This article was downloaded by: [Duzce University]

On: 16 January 2015, At: 08:27

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Journal of the Air & Waste Management Association

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uawm20

Identification of the sources of PM10 in a subway tunnel

using positive matrix factorization
a

ac

ac

Duckshin Park , Taejeong Lee , Doyeon Hwang , Wonseok Jung , Yongil Lee , KiChul Cho ,
b

Dongsool Kim & Kiyoung Lee

a

Eco-Transport Research Division, Korea Railroad Research Institute, Uiwang, Korea

Department of Environmental Science and Engineering, Kyung Hee University, Yongin,

Korea
c

Mechanical Engineering Department, Sungkyunkwan University, Suwon, Korea

Department of Bio and Environmental Science, Dongnam Health College, Suwon, Korea

Click for updates

Department of Environmental Health and Institute of Health and Environment, Graduate

School of Public Health, Seoul National University, Seoul, Korea
Accepted author version posted online: 12 Aug 2014.Published online: 14 Nov 2014.

To cite this article: Duckshin Park, Taejeong Lee, Doyeon Hwang, Wonseok Jung, Yongil Lee, KiChul Cho, Dongsool Kim &
Kiyoung Lee (2014) Identification of the sources of PM10 in a subway tunnel using positive matrix factorization, Journal of the
Air & Waste Management Association, 64:12, 1361-1368, DOI: 10.1080/10962247.2014.950766
To link to this article: http://dx.doi.org/10.1080/10962247.2014.950766

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

TECHNICAL PAPER

Identification of the sources of PM10 in a subway tunnel using positive

matrix factorization
Duckshin Park,1 Taejeong Lee,2 Doyeon Hwang,1 Wonseok Jung,1,3 Yongil Lee,1,3
KiChul
Cho,4 Dongsool Kim,2 and Kiyoung Lee5,
1

Eco-Transport Research Division, Korea Railroad Research Institute, Uiwang, Korea

Department of Environmental Science and Engineering, Kyung Hee University, Yongin, Korea
3
Mechanical Engineering Department, Sungkyunkwan University, Suwon, Korea
4
Department of Bio and Environmental Science, Dongnam Health College, Suwon, Korea
5
Department of Environmental Health and Institute of Health and Environment, Graduate School of Public Health, Seoul National University,
Seoul, Korea
Please address correspondence to: Kiyoung Lee, Graduate School of Public Health, Seoul National University, 1 Gwanak-ro Gwanak-gu, Seoul
151-470, Korea; e-mail: cleanair@snu.ac.kr

Downloaded by [Duzce University] at 08:27 16 January 2015

The level of particulate matter of less than 10 mm diameter (PM10) at subway platforms can be significantly reduced by installing
a platform screen-door system. However, both workers and passengers might be exposed to higher PM10 levels while the cars are
within the tunnel because it is a more confined environment. This study determined the PM10 levels in a subway tunnel, and identified
the sources of PM10 using elemental analysis and receptor modeling. Forty-four PM10 samples were collected in the tunnel between
the Gireum and Mia stations on Line 4 in metropolitan Seoul and analyzed using inductively coupled plasmaatomic emission
spectrometry and ion chromatography. The major PM10 sources were identified using positive matrix factorization (PMF). The
average PM10 concentration in the tunnels was 200.8
22.0 mg/m3. Elemental analysis indicated that the PM10 consisted of 40.4%
inorganic species, 9.1% anions, 4.9% cations, and 45.6% other materials. Iron was the most abundant element, with an average
concentration of 72.5
10.4 mg/m3. The PM10 sources characterized by PMF included rail, wheel, and brake wear (59.6%), soil
combustion (17.0%), secondary aerosols (10.0%), electric cable wear (8.1%), and soil and road dust (5.4%). Internal sources
comprising rail, wheel, brake, and electric cable wear made the greatest contribution to the PM10 (67.7%) in tunnel air.
Implications: With installation of a platform screen door, PM10 levels in subway tunnels were higher than those on platforms.
Tunnel PM10 levels exceeded 150 mg/m3 of the Korean standard for subway platform. Elemental analysis of PM10 in a tunnel showed
that Fe was the most abundant element. Five PM10 sources in tunnel were identified by positive matrix factorization. Railroad-related
sources contributed 68% of PM10 in the subway tunnel.

