Evaluation of the impact of coal mining on surface water in the Boesmanspruit, Mpumalanga, South Africa

The Boesmanspruit falls under the Inkomati–Usuthu Water Management Area in the Mpumalanga province. The part of the stream under consideration lies within the X11B quaternary catchment in the upper Komati catchment as indicated in Fig. 1. The stream feeds two important dams, Boesmanspruit and Nooitgedacht dams, before it drains into the Komati River, a transboundary river between South Africa, the Kingdom of eSwatini, and Mozambique (McCarthy and Humphries 2013). The geographical coordinates of the study area encompass the latitudes 26°04′09.80″ S and longitudes 30°10′31.53″ E.

The study area is dominated by sandy and loamy soil types and the seasons alternate between dry and moist conditions (Golder Associates 2014). The town is characterized by dry, mild-to-warm winters and humid, warm summers (McCarthy and Humphries 2013). The average annual precipitation falls between 700 and 800 mm, and the average yearly evaporation falls between 1650 and 1900 mm (Keighley 2017). Temperatures can vary between 32.5 °C (maximum) and 7 °C (minimum) in summer and 21.9 °C (maximum) to − 6 °C (minimum) in winter (McCarthy and Humphries 2013). The geology of the study area is found in the Ermelo Coal Field. Underlying the coalfield are glacial pre-Karoo rock layers, including the Dwyka tillite (Keighley 2017). The coal deposits in the Carolina region are found within the Vryheid Formation, a constituent of the Ecca Group that is a component of the Karoo Supergroup (Keighley 2017).

The aquatic ecology is protected in the catchment by the gazetted Resource Quality Objectives (RQOs) that set limits on water quality and quantity needed for the aquatic life to survive (RSA DWS 2016). The Boesmanspruit is one of the tributaries of the Komati River, an internationally shared water resource that flows into eSwatini and Mozambique and is managed by the Inkomati–Usuthu Catchment Management Agency. Therefore, international treaties and committees have been established to manage this river. Limits have been set on what South Africa can use and the amount of water needed to be released to the two countries (Pollard et al. 2014).

The town’s population is approximately 23,000 (McCarthy and Humphries 2013). The study area consists of a variety of land use activities such as coal mining, residential, cattle grazing and dry crop farming that have the potential to impact the water quality of rivers (McCarthy and Humphries 2013). Due to Carolina’s long history as an agricultural town, the X11B catchment contains a large amount of agricultural land, and 40% of the households depend on agriculture as their primary source of income (Keighley 2017).

The primary objective and activities in the catchment informed the site selection for this study. The sampling points were chosen based on the available historical data and where there are sources of pollution. Location factors such as ease of access and security were also considered. Six monitoring points were chosen to evaluate the status of Boesmanspruit based on the available data. Location with respect to upstream impact and downstream users determines optimal monitoring points (SA DWAF 1996a, b). Table 1 provides more detail of the selected monitoring points and Fig. 2 shows a map that represents the location of the monitoring points of the Boesmanspruit in Carolina, Mpumalanga.

This study used a quantitative research methodology. To accomplish the research objectives, historical data analysis was used. This study evaluated the existing water quality monitoring data from a mine in Carolina from 2017 to 2021. The act of monitoring water quality involves a systematic and numerical sampling process aimed at collecting information on the chemical, biological and physical properties of water (Yadav and Jamal 2018). The primary objective of assessing the quality of the aquatic environment has traditionally been to determine whether the observed water quality is appropriate for its designated purpose (SA DWAF 1996b).

The study used water quality data from samples as per the South African Water Quality Guidelines (SA DWAF 1996b). The research utilized historical data over the five years from 2017 to 2021 for six surface water points covering the dry and wet seasons. The data was collected from January to December of each year. The historical data samples were collected monthly, using the grab sampling technique by the mine. These samples were collected into clearly marked sterile two-liter polyethylene bottles. Using a permanent marker, the site code, date and time of sample collection were written on the water quality sample bottles. The caps of the bottles were left on until it was time to collect the sample. Samples were preserved using ice cubes and kept in cooler boxes. Samples were taken within 24 h after sampling, to a laboratory accredited with the South African National Accreditation System.

