Discard Coal as Filter Bed Material in Horizontal Subsurface Flow Constructed Wetlands: a Preliminary Study

Constructed wetlands (CWs) are engineered systems that use the natural functions of vegetation, substrate and microorganisms to treat wastewater. In coal mining regions, low calorific coals are dumped as discard. Left unattended, discard and slurry ponds contaminate surface and groundwater, cause erosion and sedimentation of particulates into nearby rivers and dams and contribute to atmospheric pollution and landslides. This study sought to investigate the use of South African bituminous discard as filter bed material for CW. A laboratory-scale horizontal subsurface flow (HSF) CW was supplied either nutrient-poor tap water (TW) or nutrient-rich advanced facultative pond (AFP) effluent, and quality of the treated water monitored over 6 months. Additionally, residual material from the discard coal filter bed was assayed after 6 months to establish substrate stability and to assess the contribution of phyto-biodegradation. Results showed successful establishment of P. australis on discard coal, better plant performance (measured as PSII quantum yield and biomass accumulation) and greater nutrient removal when fed AFP effluent. Discard coal filter bed material had greater ash content, sustained fixed carbon and C/N ratio with unchanged electrical conductivity (EC) and sulphate and phosphate concentration, indicative of balanced ion exchange. This, along with a > 70% reduction in NH4+-N concentration, yielded a final effluent within the general limit set by the South African authority for either irrigation or discharge, into a water resource that is not a listed water resource, for volumes up to 2000 m3 on any given day.


