Assessment of changes in concentrations of selected criteria pollutants in the Vaal and Highveld priority areas

Introduction In the time since the promulgation of the National Environmental Management Air Quality Act (NEM:AQA) (DEA, 2004) considerable changes have occurred in the air quality management landscape in South Africa (Tshehla and Wright, 2019), these include: • The introduction of National Ambient Air Quality Standards (NAAQS) (DEA, 2009, 2012b), • The identification of activities resulting in atmospheric emissions from industry (Section 21 of the Act) (DEA, 2010) and the subsequent setting of atmospheric emission limits (DEA, 2013a, 2015a), including certain printing industry activities (DEA, 2016). • The promulgation of dust control regulations (DEA, 2013b), • The identification and development of emission standards for controlled emitters as described under section 23 of the NEMAQA including; small scale char and charcoal plants (DEA, 2015b), temporary asphalt plants (DEA, 2014b), small scale boilers (DEA, 2013c), • The declaration of greenhouse gases as priority air pollutants (DEA, 2014a)


Introduction
In the time since the promulgation of the National Environmental Management Air Quality Act (NEM:AQA)  considerable changes have occurred in the air quality management landscape in South Africa (Tshehla and Wright, 2019), these include: • The introduction of National Ambient Air Quality Standards (NAAQS) (DEA, 2009(DEA, , 2012b • The identification of activities resulting in atmospheric emissions from industry (Section 21 of the Act) (DEA, 2010) and the subsequent setting of atmospheric emission limits (DEA, 2013a(DEA, , 2015a, including certain printing industry activities (DEA, 2016). • The declaration of greenhouse gases as priority air pollutants (DEA, 2014a) While this legislation has been enacted, certain areas have been identified as being of particular concern and have been declared as air quality priority areas, these include: the Vaal Triangle Airshed Priority Area (VTAPA; DEAT, 2006), the Highveld Priority Area (HPA; DEAT, 2007) and the Waterberg-Bojanala Priority Area (DEA, 2012a). Within these areas, particular concern has been placed in the development and review of management plans, including the VTAPA in 2009, the HPA in 2012 and the Waterberg in 2015 (DEA, 2015c).
Beyond the scope of changing legislation, there have been changes in the emissions profiles within the priority areas that have had significant impacts on emission profiles in the priority areas (Pretorius et al., 2015). These include: • The increase in the number of vehicles in the country has contributed to changes in the emissions profiles, including an increase in vehicle emissions. Between January 2009 and June 2018, South Africa has seen an increase in motor vehicles, rising from ~9 to ~12 million. Notable increases have been seen in the Gauteng and Mpumalanga provinces (where VTAPA and HPA are located), with increases from ~3.5 to ~4.7 million in Gauteng, and ~600 000 to ~900 000 motor vehicles in Mpumalanga have been recorded (ENATIS, 2018).
• The closure of Highveld Steel in February 2015 (Goldswain, 2016), which resulted in a decrease in industrial emissions in the HPA.
• The large scale domestic electrification programmes in the priority areas which resulted in the rapid electrification, across South Africa more than 5 million households received access to electricity between 1990 and 2007 in South Africa (Bekker et al., 2008) which resulted in a reduction in household emissions but an increase in electricity consumption and associated emissions at the power generation plants.
Although changes in power generation including the recommissioning of Camden, Grootvlei and Komati power stations to counter the electricity shortages experienced from 2007 (ESKOM, 2011), and the introduction of the Kusile and Medupi power stations into the national grid have improved the output capacity, this only sustains the country's reliance on coal, further delaying emission reductions, especially with noncompliant ageing power plants (Pretorius et al., 2015).
Concurrent with the development of legislation and the rollout of air quality management plans, there has been the installation of air quality monitoring infrastructure. Starting in 2007, the Vaal Triangle Airshed Priority Area Ambient Air Quality Monitoring Network was established with monitoring sites located at identified air pollution hotspots, followed by the Highveld Priority Area ambient air quality monitoring network in 2008. These networks have approximately 10 years of data that are available through the South African Air Quality Information System (SAAQIS). These data can be used to identify trends in ambient concentrations in order to identify if the interventions made have been effective, and if there are changing priorities in which pollutants are of greatest concern.