The following parameters were analyzed during this project: pH, electrical conductivity (EC), sulfate (SO₄), total dissolved solids (TDS), sodium (Na), magnesium (Mg), calcium (Ca), aluminum (Al), iron (Fe), manganese (Mn). The parameters were selected according to the study's objectives and land use types. The pH was measured on-site using a pH meter (American Public Health Association [APHA] 2005). All the remaining physiochemical parameters and metals, EC, TDS, SO₄, Na, Mn, Al, Fe, Mg and Ca, were analyzed by Yanka Laboratories in Emalahleni, Mpumalanga. SO₄ was detected using the gravimetric techniques as specified in the APHA guidelines, and Mn was analyzed using the Permanganate method (APHA 2005). Na was analyzed using the bismuthate method (APHA 2005). Ca was analyzed using the complexometric method; Fe was analyzed using the bipyridine method; Mg was analyzed using the pyrophosphate method; and TDS were analyzed using the residue-on-evaporation method (Kalkhajeh et al. 2019).

The evaluation of water quality necessitates the formulation of catchment strategies and the establishment of objectives for management of the resource (Olds et al. 2011). It was necessary to establish the limits of the catchment that will be evaluated and, if necessary, to break it into smaller, comparable management areas (Sivaranjani et al. 2015). The natural features and attributes of the catchment area that could impact the water quality of the water resource were also considered (Olds et al. 2011). The assessment also considered anthropogenic pollution sources and existing problems with water quality. In the end, it was important to comprehend how natural and man-made processes in the watershed affect the water quality and how they are related to those impacts (SA DWAF 1996a, b). To understand how water quality changes with the seasons, all the data on water quality were split into data sets for the dry (April–September) and wet (October–March) seasons.

The data analyzed were compared to the following water quality standards:

The CCME-WQI offers a practical method for reducing complicated water quality data and making it easier to communicate it to a broad audience. This index comprises three constituent parts: the scope refers to the quantity of parameters that fail to meet the established water quality standards, while the frequency denotes the rate at which these standards are not achieved, and the amplitude pertains to the extent to which the standards are not met. The CCME-WQI combines three measures of variation (scope, frequency and amplitude) into a single figure that quantifies water quality at a location in contrast to a set benchmark (Tyagi et al. 2013). Each of these three components of the CCME-WQI’s must be calculated after the selection of the water body, the period, the parameters and standards have been set. The CCME-WQI can be expressed mathematically in three steps as follows:

Step 1 was to compute the scope: where F₁ (scope) = the number of variables whose objectives are not met:

Step 2 was to compute the frequency, F₂ (frequency) = the number of individual tests that do not meet the objectives:

Step 3 was to compute the amplitude, F₃ (amplitude) = amount by which failed test values do not meet their guidelines and are calculated in the different steps.

When the test value must not be lower than the recommended guidelines values:

Once the factors have been determined, it is possible to calculate the index by adding the three factors together as though they were vectors. This was done using the CCME-WQI, as stated in Eq. (1).

The CPI is the proportion of the concentration of each analyzed parameter to the specified standards (Odipe et al. 2020). The CPI seeks to find a single value that will cut down on the number of parameters and presents data in a manner that is easy to understand (Tanjung and Hamuna 2019). The CPI can be expressed mathematically as:where

PI = pollution index of the induvial parameter, N = number of monitoring parameterswhere

C_i = observed value of ith parameter, S_i = RQO standard value of ith parameter.

The water quality indices (mentioned above) and their values were classified into different categories, and the CPI was then divided into five groups (Ramakrishnaiah et al. 2009; Son et al. 2020), as tabulated in Table 2.

To determine the status of the water quality, the mean and standard deviation of each parameter were calculated using Microsoft Excel. Microsoft Excel was also used to analyze the trend analysis for different concentrations at each monitoring point. To identify the spatial and temporal trends among the water quality set from the Boesmanspruit monitoring points and their possible sources, multivariate approach such as the Pearson correlation matrix, cluster analysis and principal component analysis (PCA) were performed using the commercial statistics software package IBM SPSS version 19. The application of multivariate statistical methods can aid in the interpretation of intricate data matrices, thereby enhancing comprehension of the water quality and ecological condition of the area being studied (Kazi et al. 2009). It will identify potential factors influencing the water system and provide a useful tool for dependable problem-solving (Kazi et al. 2009). Prior to conducting the PCA, the z-scores were adjusted to mitigate any potential bias that may have arisen from the utilization of differing scales. Determining the degree of association between two variables can be calculated using correlation analysis. The degree and significance of a correlation between two variables can be used to evaluate the nature of that relationship. The correlation, denoted by r, is taken to be a measure of strength, whereas the significance is quantified in terms of the probability levels (p values) (Maiolo and Pantusa 2021).

