The effect of coal-fired power plants on ambient air quality in Mpumalanga province, South Africa, 2014–2018

ABSTRACT

Several coal-fired power plants (CFPPs) were built in South Africa, mainly in the central Mpumalanga Province, due to an increase in the demand for Eskom, the national power utility, to keep up with socio-economic growth. The CFPPs, of which 90% are owned by Eskom, generate a significant share of the country’s electricity but contribute to the air pollution experienced in the country. The paper discusses sulphur dioxide (SO2), nitrogen dioxide (NO2) and particulate matter of size less than 10 micrometre (μm) in diameter (PM10), using data from 2014 to 2018. The statistics revealed higher PM10 concentrations during winter than in summer and spring at the Kriel and Komati sites; associated with the higher contribution of domestic burning. The study’s results could influence legislation and policies and help to understand the source of poor ambient air quality by assessing the three pollutants within the area of the selected power plants.

KEYWORDS:

• Air pollution
• sulphur dioxide
• particulate matter
• komati
• kriel

Introduction

Air quality deterioration has become an increasing global concern due to its impacts on public health and the environment from the Industrial Revolution in the 18th century to modern times (Olufemi et al. 2018; Xiong and Xu 2021; Zhang et al. 2022). Today, industries require considerably more energy compared to the 1800s. This compounded with industry’s heavy reliance on fossil fuels, which significantly contributed to poor ambient air quality (Naidoo et al. 2014; Moletsane et al. 2021; Walton et al. 2021). Several coal-fired power plants (CFPPs) were built in South Africa (SA), mainly in the central Mpumalanga Province, due to an increase in the demand for the national power utility, Eskom, to keep up with socio-economic growth in the country (Rorich and Galpin 1998). These CFPPs are mostly (90%) owned and operated by the power utility, and 80% are located within Mpumalanga.

The National Environmental Management: Air Quality Act (NEMAQA) (No. 39 of 2004) air quality regulations provide the minimum emission standards (MES) for CFPPs, regulating the pollution that may be emitted from point sources, such as stacks. Compliance with existing and new emission limits was specified in 2015 and 2020, respectively, depending on the date of construction (DoE Department of Energy 2019). The emission reduction initiatives in SA are also in line with the 2015 adoption of the United Nations’ 2030 Agenda for Sustainable Development Goals (SDGs), which the country has signed as outlined in the study by Udeagha and Muchapondwa (2023). To reduce the impact of air pollution due to CFPPs, the power utility has developed an Emission Reduction Plan, focusing on reducing SO₂, among other pollutants (Eskom 2021). Particulate emission reduction initiatives have been implemented at some of the CFPPs to reduce emissions at the existing coal-fired fleet (as seen in Figure 1–Matimolane et al. 2021), such as the retrofit of electrostatic precipitators (ESPs), flue gas desulphurisation (FGD) and fabric filter plants (FFPs). Currently, the FGD is used only at the Kusile Power Plant to reduce SO₂ emissions, although it is currently not fully operational due to infrastructure breakdowns. The retrofitting with FGD at Medupi Power Plant will be implemented in 2025.

Figure 1. Timescale in relative emissions of particulates (ash) from Eskom’s power plants from the early 1980s to present (Matimolane et al. 2021).

The power utility, in compliance with its atmospheric emission license (AEL), has strategically installed ambient air quality stations to attain continuous air pollution parameter measurements to represent the ambient air of the areas potentially impacted by the power plant emissions (Eskom 2013). The measuring stations are distributed and located in residential areas to measure and understand the impact of air pollution on health (Stolz et al. 2020) and the environment in general. The burning of solid fuels is the primary source of domestic energy for people living in low-income settlements in developing countries, which causes gaseous and particulate emissions (Xie et al. 2015).

Appreciable reductions in air pollutant concentrations and emission factors from CFPPs in China, India, Japan, Europe and the USA were observed over the past two decades (Wang et al. 2020). The combination of control measures such as the application of removal technologies, stringent emission standards, equipment improvement and the new perspective of policymaking as well as effectiveness of government policies to reduce emissions (Filonchyk and Peterson 2023a; Wan et al. 2023). However, from the listed control measures, technological upgrades are the most essential for addressing further emission control issues to improve air quality (Weng et al. 2023). The current study is of significance since there is continued pressure to move away from fossil fuel power generation. Furthermore, the assessment and understanding of the gaseous and particulate emission impacts from CFPPs relates to how the changes in air quality are managed.

