See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/233925827

Source Apportionment of Atmospheric PM 10 in

Kanpur, India
Article in Environmental Engineering Science July 2008
DOI: 10.1089/ees.2006.0275

CITATIONS

READS

21

209

2 authors, including:
SP Shukla
Institute of Engineering & Technology - Luck
49 PUBLICATIONS 137 CITATIONS
SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

Available from: SP Shukla

Retrieved on: 20 August 2016

ENVIRONMENTAL ENGINEERING SCIENCE

Volume 25, Number 6, 2008
Mary Ann Liebert, Inc.
DOI: 10.1089/ees.2006.0275

Source Apportionment of Atmospheric

PM10 in Kanpur, India
Sheo Prasad Shukla1 and Mukesh Sharma2,*

Abstract

Atmospheric particle characteristics and concentrations are critical in source apportionment analysis. Particulate matter (PM) levels (in Indian conditions) suggest significant seasonal variability. There could be several
reasons for this variability, including the variable meteorology, wind-driven dust, and atmospheric chemistry,
which is responsible for formation of secondary particles. To clearly establish the causeeffect relationship of
particulate levels and sources, PM10 (n  100) samples were collected at Kanpur, India, during 2000 to 2001.
The collected samples were analyzed for metals and water-soluble ions, and results were utilized for factor
analysis and source apportionment estimations with specific focus on secondary particles of fine mode. In the
three seasons studied, two important PM10 sources were soil-road dust (1547%) and inorganic secondary particles like (NH4) 2SO4 and NH4NO3 (2126%). Winter and monsoon conditions were conductive for formation
of secondary particles due to moderate temperatures and high relative humidity. The role of ambient ammonia was vital in overall atmospheric chemistry of secondary particulate formation. It is thus clear that control
of PM will require control of primary precursor gases like SO2 and NOx, including NH3.
Key words: PM10; atmospheric particle characteristics; apportionment analysis

Introduction

NDIA, THE SECOND FASTEST GROWING ECONOMY in the world

(GDP growth  8%), has set a target of 215,804 MW power
generation capacity by March 2012 from the present level of
100,010 MW (Abbi, 2001). Thermal power generation will be
the main contributor to the overall power generation (approximately 68%). The proposed new power plants and
other industries will add to the existing emissions of SO2,
NOx, and PM (particulate matter) significantly. In view of
this phenomenal growth, the situation warrants a closer examination of the impacts of emission of SO2, NOx, PM, and
their interactions. To connect SO2/NOx emissions, its implications in understanding of secondary particle formation
over Indian subcontinent is quite complex, primarily for two
reasons. First are significant photochemical activities at high
temperature, and second, the atmosphere is quite dusty with
background particles (e.g., average PM10 particle size less
than or equal to 10 m) levels around 350 g m3 in North
India) (Kishore, 2005), largely due to arid conditions.

We could not find any study in India to examine the formation of secondary particles and identify their contribution
to PM10, and more importantly, identify the precursor
gaseous pollutants responsible for formation of secondary
particles. Because most of the secondary particles are formed
in fine mode, their associated health risk is high and their
control will require control of primary precursor gaseous
pollutants. The problems of apportionment of secondary
particles is further compounded, as no measurements of
PM2.5, left beside their composition, are done in India, which
must be true in most developing countries. If one could apportion the sources of secondary particles of fine mode from
the measurements of PM10 (which is routinely done) this
knowledge could be of immense utility for the control of
PM2.5.
The winter levels of PM were found to be less than those
in summer at each of the seven sampling locations in Delhi
during the year 1990 (Sharma et al., 1995). A 5-year-long
study (19811985) by the Central Pollution Control Board
(CPCB) at Agra concludes that PM levels were significantly

1Department

of Civil Engingeering, IET Lucknow, Lucknow, India.

of Civil Engineering, IIT Kanpur, Kanpur, India
*Corresponding author: Department of Civil Engineering, IIT Kanpur, Kanpur 208016, India. Phone: 91-512-2597759; Fax: 91-5122597395l; E-mail: mukesh@iitk.ac.in
2Department

849

850

SHUKLA AND SHARMA

less (270 g m3 5-year average) in winter months compared

to summer months (400 g m3 5-year average). The anthropogenic emission of PM in summer and winter months
are likely to be the same; therefore, the variation in PM levels in the winter months could be due to the absence or near
absence of soil dust in the atmosphere caused by calm atmospheric conditions and/or photochemistry responsible for
formation of secondary particles. Seasonal variation in temperature, humidity, wind speed, mixing height, soil moisture content, emission of ammonia, etc., affects the atmospheric chemistry and particle generation (from soil)/
formation. Hence, one can argue that to understand the issue of particle formation and source apportionment analysis, it is of great significance that seasonal variability of particulate concentration, their characterization, and their
sources be studied together.
In view of the discussions so far, the objective of present
study was to understand the seasonal variability in terms of
characteristics and concentrations of PM and understand
their sources in the atmosphere. For this purpose, a longterm study was designed and completed during June 2000

FIG. 1.

to December 2001 at Kanpur (longitude 88 22 E and latitude 26 26 N). Finally, efforts have been made to quantitatively apportion the sources contributing to the PM levels
and subsequently defining the role of NH3, SO2, and NOx in
formation of PM. The sampling site was a relatively clean
site in the city of Kanpur.
The technique of factor analysis-multiple regression (FAMR) (Kumar et al., 2001), a receptor model, was employed
to apportion the sources for PM. It may be noted that techniques like chemical mass balance will not work for apportioning secondary particles as these particles have no specific signatures at the source.
Study Design and Materials and Methods
Study area
The study area of Kanpur is situated at an approximate
distance of 435 km SE from Delhi on the right bank of river
Ganges. Kanpur is a representative site of the Ganga basin
(where new power plants are expected to come) in terms of
the weather conditions and atmospheric seasonal variability.

