on Wednesdays and Saturdays using Interagency Monitoring of Protected Visual Environments ( IMPROVE) samplers. Particulate carbon was analyzed using the thermal optical reflectance method that divides carbon into four organic carbon (OC), pyrolyzed organic carbon (OP), and three elemental carbon (EC) fractions. A total of 384 samples that were analyzed for 36 species were collected between March 1996 and February 2000. These data were analyzed with the standard factor analysis model using the Multilinear Engine ( ME). Eleven sources were identified: sulfate-rich secondary aerosol (26%), diesel emissions (22%), wood smoke (16%), gasoline vehicle (10%), aged sea salt (8%), airborne soil (7%), nitrate-rich secondary aerosol (5%), sea salt (4%), oil combustion (3%), paper mill (2%), and ferrous metal processing (1%). The use of ME provided enhanced source separations, including the nitrate-rich aerosol source and two industrial sources that were not deduced in a previous PMF2 solution. Conditional probability functions using surface wind data and resolved source contributions aid in the identifications of local sources. Potential source contribution function analysis tentatively shows southern Washington State, along the Canadian border, and southwestern British Colombia, Canada as the possible source areas and pathways that give rise to the high contribution of the sulfate-rich secondary aerosol.
Receptor and dispersion modeling techniques are extended to develop a health risk apportionment concept in which inhalation exposure to emission sources of ambient element pollutants are estimated. A preliminary demonstration of the concept is performed using ambient and emission inventory data from an industrial air shed located in a residential area. It is shown that risks from identified emission sources can be quantified and that a total, additive risk can be estimated for the sources in the air shed. Potential risk reduction measures can then be considered on the main risk sources without arbitrarily reducing risk for all existing sources in the air shed. Dispersion modeling is utilized from emission inventory data so that risk estimates for the primary sources can be modeled and compared with both ambient and receptor model risk estimates.
Sumario: Introduction -- Methods -- Modeled estimates of outdoor air toxics concentrations -- Application of toxicity information to assess health risks -- assessment of cancer risks -- Assessing noncancer risks -- Results -- Geographic Distribution of air toxics concentrations and cumulative health risks by county -- Cumulative cancer and noncancer hazard estimates by Census Tract -- Pollutant-specific health risk estimates -- Discussion -- Limitations of toxicity information and uncertainties in science-policy assumptions -- Performance of dispersion model -- Emission source allocations -- Conclusions
A public health concern regarding hazardous air pollutants (HAPs) is their potential to cause cancer. It has been difficult to assess potential cancer risks from HAPs, due primarily to lack of ambient concentration data for the general population. The Environmental Protection Agency's Cumulative Exposure Project modeled 1990 outdoor concentrations of HAPs across the United States, which were combined with inhalation unit risk estimates to estimate the potential increase in excess cancer risk for individual carcinogenic HAPs. These were summed to provide an estimate of cancer risk from multiple HAPs. The analysis estimates a median excess cancer risk of 18 lifetime cancer cases per 100,000 people for all HAP concentrations. About 75% of estimated cancer risk was attributable to exposure to polycyclic organic matter, 1,3-butadiene, formaldehyde, benzene, and chromium. Consideration of some specific uncertainties, including underestimation of ambient concentrations, combining upper 95% confidence bound potency estimates, and changes to potency estimates, found that cancer risk may be underestimated by 15% or overestimated by 40-50%. Other unanalyzed uncertainties could make these under- or overestimates larger. This analysis used 1990 estimates of concentrations and can be used to track progress toward reducing cancer risk to the general population.
Of the 188 hazardous air pollutants (HAPs) listed in the Clean Air Act, only a handful have information on human health effects, derived primarily from animal and occupational studies. Lack of consistent monitoring data on ambient air toxics makes it difficult to assess the extent of low-level, chronic, ambient exposures to HAPs that could affect human health, and limits attempts to prioritize and evaluate policy initiatives for emissions reduction. Modeled outdoor HAP concentration estimates from the U.S. Environmental Protection Agency's Cumulative Exposure Project were used to characterize the extent of the air toxics problem in California for the base year of 1990. These air toxics concentration estimates were used with chronic toxicity data to estimate cancer and noncancer hazards for individual HAPs and the risks posed by multiple pollutants. Although hazardous air pollutants are ubiquitous in the environment, potential cancer and noncancer health hazards posed by ambient exposures are geographically concentrated in three urbanized areas and in a few rural counties. This analysis estimated a median excess individual cancer risk of 2.7E-4 for all air toxics concentrations and 8600 excess lifetime cancer cases, 70% of which were attributable to four pollutants: polycyclic organic matter, 1,3 butadiene, formaldehyde, and benzene. For noncancer effects, the analysis estimated a total hazard index representing the combined effect of all HAPs considered. Each pollutant contributes to the index a ratio of estimated concentration to reference concentration. The median value of the index across census tracts was 17, due primarily to acrolein and chromium concentration estimates. On average, HAP concentrations and cancer and noncancer health risks originate mostly from area and mobile source emissions, although there are several locations in the state where point sources account for a large portion of estimated concentrations and health risks. Risk estimates from this study can provide guidance for prioritizing research, monitoring, and regulatory intervention activities to reduce potential hazards to the general population. Improved ambient monitoring efforts can help clarify uncertainties inherent in this analysis.