Introduction
Coal remains South Africa's prime energy source with coal-fired power generation (8.74 × 10 −6 t) and gasification (4.88 × 10 −6 t) leading coal consumption locally. Raw coals are treated/washed to achieve a high calorific coal product for export (6.9 × 10 −6 t) and a middle calorific quality coal for local power and heat production (LHV ~ 18-30 MJ/kg). Lower calorific coals are dumped as discard although this is changing. Nowadays, washing and sorting have been optimized to higher efficiencies so that the resulting discard is of calorific values often lower than ~ 5-6 MJ/kg. Even so, as at 2001, discard coal and slurry were being produced at annual rates in excess of 4.2 × 10 −6 and 1.1 × 10 −6 t respectively (Lloyd, 2000;National Inventory Discard and Duff Coal, 2001). Thus, annual discard produced in South Africa increased from 4.36 × 10 −6 t in 1985 to 6.62 × 10 −6 t by 2001. The total land area covered by discard and slurry disposal at that time amounted to 4.011 × 10 7 m 2 , primarily located in the Emalahleni Coalfields of South Africa, which continues to increase annually.
If left unattended, discard dumps and slurry ponds contaminate surface and ground water by acid leachate and runoff, cause erosion and sedimentation of particulates into nearby rivers and dams and contribute to atmospheric pollution and landslides (Truter et al., 2009). While feasibility of electricity generation from discard coal has been considered (North et al., 2015), the better approach is to mitigate environmental degradation resulting from stockpiled discard coal and slurry by utilising this material in bioremediation (Mxinwa et al., 2020;Sekhohola-Dlamini et al., 2022).
For rehabilitation, discard dumps are typically limed, covered with a 30-100-cm layer of topsoil, fertilizer applied and the topsoil seeded with the appropriate grass species (Claassens et al., 2006;Cowan et al., 2016;Sekhohola & Cowan, 2017). The drawbacks of this technology are (1) topsoil if not readily available locally and must be sourced and transported sometimes over long distances to sites under rehabilitation and, at great cost; (2) areas of top soil excavation are prone to environmental damage including erosion and accumulation of particulates in water ways; and (3) borrow pits will also eventually require rehabilitation and/or restoration. For coal dump rehabilitation, even after successful re-vegetation, the approach does not necessarily result in breakdown of underlying carbonaceous material and indeed, may only mask and delay future environmental catastrophes (Cowan et al., 2016). Nevertheless, remediation of such large quantities of solid waste to mitigate environmental impact should occur in situ, be passive, and if possible, lead to formation of carbon rich soil-like material (Sekhohola & Cowan, 2017;Sekhohola-Dlamini et al., 2022). As stated by Sourkova et al. (2005), rehabilitation of opencast spoil and discard dumps should focus on two aspects: (1) transformation of the carbonaceous waste to a soil-like material through abiotic and biotic weathering and (2) successful (re-)vegetation.
In addition to the above, another approach with potential to mitigate environmental degradation resulting from stockpiled discard might be to consider tailings dams or slurry ponds as non-vegetated constructed wetlands (CWs) in need of nutrient-rich water and a suitable macrophyte to act as biocatalyst. A CW is an artificial wetland for the purpose of treating municipal or industrial wastewater, greywater or storm water runoff. It may also be designed to facilitate reclamation of land disturbed by mining, or as a strategy to mitigate loss of natural areas to development. As opposed to dryland bioremediation, treating discard coal dumps as CW could be a more economical and environmentally sound option as no topsoil excavation and transportation is necessary. Additionally, replacing gravel with discard coal as filter bed for CW would reduce capital costs associated with CW construction.
Constructed wetlands are engineered systems that use natural functions of vegetation, soil and microorganisms to treat wastewater (Vymazal, 2008;Parde et al., 2021;Tebitendwa & Cowan 2021). Although used as a sanitation technology, CW is not specifically designed for pathogen removal. Similar to natural wetlands, a CW acts as a biofilter and either reduces or removes pollutants including suspended solids, organic matter, nutrients (nitrogen and phosphorus) and heavy metals (Gorgoglione & Torretta, 2018). There are essentially two main types of CW: subsurface flow and surface flow CW (Tebitendwa & Cowan 2021) capable of treating a range of different waste streams (Vymazal et al., 2021). The planted vegetation or macrophyte plays an important role in contaminant removal and is potentially a source of renewable energy (Avellána & Gremillion, 2019) while the filter bed, consisting 1 3 Vol.: (0123456789) usually of sand and/or gravel, has an equally important role to play (Tebitendwa & Cowan, 2021). Here, we describe the outcome of a study conceived and carried out to determine the potential of South African waste bituminous coal (i.e., discard) for use as filter bed material in CW. The intention was to first ascertain whether discard can indeed support growth and proliferation of the wetland macrophyte Phragmites australis, second, establish the extent to which discard is leached and/ or decomposed when used as a CW filter bed material planted with Phragmites australis and third, determine the efficacy of a discard coal-Phragmites australis horizontal SF-CW for water reparation. Results are discussed in terms of the potential of discard coal-containing CW to mitigate pollution and the suitability of discard coal as an alternative treatment media to gravel in HSF CW for domestic wastewater treatment.

