Introduction

Acid mine drainage (AMD) is a serious environmental problem that affects mining areas and its surroundings around the world (Abrahams and Carranza 2023; Adeniyi et al. 2022; Cánovas et al. 2022; Kumar et al. 2023; Nascimento et al. 2023; Ojonimi et al. 2021). It is an acidic effluent generated as a result of metal-sulphide oxidation in the presence of air and water and it is rich in sulfates and metals (Skousen 2014; Skousen et al. 1999). Acidic conditions can favor the leaching of trace metals from surrounding host rocks and waste materials into surface waters (Masindi et al. 2017; Skousen et al. 2019). Elevated trace metal concentrations are among the biggest concerns associated with AMD. Unlike many organic compounds, trace metals are non-biodegradable and continue to accumulate within the environment over time (Manahan 1991; Uysal et al. 2009). Several trace metals, such as Cu, Zn, and Co, are considered crucial for the maintenance of health in biological systems (Ali et al. 2019), but can become toxic to the environment and humans at elevated concentrations (Alomary and Belhadj 2007). In addition, trace metals such as Pb and Cd are toxic at even very low levels of exposure (Tchounwou et al. 2012). Thus, the detection and monitoring of trace metal concentrations within the environment is crucial to mitigate the associated health risks (Bakirdere and Yaman 2008).

Streambed and overbank sediments along streams and rivers draining mines can act as important sinks and secondary sources of trace metals (Galán et al. 2003; Kos et al. 2022; Schulz-Zunkel and Krueger 2009). This is largely due to the presence of secondary oxides and oxyhydroxides (Campaner et al. 2014; Parker et al. 2007; Schaider et al. 2014; Zhao et al. 2012), which are known to concentrate trace metals in soils and sediments because of their high specific surface areas, high cation exchange capacities and affinity for forming colloids and coatings on other minerals (Sparks 2002; Wilkin 2008). Thus, they play a crucial role in regulating dissolved trace metal concentrations along streams and rivers (Webster et al. 1998).

Conventional geochemical techniques for detecting and monitoring trace metal contamination in the environment typically require: extensive field sampling, thorough sample preservation, the use of strong acid digestions, and expensive (priced per element analyzed) laboratory analysis, all of which can be very time-consuming (Pandit et al. 2010). As a result, conventional geochemical techniques can often be inefficient and expensive when performed on a large scale (Ren et al. 2009) and in investigations requiring rapid data analysis (Kemper and Sommer 2002, 2003; McCarthy and Humphries 2012). Thus, there is a need to develop a simple method for detecting trace metal contamination (considering as few environmental factors as possible) to guide more extensive and costly geochemical analyses.

Here, we hypothesized that dissolved—relative to sediment—concentrations of major elements associated with secondary oxides and oxyhydroxides in AMD-contaminated environments, could serve as good proxies for detecting AMD-related trace metal contamination in streams. To investigate our hypothesis, we calculated Fe and Al ‘dilution factors’ by dividing stream water concentrations (mg/L) of Fe and Al by their respective overbank sediment concentrations (mg/L), and then we mapped these [stream water/overbank sediment] dilution factors with dissolved trace metal concentrations in the study site. Fe and Al were selected for the calculation of dilution factors because they are the most common major elements associated with AMD (Alpers et al. 1994; Bigham and Nordstrom 2000; Nordstrom 2020) and were dominant in overbank sediments in the study site to a much larger extent than Mn.

The objectives of this study were to: characterize stream water chemistry at the study site, assess the extent of metal contamination in the study site based on enrichment factors (EFs) and contamination factors (CFs), and evaluate the potential of Fe and Al dilution factors for detecting trace metal contamination in the study site. The advantages of using Fe and Al to detect AMD-related trace metal contamination in streams are that they are major elements in AMD-contaminated environments and, thus, can be analyzed with greater precision than trace metals, which are typically less precise because they are closer to the detection limit of the analytical equipment (Hall 1998) and would require the analysis of dissolved and sediment concentrations of only two elements (i.e. Fe and Al), as opposed to a number of AMD-related trace metals, thus, potentially and markedly reducing the cost and time of laboratory analysis.

