Attenuation of acid rock drainage by stimulating sulfur-reducing bacteria

Iron-sulfide minerals found in shale formations are stable under anaerobic conditions. However, in the presence of oxygen and water, acid-loving chemolithotrophic bacteria can transform the iron-sulfide minerals into a toxic solution of sulfuric acid and dissolved iron and minerals known as acid rock drainage (ARD). The objective of this study was to disrupt chemolithotrophic bacteria responsible for ARD using chemical treatments and to foster an environment favorable for competing microorganisms to attenuate the biologically induced ARD. Chemical treatments were injected into flow-through microcosms consisting of 501 g of pyrite-rich shale pieces inoculated with ARD bacteria. Three treatments were tested in the microcosms: (1) a sodium hydroxide-bleach mix, (2) a sodium lactate solution, and (3) a sodium lactate-soy infant formula mix. The effectiveness of the treatments was assessed by monitoring pH, dissolved iron, and other geochemical constituents in the discharge waters. The optimal treatment was a sequential injection of 1.5 g sodium hydroxide, followed by 0.75 g lactate and 1.5 g soy formula dissolved in 20 mL water. The pH of the discharge water rose to 6.0 within 10 days, dissolved iron concentrations dropped below 1 mg/L, the median alkalinity increased to 98 mg/L CaCO3, and sulfur-reducing and slime-producing bacteria populations were stimulated. The ARD attenuating benefits of this treatment were still evident after 231 days. Other treatments provided a number of ARD attenuating effects but were tempered by problems such as high phosphate concentrations, short longevity, or other shortcomings. The results of these laboratory microcosm experiments were promising for the attenuation of ARD. Additional investigations and careful selection of treatment methods will be needed for field application.


Introduction
Acid rock drainage (ARD) occurs when geologic deposits rich in iron-sulfide minerals are exposed to oxygen and water through excavation or landslides, transforming the sulfide into sulfuric acid and dissolving the iron . Other trace metals such as aluminum, zinc, copper, and other minerals often associated with iron-sulfide minerals co-dissolve into the acidic waters and exacerbate ARD environmental problems (Hammarstrom et al. 2005a). In Tennessee, the black shale in the Highland Rim and the Valley and Ridge provinces, the shale and coal formations along the Cumberland Plateau, and metamorphic rocks and shale in the Blue Ridge are rich in iron-sulfide minerals such as pyrite and FeS 2 (Fig. 1, Hardeman 1966). These geologic units are susceptible to forming ARD if disturbed by landslides, mining or construction excavation. ARD runoff, if left untreated, can have unintended and negative consequences 1 3 237 Page 2 of 18 on local stream aquatic biodiversity and environmental conditions (Kucken et al. 1994;Schorr et al. 2013;Betrie et al. 2015).
The chemical reaction responsible for ARD is summarized in Eq. 1. This reaction has a Gibbs free energy of − 17.68 kiloJoules per gram atomic mass and is thermodynamically favorable (Curtis and Morris 2013). Although this reaction is thermodynamically favorable, it is relatively slow under abiotic conditions due to the energy of activation: Equation 1 simplifies several biochemical steps used by acid-loving chemolithotrophic bacteria to catalyze the oxidation of iron-sulfide minerals into the ARD products and derive energy (Bacelar-Nicolau and Johnson 1999;Johnson and Hallberg 2005;Steinhower et al. 2010). The rate of biologically catalyzed iron-sulfide oxidation into H 2 SO 4 , dissolved iron and other minerals is enhanced more than 1000fold by the action of acid-loving chemolithotrophic bacteria Suzuki 1989, 1991). The equation highlights two essential components required by the chemolithotrophic bacteria to transform iron-sulfide minerals into ARD, oxygen (O 2 ) and water (H 2 O). Depriving the chemolithotrophic bacteria of either component will greatly reduce the biologically (1) 12FeS 2 + 45O 2 + 34H 2 O → 4[H 3 OFe(SO 4 ) 2 − 2Fe(OH) 3 ] + 16H 2 SO 4 derived ARD. For example, extended drought will hinder ARD development by limiting water required by the bacteria for the reaction. Under temperate, alternating wet-dry conditions, the moisture retained in the primary and secondary pores may be sufficient for the chemolithotrophic bacteria to transform FeS 2 into acidic water with high concentrations of dissolved minerals. As the weather conditions change from wet to dry, the seepage waters may become supersaturated with dissolved minerals and acid. Sustained drying conditions initiate acid-sulfate minerals to precipitate into efflorescent blooms of hydrated metal salts known as secondary sulfate minerals (Hammarstrom et al 2005b). The secondary sulfur minerals tend to be highly soluble during rain events, producing extremely acidic waters rich in dissolved metals and minerals . Secondary sulfate minerals were observed at the shale-source site in Middle Tennessee during the early stage of this investigation (Bradley et al. 2015b).