Constructed Wetland Configuration and Operation
A set of laboratory-scale horizontal subsurface flow (HSF) CW were established by packing plastic containers (0.4 × 0.3 × 0.25 m) with either discard coal or gravel to a depth of 0.2 m. A schematic illustrating the setup and process flow used in the configuration of the HSF CW is shown in Fig. 1. Waste bituminous coal, referred to as discard or gangue, typically comprised a mixture of roof coal and bituminous waste (calorific value of 8-10 MJ/kg) from the void after removal of high-grade coal, and was sourced from mines in the Emalahleni coalfields, Mpumalanga Province, South Africa. Gravel of particle size 1-8 mm and 8-14 mm was obtained from Amatola Quarry Products, Tempe Farm, Grahamstown, washed to eliminate fines and sorted. Bulk volume (V b ), pore volume (V p ) and porosity (n = V p / V b ) of the gravel and discard coal were derived for non-mechanistic sizing of the HSF CW (Tousignant et al., 1999;Vymazal, 2005). Filter beds were prepared by first packing particles of 1-8 mm followed by particles of 8-14 mm to depths of 8 and 12 cm respectively. The small-scale HSF CW was positioned outdoors under ambient conditions at the Belmont Valley Wastewater Treatment Works (WWTW), Grahamstown, South Africa (33° 19′ 07″ South, 26° 33′ 25″ East) and operated from November 2015 to July 2016.
One gravel-and one discard coal-containing HSF CW were supplied with nutrient poor tap water (TW). The remaining gravel-and discard coal-containing wetlands received nutrient rich water from an advanced facultative pond (AFP) attached to the Belmont Valley WWTW integrated algae pond system (IAPS). Details on the configuration and operation of this IAPS have been described elsewhere (Mambo et al., 2014;Jimoh & Cowan, 2017;Laubscher & Cowan, 2020;Titilawo et al., 2021). Tap and AFP water was from 20 L reservoirs fed continuously and positioned to gravity-feed each of the four HSF CWs at a rate of 43 L/day. Water within the HSF CW was maintained at 0.15 m below surface of the treatment bed by correct positioning of an outlet pipe. Into each of two gravel-and two discard coal-containing HSF CWs was planted similar-sized shoots of locally sourced Phragmites australis (Cav.) Trin. ex Steud. Fig. 1 Schematic of process flow and experimental set-up used to assess discard coal as a filter bed medium in HSF CW supplied with either TW or AFP effluent located at the Belmont Valley WWTW, Grahamstown, South Africa and the filter beds immediately wetted with the respective influent water and allowed to equilibrate following initiation of flow for a period of 10 days.

Macrophyte Growth and Productivity
Establishment, growth and productivity of Phragmites australis were assessed by determining shoot density, monitoring plant health by measurement of leaf chlorophyll fluorescence during the course of CW operation and by quantification of above ground biomass. Shoot density was determined by counting the number of shoots recruited in each experimental HSF CW monthly from November 2015 to July 2016 and is expressed per unit area. Chlorophyll fluorescence, measured as F v /F m , was analysed following dark phase adaptation (30 s) of at least 10 different leaves per treatment using a plant efficiency analyser (PEA, Hansatech Instruments Ltd., Norfolk, UK). Plant biomass was assessed after destructive sampling carried out at the end of July 2016. All aboveground plant material from each experimental CW was harvested, sun dried (1 week), and then oven-dried at 60 °C to a constant weight after which the mass was determined and productivity expressed as g dry mass per unit area of HSF CW.

Water Sampling and Analysis
Monitoring of water quality from laboratory-scale HSF CW commenced in March 2016 and was carried out monthly until July 2016. Sampling points were located at the outflow from each of the four reservoirs and at the point of exit from each of the four HSF CW. Flow rate into each CW was determined using the volumetric method (Talbot-Smith, 2013). Parameters measured on-site included pH using a model HI 8424 meter (Hanna Instruments Pty. Ltd., Bedfordview), electrical conductivity (EC; mS/m) and temperature (°C) using an Oakton EC Testr 11 conductivity meter (Eutech Instruments, Singapore).
For nutrient analysis, duplicate grab water samples were collected monthly from sampling ports positioned before and after each HSF CW. Nutrients monitored included NH 4 + -N, PO 4 3− -P and SO 4 2− , and analyses were either by standard method (APHA, 1998) or carried out according to the manufacturer's instructions using test kits purchased from Merck Chem. Co., Darmstadt, Germany. Where necessary, an AquaMate UV/VIS spectrophotometer (Thermo Spectronic, Cambridge, UK) was used to construct calibration curves for interpolation purposes. NO 3 − N and NO 2 − -N were not analysed since concentration of these nutrients in effluent from HSF CW is typically < 5 mg/L (Mburu et al., 2013), well below the general limit of 15 mg/L for South Africa (Department of Water Affairs, 2013). Furthermore, the nitrogen content of coal is low, typically ~ 1%, and the organic nitrogen exists as heterocyclic nitrogen, which is largely inaccessible to plants and microorganisms (Bates et al., 2011;Emmert et al., 2020).
For determination of Al, Fe, Mg and Cl, water samples were collected, cooled and shipped on the same day for analysis (Bemlab, Gant's Centre, Strand, South Africa), in a SANAS-accredited testing laboratory in accordance with ISO 17025:2005, for chemical analyses of soil, leaves, fruit and water, as well as for microbiology on water and fruit.