Map of Kanpur showing location of sampling site.

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

The Ganga basin experiences three dominant seasons each
year: winter (NovemberFebruary), summer (MarchJune),
and monsoon (JulyOctober). The Indian Institute of Technology Kanpur (IITK), an educational institute having a residential campus with no commercial or industrial activities,
was taken as the sampling site (Fig. 1). The campus lies at
about 15 km north of city downtown area with minimum
emissions. Within the campus, vehicular population mainly
comprises two-wheelers and cars. The heavy-duty vehicle
population is negligible. For most of the year, the campus
lies on the upwind side and receives no air pollution from
Kanpur city. Upwind (beyond 4 km) of this site, there are
villages having extensive agricultural activities.
Sampling and quality control
PM10 measurements were carried at a height of about 10
m (above the ground), at the roof of environmental engineering laboratory of IITK during the period June 2000 to
December 2001. The frequency of sampling was two 24-h (6
am6 am) samples in a week in accordance with National
Air Quality Monitoring Program in India. The collected
particulates were subjected to chemical analysis (details in
Table 1).
All initial and final weighing (using 440 Mettler Balance
with sensitivity 0.00001 g) of filter papers was done in a humidity-controlled room, and filters were conditioned in desiccators for 24 h before and after the sampling. High-volume
sampler (APM 450 Envirotech, New Delhi) was used for
sampling PM10 on Whatman GF/A 8"  10" size filter paper
at a flow rate of 1 m3 min1.
Estimation of heavy metals
After particulate collection, chemical speciation is the next
step in air quality assessment. For this purpose, metals (Al,
As, Ba, Ca, Fe, K, Mg, Mn, and Pb) were analyzed in PM10.
Half of the filter paper was used for metal analysis, and the
remaining half of the filter paper was used for analysis of
water-soluble ions.
The extraction and analysis of metals was carried out as
per the U.S. EPA Method IO-3.2 (U.S. EPA, 1999). As per the
method, PM10 samples collected on GF/A filters may be digested either by hot acid digestion or by a microwave-assisted digestion system. Half of the filter paper was cut into
small pieces by plastic scissors into Griffin beaker and 80 mL
of extraction solution (5.55% HNO3/16.75% HCl) was added.
These beakers were placed on a hot plate at a temperature
of 170C for 2530 min. Care was taken so that samples do

TABLE 1.
Sampling
seasons
Summer
Monsoon
Winter

851
not become dry. The walls of beaker were rinsed with 20 mL
Milli-Q water (conductivity 0.055 S cm1) and allowed to
stand for at least 30 min to allow the diffusion of acid from
the filter into the rinse. The sample was filtered through 0.22
micron filter paper and the filtrate was made up to 100 mL.
This sample was transferred to plastic bottles and refrigerated until further analysis.
Filter paper blanks and acid blanks were also taken
through the same procedure simultaneously. Al, As, Ba, Ca,
Fe, K, Mg, Mn, and Pb were determined in these acid-digested samples using flame atomic absorption spectrophotometer (AAS) (Model Varian SpectrAA 220FS and GBC
Avanta , Australia). As metal was determined using a
hydride generator (Model GBC HG3000) coupled with AAS.
Before analyzing the samples, the instrument was calibrated
for Al, As, Ba, Ca, Fe, K, Mg, Mn, and Pb. As per the U.S.
EPA method, a stock solution (of 1,000 ppm) was prepared
and diluted to the range of working standards for individual metals. The calibration graphs were prepared using these
working standards in the linear range of the optical density
(0.04 to 0.8). The instrument was calibrated at three different levels for each metal.
To examine the background metal contents of blank filter
paper (5% of total samples), exactly similar extraction and
analysis procedure was employed for blank filter papers. It
was found there was a large variation in the characteristics
of blank filter papers. To overcome the impact of background
elemental variability in the filter paper, the blank portion
(the border of filter paper unaffected by sampling due to top
plate holding the filter paper to the filter assembly) of each
PM10 laden filter paper was separately analyzed for getting
the concentrations of metals in the blank filter paper. For this
purpose, the blank portion of each PM10 laden filter paper
was cut and weighed separately. Corresponding blank corrections were incorporated in the data. The blank value, recovery, and detection limit of metals and ions in filter blank
is given in Table 2.
Analysis of ions
The remaining half portion of PM10 filter papers were analyzed for water-soluble ions following the prescribed extraction procedure (Lodge, 1989). Half of the filter paper used
in the high-volume sampler was cut into small pieces into a
conical flask, and 80 mL of Milli-Q water was added. The
sample was agitated for 30 min and then digested at a temperature of 65 C for 3 h. Digested sample was again agitated for 30 min. Thereafter, the walls of beaker were rinsed

SAMPLING DETAILS

Sampling months
June, 2000
MarchJune
JulyOctober, 2000
JulyOctober, 2001
November, 2000February, 2001
NovemberDecember, 2001

OF

PM10
Parameter
measured

PM10 concentrations,
Al, As, Ba, Ca
Fe, K, Mg,
Mn, Pb, NH4,
F, Cl, NO3, and SO42

Number of
samples
26
41
33

852

SHUKLA AND SHARMA

TABLE 2. METAL AND IONS IN FILTER BLANK,
RECOVERIES, AND DETECTION LIMITS.