Materials and Methods

Study Area

Description of the Study Site

The Blesbokspruit River (Fig. 1) is situated in the Witbank Coalfield and forms part of the Olifants River catchment in South Africa. It is located ≈5 km northwest of the town of Emalahleni, Mpumalanga. The Blesbok Colliery, which has been mined by CCCR Commodities (2022) since 2020, is currently operating there. The main features of the study site include four upstream acid ponds and a wetland ≈3 km downstream of the acid ponds. The acid ponds were constructed to reduce the impact of polluted underground mine-water on the Blesbokspruit River (Bell et al. 2001) and are considered to be a source of AMD. The Blesbokspruit River was deemed suitable for study based on previous AMD studies at this site (Bell et al. 2001, 2002; Netshitungulwana et al. 2013). The climate of the area and its surroundings, based on the Köppen climate classification, is Cwb (temperate, dry winter, warm temperature). The average temperature during autumn (the season when we did our sampling) is 13–16 °C and average rainfall is 15–45 mm (Climate-data.org 2022). Stream flow, at the time of sampling, was laminar. The river showed no signs of aquatic life and, with the exception of the wetland and acid pond areas, was only sparsely vegetated. Thus, the influence of biota and organic matter on trace metal attenuation was not considered.

Fig. 1
figure 1

Localities for stream water sampling with- (red dots) and without (purple dots) overbank sediment samples along the Blesbokspruit River, Mpumalanga, South Africa. Stream water samples were collected at two sites approx. 5 m apart at each of the eleven different localities. Similarly, overbank sediment samples were collected at two sites roughly 5 m apart at each of the six different localities. Flow direction is indicated by the black, dashed arrow. Also shown is a wetland (green dash lines) and acid ponds (yellow rectangle)

Geology

The Witbank Coalfield is located within the northern portion of the Main Karoo Basin and by 2014, it was one of the major suppliers (yielding  > 50%) of saleable coal in South Africa (Hancox and Götz 2014). The study area is characterized largely by outcrops of the Vryheid Formation, which forms part of the Permian Ecca Group of the Karoo Supergroup (SACS 1980), and several intrusive dolerite dykes and sills (Cairncross 2001). Rock sequences of the Vryheid Formation in the study area consists mainly of siltstone, sandstone, carbonaceous shale, five bituminous coal seams, and minor amounts of conglomerate (Cairncross 2001). The strata of the Witbank Coalfield are dominated by quartz and kaolinite, with trace to minor concentrations of pyrite, calcite, and dolomite (Azzie 2002; Pinetown 2003). Pyrite, which was present in percentages between 0.1 and 37.1%, was determined to be the dominant sulfide in these strata (Pinetown 2003). In terms of major element geochemistry, the Witbank Coalfield strata are dominated by SiO2 and Al2O3, with minor to trace amounts of Fe2O3, S, and CaO. With respect to trace element geochemistry, the strata are dominated by Cr with minor to high concentrations of Ni, Zn, Co, and Cu, and trace to minor concentrations of Pb (Cairncross 2001; Pinetown 2003).

Field Sampling and Analyses

Sampling was conducted during autumn (one of the driest seasons in the study area and its surroundings) and, thus, under base flow hydrological conditions. Base flow is defined as the portion of stream flow not directly generated by excess precipitation (Price 2011). The sampling was conducted in the dry season to minimize the influence of rainfall on the dilution factors. In addition to dilution, increased rainfall can also mobilize stored AMD (and related trace metals) in surface waters in the vicinity of coal mines (Jewiss et al. 2020) and thus undermine the utility of dilution factors when sampling is done during wet seasons. The irregular sampling interval was due mainly to water-logged sediments in the wetland between sample localities #3 and #6, sewerage contamination, and, at the time of sampling, bush burning between sample localities #6 and #7 (Fig. 1). Thus, water and sediment samples were sparse.

Data sparsity is occasionally unavoidable in geological studies (Davis 2002) and is a common scenario in environmental studies (e.g. n = 6 in Ayora et al. 2022; n = 5 in Fitzsimons and Courtney 2022; n = 10 in Petrini et al. 2016). However, Davis (2002) argued that sparse data can still be useful when handled with suitable statistical analyses and assessments of uncertainty. Therefore, the sparsity of data in our case was addressed by applying statistical analyses that are suitable for small datasets and the associated uncertainty was quantified using statistical significance (p) and confidence intervals (CIs).