Proximate and Ultimate Analysis of Discard Coal
Proximate and ultimate analysis of discard coal and residual carbonaceous material from filter beds following conclusion of the experiment included determination of fixed carbon, moisture, volatile matter, ash content and elemental composition, and each parameter is expressed on a mass percentage basis. Moisture content was determined after exposure of discard (1 g) to 105 °C for 1 h and the relative loss of mass reported as percentage moisture. Ash content was determined after combustion of discard in a muffle furnace (Gallenkamp Model, Gallenkamp Muffle Furnace Co., London) at 815 °C . Heating was first by an increase in temperature to 400 °C in 30 min, then to 815 °C in 30 min and isothermal for 2 h before determination of mass of the residue which is reported as percentage ash. Volatile organic matter of discard was quantified after heating to 910 °C for 7 min and the change in mass reported as percentage volatile matter. Percentage fixed carbon of discard was calculated using the equation 100 − (% moisture + % volatile matter + % ash).

Statistical Analysis
Data were analysed using the statistics function in Sigma Plot version 11 (Systat Software Inc., San Jose, CA). Where necessary, results were analysed by one-way analysis of variance and significant differences between measurements for each treatment determined (Holm-Sidak method; P < 0.05). Unless otherwise stated, data are presented as the mean of at least three determinations ± standard error (SE).

Results and Discussion
In this study, we set out to investigate the potential of discard coal as treatment media for use in CW. Experiments were designed using a series of laboratory-scale HSF CW containing either discard coal or gravel as media.
3.1 Growth and Performance of P. australis Table 1 provides an overview of the response of the macrophyte P. australis in a HSF CW containing discard coal as filter bed material in comparison to a control system with gravel as filter bed material.
Recruitment of shoots and hence shoot density of the HSF CW was, as might be expected, significantly lower for plants supplied TW than for plants fed nutrient-rich AFP effluent. And, mean shoot density in the discard coal-containing HSF CW (124 ± 19.8 shoots/m 2 ) was significantly greater than that of the gravel-containing HSF CW (109 ± 17.3 shoots/m 2 ) as confirmed by oneway analysis of variance (0.01 at P < 0.05).
In HSF CW fed with AFP effluent, shoot density increased from 33 to 258 shoots/m 2 (with mean shoot density = 158 ± 25.6 shoots/m 2 ) and from 33 to 292 shoots/m 2 (with mean shoot density = 178 ± 28.9 shoots/m 2 ) in discard coal-and gravel-containing HSF CW respectively at conclusion of the experiment (Table 1). Surprisingly, plants in the gravelcontaining HSF CW registered higher mean shoot density than those of the discard coal-containing wetland suggesting that shoot recruitment and growth of P. australis were different (P = 0.02). However, where discard had been used as filter bed material, accumulation of aboveground biomass was enhanced (Table 1). And, relative to the CW-supplied TW, nutrient-rich AFP-water further increased accumulation of aboveground biomass by the macrophyte indicative of higher rates of productivity. Confirmation of the latter was obtained by measurement of chlorophyll fluorescence which showed that there was significant difference between mean F v /F m values for P. australis across the four treatments (Table 1).
Chlorophyll fluorescence, measured as F v /F m , is used to assess plant health. Typically, F v /F m indicates the maximum quantum yield of photosystem II (PSII) for all photochemically active reaction centres (relative units) when plants are exposed to light. Values close to one (i.e. ≤ 1) generally indicate healthy plants with robust photosynthetic competence and maximum quantum yield of PSII chemistry (Murchie & Lawson, 2013). Thus, for unstressed plants, leaf Table 1 Shoot recruitment, plant health and productivity of P. australis on filter beds of either gravel or discard coal in HSF CW fed either TW or AFP effluent. Growth of P. australis measured as shoots recruited, plant performance as chlorophyll fluorescence and productivity as dry matter accumulated (i.e. aboveground biomass harvested) were used to assess the effect of discard coal on macrophyte growth and performance. Data are from a single experiment conducted from November 2015 to July 2016, monitored monthly, and are the mean ± SE (n = 9). Values followed by different letters are significantly different (P ≤ 0.05). AFP, advanced facultative pond; TW, tap water F v /F m is decidedly consistent and values of ~ 0.8 correlate to maximum quantum yield of photosynthesis (Demmig & Björkman, 1987). In contrast, any type of 'stress' that results in inactivation or damage to PSII (e.g. photoinhibition) or induction of sustained quenching (Demmig-Adams and Adams III, 2006) will show proportionately lower leaf F v /F m values after an appropriate period of dark adaptation and be reflected as lower plant productivity, i.e. lower photosynthetic competence, reduced carbon assimilation and lower biomass accumulation.
As seen from the data in Table 1, the most pronounced difference in P. australis leaf F v /F m was between TW and AFP effluent-fed discard coal-containing CW indicating nutrient load as a major factor in determining growth and dry biomass acquisition by this macrophyte. Second, although quantum yield of photosystem II (PSII) for P. australis in a gravel medium was similar irrespective of nutrient load, plants accumulated less biomass presumably due to insufficient nutrient. Even so, a decline in leaf F v /F m for P. australis in the TW-fed discard coal-containing CW may indicate either substrate instability or phytobiodegradation or a combination of both.