Methods and Materials
Data for the 11 monitoring stations in the VTAPA and HPA priority area ambient air quality monitoring networks were requested from SAAQIS at an hourly temporal resolution ( Table 1). The data were provided in a quality controlled form, but a subsequent quality control of the data was conducted to remove a limited number of negative concentration values and the removal of a significant number of values below the detection limit of the instruments (typically zero values) and other anomalous measurements. The data were analysed using the R language for statistical computing (R Core Team, 2013), Research article: Changes in criteria pollutants in the Vaal and Highveld priority areas Page 2 of 13 specifically using the Open Air Package (Carslaw and Ropkins, 2012). The trends in the concentrations of PM 10 , PM 2.5 , and SO 2 were calculated for each station for the period of available data using the Theil-Sen trend analysis following a deseasonilisation step as recommended, which uses the Loess method. The Theil-Sen Estimator approach uses monthly averages, for which an 80% data availability threshold was set. A comparison of the trends in the continuously monitored SO 2 concentrations can be made with a dataset of measurements conducted by the CSIR extending from 1959-1968(Kemeny and Halliday, 1972Kemeny, 1980;Kemeny and Vleggaar, 1983;Walker, Ellerbeck and Kemeny, 1986;Walker, Galpin and Pienaar, 1987). The historical CSIR measurements were made using the Hydrogen Peroxide Method, which consisted of passing the air volume through a dilute solution of hydrogen peroxide and measuring the change in the pH of the solution through titration with a sodium borate solution (Kemeny, 1980). Comparison with these historical results serves as a valuable benchmark as to how the ambient concentrations of SO 2 have changed in the last 60 years.

PM 10
Using a 75% data capture threshold, it is clear that all the monitoring stations in the VTAPA and HPA have been out of compliance with the historical (red line) and/or current (orange line) annual NAAQS (Figure 1). This corresponds to the results presented by the National Air Quality Officer in the annual State of the Air Report (individual values presented may differ slightly based on the data completeness requirements or the data cleaning protocol that have been used). As of 2016 (the last year with full data used in this assessment), it was only the Hendrina and Middelburg sites that complied with the annual standard, and both of these sites are located in middle income communities. The Zamdela, Witbank, Three Rivers and Sharpeville sites were non-compliant with the historical standard. From this it is clear that significant problems related to the concentrations of PM 10 occur in the majority of the monitoring sites in the VTAPA and HPA.
The monthly PM 10 concentrations have shown a general decrease at all the sites in the VTAPA and HPA (Figure 2)   Using the trend in the change in PM 10 concentrations over the measurement period, a simplistic assessment was made of how long it would take for each of the monitoring stations to reach compliance with the national PM 10 standard based on the annual average concentration for the last full year with data (Table 2), it is acknowledged that this is an overly simplistic approach as it is not expected that the ambient concentrations of PM 10 will follow a continuous linear trend. However, this approach is illustrative as to how long it may take to get to compliance following the long term current trend. Currently three stations are in compliance with the PM 10 annual standard for the remainder it is (simplistically) expected that compliance will be reached between 2018 for Secunda and 2065 for Sharpeville. In the Sharpeville case, the annual concentration for 2016 (57.2 µg/m³) was considerably higher than for 2015 (22.78 µg/m³) in this case if the 2015 annual value is used with a continuation of the historical trend, compliance is expected by 2035.
PM 2.5 Using the 75% data capture threshold, it is clear that most of the monitoring sites are in exceedance of the current annual standard for PM 2.5 , it is only Hendrina and Middelburg that are below the current annual NAAQS (Figure 3). From this it is clear that there are still significant problems relating to the concentrations of PM 2.5 over the VTAPA and HPA.
Monthly PM 2.5 concentrations show a decreasing trend for all the sites, except for Zamdela (a non-significant increase in PM 2.5 ) ( Figure 4). A decreasing concentration at a confidence interval p≤0.001 was found for: Diepkloof (-2.33 µg/m³/year), Middelburg      Using a 75% data capture requirement, it is only the Witbank site that exceeds the annual SO 2 standard within the priority areas ( Figure 5). SO 2 has not proven to be out of compliance with the NAAQS.
In comparison to the PM measurement where there was a significant decreasing trend in almost all the monitoring sites, the trends in SO 2 concentrations were only significant at five of the 11 sites ( Figure 6). A significant decrease in SO 2 concentrations at the p<0.001 level was found at Hendrina (-0.73 ppb/year), Ermelo (-0.52 ppb/year), Middelburg (-0.52 ppb/ year), and Diepkloof (-0.3 ppb/year). At the Three Rivers site an increasing trend of 0.08 ppb/year in the SO 2 concentration was observed at the p<0.05 confidence level. No significant trends were observed at the other sites (Table 4).