Water Samples

Twenty-two stream water samples were collected (at a depth of ≈10 cm) at 11 different locations ≈0.2–1.5 km apart (i.e. two samples collected ≈5 m apart at each location). This was done because ion concentrations at adjacent locations along a stream can vary considerably during periods of low flow (i.e. during dry seasons; Floriancic et al. 2019). Bell et al. (2001) found that dissolved metal concentrations more than 6 km downstream of the acid ponds were almost negligible due to the alkalizing and neutralizing effect of the Prison Stream tributary, so only the first 6 km of the Blesbokspruit River downstream of the acid ponds were sampled. Sampling was conducted upstream from sample location #1 towards the acid pond contamination source (Fig. 1) to ensure minimal disturbance and cross-contamination downstream.

Physico-chemical parameters (pH and EC) were measured in situ using a HANNA 9828 multisensor probe and Mettler Toledo portable pH meter. Stream water samples were filtered using 0.45 µm nylon syringe filters, preserved using 10% HNO3 and kept in cool storage (U.S. EPA 1983) for analysis by ICP-AES/MS. ICP-AES/MS was used for elemental analysis because it provides low to very low (μg/L to ng/L) detection limits, which are crucial for analysis of trace metals in environmental studies. Elements were measured using prepared calibration solutions and quality control methods based on U.S. EPA guidelines (Stellenbosch University 2022). Al, Fe, Mn, Co, Ni, Cu, Zn, Pb, Cr, and Cd were selected for analysis because they are commonly associated with AMD (España 2007; Nieto et al. 2007; Sengupta 1993). Samples meant for sulfate analysis were filtered using 0.45 µm nylon syringe filters and kept cool (U.S. EPA 1983) prior to analysis by the Skalar Bluevision™ discrete analyzer.

Overbank Sediment Samples

Twelve overbank sediment samples were collected at six different locations (i.e. two samples collected ≈5 m apart at each location) along the Blesbokspruit River (Fig. 1). We collected samples at each location because Walling and He (1998) found considerable spatial variability of overbank sediment. Samples were collected at a depth of ≈10 cm, ≈1 m distance from the active stream. The advantages of using overbank (floodplain) sediments are that large quantities of sample are easy to collect and that in mine and industrially polluted areas, they may represent the most important secondary source of metal pollutants (Macklin and Klimek 1992). The overbank sediment samples were air-dried and stored in high density polyethylene plastic bags at room temperature. Samples meant for XRD Rietveld analysis were milled to size  < 75 µm and analyzed using the Bruker D8 Advance Diffractometer (Brime 1985).

Samples meant for chemical analysis were sieved to  < 63 µm size fraction because the trace metals of interest (Co, Ni, Cu, Zn, Pb, Cr, and Cd) are typically concentrated in this size fraction (Förstner and Salomons 1980). Sediment samples were treated with reverse aqua regia (3 HNO3: 1 HCl) and were digested using a Mars™ microwave digestion system to determine ‘near-total’ metal contents. ‘Near-total’ digestion (which does not include trace metals hosted in the crystalline lattice of primary minerals) was preferred because it represents the proportion of metals that is environmentally extractable (Shahbazi and Beheshti 2019). ICP-AES was used to analyze Fe and Al in sediments because it is suitable for major to minor elements while ICP-MS was used to analyze Mn, Sr, Ni, Cu, Zn, Pb, Cr, and Cd (which were present in much lower concentration ranges than Fe and Al) because it is more suitable for analysis of trace to ultra-trace elements (Stellenbosch University 2022).

Quality Control and Censored Values

Quality control and quality assurance (QC/QA) methods for stream water included the use of laboratory blanks and field- and analytical duplicates. The QC/QA methods for sediments included the use of field-, laboratory- and analytical duplicates as well as soil certified reference material (CRM). Elements with precision of 20% or better were included in further data analysis (Ramsey 1998). Censored data were substituted with 1/2 the detection limit as recommended by the U.S. EPA (1998, 2000). The detection limits of the analysis per metal via ICP-AES were (in mg/L): Al (13.50) and Fe (4.50); and via ICP-MS were (in μg/L): Al (0.83), Fe (0.90), Mn (0.04), Co (0.03), Ni (0.34), Cu (0.47), Zn (1.85), Pb (0.02), Cr (0.62), Cd (0.003) and Sr (0.07).

Data Analysis

Summary Statistics

The median and median absolute deviation (MAD) were used as measures of central tendency and variance, respectively. The advantage of using the median and MAD, as opposed to the mean and standard deviation, is that they do not assume a normal distribution and, thus, are robust against outliers (Reimann and Filzmoser 2000). Robust methods were necessary for this work because of the presence of outliers and the sparse number of stream water (n = 22) and sediment (n = 12) samples.