Cluster analysis and PCA are two examples of multivariate statistical methods that have been shown to be useful for analyzing and interpreting large and complicated water quality data sets (Liu et al. 2021). These methods can be used to pinpoint the origins of pollution, categorize sample locations, and quantify the spatial and temporal variability in water quality (Sharma et al. 2021). PCA is used to characterize the correlated data as a linear combination of independent variables (Pratama et al. 2020). This technique is popular because of its ability to both extract data and lower the dimensionality of the system. The correlation is a statistical measure that quantifies the degree of similarity between two variables. Cluster analysis is a multivariate technique for data analysis that utilizes statistical methods to divide a set of data points into smaller, more manageable groups (clusters) by maximizing the degree of similarity between items within each cluster and minimizing the degree of similarity between items in different clusters (Lei 2014). The method begins by classifying items into clusters based on their similarities (Lei 2014).

The water quality assessment results for the selected monitoring points from January 2017 to December 2021 are shown in Table 3. The findings that are highlighted in red shows where the measured water quality was found to be exceeding the set guidelines, and the ± symbol indicates the standard deviation. Out of the water quality parameters chosen to calculate the CCME-WQI and CPI indices in the study area, only six out of ten of the parameters were above the prescribed standards. Throughout the study period, it was observed that the levels of pH, EC, TDS, SO₄, Fe, and Mn exceeded the standards established by the RQOs (RSA DWS 2016), the target water quality range (TWQR) for aquatic ecosystems and agricultural use as outlined in the South African Water Quality Guidelines (RSA DWAF 1996a), and the CCME (2012) for guidelines for aquatic ecosystems for the preservation of aquatic areas. The standard guideline limits were met by Ca, Mg, Na, and Al, exclusively.

The results of the water quality trends are indicated in Fig. 3a–s. The pH values, which represent the levels of acidity and alkalinity, ranged from 4.2 to 10.2, except for site GR S12 in the wet period, which had a consistent increase with a pH above the upper limit of 8.8 between January 2019 and December 2020. All the points were within the RQOs upper limit during the monitoring period. The causes of this pH rise could be natural or geological factors. The pH dropped below the lower limit of 8.0 at monitoring points GR S26, GR S21, GR S13, GR S23 and TD SW01 between January 2017 and December 2021 for both seasons. These monitoring points are situated downstream of the coal mine. The EC values ranged from 8.8 to 393 mS/m. At site GR S26, the greatest EC value was recorded in July 2018 for both seasons. Throughout the observed period, the EC at site GR S13 remained stable and below the 30 mS/m limit. The graph shows that the RQO limit was exceeded at points GR S26, GR S12, GR S23, TD SW01 and GR S21 for both seasons. The TDS values ranged from 59.8 to 3 344 mg/l. At site GR S26, the TDS value spiked above the limit in January 2017 and dropped below the limit in October 2018–January 2020, and then it spiked again during the dry and wet season. Throughout the observed period, the TDS remained stable and below the 1 000 mg/l limit for the TWQR, except for spikes above the limit in GR S21 from July 2020 to January 2021 during the wet period. During the monitoring period, the concentration range of Ca was observed to be between 4.5 and 384 mg/l. Notably, all the recorded values were found to be within the limit of 1 000 mg/l, as stipulated by the TWQR for both seasons. The most prominent trend was observed at GR S26 in January 2017, with a minor decline noted between October 2018 and April 2019 during the wet season.