Historical SO 2 concentrations
Long term historical data are not available for many of the sites under consideration in this study, however the measured SO 2 concentrations reported for the major metros in South Africa during the 1960s show that historically the ambient concentrations of SO 2 were considerably higher than is currently recorded. Within Johannesburg, Durban and East London annual average SO 2 concentrations above 20ppb were common (Figure 7).

Discussion
Considerable efforts have been expended over the last decade and a half in the development of the legislative framework to govern air quality management in South Africa, however it is frequently stated that the strategic objectives have not been met (Tshehla and Wright, 2019). It was reported that with 70 % of the planned interventions from the Vaal Triangle Priority Area Air Quality Management Plan having been implemented, there was no proportional improvement in air quality (Senene, 2018). However, these reports are often based on a fairly cursory analysis of whether compliance criteria have been met. The long term dataset of reasonably comprehensive and good quality data that is growing in the VTAPA and HPA provides a good opportunity to assess the long term trends in pollution in these areas.
As stated by the National Air Quality Officer, in previous State of Air Reports, the compliance with the PM 10 and PM 2.5 standards is a significant problem over the Vaal and Highveld regions with almost all of the monitoring sites being non-compliant with the current NAAQS this is well known and has been previously extensively documented (Venter et al., 2012;Hersey et al., 2015;Wernecke et al., 2015;Feig, et al, 2016;Garland et al., 2017). However, annual SO 2 is generally (with the exception of Witbank) considerably lower than the NAAQS. These results are similar to those reported by Feig et al, (2016) for the Waterberg priority area and Venter et al. (2012), for the Western Bushveld Igneous complex, which identified the major sources of SO 2 as high stack emissions and PM 10 from domestic combustion emissions. Across the Highveld annual average concentrations of SO 2 were reported as being below the NAAQS (Lourens et al., Figure 6: Thiel-Sen trend in SO 2 concentration (ppb) over the VTAPA and HPA *** indicates significance at the p<0.001 confidence level, ** indicates significance at the p<0.01 confidence level and * indicates significance at the p<0.05 confidence level 2011). At the KwaDela site (located in the Highveld between Ermelo and Secunda), health risks associated with both indoor and outdoor exposure to particulate matter was identified as a risk (Wernecke et al., 2015).
The trend in the concentrations of particulate matter is largely negative, with the exception of Three Rivers for PM 10 and Zamdela for PM 2.5 (which did not show any statistical significance).
Previous studies into the long term temporal and spatial analysed of pollutants in the priority areas has been done previously including the work done by Sangeetha and Sivakumar, (2019), where monitoring stations were grouped and averaged according to broad spatial location. In this study the seasonal variability of measured SO 2 was discussed, but the long term trend was not analysed. Further comparison between ground based measurements of SO 2 and satellite based estimates were performed for the Sharpeville site in the VTAPA (Sangeetha et al., 2017). In a remote sensing based study that looked at SO 2 concentrations over the HPA and increasing trend in SO 2 emissions was reported (Shikwambana and Tsoeleng, 2019) it is however unclear how an emissions value was obtained from the ambient data that was used in the study. In addition to these recent studies the long term trends in SO 2 obtained from the historical CSIR studies in these it showed a reduction in the SO 2 concentrations between the 1960s and 1970s and a plateau in concentration in the 1980s (Kemeny, 1980;Kemeny and Vleggaar, 1983).