Introduction
In urban areas, the subway system is the most efficient and
convenient mode of commuter transportation. In Seoul, the subway system is a major form of transportation and accounts for
~36.2% of the total transportation in the city (Seoul, 2010).
Seoul is served by nine subway lines with a total of 358 stations
and transports 6.7 million passengers daily (Jung et al., 2010).
Subway passengers and workers can develop air pollution problems because the subway system is typically a confined space
exposed to pollution from various sources. The health of subway
passengers and employees is of great concern, considering the
potentially high levels of air pollution combined with prolonged
exposure times (Kim et al., 2006).
Particulate matter with an aerodynamic diameter less than 10
mm (PM10) is one of the major pollutants in subway environments. Many studies have reported that the PM concentrations in

subway systems were significantly higher than those measured

in ambient air (Aarnio et al., 2005; Kim et al., 2008; Salma et al.,
2007, 2009; Ripanucci et al., 2009; Jung et al., 2010; Kam et al.,
2011; Querol et al., 2012). Studies show that the PM in tunnels
can move into the cabin of a train travelling on the subway
(Seaton et al., 2005; Gustavsson et al., 2008). Particulate matter
in a subway originates mainly from the movement of trains and
passengers and accumulates in the closed environment (Park
et al., 2013).
Recently, a platform screen door (PSD) system has been
installed in many metropolitan subway systems worldwide to
reduce the risk of passenger accidents and to improve safety in
the event of a train fire (Yang et al., 2011). In addition, the PSD
could improve the air quality on indoor platforms by blocking
the train-induced airflow in subway tunnels. The average PM10
levels at the subway platform were reduced significantly after
installing a PSD in Seoul, Korea (Lee et al., 2010; Kim et al.,

1361
Journal of the Air & Waste Management Association, 64(12):13611368, 2014. Copyright 2014 A&WMA. ISSN: 1096-2247 print
DOI: 10.1080/10962247.2014.950766 Submitted April 10, 2014; final version submitted July 4, 2014; accepted July 26, 2014.

1362

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

2012). However, the tunnel air quality after installation of the

PSD has not been investigated. Since the subway tunnel becomes
a more enclosed environment as a consequence, both subway
workers and passengers might be exposed to more airborne
particulate matter in the tunnel and inside the subway passenger
cabins. This study measured the PM10 levels in the tunnel of
Seoul Metropolitan Subway Line 4 for 12 days and identified the
sources. The chemical properties of the PM10 samples were
determined by elemental analysis and the local particulate
sources were identified using receptor modeling.

Downloaded by [Duzce University] at 08:27 16 January 2015

Materials and Methods

Seoul Metropolitan Subway Line 4 was selected for evaluation of tunnel air quality. Line 4 has a total length of 31.7 km
between Sadang and Danggogae, an average depth of ~14.0 m,
and 25 stations. Travel from one end of the line to the other takes
53 min. In total, 47 trains with 470 cabins operate on the line.
The travel time between stations ranges from 2.5 to 5.5 min. As
of 2012, the line carried ~306 million passengers annually and
an average of 840,000 passengers daily (Seoul Metro, 2012).
Forty-eight PM10 samples were collected in the tunnel
between the Gireum, Miasageori, and Mia stations on Line 4 in
metropolitan Seoul from October 12 to 24, 2012, by sampling for
24 hr on each of the 12 days at four points in the tunnel. However,
four samples were damaged, leaving 44 for analysis. The tunnel
is a single wide tunnel with two tracks. The total length from
Gireum to Mia was 1,440 m at an average depth of ~25 m. PSDs
had been installed in all of the stations. The PM10 samples were
collected on 47-mm quartz filters (QMA filter 47 mm,
Whatman) using a mini-volume air sampler (model PAS 201,
Air Metrics). The flow rate was adjusted to 5 L min1 based on
the ambient temperature and pressure in each 24-hr period. All of
the filters were weighed before and after sampling with an
analytical balance (A&D, model HM-202) with 0.01 mg sensitivity after stabilizing the filter under a constant temperature and
humidity for 3 days in an electronic desiccator (model Oyin
09678BN, Sanplatec).
The filter paper was cut in half using ceramic scissors. Each
half of the filter paper was preprocessed for analysis of the
inorganic components and iron content. Preprocessing for the
inorganic component (IC) analysis complied with the preprocessing standard of the US Clean Water Act. We used the nitric acid/
hydrochloric acid preprocessing method with a microwave
(model MARS5, CEM). The sample was put in a perfluoroalkoxy polymer resin (PFA) liner and mixed with 7 mL of 61%
nitric acid and 3 mL of 35% hydrochloric acid. This mixture was
maintained at 150 psi for 10 min to extract the inorganic components. The extracted solution was filtered using filter paper
(number 5B, 110 mm, Advance MFS), diluted with 50 mL of
ultrapure water, and then refrigerated at 4 C until analysis. The
solution was analyzed for 13 inorganic components, namely Mg,
Al, Si, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Ba, and Pb, using inductively coupled plasmaatomic emission spectrometry
(ICP-AES; DRE ICP, Leeman Labs).
Preprocessing for analysis of the water-soluble ion content
involved soaking the filter paper in 30 mL of ultrapure water and
extracting it with an ultrasonic extractor (Model 5210R-DTH,