Contamination Assessment Indices

Metal contamination along the Blesbokspruit River was assessed using enrichment factors (EF) (Hakanson 1980), contamination factors (CF) (Hakanson 1980), and element (Fe and Al) ‘dilution factors’ (introduced as part of this study).

Enrichment Factor (EF): The EF technique assesses the degree of metal enrichment in soils, sediments, and surface waters relative to background metal concentrations and a conservative element. The goal is to distinguish between enrichment resulting from natural processes and anthropogenic impact (El-Kady et al. 2019). EF was defined by Hakanson (1980) as:

$$EF = ({\text{M}}/{\text{X}})_{{{\text{sample}}}} /({\text{M}}/{\text{X}})_{{{\text{background}}}} ,$$
(1)

where Msample and Mbackground, respectively, are the concentrations of the metal of interest in the sample and background data; and Xsample and Xbackground, respectively, are the concentrations of a conservative element in the sample and background data. Contamination classes based on the EF were defined according to Birch (2003) as: no enrichment (< 1); minor enrichment (1–3); moderate enrichment (3–5); moderately severe enrichment (5–10); severe enrichment (10–25); very severe enrichment (25–50), and extremely severe enrichment (> 50).

Here, the MAD was used to represent background concentrations (Esmaeili et al. 2014; Rezapour et al. 2022). Use of the MAD results in site-specific enrichment factors and, thus, may address some of the concerns around the use of regional or global backgrounds in calculations of the EF (e.g. Reimann and de Caritat 2000, 2005). A conservative element is one that is generally stable in soils and sediments and is largely of lithogenic provenance (Dan et al. 2014). The EF can generally be calculated using Al, Fe, Mn, Ti, Sc, or Sr as a conservative element (Altıkulaç and Turhan 2023; Fan et al. 2019; Gaberšek et al. 2022; Kicińska and Wikar 2021; Kowalska et al. 2018; Kükrer et al. 2020). Unlike Al, Fe, and Mn which are commonly associated with coal mine drainage (Equeenuddin et al. 2013; Seo et al. 2017; Silva et al. 2011), aqueous Sr concentrations are generally associated with the natural weathering of geogenic materials along a river or streambed (Cánovas et al. 2007; Cao et al. 2020; Lenoble et al. 2013). Thus, Sr was considered a suitable conservative element.

Contamination factor (CF): The CF evaluates the anthropogenic component of trace metal concentrations in a study site based on the ratio of metal concentration in the sample to background concentrations of that metal. The CF was defined by Hakanson (1980) as:

$$CF = {\text{C}}_{{{\text{sample}}}} /{\text{C}}_{{{\text{background}}}} ,$$
(2)

where Csample and Cbackground represent the concentration of a dissolved metal in the sample and background concentrations (here, the MAD) of that metal, respectively. Contamination classes based on the CF were defined according to Hakanson (1980) as: no to low contamination (< 1.5); low contamination (1.5–2); moderate contamination (2–4); high contamination (4–8); very high contamination (8–16); extremely high contamination (16–32), and ultra-high contamination (> 32).

Element (Fe and Al) dilution factor: Dilution factors of Fe and Al are proposed in this present study and are defined here as:

  • Fe dilution factor = Festream water/Fesediment, and

  • Al dilution factor = Alstream water/Alsediment, respectively,

where Festream water and Alstream water represent Fe and Al concentrations in the stream water; and Fesediment and Alsediment represent Fe and Al concentrations in the overbank sediments. Contamination classes were defined according to the median ± MAD concentrations of the Fe and Al dilution factors and are represented as percentages (Supplemental Table S1).

Spatial distribution maps were generated using untransformed data of single element concentrations overlain with contamination classes based on the dilution factors of Fe and Al for comparison. The Fe and Al dilution factors and individual trace metal concentrations were not transformed because they do not originate from the same, closed composition and, thus, were considered independent (Reimann and de Caritat 2017; Reimann and Filzmoser 2000). Correlations among the calculated (stream water/overbank sediment) Fe and Al dilution factors and dissolved trace metal concentrations were evaluated using Spearman’s (1904) rank correlation analysis (Reimann and Filzmoser 2000). The CIs of correlations and hierarchical cluster analysis (HCA) based on the average linkage (between groups) method and squared Euclidean distance of trace metal concentrations were used to assess the robustness of the median ± MAD concentrations of the Fe and Al dilution factors for classifying AMD-related trace metal contamination in the streams. Centered log-ratio (clr) transformation was performed using CoDaPack (Comas-Cufí and Thió-Henestrosa 2011) and used to ‘open’ the ‘closed’ geochemical data (Aitchison 1981, 1986) prior to cluster analysis. Clr transformation was considered suitable because it produces results of better meaning (Carranza 2011) without requiring back-transformation to the original values for interpretation, as is necessary for isometric log-ratio (ilr) transformation (Egozcue and Pawlowsky-Glahn 2005).