Species

Blank value
(mg g1)

Recovery (%)

Detection limit
(mg L1)

Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
SO42
NO3

5.4
1.4
23.6
3.9
8.1
5.1
13.3
BDLa
BDLa
BDLa
BDLa
BDLa
BDLa
BDLa

92
85
88
91
96
86
94
97
97
85
97
97
97
90

0.020
0.035
0.045
0.002
0.006
0.002
0.001
0.003
0.020
0.10
0.10
0.10
0.25
0.20

aBDL,

below detection limit.

with Milli-Q water and the sample was filtered through a

0.22-micron filter paper to remove insoluble matter. The filtrate was made up to 100 mL and transferred to plastic bottles. This extracted solution was refrigerated until further
analysis. The water-soluble cation NH4 and anions F, Cl,
NO3 and SO42 were analyzed by an ion chromatograph
(IC) (model: Metrohm 761 compact).
Factor analysis
FA-MR, a receptor model technique, can be employed to
apportion the sources for PM10, and it can indicate formation of secondary particles. One such technique of factor
analysis is the principal component analysis (PCA). The objective of applying PCA is to derive a small number of components, which explain a maximum of the variance in the
data. Initially, the PCA results in as many PCs as there are
original variables. Usually, however, only a limited number
of these uncorrelated PCs are required to explain virtually
all of the variance in a data set of 15 or more original (intercorrelated) variables. For this reduction in the dimensionality to be useful, the newly obtained components (or
factors) must have simple substantive interpretations
(Thurston and Spengler, 1985). To be able to interpret the
component in terms of physical meaning, loadings of variables on the component is estimated. Loading represents the
degree and direction of relationship of the variables with a
factor. The loadings measure, which variables are associated
with, which factor, and to what degree. They can be interpreted like correlation coefficients. The square of the loading multiplied by 100 equals the percent variation that a variable has common with a factor. By comparing the factor
loadings for all factors and variables, those particular variables involved in an independent factor can be defined.
An examination of the PC loadings on the original variables allowed the identification of the PCs as pollution
sources affecting the data. Subsequently, the mass contribution (in g m3) of each source component to the total PM10
pollution levels was estimated.

Varimax rotated PCA was employed to apportion sources

of PM10 using a statistical package, SYSTAT (2004). The PCA
has been conducted using a conventional R analysis of elemental correlations about their means. Initially PC loadings
and PC scores were determined on the basis of measured
variables of data set for each season. However, the absolute
zero PC scores (APCS) have subsequently been estimated
for each PC by subtracting the PC score of an extra day
(wherein all the elemental concentrations are assumed to be
zero) from PC scores of each day (Thurston and Spengler,
1985). Regressing observed PM10 data on these APCS gave
estimates of the regression coefficients, which convert the
APCS into pollutant source mass contributions (in g m3)
for each sampling day. Mathematically for the jth observation,
n

[PM10]j  C0   Ci [APCSi]j

(1)

i1

where, n is the number of PCs, j is the 1, 2, . . . , m, m is the

number of observations, and C0, Ci are the constants. Effectively, the product of corresponding APCSi for observation
j and estimated constant (Ci) will give the contribution of the
ith source in PM10.
Henry et al. (1984) suggested that the minimum number
of samples (n) for FA should be such that n  30  (V 
3)/2, where V represents the number of variables. An important aspect in any multivariate analysis is the selection of
proper data set. Outliers in PM10 concentrations were removed before carrying out factor analysis, as they may not
represent the normally observed situation (Hopke, 2000). We
found the PM10 range of 45 to 235 g m3 to fit well into
model interpretations. It was observed that PM10 values out
of the above-mentioned range resulted in an unrealistic estimate of constant, C0 [in Equation (1)]. Overall, only five
data points of the PM10 concentrations (out of 105) were

FIG. 2. Twenty-four-hour average ambient air PM10 concentration (g m3) at IITK (after removing outliers).

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

TABLE 3.
Species
Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
NO3
SO42

MEAN AMBIENT AIR (24-H) PM10 CONCENTRATIONS

Summer (n  26)
(g m3)
22.15
0.54
9.10
21.33
2.08
0.62
5.19
0.16
0.22
1.76
0.76
2.06
7.52
10.25

853

7.23
0.44
4.85
8.65
0.87
0.18
1.81
0.05
0.11
0.83
0.32
1.26
2.64
2.54

Monsoon (n  41)
(g m3)
5.89
0.69
4.02
6.45
1.16
0.60
3.60
0.08
0.22
0.87
0.26
1.12
4.50
7.07

2.20
0.59
1.69
2.25
0.57
0.38
0.88
0.03
0.10
0.65
0.17
0.59
2.03
2.18

OF

VARIOUS SPECIES

AT

IITK.

Winter (n  33)
(g m3)

Other studies
(g m3)

0.45,b 5.8,c 0.2d

0.008d

14.67
0.37
3.24
14.15
1.15
1.08
2.78
0.09
0.32
5.52
0.30
1.33
18.02
10.10

5.08
0.29
1.26
3.79
0.45
1.89
0.91
0.03
0.13
1.90
0.09
0.49
5.91
2.74

2.7,b 23.8,c 1.2d

1.6,b 38.9,c 2.1d
0.6,b 0.05,c 0.3d
1.1,b 6.6,c 0.09d
0.004,a 40.59,b 1.2,c 0.02d
0.03,a 0.10,b 0.1,c 0.05d
0.8,d 1.3e
0.0628.4f
0.122.4f

aAbdul-Wahab and Yaghi (2004); bTripathi et al., (2004); cQuieterio et al., (2004); dHueglin et al., (2004); eChan et al., (1999); fTripathi et al.,
(1996).

