enhanced al and zn removal from coal mine drainage during rapid oxidation and precipitation of fe oxides at near neutral ph

Sensitivity analysis was conducted 1 to determine the level of control exerted by different surface characteristics of HFO surface area, number of weak and strong sites, log K values, 2

PII: S0883-2927(16)30603-5

DOI: 10.1016/j.apgeochem.2016.12.019

Received Date: 8 March 2016

Revised Date: 23 September 2016

Accepted Date: 22 December 2016

Please cite this article as: Burrows, J.E., Cravotta III., , C.A., Peters, S.C., Enhanced Al and Zn removal

from coal-mine drainage during rapid oxidation and precipitation of Fe oxides at near-neutral pH, Applied

Geochemistry (2017), doi: 10.1016/j.apgeochem.2016.12.019.

This is a PDF file of an unedited manuscript that has been accepted for publication As a service to our customers we are providing this early version of the manuscript The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

16.8 mL/s; Aer 3 25.0 mL/s), and hydrogen peroxide (H2O2) to test the hypothesis that aeration

increases pH, FeII oxidation, hydrous FeIII oxide (HFO) formation, and trace-metal removal

through adsorption and coprecipitation with HFO During 5.5-hr field experiments, pH increased from 6.4 to 6.7, 7.1, 7.6, and 8.1 for the control, Aer 1, Aer 2, and Aer 3, respectively, but

decreased to 6.3 for the H2O2 treatment Aeration accelerated removal of dissolved CO2, Fe, Al,

and Zn In Aer 3, dissolved Al was completely removed within 1 hour, but increased to ~20% of the initial concentration after 2.5 hours when pH exceeded 7.5 H2O2 promoted rapid removal of

all dissolved Fe and Al, and 13% of dissolved Zn

Kinetic modeling with PHREEQC simulated effects of aeration on pH, CO2, Fe, Zn, and

Al Aeration enhanced Zn adsorption by increasing pH and HFO formation while decreasing aqueous CO2 available to form ZnCO30 and Zn(CO3)22- at high pH Al concentrations were

inconsistent with solubility control by Al minerals or Al-containing HFO, but could be simulated

by adsorption on HFO at pH<7.5 and desorption at higher pH where Al(OH)4- was predominant

Thus, aeration or chemical oxidation with pH adjustment to ~7.5 could be effective for treating high-Fe and moderate-Zn concentrations, whereas chemical oxidation without pH adjustment may be effective for treating high-Fe and moderate-Al concentrations

Keywords: PHREEQC modeling; Fe oxidation kinetics; CO2 outgassing; metals adsorption;

coprecipitation; carbonate complexing

near neutral (6 to 8) and elevated concentrations of sulfate and metal(loid) constituents such as

Fe, Mn, Al, Zn, Ni, and As (Rose and Cravotta, 1998; Kirby and Cravotta, 2005a, 2005b;

Cravotta, 2008b) CMD can degrade the aquatic environment through chemical, physical,

biological, and ecological pathways because of the toxicity of the dissolved constituents and the accumulation of ochreous precipitates (Soucek et al., 2000; Battaglia et al., 2005) While FeII,

MnII, and Al tend to be the most abundant dissolved metals in CMD, associated trace elements,

such as Zn, Ni, Pb, Cu, Cd, Cr, Co, and As, can be present at concentrations that approach or exceed aquatic toxicity thresholds (Smith and Huyck, 1999; Cravotta, 2008a) Near-neutral CMD tends to have low dissolved Al concentrations (< 0.1 mg/L), presumably because of solubility limitation with respect to Al hydroxides; however, FeII, MnII, and associated trace metals

concentrations generally are not limited by hydroxide mineral solubility (Cravotta, 2008b) With few exceptions, concentrations of potentially toxic trace elements in CMD tend to decrease with increased pH, despite being orders of magnitude below theoretical solubility limits (Cravotta, 2008b) Such trends can be explained by the removal of trace ions by adsorption to or

coprecipitation with hydrous FeIII oxides (HFO) (Dzombak and Morel, 1990; Kooner, 1993;

Webster et al., 1998; Lee et al., 2002; Kairies et al., 2005; Sparks, 2005; Burgos et al., 2012) and,

to a lesser extent, hydrous MnIII-IV oxides (HMO) (Loganathan and Burau, 1973; McKenzie,