The unique topography, hydrology, geochemistry, and biology of each site influences acid production as well as the fate and transport mechanisms of dissolved minerals and metals. The site hydrogeologic conditions and weather patterns will determine if the recharging groundwater is relatively stagnant, persistently flowing, or pulse driven by rains and will influence the contact time between the water Fig. 1 Location of pyrite-bearing formations in Middle and East Tennessee and exposed iron-sulfide minerals (Sandlin et al. 2020). The exposed surface area and geochemistry of the site determines the bioavailability of the iron-sulfide minerals; for example, an exposed pile of fragmented shale pieces rich in iron-sulfide will provide greater surface area and bioavailability than an intact shale formation (Bradley et al. 2015a). The population of chemolithotrophic bacteria on a freshly exposed FeS 2 -rich geologic surface may initially be limited by the quantity and type of bacteria present. However, after sufficient time and natural selection, microbial communities acclimate to the local environmental conditions, ultimately reaching a natural carrying capacity.
Strategies for the remediation of ARD can be categorized as either on-site or off-site treatment methods. Examples of treating acidic drainage off-site include use of constructed wetlands to treat AMD (Whitehead and Prior 2005), diverting AMD to engineered anaerobic, sulfur-reducing-bacteria bioreactors (Sheoran et al. 2010;Hessler et al. 2020), or neutralizing ARD by directing flow-through reactive alkaline materials (Egiebor and Oni 2007;Li et al 2007)) followed by a bioreactor (Mendez et al 2021)). On-site treatment includes modifying the hydrology, reducing the oxygen availability, or altering the microbiology at the site. The site hydrology can be modified through physical alterations that cut off percolating rainwater. Reducing water flow may also reduce the oxygen levels at the site of the ARD reaction (Ma et al 2020). Johnson and Hallberg (2005) described using caps and other sealing methods to prevent groundwater recharge at mine spoil piles. One concern about this approach is the longevity of a sealant. Caps and sealants are vulnerable to weathering, biological perturbations, or groundwater flowing into and contacting iron-sulfide from an unprotected/ unsealed side. Rather than altering the hydrology, Janssen and Temminghoff (2004) neutralized an industrial sulfuric acid spill through stimulation of sulfur-reducing bacteria. They applied sugar-rich molasses to stimulate microbial respiration, consume the oxygen in the fluvial aquifer, generate sulfur-reducing conditions, precipitate zinc as zincsulfide, and neutralize the spilled sulfuric acid. Additional research is needed to determine if a similar strategy can be used to stimulate sulfur-reducing conditions at the source of ARD and deprive the chemolithotrophic bacteria of oxygen, thereby attenuating ARD problems. One of the challenges for any man-made anaerobic bioremediation system is the longevity of treatment because supplements can be consumed or carried away in rains and oxygen can be replenished during recharge (Sandlin et al. 2020).
The U.S. Geological Survey (USGS), in cooperation with the Tennessee Department of Transportation, conducted a study of the occurrence and geochemistry of ARD at road cuts in Tennessee (Bradley et al. 2014). A bibliography compiled for the investigation includes applicable references for acid rock drainage and remediation with emphasis on acid rock drainage related to transportation activities (Bradley and Worland 2015). One objective of the overall investigation was to attenuate ARD by manipulating the indigenous microbial community through different treatment injections, ultimately generating anaerobic sulfur-reducing conditions. Passive, on-site bioattenuation and the longevity of the treatment were also considered. The scope of the study included establishing flow-through microcosms constructed with shale from the Chattanooga Shale formation rich in pyrite collected from an ARD site in Middle Tennessee. The microcosms were subjected to various treatments and evaluations included monitoring pH and additional geochemical and microbial constituents in the effluent waters.