Results and Discussion

Mineralogy

The results of XRD Rietveld analysis and ICP-AES/MS analysis of overbank sediments along the Blesbokspruit River are shown in Table 1. According to Table 1, the Blesbokspruit River overbank sediments were dominated by quartz (95.5–100%), with minor kaolinite proportions (0–4.5%), which is in agreement with Pinetown (2003). In addition, the overbank sediments showed elevated concentrations of amorphous Al (median = 4.18%) and Fe (median = 2.46%) oxides and oxyhydroxides, with minor concentrations of amorphous Mn (median = 201 mg/L) oxides and oxyhydroxides. Increased Fe (6.21%) and Mn (885 mg/L) content in sediments at sample location #1, compared to sample locations further upstream, is consistent with the precipitation of iron and manganese oxides and oxyhydroxides following the confluence of the Blesbokspruit River with the uncontaminated Prison Stream (Bell et al. 2001). The very high Al content (7.79%) in the acid pond (sample location #11) is consistent with periodic treatment of the ponds with soda ash (sodium carbonate) (Janse van Rensburg 2003), which can very efficiently precipitate Al from acidic waters (Masindi et al. 2017). Overall trends in metal content in sediments along the study site are consistent with the findings of Equeenuddin et al. (2013), Sahoo et al. (2017), and Santos et al. (2015), who noted much higher Fe and Al content in sediments in the vicinity of coal mines, compared to Mn contents.

Table 1 XRD Rietveld analysis and ICP-AES/MS analysis of twelve overbank sediment samples collected at six different locations along the Blesbokspruit River

Stream Water Chemistry and Metal Enrichment

According to Table 2, the Blesbokspruit River water had pH values ranging from acidic (< 3) to near-neutral (6.8) and EC values between 180 and 2240 (μS/cm). Dissolved sulfate ranged from 73.7–1530 (mg/L). Decreasing median major and trace metal concentrations in the Blesbokspruit River water were: Al (7.41 mg/L) > Mn (2.30 mg/L) > Fe (0.50 mg/L) > Sr (372 μg/L) > Zn (328 μg/L) > Ni (97 μg/L) > Co (71 μg/L) > Cu (2.64 μg/L) > Cr (0.86 μg/L) > Pb (0.71 μg/L) > Cd (0.58 μg/L). Median trace metal concentrations in the Blesbokspruit River waters appeared consistent with those documented by Miranda et al. (2022), who suggested that Zn and Ni typically have the greatest potential for deteriorating water quality because of their strong affinity for the aqueous phase. Among the elements typically considered conservative (Al, Fe, Mn, and Sr), Sr showed the lowest variability (MAD = 127 μg/L), supporting its use as a conservative element in the calculation of EFs (Loska et al. 2003). In addition, the higher Sr content in the acid pond (sample location #11; pH = 3.5), compared to sample locations #7, #8, #9, and # 10 (pH < 3.0), suggests that increased Sr content is less related to pH but is likely more associated with prolonged contact of stream water with streambed materials in the pond, as suggested by Cánovas et al. (2007).

Table 2 Physico-chemical parameters pH, EC (µS/cm), average dissolved sulfate (mg/L), dissolved major metal (Al, Fe, Mn) content shown in mg/L, trace metal (Co, Ni, Cu, Zn, Pb, Cd and Cr) and Sr content shown in µg/L in water samples collected at eleven different localities along the Blesbokspruit River

Bell et al. (2001) reported median dissolved Al (113 mg/L), Fe (26.4 mg/L), Mn (12.4 mg/L), Cu (1.1 mg/L) Ni (1.9 mg/L), Pb (1.4 mg/L), and Zn (2.4 mg/L) concentrations for roughly the same portion of the Blesbokspruit River. When the water chemistry from the present study (Table 2) was compared with that of Bell et al. (2001), it is clear that dissolved metal concentrations in the Blesbokspruit River have markedly decreased over the last 20 years. A possible reason for this decrease may be the refurbishment of the Brugspruit Water Pollution Control Works treatment plant in 2014 (Rand Water 2014). The plant was originally built in 1997 by the South African Department of Water Affairs and Forestry to treat the acidic waters of the Brugspruit and Blesbokspruit Rivers. However, due to equipment theft, insufficient staff, and lack of maintenance, the plant had been non-operational for lengthy periods of time (Hobbs et al. 2008; McCarthy and Pretorius 2009).