1980; Harvey and Fuller, 1998; Trivedi and Axe, 2001) or hydrous Al oxides (HAO) (Coston et al., 1995; Lee et al., 2002; Munk et al., 2002; Karamalidis and Dzombak, 2010) Furthermore, Al and Mn adsorption to HFO has been hypothesized for CMD in which Al and Mn are

undersaturated with respect to Al(OH)3(a) and Mn(OH)2(a), and Fe is oversaturated with respect to

Fe(OH)3(a) (Lee and Faure, 2007) Therefore, enhancing the potential for the precipitation of

HFO, HMO, and/or HAO with the subsequent adsorption of trace elements can be a critical factor in treatment-system design to decrease environmental impacts from CMD

The optimal treatment strategy for CMD can be influenced by water chemistry, discharge rate, site characteristics, and financial resources to construct and maintain a treatment system(s) Typical treatment steps involve acid neutralization, metals precipitation, and solids removal CMD treatment systems can use various processes to increase pH to target values of 6 to 9 and induce the precipitation of HAO, HFO, and HMO, which ideally will accumulate as sludge within the treatment system For example, aeration can be an effective treatment strategy to increase the pH of poorly oxygenated, net-alkaline, FeII laden CMD, partially treated effluent

discharged from an anoxic limestone drain (ALD), or a reducing and alkalinity producing system (RAPS) (e.g Hedin et al., 1994; Johnson and Hallberg, 2005) Aeration increases dissolved O2

needed for FeII oxidation and promotes CO2 outgassing, thereby increasing the pH, the rate of

FeII oxidation, and the formation of HFO (Singer and Stumm, 1970; Lee et al., 2002; Cravotta,

2007; Kirby et al., 2009; Geroni et al., 2012; Cravotta, 2015) However, the potential effects of

CO2 outgassing and FeII oxidation kinetics on HAO or HMO formation and on trace-metal

adsorption or coprecipitation processes in CMD treatment systems generally has not been quantified

Ochres associated with CMD tend to be enriched in Fe but generally contain varying proportions of hydrous FeIII, MnIII-IV, and Al oxides plus other compounds including silicates,

carbonates, and sulfates The HFO at CMD discharge and treatment sites can be described as poorly crystalline to microcrystalline FeIII solids including ferrihydrite, lepidocrocite,

schwertmannite, and goethite (e.g Winland et al., 1991; Bigham et al., 1996; Kairies et al., 2005;

whereas the HAO generally can be characterized as amorphous Al(OH)3, microcrystalline

gibbsite, and basaluminite (Nordstrom, 2011) HFO, HMO, and HAO in mine drainage tend to contain trace metals For example, Kairies et al (2005) reported Zn, Co, Ni, and Mn adsorption

to HFO and HMO in CMD treatment systems, and Munk et al (2002) measured high concentrations of Zn, Pb, Cu, and Ni associated with HAO precipitates in a stream contaminated

by acid rock drainage However, studies have indicated that the extent of trace-metal accumulation with the CMD ochres can vary within and between treatment systems (Cravotta and Trahan, 1999; Kairies et al., 2005; Burgos et al., 2012)

Trace-metal adsorption to HFO, HMO, and HAO is affected by the solution pH and ionic composition plus various other factors Numerous studies have indicated enhanced cation adsorption and anion desorption in response to increasing pH values (e.g., Dzombak and Morel, 1990; Lee et al., 2002; Karamalidis and Dzombak, 2010) Adsorption can also be affected by aging, crystallinity, and particle size of the sorbent, competition for surface sites, and the ionic strength and formation of aqueous metal complexes with organic acids and carbonate, hydroxide, and sulfate ions (Davis and Leckie, 1978; Theis and Richter, 1979; Swallow et al., 1980; Trivedi and Axe, 2001; Villalobos et al., 2001; Grafe et al., 2002; Kim et al., 2011; Antelo et al., 2005; Kairies et al., 2005; Xu et al., 2006) Compared to the free metal ions, the aqueous complexes have decreased ionic charge, which decreases the potential for the metal complexes to be adsorbed to negatively charged HFO, HAO, and HMO surfaces (McLean and Bledsoe, 1992; Langmuir, 1997)

hypotheses: (1) aeration will promote CO2 outgassing, with consequent increases in pH, FeII