Stimulating indigenous sulfur-reducing bacteria under different geologic settings has been documented with regards to bioremediation of halogenated solvents and metals (Byl and Williams 2000;Lovley 2002;Gadd 2004;Zhuang et al 2012;Rambabu et al. 2020). Additional research identified potential food supplements and optimal temperature and pH to stimulate indigenous sulfur-reducing bacteria in the soil or aquifer (Fennell et al. 1997;Ellis et al. 2000;Bario-Lage et al. 1990). The most favorable foods were volatile fatty acids, such as acetate, lactate, or propionate, in conjunction with vitamins and minerals (Glod et al. 1997a, b) at stimulating environmental sulfurreducing bacteria (Bradley 2003;Carr and Hughes 1998). In each of these studies, the pH of the water was generally between 6 and 8. Sheoran et al. (2010) reported that axenic cultures of sulfur-reducing bacteria isolated from the environment were suppressed at pH lower than 5.5. Gupta and Sar (2020) had some success reversing the pyrite oxidation and acidification process in acid impacted soils (pH of 3.21) using an enriched bacteria consortium and cellulose as a food. Similarly, Jimenez-Castneda was able to stimulate sulfur-reducing bacteria using a local manure source (Jimenez-Castaneda et al 2020). Others reported that sulfur-reducing bacteria collected from anaerobic acid mine drainage and raised as part of a consortia were able to function at pH 4 and even pH 2.9 (Hwang and Jho 2018;Church et al. 2007;Bijmans et al. 2010;Miao et al. 2012). However, those investigations did not address whether these same sulfur-reducing bacteria could be nurtured and compete in an active iron-sulfide oxidizing environment to bio-attenuate the ARD problem. The study described in this paper addresses this oversight by establishing aerobic, flow-through, ARD microcosms using geologic materials from an active ARD site and then testing different treatments to influence the indigenous microbial community.

Materials and methods
The ability to shift the microbial community in microcosms from a pyrite-oxidizing, acid-producing, microbial community to an anaerobic, sulfur-reducing microbial consortium was tested using materials collected from an ARD site, and exposing the microcosm to different treatments under controlled lab conditions. Microcosms were established using crushed shale from a Middle Tennessee roadcut with active ARD conditions (water pH less than 3 and high concentrations of dissolved iron). The flow-through microcosm treatments and the longevity were assessed through monitoring activities as described below.

Source of shale and bacteria
The source of the pyrite-rich shale and the ARD bacteria inoculum was the Chattanooga Shale formation at a roadcut that was less than 5 years old in south Williamson County, Tennessee (latitude 35.814339 and longitude − 86.975004). Williamson County, Tennessee, in the southeastern United States, has a humid, subtropical climate with average monthly temperatures ranging from 27 to 89 F. Average annual precipitation is about 50 in. (NOAA 2021). The formations exposed in the roadcut include the upper Ordovician limestones at road level, overlain by the Devonian Chattanooga Shale, the Mississippian Maury Formation, with the Mississippian Fort Payne formation at the top of the roadcut. The Chattanooga Shale, about 6 ft thick, is composed of a fissile black shale with abundant pyrite mineralization. The pyrite occurs as small, distributed crystals ranging from < 0.25 to 1.5 mm and nodules ranging from about 0.5 to 2.5 cm. The Chattanooga Shale is a regional confining unit and several small springs discharge from near the top of the Chattanooga Shale. The water from the springs seep down the face of the Chattanooga Shale and flows through spoil piles consisting of broken and collapsed Chattanooga Shale.
The roadcut site had wet yellow-orange stains and other signs of ARD associated with the pyrite-rich shale and spoil piles (Fig. 2). Water that had not been in contact with the pyrite-rich shale or spoil piles at the site had a pH range of 6.40 to 7.83, and a specific conductance range of 306 to 436 microSiemens per centimeter (µS/cm). Water draining from the spoil piles, containing a mix of fragmented shale and limestone, and water dripping from openings on the upper surface of the Chattanooga Shale formation had a pH ranging from 2.45 to 4.00. The specific conductance of these acidic waters ranged from 926 to 2266 µS/cm, indicating the water was rich in dissolved minerals. During dry, hot intervals in the summer of 2014, many of the ARD seeps dried up and produced secondary sulfur minerals along the desiccated flow path.
The USGS Central Mineral and Environmental Resources geologic laboratory (Denver, CO) confirmed the shale was rich in iron-sulfide minerals with 1.67 mol (93.1 g) of iron and 3.19 mol (102 g) of sulfide per kilogram of shale, respectively. This approximate 1:2 molar ratio of iron to sulfide is equivalent to that of pyrite (FeS 2 ). Approximately 30 kg of this pyrite-bearing fresh, unweathered shale was excavated from the site and placed in clean, 5-gallon plastic buckets, and transported to the lab to be processed. A small portion of the shale material was shipped to the USGS laboratory for analysis and the rest was readied for use in the microcosms. In addition, 1 kg of gravel-sized, active ARD material from the spoil pile was collected in a separate plastic container to provide an inoculum of indigenous ARD bacteria. Sixty liters of water was collected from neutral pH seeps above the shale formation for the dilute reservoir water.

