Optimising Nature-Based Treatment Systems for Management of Mine Water

Abstract

Deployment of nature-based systems for mine water treatment is constrained by system size, and the evidence suggests decreasing hydraulic conductivity (K_sat) of organic substrates over time compromises performance. In lab-scale continuous-flow reactors, we investigated (1) the geochemical and hydraulic performance of organic substrates used in nature-based systems for metals removal (via bacterial sulfate reduction) from mine water, and then (2) the potential to operate systems modestly contaminated with Zn (0.5 mg/L) at reduced hydraulic residence times (HRTs). Bioreactors containing limestone, straw, and wood chips, with and without compost and/or sewage sludge all achieved 88%–90% Zn removal, but those without compost/sludge had higher K_sat (929–1546 m/d). Using a high K_sat substrate, decreasing the HRT from 15 to 9 h had no impact on Zn removal (92.5% to 97.5%). Although the sulfate reduction rate decreased at a shorter HRT, microbial analysis showed high relative abundance (2%–7%) of sulfate reducing bacteria, and geochemical modelling pointed to ZnS(s) precipitation as the main attenuation mechanism (mean ZnS saturation index = 3.91–4.23). High permeability organic substrate treatment systems operated at a short HRT may offer potential for wider deployment of such systems, but pilot-scale testing under ambient environmental conditions is advisable.

1. Introduction

Mine discharges containing elevated concentrations of metals (e.g., Zn, Pb, Cd) and sulfate are of global environmental concern due to the significant risks they pose to receiving watercourses [1,2,3]. In the UK, some 338 such discharges have been identified in England and Wales [4], contributing over 250 t/annum of Zn to freshwater [5]. There is a large body of literature on mine water discharges that are acidic (acid mine drainage, AMD) due to the oxidative dissolution of the disulfide mineral pyrite (FeS₂) [1,2,6]. However, most metals in freshwater of the UK arise from abandoned mines in which the oxidative dissolution of monosulfide minerals (e.g., ZnS, PbS) predominates, whilst pyrite oxidation is limited. Consequently, many UK metal mine discharges are circumneutral, and contain high concentrations of metals, such as Zn and Pb, but low concentrations of Fe [7]. Zn is the most common contaminant metal encountered and is one of six target substances (the others are As, Cd, Cu, Pb, Ni) of the Environmental Targets (Water) (England) Regulations 2023. These new regulations require that the length of rivers and estuaries in England polluted by such substances from abandoned metal mines is reduced by at least 50% by 31 December 2038 [8]. Although most UK mine water is only modestly contaminated with Zn (over 30% have concentrations ≤ 0.1 mg/L) (Figure S1), many receiving watercourses still breach freshwater environmental quality standards (EQS) (10.9 µg/L bioavailable for Zn), which is important given the strong association of Zn with freshwater macroinvertebrate biodiversity [9]. Zn is therefore the main target metal for remediation.

Nature-based (passive) systems are the favoured approach to metal mine drainage management. Horizontal- and vertical-flow wetlands have both found application to mine water treatment, operating under aerobic (usually surface flow) and anaerobic (subsurface flow) conditions [2,6], but current UK practice is to use systems that harness bacterial sulfate reduction (BSR) [10,11,12,13,14]. Under anaerobic conditions, sulfate reducing bacteria (SRB) reduce sulfate (reaction (1)) and immobilise metals as their sulfides (reaction (2)) [15]. The anaerobic conditions required for BSR are achieved by passing mine water through an organic substrate, typically compost.

$$2\mathrm{C}{\mathrm{H}}_2\mathrm{O}+\mathrm{S}{\mathrm{O}}_4^{2-}\to {\mathrm{H}}_2\mathrm{S}+2\mathrm{H}\mathrm{C}{\mathrm{O}}_3^{-}$$

(1)

$${\mathrm{H}}_2\mathrm{S}+{\mathrm{Z}\mathrm{n}}^{2+}+2\mathrm{H}\mathrm{C}{\mathrm{O}}_3^{-}\to \mathrm{Z}\mathrm{n}\mathrm{S}\left(\mathrm{s}\right)+2{\mathrm{H}}_2\mathrm{O}+2\mathrm{C}{\mathrm{O}}_2$$

(2)

A major constraint to deployment of such nature-based treatment systems is size, particularly in the UK, where land availability is limited due to the upland locations in which most metal mine discharges occur [16]. Typical hydraulic residence times (HRTs) on the order of days [12,13,17] are therefore not feasible in the UK, but recent research has demonstrated successful metal removal in much shorter HRTs; less than 24 h [10,11,18,19].