Stability of Discard Coal as HSF CW Filter Bed Material
To establish the stability of discard coal as HSF CW filter bed material, and in an effort to assess the contribution of phyto-biodegradation during prolonged operation of the HSF CW and in the presence of P. australis, residual carbonaceous material was subjected to proximate and ultimate analysis, and the results are presented in Table 2. Analysis of the raw discard revealed a material with characteristics similar to those of bituminous coal (Given, 1984;O'Keefe et al., 2013;Coppola et al., 2015). And, with only ~ 55% fixed carbon, an ash content well above 20% and elemental C under 70%, the coal used in this study most closely resembled sub-bituminous discard material (Table 2). During operation of the HSF CW, nutrient-poor TW caused a decline in fixed carbon content of the discard filter bed, presumably due to leaching and/or bio-conversion, to yield a residue with lower moisture retention and increased volatile matter and ash content, the latter most likely due to deposition of plant and microbial material. By comparison, the HSF CW Table 2 Proximate and ultimate analysis of discard coal from filter beds before and after operation of laboratory-scale HSF CW fed either TW or AFP effluent. Composite samples were prepared from raw discard coal prior to initiation of the experiment and from the carbonaceous material remaining in the filter bed of the HSF CW after cessation of operation. Subsamples were analysed as described in "Materials and Meth-ods" and the mass fraction for each parameter determined, multiplied by 100 and presented as mass percentage. Data are from a single experiment conducted from November 2015 to July 2016 and are the mean ± SE (n = 4). Values in rows followed by different letters are significantly different (P ≤ 0.05). AFP, advanced facultative pond; n.s., not significant; TW, tap water fed AFP water showed a similar but less pronounced increase in ash content and volatile matter, reflective of the more nutrient-dense AFP effluent feed and, likely, greater deposition of plant and microbial matter (Wang et al., 2022). Ultimate analysis of the carbonaceous residue from horizontal filter beds containing discard coal confirmed that diminution of elemental C and N was significant (P = 0.001) where nutrientpoor TW had been used as feed (Table 2). Reduced C and N along with limiting P content of the carbonaceous residue from TW-fed discard may also explain the apparent increase in shoot recruitment observed, viz. 124 shoots/m 2 versus 109 shoots/m 2 (see Table 1), and enforce potential of discard as a material suitable for supporting P. australis in filter beds of HSF CW. Taken together, the results support the idea that waste coal (also referred to as discard or gangue) may have potential for use as filler for filter beds in HSF CW to support macrophyte growth, shoot recruitment, nutrient abstraction and wastewater treatment assuming that the wastewater to be treated contains sufficient nutrient. The results however contrast with those from a similar recent study which indicated that manganese ore and iron ore-based CW appear better and are relatively more stable in terms of treatment performance for N and P (Wang et al., 2022). Substrates used as filter bed material in CW play a major role in facilitating pollutant removal from wastewater. While inorganic media such as sand, zeolite and gravel has been the norm Kadlec & Wallace, 2009), in recent years, support for use of readily-available, low-cost materials as alternative treatment media for CW has increased. Examples include construction waste, rice straw, bamboo rings, dewatered alum, palm tree mulch and sludge, and all serve to improve the sustainability of this wastewater treatment technology by reducing import of natural sand and gravel (Kadlec & Wallace, 2009;Mateus & Pinho, 2020;Hamada et al., 2021). Results from the present work show that establishment of P. australis on coal discard fed AFP effluent was better and macrophyte growth superior, including quantum yield of photosynthesis and accumulation of dry biomass, in comparison to plants on a gravel-containing CW fed with AFP effluent or the gravel and coal discard HSF CW fed TW. Table 3 provides a summary of water quality data from operation of the four laboratory-scale HSF CWs fed either TW or AFP water.

