Journal of Atmospheric and Solar–Terrestrial Physics 191 (2019) 105044

Contents lists available at ScienceDirect

Journal of Atmospheric and Solar-Terrestrial Physics

journal homepage: www.elsevier.com/locate/jastp

Long-term temporal and spatial analysis of SO2 over Gauteng and

Mpumalanga monitoring sites of South Africa
S.K. Sangeetha a, *, V. Sivakumar b
a
School of Agricultural, Earth and Environmental Sciences, College of Agriculture, Engineering and Science, University of KwaZulu-Natal, Westville Campus, Private Bag
X5400, Westville, Durban 4000, South Africa
b
School of Chemistry and Physics, College of Agriculture Engineering and Science, University of KwaZulu-Natal, Westville Campus, Private Bag X5400, Westville, Durban
4000, South Africa

A R T I C L E I N F O A B S T R A C T

Keywords: This paper focused on the temporal and spatial variability of sulphur dioxide (SO2) from 2004 to 2013 over 36
Temporal variations ground-based (GB) stations located in Gauteng and Mpumalanga provinces of South Africa. The assessment was
Spatial analysis based on the in situ SO2 data, where the GB stations were sorted into different groups and in addition, kriging-
SO2
based analysis and OMI SO2 data (2004–2013) were deployed to study their seasonal spatial variability. It was
Remote sensing
Seasonal variability
observed that Pretoria west, Witbank and Randwater had high SO2 levels and standard deviations. Although,
winter played a major role in having peak SO2 values in most of the stations, it was surpassing that in spring the
SO2 levels of Witbank and Pretoria west stations dropped and Randwater had the highest standard deviation of
11.2 ppbv in October. A prominent seasonal variation was seen in all the groups, excluding Mpumalanga power
stations (MP) group. This was particularly evident in the Vaal Triangle group. Highveld group ranked topmost
level with elevated SO2 values proceeded by the MP group in all the temporal time scales. The spatial based
analysis further proved that the highest SO2 levels were centred around major industrial regions of Mpumalanga,
regardless of all the seasons.

1. Introduction Municipality and Private sectors (Pretorius et al., 2015; Spalding-Fecher

Air pollutants, particularly sulphur dioxide (SO2) has become an
important concern as it impacts on human health in many ways, like
chronic obstructive pulmonary diseases (COPD) (Ghozikali et al., 2015),
early asthmatic attacks in children (Clark et al., 2015) and also on the
environment such as acid deposition and sulphate aerosol formation. In
the past few decades, the developing countries in Asia, particularly
China and India, increased the global SO2 emissions by � 60% because
of their rapid industrial growth (Smith et al., 2011). African countries
provide � 5% of the total global SO2 emissions and this will increase to
about 30% by 2030 (Liousse et al., 2014). South Africa as a developing
nation rich in coal mines, with progressive industrial growth still facing
challenges regarding air pollution. The major percentage (� 93%) of
electricity generations is based on coal based power plants in the
country (Menyah and Wolde-Rufael, 2010), due to the plentiful coal
reserves in the country and releases nearly 1.5 million tons of SO2
annually (Lloyd, 2002). Nearly � 95% of electricity generation is owned
by ESKOM and the remaining percentage is maintained by the
and Matibe, 2003). The health outcomes from these municipalities
owned power station emissions, however might be more vulnerable than
that of ESKOM, as they are old and have low stack heights (Spalding-­
Fecher and Matibe, 2003). Thus in order to gain insight of SO2 over a
specific region, it is worth to analyse its trend both on temporal and
spatial basis. A continuous monitoring and reporting of air quality data
is essential for effective air quality management practices to ensure that
the pollutants are maintained within the standards (Macpherson et al.,
2017). Spatial distribution of air pollutants and determining their sea­
sonal trends using GIS has become a helping tool in this era. A precise
regional mapping of air quality is important for a constructive envi­
ronmental mitigation and monitoring plan (Hamm et al., 2015). The
temporal and spatial assessments of pollutants have been made in many
countries over the past years. Zou et al. (2011) pursued AERMOD
dispersion modelling and applied Ordinary Kriging (OK) method for the
period from 1996 to 2002, to study the spatial and temporal trend of SO2
in Dallas, USA. In addition, they mapped the variations of annual SO2
concentrations according to specific sources and found that the traffic

* Corresponding author.
E-mail address: trijashan@yahoo.com (S.K. Sangeetha).

https://doi.org/10.1016/j.jastp.2019.05.008
Received 12 December 2018; Received in revised form 15 March 2019; Accepted 13 May 2019
Available online 23 May 2019
1364-6826/© 2019 Elsevier Ltd. All rights reserved.
S.K. Sangeetha and V. Sivakumar
Fig. 1. Geo-location of 36 monitoring stations with power stations located near by some stations. The station names are abbreviated as in Table 1 and the group
names given in brackets. Each group has been given specific colours. (For interpretation of the references to colour in this figure legend, the reader is referred to the
Web version of this article.)
emission sources were the main contributor for significant variations in
annual SO2 concentrations. Wang et al. (2014) reported a study on
spatial and temporal variations of six criteria pollutants including SO2 in
31 cities in China utilising the hourly downloaded data from March
2013 to February 2014 and found that high SO2 levels in the northern
region during winter were due to coal consumption for domestic heat­
ing. In South Africa, Lourens et al. (2011), used passive sampling
method to conduct the spatial and temporal distributions of gaseous
pollutants for a year from August 2007–July 2008, over eight repre­
sentative sampling sites for the Highveld region and perceived the
contribution of industrial sources or the sites located downwind from
these sources for high SO2 concentrations. A decadal study on monthly
mean SO2 concentrations from 1995- to 2005 time period at four remote
sites in southern Africa also confirmed that the highest SO2 levels are
concentrated in the industrial hot spot region of Mpumalanga (Martins
et al., 2007). Venter et al. (2012), assessed a two year time period
(2008–2010) of study on the air quality in the Bushveld Igneous Com­
plex of South Africa, where they had selected Marikana in Northwest
province as the representative site and explored the tall stack industrial
sources for SO2 emissions. Although a comprehensive exposition on SO2
can be drawn from these studies, none of them deal with a complete
spatial representation of SO2, as conventional passive sampling and
either a single representative or a remote site was used.
The main aim of the current study is to investigate the temporal and
spatial variations of SO2 derived from ground-based (GB) data over 36
monitoring stations in South Africa. Since these GB measurements pro­
vide vital information even on small temporal resolutions, the first part
analyzes the temporal variation of the selected stations starting from the
smaller resolutions (diurnal variations) through the daily, monthly
variations and finally ending with the annual basis. Second part focusses
on the spatial aspect, utilising the Ozone Monitoring Instrument (OMI)
data and Geographic Information System (GIS) based approach.
2. Data and methodology
2.1. Ground-based data
South African government has a wide variety of air quality moni­
toring stations, which are under the control of several network systems
2
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
across the nation. The air quality data are collected, maintained, ana­
lysed and controlled by the South African Weather Service (SAWS)
through the South African Air Quality Information System (SAAQIS).
The raw data for this study was provided by the respective networks that
are under the SAAQIS system. All the GB monitoring stations follow a
similar method of SO2 measurement, i.e., the principle of absorbing
ultraviolet light commonly known as the pulsed fluorescence SO2 ana­
lyser, has been employed. Almost all the monitoring stations are
managed by South African National Accreditation System (SANAS),
which provides quality assurance of test and calibration of data. The SO2
data were collected across 36 monitoring stations of different network
providers from a time period of 2004–2013. Of these, 21 were from
Gauteng, 14 from Mpumalanga and 1 (Zamdela) from the Free State
province. Fig. 1, shows the Geo Location of monitoring stations with
some major power stations. These stations cover major coal industries,
power stations, air shed priority areas namely Vaal Triangle and High­
veld priority areas and traffic related emissions in the metro cities of
Pretoria and Johannesburg, which are of primary concern in relation to
major pollutants. Eleven stations monitor the urban pollution, six in
industrial areas and the rest falls in semirural or rural and mixed up
categories.
2.2. Methodology
It has been noted that there was no consistency in the time period of
the stations and therefore, they were grouped according to the time
period of SO2 data available and also on the type of network. For
example, it has been found that some of the Johannesburg stations had
two different time periods although the network provider is the same for
both (City of Johannesburg), so they were grouped into two categories
Johannesburg 1 and 2. On the other hand, the miscellaneous group
contains stations from the Tshwane Air Quality and ESKOM networks.
Such groupings of data was done by Qu et al. (2010), in their study on
spatial and temporal analysis of PM10 concentrations in 86 cities of
China, where they have grouped the cities according to the time period
of Air Pollution Index data and the geographical location of the cities.
Table 1 shows the details of each group with their station names and
time period. The abbreviations for each group and their station are given
in brackets. It is to be noted that the stations under a specific group
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Table 1
Details of each grouping of stations with their time period, Group Name, Station Names, Network provider, Province and Site classification where Urb: Urban, Res:
Residential, Ind: Industrial, Com: Commercial.
Group Name Stations Network Province Source Type