Evidence from the UK’s first full-scale nature-based treatment system for metal mine drainage suggests that hydraulic conductivity may also be a major constraint to system longevity [20]. A concurrent decline in treatment efficiency with a rise in water level in the compost bioreactors after around 1.5 years of operation indicated a reduction in substrate permeability, resulting in hydraulic short-circuiting [20]. Other authors have similarly noted reduced permeability in organic substrates, such as those containing compost, manure, and wood waste, over time [2,21]. However, whilst many authors have investigated the geochemistry and microbiology of passive bioreactors, few have specifically studied their hydraulic performance. In those studies where saturated hydraulic conductivity (K_sat) has been measured, a decrease over time was noted [17,18,22] due to factors such as substrate compaction and clogging of pore spaces with metal precipitates and biological growths. The HRTs in these systems ranged from 0.5 days to 10 days. Substrates containing fine sand [18] and manure [22,23] were found to have the lowest permeabilities whilst mixtures containing straw and wood shavings had higher permeabilities [22]. Larger particles, such as gravel, coarse sand, walnut shells, and wood chips, are commonly added to organic substrates to minimise compaction and to maintain permeability [2].

The choice of reactive substrate used in nature-based treatment systems for the remediation of metal mine drainage must therefore consider both the biogeochemical performance, to achieve the desired effluent water quality, and the potential for the deterioration of hydraulic properties affecting long-term operation. For short HRT systems, this is especially important, since hydraulic and contaminant loading rates to such systems are higher, and the rates of favourable biogeochemical reactions must be maximised. A hydrogeochemical investigation has therefore been undertaken to specifically test, for the first time, the geochemical, microbial, and hydraulic performance of a substrate selected for its high permeability for treatment of modestly contaminated mine water in a short HRT. A key element of the research was to ascertain whether low concentrations of Zn, typical of many UK mine water, can be attenuated in short HRT sulfate reducing bioreactors, facilitating the remediation of mine drainage at locations with limited land availability. Whilst Zn was the main focus of this research, metal monosulfides are thought to precipitate sequentially according to their solubility product, and the lower solubility products of the sulfides of other metals, such as Cu, Cd, and Pb, means that these metals will likely form a sulfide more readily than Zn [16], and would therefore be removed to an equal or greater extent.

This study aims to (1) evaluate the performance of different substrate mixes in terms of Zn removal, sulfate reduction, and hydraulics, and (2) investigate the potential to utilise short HRT systems containing a substrate mix with good hydraulic and geochemical properties for the treatment of modestly contaminated metal mine drainage.

2. Materials and Methods

2.1. Experimental Configuration

Three sets of laboratory-scale, continuous-flow bioreactors (internal diameter 105 mm, length 500 mm), each containing a different substrate mix (

Table 1. Substrate mixes used in continuous-flow laboratory bioreactors to investigate hydraulic and geochemical performance, including percentage volume of each component.

Substrate	Limestone Gravel	Wood Chips	Straw	PAS 100 Compost	Municipal Sewage Sludge
L-W-C-SS	50%	25%	0%	15%	10%
L-W-S-SS	50%	25%	15%	0%	10%
L-W-S	50%	25%	25%	0%	0%