VTAPA and HPA indicates that improvements in air quality are occurring; however, there is still considerable variability in the rate of improvements between sites. For PM 10 , the greatest rate of improvement in ambient concentration is seen for the Secunda and Zamdela sites at -4.35 and -4.68 µg/m³/year respectively, while at Kliprivier and Witbank, no significant changes were observed; and a significant increasing trend was observed at Three Rivers. Based on the initial ambient concentrations and the rate of decrease in the PM 10 concentration, the expected time until compliance with the annual NAAQS ranges from less than three years for Sebokeng to 49 years for Sharpeville, this is dependent on the assumption that the current linear trend is to continue. It is acknowledged that this assumption is a simplification, and thus is used for illustrative purposes in this context. Similarly, for PM 2.5 , the rate of decrease in ambient concentrations ranges from -2.3 for Diepkloof to -0.3 µg/m³/ year for Three Rivers. Compliance with the annual NAAQS is expected to take between 1 year and 39 years for Ermelo and Sharpeville, respectively. By the time the 2030 NAAQS for PM 2.5 comes into effect, it is expected that only Kliprivier, Sharpeville and Three Rivers will still be out of compliance, based on the current trends.
For both PM 2.5 and PM 10 , the concentrations at Sharpeville showed a strong increase in 2016. It is not known whether this is the result of a short term temporary localised event or if it is indicative of a more significant local change in the emissions profile at the site.
As recently demonstrated through an assessment and cost performed for the entire country using the BENMAP model, significant impacts on health and associated economic costs for South Africa from not meeting the NAAQS for PM 2.5 (Altieri and Keen, 2019). Changes in the ambient concentration of PM 10 and PM 2.5 are expected to have an impact on the ambient concentrations of co-emitted pollutants such as black carbon (Feig et al., 2015;Kuik et al., 2015).
In contrast to the decreasing trend in particulate matter, there is little to no trend in the concentrations of SO 2 over the VTAPA and HPA monitoring stations, where a significant increasing trend was found in five of the 11 sites. Four of the sites showed a negative trend in SO 2 (Diepkloof, Ermelo, Hendrina and Middelburg) which ranged between -7.3 ppb/year for Hendrina to -0.3ppb/year for Diepkloof. Supporting the previous findings, the ambient concentrations of SO 2 at the annual averaging period are in compliance with the NAAQS (Lourens et al., 2011;Venter et al., 2012;Wernecke et al., 2015).
Long term historical measurements at some of the major South African cities in 1960s to 1980s show a decreasing trend in SO 2 concentration during the 1960s and 1970s however it levelled out during the 1980s at the end of the measurement time series (Kemeny and Halliday, 1972;Kemeny, 1980;Kemeny and Vleggaar, 1983). During the time period between the observations reported in the Kemeny papers and the beginning of the observations in the VTAPA and HPA reported here there has been a considerable decrease in the ambient concentrations, once again highlighting the importance of continuous observation record.
With the availability of a long term data set of ambient air quality concentrations, it is valuable to assess the state of air quality not just in terms of the compliance in terms of the annual standards, but also to examine the trend in concentration to determine how quickly sites are moving towards compliance and to focus attention on the locations where both the ambient concentrations are out of compliance and where little progress is being made towards meeting the NAAQS.

Conclusion
Despite the existence of the current air quality management regime for a period of 15 years (since the promulgation of the NEM:AQA in 2004) the air quality in the VTAPA and HPA is still considered to be poor and these areas are out of compliance with the PM 10 and PM 2.5 NAAQS. However, in most instances over the monitoring time period, the ambient concentrations of particulate matter are improving and in some cases are improving fairly rapidly. These trends are not as evident for SO 2 concentrations, however, in contrast to PM, there is only one station where ambient SO 2 concentrations exceeded annual NAAQS.
This study is intended to provide a simple approach to identify where and at what rate the ambient air quality is improving, or to identify locations where improvements are not being observed. This serves as a guidance for air quality managers to consider and focus their management interventions. This analysis can be updated annually to continue to quantify trends, as long as data quality and data capture rates are high.

Author contributions
Gregor Feig conceptualised the study and wrote most of the text. Rebecca Garland contributed to the conceptualisation of the study and the review and drafting of the text. Seneca Naidoo contributed to the data analysis. Amukelani Maluleke contributed to the capture and analysis of the historical data. Marna Van der Merwe provided guidance on the Thiel-Sen analysis and provided valuable input in the review and editing of the paper.