deleted from the observed data points. Box-whisker plot

technique was used to identify outliers (Tukey, 1977). These
plots show median, hinges, and whiskers (lower fence and
upper fence), which are defined as:
lower fence  upper hinge  (1.5*difference in hinges)
upper fence  lower hinge  (1.5*difference in hinges)
The values less than the lower fence and more than the upper fence, are regarded as outliers. The outliers are also classified as outside values or far outside values depending on
how far one value is from the whiskers.
Results and Discussions
Figure 2 presents 24-h average PM10 concentrations in ambient air measured at the sampling site. As sample size is
large in each season, it can be concluded that PM10 levels, as
expected, show statistical significant seasonal variations (at
a 5% level of significance). Levels are lowest in the monsoon
season and show higher variability in summer due to high
speed winds prevalent in the summer season.
What is of even more significance is the variability in composition of particles from one season to another (Table 3). As
expected, elemental and ionic composition show seasonal
variability; especially, the levels of soil-derived elements like
Ca, Al, and Mg are highest in summer due to high speed
winds and dry soil, which gets airborne. SO42 and NO3
ions (indicator of secondary particles) have shown marked
TABLE 4.
Average
meteorological
parameters
Temperature (C)
Relative humidity (%)
Wind speed
(m s1)

variation in winter compared to other seasons. It is a little

surprising that levels of Ba were high in all seasons, which
is generally taken as an indicator of oil or diesel combustion.
The variation in elemental and ionic concentration (of PM10)
will obviously impact the neutralization of acid (gas phase)
present in atmosphere to form secondary particles. A comparison of levels (of metals and ions) obtained in this study
(after removing outliers) with levels found in other studies
is presented in Table 3.
During the sampling period, meteorological parameters
(temperature, relative humidity, wind speed, and wind direction) were recorded. Temperature and relative humidity
were measured by a temperature and relative humidity sensor, whereas wind speed and direction was measured by a
wind vane and a three-cup anemometer (wind monitor
WM251 Envirotech, New Delhi). Summary of the meteorological parameters is presented in Table 4:
Results and interpretation of PCA
Seasonwise as well as combined PM10 data were analyzed
to get information of seasonal/combined dominant sources
contributing toward PM10. In this study, 14 variables and 26,
41, and 33 PM10 samples (after removing outliers) for summer, monsoon, and winter seasons, respectively, were considered for PCA. Results of varimax rotated PCA carried out
on various ambient air components, and corresponding
sources are presented in Tables 5, 6, and 7 for summer, monsoon, and winter seasons, respectively, and in Table 8 for

METEOROLOGICAL PARAMETERS DURING SAMPLING PERIOD

AT

IITK.

Summer
(MarchJune)

Monsoon
(JulyOctober)

Winter
(NovemberFebruary)

30.60
51.30
0.67

29.70
73.80
0.50

18.00
62.80
0.43

854

SHUKLA AND SHARMA

TABLE 5.

VARIMAX ROTATED PC MATRIX

FOR

PM10 DURING SUMMER SEASON

AT

IITK.

Identified emission sources

Species

Soil and road dust

Secondary particles

Coal combustion

0.85
0.02
0.63
0.88
0.84
0.55
0.77
0.62
0.34
0.35
0.52
0.02
0.34
0.37
4.65
33.24
33.24

0.39
0.19
0.17
0.36
0.33
0.62
0.02
0.35
0.16
0.89
0.10
0.71
0.88
0.83
3.76
26.83
60.07

0.02
0.76
0.46
0.00
0.05
0.47
0.14
0.02
0.59
0.17
0.55
0.08
0.19
0.04
1.77
12.62
72.69

Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
NO3
SO42
Eigenvalue
% variance
Cumulative % variance

combined samples (100). The number of factors (PCs) was

selected such that the cumulative percentage variance explained by all the chosen factors is more than 70%. Only the
factors with eigenvalue more than 1 were chosen. The argument is that the normalized variables each carry one unit of
variance. Thus, if an eigenvalue is less than 1, then it carries
less information than one of the initial variables and is therefore not needed (Maenhaut and Cafmeyer, 1987). Because
higher factor loading of particular elements/ions (marker
species) in a factor can help in identifying the possible
sources (Henry et al., 1984), the number of factors identified
should represent the sources, which are relevant in the receptor domain.

TABLE 6.

VARIMAX ROTATED PC MATRIX

The basis of identifying a factor to a specific source type

(e.g., soil and road dust, secondary particles, etc.) is from
Table 9. Based on strong factor loadings of various variables
in factors, the sources have been identified in Tables 5, 6, 7,
and 8. Strong loading of Al, Ca, Fe, and Mg (crustal elements)
in all seasons/combined samples can be identified as originating from a crustal contribution that can be identified to
soil dust emission.
A study of the particulate sources in the receptor domain
indicates that there is no separate crustal source, other than
the natural wind-blown dust in ambient aerosol. However,
a significant amount of dust is present on the road and also
on the road shoulders, and it is kept in suspension by ve-

FOR

PM10 DURING MONSOON SEASON

AT

IITK.

Identified emission sources

Species
Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
NO3
SO42
Eigenvalue
% variance
Cumulative % variance

Soil and road dust

Secondary particles

Fuel combustion

Miscellaneous

0.85
0.00
0.03
0.95
0.88
0.26
0.84
0.08
0.07
0.33
0.02
0.19
0.36
0.34
3.57
25.50
25.50

0.37
0.02
0.04
0.19
0.32
0.01
0.06
0.26
0.68
0.90
0.13
0.01
0.89
0.90
3.25
23.18
48.68

0.08
0.93
0.12
0.02
0.04
0.21
0.05
0.83
0.33
0.17
0.91
0.30
0.12
0.15
2.74
19.57
68.25

0.20
0.09
0.84
0.14
0.11
0.56
0.36
0.06
0.10
0.02
0.05
0.75
0.02
0.01
1.80
12.87
81.12

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

TABLE 7.