oxidation, HFO precipitation, and coprecipitation or adsorption of trace metals to HFO; whereas (2) H2O2 addition, without aeration, will accelerate the oxidation of FeII, the precipitation and

settling of HFO, and coprecipitation of trace metals, but will not increase pH or enhance metal adsorption Kinetic and surface complexation modeling with PHREEQC was used to evaluate the interrelations among the pH, the rates of FeII oxidation, and the extent of Al and Zn

trace-removal during aeration experiments and during transport downstream from the CMD source, and to determine the partitioning of Al and Zn between the free ion, aqueous complexes, and surface complexes on HFO and HAO Sensitivity analysis was conducted (1) to determine the level of control exerted by different surface characteristics of HFO (surface area, number of weak and strong sites, log K values), (2) to evaluate the relative importance of adsorption versus precipitation or coprecipitation of dissolved Al and Zn with FeIII, (3) to demonstrate the effects

of the initial concentration of Fe, Al, Mn, and Zn on adsorption of Al and Zn, and (4) to address uncertainty in the aqueous concentration measurements in this study

2 Methods 2.1 Site Description This study evaluated anoxic CMD from the Oak Hill Boreholes (40°42'11.73"N, 76°15'4.14"W) northwest of Pottsville, PA in the West Branch Schuylkill River watershed The

and MnII (~4 mg/L), the near-neutral CMD at the Oak Hill Boreholes generally has low

concentrations (medians 0.04 to 0.06 mg/L) of dissolved Al, Zn, Ni, and Co (Cravotta et al., 2014) However, measured concentrations of these trace metals occasionally approach or exceed aquatic toxicity thresholds (shown in Supplementary Table S.1) These chemical characteristics

of the Oak Hill Boreholes are consistent with many other net-alkaline discharges in the anthracite and bituminous fields in Pennsylvania (Cravotta, 2008a, 2008b) For example, 15 of 24 samples from anthracite discharges in 2012 (not including Oak Hill Boreholes) were near-neutral (pH 6 to 8), with 10 of those found to be net alkaline (Burrows et al., 2015)

2.2 Experimental Procedure Batch aeration experiments (Figure 1) were conducted in the field for a duration of 5 to 5.5 hours using freshly discharged CMD from the Oak Hill Boreholes (Supplementary Tables S.1 and S.2) The experiments were conducted using 40-L insulated containers filled with 36 L CMD open to the air, out of direct sunlight; the water temperature variation between the start and finish of each experiment did not exceed 2.5°C Three of the experiments involved aeration at different rates (Aer 1: 12.6 mL/s; Aer 2: 16.8 mL/s; Aer 3: 25.0 mL/s) by continuously bubbling air through porous stone diffusers immersed in the CMD For comparison, a control with no mechanical aeration or chemical addition, and a hydrogen-peroxide treatment (5 mL of 3% H2O2

added to 20 L CMD in a 24-L insulated container) were also monitored The dosage of H2O2 was

estimated on a stoichiometric basis (2 Fe2+ + H2O2 + 2 H+ = 2 Fe3+ + 2 H2O) for an initial FeII

concentration of 20 mg/L (0.358 mmol/L) in the 20-L sample, and assuming that H2O2 would not

be effective for oxidizing MnII (Cole et al., 1977; Sato and Robbins, 2000) The 5 mL of 3%

H2O2 added to the 20 L sample equates to 0.22 mmol/L H2O2, which would be sufficient to

oxidize 0.44 mmol/L FeII (20 percent greater than the initial FeII concentration) Thus, the batch

experiments demonstrate potential effects of common CMD treatment strategies involving settling ponds (control), active aeration through cascades or mechanical mixing (Aer 1, Aer 2, Aer 3), and chemical oxidation (H2O2) The experiments were conducted in two batches, one in