Branson Ultrasonics). To prevent the column from clogging

during the IC analysis, the extracted solution was filtered using
sterile 47-mm-diameter membrane filter paper (Whatman) with
0.45-mm pores; the filtered solution was refrigerated in a 60-mL
narrow mouth bottle (Nalgene) at 4 C. We used ion chromatography (model DX-400, Dionex) to evaluate the extracted watersoluble ion content. Three anions (NO3, SO42, and Cl) were
analyzed using an AS12A column.
Prior to use of ICP-AES and IC, several standard solutions
with similar level of sample concentration were injected to
examine quality assurance. An accuracy check was performed
by calculating a relative error (RE) resulting from standard
solutions, and a precision check was performed by obtaining
relative standard deviation and a coefficient of variation (CV)
based on three repetitive measurements.
For positive matrix factorization (PMF), the calculated factor
loading is always positive and does not depend on the correlation
matrix data, but instead on the least-square minimization of the
individual data value. Therefore, PMF depends on the error
estimation of the individual data value. DOS version PMF2
software was applied to identify the sources. The PMF model
equation can be expressed as
Xp
Xij
G F Eij
(1)
k1 ik kj
Here, X (m  n) is a matrix of m analytical chemical species and
n samples. Matrix G (n  p) is the factor score and represents the
contribution of the pollution sources to each sample. Matrix F
(pm) represents the source profiles, while E is the residual
matrix (Paatero and Tapper, 1994).
In a PMF model, the values of the G and F series are always
positive and can be obtained by minimizing the Q value. The Q
value is useful for determining the optimal number of factors
(Paatero, 1997) and can be calculated with
Q

Xn Xm Xij  Gik Fkj 2 Xn Xm Eij 2

(2)
i1
j1
i1
j1 s
s2ij
ij

where sij is the uncertainty (Paatero and Tapper, 1994).

An important variable in PMF analysis is the nondimensional
explained variation (EV). The EV of the jth chemical species of
the kth element can be calculated with
Pn
jGik Fki j=Sij
 
EVkj Pn Pn i1
(3)
 
i1 h1 jGih Fhi j Eij =Sij
Here, k 1, . . ., p, and Sij represents the standard deviation of
Xij, which is the ith chemical species of the jth sample. EV ranges
between 0 and 1. As the value of EVkj approaches 1, the jth
chemical element of matrix X more accurately describes the kth
pollution source. As EVkj approaches 0, identification of the
pollution source is not accurate.
To evaluate the PM10 pollution sources in subway passenger
cabins quantitatively using PMF, the raw data must be reorganized and input into the model. Here, the data for PMF modeling
were organized into a 30  16 matrix composed of 13 inorganic
components (Mg, Al, Si, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Ba, and

1363

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

Downloaded by [Duzce University] at 08:27 16 January 2015

Pb) and three anions (Cl, NO3, and SO42). Q-mode was used
to organize the samples, with the samples given in the rows and
the concentration of each sample in the columns. When there are
incomplete data sets containing data below detection limit or
missing data, there are several methods to substitute by approximated values. For this study, values below the detection limit
were replaced by of the detection limit, while unmeasured
values were replaced by the geographical average of the particular chemical species described in the previous reports (Polissar
et al., 1998; Lee et al., 2002). This individual data point weighting permits the influence of the values to be related to the level of
confidence the analyst has in the data.
To perform PMF modeling, the uncertainty of the measurements, in addition to the values discussed already, is required.
The uncertainty Sij (error) can be calculated using eq 4 (Polissar
et al., 1998):
Sij MDL=3 k  Xij

(4)

The fraction error, k, in eq 4 is used to calculate the uncertainty.

Various species, instruments, and laboratories show different
analytical uncertainty, k. The specific values (%) of k for each
element were Al 1, Ba 5, Cr 5, Cu 5, Fe 5, Mn 1, Ni 1, Pb 5, Si 5,
Ti 3, Zn 5, Cl 3, NO3 7, SO42 3, Na 3, K 1, Mg2 3, and
Ca2 1. MDL represents the detection limit and Xij represents the
concentration of the jth chemical species of the ith sample. Since
the analytical measured uncertainty generally increases with the
concentration, k can be deduced based on the linear relationship

between these two values of measured concentration and analytical uncertainty (Kim et al., 2005).

Results and Discussion

Table 1 lists the PM10 and concentrations of 18 elements in
the tunnel of the Seoul Metropolitan Subway Line 4. The average
PM10 concentration in the subway tunnel was 200.75
21.99
(range 153.4252.1) mg/m3. The PM10 concentrations in the
tunnel were 2.6- and 2.9-fold higher than the levels on the subway platform (77.4 mg/m3) and in the cabins (68.4 mg/m3),
respectively (Park et al., 2012; Park et al., 2013). The levels in
the subway tunnel always exceeded the 150 mg/m3 Korean
standard for subway platforms.
Of the inorganic components, iron was present at the highest
concentration, with an average of 72.51
10.37 (range 49.14
97.66) mg/m3. The average concentrations of the other inorganic
components in mg/m3 were Si 3.03
1.18, Cu 2.35
0.38, Ba
1.19
0.25, Mn 0.60
0.12, Zn 0.51
0.34, Al 0.36
0.11,
Pb 0.23
0.13, Cr 0.22
0.30, Ni 0.15
0.15, and Ti 0.04