According to the EFs in Table 3, waters along the Blesbokspruit River were characterized by: no to minor enrichment of Al, Mn, Co, Ni, Zn, and Cd; no to moderately severe enrichment of Pb, Cu, and Cr; and no to very severe enrichment of Fe. According to the CFs in Table 3, the study site was characterized by: no to moderate Mn, Co, Ni, Zn, and Cd contamination; no to high Al contamination; no to very high Cu, Pb, and Cr contamination; and no to ultra-high Fe contamination. Low pH, high EC, high sulfate, minor to moderately severe enrichment of trace metals, and ultra-high degrees of Fe contamination are clear indications that the Blesbokspruit River is still affected by AMD.

Table 3 Enrichment factors (EFs) and contamination factors (CFs) for Al, Fe, Mn, Co, Ni, Cu, Zn, Pb, Cd and Cr

Spatial Variability of Metals

Concerning the spatial variability of dissolved trace metals (Supplemental Fig. S1), maximum trace metal concentrations (μg/L) occurred within 0.2–0.5 km downstream of the acid ponds for all of the studied metals except Cu (Supplemental Fig. S1f), which showed maximum dissolved trace metal concentrations ≈1.1 km downstream of the acid ponds (corresponding with the maximum kaolinite proportions in Table 1). The consistency between kaolinite and Cu was also documented by González Costa et al. (2017), who stated that higher clay contents in soils increase the bioavailability of Cu.

Trace metal (Co, Ni, Cu, Zn, Pb, Cr, and Cd) concentrations were notably lower at ≈2.8 km downstream of the acid ponds and lowest more than 4.6 km downstream of the acid ponds. Overall, dissolved trace metal concentrations typically decreased with increasing distance from the acid ponds, which is consistent with the findings of Bell et al. (2001). However, sample locality #6 showed dissolved trace metal concentrations lower than sample localities #5 and #4, which were located further downstream from the acid ponds. This is likely because samples at locality #6 were collected on the final day of sampling, following a day of rain that may have subsequently diluted the samples. The increased Mn, Co, and Ni content noted at sample location #1 is likely due to the dissolution of Mn, Co, and Ni that were scavenged by Fe oxide and oxyhydroxides in sediments (Table 1) at this location (Teixeira et al. 2001).

The spatial variability of metal concentrations in the waters of the Blesbokspruit River appeared to be strongly influenced by the pH, dissolved sulfate concentrations, the presence of a wetland, and the density and health of surrounding vegetation.

Role of pH: Trace metal concentrations (Supplemental Fig. S1d–j) exhibited a strongly inverse relationship with pH. This was expected because the hydroxyl groups on variable-charge oxide and clay mineral surfaces become deprotonated at near-neutral pH, thereby facilitating cation adsorption. At lower (acidic) pH levels, they become protonated, thereby facilitating cation desorption (Dzombak and Morel 1990; Sparks 2002). The relatively inverse relationships among dissolved Al, Fe, and Mn concentrations and pH (Supplemental Fig. S1a–c) suggest that these metal concentrations were strongly influenced by the solubility of hydroxides at low pH and, thus, they were likely in the form of Al3+, Fe3+, and Mn3+ (Cravotta 2006).

Role of dissolved sulfate: Trace metal concentrations (Supplemental Fig. S1d–j) in the acid ponds (sample locality #11 and in the study site, the supposed source of contamination) were notably lower than expected. This may be related to the much lower dissolved sulfate concentrations in the acid ponds compared to sample locality #10, 0.2 km downstream of the acid ponds (Table 2) since elevated concentrations of dissolved sulfate can increase dissolved metal concentrations via the formation of soluble metal sulfate complexes (Cravotta 2006). The higher pH (3.5) and lower dissolved sulfate content (216 mg/L) in the acid ponds compared to sample location #10 (pH = 2.6 and 1530 mg/L), is consistent with periodic treatment of the acid pond with soda ash (Janse van Rensburg 2003). Soda ash can cause the formation of insoluble sulfate complexes, which precipitate and settle out of solution (Masindi et al. 2017).