Preparing the microcosms
The shale material was brittle and relatively easy to break with a hammer into 0.75 in. or smaller pieces. A series of sieves was used to separate the fragmented shale into discrete sizes: 0.5 to 0.75 in., 0.25 to 0.5 in., 0.125 to 0.25 in., and less than 0.125 in. The 12 microcosm containers consisted of clean, sterile 500-mL Erlenmeyer side-arm flasks. Each microcosm flask was filled with the following quantities of shale fragments: 1 g of inoculum material from the spoil pile, 180 g of 0.5-0.75 in., 140 g 0.25-0.5 in., 80 g 0.125-0.25 in., and 100 g of less than 0.125 in. fragments for a total of 501 g of shale. The shale was placed in the flask with the largest pieces on the bottom and progressively smaller sized pieces stacked on top. A quarter-inch diameter polytetrafluoroethylene (PTFE) tube that touched the bottom of the Erlenmeyer flask was in place prior to adding the shale fragments to the flask. After the shale fragments were in place, a rubber stopper was lowered down the PTFE tube and seated into the mouth of the flask. Silicone sealant was used to prevent gas and water leaks around the PTFE tubing and rubber stopper. The PTFE tubing was attached to a 12-channel peristaltic pump (model Watson-Marlow 205u). Additional PTFE tubing was inserted into the side arm for the discharge water and sealed with Parafilm™. A 16-L carboy filled with purified water (16 Ohm resistivity), continuously stirred to aerate the water, served as the reservoir (Fig. 3). The initial pump rate was 2 mL per minute to remove dust and colloidal particles. Air bubbles were removed from the microcosms as they slowly filled by gently tilting and tapping the flasks to allow the bubbles to escape via the side-arm discharge. Each microcosm required approximately 360 mL of water to fill. After running five volumes of water (1800 mL) through each microcosm to remove dust and colloidal particles, the pump speed was lowered to 1 mL per minute. One liter of raw, unfiltered water collected from neutral pH seeps at the site was blended into 15 L of purified reservoir water. This blended reservoir water provided another inoculum with a mixed bacteria consortium indigenous to the site to improve ARD simulation (Merchand and Silverstein 2002). The inoculated microcosms were allowed to sit for 12 h before pumping resumed at 1 mL/min with the 1:15 blended reservoir water. The reservoir water was pumped through the microcosms 8 h/day at 1 mL/min for 5 consecutive days per week. Over the course of this project, the blended waters in the reservoir had a temperature of 25 °C (± 1.5), dissolved oxygen ranging from 6 to 10 mg/L, specific conductance ranging from 21 to 55 µS/cm, and pH ranging from 6.4 to 7.8.

Treatments
Different supplements have been used to stimulate sulfurreducing bacteria. Each supplement had success stimulating sulfur-reducing bacteria under their respective environmental conditions, but none of the treatments were evaluated under active ARD conditions. Stimulating sulfur-reducing bacteria in situ at an ARD site must consider the low pH environment, the oxygen input, as well as how to prolong the effect of the supplements in an environment that is subject to intermittent flow and stagnation. To address these concerns, different treatments were evaluated with three replicate microcosms per treatment. The study consisted of three experiments with four treatments per experiment. Treatment modifications built upon previous results. The treatments in the first experiment included: 1. Reference controls that consisted of reservoir water with no supplements or treatment injection, 2. Chemical shock treatment with 1.5 mL of 6% bleach and 1.5 g sodium hydroxide blended into 20 mL of water to knock down the ARD microbial consortia and the pH, followed by non-sterile reservoir water, 3. Initially injecting 0.75 g of sodium lactate syrup followed by non-sterile reservoir water, 4. Initially injecting 0.75 g sodium lactate syrup and 1.5 g soy-based infant formula followed by non-sterile reservoir water. The treatments used in the experiment were selected to either prepare the microcosm or provide a source of food and vitamins to stimulate microbial action. The chemical shock treatment in treatment #2 utilized bleach and sodium hydroxide to kill the bacteria in the microcosm to provide a near zero starting base. The shock treatments were not intended for a possible field application but were used to provide a zero-population control. Water containing live bacteria was pumped through the shocked microcosm allowing the bacteria to re-inoculate the microcosm. The third treatment used an injection of sodium lactate to provide a food source for sulfur-reducing bacteria. In the fourth treatment, a soy-based infant formula was injected to provide vitamins, especially vitamin B, and essential amino acids to enhance the microbial activity. The soy formula also curdles and "sticks" in the pore spaces providing a longer lasting effect on the microbial activity and improving the cost effectiveness of the treatment for possible field application. The persistence of the soy formula in the subsurface resulted in continued bioremediation of trichloroethylene for more than 8 months in a karst aquifer impacted by landfill waste (Hileman et al. 2004).