Water Quality
For both gravel-and discard coal-containing HSF CWs fed TW, values for the water quality parameters monitored (i.e., pH, EC, DO, NH 4 + -N, PO 4 3− -P and Fe) were within the general limit set by the South African authority for either irrigation or discharge into a water resource that is not a listed water resource for volumes up to 2000 m 3 /day (Department of Water Affairs, 2013). Table 3 Summary data on effluent quality and pollutant removal efficiency from laboratory-scale HSF CW filter beds comprising either gravel or discard coal and fed either TW or AFP effluent. Where specified, values in rows followed by dif-ferent letters are significantly different (P ≤ 0.05). Analysis of final effluent quality was per month from May to July. AFP, advanced facultative pond; n.d., not detectable; TW, tap water Values are the mean ± SE. For discard coal and gravel CW fed TW: n = 5, 4, 5, 5, 5, 5, 5, 5, 4 and 5 and for discard coal and gravel CW fed AFP effluent: n = 5, 5,4,5,5,4,4,5,4  On the other hand, in gravel and discard coalcontaining CW fed with AFP water, all parameters (i.e. pH, EC, DO, PO 4 3− -P and Fe) except NH 4 + -N were within the national effluent discharge standards. Even so, NH 4 + -N concentration was reduced by > 70% in effluent from the AFP-fed discard coal HSF CW. And, NH 4 + -N concentration, recorded over the course of the 6-month experiment, did not differ significantly from the 6 mg/L limit for discharge of treated water from a WWTW (i.e. P = 0.320 and P = 0.475, for gravel-and discard coal-containing CWs respectively).
The reduction in PO 4 3− -P and NH 4 + -N concentration in AFP effluent after treatment by the discard coal-containing HSF CW was surprising in the present study since it was expected that the presence of P and N in discard coal (Table 2) would result in an increase in these in the treated water. Although the main mechanisms of P removal from subsurface flow (SF) CW are believed to be adsorption and precipitation (Kadlec & Wallace, 2009), several studies have also reported an increase in PO 4 3− -P concentration in treated water (Mburu et al., 2013), which is attributed to release of adsorbed PO 4 3− -P back into the water column due to the anaerobic conditions that prevail in SSF CW (Cramlet & Turyk, 2002). The general reduction of PO 4 3− -P from discard coal-containing HSF CW contrary to an increase reported in the literature could be attributed to (i) the high planting density of P. australis of ~ 33 plants/m 2 compared to a recommendation of between 4 and 14 plants/m 2 (UN-Habitat, 2013;US EPA, 2000). While nutrient removal is considered insignificant in SSF CW , the high planting density adopted in the present study presumably increased removal of PO 4 3− -P by offering a larger root surface area for adsorption of P. Also, the presence of a myriad of metals (Table 2) in discard coal could have facilitated precipitation of PO 4 3− -P and its retention in the CW. Interestingly, the NH 4 + -N concentration obtained from discard coal-containing HSF CW fed with AFP effluent was significantly lower than values reported for HSF CW that employed gravel as a treatment media (Table 3; Okurut, 2000;Mburu et al., 2013;Almuktar et al., 2018). The major pathways responsible for nitrogen removal from SF CW are nitrification and denitrification (Lee et al., 2009;Ma et al., 2021). Nevertheless, deficiency of a carbon source in domestic wastewater is believed to limit these processes and result in poor nitrogen removal from SF CW Lee et al., 2009). The presence