Role of the wetland and surrounding vegetation: The wetland was crucial in trace metal attenuation along the Blesbokspruit River, having recorded some of the lowest dissolved trace metal concentrations in the study site (Table 2). These findings were similar to those of Bell et al. (2001), who observed subdued trace metal concentrations in the Blesbokspruit wetland, relative to measurements further upstream. Metal attenuation in the wetland may be attributed to several processes, including settling and sedimentation of particulate‐bound metals, phytoextraction (uptake by and accumulation in plants), reduction and oxidation by bacteria, and adsorption to oxides, oxyhydroxides, clays, and organic matter present in the wetland (Sheoran and Sheoran 2006).

Wetland plants at sample locality #4 were noticeably drier than plants along other sections of the wetland. Water deficiency in plants can negatively affect their photosynthetic efficiency and root activity (Oguz et al. 2022; Peng et al. 2022), both of which play important roles in phytoextraction (Cassina et al. 2012; Wu et al. 2017). Thus, the dryness of wetland plants at sample locality #4 provides a possible explanation for the increase in trace metal concentrations at this location.

The increased biomass of plants is known to enhance the phytoextraction of trace metals and other pollutants (Sheoran et al. 2016). Thus, the dense vegetation around the acid ponds (sample locality #11) may provide an additional explanation for why the acid ponds themselves do not record the highest dissolved trace metal concentrations in the study site. Dense vegetation around the acid ponds may also explain why dissolved Pb (Supplemental Fig. S1h) (which unlike metals such as Co, Ni, Cu, Zn, Cd, and Cr, forms insoluble sulfate complexes and, thus, was expected to increase due to low sulfate concentrations in the acid ponds) showed a decrease similar to Co, Ni, Cu, Zn, Cd, and Cr in the acid ponds.

Element (Al and Fe) Dilution Factors

Contamination Assessment Indices

Using median ± MAD concentrations as the basis for classification (Supplemental Table S2), sample localities #1, #7, #8, #9, #10, and #11 (for which Al and Fe dilution factors were calculated) were grouped according to Al and Fe dilution factors, respectively, into levels of low (< 0.05%), moderate (0.05–0.13%), and high (> 0.13%) levels of trace metal contamination according to Al dilution factor, and into low (< 0.10%), moderate (0.10–0.13%), and high (> 0.13%) levels of trace metal contamination according to Fe dilution factor. Thus, according to the median ± MAD concentrations of the Al dilution factors, sample localities #1 and #11 were classified as ‘low’ trace metal contamination, sample localities #7, #8 and #9 as ‘moderate’ trace metal contamination and sample locality #10 as ‘high’ trace metal contamination (Fig. 2a). Similarly, according to median ± MAD concentrations of the Fe dilution factors, sample localities #1 and #11 were classified as ‘low’ trace metal contamination, sample localities #7, #8 and #10 as ‘moderate’ trace metal contamination, and sample locality #9 as ‘high’ trace metal contamination (Fig. 2b). Additional dilution factor maps are provided in Supplemental Fig. S2.

Fig. 2
figure 2

Spatial distribution of dissolved (i) Co in μg/L along the Blesbokspruit River, classed according to (a) Al and (b) Fe dilution factors as low, moderate or high contamination

When the results of the classifications are compared with the HCA performed on dissolved trace metal concentrations (Fig. 3), samples classified by both the Fe and Al dilution factors as ‘low’ trace metal contamination correspond with cluster one, while samples classified as ‘moderate’ and ‘high’ trace metal contamination correspond with cluster two.

Fig. 3
figure 3

Dendrogram of cases (samples) depicting results of the HCA based on the average linkage (between groups) method and squared Euclidean distance of clr-transformed trace metal concentrations in stream water along the Blesbokspruit River

Correlation Analysis

Table 4 shows the Spearman’s rank correlations of aqueous trace metals with dilution factors, stream pH, and aqueous Fe and Al. The strengths of the correlations (r) were interpreted according to Dancey and Reidy (2017); thus, r = 1 indicates perfect correlation, r = 0.7–0.9 strong correlation, r = 0.4–0.6 moderate correlation, r = 0.1–0.3 weak correlation, and r = 0 no correlation. Strong positive correlations of the Al dilution factors with Co (r = 1.00, CI > 99%), Ni (r = 0.89, CI > 95%), Pb (r = 0.77, CI > 93%), Cr (r = 0.77, CI > 93%), Zn (r = 0.71, CI > 88%), and Cd (r = 0.71, CI > 88%) in Table 4, suggest the usefulness of Al dilution factors for detecting not only other lithophile metals (such as Cr), but also siderophile metals (such as Ni and Co) and chalcophile metals (such as Cd, Zn, and Pb). Similarly, the strong positive correlations of the Fe dilution factors with Cd (r = 1.00, CI > 99%), Ni (r = 0.94, CI > 99%), Co (r = 0.71, CI > 88%), and Zn (r = 0.71, CI > 88%) suggest the usefulness of Fe dilution factors for detecting not only siderophile metals (such as Ni and Co), but also chalcophile metals (such as Cd and Zn).