0.01. The average concentrations of the ions in mg/m3were
NO3 10.17
7.68, SO42 6.54
2.07, Ca2 6.15
3.03,
Cl 1.45
1.47, K 1.40
1.33, Na 1.32
1.67, and Mg2
0.95
0.54. The proportion (%) of each species in inorganic
component was Fe 89.3, Si 3.7, Cu 2.9, Ba 1.5, Mn 0.75, Zn
0.63, Al 0.44, Pb 0.29, Cr 0.27, Ni 0.19, and Ti 0.05. The
proportion (%) of each species in ion components was NO3
36.3, SO42 23.4, Ca2 30.0, Cl 5.2, K 5.0, Na 4.73, and

Table 1. Summary of the PM10 concentration and elemental composition in the subway tunnel

Concentration (mg/m3)
Percentile
Chemical species
PM10
Al
Ba
Cr
Cu
Fe
Mn
Ni
Pb
Si
Ti
Zn
Cl
NO3
SO42
Na
K
Mg2
Ca2

Average

Minimum

25

50

75

Maximum

200.75
21.99
0.36
0.11
1.19
0.25
0.22
0.30
2.35
0.38
72.51
10.37
0.60
0.12
0.15
0.15
0.23
0.13
3.03
1.18
0.04
0.01
0.51
0.34
1.45
1.47
10.17
7.68
6.54
2.07
1.32
1.67
1.40
1.33
0.95
0.54
6.15
3.03

153.40
0.18
0.79
0.00
1.56
49.14
0.35
0.00
0.03
0.47
0.03
0.15
0.05
0.95
2.58
0.05
0.04
0.20
2.05

183.66
0.28
1.00
0.04
2.05
65.72
0.52
0.08
0.10
2.43
0.03
0.32
0.81
4.80
4.40
0.66
0.61
0.62
4.24

199.21
0.33
1.14
0.08
2.35
72.67
0.60
0.08
0.23
3.02
0.04
0.43
1.28
9.05
5.35
0.77
0.94
0.84
5.25

213.78
0.39
1.41
0.21
2.66
79.15
0.67
0.17
0.29
4.12
0.04
0.58
1.38
13.88
9.14
1.31
1.24
1.12
7.03

252.11
0.75
1.80
1.18
3.31
97.66
0.90
0.62
0.59
5.22
0.06
1.99
8.84
31.62
13.95
10.23
6.03
2.79
16.69

Downloaded by [Duzce University] at 08:27 16 January 2015

1364

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

Mg2 3.4. The analysis indicated that the PM10 consists of

40.4% inorganic species, ions 13.9%, and 45.6% other materials,
including organic and inorganic carbons and other unmeasured
components. By the result of other studies, the chemical composition of subway tunnel particles by the XRD method is the
oxidation form, SiO2, CaCO3, TiO2, FeOx, and so on (Jung
et al., 2010).
At 36.1%, the iron content comprised the greatest fraction of
the PM10 in the subway tunnel. A large Fe fraction was also
reported in PM10 near urban railways. The Fe content in PM10 at
above1ground platforms, underground platforms, and electric
cabins ranged from 6471% (Colombi et al., 2013; Park et al.,
2013). Fe-containing subway particles are generated mainly by
mechanical wear and friction processes at the rail-wheel brake
interfaces, and at the interface between the catenaries providing
electricity to the subway trains. The major source of Fe particles
is the movement of the train on the rails (Jung et al., 2010).
Table 2 shows the correlation analysis of 18 elements. Fe was
significantly correlated with Mn (r 0.86), Cu (r 0.62), Al
(r 0.45), and Ba (r 0.44). Ba was also correlated with Cu
(r 0.78) and Ti (r 0.64). Fe, Ba, and Mn are present in the
materials used to manufacture rails and wheels (Chan and
Stachowiak, 2004), and iron and manganese dust accounted for
the measurable, albeit small, increases in PM concentrations
observed near railway lines (Gehrig et al., 2007; Bukowiecki
et al., 2007).
The Fe/Mn ratio was used to evaluate the impact of a railroad
on atmospheric PM10 (Bukowiecki et al., 2007). In this study, the
mean Fe/Mn concentration ratio was 121, which was similar to
the ratio of 122 determined on a platform (Park et al., 2013). The
Fe/Mn ratio is comparable to the ratios of 107 obtained for New
York (Chillrud et al., 2004), 92 for Helsinki (Aarnio et al., 2005),
and 108 for Budapest (Salma et al., 2007). These ratios are
similar to the typical elemental compositions of the rails and
electric sliding collectors (Salma et al., 2007). This suggests that
the contributions of railroad sources in subways are similar in
most countries.
The PMF analysis identified five factors as sources of PM10
pollution in the subway tunnel. A method to find out the number
of optimal factor is using the Q value described in eq 2. This step
minimizes the difference between true measured value and theoretical value. After choosing the number of factors, modeling
was conducted by repeatedly increasing 0.1 of FPEAK between
1.0 and 1.0. The parameter FPEAK was used to control rotations. The rotational matrix explaining uncertainty of factor
rotation is used as a criterion by determining degree of freedom.
Generally, optimum degree of rotational freedom can be determined with preserving a certain limit of Q values (Han et al.,
2006; Kim et al., 2003). Optimum modeling conditions for
FPEAK and Q value used in our study are 0.2 and 427.9.
Figure 1 shows the correlation between the measured PM10
mass concentration and the PM10 concentration predicted by
standardization after PMF modeling. The coefficient of determination (R2) between the PM10 mass concentrations was 0.87,
indicating that the model accurately predicted the measurements.
Figure 2 shows the subway tunnel PM10 pollution source
profile and EV calculated using PMF modeling. For the first
pollution source, the Ni, Cr, SO42, and Ca2 EV values were