June 2013 (control, Aer 3, and H2O2) and one in July 2013 (control, Aer 1, and Aer 2) Chemical

characteristics of the initial CMD and experimental procedures were similar for the two batches For this paper, the control, Aer 1, and Aer 2 data of July and the Aer 3 and H2O2 data of June are

evaluated as parallel tests

The temperature, pH (gel-filled electrode), specific conductance (SC), oxidation reduction potential (ORP), and dissolved oxygen (DO) were measured at 1 minute intervals by using a dedicated YSI 556 multiparameter sonde for each experiment The pH was also measured at less frequent intervals during alkalinity analysis by using an Orion 8165BNWP Ross (liquid-filled) electrode The alkalinity was determined in the field by titrating 50-ml aliquots of unfiltered serial samples with 0.16 N H2SO4 to an endpoint pH of 4.5 Although the YSI pH

electrodes and the Orion Ross electrode were simultaneously calibrated in the same buffers before starting the parallel tests, there were differences in the absolute value of pH recorded over time for the different electrodes (as shown in Supplementary Table S.2) Thus, the “observed”

pH values considered for model calibration were estimated by adjusting the continuous pH data

to be comparable to the Orion Ross values The [CO2] (shown in Supplementary Table S.2) was

computed as [H2CO3*] (= 10-pH × γHCO3- × [HCO3-] / K1) from observed pH and alkalinity ([HCO3-] = ALK / 50,000), with SC used to estimate ionic strength (I = SC×1.88×10-5) for the

activity coefficient (γHCO3-) and with temperature used to adjust the first-dissociation constant

of H2CO3* (K1) by the van’t Hoff equation (e.g Langmuir, 1997) The Eh and pe were

calculated from ORP by using the temperature correction method of Nordstrom (1977) Serial samples for dissolved cation concentrations were filtered (0.45 µm, Whatman nylon membrane with polypropylene prefilter and housing) and acidified to a pH < 2 in the field using HNO3, and

analyzed at Lehigh University by using an Inductively Coupled Plasma Mass Spectrometer

(ICP-MS, ThermoElemental X-Series, Winsford, UK) The concentrations of FeII analyzed for

selected samples in the field by the phenanthroline method were equivalent to the dissolved Fe concentrations determined by ICP-MS; therefore, ICP-MS values for the filtered samples are used hereinafter to represent FeII Analytical error and precision were determined by analyzing

duplicates of U.S Geological Survey (USGS) standard reference water samples (reported in Supplementary Table S.2) While the batch experiments were conducted on two separate dates and combined into one dataset, the small differences in the initial solution chemistry

(Supplementary Tables S.1 and S.2) were consistent with the analytical uncertainty

Figure 1 CMD treatment methods used for batch experiments Aeration rate is indicated below experiment name

A coupled model of kinetic reactions (O2 ingassing, CO2 outgassing, and FeII oxidation)

and surface complexation (Al and Zn adsorption) was created by using PHREEQC version 3.2.0 (Parkhurst and Appelo, 2013) The kinetic model was previously calibrated by Cravotta (2015)

to fit the data for pH, FeII, dissolved O2 and CO2, and alkalinity measured during the aeration

experiments described in this paper by applying 1st-order rate equations for O2 ingassing and

CO2 outgassing and the pH-dependent homogeneous FeII oxidation rate law:

-d[FeII]/dt = k[FeII][O2][H+]-2 (1)

where k is the FeII oxidation rate constant and square brackets indicate concentration (Singer and

Stumm, 1970; Stumm and Morgan, 1996) The simulated pH varies in response to CO2

outgassing, FeII oxidation plus aqueous CO2 speciation, and precipitation reactions The

simulation of the three aeration test results use the measured anion, cation, pH, alkalinity, and temperature data for initial conditions; previously calibrated rate constants for CO2 outgassing,

O2 ingassing, and homogeneous FeII oxidation (Cravotta, 2015) for the kinetic parameters;

solubility control by Fe(OH)3(a) and Al(OH)3(a); and precipitation of calcite at saturation index

(SI) ≥0.3 To simulate the control and H2O2 tests, additional rate estimates for gas exchange and

FeII oxidation were added For the control, the homogeneous FeII oxidation rate constant

determined for Aer 1 was used; however, more than an order of magnitude lower rates for O2

ingassing [kL,O2 (mass-transfer coefficient) = 0.000054 min-1], and CO2 outgassing (kL,CO2 =

0.0013 min-1) were used To simulate the instantaneous oxidation of FeII with H2O2 addition, a 1st

order kinetic rate model dependent on time and H2O2 concentration (zero-order with respect to