During the period of monitoring, the recorded Mg values remained below the limit of 500 mg/l as per the TWQR limits for all seasons. The Mg concentration exhibited a range of values spanning from 2.8 to 428 mg/l. The most notable trend was observed at GR S26 in January 2017, with a minor decline between October 2018 and April 2019, followed by a subsequent increase from January 2020 during the dry and wet season. During the monitoring period, the concentration range of Na was observed to be between 2.7 and 72.1 mg/l. Additionally, all the recorded values were found to be within the TWQR limit of 2 000 mg/l. During both seasons, the SO₄ concentration was between 7.1 and 2 993 mg/l. Points GR S26 and GR S21, which are located downstream of the mine, had concentrations that were significantly higher than the RQO limit of 80 mg/l. Throughout the monitoring period, the concentrations of SO₄ at points GR S13, TD SW01 and GR S23 remained stable and were found to be below the limit. However, at GR S23, minor spikes of above the TWQR limit of 80 mg/l were observed during the months of October 2017, July to August 2019, and July to November 2021.

The concentration range of Al was observed to be between 0.01 and 41.4 mg/l. During the monitoring period from March 2020 to January 2021, point GR S21 experienced a surge in both seasons. However, it is noteworthy that the Al readings remained within the TWQR limit of 5 mg/l throughout the monitoring period. The results for Fe ranged from 0.01 to 18.1 mg/l. The CCME limit of 0.3 mg/l for Fe was surpassed at all points throughout the monitoring period, except for point GR S26 from March 2021 to November 2021, and point GR S23 from December 2018 to February 2019 for both seasons. Between May and August of 2017, the greatest levels were recorded at point GR S21, and another surge was observed between May 2020 and January 2021, except for GR S13 and TD SW01. The results of the Mn results show that the Mn concentrations were greater than the of limit of 0.180 mg/l that applies to the TWQR for aquatic ecosystems. The greatest concentration values (62.43 mg/l) were found downstream of the mine at point GR S26 between May and September of 2018 and October 2020 to February 2021.

The water quality of the monitoring points indicated that the parameters exceeded the set thresholds (CCME 2012; SA DWAF 1996a, b; SA DWS 2016), except for Ca, Na, Mn and Al. The monitoring site GR S21 met the pH value for the upper limit of 8.8 as guided by the RQO for the catchment, but this point also had some spikes for mostly the wet season, indicating that the water was moderately alkaline. The lower limit of a pH of 8, as guided by the RQO, was not met by all monitoring points, which indicate that the water around the monitoring points were acidic. Monitoring site GR S26 had a low pH of 4.3 during the dry season of April 2019, whereas GR S21 and GR S23 had the lowest pH of 3.3 during January 2020–December 2020. The acidity of the water may be related to the breakdown of sulfide minerals in the coal stockpiles and runoff.

The EC in the study area showed an increasing trend for all monitoring points except for point GR S13, which remained under the RQO limit. The points that showed a high spike in the EC were GR S26, GR S21, and GR S23. This was mainly caused by the runoff from the mining operations and leaching of minerals from rock formations. Since the EC concentration is a reliable predictor of the quality of river systems, it appears that the water quality of the study area, particularly regarding this sampling points, has not been good. Possible causes for the rise in EC include the ongoing introduction of salts from both natural and anthropogenic sources.

The TDS has shown a high increase at point GR S26 for both the dry and wet seasons as regulated by the TWQR for livestock watering. The increase was caused by the presence of harmful substances such as heavy metals that are harmful to the aquatic life.

The SO₄ trend for the study area indicate a high elevation of the concentration above the RQO of 80 mg/l for point GR S26, which showed a spike up to 2 999 mg/l. The spike in the concentration was seen for both the dry and wet seasons. Point GR S21 also showed a spike from January 2019 to February 2021, which was higher than the 80 mg/l. The presence of high SO₄ indicated that there was a formation of acid mine drainage (AMD) in the vicinity, emanating from coal mining activities.

The Fe concentration was set for 0.3 mg/l, using the CCME guideline. The concentration of Fe was exceeded by all monitoring points. The highest concentration was at monitoring point GR S21 and the trend was mostly picked up during the dry season.