of > 60% elemental carbon in discard coal (Table 2) suggests that it is a better treatment media than gravel for providing the carbon required for nitrification and denitrification processes and hence a major contributing factor to high NH 4 + -N removal recorded from discard coal-containing HSF CW fed with AFP effluent. Moreover, we recommend that metagenomic analyses be carried out across the horizontal subsurface to monitor microbial colonisation and population dynamic within the discard coal-HSF CW microbiome in an effort to elucidate the various bacterial species that play a role in the nitrification and denitrification processes. Furthermore, since lignite, sub-bituminous and bituminous coals are hydrophobic and negatively charged (Ye et al., 1989;Maršálek, 2009), retention of nitrate and other oxyanions by electrostatic interaction is expected to be very low. Indeed, Nunes et al. (2020) needed to chemically modify the surface of 'coal beneficiation tailings' using hexadecyltrimethylammonium bromide to derive a solid sorbent for treatment of nitrate-contaminated wastewater and achieved a removal efficiency of 93%.
As outlined in Table 3, discard coal-containing HSF CW failed to reduce SO 4 2− and metals (Fe and Mg) which was expected due to leaching of these minerals from filter bed material. Nonetheless, mean SO 4 2− and metal ion concentration recorded in this study after treatment of TW and AFP effluent by a discard coal-containing HSF CW was lower than that reported for coalmine water (Sigh, 1988;Annandale et al., 2009). This suggests a role for plant uptake of SO 4 2− and metals as a major removal mechanism by HSF CW. In fact, various studies report that wetland macrophytes absorb and accumulate metals in tissue (Matagi et al., 1998;Basile et al., 2012;Matache et al., 2013), hence the potential of HSF CW to maintain concentrations of these metal ions at levels below those reported for coalmine water.

Conclusion
Performance of the concept pilot-scale HSF CW with discard coal as filter bed material fed AFP effluent revealed, in comparison to gravel-containing CW, minimal leaching of the substrate, successful establishment of P. australis, better plant performance (measured as the quantum yield of PSII and accumulation of biomass) and greater nutrient removal. After operation for 6 months, analysis of filter bed material from the CW revealed greater ash content with little change in fixed carbon in coal-containing HSF CW-fed AFP effluent. Additionally, elemental analysis revealed maintenance of the C/N ratio of discard coal substrate in the AFP-fed HSF CW at the end of the experiment but more importantly little or no change in EC and the concentration of either sulphate or phosphate indicating balanced ion exchange. This, along with a > 70% reduction in NH 4 + -N concentration, yielded a final effluent within the general limit set by the South African authority for either irrigation or discharge, into a water resource that is not a listed water resource, for volumes up to 2000 m 3 on any given day.
Funding Open access funding provided by Rhodes University. The study was financially supported by Rhodes University via an annual grant to EBRU. Professor A. K. Cowan was supported by a grant from the National Research Foundation, South Africa (IFR1202220169) Data Availability Datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.
Code Availability Not applicable.

Declarations
Ethics Approval Not applicable.

Conflict of Interest
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.