Table 4 Spearman’s rank correlation coefficient analysis of (i) Al and Fe dilution factors, (ii) stream pH and (iii) aqueous Al and Fe with Co, Ni, Cu, Zn, Pb, Cr and Cd

The relatively weak correlation between dissolved Cu and the Al dilution factors (r = 0.49) compared to the correlation between dissolved Cu and Fe dilution factors (r = 0.60) may be related to the influence of kaolinite (Table 1), which was fairly consistently associated with dissolved Cu at the study site (Supplemental Fig. S1f). González Costa et al. (2017) found that, compared to other clays, kaolinite has a high affinity for Cu. The increased affinity of kaolinite for Cu when coated with Fe-oxides (Osei and Singh 2000; Zhuang and Yu 2002) provides a possible explanation for the higher correlation observed between Cu and the Fe dilution factors (r = 0.60), compared to the Al dilution factors (r = 0.49).

Trace metal correlations with stream pH and aqueous Fe and Al are generally stronger than those with the dilution factors. However, compared to stream pH and aqueous Fe and Al, the dilution factors show stronger correlations with aqueous Co and Cd and comparable correlations with aqueous Ni. Strong correlations with Co, Cd, and Ni suggest that the dilution factors have a better potential for detecting mobile metals than relatively immobile metals such as Cu, Pb, and Cr (Covelo et al. 2008). Efficient detection and monitoring of Cd, in particular, is an important benefit of the ‘dilution factor’ method because of its mobility and hazardous effects on environmental and human health (Kicińska 2019, 2021).

Robustness of the Fe and Al Dilution Factors

Calculated Fe and Al dilution factors were moderate to strongly correlated with dissolved trace metals and showed CIs above 88% for all but Cu, which showed weaker correlation (Table 4). In addition, ‘low’, ‘moderate,’ and ‘high’ trace metal contamination groupings based on median ± MAD concentrations for both Fe and Al correspond well with sample groupings determined by the HCA based on clr-transformed dissolved trace metal concentrations (Fig. 3). This similarity in sample classifications suggests that the Fe and Al dilution factors and their median ± MAD statistics are fairly robust for the classification of AMD-related trace metal contamination in stream water affected by coal mining. However, the robustness of the dilution factors for detecting aqueous trace metal contents may be affected by increased rainfall during wetter seasons and by associated dilution and mobilization processes, as mentioned in the methodology. Thus, further investigation into the effects of seasonal change on the robustness of dilution factors for detecting aqueous trace metal contents is warranted.

Conclusions

The Blesbokspruit River was characterized by low pH, high EC, high sulfate, minor to moderately severe enrichment of trace metals, and ultra-high degrees of Fe contamination, all of which are strong indications that the river is affected by AMD. However, compared to previous studies, the quality of the Blesbokspruit River waters has significantly improved over the last 20 years, with metal concentrations showing a significant decrease, overall.

The spatial variability of trace metals was seemingly largely dependent on the pH of the river water, dissolved sulfate concentrations, and the health and density of the wetland and surrounding vegetation. Dissolved trace metal concentrations were generally inversely related to pH and positively related to dissolved sulfate concentrations. In addition, trace metal concentrations were seemingly elevated near sparse, dry vegetation and considerably more attenuated near dense, lush vegetation.

Al and Fe dilution factors showed moderate to strong positive correlations with dissolved Co, Ni, Zn, Pb, Cr, and Cd. The consistency between sample classifications based on the median ± MAD concentrations of the Fe and Al dilution factors and the HCA suggests that dilution factors are relatively robust and, thus, may serve as potential proxies for trace metal contamination in an AMD-affected stream during the dry season. Thus, it is recommended that future studies explore the use of Fe and Al dilution factors for detecting trace metal contamination in other AMD-affected streams with some additional considerations such as the influence of variable mineralogical contents in overbank sediments and climate and seasonality on the usefulness of Fe and Al dilution factors in the study area.