McCarthy and Humphries (2013) conducted a study to assess the contamination of the water supply in the town of Carolina at the Boesmanspruit Dam. The abrupt decrease in pH to 3.7, coupled with increased concentrations of Fe, Al, Mn, and SO₄, resulted in the water becoming hazardous and inappropriate for utilization. The origin of the contamination remains ambiguous as it is not evident how the dam has undergone rapid pollution. The present study was conducted to examine the circumstances that led to the pollution of Carolina’s water supply. The primary objectives were to determine the potential cause of the incident and to evaluate its implications for other dams situated in the Vaal River system. The results of the chemical analyses conducted on water samples indicated that the source of pollution can be traced back to the Witrandspruit sub catchment. The accumulation of seepage from coal mines in a wetland located upstream of the dam is believed to be the cause. In the context of an unusually intense precipitation event, storage facilities that contained contaminated runoff originating from coal management infrastructures, exceeded their capacity and consequently discharged the contents of the wetland into the Boesmanspruit Dam.

From this study, it has been discovered that a point can exhibit different results from the CCME and CPI indices. A point that meets the moderate pollution threshold according to the CCME-WQI may be classified as heavily polluted according to the CPI. The reduction of various variables to a smaller number can be achieved through the assessment of water quality levels. Likewise, CPI revealed the comparatively reliable state of water quality for the remaining ranking classifications during both seasonal periods. The utilization of multiple variables and a vast dataset in CCME-WQI may result in the potential loss of information. The CCME-WQI values for the monitoring points, which included parameters such as Fe, Mn, Al, pH, SO₄ and EC, which exceeded the permissible limit for five parameters in most monitoring points. Notably, the values for CCME-WQI were nearly comparable across all points, whereas the CPI values exhibited variability. Comparable results were observed in the case of the Al and Fe samples in monitoring point GR S13 during both the wet and dry seasons, wherein a single parameter surpassed the recommended threshold. The matter at hand originated from the fact that the CCME-WQI solely takes into account the exceeded parameter in its computation, thereby disregarding the potential integrated impact of all parameters. On the contrary, the CPI recognized each individual metric and its respective function in the context of integrated water quality. A limitation that pertains to the calculation of CCME-WQI, is that it cannot be computed for a singular parameter, given that F1 would constantly obtain a perfect score of 100, which is a false interpretation. Additionally, it should be noted that the CCME-WQI score of 100 may be misleading as it does not consider the lack of relationship among the parameters and their allowable ranges. The integrated water quality measurement, such as CCME-WQI, necessitates a non-zero value; a value of zero is not acceptable. In addition, it is noteworthy that while the CPI identified the presence of pollutants in the form of Al and Fe at monitoring point GR 13, the CCME-WQI also classified the water at that location as good, despite the slight pollution detected by CPI. The collective effects of the input parameters were responsible for this outcome. The study found that if certain variables, including SO₄, Al, Fe, EC and Mn, exceeded the permissible range, the water quality would deteriorate in accordance with the CPI classification.

The findings indicate that CPI may yield more favorable results and productivity in comparison to CCME-WQI, which employs a more comprehensive approach. The use of the CCME-WQI in water quality assessment found that water samples taken close to the mining activities, specifically GR S26 (41.67 for the dry season and 41.70 for the wet season) and GR S21 (50.13 for the dry season and 44.3 for the wet season), exhibited poor water quality. The Boesmanspruit is constantly at risk of being threatened because of continuous mining operations, and in this instance, it may have been the runoff from the area where coal stockpiles are located. The excessive levels of EC and Fe, in addition to the high concentrations of SO₄, TDS and Mn, may be responsible for the poor water quality. At monitoring points GR S26 and GR S21, the most failed parameters and tests that failed to accomplish their objectives were observed. The CCME-WQI found that the water quality at points that are the furthest from the mining operation (TD SW01, GR S13, GR S23, and GR S12) ranged from good to marginal. The calculated results of the CPI indicated that the monitoring points that are closer to the mining operations—in this case, the GR S26 (4.891 for the dry season and 4.127 for the wet season) and GR S21 (2.484 for the dry season and 2.85 for the wet season—were heavily polluted during the dry and wet seasons. These points are significantly contaminated and poses a great threat for aquatic life because of the wastewater produced during coal mining. Based on the results of the calculations, it appears that many anthropogenic pollutant sources may also be to blame for water pollution at other monitoring locations. The GR S23 (1.371 for the dry season and 1.626 for the wet season) indicated that this monitoring point was moderately polluted for both the dry and wet seasons, whereas monitoring points GR S12 (0.435 for the dry season and 0.585 for the wet season), GR S13 (0.510 for the dry season and 0.541 for the wet season) and TD SW01 (0.473 for the dry season and 0.594 for the wet season) indicated that the points are slightly polluted.