high, and the Cr and Ni EV were 0.7 and 0.5 respectively. The
first source could be identified as various types of oil combustion. Cr and Ni are markers of oil combustion in the fine fraction
(Lee et al., 2002; Morawska and Zhang, 2002). These markers
are emitted mainly from diesel oil or residual oil combustion. In
the profiles, Ca2 and SO42 were present at high mass fractions.
Calcium is generally emitted by the wear of brake linings, tires,
and clutch plates and by asphalt road surfaces, mainly in the
coarse mode (Monte and Rossi, 2000; Kupiainen et al., 2005).
These elements can be generated by the diesel vehicles used to
repair the subway rails at night and can enter underground subway tunnels through vents. The contribution of oil combustion to
the PM10 level in the subway tunnel was 17.0% (33.8 mg/m3).
This is similar to the contributions of outdoor-related sources,
including oil combustion and vehicles (4.0% and 6.0%, respectively) reported for PM10 in the subway tunnel (Park et al., 2013).
The second source, soil and road dust, included Ca2, K,
Mg2, Si, and SO42, which had EV > 0.5. Si was reported to be
present as a fraction >10% of a typical soil source for coarse
particles (Chow, 1995; Watson and Chow, 1994; Watson et al.,
2002). These elements in outdoor soil and road dust can enter a
subway tunnel through vents. The crustal particles originate
from unpaved roads, construction sites, and windblown soil
dust. As shown in Figure 2, soil and road dust contributed
5.4% (10.7 mg/m3) of the PM10. A previous PMF study (Jung
et al., 2010) reported a 7.210.6% contribution to the PM10 mass
concentration in subway tunnels.
The third source was rail, wheel, and brake wear, originating
mostly from indoor emissions due to the movement of trains. Fe,
Mn, Si, Ba, and Ca2 are markers of railroad operations and are
generated by the abrasion of the railroad tracks and brakes during
subway operation. Pollution related to rail, wheel, and brake
wear accounted for 59.6% (118.9 mg/m3) of the PM10 in the
subway tunnel, which was the highest of all sources identified in
this study. Rail-, wheel-, and brake-related sources have been
reported to contribute to the PM10 detected in subway tunnels
and on platforms (Gehrig et al., 2007; Jung et al., 2010; Lee
et al., 2010; Colombi et al., 2013; Park et al., 2013).
Zn, Cu, Cl, and Ca2 accounted for a large proportion of the
fourth source profile. Zn and Cu were present at relatively high
levels and were classified as being related to electric cables. Zn
and Cu are markers of subway operations and are generated by
abrasion of the power supply lines during subway operation
(Park et al., 2012). Pollution related to electric cable wear
accounted for 8.1% (16.2 mg/m3) of the PM10 in the subway
tunnel.
The fifth factor was characterized as secondary aerosols
based on presence of NO3 and SO42. This source might
include local and regional sources because formation of particulate SO42, NO3, and NH4 from gaseous SO2, NO2, and NH3
requires time. In this study, the NH4 content was not measured,
but it is likely that the NO3 and SO42 ions bonded with NH4,
resulting in formation of nitrate and sulfate particles. Once
released to the ambient atmosphere, these are converted predominantly to NH4NO3 and (NH4)xSO4 (x 0 to 2) via homogeneous or heterogeneous photochemical processes (Watson and
Chow, 1994; Khoder, 2002; Wilson et al., 2002). Nitrate and
sulfate aerosols were also present in significant amounts in

0.23*
0.09*
0.38*
0.45*
0.48
0.25
0.28
0.08*
0.67*
0.66*
0.22*
0.21*
0.25*
0.14*
0.16*
0.13*
0.23*

0.03
0.78*
0.44*
0.05*
0.11
0.03*
0.19*
0.64*
0.15
0.22
0.03*
0.14*
0.15
0.11*
0.01*
0.12*

Ba

Notes: *p < 0.01. p < 0.05.