VARIMAX ROTATED PC MATRIX

855
FOR

PM10 DURING WINTER SEASON

AT

IITK.

Identified emission sources

Species
Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
NO3
SO42
Eigenvalue
% variance
Cumulative % variance

Soil and road dust

Secondary
particles

Vehicular and biomass

burning emission

0.84
0.08
0.21
0.77
0.87
0.12
0.76
0.20
0.24
0.48
0.17
0.80
0.48
0.35
4.05
28.94
28.94

0.29
0.73
0.17
0.42
0.30
0.03
0.40
0.82
0.02
0.72
0.79
0.09
0.71
0.77
4.00
28.58
57.52

0.33
0.23
0.62
0.34
0.28
0.84
0.40
0.36
0.86
0.36
0.13
0.04
0.36
0.42
2.92
20.82
78.34

hicular movement. Hence, such a source can be identified as

the combination of soil and road dust. A major portion of
the percentage variance (33.2% during summer, 25.5% during the monsoon, 28.9% during winter, and 38.7% for combined samples) was explained by soil and road dust.
The second factor explained percentage variance of 26.8,
23.2, and 28.6% during summer, monsoon, and winter, respectively, and 21.5% for combined samples with high loading of NH4, NO3, and SO42. These ions (NH4, NO3,
and SO42) in the atmosphere are generally assumed to be
secondary fine particles produced from gas to particle con-

TABLE 8.

VARIMAX ROTATED PC MATRIX

version of NH3, NOx, and SO2 and, hence, this source can be
identified as the secondary particles.
Pb is associated with leaded gasoline (used at the time
of this study period) vehicular emissions. During the sample collection period, gasoline supply was not fully unleaded in Kanpur city. Hence, the factor comprising Pb can
be identified with vehicular emissions. Mn is normally associated with ferrous metal processing. Hence, cluster containing Mn can be traced as a complex mixture of metal
working emissions. There are several small- and mediumscale industries located in the receptor domain (although

FOR

PM10 COMBINED SAMPLES

AT

IITK.

Identified emission sources

Species
Al
As
Ba
Ca
Fe
K
Mg
Mn
Pb
NH4
F
Cl
NO3
SO42
Eigenvalue
% variance
Cumulative % variance

Soil and road dust

Secondary
particles

Vehicular and biomass

burning emission

0.84
0.08
0.81
0.87
0.87
0.10
0.83
0.81
0.04
0.02
0.75
0.61
0.01
0.50
5.42
38.71
38.71

0.41
0.07
0.21
0.37
0.07
0.13
0.18
0.21
0.41
0.96
0.07
0.13
0.95
0.74
3.02
21.54
60.25

0.14
0.53
0.07
0.13
0.16
0.75
0.05
0.35
0.70
0.22
0.34
0.05
0.23
0.14
1.76
12.60
72.84

856

SHUKLA AND SHARMA

TABLE 9.

INTERPRETATION

OF

FACTOR LOADINGS

Strong factor loadings

Al, Ca, Fe and Mg
NH4, NO3and SO42
Pb
As
Ba
K
Cl
Mn

TO

PM10 DURING VARIOUS SEASONS.

Identified sources

Soil & road dust (Kumar et al., 2001; Salvador et al., 2004)
Secondary particles (Salvador et al., 2004)
Vehicular emission (Kumar et al., 2001)
Coal combustion (Kumar et al., 2001)
Diesel combustion (USEPA, 1993)
Biomass burning (Godoy et al., 2005)
Marine aerosol (Harrison et al., 1996)
Metallurgical aerosol (Harrison et al., 1996)

Liu et al. (2003) classified the factor (PC) loadings as strong, moderate, and weak corresponding to
absolute loading values of 0.75, 0.750.50, and 0.500.30.

in the downwind direction). As is emitted during combustion of coal, firewood, and refuse burning (Sadasivan
and Negi, 1990; Kumar et al., 2001). A coal-based thermal
power plant (200 MW) is situated at a distance of about 3

FIG. 3.

km, and the presence of As may indicate contribution

from the power plant; however, it is a weak source, as wind
direction has never been from the power plant to the sampling site.

PM10 observed vs. PM10 estimated by regression model at IITK.

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

Soil and road dust and secondary particles are invariably
present in all three seasons/combined samples as major
sources contributing to PM10. A large chunk of the percentage variance (60, 49, and 57.5% during summer, monsoon,
and winter seasons, and 60% in combined samples) is being
explained by these two sources, indicating that these are the
prime sources contributing toward PM10 concentrations at
the sampling site. Other sources, such as vehicular emission,
oil combustion, coal combustion, refuse burning, and industrial source, are also contributing to PM10, but their contributions are small compared to combination of first two
sources This source identification is compatible with a general land use pattern of the sampling site.
In summary, total variance explained by various sources
was about 73% (during summer), 81% (during monsoon),
78% (during winter), and 73% (combined samples).
Quantification of source apportionment.
Further, there is a need to quantify the percentage contribution of these sources through multiple regression as per
the model given in equation (1). APCS were regressed on the
observed PM10 concentration to apportion the PM10 concentrations to each sample, and the following seasonal/combined samples models were obtained (after taking only statistically significant coefficients (p  0.05) of factors):
For summer season:

PM10  46.5  29.1*APCS1 

17.6*APCS2

For monsoon season:

PM10  27.7  17.6*APCS1 

13.9*APCS2

For winter season:

PM10  58.5  20.2*APCS1 

14.7*APCS2  11.8*APCS3

FIG. 4.