Pretoria (PR) Booysens (Boo) Tshwane Air Quality Gauteng Urb

2009–2013 Mamelodi (Mam) Tshwane Air Quality Gauteng Suburb/Res
Olievenhoutbosch (Oli) Tshwane Air Quality Gauteng Urb
Pretoria West (Pre) Tshwane Air Quality Gauteng Urb/Ind
Rosslyn (Ros) Tshwane Air Quality Gauteng Ind
Johannesburg (J1) Diepsloot (Die) City of Johannesburg Gauteng Township
2009–2011 Ivory Park (Ivo) City of Johannesburg Gauteng Urb
Highveld (HV) Ermelo (Erm) DEA Mpumalanga Res & Ind
2008–2013 Hendrina (Hen) DEA Mpumalanga Res
Middleburg (Mid) DEA Mpumalanga Urb
Secunda (Sec) DEA Mpumalanga Res
Witbank (Wit) DEA Mpumalanga Urb
Vaal Triangle (VT) Diepkloof (Dpk) DEA Gauteng Res
2007–2013 Kliprivier (Kli) DEA Gauteng Semirural
Sebokeng (Seb) DEA Gauteng Res
Sharpeville (Sha) DEA Gauteng Res
Three Rivers (Thr) DEA Gauteng Res
Zamdela (Zam) DEA Free State Res
Johannesburg (J2) Alexandra (Ale) City of Johannesburg Gauteng Township
2004–2010 Buccleuch (Buc) City of Johannesburg Gauteng Urb
Jabavu (Jab) City of Johannesburg Gauteng Township
Orange Farm (Ora) City of Johannesburg Gauteng Township
Mpumalanga power stations (MP) Camden (Cam) ESKOM Mpumalanga Urb
2011–2013 Elandsfontein (Ela) ESKOM Mpumalanga Rural
Grootdraaidam (Gtd) ESKOM Mpumalanga Urb
Komati (Kom) ESKOM Mpumalanga Rural
Leandra (Lea) ESKOM Mpumalanga Urb
Majuba (Maj) ESKOM Mpumalanga Rural
Phola (Pho) ESKOM Mpumalanga Urb
Verkykkop (Ver) ESKOM Mpumalanga Rural
Miscellaneous group (MG) Grootvlei (Gtv) ESKOM Mpumalanga Urb
2012–2013 Randwater (Rdw) ESKOM Gauteng Ind
Bodibeng (Bod) Tshwane Air Quality Gauteng Res & Com
Ekandustria (Eka) Tshwane Air Quality Gauteng Res & Ind
West Rand (WR) Mogale city (Mog) West Rand District Municipality Gauteng Semirural
2013 Randfontein (Rnf) West Rand District Municipality Gauteng Ind

Fig. 2. Overall monthly mean variations of meteorological parameters (a: Relative Humidity; b: Temperature; c: Rainfall and d: Wind speed) for Highveld, Vaal
Triangle and West Rand regions.

(except miscellaneous group) were under the control of same network
provider, which also added a reasonable justification for grouping the
stations and result analysis.
2.3. Background meteorological observations
The atmospheric background meteorological parameters play a vital
role on the seasonal spatial variability of SO2. Thus in order to study
their relationship with SO2 concentration, surface meteorological data
namely rainfall (RF), relative humidity (RH), temperature (temp), wind
speed (WS) and wind direction (WD) provided by the South African
Weather Service (SAWS) over Highveld, Vaal Triangle and West Rand
networks were analysed. Fig. 2 (a, b, c and d) shows the overall monthly
variations of RH, temp, RF and WS respectively for the three networks. It
should be noted here that there were some data gaps either due to
missing data or due to the data filtering method. Hence, there was an

3
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Table 2 networks, where it was slightly warmer during mid-winter period,

Seasonal variations of dominant wind directions for the selected group. although its average temperature was significantly below the latter over
Network Winter Spring Summer Autumn the rest of season (see. Fig. 2b). There were two enhancements in Vaal
\Season Triangle rainfall data (Fig. 2c) during the transitional periods with its
Highveld NNE (7%) & NNE (6.8%); NE (6.8%); NNE (6.8%); NE greatest value in October and it can be inferred that Highveld had lower
NNW (5%) NE (6.6%) ENE (6.4%) (5.6%) & NW RF values in comparison with the former, even though it had missing
(4%) data. Finally, there was a visible seasonal variation in WS data over Vaal
Vaal- NNE (6.6%) NNE (7.7%); NE (7.7%); NNE (6.6%); Triangle and West Rand networks as outlined in Fig. 2d, where the WS
triangle & NNW (5%) NNW (5.7%) NNE (6.4%) NNW (5.5%)
West-rand NNE (3.5%); NE (3.4%); NNE (2.4%) NNE (1.9%); NW
started to decrease from late summer to late winter, but peaked off in
NW (2.6%) NNE (3.2%) (1.3%); WNW spring. Contrary to this, Highveld did not show any seasonal variation
(1.4%) and apart from this, its WS value was always high above the other two
N-North, E-East, W-West and S-South (NNE: North-North East Directions, same
networks from autumn to early spring. Another distinctive point noticed
for all other combinations). was that Highveld WS value dropped off in spring and summer as against
the Vaal Triangle and West Rand WS data in the corresponding periods.
unavailable RF data in West-Rand and there were data gaps within the Table 2 below exhibits the details of seasonal change in wind di­
Vaal Triangle network. Relative humidity (RH) displays that of the three rection for the three networks. The major dominant wind direction were
networks, Vaal Triangle is the driest and all of them showed a noticeable tabulated for each network based on the wind-rose plot (Figures are not
seasonal variation of elevated values from late spring to summer shown). It was noticed that the dominant WD for the three networks was
thereafter dropped off from autumn until late winter (see. Fig. 2a). The North-Northeast, irrespective of all seasons and North-Northwest was
difference seen was that Highveld had the driest time in winter, whereas the second predominant WD, except in summer. Generally, winter and
Vaal Triangle and West-Rand was in September. Temperature mea­ autumn were also influenced by winds originated from the west,
surements show that Highveld behaved uniquely from the other two whereas for spring and summer it was from the eastward direction.

Fig. 3. Diurnal variation of SO2 for each group and their respective stations.