Ba
Cr
Cu
Fe
Mn
Ni
Pb
Si
Ti
Zn
Cl
NO3
SO42
Na
K
Mg2
Ca2

Al

0.11
0.16*
0.27*
0.87*
0.01
0.57
0.07*
0.14
0.02
0.06
0.02
0.14*
0.26*
0.16
0.07

Cr

0.62*
0.37*
0.03
0.21*
0.28*
0.63*
0.06
0.03
0.08
0.13*
0.07
0.16*
0.05
0.03*

Cu

0.86*
0.17*
0.04*
0.21*
0.51*
0.04*
0.01*
0.12
0.17*
0.08
0.09
0.27
0.12

Fe

Mn

0.25
0.15
0.03*
0.33*
0.12*
0.17*
0.06*
0.11*
0.05*
0.08
0.26
0.13

Table 2. Correlation analysis of the elemental species and ions in PM10

0.03
0.55
0.21*
0.15*
0.04
0.11*
0.11*
0.09*
0.18*
0.10
0.03

Ni

0.07
0.15*
0.16*
0.01*
0.05*
0.43*
0.19*
0.11
0.22
0.11

Pb

0.16*
0.00*
0.03
0.07*
0.10*
0.11
0.01
0.22*
0.17*

Si

0.40*
0.05*
0.32*
0.42*
0.09
0.07*
0.10*
0.16*

Ti

0.35*
0.32*
0.36*
0.21*
0.08*
0.15*
0.25*

Zn

0.16*
0.20*
0.63
0.04
0.11*
0.12*

Cl

Downloaded by [Duzce University] at 08:27 16 January 2015

0.74*
0.14*
0.13
0.13
0.04

NO3

0.27*
0.10
0.04*
0.06

SO42

0.15
0.08*
0.26*

Na

0.57*
0.55*

0.76*

Mg2

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

1365

1366

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

PMF analysis of PM in the subway tunnel was used to determine the importance of internal sources. The sources can be
apportioned by comparing outdoor and tunnel PM. In further
studies, outdoor aerosols should be collected near the subway
stations to compare the composition of indoor and outdoor
aerosols. To understand subway tunnel emission sources in
more detail, it is necessary to investigate additional chemicals,
such as carbon compounds and other inorganic elements, and to
develop more marker species for specific sources. Although we
identified five sources of PM10 in the subway tunnel using PMF
modeling, this was based on measurements made during the fall,
not the other three seasons. As the primary sources of PM10
might differ seasonally, further research on the sources of PM10
in the tunnel according to season is needed.

Downloaded by [Duzce University] at 08:27 16 January 2015

Conclusion

Figure 1. Comparison of the measured PM10 concentrations and the values

predicted by PMF.

subway aerosol samples, and contributed 10.0% (19.9 mg/m3) of

the PM10. These secondary aerosols containing nitrates and
sulfates must originate from the outdoor environment, because
they are abundant in samples taken outdoors (Jung et al., 2010;
Lee et al., 2010; Park et al., 2013).

Figure 2. The source profile and EV calculated using PMF modeling.

The average PM10 concentration in the tunnels of Seoul

Metropolitan Subway was significantly higher than on the subway platform. The PM10 in the tunnel consisted of 40.4% inorganic species, 9.1% anions, 4.9% cations, and 45.6% other
materials. The iron fraction was the highest, contributing
36.1% of the PM10 concentration in the subway tunnel. Major
contributors in the subway tunnel were rail, wheel, and brake
wear (59.6%), oil combustion (17.0%), secondary aerosols
(10.0%), electric cable wear (8.1%), and soil and road dust
(5.4%). The internal sources, that is, rail, wheel, brake, and
electric cable wear, were the major contributors of PM10 in the

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

subway tunnel, at 67.7%. To reduce PM levels, it is important to

manage internal sources in subway tunnels. The source apportionment study based on PMF was useful to control specific
emission sources in subway tunnel to provide recommendations
for reducing PM10 levels in the subway tunnels so as to meet the
Korean standard for subway platforms.

Funding
The work described in this paper was supported by research
grants on the Basic Research Project of Korea by Ministry of
Science, ICT, and Future Planning, and Railway Technology
Research Project by Minister of Land, Infrastructure and
Transport (14RTRP-B081249-01).