), were operated in duplicate. The components of each substrate mix were not rigorously specified but comprised a combination of locally sourced materials (Table S1) that were chosen, following a literature review [24], for their potential geochemical, microbial, and hydraulic performance in treating metal mine water. Limestone gravel was included in all substrate mixes at 50% v/v to add structure and to aid permeability. The L-W-C-SS substrate mix was based on previous pilot-scale research [9], whilst the proportions of PAS 100 compost and sewage sludge were incrementally reduced in the others to exclude them entirely in the L-W-S substrate mix, due to their potential impact on permeability [2,21]. Limestone gravel (4–10 mm diameter) was placed at the base of each bioreactor, to a depth of approximately 30 mm, and was overlain by alternating layers of limestone and substrate mix (depth 400 mm). Each substrate mix, except for limestone, was thoroughly mixed before placement in the bioreactors. The bioreactors were configured such that synthetic mine water entered at the top of each bioreactor and passed downwards, by gravity, through the reactive media (Figure S2). Effluent water was discharged by variable height outlet pipes which ensured that the substrate remained saturated with a 40 mm cover of water above the substrate. Prior to bioreactor start-up, the substrates were saturated to enable the calculation of porosity (L-W-C-SS = 0.46; L-W-S-SS = 0.50; L-W-S = 0.56). A Watson Marlow 300 series peristaltic pump was used to transfer mine water to the bioreactors, with mean flow-rate ranging from 1.89 mL/min to 2.00 mL/min, which equated to a mean residence time of 13 to 17 h.

Following selection of the best substrate mix, in terms of Zn removal, sulfate reduction, and hydraulics (L-W-S, Table 1), an additional set of bioreactors was operated in triplicate to evaluate the potential to utilise an organic substrate with good hydraulic properties in short HRT systems. The bioreactors were configured to be identical to those described above. Mean initial porosity in the bioreactors was 0.55.

2.2. Bioreactor Operation

The bioreactors to evaluate the performance of different substate mixes were operated for 199 days. Synthetic mine water was produced by dissolving laboratory grade salts in deionised water (mean 0.5 mg/L Zn, 22 mg/L sulfate, pH 7.6). This water quality was based on the water from a mine in Northern England (known as Park Level; Table S2) that is typical of the water discharging from the mine that is contaminated, not the mine itself in the UK.

After 28 days of operation, constant head permeameter tests were conducted on the 3 sets of bioreactors, together with a bioreactor containing limestone only as a pseudo-control, to determine the K_sat of each of the substrates. Hydraulic testing was repeated at the end of the trial to assess any changes in the K_sat during operation.

The addition of liquid carbon to enhance sulfate reduction rates was trialled for a period of 39 days. Propionic acid (13.4 M) addition to each bioreactor, previously shown to improve performance in laboratory-scale bioreactors receiving a high Zn concentration [11], commenced on day 143 at a rate of 1 mL per 40 L of influent water, and stopped on day 182.

The additional set of bioreactors, to assess the longer-term performance of a substrate with good hydraulic and geochemical properties (L-W-S, Table 1) at a reduced HRT, was operated for 324 days. Synthetic mine water (Table S2) was initially passed through at a mean flow-rate of 1.81 mL/min, equating to an HRT of 15 h. Flow-rate was incrementally increased to a mean of 2.27 mL/min (HRT 12 h), after 119 days, and a mean of 2.97 mL/min (HRT 9 h), after 253 days, to evaluate performance at shorter HRTs.

Constant head permeameter tests were conducted on the additional set of bioreactors at the beginning and end of the trial to determine the K_sat of the substrate. Due to uncertainties over the effect of permeability tests on treatment performance, initial constant head permeameter tests were conducted on a separate triplicate set of bioreactors, containing the identical substrate, set up for the sole purpose of testing the K_sat. Hydraulic testing of the long-term bioreactors was carried out at the end of the trial to assess any changes in the K_sat during operation.

2.3. Water Sampling and Analysis

Three 30 mL samples were collected fortnightly in polypropylene bottles from the influent mine water and the effluent of each bioreactor for total and filtered cation analysis and filtered anion analysis. Filtration was undertaken using 0.45 µm cellulose acetate filters (Sartorius Stedim Biotech GmbH, Göttingen, Germany). Samples for cation analysis were acidified with 1% v/v concentrated nitric acid. All samples were stored at 4 °C prior to analysis. Cation analysis was performed using a Vista-MPX Inductively Coupled Plasma-Optical Emission Spectrometer (ICP-OES) (Varian, Mulgrave, Australia). Anion concentrations were determined using a DX320 Ion Chromatograph (IC) (Dionex Corporation, Sunnyvale, CA, USA). Measurements of water temperature, pH, oxidation-reduction potential (ORP), and electrical conductivity were recorded using a pre-calibrated Myron L 6P Ultrameter (Myron L Company, Carlsbad, CA, USA). Total alkalinity was determined using a digital titrator (Hach Lange GmbH, Düsseldorf, Germany) with 1.6 N sulfuric acid and bromcresol-green methyl-red indicator. Flow-rate was calculated at the time of sample collection by measuring the volume of sample collected over a specified time period.