4
S.K. Sangeetha and V. Sivakumar
Fig. 4. Winter diurnal variation of SO2 for each group and their respective stations.
There was a distinctive feature in spring over Vaal Triangle network,
where the North-Northwest winds were sustained in that period. In
summer, the wind was slightly eastward in Highveld network, where it
was East-Northeast, besides Northeast. The above information is used to
interpret the results, which are discussed in the current study.
2.4. Spatial interpolation
In this study, ordinary kriging (OK) method has been used to inter­
polate the spatial distribution of SO2 concentration between the moni­
toring stations, based on the GBdata for each season. Kriging is a geo-
statistical approach that is widely used in different fields such as
remote sensing, environmental science and mining (Bayraktar and
Turalioglu, 2005). It depends on spatial autocorrelation that not only
predicts the spatial measurement of values at certain points, but also
measures the accuracy of the predictions. Numerous studies have been
based on spatial dispersion of atmospheric pollutants using the
kriging-based approach. Xia et al. (2016) used OK for seasonal
spatio-temporal variations of major pollutants during 2013 in Shenzhen
city, China using GB data from 19 monitoring sites. Similarly, Feng et al.
(2014) used the OK method and population distribution data to analyse
the spatial distribution of domestic heating and found that the SO2 were
5
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
highly concentrated in the suburbs of Beijing during winter.
2.5. OMI SO2 data
The OMI SO2 data over planetary boundary height level (PBL) from
2004 to 2013 has been employed in order to better understand the
spatial variation of SO2 over the GB sites. It is a level 2 product with
version 003 based on Principal Component Analysis (PCA) algorithm.
For more information about OMI features and PCA algorithm one can
refer to OMI data User’s Guide (2012) and Li et al. (2013) respectively.
Our earlier research work had also employed OMI SO2 measurements
over South Africa where we focussed on a particular station and a case
study (see. Sangeetha et al., 2017, 2018).
3. Results
3.1. Ground-based SO2 temporal variations
3.1.1. Diurnal variation
3.1.1.1. Mean diurnal variation. In accordance with the results from
Fig. 3, it can be seen that Pretoria west of Pretoria group, showed both
S.K. Sangeetha and V. Sivakumar
Fig. 5. Summer diurnal variation of SO2 for each group and their respective stations.
high SO2 levels and standard deviation up to even � 14 � �15 ppbv as
the site was impacted by both industrial and residential sources. Similar
to this, the stations in the Highveld group especially Witbank, Secunda
that were representative of both type of emissions, showed double peaks
although not prominently and also high SO2 levels were maintained all
throughout the entire day starting from morning until late evening (10 a.
m. - 7 p.m.). Stations like Witbank, Hendrina in this group even reached
above 20 ppbv. An exclusive industrial signature was shown by Mpu­
malanga power stations group with constant high levels from morning
until evening (local time from 10 a.m. to 5 p.m.) where high SO2 levels
of � 20 ppbv were seen in some stations like Komati and Grootdraaidam.
This could be a combination of both industrial sources and the
mixing of pollutants after the inversion layer breakage in the morning
hours (Sundstro €m et al., 2015; Vakkari et al., 2013; Venter et al., 2012).
Groups such as Johannesburg 1 & 2, Vaal Triangle (both industrial and
household combustion) (Feig et al., 2014) and stations like Mamelodi
and Olievenhoubosch in the Pretoria group with predominance of do­
mestic fuel burning, showed double peaks. However, stations like
Alexandra, Jabavu and Orange farm of Johannesburg 2 group could also
be associated with traffic, as they peaked at 7 a.m. and 6–7 pm, which
correlated better with traffic hours. It should be noted that the Buc­
cleuch (Johannesburg 2) impacted by vehicular emissions (Department:
6
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
Environmental Affairs, 2012) also had higher values (~ 12.5 ppbv)
comparable to industrial group, as it lies in close proximity to the N1 and
M1 highways (Kgabi and Sehloho, 2012). Johannesburg 1 and the West
Rand showed the lowest SO2 value of all groups. Miscellaneous group
stations like Rand water and Grootvlei, showed high SO2 levels only in
the midday (11am-3pm), where Rand water had higher standard de­
viations than Pretoria west (� 22 ppbv) during that hours. West Rand
group comprising of Randfontein and Mogale city did not show any
prominent diurnal variation, as their SO2 levels were stable throughout
the entire period.
3.1.1.2. Seasonal diurnal variations. Winter: The winter diurnal varia­
tion of each group and their stations are shown in Fig. 4. It can be
deduced from Fig. 4 that a single peak was seen in Pretoria (morning)
except Pretoria west which had 3 peaks. The West Rand group showed
an evening peak, whereas Johannesburg 1 & 2, Vaal Triangle, Highveld
groups, Bodibeng station of Miscellaneous group showed double peaks
which might be due to domestic burning of coal products for space
heating in winter. Industrial signature was shown in Mpumalanga power
stations group, where high values were maintained from late morning
until evening due to the tall stack industrial emissions (Mugabo, 2011).
In addition, some stations like Komati and Elandsfontein showed minor
S.K. Sangeetha and V. Sivakumar
Fig. 6. aDaily mean variation of SO2 for each group (PR: Pretoria, J1: Johannesburg group 1, HV: Highveld, VT: Vaal Triangle, J2: Johaanesburg group 2, MP:
Mpumalanga power stations, MG: Miscellaneous group, WR: West Rand). b: Boxplot representing the spatial variation of each group.
night peaks around 8 p.m., which might be due to the anti-cyclonic
stability over the Highveld which played a useful part in the accumu­
lation of pollutants (Collett et al., 2010). The SO2 levels of these Mpu­
malanga power stations, Highveld groups and the Pretoria west station
were the highest as they rose up to � 18 ppbv, especially Witbank, which
was � 29 ppbv in the evening. This high SO2 peak values in winter could
be the influence of more prominent multi-inversion layers when
compared to other seasons (Venter et al., 2012). Similar to the average
diurnal variation, here also the standard deviations of Pretoria west and
Rand water stations were more � 20 ppbv. Zamdela showed higher
values when compared to other stations in the Vaal Triangle group and
this could be due to the neighbouring industrial emissions from the
north of the site (Feig et al., 2014) like Sasolburg, NATREF and domestic
fuel burning.
Summer: The summer diurnal variation of each group and their
stations are depicted in Fig. 5 and where, Pretoria and Johannesburg 1
7
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
groups did not show any peak values, especially Johannesburg 1 where
the lowest SO2 levels of � 2 ppbv were recorded. Johannesburg 2, Vaal
Triangle, Miscellaneous, West Rand groups showed single peak in the
morning after that the SO2 levels got reduced. This might be due to the
increase in the mixing height of the pollutants in summer (Korhonen
et al., 2014) as the frequency of occurrence (40%) of subsidence
inversion layer was lower (Thomas and Scorgie, 2006) and thus enabling
dispersion of pollutants (Lourens et al., 2011). Although the SO2 levels
of Highveld and Mpumalanga power stations group were within
12 ppbv, few stations like Witbank, Hendrina, Elandsfontein, Groot­
draaidam and Komati showed high values � 18 ppbv during the oper­
ating hours of the industries.
Spring and autumn: The variations in spring were almost similar to
those exhibited in summer in all the groups (see Appendix A,
Figure A.1). A remarkable difference was that Witbank and Pretoria west
showed lower SO2 levels in spring compared to other seasons, when
S.K. Sangeetha and V. Sivakumar
Fig. 7. Intra-seasonal variation of SO2 for each group and their respective stations.
Ermelo and Hendrina SO2 values were higher than that of Witbank and
similarly Rosslyn’s SO2 concentration was more than Pretoria west. It
could be assumed that the westerly disturbances and anticyclonic cir­
culations that were the dominant features in spring (Garstang et al.,
1996; Tyson et al., 1996a) that could favour for the dispersion of pol­
lutants away from the source region. During spring, there were unstable
atmospheric conditions with extra-tropical nature in October over
Gauteng, leading to strong wind shears (Dyson et al., 2015). Earlier
study on aerosol variations also noticed that biomass-burning activity
transporting smoke particles from Mozambique and Madagascar was
one of the major aerosol sources in Pretoria during spring (Adesina et al.,
2014). Hence, there might be a huge conversion of SO2 into sulphate
aerosols during this period. Apart from this, wind direction plays a
significant role with regard to this. During winter and spring, the air
masses over Pretoria was NW originating from desert and arid regions
(Kumar et al., 2017) and thus had dearth of industrial plumes. All these
factors could have led to a decline in SO2 levels in Pretoria west in
spring. Considering the Witbank WD, it seems that the dominant WD
was mostly eastward as ENE in winter, ENE & ESE in summer and
autumn (Figures are not given, ref. Table 2). Therefore, it carried SO2
laden plumes from neighbourhood industrial regions during this time. In
contrary to this, the WD in spring was mostly northward as NE, where it
8
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
sparingly collected industrial plumes. Also, other conditions like high
WD and RH in spring still intensify the condition, thus leading to further
decrease in the amount of SO2 during spring over Witbank. In autumn,
similar trends were seen in all the groups as in summer, however the SO2
levels and standard deviations of Witbank (� 24 � 20 ppbv), Rand water
(� 20 � 20 ppbv) and Komati (� 21 � 17 ppbv) were as high as in winter
(see. Appendix A, Figure A.2).
3.1.2. Daily mean variations
The daily mean variations of each group and the representation of
their spatial variations in the form of boxplots are shown in Fig. 6 (a &
b). Fig. 6a proved that in the beginning (2009) Pretoria group showed
higher levels when the SO2 values almost reached the National Ambient
Air Quality Standards (NAAQS) of 48 ppbv. From 2012, it was below
18 ppbv. This was the impact of high SO2 surface loadings of Pretoria
west (See. Appendix A, Figure A.3). Secondly, Pretoria group had the
highest spatial variations of all groups followed by Johannesburg 2 and
Highveld groups as seen in Fig. 6b. The Vaal Triangle (0.5–15 ppbv),
Miscellaneous (0.5–18 ppbv), Johannesburg 2 (0.5–20 ppbv) and West
Rand (1–13 ppbv) groups generally showed similar ranges, whereas the
ranges of the Highveld and Mpumalanga power stations groups were
double that of these groups (Fig. 6a). The Highveld group showed stable
S.K. Sangeetha and V. Sivakumar
Fig. 8. Inter-seasonal variation of SO2 for each group and their respective stations.
high SO2 values exceeding 18 ppbv throughout the entire period.
Especially Witbank, Ermelo and Hendrina stations had consistent high
values of above 30 ppb in the middle of each year (nearly twice that of
the Vaal Triangle group) (See. Appendix A, Figure A.3), which again
proved the dominance of coal based industries in this group. Similarly,
Mpumalanga power stations group were at higher levels (0.5–18 ppbv),
which were comparable to those of the Highveld group. The boxplot also
proved that the excessive SO2 amounts were seen in the Highveld group
where the median was ~10 ppbv. The 75th percentile value of Highveld,
Mpumalanga power stations and Johannesburg 2 groups was 13.2 ppbv,
10.29 ppbv and 8.8 ppbv respectively (Fig. 6b).
Contrary to this, it was found that Johannesburg 1 had the lowest
SO2 values of all the groups and did not exceed 10 ppbv in all 3 years as
shown in Fig. 6a (2009–2011). When compared it with the Highveld
group, it was noticed that the 75th percentile value of Johannesburg 1
group (3.65 ppbv) was significantly lower than the 25th percentile of
Highveld group (7.11 ppbv) (Fig. 6b).
Thus, the daily mean variations confirmed that the Highveld and the
Mpumalanga power stations group had the highest SO2 level. This was
consistent with the study conducted on major sources of air pollution in
South Africa using air borne instruments between 2003 and 2006, where
it was proved that the plumes from coal fired power plants and
9
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
petrochemical facilities typically have higher SO2 levels (Burger, 2016).
Secondly, the SO2 levels of some stations near Johannesburg and Pre­
toria like Jabavu, Orange farm, Alexandra and Pretoria west were
sometimes comparable with that of major industrial based regions of the
Highveld and Vaal Triangle groups. This could be that the cities were
impacted by the air masses from the Highveld region rich in SO2 (Beukes
et al., 2013). Additionally, other activities like burning of low grade coal
for cooking and heating purposes and vehicular emissions could
contribute to the pollution episodes in the megacities (Lourens et al.,
2012).
3.1.3. Seasonal variations
The seasonal variations of each group and their stations were broadly
categorised into intra-seasonal (Fig. 7) and inter-seasonal variations
(Fig. 8). A more detailed study of the overall monthly averages and the
standard deviations can be found in Appendix B (Table B.1). It should be
noted that since the West Rand group had only 2013 data, its Inter-
seasonal variations were discussed only.
3.1.3.1. Intra-seasonal variations. In the beginning of the time period i.
e. from 2009 to 2010, concentrations at all the stations in the Pretoria
group, except Pretoria west, ranged from 1.5 to 15 ppbv, while Pretoria
S.K. Sangeetha and V. Sivakumar
west had a few higher SO2 values and standard deviations � 20 ppbv �
10 ppbv both in winter and summer of these years. From 2012, all the
stations in this group were below 10 ppbv. It was found that the
Johannesburg 1 group had low SO2 values. Both Diepsloot and Ivory
Park stations started increasing from autumn, however Diepsloot
remained below 5 ppbv even in winter of all the years (2009–2012),
while Ivory Park were slightly higher than 5 ppbv in winter. Similarly,
the Miscellaneous group had low ranges, where they were below
10 ppbv. Nonetheless, the standard deviation for Rand water was
sometimes higher than 5 ppbv even reaching up to 11 ppbv. It could be
seen that prominent winter peaks were shown in all the stations in the
Vaal Triangle for the entire period (2007–2013) and ranged from 1.5 to
10 ppbv in all the seasons. Sharpeville and Zamdela, on the other hand,
slightly exceeded 10 ppbv in winter. This was in agreement with the
study conducted on air pollutant concentrations where it was found that
only the Vaal Triangle area showed unique winter extremities when
compared to other hot spot regions, which could be due to the presence
of inversion layers at 750-700 hpa and calm winds (Ncipha, 2011). The
Highveld and Mpumalanga power stations groups showed slightly
higher ranges than other groups as Middleburg and Secunda stations in
the Highveld group ranged between 2 ppbv and 15 ppbv, while Hen­
drina, Ermelo and Witbank showed even higher values (>15 ppbv)
irrespective of the seasons, sometimes even 20 ppbv. The standard de­
viations of Witbank and Ermelo were high � 10 ppbv, while Middleburg
had the lowest value (below 15 ppbv even in winter). There was no
decrease seen in the Highveld group. The Johannesburg 2 group
resembled the Vaal Triangle group for its ranges and prominent winter
peaks. The ranges were from 2 to 12.5 ppbv, but Buccleuch had slightly
higher ranges from 3 to 17 ppbv. Winter peaks were seen for all the
stations, especially Buccleuch, where the SO2 values exceeded 15 ppbv,
however from 2008, the peak values started to decrease for all the sta­
tions. Mpumalanga power stations as said earlier had high ranges (from
2.5 to 15 ppbv), in spite of the fact that Majuba and Verkykkop stations
had the lowest values (below 10 ppbv). Grootdraaidam had the highest
range especially in 2013, when it had high SO2 values and a high
standard deviations of > 20 � 10 ppbv irrespective of all the seasons.
Leandra had an intermediate range and other stations like Camden,
Komati, Elandsfontein and Phola occasionally exceeded 15 ppbv. Some
stations in this group did not show any seasonal variations as they
sometimes peaked even in summer periods also.
3.1.3.2. Inter-seasonal variations. Almost all the groups followed a
general seasonal pattern of increasing trend from autumn, reached their
maximum level in winter, and later on decreased slowly from spring.
Lourens et al. (2011) and Martins et al. (2007) also saw such homoge­
nous seasonal patterns, where the SO2 values displayed maximum level
in winter, which were probably related to the increase in electricity
demand and trapping of pollutants near the surface due to surface
inversion during winter. Furthermore, local climatic conditions played a
significant role as seen from the meteorological data above in section
2.3, when the winds were calm and RH was low in winter. Calm winds
inhibit the transport of pollutants and the low RH value restricts the
pollutant absorption (Xue and Yin, 2014). A vice versa situation
happened in spring which had high wind speeds and high RH amount,
which led to low SO2 concentrations in spring, which continued until
early summer. However, Pretoria West (Pretoria group), showed regular
higher SO2 levels and deviations of �16 � 12 ppbv even in summer. A
probable explanation could be because of the low stack emission height
of the Pretoria west power station, owned by the Tshwane municipality
(Spalding-Fecher and Matibe, 2003). The SO2 must have been released
well below the mixing height and thus led to downwash (Thomas et al.,
1963). Witbank in the Highveld group had the highest SO2 value of all
36 stations, especially in winter, when it reached up to � 20 ppbv.
Likewise, Hendrina in the same group had SO2 values of 14 � 2 ppbv
maintained even in spring, apart from autumn and winter, whereas the
10
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
Middleburg and Secunda stations had low values (below 10 ppbv)
except in June and July. Ermelo station exhibited two peak values, one
in April and the other in July (� 20 ppbv). The standard deviation of this
Highveld group (Appendix B, Table B.1) was high all throughout the
entire season, which could probably be related to climatological factors.
This was prominently reflected in Fig. 2, where Highveld had higher
standard deviations with respect to meteorological conditions especially
RH and WS. The sudden change in these factors could also contribute to
high spatial variations of SO2 over this group, as these are the essential
components of SO2 sinks in the earth’s atmosphere.
The Johannesburg 1 group had low SO2 values, even in winter, as the
SO2 levels were maintained below 5 ppbv, which could because a
resultant of two main reasons. Firstly, the stations were at urban location
where the consumption of electricity was preferred over coal. Secondly,
the use of Basa-njengo-Magogo (BnM) techniques which is a top down
fire lightning procedure, introduced in the low income settlements of
Diepsloot and Ivory Park (2012–2013 National Air Quality Officer’s
Annual Report on Air Quality Management, 2014). Moderate levels were
seen in the Vaal Triangle group as their SO2 values were below 8 ppbv
during all seasons, except Sharpeville and Zamdela, which were �
8.5 ppbv in July and August. Buccleuch, Orange farm and Jabavu from
the Johannesburg 2 group, showed a gradual increasing trend in autumn
and winter and a gradual decreasing trend in spring, whereas Alexandra
increased sharply in July to reach a maximum value of 10.5 ppbv. All the
stations in this group, except Buccleuch, maintained a concentration
below 11 ppbv, even in winter. The stations in the Mpumalanga power
stations group did not show a steady trend, as they increased or
decreased suddenly, especially Komati, Grootdraaidam and Elandsfon­
tein, but Majuba showed the lowest value. Rand water in the Miscella­
neous group, had unusually low values in winter and increased in spring
when it had a higher standard deviation of 11.7 ppbv in October and to a
less extent in November and April. Even Grootvlei had lower values in
winter and higher values in spring and summer. The West Rand group
was similar to that of the Johannesburg 1 group, where Randfontein had
smaller ranges (from 4.8 to 5.9 ppbv) and Mogale city was below 7 ppbv,
except in January where it increased to 11.2 ppbv.
Overall, from the seasonal variations, it can be concluded that all the
groups followed a general seasonal pattern, but Mpumalanga power
stations did not show any consistency in their trends, as they were
dominated by power stations. Secondly, the Johannesburg 1 group had
the lowest (� 5 ppbv) and Highveld group had the highest SO2 values
followed by Mpumalanga power stations group (17 ppbv). Pretoria west,
Witbank and Rand water (in spring) had high standard deviations
(greater than 10 ppbv).
In addition to this, a study on the variations among the groups were
investigated. Their difference in inter-seasonal averages was tabulated
in Table B.2 in Appendix B. The table shows that despite Mpumalanga
power stations group had higher ranges than any other groups, it
exhibited low SO2 values when compared with Highveld group, espe­
cially in July (~5.52 ppbv). A possible reason might be due to the in­
clusion of Witbank’s high SO2 level in the latter. Another key
observation was that urban and township predominant groups namely
Pretoria and Johannesburg 2 had higher degrees of SO2 values than Vaal
triangle group. Thus proving that high vehicular emissions and air
pollution from the city could also pose a serious threat.
3.1.4. Inter-annual variation
The inter-annual averages for each group calculated from the overall
mean of their respective stations are depicted in Fig. 9 and detailed in
Appendix B (Table B.3). [Insert Fig. 9 here]
It can be deduced from the above figure that Highveld group showed
the highest SO2 value of all groups of ~12 ppbv during 2009–2011. This
was equivalent to the ambient SO2 concentration inferred from the
multi-scale atmospheric transport and chemistry (MATCH) model which
was based on an emission inventory provided by the South African
Department of Environmental Affairs and Tourism (DEAT), where the
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Fig. 9. Inter-annual averages of SO2 for each group.