The second and third experiments used information gathered from the preceding study and modified the supplements to improve the outcome. The modifications to the four primary treatments and rationale are listed in Table 1. The supplements were pumped into the microcosms using the peristaltic pump at a rate of 2 mL/min. The supplements were dissolved in 20 mL of purified water in 12 large test tubes with pump intake tubes inserted directly into the test tube containing the assigned treatment. Using individual tubes to deliver the supplements provided an opportunity to confirm uniform pumping rates. As the supplement level declined in the test tubes, the tubes were refilled two times with reservoir water to rinse the treatment residue into the microcosms. After two rinses, the intake tubes were returned to the reservoir and the pump flow rate was adjusted to 1 mL/min.
The duration of the experiments varied based on how effective the treatments appeared to be working as assessed by the geochemistry of discharge waters. Geochemical constituents that factored into the decision to terminate a particular experiment included pH, dissolved iron, and phosphate concentration in the discharge waters. The first experiment was terminated after 30 days, the second experiment was terminated after 40 days, and the third experiment was terminated after 231 days. The pump flow rate for experiments 1 and 2 was 1 mL/min for 8 h a day, for 5 days a week for the entire study period. Experiment 3 started with the same flow rate as experiments 1 and 2 for the first 50 days, after which, the flow was reduced to two 8 h days per week. The reduction in pumping time after 50 days was done because a greater residence time was needed to maintain ARD conditions in the reference control microcosms.
The amount of water pushed through each microcosms (including pre-treatment injection water) for each experiment expressed in pore volumes was: 30 volumes in the first experiment, 45 volumes in second experiment, and 90 volumes in experiment 3.

Sampling procedure, geochemical and biological analysis
Prior to collecting the water samples for geochemical analysis, the pumps were run for 6 h at 1 mL per minute, which is equivalent to 360 mL per microcosm or one pore volume. The discharge waters were collected in clean 40 mL sample vials. The vials were allowed to overflow for 20 min to minimize water interaction with the atmosphere. Water quality readings were taken in the following order using calibrated meters: dissolved oxygen (Orion meter, model 820), pH (Hach SensIon1), and specific conductance (Hach, SensIon5). Sulfide concentration in the discharge water was also immediately measured after taking the meter readings. The vials were allowed to re-fill, capped without air bubbles, and stored in a sample refrigerator at 5 °C (± 1) until the iron, phosphate, and sulfate could be analyzed within a few days. Geochemical analytical methods and detection limits are listed in Table 2. The multiple microcosms and treatments with multiple sampling through time resulted in about 3000 samples for analysis. The geochemical analytical methods listed in Table 2 were utilized instead of laboratory analytical methods since the methods listed provide the level of accuracy needed at much reduced cost. All water-quality meters were calibrated to standards as described by the manufacturer. Samples were analyzed in triplicate providing additional quality assurance/quality control.
Microbial types were characterized using Biological Activity Reaction Tests™ (BART). The BART assays were run near the end of each study when the pH and iron concentrations appeared to be stable which suggested the microcosms had reached equilibria. The waters for the BART assays were pumped at 1 mL/min directly into the BART tubes from the microcosms, incubated in the dark at 25 °C and monitored each day for 7 days (Culimore 2007). The BART assays provided data on the general types of active bacteria present and the response of the bacterial communities to the various treatments. For the purpose of these microcosms, the identification of specific bacteria was not needed.
The geochemical data were arranged in a computer spreadsheet for statistical analysis (Helsel and Hirsch 2002). Due to the nature of the experiments and the evaluation needed to determine effects of the treatments on dissolved iron and pH relative to the control strict statistical analysis were not needed. The change in pH and dissolved iron were evaluated in terms of change through time by graphical analysis. Plots of pH, specific conductance and dissolved iron concentrations through time were created to compare different treatments. The remaining geochemical data were summarized in tables providing the median and the interquartile range to provide a measure of spread. The BART results were also plotted to show differences in microbial communities between treatments.

Results
Prior to starting the experiments, all the microcosms had to exhibit ARD conditions in the discharge water, including a pH of 3 or less, dissolved iron concentrations ranging from 75 to 100 mg/L, and specific conductance ranging from 1500 to 2500 µS/cm. In addition, dissolved sulfide concentrations were below detection (< 0.01 mg/L) and sulfate concentrations were high (> 500 mg/L) in the ARD discharge waters. These geochemical values were similar to field ARD conditions and provided evidence that the bacteria responsible for ARD were active in the microcosms. Once stable, ARD conditions were observed for several days, the experimental treatments began and monitoring continued.