857
For combined samples: PM10  28.8  27.6*APCS1 
23.3*APCS2  9.4*APCS3
Percent contribution from each source (factor) can be calculated from the above models. For example, for the summer
season percent contribution from first factor (soil and road
dust) for the jth observation will be [ (29.1*APCS1)j*100/
(PM10)j]. The average percent contribution by a source in a
season can be calculated by taking average of the percent
contribution for all observations for that source.
Based on the above models, the estimated PM10 was plotted vs. observed PM10 (Fig. 3a to 3d). These figures signify
that computed and observed data of PM10 compare well
(R2  0.86). It may be seen that constant [C0 in equation (1)]
is estimated as a significant number indicating 26 to 43% contribution from the sources not apportioned by this exercise
these are referred to as other sources. These may include
long-range transport, refuse, and agricultural residue burning, tyre wear, industrial emission, etc.
Quantitatively, the percentage contribution of sources/
factors for different seasons and for combined samples was
calculated by seasonal models mentioned above, and the
same is presented graphically in Fig. 4a to 4d.
Figure 4 signifies that local soil and road dust (15 to 47%)
and secondary particles (21 to 26%) are the key sources in
all three seasons. It is interesting to note that in the winter
season when the PM10 levels are relatively lower than summer, the contribution of soil and road dust becomes smaller,
and the contribution of secondary particles and vehicular
source tend to become important (18%).
Formation of secondary products like SO42 and NO3 is
quite complex, as it depends not only on the concentration
of precursor pollutants but also on temperature and relative

Percentage contribution of different sources toward PM10 levels at IITK.

858

SHUKLA AND SHARMA

FIG. 5. (a) Contribution of soil and road dust (combined samples). (b) Contribution of secondary particles (combined
samples). (c) Contribution of vehicular and biomass burning emission (combined samples).

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

humidity. Earlier studies (Mszros and Horvath, 1984;
Willison et al., 1985; Keneyasu et al., 1995; Utsunomiya and
Shinzi, 1996) have reported high wintertime concentrations
of NO3 (compared to summer). Under the conditions of
high temperatures and low relative humidity in the summer,
particulate ammonium nitrate is volatile, and as a result,
NH4NO3 will be in the gaseous phase. Adams et al. (1999)
reported that the temperature 25C prevents formation of
a significant amount of particulate NH4NO3. In this study as
well, wintertime NO3 are higher than summertime levels
(Table 3). This implies formation of the inorganic secondary
fine particulate in winter will be high. This also signifies the
role of NH3, which is a precursor pollutant for formation of
NH4NO3 and (NH4)2SO4.
It is expected that SO42 concentrations increase in summer due to a high rate of photochemical activity and high
OH concentrations, which increases the oxidation of SO2
and its conversion rate to SO42 (Stockwell and Calvert, 1983;
Cadle, 1985; Utsunomiya and Shinzi, 1996; Khoder, 2002;
Gupta et al., 2003). In this study, levels of SO42 in summer
and winter are about the same (Table 3). The probable reason for similar levels of SO42 (in winter and summer) include better dispersion in summer, as both wind speed and
mixing height is higher in summer and can provide considerable dilution in summer. In other words, even if the production of SO42 may be high in summer, concentrations (in
g m3) may not be high compared to winter levels. During
the monsoon, concentrations of NO3 and SO42 were lowest due to washout of particles as a result of rain events.
Combined samples
Based on the model for combined samples obtained by regressing APCS on the observed PM10 concentrations, contribution for each identified source type toward PM10 concentrations is shown in Fig. 5ac. The rectangular boxes have
been drawn in Fig. 5ac to represent seasons (mentioned earlier) and to show seasonal variation in contribution of identified sources toward PM10 concentrations.
Soil and road dust
Figure 5a clearly suggests seasonal variability as observed
by analyzing the segregated seasonal observation (Fig. 4a).
Summer months show highest contribution of soil and road
dust toward PM10 concentrations due to high wind speeds
and dry conditions.
Secondary particles
Figure 5b also suggests seasonal variability as observed by
analyzing the segregated seasonal observation (Fig. 4b). Winter months show the highest contribution of secondary particles toward PM10 concentrations, which is similar to findings reported by Sharma et al. (2007).
Vehicular and biomass burning emission
Both Pb (indicator of vehicular emission) and K (indicator
of biomass burning) were found to be strongly loaded to the
third factor (identified as vehicular and biomass burning
emission) in Table 8. These two sources could not be separated out. Figure 5c shows contribution of these two sources
(samples contributing negatively (16%) were deleted). It is

859
argued that within a season vehicular contribution should
remain quite similar. No specific inventory is available, but
for sure, biomass burning in the urban area and burning of
agricultural residue in the field (to clear the field for next
crop) is a common practice, and thus emissions due to biomass burning in winter are going be significantly large. This
is reflected in high peaks of the contribution in the winter
season, which is in line with the segregated seasonal observation (Fig. 4c).
There are not many studies in India on source apportionment of PM10. The study by Kumar et al. (2001) in Mumbai,
India, for source apportionment of PM was conducted for
traffic intersection. Their study indicated a major contribution of 3341% from road dust or soil, followed by contribution from vehicles and marine aerosols. Similar results are
found in this study that soil and road dust contribute about
15% (in winter), 46% (in summer) to ambient air PM10 levels.
Kumar et al. (2001) could not resolve the contribution of
secondary particles, as that will require measurements of water-soluble ions like NH4, NO3, and SO42 in atmospheric
particles. Although SO42 levels were measured, in the absence of other ions it was weakly related to the road dust
component and the authors enquired Sulphate in the atmosphere is generally assumed to be a secondary fine particle
aerosol component produced from gas to particle conversion
of SO2 and hence its association with crustal elements is
questionable and should have been identified as a separate
source.
In the research presented here, focus has been to understand the formation of secondary particles and their quantitative contribution to ambient levels of PM10. The NH4 ion
has correlated strongly with both SO42 and NO3 in all seasons, and through FA-MR, a group of these ions is identified as one of the main components (of PM10) in all seasons.
Seinfeld and Pandis (1986) have explained the vital role that
NH3 plays in neutralizing acidic substances such as H2SO4
and HNO3, mostly as a sole neutralizer of acidic species,
which leads to formation of secondary particles. It is also reported that available NH3 will be first taken up by H2SO4
and then the remaining NH3 will be available to react with
HNO3 to produce ammonium nitrate.
Soil analysis from the region showed Ca (0.3%), Al (1.5%),
Na (0.4%), Mg (0.08%), NH4 (0.01%), SO42 (0.02%), and
NO3 (0.002%). This soil analysis signifies two things: (1) the
markers taken for soil dust (Ca, Mg, Al) in identifying
sources for PM10 are from the local soil, and (2) contribution
of soil to NH4, SO42, and NO3 concentrations in PM10 is