Fig. 10. Spatial interpolation of SO2 over 36 monitoring stations in each season with colour codes below representing SO2 ranges in ppbv. (For interpretation of the
references to colour in this figure legend, the reader is referred to the Web version of this article.)

11
S.K. Sangeetha and V. Sivakumar
Fig. 11. Longitude-Latitude-OMI SO2 variations in DU at Planetary Boundary Layer (PBL) height over South Africa for different seasons.
mean SO2 levels were higher than 10 ppbv near the industrial sources
(Zunckel et al., 2000). Although the air quality of monitoring stations
within the Mpumalanga power stations group was impacted by power
generation, still the SO2 level of this group was maintained at
~8.5 ppbv. This might be due to the high-altitude stack emission, which
disperse the pollutants above the inversion layer (Lloyd, 1987; Spal­
ding-Fecher and Matibe, 2003). Compared to the above two groups, the
Pretoria group also showed a similar high annual average and high
standard deviation of ~9 ppbv in 2009, but later on showed a declining
trend. The reason was that there was a sharp decline in the SO2 values of
Pretoria west from 2009 to 2011, where � 89% decrease was seen be­
tween 2009 and 2013 (Table B.3).
3.2. Spatial variations
The spatial interpolation of SO2 by means of Kriging in each season
using the GB data from 36 monitoring stations (Fig. 10) had been
compared with the spatial mapping of OMI SO2 data in Dobson Units
(DU) (Fig. 11) over the same time period (2004–2013) for a more
effective interpretation of the spatial variability, which are as follows:
Autumn: In autumn, kriging-based modelling showed that the re­
gions located near the power stations like Witbank, Komati and
Elandsfontein, had the highest SO2 levels (� 10–12.5 ppbv ranges),
whereas their outskirts had 7.5–10 ppbv. However, the major portion of
the Gauteng province, which covers mega cities such as Pretoria and
Johannesburg and the southern part of the Mpumalanga Province, with
Verkykkop, Majuba and Grootvlei stations, had a low SO2 range of
2.5–5 ppbv. The OMI SO2 measurements at Planetary Boundary Layer
(PBL) also confirmed high SO2 columnar values of ~1.5 DU to 2 DU
around industrial zones.
Winter: The spatial interpolationin winter was more or less similar
to that in autumn, except high ranges were seen in the regions near to
power stations like Komati, Elandsfontein etc., (12.5–15 ppbv) and the
northern part of Gauteng covering Booysens, Rosebank, Pretoria west
(7.5–10 ppbv) when compared to that in autumn. This could be the high
SO2 emissions because of energy demand in winter from the power
plants like Hendrina, Komati, Kendal and Pretoria west, Rooiwal power
stations in Mpumalanga and Gauteng provinces respectively. However,
other factors like persistence of stable inversion layers, low wind speed
12
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
and low mixing depth also play a vital role, thereby hindering the
dispersion of pollutants in winter. However, OMI SO2 measurements
showed the highest number of scattered plumes in winter, which might
be due to the combination of subtropical anticyclones and transient
ridging anticyclones prevailing in winter, both contributing to 80% of
anti-cyclonic circulations in mid–winter (Tyson et al., 1996a).
Spring: Both OMI SO2 measurements and ground-based measure­
ments (spatially interpolated by kriging-based approach) showed a well-
marked discrimination between the eastern and western parts of Gau­
teng province, where the former showed a higher SO2 level than the
latter. Another point, which was noticed by both observations, was that
the air parcels were more inclined towards an eastward direction. This
feature was seen in all the parts including the central and peripheral
regions of the selected provinces. It could be a result of the westerly
wave disturbances, which is a spring phenomenon with a frequency of
41% occurrence in October (Tyson et al., 1996b).
Summer: In summer, the two datasets showed a controversial maps
around the industrial hub region, where OMI data showed the highest
SO2 value, whereas the spatial interpolation had the lowest SO2 value of
all the seasons. This could be due to the mixing height level increase in
summer, where the vertical mixing coefficient increases. In addition, the
maximum rainfall in summer (Davis, 2011), could have led to a
large-scale wet deposition of SO2 in the summer season, thus lowering
the surface SO2 levels. Another key concept noticed was that OMI data
showed a more confined air parcel than any other season. Nevertheless,
they both confirmed that the western parts of Gauteng experienced
higher SO2 values in summer than spring, as entirely Northeast plumes
(Table 2) were seen in summer over the central industrialised zones
moving towards the West.
4. Summary
In this paper, we have investigated the temporal and spatial varia­
tions of SO2 by utilising GB instruments and OMI SO2 data. The temporal
variations showed that Pretoria west, Witbank and Rand water had both
consistent high SO2 and high standard deviations based on diurnal, daily
and seasonal analysis. The seasonal diurnal variations inferred that
double peaks were shown in most of the groups in winter, except the
Mpumalanga power stations group and during spring, the SO2 values of
S.K. Sangeetha and V. Sivakumar
Witbank and Pretoria west had been underscored by other stations in
their respective group. The overall temporal based analysis showed that
the Highveld group had the highest SO2 values followed by the Mpu­
malanga power stations group, whereas the Johannesburg 1 group had
the lowest SO2 values. The seasonal variations proved that all the groups
followed a general seasonal variation except Mpumalanga power sta­
tions group, as they were at high levels even in summer. Rand water had
a peculiar low level in winter and showed high SO2 levels and standard
deviations in October (spring).
The spatial analysis based on the kriging method and OMI spatial
maps, showed that high SO2 levels were found in the major industrial
areas of the Highveld, irrespective of the season. Thus from both the
temporal and spatial based analysis, it was shown that the highest SO2
levels were centralised in the industrial zones of the Mpumalanga
province. However, there was a disagreement between the two spatial
measurements in summer, when OMI measurements showed extreme
SO2 values as against kriging approach using ground-based measure­
ments. Furthermore, it was found that the SO2 plumes were under the
influence of wind speed and direction. This was in agreement with the
Appendix A
13
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
present study on meteorological data and previous studies on wind
climatology, which proved that the SO2 level not only depends on the
emission sources, but also on the meteorological conditions, which need
to be explored further in future studies.
Declarations of interest
The author declare that there are no conflicts of interest.
Acknowledgements
One of the author, S.K. Sangeetha is thankful to South African Na­
tional Space Agency (SANSA) for providing Ph.D study bursary support.
The authors would like to thank South African Air Quality Information
System (SAAQIS) for providing the raw data in the Ground based
monitoring stations and South African Weather Service (SAWS) for
utilising the meteorological data. We are thankful to the two anonymous
reviewers for their helpful comments and suggestions to improve the
quality of the research work presented here.
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Fig. A.1. Spring diurnal variation of SO2 for each group and their respective stations