Experiment 1 consisted of four treatments: reference controls with no supplements, a chemical shock treatment combining bleach and sodium hydroxide, a single supplement of sodium lactate, and a mixed supplement of sodium lactate and soy formula. The control microcosms maintained a low pH of 3.0 to 3.4 the entire 30 days (Fig. 4a). The pH of microcosms treated with bleach and sodium hydroxide initially climbed to pH 8.8, followed by a drop to pH 5.1 by day 30. The microcosms supplemented with lactate or lactate + soy formula initially experienced rising pH in the discharge waters indicating that non-ARD bacteria stimulation was occurring. However, after 20 days, the pH in both treatments stabilized for 10 days at approximately 4.3 for lactate and 5.4 for lactate + soy formula, respectively. The specific conductance values, an indicator of dissolved solids concentrations, were very similar for all treatments except for the microcosms treated with bleach and NaOH (Fig. 4b).
By day 30, all the treatments had similar specific conductance values. The dissolved iron concentrations in the discharge waters dropped for the control and the microcosms treated with bleach and NaOH, while remaining high in the microcosms treated with lactate and lactate + soy formula, approximately 50 and 60 mg/L, respectively (Fig. 4c).
Geochemical properties, including alkalinity, dissolved oxygen, phosphate, sulfide, and sulfate, of the discharge waters varied among treatments (Table 3). In the first experiment, only the microcosms treated with bleach and sodium hydroxide had any measurable alkalinity. The median dissolved oxygen was lowest in the microcosms treated with lactate and soy formula. The median phosphate was below 0.5 mg/L for all the experiment 1 treatments. Median sulfide concentrations were greatest in the microcosms spiked with lactate and soy, while median sulfate was greatest in the microcosms treated with bleach and NaOH.
Experiment 2 was based on results of experiment 1. From the results in experiment 1, it was clear that microcosms shocked with bleach and sodium hydroxide attenuated two ARD symptoms (lowered iron, raised the pH) as compared to the control and other treatments. However, treating an ARD site with sufficient bleach and sodium hydroxide to replicate the laboratory chemical shock is not a reasonable remediation treatment for a field application. For experiment 2, 10 mL of 1 molar potassium dibasic phosphate buffer (K 2 HPO 4 ), pH 9.1, was injected as part of each treatment. It was conjectured that reactive phosphate would help remove dissolved iron through precipitation (Seida and Nakano 2002) and that it would also act as a buffer to raise the pH slightly.
The results of adding phosphate buffer to the treatments helped to moderate the pH spike observed in the bleach and NaOH treatment of Experiment 1 (Fig. 5a). The addition of  phosphate than those without in experiment 1 (Fig. 5b). The dissolved iron concentrations dropped at a faster rate when phosphate was added to the microcosms as compared to experiment 1 (Fig. 5c). Other geochemical indicators, such as lower dissolved oxygen concentrations and an increase in sulfide concentrations in the discharge water, support the premise that addition of pH-neutralizing phosphate buffer enhanced the suitability of the environment for anaerobic bacteria and reduced the ARD symptoms (Table 3). Unfortunately, in addition to mitigating the ARD symptoms, there was a 100-fold increase (average 51 mg/L after 12 days) in reactive phosphate in the discharge waters. Applying the use of phosphate buffers at an ARD field site would require scaling up the volume and would result Fig. 5 Temporal changes in geochemistry of microcosm discharge water (experiment 2). Treatments included control with potassium phosphate (K 2 PO 4 ), addition of bleach + NaOH + K 2 PO 4 , addition of lactate + K 2 PO 4 , and addition of lactate and soy formula + K 2 PO 4 . (a, Top) Average pH; (b, middle) average specific conductance; (c, bottom) average dissolved iron concentration. n-number of samples per treatment 1 3 237 Page 12 of 18 in unacceptably high phosphate concentrations in the runoff. Although the injection of phosphate buffer in conjunction with the supplements resulted in nutrient complications, it did improve attenuation of several ARD traits. It was hypothesized that the neutralizing capacity provided by the phosphate buffer enhanced the colonization and growth Fig. 6 Temporal changes in geochemistry of microcosm discharge waters (experiment 3). Treatments included controls with nothing added, sequential addition of NaOH followed by bleach, NaOH followed by lactate, and addition of NaOH followed by lactate and soy formula. (a, Top) Average pH; (b, middle) average specific conductance; (c, bottom) average dissolved iron concentration. n-number of samples per treatment Fig. 7 The mean number of iron related bacteria (top, a), sulfur-related bacteria (middle, b) and slime-producing bacteria (bottom, c) in discharge water derived using Biological Activity Reaction Tests. [Y-axis labels: numbers refer to experiment; background = neutral pH waters from site used in reservoir; Ref = reference control; Shock = NaOH and bleach shock treatment; Lact = Lactate treatment; L + Soy = Lactate and soy formula treatment; experiment 2 included phosphate buffer; experiment 3 included sequential injection, beginning with NaOH, Number of samples for each average, n = 3] of non-ARD bacteria, especially in the lactate and the lactate + soy formula treatments.