In a study carried out by Son et al. (2020), the researchers utilized five different water quality indices, one of which was the CPI, to evaluate the water quality during both the dry and wet seasons in the Cau River located in the northern region of Vietnam. The CPI results showed that there were no substantial variations between the dry and wet seasons. The river’s CPI varied from 0.50 to 1.57 during the dry season, indicating that it was slightly to moderately polluted. The CPI ranged from 0.66 to 1.37 during the wet season, indicating that there were no variances due to seasonal changes. The results from the study by Son et al. (2020) also corroborate the results from this study, which also indicated that there were no seasonal variations from the results for all the monitoring points.

The results obtained from the CCME-WQI and the CPI indicate that the points located downstream from the mining activities exhibited higher levels of pollution in comparison to the areas that are situated further away from the mining activities. The results of the water quality indices indicate that the optimal approach for categorizing the water quality of the Boesmanspruit is through the utilization of the CPI.

The use of cluster analysis was effectively utilized to successfully discover three groups of variables that shared similar characteristics, and the outcomes of PCA additionally proved the reliability of the clustering result. The dry season was found to comprise three clusters, each consisting of three components. Two primary sources have been found for the three significant components based on their analysis: (1) The study area was found to have anthropogenic sources of EC, TDS, Ca, Mg, Na, SO₄, and Mn, which were attributed to coal mining activities; (2) pH and Fe, on the other hand, were found to originate from natural sources, (3) while Al was found to originate from other natural sources. The analysis revealed that the wet season consisted of two components that were categorized into three distinct clusters and had two primary sources: (1) The presence of TDS, Mg, Ca, EC, SO₄, Na, and Mn in the area can be attributed to coal mining activity. (2) The appearance of Al and Fe, on the other hand, seems to have originated from natural sources other than coal mining.

The Pearson correlation coefficient analysis showed that the variables EC, TDS, Ca, Mg, Na, SO₄, and Mn all shared a common source, whereas Al, Fe and pH had a different source. The results of both the PCA and the cluster analysis agree with these views. The correlation results, PCA and cluster analysis identified three sources of pollution from the study area. During both the dry and wet seasons, the PCA results indicated that EC, TDS, Ca, Mg, Na, SO₄, and Mn come from one source. The first group of elements EC, TDS, Ca, Mg, Na, SO₄, and Mn had a strong positive correlation with PCA for both the dry and wet seasons.

The first factor explains 58.19% of the total variance for the dry season and 58.75% for the wet season. The results indicate a significant presence of pollutants such as EC, TDS, Ca, Mg, Na, SO₄ and Mn across all monitoring points in the Boesmanspruit during both the wet and dry seasons. This group for both the dry and wet seasons was contributed by the anthropogenic sources emanating from the coal mining activities at the study area. The surrounding mining industries contributed a high concentration of ions, which contributed to the relatively high EC value in the study area. Mining operations in the catchment area may be to blame for the high levels of TDS due to the increased concentration of dissolved and suspended solids. The coal mining and coal storage areas are likely to be the source of AMD that has contaminated the water of the Boesmanspruit, with an excessive level of SO₄. Fe and pH coming from the same source, and they were all confirmed by the three statistical analyses done for this study.

The second group of elements extracted by the PCA loading was pH and Fe for the dry season and accounts for the 12.25% of the total variance. The wet season extracted the following elements dominated by Al and Fe that accounts to 12.08% of the total variance. The group of elements extracted by the PCA emanated from the natural sources. The third group of extracted elements using PCA loading for the dry period was Al, which accounted for 10.23% of the total variance. This component appeared to have originated from other natural sources, namely weathering of rocks. This can be compared to a study was conducted by Novhe et al. (2016), aiming at examining the mined watershed located in the Ermelo coalfield in the Mpumalanga province, with the objective of identifying the sources of metal pollution and the pathways associated with it in the Boesmanspruit dam in Carolina. The results to the study found that the pollution was attributed to both abandoned and operational mines and manifested as AMD with a low pH and elevated levels of Fe, Al, Mn, and SO₄. Additionally, it was discovered that streams serve as significant pathways for pollutants.