Fig. A.2. Autumn diurnal variation of SO2 for each group and their respective stations

14
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Fig. A.3. Daily mean variation of SO2 for each group and their respective stations.

Appendix B

15
 Table B.1Inter-seasonal average values of SO2 and standard deviations of each group with their corresponding stations.

Stations\ Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Pretoria group
Booysens 4.22 � 2.70 4.26 � 3.23 4.69 � 2.63 5.60 � 3.40 6.97 � 3.67 10.69 � 2.30 6.47 � 2.23 8.30 � 5.23 2.67 � 2.12 3.60 � 2.61 2.68 � 1.45 3.13 � 2.03
Mamelodi 4.44 � 2.19 4.99 � 2.10 4.55 � 2.45 3.19 � 2.11 5.14 � 3.01 6.80 � 2.95 5.06 � 2.65 4.68 � 2.24 6.64 � 4.38 3.38 � 3.68 2.96 � 1.60 2.64 � 2.04
Olievenhoutbosch 4.38 � 3.93 5.66 � 2.99 5.56 � 2.95 3.81 � 3.17 6.50 � 3.28 7.31 � 3.03 7.93 � 3.08 5.80 � 2.24 5.19 � 2.06 3.40 � 2.15 3.34 � 1.7 3.85 � 2.64
Pretoria west 15.07 � 12.06 14.86 � 12.55 12.07 � 8.34 9.16 � 5.80 13.92 � 9.65 14.71 � 10.74 16.49 � 11.67 9.64 � 10.03 10.74 � 8.20 3.07 � 2.10 3.58 � 2.73 6.17 � 9.34
Rosslyn 7.19 � 3.31 5.99 � 2.46 6.59 � 3.25 7.39 � 3.23 8.36 � 4.09 8.16 � 3.96 11.59 � 4.80 6.89 � 3.34 8.38 � 3.35 6.97 � 2.70 5.81 � 2.30 6.09 � 1.87
S.K. Sangeetha and V. Sivakumar

Pretoria 7.06 � 4.64 7.15 � 4.36 6.69 � 3.11 5.83 � 2.48 8.18 � 3.40 9.53 � 3.25 9.51 � 4.59 7.06 � 1.96 6.73 � 3.06 4.09 � 1.62 3.68 � 1.24 4.38 � 1.65
Johannesburg 1
Diepsoot 0.98 � 0.56 1.52 � 1.18 2.57 � 1.17 1.80 � 1.12 2.46 � 1.45 3.11 � 1.67 3.42 � 1.30 3.03 � 0.96 2.69 � 0.98 1.72 � 0.60 0.73 � 0.29 0.92 � 0.56
Ivory park 1.12 � 0.67 1.21 � 0.57 1.40 � 1.04 2.15 � 1.14 3.58 � 1.91 5.20 � 1.93 5.30 � 1.76 4.34 � 1.73 4.07 � 1.19 3.21 � 1.48 1.93 � 1.04 0.95 � 0.64
Johannesburg 1 1.05 � 0.09 1.37 � 0.22 1.99 � 0.83 1.97 � 0.24 3.0 � 0.79 4.16 � 1.47 4.36 � 1.33 3.69 � 0.92 3.38 � 0.97 2.47 � 1.05 1.33 � 0.85 0.93 � 0.02
Highveld group
Ermelo 8.55 � 5.33 6.57 � 4.88 10.41 � 6.90 17.11 � 5.21 10.67 � 8.29 15.45 � 11.21 20.83 � 4.62 16.56 � 6.31 14.09 � 7.64 12.32 � 4.79 8.77 � 3.92 7.87 � 4.62
Hendrina 8.87 � 5.60 9.87 � 3.85 8.93 � 3.98 13.67 � 6.14 15.64 � 6.91 11.0 � 7.73 14.01 � 5.80 15.22 � 6.44 14.59 � 5.34 15.63 � 5.70 14.12 � 6.03 14.14 � 5.50
Middleburg 5.37 � 3.12 4.68 � 3.14 4.81 � 2.76 5.89 � 3.32 7.83 � 3.79 9.65 � 3.70 7.73 � 3.89 8.02 � 4.47 5.42 � 4.33 6.47 � 5.88 2.64 � 2.25 4.62 � 2.56
Secunda 6.49 � 3.10 6.86 � 3.04 6.08 � 3.16 6.17 � 3.20 7.65 � 4.10 8.51 � 3.21 11.93 � 5.15 9.06 � 4.49 7.65 � 3.19 7.25 � 4.27 7.15 � 5.05 6.07 � 3.09
Witbank 16.94 � 7.15 15.53 � 5.37 13.26 � 5.35 19.84 � 8.80 20.06 � 7.04 20.76 � 8.58 20.28 � 10.07 18.03 � 8.32 12.61 � 5.52 10.01 � 5.58 9.46 � 4.81 14.34 � 7.75
Highveld 9.24 � 4.53 8.70 � 4.24 8.70 � 3.38 12.54 � 6.33 12.37 � 5.37 13.08 � 5.04 14.95 � 5.59 13.38 � 4.54 10.87 � 4.10 10.34 � 3.75 8.43 � 4.14 9.41 � 4.55
Vaal triangle group
Diepkloof 4.47 � 3.05 4.32 � 2.93 5.32 � 2.85 5.39 � 2.63 6.64 � 2.94 6.05 � 3.01 6.80 � 2.97 5.61 � 2.68 4.18 � 2.40 3.56 � 2.66 3.36 � 2.02 2.98 � 2.14
Kliprivier 3.05 � 2.09 3.63 � 2.10 3.41 � 1.81 3.39 � 1.77 4.19 � 1.90 4.35 � 1.90 5.22 � 2.14 4.62 � 1.98 4.26 � 1.60 3.04 � 1.55 2.62 � 1.58 2.77 � 1.93
Sebokeng 2.55 � 2.05 4.41 � 3.23 4.83 � 2.70 3.91 � 2.37 5.44 � 3.20 6.29 � 3.08 6.08 � 2.77 4.81 � 2.54 4.20 � 1.93 3.20 � 1.89 3.45 � 2.12 2.52 � 1.81
Sharpeville 5.17 � 3.65 5.68 � 3.06 5.67 � 3.27 5.75 � 2.94 7.47 � 3.87 7.39 � 3.61 8.97 � 5.01 7.50 � 4.24 5.94 � 3.01 4.29 � 2.83 4.93 � 2.87 3.81 � 2.27
Three rivers 3.38 � 2.39 4.59 � 2.20 4.01 � 1.98 3.99 � 1.97 4.57 � 1.93 4.73 � 1.86 4.87 � 1.83 5.08 � 2.39 4.59 � 1.84 3.68 � 2.27 2.81 � 1.66 3.47 � 1.97
Zamdela 5.27 � 2.50 6.02 � 3.07 5.85 � 3.19 5.67 � 3.17 6.78 � 3.79 7.65 � 3.99 7.36 � 4.01 8.46 � 4.16 6.94 � 3.62 5.82 � 3.27 6.09 � 3.28 5.69 � 3.08
Vaal triangle 3.98 � 1.14 4.78 � 0.90 4.85 � 0.96 4.68 � 1.03 5.85 � 1.31 6.08 � 1.34 6.55 � 1.50 6.01 � 1.58 5.02 � 1.15 3.93 � 1.02 3.88 � 1.35 3.54 � 1.15