TABLE 10.

SEASONAL LEVELS OF PARAMETERS RELEVANT

SECONDARY PARTICLE FORMATION.

TO

Pollutant
NO3
SO42
NH4
NO2
NH3
HNO3

Winter (g m3)

15.57
12.76
6.77
12.9
19.09
2.30

5.95
5.11
4.10
3.81
5.70
0.74

SO2 levels were always below 5 g m3.

Summer (g m3)

7.02
9.41
1.43
12.87
18.69
6.90

1.66
1.57
0.99
3.89
6.0
2.16

860
insignificant. Thus, it can be said that NH4, SO42, and
NO3 are produced in the atmosphere (in fine mode) as secondary particles due to emissions from anthropogenic activities.
Further, the reaction rate constant of formation of ammonium nitrate aerosol is 2.04  104 m3 mol1 s1 (Baek and
Aneja, 2005) and for ammonium sulfate aerosol is 1.5  104
s1 (Harrison and Kitto, 1992). The rate constant of CaSO4
and MgSO4 formation is reported as 1.73  109 s1 by Pervez and Pandey (1992) while studying the marble damage
by SO2 acidity. As the reaction rate of CaSO4 and MgSO4 is
less (by several orders of 10) than that of ammonium nitrate
and ammonium sulfate aerosols, major portion of secondary
atmospheric particles will be (NH4)2SO4 and NH4NO3 and
in fine mode.
Because NH3 plays an important role, a study was taken
up recently at the same sampling location including the measurement of NH3 (Kishore, 2005). Table 10 presents the results of Kishore (2005) for parameters relevant to formation
of secondary particles.
In Table 10, NO3, SO42, and NH4 levels were statistically higher in the winter season compared to levels in summer, and HNO3 levels were higher in summer than in winter. The correlation between NH4 and SO42 and NH4 and
NO3 was in the range of 0.560.79 in winter (day and night)
and 0.750.88 (summer night); no correlation was found in
summer daytime. This study once again signifies the formation of secondary particles and vital role played by ammonia.
It may, therefore, be concluded that NH3 is playing a
significant role in formation of secondary particles that has
been seen through air quality sampling, receptor modeling, and also through the atmospheric chemistry. NH3
emitted due to anthropogenic activities is the major source
of NH4 in the atmosphere. The major sources responsible for NH3 emission are from the agricultural activities
(urea application) and intensive animal production facilities. India is primarily an agricultural economy, and it has
highest number of cattle population (an estimated 219.6
million cattle) in the world (http://dahd.nic.in/stat.htm).
No estimates of ammonia emission are available from the
study area. This study recommends that there is a need to
develop proper inventory of ammonia emission in India,
especially in the area where particulate pollution is high
to understand fully formation of secondary particles and
plan their control.
Conclusions
Simple source apportionment technique, factor analysismultiple regression, was applied to apportion the
sources of PM10 at a relatively less polluted site in Kanpur,
India. In three seasons studied, two important PM10
sources were soil-road dust (1547%) and inorganic secondary particles like (NH4)2SO4 and NH4NO3 (2126%).
Winter and monsoon conditions were conducive for formation of secondary particles due to moderate temperatures and high relative humidity. The main conclusion of
the research was that NH4 is a far more dominating ion
responsible for secondary particle formation; thus, the role
of ambient ammonia is most vital in overall atmospheric
chemistry of particulate formation.