Experiment 3 built on the results of experiment 2 and was designed to test the hypothesis that increasing the pH in the microcosm with a pretreatment, prior to injecting the food supplements, would enhance colonization and growth of non-ARD bacteria and aid in attenuation of ARD. In experiment 3, sodium hydroxide (NaOH) was used as a pretreatment in all treatments. As a result of the NaOH injection, the water in the microcosms treated with lactate and soy formula rose to pH 6 in less than 10 days (Fig. 6a). The concurrent rise in median alkalinity and sulfide concentrations, and drop in dissolved oxygen concentrations, implies that there was a substantial rise in bacteria activity in the microcosms treated with NaOH followed by lactate and soy formula. This rise in pH was faster and higher than the rise triggered by a sequential injection of NaOH followed by bleach.
The injection of NaOH caused a slight rise in specific conductance initially, but the specific conductance dropped to less than 500 µS/cm by day 25 (Fig. 6b), with the NaOH + lactate + soy formula treatment resulting in the lowest rise in specific conductance. The dissolved iron concentrations dropped quickest in the NaOH + lactate + soy formula treatment (Fig. 6c). This reduction in dissolved iron may have been due to elevated sulfides (Table 3) precipitating the iron. The median phosphate levels were less than 1 mg/L in all the treatments. The benefits of the sequential injection of NaOH, followed by lactate and soy formula, were still evident after 231 days.
The BART assays provide a population estimate of different bacteria types based on changes observed in the incubation tubes containing microcosm water. The BART assays were conducted toward the end of each experiment to provide an evaluation of how the microbial community responded to the treatments after significant volumes of artificial recharge water have percolated through the microcosms. The microcosms dosed with phosphate as part of Experiment 2 had a distinct increase in numbers of iron related bacteria as compared to the other treatments (Fig. 7a). The sulfur-related bacteria appeared to respond positively to the lactate and lactate + soy formula treatments when there was an increase in pH brought about by phosphate buffer or NaOH (Fig. 7b). The population of slime-producing bacteria tended to stabilize at densities similar to the background water in the reservoir with the exception of a depressed population in the chemical shock treatment (experiments 1 and 2) and an increase due to pH adjustments (experiment 2) and lactate and soy formula (experiment 3) (Fig. 7c).

Discussion
Geochemical and biological indicators were used to evaluate treatments with the goal of attenuating microbialderived ARD. The treatment strategies included disrupting the microbial population through chemical shock or shifting the microbial population from iron-sulfide oxidizing bacteria to sulfur-reducing bacteria with supplements. In addition to evaluating effectiveness of the treatment, consideration was given to the longevity of the treatment. The optimum attenuation and longevity treatment was a sequential injection of NaOH, followed by sodium lactate and soy infant formula. This sequential treatment promoted the highest rise in pH, alkalinity, sulfide, slime-producing and sulfur-reducing bacteria in the discharge waters as compared to the other treatments and reference controls.
The initial injection of NaOH probably neutralized the pH long enough to allow heterotrophic, aerobic and anaerobic bacteria to colonize and feed on the lactate and soy formula. The dissolved soy formula is rich in nutrients, such as protein, carbohydrates, fats, and B vitamins. These nutrients stimulate heterotrophic aerobic bacteria, such as the slime-producing bacteria, responsible for rapidly consuming oxygen and establishing an anaerobic environment. After the oxygen is consumed, the facultative anaerobic and obligate anaerobic bacteria begin to dominate the microbial community. Lactate was included in the supplement treatment because it is the preferred food of sulfur-reducing bacteria (Carr and Hughes 1998) and has been used to stimulate their growth in groundwater systems (Bradley 2003). The B vitamins present in the soy formula are essential to anaerobic respiration enzymes and would normally take a long time to synthesize (Ellis et al. 2000).