16
Johannesburg 2 group
Alexandra 5.65 � 3.28 7.98 � 5.11 8.18 � 5.02 6.49 � 5.10 7.29 � 4.06 8.05 � 4.26 10.58 � 4.78 5.47 � 2.91 5.36 � 2.26 5.68 � 2.45 4.87 � 2.80 5.37 � 3.30
Buccleuch 7.38 � 4.26 9.02 � 3.97 10.44 � 3.84 10.68 � 4.62 12.96 � 4.86 12.31 � 4.61 11.47 � 4.24 9.85 � 3.83 9.13 � 4.15 7.35 � 3.76 7.06 � 4.14 6.45 � 3.63
Jabavu 2.82 � 1.90 5.39 � 3.14 6.32 � 3.12 9.35 � 7.36 8.18 � 3.64 9.67 � 3.90 9.66 � 3.77 7.41 � 3.45 5.81 � 2.79 4.23 � 2.35 3.95 � 2.38 3.97 � 2.66
Orange farm 2.66 � 2.16 3.16 � 2.26 3.71 � 2.45 3.27 � 2.10 5.22 � 3.34 6.29 � 3.45 6.85 � 3.55 5.08 � 2.46 3.37 � 1.90 2.31 � 1.61 2.13 � 1.49 2.12 � 1.62
Johannesburg 2 4.63 � 2.29 6.39 � 2.63 7.16 � 2.85 7.45 � 3.28 8.41 � 3.27 9.08 � 2.55 9.64 � 2.0 6.95 � 2.18 5.92 � 2.39 4.89 � 2.14 4.50 � 2.05 4.48 � 1.87
Mpumalanga power stations
Camden 4.79 � 3.63 5.40 � 3.50 7.64 � 4.89 10.58 � 5.55 11.39 � 6.06 11.38 � 4.79 9.90 � 6.21 7.96 � 3.93 7.64 � 4.84 4.50 � 3.80 10.95 � 8.01 6.27 � 3.78
Elandsfontein 10.25 � 3.95 8.94 � 4.95 9.03 � 4.74 12.50 � 7.06 10.30 � 5.65 13.22 � 5.53 10.54 � 4.80 12.05 � 4.92 8.87 � 5.05 15.21 � 6.93 11.49 � 4.42 12.68 � 5.80
Grootdraaidam 16.11 � 6.89 15.47 � 8.52 10.32 � 4.08 11.30 � 5.30 15.84 � 6.29 11.69 � 5.11 15.25 � 12.52 16.25 � 10.93 10.26 � 6.47 11.53 � 6.09 10.28 � 7.20 12.19 � 8.73
Komati 8.91 � 4.98 8.97 � 6.80 10.06 � 4.87 9.75 � 4.39 11.77 � 5.94 15.98 � 7.69 12.58 � 8.51 13.88 � 5.72 11.91 � 5.31 10.06 � 5.93 11.31 � 4.89 9.41 � 3.67
Leandra 6.34 � 3.44 6.84 � 3.47 6.56 � 3.17 6.39 � 3.09 7.76 � 4.19 6.55 � 4.91 7.59 � 3.37 7.94 � 5.63 7.12 � 3.47 7.75 � 5.34 11.45 � 5.95 7.20 � 4.76
Majuba 2.65 � 2.32 3.16 � 2.26 2.50 � 2.02 2.48 � 2.04 1.89 � 1.56 2.28 � 1.58 1.90 � 2.03 2.92 � 2.42 3.73 � 4.03 3.10 � 2.48 3.40 � 3.02 3.84 � 3.11
Phola 10.87 � 5.48 10.46 � 5.01 11.19 � 4.81 10.30 � 4.80 13.47 � 4.22 15.28 � 5.59 14.49 � 4.73 12.21 � 4.03 11.69 � 4.62 10.58 � 5.72 7.46 � 2.68 8.94 � 4.14
Verkykkop 4.23 � 3.41 2.57 � 2.42 3.10 � 1.95 2.80 � 1.84 4.57 � 3.26 5.86 � 5.49 3.27 � 2.34 4.96 � 2.69 3.44 � 3.78 3.92 � 2.38 2.97 � 2.83 3.82 � 3.92
Mpumalanga power 8.02 � 4.40 7.72 � 4.20 7.55 � 3.28 8.26 � 3.88 9.62 � 4.63 10.28 � 4.88 9.44 � 4.91 9.77 � 4.57 8.08 � 3.26 8.33 � 4.26 8.67 � 3.62 8.05 � 3.39
Miscellaneous group
Grootvlei 7.55 � 3.80 8.15 � 3.37 7.80 � 3.52 5.92 � 3.53 6.71 � 3.46 7.05 � 3.68 7.17 � 3.15 6.40 � 3.97 6.35 � 3.38 6.96 � 3.06 8.02 � 3.75 9.54 � 3.41
Randwater 3.26 � 3.28 5.92 � 4.12 5.10 � 3.82 8.35 � 7.95 5.24 � 5.10 4.50 � 4.55 3.62 � 2.28 4.64 � 3.97 4.26 � 4.96 8.49 � 11.73 8.77 � 6.51 6.65 � 4.43
Bodibeng 5.63 � 4.10 3.45 � 2.55 4.02 � 2.85 4.97 � 2.96 5.94 � 2.80 7.25 � 3.38 9.18 � 5.41 7.64 � 3.25 5.38 � 3.27 3.63 � 3.49 1.95 � 1.54 4.09 � 2.93
Ekandustria 4.59 � 2.38 6.24 � 2.38 6.45 � 2.19 6.05 � 2.89 3.91 � 1.76 3.30 � 1.48 2.74 � 1.90 3.02 � 1.53 3.35 � 1.83
Miscellaneous group 5.48 � 2.14 5.84 � 2.35 5.64 � 1.94 5.96 � 1.68 6.03 � 0.61 6.32 � 1.25 6.50 � 2.31 5.65 � 1.69 4.82 � 1.32 5.46 � 2.71 5.44 � 3.45 5.91 � 2.80
West rand group
Mogale city 11.11 � 4.07 6.56 � 2.61 7.49 � 2.55 7.31 � 2.38 6.47 � 2.42 8.06 � 2.84 5.26 � 2.05 5.74 � 2.60
Randfontein 5.23 � 2.88 5.58 � 2.38 5.14 � 2.19 5.11 � 2.0 5.94 � 2.37 4.85 � 2.36 5.00 � 2.26
West Rand 8.17 � 4.15 6.07 � 0.69 5.14 � 0 6.30 � 1.68 7.31 � 0 6.47 � 0 7.0 � 1.50 5.06 � 0.28 5.37 � 0.52
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

Table B.2Inter-seasonal average differences of SO2 between each group.

Difference in inter-seasonal averages between Pretoria group and other groups

Pretoria - Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Johannesburg 1 6.01 5.78 4.70 3.85 5.15 5.37 5.15 3.38 3.35 1.62 2.35 3.44
Highveld 2.18 1.55 2.01 6.71 4.19 3.54 5.44 6.31 4.14 6.25 4.75 5.03
Vaal triangle 3.08 2.37 1.84 1.14 2.32 3.45 2.95 1.04 1.71 0.1573 0.20 0.84
Johannesburg 2 2.43 0.76 0.47 1.62 0.24 0.45 0.13 0.11 0.81 0.81 0.83 0.10
Mpumalanga power 0.96 0.58 0.86 2.43 1.44 0.75 0.06 2.70 1.36 4.24 4.99 3.67
Miscellaneous group 1.58 1.31 1.05 0.13 2.15 3.22 3.00 1.42 1.90 1.37 1.77 1.54
West rand 1.11 1.08 1.55 0.46 0.87 3.06 2.51 2.01 1.36 NAN NAN NAN
Difference in inter-seasonal averages between Johannesburg 1 group and other groups
Johannesburg 1 -
Highveld 8.19 7.33 6.71 10.56 9.35 8.92 10.60 9.69 7.50 7.87 7.10 8.47
Vaal triangle 2.93 3.41 2.86 2.71 2.83 1.92 2.19 2.33 1.64 1.46 2.55 2.61
Johannesburg 2 3.58 5.019 5.17 5.48 5.40 4.92 5.28 3.27 2.54 2.43 3.17 3.55
Mpumalanga power 6.97 6.36 5.56 6.30 6.60 6.12 5.08 6.08 4.71 5.86 7.33 7.11
Miscellaneous group 4.43 4.47 3.65 3.99 3.01 2.16 2.14 1.96 1.44 2.99 4.11 4.98
West rand 7.12 4.70 3.15 4.32 4.29 2.32 2.64 1.37 1.99 NAN NAN NAN
Difference in inter-seasonal averages between Highveld group and other groups
Highveld -
Vaal triangle 5.27 3.92 3.85 7.85 6.52 6.99 8.40 7.36 5.85 6.40 4.55 5.86
Johannesburg 2 4.61 2.31 1.54 5.09 3.95 3.99 5.32 6.43 4.95 5.44 3.93 4.93
Mpumalanga power 1.22 0.98 1.15 4.27 2.74 2.80 5.52 3.62 2.79 2.01 0.2369 1.3610
Miscellaneous group 3.77 2.86 3.06 6.5772 6.34 6.76 8.45 7.73 6.05 4.88 2.99 3.50
West rand 1.07 2.63 3.56 6.24 5.06 6.60 7.95 8.32 5.50 NAN NAN NAN
Difference in inter-seasonal averages between Vaal triangle group and other groups
Vaal triangle -
Johannesburg 2 0.65 1.61 2.31 2.77 2.56 3.00 3.09 0.94 0.90 0.96 0.63 0.94
Mpumalanga power 4.04 2.95 2.70 3.59 3.77 4.20 2.89 3.76 3.06 4.40 4.79 4.51
Miscellaneous group 1.50 1.06 0.79 1.28 0.18 0.24 0.05 0.37 0.20 1.53 1.57 2.37
West rand 4.19 1.30 0.29 1.61 1.46 0.40 0.45 0.96 0.35 NAN NAN NAN
Difference in inter-seasonal averages between Johannesburg 2 group and other groups
Johannesburg 2 -
Mpumalanga power 3.40 1.34 0.39 0.82 1.21 1.20 0.20 2.82 2.16 3.44 4.16 3.57
Miscellaneous group 0.85 0.54 1.52 1.50 2.39 2.77 3.14 1.30 1.09 0.56 0.94 1.43
West rand 3.54 0.32 2.02 1.15 1.10 2.60 2.64 1.90 0.55 NAN NAN NAN
Difference in inter-seasonal averages between Mpumalanga power group and other groups
Mpumalanga power -
Miscellaneous group 2.54 1.88 1.90 2.30 3.60 3.96 2.94 4.12 3.26 2.87 3.22 2.13
West rand 0.15 1.66 2.41 1.97 2.32 3.81 2.44 4.71 2.72 NAN NAN NAN
Difference in inter-seasonal averages between Miscellaneous group and West rand group
Miscellaneous group-
West rand 2.70 0.22 0.51 0.34 1.28 0.16 0.50 0.59 0.54 NAN NAN NAN