The aim of the current study was to analyse the air quality data from 2014 to 2018 obtained from the power utility to determine ambient air quality trends in SO₂, NO₂ and PM₁₀ from each community neighbouring the selected CFPPs. The present study, therefore, selected one of the previously mothballed CFPPs, namely the Komati Power Plant, which was returned to service in 2013 and one large 3000 MW Kriel Power Plant. The data from ambient monitoring stations near these two CFPPs were used to assess compliance of ambient pollution concentrations with the National Ambient Air Quality Standards (NAAQS), to determine ambient air quality trends of these pollutants and, lastly, to determine the correlation between the three selected criteria pollutants.

Materials and methods

Study area

Two ambient air quality monitoring sites were selected within a monitoring network of 600 km² as observed in Figure 2 and were considered to be representative of the study area. These monitoring sites are located at the Kriel and the Komati villages corresponding to the two CFPPs selected for the current study. The Kriel village site commenced in June 2006 and the monitoring site, as observed in Figure 2, is located ±7.8 km east of this power plant. The Komati site (Figure 2) is situated 2.2 km west-south-west of the Komati Power Plant and monitoring commenced on 1 June 2006 (Eskom 2013). The sites represented a range of highly industrialised areas such as Secunda, where the Kriel village monitoring site is located, to the more rural area of Komati, where the Komati site is located. The air pollution sources in the area include a nearby coal mining and handling company, along with haul truck exhaust emissions; domestic fuel use; agriculture and biomass burning as well as small industries. Both sites are affected by CFPPs and, further afield, a large petrochemical refinery.

Figure 2. Map of the South African Highveld including major towns, power stations and monitoring locations (Lourens et al. 2011)

Data set description

The Kriel village and the Komati sites are equipped to continuously monitor ambient concentrations of SO₂, NO₂, NOx, and coarse particulate matter of particulate size less than 10 μm in diameter (PM₁₀). The standard specifications, as well as the equipment/techniques used for the measurement of SO₂ and NOx, conform to the US-EPA equivalent method numbers EQSA-0486-060 and RFNA-1289-074, respectively (Komane 2019). The meteorological parameters of wind velocity, wind direction, ambient pressure, ambient temperature, humidity, rainfall and solar radiation are also recorded but not analysed in this discussion.

Data analysis

Various descriptive statistics were calculated from hourly (where possible), daily, weekly, monthly and annual measures of ambient SO₂, NO₂ and PM₁₀. The reason for focusing on these three pollutants is because of their cosmopolitan nature and the large quantities released from the coal combustion process for the generation of electricity, as well as for industrial and domestic use. The power utility’s in-house personnel from Research, Testing and Development (RT&D) provided the SO₂, NO₂ and PM₁₀ data used in the current study. Statistical analyses were performed using R (version 4.1.0), and the R Openair package version 2.11 (https://bookdown.org/david_carslaw/openair/). The R statistical package allows for easy importing of the data and descriptive analysis of the data (R Core Team 2021). Note that the data integrity was accepted since it was assumed that it was already completed by the RT&D team. Outliers were not removed, but those were incomplete data for the pollutants at various timestamps that were removed in order to complete the statistical analyses. Incomplete data were due to plant unavailability reported as the percentage of time that electrical power was unavailable to the monitoring station. In addition, the analysers installed during mid-month sometimes became faulty and were then removed from the site for maintenance. Considering hourly data, the percentages of missing values were about 49% for the Kriel village site (from a total of 37,944 observations), and 43% for the Komati site (from a total of 29,928 observations). The measured data from 2014 to 2018 were used to assess compliance of ambient pollution concentrations with the NAAQS to determine ambient air quality trends and to assess correlations between the three ambient pollutants. In this paper, the Theil-Sen analysis was used to estimate the trends of the pollutants for the given period as it tends to yield accurate confidence intervals and is robust against outliers as it is based on the median of the slopes (Carslaw and Ropkins 2012; Munir et al. 2013). Note that this study used the 95th percentile as an index that captures the range in the distribution of daily observed criteria pollutants selected during the study period.