FeII) was used to oxidize two moles of FeII per one mole of H2O2, and combined with the same

low O2 ingassing rate (kL,O2 = 0.000052 min-1) and CO2 outgassing rate (kL,CO2 = 0.0013 min-1)

used for the control It should be noted that the CO2 outgassing rates estimated for the batch

experiments reported herein are consistent with values for full-scale CMD treatment systems in Pennsylvania Specifically, the kL,CO2 value estimated for the control and H2O2 simulations is

consistent with values estimated for passive, aerobic CMD treatment ponds and wetlands, and the kL,CO2 value estimated for the Aer 3 simulations is consistent with values estimated for

surface aerators and continuous stirred reactors at active CMD treatment facilities (B Means, U.S Office of Surface Mining, Reclamation and Enforcement, 2016, written commun.).The kinetics model was expanded to include reactions for cation and anion adsorption to HFO based on the PHREEQC database for surface-complexation (Parkhurst and Appelo, 2013), with the addition of dissolved Zn adsorption to HAO (Karamalidis and Dzombak 2010), and dissolved Al adsorption to HFO (calibrated for this study) The HFO surfaces available for trace-metal adsorption accumulate as FeII is oxidized and amorphous Fe(OH)3 precipitates to maintain

chemical equilibrium Likewise, HAO surfaces may accumulate as amorphous Al(OH)3

precipitates Two types of adsorption sites (strong and weak) on HFO (Dzombak and Morel, 1990) and one type of adsorption site on HAO [Al(OH)3(a)] (based on the surface parameters for

gibbsite given in Karamalidis and Dzombak, 2010) were included Thermodynamic data for aqueous speciation and solubility calculations were obtained from an extended PHREEQC database (Parkhurst and Appelo, 2013) (Supplementary Table S.3)

The PHREEQC adsorption model was based on the diffuse double-layer complexation equilibrium model (Dzombak and Morel, 1990) with multiple binding sites on surfaces, and the average concentrations in the diffuse layer calculated using the Donnan option

(Parkhurst and Appelo, 2013) The surface area of HFO was specified as 5.33x104 m2/moleHFO

(600 m2/g as FeOOH) with 0.005 molesites/moleHFO strong sites, and 0.2 molesites/moleHFO weak

sites (Dzombak and Morel, 1990) The surface area of HAO was specified as 2.495x103

m2/moleHAO (32 m2/g) with 0.033 molesites/moleHAO (Karamalidis and Dzombak, 2010) Although

Al adsorption by HFO is not included in Dzombak and Morel (1990), we adapted their Cr3+

model and estimated equilibrium constants for Al3+ adsorption to simulate empirical data for

surface-was also evaluated by incorporating Al and Zn as trace components in the HFO formula [e.g.,

Fe0.9985Al0.0009Zn0.0009(OH)3] (Supplementary Table S.3) With respect to the surface parameters,

surface area was increased or decreased by multiplying the initial value by factors of 0.01, 100, and 10,000 to determine the influence of HFO surface area on Al and Zn adsorption The number

of strong and weak sites were increased or decreased by multiplying the initial value by factors

of 0.1, 5, and 10 The supplementary data includes an evaluation of the effects of competing cations on Al and Zn adsorption, and address uncertainty in observed values The concentrations

of Al and Mn in the initial solution were increased or decreased by multiplying the initial value

by factors of 0.5 and 1.5 and the initial concentrations of Fe and Zn were increased or decreased

by multiplying the initial value by factors of 0.5, 0.75, 1.25, and 1.5 (encompassing all variation

in initial values, Supplementary Table S.1, Supplementary Data S.6) The applicability of the kinetics-adsorption model to natural in-stream chemical changes was assessed by comparing the simulated pH, alkalinity, and dissolved concentrations of O2, CO2, Fe, Al, and Zn to results for

samples from the West Branch Schuylkill River downstream of the Oak Hill Boreholes outfall (Cravotta et al, 2014; Cravotta, 2015)

3 Results 3.1 Water Quality The concentrations of dissolved Fe, Al, Mn, Zn, Co, Ni, and associated water-quality parameters (pH, DO, Eh, alkalinity, specific conductance, Pco2) changed to different extents

during the batch experiments (Figures 2 and 3) Over the course of the aeration experiments, substantial decreases of 60% or more were observed relative to initial dissolved concentrations

of Fe, Al, and Zn (Figure 2A, 2B, 2D) In contrast, minor decreases of 20% or less were observed relative to initial dissolved concentrations of Mn, Ni, and Co (Figure 2C, 2E, 2F) Although the H2O2 treatment caused the most rapid changes in dissolved Fe and Al

concentrations and pH, the overall changes in water quality were the most extensive for Aer 3, which exhibited the greatest increases in pH and DO and decreases in Zn and CO2 (Figures 2 and