Downloaded by [Duzce University] at 08:27 16 January 2015

References
Aarnio, P., T. Yli-Tuomi, A. Kousa, T. Mkela, A. Hirsikko, K. Hmeri, M.
Risnen, R. Hilamo, T. Koskentalo, and M. Jantunen. 2005. The concentration
and composition of and exposure to fine particles (PM2.5) in the Helsinki subway
system. Atmos. Environ. 39:50595066. doi:10.1016/j.atmosenv.2005.05.012
Bukowiecki, N., G. Robert, H. Matthias, L. Peter, N.Z. Christoph, B. Brigitte, W.
Ernest, and B. Urs. 2007. Iron, manganese and copper emitted by cargo and
passenger trains in Zurich (Switzerland): Size-segregated mass concentrations in ambient air. Atmos. Environ. 41:878889. doi:10.1016/j.
atmosenv.2006.07.045
Chan, D., and G.W. Stachowiak. 2004. Review of automotive brake friction
materials. Proc. Inst Mech. Eng. Part D J. Automobile Eng. 218:953966.
doi:10.1243/0954407041856773
Chillrud, S.N., E. David, M.R. James, N.S. Sonja, P. Dee, D.S. John, and L.K.
Patrick. 2004. Elevated airborne exposures of teenagers to manganese, chromium, and iron from steel dust and New York Citys subway system. Environ.
Sci. Technol. 38(3): 732737. doi:10.1021/es034734y
Chow, J.C. 1995. Measurement methods to determine compliance with ambient
air quality standards for suspended particles. J. Air Waste Manage. Assoc.
45:320382. doi:10.1080/10473289.1995.10467369
Colombi, C., S. Angius, V. Gianelle, and M. Lazzarini. 2013. Particulate matter
concentrations, physical characteristics and elemental composition in the
Milan underground transport system. Atmos. Environ. 70:166178.
doi:10.1016/j.atmosenv.2013.01.035
Gehrig, R., M. Hill, P. Lienemann, C.N. Zwicky, N. Bukowiecki, E. Weingartner,
U. Baltensperger, and B. Buchmann. 2007. Contribution of railway traffic to
local PM10 concentrations in Switzerland. Atmos. Environ. 41:923933.
doi:10.1016/j.atmosenv.2006.09.021
Gustavsson, P., C. Bigert, and M. Polln. 2008. Incidence of lung cancer among
subway drivers in Stockholm. Am. J. Ind. Med. 51(7): 545547. doi:10.1002/
(ISSN)1097-0274
Han, J.S., K.J. Moon, S.J. Lee, Y.J. Kim, S.Y. Ryu, S.S. Cliff, and S.M. Yi. 2006.
Size-resolved source apportionment of ambient particles by positive matrix
factorization at Gosan background site in East Asia. Atmos. Chem. Physics
6:211223. doi:10.5194/acp-6-211-2006
Jung, H.J., B.W. Kim, J.Y. Ryu, S. Maskey, J.C. Kim, J. Sohn, and C.U. Ro. 2010.
Source identification of particulate matter collected at underground subway
stations in Seoul, Korea using quantitative single-particle analysis. Atmos.
Environ. 44:22872293. doi:10.1016/j.atmosenv.2010.04.003
Kam, W., K. Cheung, N. Daher, and C. Sioutas. 2011. Particulate matter (PM)
concentrations in underground and ground-level rail systems of the Los Angeles
Metro. Atmos. Environ. 45:15061516. doi:10.1016/j.atmosenv.2010.12.049
Khoder, M.I. 2002. Atmospheric conversion of sulfur dioxide to particulate
sulfate and nitrogen nitric acid in an urban area. Chemosphere 49:675684.
doi:10.1016/S0045-6535(02)00391-0
Kim, E., T.V. Larson, P.K. Hopke, C. Slaughter, L.E. Sheppard, and C. Claibom.
2003. Source identification of PM2.5 in an arid northwest U.S. city by positive