Results and discussion

The results are presented for the three criteria pollutants of greatest concern in SA, namely SO₂, NO₂ and PM₁₀, concerning the environment in general. The raw data for annual variations of ambient air pollutants were visualised with time-series plots for each station for the full five-year period for the Kriel village and the Komati sites, as shown in supplemental material, i.e. Figures A1 and A2 for PM₁₀, NO₂ and SO₂. Note that in supplemental Figures A1 and A2, the pollutant is specified on each y-axis, and the y-axis range is in the units of the indicated pollutant (μg.m⁻³), with its minimum and maximum values specified.

Comparison with ambient air quality standards

National Ambient Air Quality Standards (NAAQS) have been established for criteria pollutants and are a combination of a limit for an average period and an allowable frequency of exceedance (supplemental material: Table A1). The study used verified daily concentrations over annual periods for each pollutant and at each site over the investigated period, and the NAAQS limit exceedances are highlighted in bold, as observed in supplemental Table A2. The PM₁₀ levels exceeded the NAAQS 24-hour standard of 75 μg.m⁻³ at both the Kriel village and the Komati sites in all the investigated years when considering the daily 95th percentile pollutant concentrations over annual periods. The significance of mentioning the PM₁₀ exceedance at both selected sites is mainly due to the domestic burning that increases in winter and therefore, the PM₁₀ exceedance is not only due to the power station emissions (Lourens et al. 2011, Vente and Lourens 2021). In addition to the domestic burning during colder months are the veld fires that happen due to the dry conditions within the environment (Venter et al. 2012, Vente and Lourens 2021). Note the NAAQS for PM₁₀’s one-year limit of 40 μg.m⁻³ for the 2014 data compared to 50 μg.m⁻³ for the 2015 to 2018 data, observed in supplemental material in Table A1.

Sulphur dioxide

The daily mean SO₂ concentrations over annual periods at the two selected monitoring sites are presented in supplemental Table A2. The ambient SO₂ concentrations for 2014 to 2018 were low at both sites with daily mean values in the range of 11.0 to 15.4 μg.m⁻³ at the Kriel village site and 11.4 to 21.0 μg.m⁻³ at the Komati site, during the investigated period. The daily mean SO₂ concentrations over annual periods at all the sites did not exceed the SO₂ 24-hour NAAQS limit of 125 μg.m⁻³ as observed in supplemental Table A2. For SO₂, the maximum concentration range was 79.1 μg.m⁻³ at the Kriel village site during the investigated period. There were zero (0) exceedances of the SO₂ daily limit at both the Kriel village and the Komati sites. Levels of SO₂ in urban areas have been linked to coal combustion in industries and the burning of coal is the main contributor of SO₂ into the atmosphere (Xu et al. 2017; Morakinyo et al. 2020).

Nitrogen dioxide

The daily mean NO₂ concentrations over annual periods at the monitoring sites are presented in supplemental Table 2. The ambient NO₂ concentrations over the period were low with daily mean values in the range of 7.2 to 12.9 μg.m⁻³ at the Kriel village site and 8.0 to 12.6 μg.m⁻³ at the Komati site. The maximum daily mean NO₂ concentration was 68.7 μg.m⁻³ at the Kriel village site during the investigated period. There are no 24-hour NAAQS limits for NO₂ according to the South African NAAQS established for criteria pollutants (Government Gazette Republic of South Africa 2009) as shown in supplemental Table A1.

Particulate matter less than 10 micrometres in diameter

The daily mean PM₁₀ concentrations over annual periods at the monitoring sites are presented in supplemental Table A2. The ambient PM₁₀ concentrations over the period were moderate to high with mean values in the range of 36.8 to 60.9 μg.m⁻³ at the Kriel village site and 53.1 to 70.8 μg.m⁻³ at the Komati site. The maximum mean PM₁₀ concentrations over the annual period were 248.1 μg.m⁻³ at the Komati site during the investigated period. At the Kriel village site, there were 236 exceedances of the PM₁₀ daily limit, while at the Komati site, there were 423 exceedances during the entire monitoring period. The significance of mentioning these PM₁₀ exceedances besides domestic burning and veld fires is that PM₁₀ is especially problematic during dry season since rain not only scrub the atmosphere but also assists in suppression; hence, an increase in PM₁₀ is typically observed between May and September annually (Lourens et al. 2011; Venter 2021).