2.4. Substrate Sampling and Analysis

Substrate samples were collected at the end of the trial from the bioreactors operating at a reduced HRT. Four samples were collected from each bioreactor at approximate depths (from top of the substrate) of 30 mm, 130 mm, 270 mm, and 380 mm. Samples for geochemical analysis were collected in pre-washed (analytical grade nitric acid, 10% v/v) polypropylene bottles, and samples for microbial analysis were collected in 50 mL sterile centrifuge tubes. Both sets of samples were filled with water from within the bioreactors and stored at minus 20 °C prior to analysis. A sample of the original substrate mix was subjected to the same geochemical and microbial analysis.

Samples for geochemical analysis were air-dried and passed through a 2 mm sieve. A 10 g subsample of the less than 2 mm fraction of each sample was then ground to a fine powder (<250 µm) using a pestle and mortar and subjected to aqua regia digestion, according to BS 7755-3.9 [25], followed by ICP-OES analysis.

Microbial community analysis consisted of DNA extraction from all samples (FastDNA™ Spin Kit for Soil, MP Biomedicals, Irvine, CA, USA). The 16S rRNA genes were amplified from total DNA and the V4 variable region was sequenced following the 16S Illumina Amplicon Protocol from the Earth Microbiome Project [26] at NU-OMICS (Northumbria University, Newcastle upon Tyne, UK). Raw sequence data (FASTQ files) obtained from the Illumina sequencing platform were demultiplexed and analysed using QIIME 2 [27]. DADA2 was used for amplicon sequence variant (ASV) selection [28]. ASVs which appeared in the negative control were removed from the other samples. The 16S rRNA gene is a universal phylogenetic marker present in all microorganisms, and sequencing of this gene provides information on the composition of the microbial communities present within each sample. Comparative analyses and statistical testing of microbial community data were performed in STAMP [29].

2.5. Hydraulics Performance Testing

Constant head permeameter tests (Figure S3) were conducted to determine the K_sat of each substrate at the beginning and end of the trials. Tap water flowed through each bioreactor (permeameter), driven by a constant head of water (h) in an adjacent reservoir. The flow-rate (Q) from the bioreactor was calculated by measuring the volume of water over a specified time, and the K_sat was calculated using Darcy’s law (Equation (3)), where A is the cross-sectional area of the substrate and L is the length of the substrate (Figure S3).

$$Q=-A{K}_{sat}{\displaystyle \frac{dh}{dL}}$$

(3)

A Watson Marlow 300 series peristaltic pump (Watson-Marlow Limited, Falmouth, UK) was used to transfer tap water to the reservoir which, in combination with an overflow outlet, ensured that the hydraulic head difference remained constant over time. The tests were operated at 4 hydraulic head values (0.075 m, 0.125 m, 0.175 m, 0.225 m) with a lab jack used to vary the hydraulic head. The mean of 3 flow-rates, measured at 5 min intervals, was used to calculate the K_sat at each hydraulic head. Constant head permeameter tests were also undertaken, at the beginning of the trials and at the end of the trial, to evaluate the performance of different substate mixes on a control bioreactor containing limestone only. It was not possible to conduct an equivalent test on the control bioreactor at the end of the trial assessing bioreactor performance at a reduced HRT due to COVID-19 restrictions.

3. Results and Discussion

3.1. Effectiveness of Different Substrate Mixes

3.1.1. Zinc and Sulfate Removal

There was no significant difference in mean effluent Zn concentration between the different substrate mixes (Mann–Whitney U test: p > 0.05) and all three were effective in their removal of Zn (mean removal efficiencies 88 to 90% for total Zn (Figure 1A) and 92 to 93% for filtered Zn). Effluent Zn concentrations remained relatively stable throughout the trial, except for a slight rise evident in all bioreactors immediately after disturbance of the substrates by hydraulic testing (day 28), and again during liquid carbon addition (days 143 to 182) (Figure 1A). It is unclear why propionic acid addition had a negative impact on Zn removal (reduction in mean removal efficiency from 91% to 83%) as it has previously been shown to successfully enhance Zn removal in similar laboratory-scale bioreactors receiving a much higher Zn concentration (45 mg/L) [11]. Given the short-term duration of the propionic acid addition, no firm conclusions can be drawn from this.