3)

Dissolved Fe concentration exhibited the most rapid decrease in the H2O2 batch, followed

by Aer 3, Aer 2, Aer 1, and the control, in the order of highest to lowest aeration rate, respectively (Figure 2A) The effect of aeration was evident within 0.5 hours of the experiment, during which time dissolved Fe concentration decreased by 12% in Aer 3, 4.2% in Aer 2, and

Observed decreases in dissolved Al concentration during the experiments followed a similar order as those for Fe, with the greatest rate and extent of Al removal exhibited by the

H2O2 treatment, followed by greatest to least aeration rate (Figure 2B) (There was one outlier in

the Aer 2 experiment at 1 hour that was approximately 30% greater than the concentration in the previous sample taken 15 minutes earlier; this value was ignored for interpretation.) Comparing unfiltered and filtered samples, approximately 25% of the total Al (unfiltered samples) was initially present in the dissolved phase (filtered samples), and both the total and dissolved Al concentrations generally decreased during the experiments The decreased concentrations of total

Al and dissolved Al could result from the precipitation and settling of HAO and/or other solids containing Al (e.g Nordstrom and Ball, 1986) and/or the coprecipitation or adsorption of Al by HFO (e.g Bertsch et al., 1989; Winland et al., 1991)

Al; C, Mn; D, Zn; E, Ni; and F, Co during batch aeration experiments (Supplementary Table S.1) Error bars were based on replicate results for analysis of USGS standard reference water samples The dashed line segment for Al during Aer 2 (Figure 2B) indicates the concentration curve when the outlier in the sample at 1 hour is excluded

Figure 3 Changes in water-quality parameters in the control, Aer 1, Aer 2, Aer 3, and H2O2

treatments: A, pH [lines indicate continuously logged data (YSI gel-filled electrode), symbols indicate the field-measured pH of alkalinity samples (Orion Ross electrode)]; B, Pco2 [log(atm),

calculated from alkalinity, specific conductance, pH, and temperature]; C, DO (% saturation, continuously logged); D, specific conductance (µS/cm, continuously logged); E, alkalinity (mg/L

as CaCO3, field-measured); and F, Eh (V, continuously logged) Solid symbols are single-point

measurements corresponding to samples for alkalinity or metals concentrations

Dissolved Zn exhibited decreases in concentration in the order Aer 3, Aer 2, Aer 1, H2O2,

and the control (Figure 2D) Although Zn did not decrease for the control solution, the H2O2

treatment promoted removal of approximately 45% of the initial Zn during the first 0.5 hours, with no additional removal thereafter In Aer 3, Zn decreased by 81% within 2 hours, after which the concentration reached a steady-state In contrast, the dissolved Zn concentrations in Aer 2 and Aer 1 did not reach steady-state concentration and continued to decrease for the duration of the experiments

The overall trends and magnitude of change in dissolved metals concentrations were related to the changes in continuously logged water-quality constituents (pH, SC, DO, and Eh) (Figure 3) The continuously logged pH (YSI electrode) increased non-linearly for the control, Aer 1, Aer 2, and Aer 3, respectively, and decreased for the H2O2 treatment (Figure 3A) The

aeration batches exhibited an initial rise in pH due to CO2 outgassing followed by an interim

plateau in pH (constant value) at approximately 1 hour, 1.5 hours, and 2 hours for Aer 3, Aer 2, and Aer 1, respectively This interim plateau is hypothesized to result from a temporary balance

in the rates of production of CO32- ions by CO2 outgassing:

2 HCO3- = CO2(g) + CO32- + H2O (2)

and consumption of CO32- ions by the oxidation and precipitation of Fe:

Fe2+ + 0.25 O2 + 1.5 H2O +CO32- = Fe(OH)3 + CO2(g) (3)

Combining the two reactions, the observed decreases in dissolved FeII and alkalinity are

indicated without explicit involvement of H+ or CO32- ions:

After FeII oxidation and FeIII precipitation were complete, the pH continued to rise due to

continued CO2 outgassing from the net-alkaline water samples Note that if the water quality

were net acidic, the pH eventually could decrease to acidic values after the alkalinity had been consumed through FeII oxidation and Fe(OH)3 precipitation [as reported by Geroni et al (2012)]