1367

matrix factorization. Atmos. Research 66:291305. doi:10.1016/S0169-8095

(03)00025-5
Kim, K.Y., D.X. Ho, J.S. Jeon, and J.C. Kim. 2012. A noticeable shift in particulate
matter levels after platform screen door installation in a Korean subway station.
Atmos. Environ. 49:219223. doi:10.1016/j.atmosenv.2011.11.058
Kim, K.Y., Y.S. Kim, Y.M. Roh, C.M. Lee, and C.N. Kim. 2008. Spatial distribution of particulate matter (PM10 and PM2.5) in Seoul Metropolitan Subway
stations. J. Hazard. Mater. 154:440443. doi:10.1016/j.jhazmat.2007.10.042
Kupiainen, K.J., H. Tervahattu, M. Raisanen, T. Makela, M. Aurela, and R.
Hillamo. 2005. Size and composition of airborne particles from pavement
wear, tires, and traction sanding. Environ. Sci. Technol. 39:699706.
doi:10.1021/es035419e
Lee, J.H., Y. Yoshida, B.J. Turpin, P.K. Hopke, R.L. Poirot, R.J. Lioy, and J.C.
Oxley. 2002. Identification of sources contributing to Mid-Atlantic regional
aerosol. J. Air Waste Manage. Assoc. 52(10): 11861205. doi:10.1080/
10473289.2002.10470850
Lee, T.J., J.S. Jeon, S.D. Kim, and D.S. Kim. 2010. A comparative study on PM10
source contributions in a Seoul metropolitan subway station before/after
installing platform screen doors. J. Kor. Soc. Atmos. Environ. 26(5):
543553. doi:10.5572/KOSAE.2010.26.5.543
Monte, M.D., and P. Rossi. 2000. Aerosol Chemical Processes in the
Environment, ed. K.R. Spurny. New York, NY: Lewis.
Morawska, L., and J. Zhang. 2002. Combustion sources of particles. 1. Health
relevance and source signatures. Chemosphere 49:10451058. doi:10.1016/
S0045-6535(02)00241-2
Paatero, P., and U. Tapper. 1994. Positive matrix factorization: A nonnegative
factor model with optimal utilization of error estimates of data values.
Environmetrics 5:111126. doi:10.1002/(ISSN)1099-095X
Paatero, P. 1997. Least squares formulation of robust non-negative factor analysis.
Chemom. Intell. Lab. Syst. 37:23e35. doi:10.1016/S0169-7439(96)00044-5
Park, D., M. Oh, Y. Yoon, E. Park, and K. Lee. 2012. Source identification of
PM10 pollution in subway passenger cabins using positive matrix factorization. Atmos. Environ. 49:180185. doi:10.1016/j.atmosenv.2011.11.064
Park, S.B., S.N. Lee, T.J. Ko, H.K. Bae, S.J. Kim, S.D. Park, D. Sohn, and D.S.
Kim. 2013. Identification of PM10 chemical characteristics and sources and
estimation of their contributions in a Seoul metropolitan subway station. J. Kor.
Soc. Atmos. Environ. 29(1): 7485. doi:10.5572/KOSAE.2013.29.1.74
Polissar, A.V., P.K. Hopke, and P. Paatero. 1998. Atmospheric aerosol over
Alaska, 2. Elemental composition and sources. J. Geophysic. Res. Atmos.
103(D15): 1904519057. doi:10.1029/98JD01212
Polissar, A.V., P.K. Hopke, R.D., and Poirot. 2001. Atmospheric aerosol over
Vermont: Chemical composition and sources. Environ. Sci. Technol. 35:
46044621. doi:10.1021/es0105865
Querol, X., T. Moreno, A. Karanasiou, C. Reche, A. Alastuey, M. Viana, O. Font,
J. Gil, E. De Miguel, and M. Capdevila. 2012. Variability of levels and
composition of PM10 and PM2.5 in the Barcelona metro system. Atmos.
Chem. Phys. 12:50555076. doi:10.5194/acp-12-5055-2012
Ripanucci, G., M. Grana, L. Vicentini, A. Magrini, and A. Bergamaschi. 2006.
Dust in the underground Railway Tunnels of an Italian town. J. Occup.
Environ. Hyg. 3:1625. doi:10.1080/15459620500444004
Salma, I., M. Psfai, K. Kovcs, E. Kuzmann, Z. Homonnay, and J. Posta. 2009.
Properties and sources of individual particles and some chemical species in
the aerosol of a metropolitan underground railway station. Atmos. Environ.
43:34603466. doi:10.1016/j.atmosenv.2009.04.042
Salma, I., T. Weidinger, and W. Maenhaut. 2007. Time-resolved mass concentration, composition and sources of aerosol particles in a metropolitan underground railway station. Atmos. Environ. 41:83918405. doi:10.1016/j.
atmosenv.2007.06.017
Seaton, A., J. Cherrie, M. Dennekamp, K. Donaldson, J.F. Hurley, and C.L. Tran.
2005. The London underground: dust and hazards to health. Occup. Environ.
Med. 62:355362. doi:10.1136/oem.2004.014332
Seoul. 2012. http://stat.seoul.go.kr (accessed April 2, 2014).
Song, X.H., A.V. Polissar, and P.K. Hopke. 2001. Source of fine particle composition in the northeastern US. Atmos. Environ. 35(31): 52775286.
doi:10.1016/S1352-2310(01)00338-7

1368

Park et al. / Journal of the Air & Waste Management Association 64 (2014) 13611368

Downloaded by [Duzce University] at 08:27 16 January 2015

Watson, J.G., and J.C. Chow. 1994. Ammonium nitrate, nitric acid and ammonia
equilibrium in wintertime phoenix Arizona. J. Air Waste Manage. Assoc.
44:405412. doi:10.1080/1073161X.1994.10467262
Watson, J.G., T. Zhu, J.C. Chow, E.J. Engelbrecht, E.M. Fujita, and W.E.
Wilson. 2002. Receptor modeling application framework for particle
source apportionment. Chemosphere 49:10931136. doi:10.1016/S00456535(02)00243-6
Wilson, W.E., J.C. Chow, C. Claiborn, W. Fusheng, E.J. Engelbrecht, and J.G.
Watson. 2002. Monitoring of particulate matter outdoors. Chemosphere
49:10091043. doi:10.1016/S0045-6535(02)00270-9
Yang, J., Y. Yang, H.L. Wang, and L. Shi. 2011. Effect of the open ways of screen
doors on fire smoke in a subway platform. Procedia Engineering 11:
416423. doi:10.1016/j.proeng.2011.04.677

About the Authors

Duckshin Park is a principal researcher and Doyeon Hwang, Wonseok Jung, and
Yongil Lee are researchers at Korea Railroad Research Institute, Uiwang, Korea.
Taejeong Lee is a research assistant professor and Dongsool Kim is a professor
at Kyung Hee University, Yongin, Korea.
KiChul Cho is an associate professor at Dongnam Health College, Suwon,
Korea.
Kiyoung Lee is a professor at Seoul National University, Seoul, Korea.