Sulfate reduction varied between the substrate mixes. In the bioreactors containing limestone, wood chips, and straw (L-W-S, Figure 1B), sulfate reduction took 10 days to become established, but mean percentage sulfate reduction (defined as the difference between influent and effluent sulfate concentrations) remained consistently above 30% for the rest of the trial. Short-lived increases in sulfate reduction were attributed to periods when the flow-rate was temporarily reduced for operational reasons (days 116 to 122), and to propionic acid addition (days 143 to 182). Minor differences in sulfate reduction between the replicates, particularly in the early stages of the trial, were likely due to heterogeneities in the physical and chemical properties of the substrates which resulted in a longer period for sulfate reduction to become fully established in one of the replicates (L-W-S (B)). Nevertheless, overall, there was no significant difference in sulfate reduction between the replicates (Mann–Whitney U test: p > 0.05), thus demonstrating good reproducibility. The consistent removal of both Zn and sulfate within these bioreactors shows that, despite the lack of activated sludge to provide an initial source of available carbon for SRB metabolism, the establishment of the reducing conditions (Figure S4) provided an optimum environment for SRB activity [30], and favoured Zn removal as its sulfide. These findings are consistent with those of other studies that have used substrates containing straw to successfully attenuate metals via BSR, albeit sulfate reduction has proven to be limited within straw-based substrates compared to compost- and manure-based substrates [31,32].

Conversely, despite initially high rates of sulfate reduction (mean 95.6%) in the bioreactors containing PAS 100 compost and/or sewage sludge (L-W-C-SS, L-W-S-SS; Figure S5), rates steadily declined following hydraulic testing. Temporary increases in sulfate reduction again occurred when the flow-rate was reduced (days 116 to 122), and during propionic acid addition (days 143 to 182), but were not sustained. It is likely that the higher flow-rates during the hydraulic testing disturbed the SRB or removed their available carbon source in the fine material washed from the bioreactors. By contrast, sulfate reduction in the bioreactors containing wood chips and straw was unaffected by hydraulic testing, likely due to the lack of fine material to be washed out. Others have also observed a decline in sulfate reduction following an increase in flow-rate in compost-based bioreactors [13], attributed to the washing out of the SRB from the reactive mixture. A simultaneous rise in the Eh following hydraulic testing, due to the increased loading rate of oxygen [13], was short-lived in the bioreactors containing wood chips and straw but led to the establishment of oxidising conditions in the bioreactors containing compost and/or sewage sludge (Figure S4). Nevertheless, a decrease still occurred between the influent (mean 155 mV) and the effluent (mean 90.6 mV in L-W-C-SS, 84.0 mV in L-W-S-SS) Eh from day 79 onwards. Whilst strongly anaerobic conditions, as observed by others [10,16], did not appear to become established in any of the bioreactors, the effluent Eh values reported here likely overestimated the actual Eh values within the substrate due to the re-establishment of oxidising conditions during the extended period of sample collection resulting from the low flow-rates of the bioreactors [11]. The effluent pH (6.8 to 8.1) remained favourable for SRB activity [30] throughout the trial (Figure S4). Whilst sufficient sulfate reduction occurred to attenuate the low concentration of Zn present within the influent water as its sulfide, it is likely that additional processes (e.g., sorption) were taking place alongside BSR, as similarly reported by others [33,34,35].