Monthly, seasonal, weekly and daily variations of ambient air pollutants

The daily 95th percentile PM₁₀ concentration monthly average observed in 2017 at the Kriel village site of 103.6 mg.cm⁻³ was among the five out of 10 lowest concentrations at both sites over the years, as observed in the box-and-whisker plot in Figure 3 (a). In contrast, the Komati site had the highest daily 95th percentile PM₁₀ concentrations per month of 144.0 mg.cm⁻³ recorded in August 2017 (winter), as observed in the box-and-whisker plot in Figure 3 (b).

The Theil-Sen graphs of weekly mean PM₁₀ concentrations have demonstrated decreasing trends at the Kriel village site of −0.67 mg.cm⁻³.y⁻¹, with a 95% confidence interval of −4.72,3.73 mg.cm⁻³.y⁻¹, as observed in Figure 4 (a). Note that if the 95% confidence interval does not contain zero then the estimated slope is statistically significant at the 0.05 level, i.e. significantly different from zero with 95% confidence. The daily mean PM₁₀ concentrations at the Kriel village site have an estimated slope of −0.36 (−5,4.12) mg.cm⁻³.y⁻¹, as observed in Figure 4 (a). However, the Theil-Sen analysis indicated increasing trends in weekly mean PM₁₀ concentrations at the Komati site of 3.1 (−1.92,6.47) mg.cm⁻³.y⁻¹ and daily mean PM₁₀ concentrations of 3.17 (−2.34,5.79) mg.cm⁻³.y⁻¹, as observed in Figure 4 (b). Despite the increase in PM₁₀ emissions from the CFPPs, there were no temporal increasing trends in PM₁₀ concentrations at the Kriel village site. Similarly, the daily SO₂ mean and weekly concentrations at the Kriel village site shown by the Theil-Sen plots in Figure 5(a) compared to the SO₂ at the Komati (Figure 5b) site do not reflect, among other things, the sulphur contents of coal that have increased due to the poor coal quality procured by the power utility through its Primary Energy department (Eskom 2021).

Figure 3. The box-and-whisker plot of PM10 at the Kriel village (a) and Komati (b) sites showing the year and month with lowest and highest pollutant values.

Figure 4. The Theil-Sen plots of PM₁₀ at the Kriel village (a) and Komati (b) sites showing trends in the daily and weekly mean pollutant concentrations.

The Theil-Sen plot showed that the weekly and daily mean NO₂ concentrations have been decreasing at both the Kriel village and the Komati sites over the five-year period of assessment, as observed in Figure 6 (a) and (b). This could be attributed to very small changes in gaseous and particulate emissions or as a consequence of weather conditions that were not considered in the current study. Although the meteorological data was recorded but not discussed in the current article, TROPOspheric Monitoring Instrument (TROPOMI) has the capability to map a multitude of trace gases such as NO₂, ozone, formaldehyde, SO₂, methane, carbon monoxide and aerosols (Filonchyk and Peterson 2023a, 2023b). The latter would have affected PM₁₀, and therefore, it is important to be aware of this aspect for all three pollutants considered in the current study. Moreover, in Kumar et al. (2021) NOx concentration was high from vehicular sources and PM₁₀ concentration was high from industrial sources at ambient concentration when using meteorological data. In the current study, the higher concentrations of PM₁₀ primarily occurred in winter (especially August), with domestic burning a major contributor.

Figure 5. The Theil-Sen plots of SO₂ at the Kriel village (a) and Komati (b) sites showing trends in the daily and weekly mean pollutant concentrations.

Figure 6. The Theil-Sen plots of NO₂ at the Kriel village (a) and Komati (b) sites showing trends in the daily and weekly mean pollutant concentrations.

Interestingly, in winter and spring seasons the surface CO₂ footprints are spatially expanded as a result of contributions of emissions from biomass and domestic fossil fuel combustion (Ncipha and Sivakumar 2022). Levels of NO₂ and SO₂ in urban areas have been linked to coal combustion in industries and biomass burning in residential areas (Morakinyo et al. 2020). In the current study, the NO₂ concentrations peaked in spring and winter (supplemental material A1) with no significant trends otherwise. The increase in the number of vehicles in the country has contributed to changes in the emissions profiles, including an increase in vehicle emissions (Pretorius et al. 2015). The NO₂ concentrations during the investigated period were variable from month to month, with lower concentrations during January, April and October as shown in supplemental material A1a (Kriel) and b (Komati). The average concentrations of NO₂, SO₂ and PM₁₀ were in the order of winter > autumn > spring > summer. These seasonal variations reflect the effects of meteorological conditions occasioned by dawdling winds and a greater percentage of wind calm hours; lower mixing of boundary heights, limiting the diffusion capacity of the atmosphere and thus confining the emitted pollutants close to the ground, less rainfall, and lower temperatures (Xu et al. 2017; Morakinyo et al. 2020).