Table B.3Inter-annual average values and standard deviations of each group with their corresponding stations.

Stations Years

Pretoria group
2009 2010 2011 2012 2013
Booysens 3.33 �1:48 3.0 � 2.21 3.77 � 2.71 6.62 � 3.92 3.03 � 2.01
Mamelodi 4.89 � 2.27 4.92 � 2.75 2.5 � 2.59 5.85 � 3.52 2.41 � 1.98
Olievenhoutbosch 4.64 � 3.05 4.75 � 3.49 4.12 � 2.73 5.72 � 2.92 4.98 � 3.07
Pretoria west 22.91 � 10.98 16.24 � 10.75 7.26 � 6.86 3.49 � 3.20 6.58 � 4.55
Rosslyn 10.62 � 5.29 7.39 � 3.44 7.26 � 3.51 7.68 � 3.62 6.31 � 2.39
Pretoria 9.28 � 8.12 7.26 � 5.25 4.98 � 2.16 5.88 � 1.54 4.66 � 1.88
Johannesburg 1 group
2009 2010 2011
Diepsloot 2.31 � 1.39 1.87 � 1.54 2.04 � 1.09
Ivory park 2.87 � 1.75 2.95 � 2.21 3.16 � 1.87
Johannesburg 1 2.59 � 0.4 2.41 � 0.77 2.61 � 0.8
Highveld group
2008 2009 2010 2011 2012 2013
Ermelo 10.20 � 4.76 11.95 � 5.17 15.78 � 7.63 13.74 � 6.60 5.63 � 7.29 9.47 � 2.29
Hendrina 10.40 � 4.97 13.26 � 7.08 15.10 � 6.0 16.94 � 5.30 10.05 � 5.25 12.68 � 6.12
Middleburg 7.67 � 4.61 8.20 � 5.10 5.89 � 3.25 7.11 � 3.43 3.92 � 3.30 5.73 � 3.89
Secunda 4.97 � 2.40 6.90 � 3.45 9.42 � 4.25 5.57 � 4.45 8.54 � 3.63 7.47 � 3.50
Witbank 16.0 � 8.81 15.0 � 7.61 18.53 � 8.21 16.32 � 8.91 14.58 � 7.43
Highveld 8.31 � 2.54 11.26 � 3.72 12.24 � 4.37 12.38 � 5.80 8.89 � 4.79 9.99 � 3.64
Vaal triangle group
2007 2008 2009 2010 2011 2012 2013
Diepkloof 5.88 � 2.97 4.80 � 2.86 5.06 � 3.04 6.38 � 2.44 4.32 � 3.12 4.02 � 2.71 3.98 � 2.65
Kliprivier 4,42 � 2.17 3.61 � 2.09 3.36 � 1.93 3.67 � 2.07 3.91 � 2.04 3.72 � 1.94 3.65 � 1.78
Sebokeng 5.33 � 2.90 4.54 � 3.06 4.06 � 2.63 4.76 � 2.94 4.56 � 3.28 3.50 � 2.17 3.97 � 2.38
(continued on next page)

17
S.K. Sangeetha and V. Sivakumar Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044

(continued )
Stations Years

Sharpeville 7.36 � 4.14 5.94 � 3.57 5.47 � 3.53 7.33 � 4.32 6.32 � 3.87 5.38 � 3.43 5.46 � 3.45
Three rivers 4.911 � 2.17 4.29 � 2.15 3.74 � 2.01 4.56 � 2.20 4.64 � 2.72 3.74 � 1.93 3.78 � 2.02
Zamdela 6.71 � 3.28 6.99 � 3.68 6.92 � 3.79 6.29 � 3.15 6.08 � 3.80 7.17 � 3.77 5.89 � 3.60
Vaal triangle 5.77 � 1.11 5.03 � 1.22 4.77 � 1.32 5.50 � 1.37 4.97 � 0.98 4.59 � 1.43 4.46 � 0.96
Johannesburg 2 group
2004 2005 2006 2007 2008 2009 2010
Alexandra 8.48 � 5.18 6.94 � 4.23 5.93 � 4.54 5.72 � 3.83 4.96 � 2.82 6.67 � 3.29 7.88 � 3.06
Buccleuch 12.15 � 3.75 11.06 � 4.28 9.54 � 4.58 10.11 � 4.87 8.62 � 4.0 6.62 � 3.64 7.9 � 3.85
Jabavu 5.98 � 3.23 8.43 � 3.86 7.41 � 4.17 6.94 � 3.81 6.30 � 3.57 5.75 � 3.56 5.85
Orange farm 6.08 � 3.60 4.46 � 3.01 3.14 � 2.50 4.71 � 3.19 3.02 � 2.11 3.35 � 2.46 2.79 � 2.31
Johannesburg 2 8.17 � 2.89 7.72 � 2.75 6.50 � 2.68 6.87 � 2.34 5.73 � 2.35 5.60 � 1.55 6.11 � 2.41
Mpumalanga power stations
2011 2012 2013
Camden 6.40 � 4.13 8.35 � 5.19 9.57 � 6.37
Elandsfontein 11.63 � 5.14 10.83 � 6.14
Grootdraaidam 11.12 � 5.28 12.65 � 6.10 16.07 � 11.47
Komati 10.77 � 5.85 11.81 � 6.10 8.69 � 5.30
Leandra 9.26 � 3.79 7.09 � 4.33 5.65 � 4.48
Majuba 1.94 � 1.56 3.26 � 2.90 2.97 � 2.66
Phola 12.36 � 5.61 11.08 � 4.81 10.90 � 4.87
Verkykkop 3.63 � 3.11 4.87 � 3.92 3.16 � 2.78
Mpumalanga power stations 8.39 � 3.93 8.74 � 3.43 8.14 � 4.66

Appendix C. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jastp.2019.05.008.