3.1.2. Hydraulic Performance

Hydraulic testing was undertaken at varying hydraulic heads (0.075 m to 0.225 m) between days 28 and 32 to determine the initial K_sat of each substrate mix and was repeated at the end of the trial to evaluate changes in the K_sat during operation of the bioreactors. The same procedure was also carried out on a control bioreactor containing limestone only. As observed in other studies that applied constant head permeameter tests to highly permeable media [36], the K_sat decreased with hydraulic head due to flow resistance within the tubing connecting the permeameter to the bioreactor (Figure 2). This resulted in an underestimation of the K_sat, which increased non-linearly with hydraulic gradient [36]. Nevertheless, differences in the initial K_sat were apparent between the substrate mixes, with a greater proportion of finer-grained compost and sewage sludge resulting in an overall reduced K_sat (mean 506–804 m/d for L-W-C-SS compared to mean 929–1546 m/d for L-W-S, Figure 2). The variation in the K_sat between the substrate mixes is related to their physical properties, with the substrate mix containing straw and wood chips (L-W-S) having a coarser grain size (Table S1), together with a higher porosity (mean 0.56) than those containing compost and/or sewage sludge (mean porosity 0.50 (L-W-S-SS) and 0.46 (L-W-C-SS)). It is also likely that some degree of compaction occurred when the substrates were initially saturated, with greater compaction in those substrates containing the finer-grained compost and/or sewage sludge. These findings are consistent with those reported in the literature in which even small proportions of fine-grained material, such as manure and compost, lowered the bulk K_sat [22,23]. However, the values for the K_sat are around an order of magnitude higher than those previously reported for organic substrate mixes [14,17,18] due to the high proportion of limestone and wood chips present.

Over the duration of the trial, the K_sat declined to a similar degree in all three substrate mixes (mean decrease 4.8% to 8.8%, Table S3), but the same pattern was observed whereby greater proportions of compost and sewage sludge led to a lower K_sat (Figure 2). Other studies have similarly reported decreases in the K_sat during bioreactor operation [14,17,18,21], attributed to substrate compaction and clogging of pore spaces with metal precipitates and biological growths. By contrast, the K_sat in the control bioreactor reduced only slightly (<1%). Whilst none of the substrate mixes demonstrated a significant deterioration in the K_sat over time, the bioreactors containing straw and wood chips (L-W-S) had a higher K_sat at the beginning and end of the trial, due to their higher porosity and coarser grain size, and were also effective in their removal of Zn and sulfate. This substrate was therefore taken forward for longer term testing of its potential for the treatment of modestly contaminated metal mine drainage in short HRT systems.

3.2. Effectiveness of Zinc and Sulfate Removal at Reduced Hydraulic Residence Time

3.2.1. Zinc and Sulfate Removal

Bioreactors containing limestone, wood chips, and straw remained effective in their removal of a low concentration of Zn (mean 0.5 mg/L) when the HRT was incrementally reduced from 15 h to 9 h (Figure 3A). There was no significant difference in effluent Zn concentration over the duration of the trial (Mann–Whitney U test: p > 0.05) and mean total Zn removal efficiency varied from 92.5% to 97.5% (filtered Zn 92.2% to 97.8%). A more informative metric to evaluate system performance, particularly given the land area constraints on full-scale systems in the UK, is the volume-adjusted removal rate [10], since treatment efficiency disregards flow-rate, and therefore the HRT. Furthermore, a synoptic mass balance analysis has demonstrated that maximising mass removal of Zn is the most cost-effective way to maximise the reduction of Zn concentration in the receiving watercourse [37]. The mean volume-adjusted removal rate for total Zn increased from 0.35 g/m³/d at an HRT of 15 h to 0.44 g/m³/d (HRT 12 h) and 0.54 g/m³/d (HRT 9 h). Whilst these values are somewhat lower than those previously reported for compost-based bioreactors [10], they are limited by the low influent Zn concentration. There is a strong positive correlation between effluent flow-rate and volume-adjusted removal rate (Spearman’s rank correlation: r_s = 0.931 (bioreactor A1), 0.960 (bioreactor A2), 0.953 (bioreactor A3); p < 0.05) (Figure S6), suggesting that there may be scope to increase the flow-rate further to maximise the attenuation of Zn.