SHUKLA AND SHARMA

Author Disclosure Statement
The authors declare that no competing financial interests
exist.
References
Abbi, Y.P. (2001). Thermal power generation: Key issues in India. TERI Newswire 7(19), 1.
Abdul-Wahab, S.A., and Yaghi, B. 2004. Total suspended dust
and heavy metal levels emitted from a workplace compared
with nearby residential houses. Atmos. Environ. 38, 745.
Adams, P.J., Seinfeld, J.H., and Koch, D.M. 1999. Global concentration of tropospheric sulfate, nitrate, and ammonium
aerosol simulated in a general circulation model. J. Geophys.
Res. 104, 13791.
Baek, B.H., and Aneja, V.P. (2005). Observation based analysis
for the determination of equilibrium time constant between
ammonia, acid gases, and fine particles. Int. J. Environ. Pollut.
23, 239.
Cadle, S.H. (1985). Seasonal variations in nitric acid, nitrate,
strong aerosol acidity, and ammonia in an urban area. Atmos.
Environ. 19, 181.
Chan, Y.C., Simpson, R.W., McTainsh, G.H., Vowles, P.D., Cohen, D.D., and Bailey, G.M. 1999. Source apportionment of
PM2.5 and PM10 aerosols in Brisbane (Australia) by receptor
modeling. Atmos. Environ. 33, 3251.
Gupta, A., Kumar, R., Kumari, K.M., and Srivastava, S.S. (2003).
Measurement of NO2, HNO3, NH3 and SO2 and related particulate matter at a rural site in Rampur, India. Atmos. Environ. 37, 4837.
Harrison, R.M., and Kitto, A.N. 1992. Estimation of the rate constant for the reaction of acid sulfate aerosol with NH3 gas from
atmospheric measurements. J. Atmos. Chem. 15, 133.
Henry, R.C., Lewis, C.W., Hopke, P.K., and Williamson, H.J.
(1984). Review of receptor model fundamentals. Atmos. Environ. 8, 1507.
Hopke, P.K. (2000). Workbook on data analysis. Prepared for
participants in the UNDP/RCA/IAEA, Subproject on Air Pollution and its Trends, RAS/97/030/A/01/18. IAEA, Vienna,
Austria.
Hueglin, C., Gehrig, R., Baltensperger, U., Gysel, M., Monn, C.,
and Vonmont, H. (2005). Chemical characteristics of PM2.5,
PM10 and coarse particles at urban, near-city and rural sites
in Switzerland. Atmos. Environ. 39, 637.
Kaneyasu, N., Ohta, S., and Murao, N. (1995). Seasonal variation in the chemical composition of atmospheric aerosols
and gaseous species in Sapporo, Japan, Atmos. Environ. 29,
1559.
Khoder, M.I. (2002). Atmospheric conversion of sulfur dioxide
to particulate sulfate and nitrogen dioxide to particulate nitrate and gaseous nitric acid in an urban area. Chemosphere 49,
675.
Kishore, S. (2005). Investigation into seasonal and diurnal formation of atmospheric nitrate. M. Tech. Thesis, Department
of Civil Engineering, Indian Institute of Technology, Kanpur.
Kumar, A.V., Patil, R.S., and Nambi, K.S.V. (2001). Source apportionment of suspended particulate matter at two traffic
junctions in Mumbai, India. Atmos. Environ. 35, 4245.
Liu, C.W., Lin, K.H., and Kuo, Y.M. 2003. Application of
factor analysis in the assessment of groundwater quality in
a blackfoot disease area in Taiwan. Sci. Total Environ. 313,
77.
Lodge, J.P. (1989). Methods of Air Sampling and Analysis. Chelsea,
MA: Lewis Publishers.

APPORTIONMENT OF ATMOSPHERIC PM10 IN KANPUR

Maenhaut, W., and Cafmeyer, J. (1987). Particle induced X-ray
emission analysis and multivariate techniques: An application
to the study of the sources of respirable atmospheric particles
in Gent, Belgium. J. Trace Microprobe Tech. 5, 135.
Mszros, E., and Horvth, L. (1984). Concentration and dry deposition of atmospheric sulfur and nitrogen compounds in
Hungary. Atmos. Environ. 18, 1725.
Pervez, S., and Pandey, G.S. (1992). Rate evaluation of marble
damage by SO2-acidity in the vicinity of stacks. Environ.
Geochem. Health 14, 103.
Quiterio, S.L., Silva, C.R.S., Arbilla, G., and Escaleira, V. 2004.
Metals in airborne particulate matter in the industrial district
of Santa Cruz, Rio de Janeiro, in an annual period. Atmos. Environ. 38, 321.
Sadasivan, S., and Negi, B.S. (1990). Elemental Characterization
of atmospheric aerosols. The Science of the Total Environment
96, 269.
Seinfeld, J.H., and Pandis, S.N. (1986). Atmospheric chemistry
and physics. New York: John Wiley & Sons.
Sharma, M., McBean, E.A., and Ghosh, U. (1995). Prediction of
atmospheric sulfate deposition at sensitive receptors in Northern India. Atmos. Environ. 29, 2157.
Sharma, M., Kishore, S., Tripathi, S.N., and Behera, S.N. (2007).
Role of atmospheric ammonia in the formation of inorganic
secondary particulate matter: a study at Kanpur, India. J. Atmos. Chem. 58, 1.

861
Stockwell, W.R., and Calvert, J.G. (1983). The mechanism of the
HO-SO2 reaction. Atmos. Environ. 17, 2231.
SYSTAT. (2004). Richmonda, CA: SYSTAT Software, Inc.
Thurston, G.D., and Spengler, J.D. (1985). A quantitative
assessment of source contributions to inhalable particulate matter pollution in metropolitan Boston. Atmos. Environ. 19, 9.
Tripathi, B.D., Chaturvedi, S.S., and Tripathi, R.D. (1996). Seasonal variation in ambient air concentration of nitrate and sulfate aerosols in a tropical city, Varanasi. Atmos. Environ. 30,
2773.
Tripathi, R.M., Kumar, A.V., Manikandan, B.,S., Mahadevan,
T.N., and Puranik, V.D. (2004). Vertical distribution of atmospheric trace metals and their sources at Mumbai, India. Atmos. Environ. 38, 135.
Tuckey, J.W. (1977). Exploratory data analysis. Reading, MA:
Addison-Wesley.
Utsunomiya, A., and Shinzi, W. (1996). Temperature and humidity dependence on aerosol composition in northern
Kyushu, Japan. Atmos. Environ. 30, 2379.
Willison, M.J., Clarke, A.G., and Zeki, E. M. (1985). Seasonal variation in atmospheric aerosol concentration and composition
at urban and rural sites in northern England. Atmos. Environ.
19, 1081.
U.S. EPA. (1999). Compendium of Methods for the Determination of
Inorganic Compounds in Ambient Air. Compendium of Method
IO-3.2, EPA/625/R-96/010a. U.S. EPA.