References Korhonen, K., Giannakaki, E., Mielonen, T., Pfüller, A., Laakso, L., Vakkari, V., Baars, H.,
Engelmann, R., Beukes, J.P., Van Zyl, P.G., Ramandh, A., 2014. Atmospheric
boundary layer top height in South Africa: measurements with lidar and radiosonde
Adesina, A.J., Kumar, K.R., Sivakumar, V., Griffith, D., 2014. Direct radiative forcing of
compared to three atmospheric models. Atmos. Chem. Phys. 14 (8), 4263–4278.
urban aerosols over Pretoria (25.75 S, 28.28 E) using AERONET Sunphotometer
Kumar, K.R., Kang, N., Sivakumar, V., Griffith, D., 2017. Temporal characteristics of
data: first scientific results and environmental impact. J. Environ. Sci. 26 (12),
columnar aerosol optical properties and radiative forcing (2011–2015) measured at
2459–2474.
AERONET’s Pretoria_CSIR_DPSS site in South Africa. Atmos. Environ. 165, 274–289.
Bayraktar, H., Turalioglu, F.S., 2005. A Kriging-based approach for locating a sampling
Liousse, C., Assamoi, E., Criqui, P., Granier, C., Rosset, R., 2014. Explosive growth in
site—in the assessment of air quality. Stoch. Environ. Res. Risk Assess. 19 (4),
African combustion emissions from 2005 to 2030. Environ. Res. Lett. 9 (3), 035003.
301–305.
Li, C., Joiner, J., Krotkov, N.A., Bhartia, P.K., 2013. A fast and sensitive new satellite SO2
Beukes, J.P., Vakkari, V., Van Zyl, P.G., Venter, A.D., Josipov, M., Jaars, K., Tiitta, P.,
retrieval algorithm based on principal component analysis: application to the ozone
Kulmala, M., Worsnop, D., Pienaar, J.J., Virkkula, A., Laakso, L., 2013. Source region
monitoring instrument. Geophys. Res. Lett. 40 (23), 6314–6318.
plume characterisation of the interior of South Africa as observed at Welgegund.
Lloyd, S.M., 1987. The control of air pollution at South African coal-fired power stations.
Clean Air Journal¼ Tydskrif vir Skoon Lug 23 (1), 7–10.
Clean Air J. 20–24. ISSN 0379-4709.
Burger, R.P., 2016. The Distribution of Aerosol and Trace Gases in the Lower
Lloyd, P.J., 2002. Coal Mining and the Environment. Energy Research Institute,
Troposphere over South Africa. Doctoral dissertation. Faculty of Science, University
University of Cape Town.
of the Witwatersrand, Johannesburg.
Lourens, A.S., Beukes, J.P., Van Zyl, P.G., Fourie, G.D., Burger, J.W., Pienaar, J.J.,
Clark, N.A., Demers, P.A., Karr, C.J., Koehoorn, M., Lencar, C., Tamburic, L., Brauer, M.,
Read, C.E., Jordaan, J.H., 2011. Spatial and temporal assessment of gaseous
2015. Effect of early life exposure to air pollution on development of childhood
pollutants in the Highveld of South Africa. South Afr. J. Sci. 107 (1–2), 1–8.
asthma. Environ. Health Perspect. 118 (2), 284.
Lourens, A.S., Butler, T.M., Beukes, J.P., Van Zyl, P.G., Beirle, S., Wagner, T.K., Heue, K.
Collett, K.S., Piketh, S.J., Ross, K.E., 2010. An assessment of the atmospheric nitrogen
P., Pienaar, J.J., Fourie, G.D., Lawrence, M.G., 2012. Re-evaluating the NO2 hotspot
budget on the South African Highveld. South Afr. J. Sci. 106 (5–6), 35–43.
over the South African Highveld. South Afr. J. Sci. 108 (11–12), 83–91.
Davis, C.L., 2011. Climate Risk and Vulnerability: A Handbook for Southern Africa.
Macpherson, A.J., Simon, H., Langdon, R., Misenheimer, D., 2017. A mixed integer
Council for Scientific and Industrial Research, Pretoria, South Africa, ISBN 978-0-
programming model for National Ambient Air Quality Standards (NAAQS)
620-50627-4, pp. 1–92.
attainment strategy analysis. Environ. Model. Softw 91, 13–27.
Department: Environmental Affairs (DEA), 2012. Republic of South Africa, 2012 South
Martins, J.J., Dhammapala, R.S., Lachmann, G., Galy-Lacaux, C., Pienaar, J.J., 2007.
Africa Environment Outlook, Chapter 5: Air Quality, Draft 2, Version 2. http://soer.
Long-term measurements of sulphur dioxide, nitrogen dioxide, ammonia, nitric acid
deat.gov.za/dm_documents/Chapter_5__Air_Quality_ 4kp8X.pdf. (Accessed 15
and ozone in southern Africa using passive samplers. South Afr. J. Sci. 103 (7–8),
February 2017).
336–342.
Dyson, L.L., Van Heerden, J., Sumner, P.D., 2015. A baseline climatology of sounding-
Menyah, K., Wolde-Rufael, Y., 2010. Energy consumption, pollutant emissions and
derived parameters associated with heavy rainfall over Gauteng, South Africa. Int. J.
economic growth in South Africa. Energy Econ. 32 (6), 1374–1382.
Climatol. 35 (1), 114–127.
Mugabo, C., 2011. Ambient Air Quality in a Low Income Urban Area on the South
Feig, G., Ncipha, X., Vertue, B., Naidoo, S., Mabaso, D., Ngcukana, N., Tshehla, C.,
African Highveld: A Case Study of Leandra Township. Doctoral dissertation. Faculty
Masuku, N., 2014. Analysis of a period of elevated ozone concentration reported
of Science, University of the Witwatersrand, Johannesburg.
over the Vaal Triangle on 2 June 2013. Clean Air Journal¼ Tydskrif vir Skoon Lug 24
2012-2013 National Air Quality Officer’s Annual Report on Air Quality Management,
(1), 10–16.
March 2014. Department: Environmental Affairs, Republic of South Africa.
Feng, X., Li, Q., Zhu, Y., Wang, J., Liang, H., Xu, R., 2014. Formation and dominant
Ncipha, X.G., 2011. Comparison of Air Pollution Hotspots in the Highveld Using
factors of haze pollution over Beijing and its peripheral areas in winter. Atmos.
Airborne Data. Masters of Science dissertation. University of the Witwatersrand,
Pollut. Res. 5 (3), 528–538.
Johannesburg.
Garstang, M., Tyson, P.D., Swap, R., Edwards, M., Kållberg, P., Lindesay, J.A., 1996.
Ozone Monitoring Instrument (OMI) Data User’s Guide, 2012. http://disc.sci.gsfc.nasa.
Horizontal and vertical transport of air over southern Africa. J. Geophys. Res. Atmos.
gov/Aura/dataholdings/additional/documentation/README.OMI_DUG.pdf.
101 (D19), 23721–23736.
(Accessed 19 August 2014).
Ghozikali, M.G., Mosaferi, M., Safari, G.H., Jaafari, J., 2015. Effect of exposure to O3,
Pretorius, I., Piketh, S., Burger, R., Neomagus, H., 2015. A perspective on South African
NO2, and SO2 on chronic obstructive pulmonary disease hospitalizations in Tabriz,
coal fired power station emissions. J. Energy South. Afr. 26 (3), 27–40.
Iran. Environ. Sci. Pollut. Res. Int. 22 (4), 2817–2823.
Qu, W.J., Arimoto, R., Zhang, X.Y., Zhao, C.H., Wang, Y.Q., Sheng, L.F., Fu, G., 2010.
Hamm, N.A.S., Finley, A.O., Schaap, M., Stein, A., 2015. A spatially varying coefficient
Spatial distribution and interannual variation of surface PM10 concentrations over
model for mapping PM10 air quality at the European scale. Atmos. Environ. 102,
eighty-six Chinese cities. Atmos. Chem. Phys. 10 (12), 5641–5662.
393–405.
Sangeetha, S.K., Sivakumar, V., Josipovic, M., Gebreslasie, M., Wright, C.Y., 2017. SO2
Kgabi, N.A., Sehloho, R.M., 2012. Seasonal variations of tropospheric ozone
seasonal variation and assessment of Ozone Monitoring Instrument (OMI)
concentrations. Glob. J. Sci. Front. Res. (GJSFR) 12 (5-B).

18
S.K. Sangeetha and V. Sivakumar
measurements at Sharpeville (27.86� E; 26.68� S) a South African ground-based
station. Int. J. Remote Sens. 38 (23), 6680–6696.
Sangeetha, S.K., Sivakumar, V., Gebreslasie, M., 2018. Long-range transport of SO2 over
South Africa: a case study of the Calbuco volcanic eruption in April 2015. Atmos.
Environ. 185, 78–90.
Smith, S.J., Aardenne, J.V., Klimont, Z., Andres, R.J., Volke, A., Delgado Arias, S., 2011.
Anthropogenic sulfur dioxide emissions: 1850–2005. Atmos. Chem. Phys. 11 (3),
1101–1116.
Spalding-Fecher, R., Matibe, D.K., 2003. Electricity and externalities in South Africa.
Energy Policy 31 (8), 721–734.
Sundstr€om, A.M., Nikandrova, A., Atlaskina, K., Nieminen, T., Vakkari, V., Laakso, L.,
Beukes, J.P., Arola, A., van Zyl, P.G., Josipovic, M., Venter, A.D., 2015.
Characterization of satellite-based proxies for estimating nucleation mode particles
over South Africa. Atmos. Chem. Phys. 15 (9), 4983–4996.
Thomas, F.W., Carpenter, S.B., Gartrell, F.E., 1963. Stacks—how high? J. Air Pollut.
Control Assoc. 13 (5), 198–204. https://doi.org/10.1080/
00022470.1963.10468165.
Thomas, R., Scorgie, Y., 2006. Air Quality Impact Assessment for the Proposed New Coal-
Fired Power Station (Kendal North) in the Witbank Area (Report No.: APP/06/NMS-
01 Rev 0.3). Airshed Planning Professionals (Pty) Ltd.
Tyson, P.D., Garstang, M., Swap, R., 1996. Large-scale recirculation of air over southern
Africa. J. Appl. Meteorol. Climatol. 35 (12), 2218–2236.
Tyson, P.D., Garstang, M., Swap, R., Kallberg, P., Edwards, M., 1996. An air transport
climatology for subtropical southern Africa. Int. J. Climatol. 16 (3), 265–291.
19
Journal of Atmospheric and Solar-Terrestrial Physics 191 (2019) 105044
Vakkari, V., Beukes, J.P., Laakso, H., Mabaso, D., Pienaar, J.J., Kulmala, M., Laakso, L.,
2013. Long-term observations of aerosol size distributions in semi-clean and polluted
savannah in South Africa. Atmos. Chem. Phys. 13 (4), 1751–1770.
Venter, A.D., Vakkari, V., Beukes, J.P., Van Zyl, P.G., Laakso, H., Mabaso, D., Tiitta, P.,
Josipovic, M., Kulmala, M., Pienaar, J.J., Laakso, L., 2012. An air quality assessment
in the industrialised western Bushveld Igneous Complex, South Africa. South Afr. J.
Sci. 108 (9–10), 1–10.
Wang, Y., Ying, Q., Hu, J., Zhang, H., 2014. Spatial and temporal variations of six criteria
air pollutants in 31 provincial capital cities in China during 2013–2014. Environ. Int.
73, 413–422.
Xia, X., Qi, Q., Liang, H., Zhang, A., Jiang, L., Ye, Y., Huang, Y., 2016. Pattern of spatial
distribution and temporal variation of atmospheric pollutants during 2013 in
Shenzhen, China. ISPRS Int. J. Geo-Inf. 6 (1), 2. https://doi.org/10.3390/
ijgi6010002.
Xue, D., Yin, J., 2014. Meteorological influence on predicting surface SO2 concentration
from satellite remote sensing in Shanghai, China. Environ. Monit. Assess. 186 (5),
2895–2906.
Zou, B., Wilson, J.G., Zhan, F.B., Zeng, Y., Wu, K., 2011. Spatial-temporal variations in
regional ambient sulfur dioxide concentration and source-contribution analysis: a
dispersion modeling approach. Atoms. Environ. 45 (28), 4977–4985.
Zunckel, M., Robertson, L., Tyson, P.D., Rodhe, H., 2000. Modelled transport and
deposition of sulphur over Southern Africa. Atoms. Environ. 34 (17), 2797–2808.