Sulfate reduction was sustained throughout the trial, but the rate of reduction decreased as the HRT was lowered (Figure 3B). Following an initial peak during the first 30 days of operation, when the percentage of sulfate reduction exceeded 83%, the rate of reduction remained relatively stable until the HRT was lowered, albeit there was some variation between the bioreactors (mean 63% (A1), 81% (A2), 78% (A3)). An instantaneous decline in the percentage of sulfate reduction, most pronounced in bioreactor A1, occurred immediately after the HRT was lowered to 12 h, and this trend continued for the remainder of the trial. A short-lived increase in the sulfate reduction in bioreactors A2 and A3 around day 183 coincided with an exceptionally high ambient temperature (due to a building-wide heating malfunction), which appeared to promote sulfate reduction. Due to technical issues, the percentage of sulfate reduction in the final four samples collected from the bioreactors is not reported here, as confidence in the results is low.

3.2.2. Zinc Removal Mechanisms

BSR has frequently proven to be the predominant metal removal mechanism in bioreactors treating metal mine drainage [13,38], but additional processes (e.g., sorption, binding to organic matter) have also been reported to take place alongside BSR [33,34,35]. For the bioreactors discussed here, which received a low influent Zn concentration (mean 0.5 mg/L), sufficient sulfide appeared to be generated to attenuate all the Zn present despite a deterioration in the sulfate reduction over time (the percentage reduced had dropped to 7% (A1), 11% (A2), and 28% (A3) by the end of the trial (Figure 3B)). In the reaction of sulfate to sulfide (reaction (1)) and the precipitation of ZnS (reaction (2)), the theoretical molar ratio of sulfate to Zn is 1:1, such that one mole of Zn is removed for every mole of sulfate reduced. Except for a single occasion towards the end of the trial, the molar reduction in sulfate consistently exceeded that of Zn (Figure S7), supporting the attenuation of Zn as a sulfide. Additionally, it is possible that higher rates of sulfate reduction took place within the bioreactors but that excess sulfide was re-oxidised to sulfate within the effluent tubing before the point at which the effluent water samples were collected.

A decrease also occurred between the influent and effluent Eh throughout the trial (Figure S8). Although strongly anaerobic conditions did not become established until the HRT was reduced to 12 h; these conditions, which are optimal for SRB activity [30], were maintained, notwithstanding the odd exception (which may be an artefact of the time needed to sample effluent water, during which the Eh may change). The effluent pH (7.6 to 8.4) was also favourable for SRB activity [30] and was consistently higher than the influent pH (7.1 to 7.7) (Figure S8). The pH and Eh values recorded during the trial plot within the ZnS phase of the pH–Eh diagram for Zn [39] which indicated that ZnS precipitation was favoured.

Geochemical modelling using PHREEQC [40] indicated that the influent water was under-saturated with respect to the solid Zn phases smithsonite (ZnCO₃), willemite (Zn₂SiO₄), and Zn(OH)₂. The effluent water was also under-saturated with respect to these phases but was super-saturated with respect to sphalerite (ZnS) (mean saturation indices 3.91 (A1), 4.23 (A2), 3.93 (A3)) (Figure S9). Despite the decreasing rate of sulfate reduction over the duration of the trial (Figure 3B), the saturation index for sphalerite remained relatively consistent, suggesting that conditions of chemical equilibria remained appropriate for immobilisation of Zn as its sulfide.

Geochemical and microbial analyses of the substrate at the end of the investigation provided further evidence for the important role of BSR as a Zn removal mechanism in these bioreactors. A substantial accumulation of Zn occurred (Table S4) with Zn concentrations at their maximum (mean 943 mg/kg) in the uppermost layers of the substrate and decreasing with depth (mean 220 mg/kg at base of the substrate). The original substrate mix prior to placement in the bioreactors contained 100 mg/kg Zn. A similar vertical profile was previously observed in a pilot-scale bioreactor [10] and can be attributed to vigorous Zn removal in the upper layers of the substrate, close to where the influent water enters. Microbial analysis demonstrated strong and consistent evidence of a redox based succession as a function of depth in all three bioreactors, with shifts from aerobic to anaerobic conditions with increasing depth. SRB were highly abundant, representing between 2% and 7% of sequences in libraries across the bioreactor samples, a similar relative abundance to that reported in pilot-scale reactors [41]. SRB observed in the bioreactors belonged to the families Desulfarculales, Desulfobacterales, Desulfovibrionales, and Desulfuromonadales. Unlike Zn concentrations, SRB increased with depth with the greatest numbers observed in the samples collected from the region around the bottom of the reactors (Figure 4). Despite this increase in SRB abundance with depth, there is evidence of intense bacterial sulfate reduction taking place throughout the bioreactors, which is consistent with the removal of Zn as its sulfide. Given the high Zn concentrations and lower abundance of SRB in the upper layer of the substrate, other Zn attenuation mechanisms, such as adsorption and binding to organic matter, may have been important here. Such processes have previously been observed in the upper substrate layers of similar bioreactors [32]. It is also possible that, initially, sorption in the upper substrate layers was the dominant Zn removal mechanism, but once sorption capacity reached saturation, BSR became more prevalent, particularly in the lower substrate where SRB proliferated [42].