Correlation analysis between ambient air pollutants

The R Openair package version 2.11 (R Core Team 2021) was used to perform a correlation analysis between measured pollutants to understand the linear relationship among air pollutants as observed in supplemental Figures B1 to B10. Moderate linear correlations in data indicate that there are potential linear relationships between the three pollutants, but further statistical analyses are required to quantify the precise relationships between the pollutants. A correlation percentage of 20% to 40% in absolute value is considered to be a weak correlation. Correlations between 40% and 60% in absolute value are considered moderate, while those below 20% in absolute value are considered very weak. Correlation percentages whose magnitudes are between 60% and 80% in absolute value indicate variables that can be considered strongly correlated. The correlation percentage of 80% and above in absolute value is considered very strong.

The linear correlations between PM₁₀, NO₂, and SO₂ were mostly moderate to high at all sites for all the years considered, as observed in supplemental Figures B1 to B10. Possible reasons for moderate-to-high linear correlations observed in this study’s pollutant data might be from coal combustion as a source of SO₂, NO₂ and PM_10. Agriculture and wind-blown dust, on the other hand, consist of only particulate matter. Additionally, the lack of flue gas desulphurisation (FGD) in all the selected stations near the monitoring sites might be a result of the atmospheric chemistry driving the conversion of SO₂ to secondary particulates (Eskom 2021).

Results showed that the concentrations of SO₂ and NO₂ changed with seasons, and they manifested the same trend, i.e. correlations in winter are greater than spring, which are greater than in autumn, which are greater than summer. In contrast, at the Komati site, in 2015 and 2016, there were very strong positive correlations between NO₂ and SO₂ in summer of 81% and 90%, respectively, as observed in supplemental Figures B1 to B10. Noticeably, a very strong linear correlation of 82% between NO₂ and PM₁₀ in September was also observed at the Komati site in 2017. However, there was a very weak negative correlation of −2% between SO₂ and PM₁₀ at the Komati site in the spring of 2016. In summary, from the current study, the main seasons of concern to control SO₂ and NO₂ were identified as winter and spring.

Conclusion

In the current study, the three criteria pollutants, namely SO₂, NO₂ and PM₁₀ were analysed from the hourly (where possible), daily, weekly, monthly and annually monitored data obtained for the period 2014 to 2018. The results showed that only PM₁₀ measured regularly exceeded the annual NAAQS at the Kriel village and the Komati sites when both daily mean and 95th percentile PM₁₀ concentrations over annual periods were considered. The Theil-Sen and box-and-whisker plots revealed higher PM₁₀ concentrations and higher exceedances during winter (especially August), and lower concentrations during summer in all the selected years, with the highest concentrations reached in 2017 at the Komati site. However, this trend could not be attributed to the operational activities at both power plants alone, as there are other CFPPs, various mines, larger industries, and other agricultural activities near these monitoring sites. Moreover, the linear correlations between PM₁₀, NO₂, and SO₂ were moderate to high in all the investigated years and selected sites. Understanding such interactions is essential for urban climate studies and our study provides a basis for legislation and policies to estimate possible changes in ambient air quality for these two CFPPs to fully comply with the NEMAQA Act.

The decrease in the emissions of the three pollutants in 2018 compared to previous years, could be explained by the retrofitting of plants with pollution abatement technologies in meeting the 2015 MES target deadline (Pretorius et al. 2015). The ambient monitoring stations are helpful in showing the impacts of several other activities directly and indirectly associated with the operation of CFPPs. These should ideally be supplemented with either dispersion plume (Morakinyo et al. 2020) or Bayesian model approaches geared at assessing the impact of the CFPP operations using historical and current data (Ngamlana et al. 2024). The daily PM₁₀ over annual period concentrations exceedances have shown that the study’s results can influence legislation and policies to estimate possible changes in particulate emissions.

Acknowledgements

This is contribution number 883 of the North-West University (NWU) Water Research Group.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data will be made available on request.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.