Another benefit associated with the soy formula was that the milky liquid flowed effortlessly into the pore spaces while adhering to the shale surfaces. Even as the soy formula began to spoil and curdle, it continued to adhere to the surfaces, providing excellent conditions for biofilm development and generating anoxic conditions on the surfaces of the shale. Despite pumping 45 pore volumes of water with relatively high dissolved oxygen concentrations (6 to 10 mg/L) through the microcosms, the treatment continued to attenuate ARD symptoms for the duration of experiment 3 (231 days). The success of the treatment was probably due in part to the biofilm development on the shale materials. Beyenal and Babauta, (2012) described how environmental bacteria such as Geobacter sulfurreducens develop biofilms on solid surfaces and establish extremely anaerobic, reducing (electron rich) conditions at the biofilm-solid interface. Such a biofilm would provide a physicochemical barrier to the oxygen and the chemolithotrophic bacteria involved in the oxidation of iron-sulfide minerals and production of acid rock drainage (conceptual model, Fig. 8).
In addition to the beneficial results from the sequential injection of NaOH, soy formula and lactate, there were some other interesting results. Use of phosphate in experiment 2 provided favorable pH control, reduced iron and sulfate, and enhanced bacteria populations for both the lactate, and the lactate + soy formula. However, the treatments in experiment 2 also released water with high levels of phosphate, median values ranged from 51 to 95 mg/L. Phosphate levels of this magnitude would not be acceptable because of the eutrophication potential for receiving streams (Tennessee Department of Environment and Conservation 2015). Perhaps treatments using phosphate buffer would work in an engineered bioreactor where flow is more tightly controlled. Additional research is needed to determine if use of phosphate buffer would be feasible in a bioreactor. As noted earlier, the results of sequential treatment using NaOH, soy formula and lactate provided a promising treatment. There was one negative side effect of the treatment, a foul odor in the discharge waters associated with the initial curdling of the formula. The odor was most noticeable after 4 days and lingered for another 5 days. This odor might be an issue in a populated area, but not in a rural areas removed from the general public. In conclusion, the results of this study indicate there is good potential that ARD can be passively attenuated using a sequential treatment of NaOH, followed by a mixture of lactate and soy formula.

Conclusions
Acid rock drainage (ARD) occurs when geologic deposits rich in iron-sulfide minerals are exposed to oxygen and water transforming the sulfide into sulfuric acid and releasing high concentrations of dissolved iron and other trace metals such as aluminum, zinc, copper, and other minerals which dissolve into the acidic waters and exacerbate ARD environmental problems. In Tennessee, the black shale in the Highland Rim and the Valley and Ridge provinces, the shale and coal formations along the Cumberland Plateau, and metamorphic rocks and shale in the Blue Ridge contain relatively high percentages of iron-sulfide minerals such as pyrite, FeS 2 . When disturbed by landslides or construction excavation, these formations are susceptible to forming ARD which can have negative consequences on stream water quality and local aquatic biodiversity.
The USGS, in cooperation with the Tennessee Department of Transportation, conducted an investigation of ARD at road cuts in Tennessee. As part of the overall investigation, experiments were conducted to evaluate the attenuation of ARD by manipulating the indigenous microbial community through different treatment injections, ultimately generating anaerobic sulfur-reducing conditions. The three experiments were conducted by establishing flow-through microcosms utilizing pyrite-rich shale from the Chattanooga Shale and injecting a series of treatments to stimulate bacterial activity.
Geochemical and biological indicators were used to evaluate the results of the treatments with the goal of attenuating microbial-derived ARD. In addition to evaluating effectiveness of the different treatments, consideration was given to the longevity of the treatment. The optimum attenuation and longevity treatment was a sequential injection of NaOH, followed by sodium lactate and soy infant formula. This sequential treatment promoted the highest rise in pH, alkalinity, sulfide, slime-producing and sulfur-reducing bacteria in the discharge waters as compared to the other treatments and reference controls. The use of dibasic phosphate buffer (K 2 HPO 4 ) to increase the initial pH of the treatment also resulted in high concentrations of phosphate in the discharge water and would not be suitable for field application. The results of this study indicate there is good potential that ARD can be passively attenuated using a sequential treatment of NaOH, followed by a mixture of lactate and soy formula. A small field application would be the next step to test the technique to reduce the impact of ARD on the environment.
Author contributions All the authors helped generate data, organize, and interpret the data.
Funding Funding was provided by Tennessee Department of Transportation and U.S. Geological Survey.
Data availability Data derived from this study are available at Bradley and Byl (2022).
Code availability Not applicable.

Conflict of interest
The author declares that they have no conflict of interest.
Ethics approval Not applicable.

Consent for publication Not applicable.
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/.