3.2.3. Hydraulic Performance

The hydraulic performance of the bioreactors was investigated by conducting hydraulic testing at varying hydraulic heads (0.075 m to 0.225 m) at the beginning and end of the trial to evaluate changes in the K_sat during operation of the bioreactors. The findings were consistent with those undertaken during the assessment of different substrate mixes, whereby the K_sat decreased with increasing hydraulic head (Figure S10) due to the flow resistance within the narrow diameter tubing connecting the permeameter to the bioreactor, which resulted in an underestimation of the K_sat [36]. Despite the three bioreactors containing the same substrate, the K_sat varied between them (Figure S10), particularly at the beginning of the trial, likely due to different packing densities. The control (limestone only) reactor had a lower K_sat (597 to 970 m/d) than the reactors containing the reactive substrate (mean 825 to 1202 m/d) due to the more uniform size of the limestone chips giving the substrate a more open structure. Over the duration of the trial, the K_sat decreased in all three bioreactors (mean decrease 12%) (Table S5) but the initial permeameter tests were conducted on a separate triplicate set of bioreactors set up for the sole purpose of testing the K_sat; therefore it was not possible to quantify the exact decrease.

4. Conclusions

An evaluation of the effectiveness of three different substrate mixes for use in nature-based (passive) systems for the remediation of metal mine drainage showed that they were all effective in their removal of Zn (88%–90% total Zn removal). Sulfate reduction varied between the substrate mixes, with those containing PAS 100 compost and/or sewage sludge impacted by hydraulic testing due to the higher flow-rates disturbing the SRB or removing their available carbon source in the fine material washed from the bioreactors. Whilst sulfate reduction was slow to become established in the bioreactors containing limestone, straw, and wood chips, it was then sustained for the remainder of the trial, with >25% decrease in sulfate concentration between the influent (22.4 mg/L) and effluent water. Hydraulic conductivity (K_sat) was consistently higher in the bioreactors without PAS 100 compost and/or sewage sludge, suggesting that reactors containing only limestone, wood chips, and straw are less likely to suffer permeability problems.

A limestone, wood chips, and straw substrate was therefore selected for performance testing at a lower HRT; initially 15 h, reducing to 12 h, and then to 9 h. There was no significant difference in Zn treatment efficiency between the HRTs, and the volume-adjusted removal rate increased as the HRT was shortened. Although the sulfate reduction rate dropped as the HRT decreased, Zn treatment efficiency remained high, suggesting sufficient sulfide was always available to immobilise Zn as ZnS(s). Geochemical modelling (PHREEQC) and microbial analysis of the bioreactor substrate also pointed to ZnS(s) precipitation, driven by a vigorous sulfate reduction with strongly positive saturation indices for ZnS and a high relative abundance (2%–7%) of SRB.

Overall, the results suggest that relatively fine-grained compost and sludge may be omitted from organic substrates that harness BSR. Substrates comprising only coarse-grained constituents, such as limestone, straw, and wood chips, are beneficial for maintaining medium- to long-term hydraulic conductivity. Furthermore, operating such treatment units at shorter HRTs may be feasible for the many modestly contaminated metal mine water discharges in the UK, without compromising treatment performance with respect to Zn removal. Operating at a shorter HRT has the benefits of (1) increasing the potential for deployment of nature-based treatment systems where land availability is limited, and (2) reducing the capital costs of such systems. Nevertheless, additional testing of the robustness of short HRT bioreactors is needed before full-scale deployment, both through longer-term testing and larger scale pilot-scale testing in ambient environmental conditions.