Figure 3B shows the change in computed Pco2 with time for each experiment The largest

changes were observed for the aeration batches, with higher aeration rates resulting in more rapid outgassing of CO2, lower concentration of CO2, and greater pH The Aer 3 batch reached a

second plateau at pH 7.75 starting at 3.5 hours (Figure 3A), possibly due to calcite precipitation and the release of H+ ions,

Ca2+ + HCO3- = CaCO3 + H+, (5)

which buffered the pH while CO2 outgassing continued

The H2O2 and control treatments, which were gently stirred only when sampled,

exhibited small increases in DO during the batch experiments In contrast, the aeration treatments exhibited rapid increases to steady-state DO concentrations of 80 to 100% saturation while the pumps were active A decrease in DO to a lower steady state of about 80% saturation took place in Aer 3 at approximately 3.75 hours into the experiment when the aerator was turned off to permit settling of HFO; all dissolved FeII in Aer 3 had been oxidized prior to this time

(Figure 2A)

The change in specific conductance (SC) during the batch experiments corresponded to a decrease in dissolved ions associated with the removal of dissolved Fe and, to a lesser extent,

CO2 For example, while the SC of the control treatment remained relatively stable for the

duration of the experiment, the SC of the H2O2 treatment exhibited a sharp decrease (Figure 3F)

corresponding with the H2O2 addition and the rapid decrease in dissolved Fe concentration

(Figure 2A), with more moderate decrease observed over the course of the experiment for Aer 1, Aer 2, and Aer 3

Alkalinity exhibited small decreases during the experiments (Figure 3E) due to its consumption as FeII oxidized and precipitated as HFO (Eq 4) The largest changes in alkalinity

occurred in the Aer 3 and H2O2 batches The abrupt decrease in alkalinity for the H2O2

experiment resulted from the rapid removal of dissolved Fe indicated by the abrupt decrease in

pH at the start of the experiment In contrast, the gradual increase in pH for the Aer 3 experiment was accompanied by a slow decline in alkalinity, indicating the pH increase during CO2

outgassing was sufficient to compensate for proton release during Fe oxidation and hydrolysis The control, Aer 1, and Aer 2 batches exhibited smaller changes in alkalinity because Fe oxidation and precipitation were less extensive and occurred at slower rates for these treatments The H2O2 treatment exhibited a dramatic increase in Eh from 0.18 V to 0.55 V during the

first 15 minutes of the experiment, and then stabilized at approximately 0.55 V (Figure 3F) In contrast, the control exhibited relatively constant Eh, while Aer 1, Aer 2, and Aer 3 exhibited declines in Eh within the first 2 hours of the experiment, with faster rates of decline and lower minimum values associated with greater aeration rate After the dissolved FeII had been

completely oxidized in the aerated systems, the Eh for Aer 3 and then Aer 2 began to increase and approach the initial Eh value (Figure 3F)

3.2 Calibration and Sensitivity Analysis of Kinetics-Adsorption Model The PHREEQC kinetics-adsorption model was used to evaluate the potential effects of changing the pH, the concentrations of dissolved CO2, FeII, and associated solutes, and the

determined by Cravotta (2015) to simulate the observed changes in pH, dissolved O2, CO2, FeII,

and alkalinity during the aeration experiments (Aer 1, Aer 2, and Aer 3) were not adjusted further To simulate observed data for the control and H2O2 treatments, additional values for the

rate constants for O2 ingassing, CO2 outgassing, and FeII oxidation were estimated

(Supplementary Table S.4) To further calibrate the model, those parameters added to simulate changes in Al and Zn concentrations were adjusted to match the observed Al and Zn data for the Aer 3 experiment (Figure 2), including precipitation of common solid phases found in CMD, coprecipitation with HFO, and adsorption to HFO (Figures 4 and 5) The values for HFO surface parameters, equilibrium constants for Al and Zn adsorption to HFO, and the stoichiometry of HFO (coprecipitate) that resulted in the best fit of the empirical results for Aer 3 experiments were considered as the calibrated values

The potential controls on dissolved Al concentration that were evaluated included precipitation of Al(OH)3(a), gibbsite, kaolinite, or basaluminite; coprecipitation with HFO; and

adsorption to HFO (Figure 4) On the basis of its saturation index, Al(OH)3(a) was indicated to be

undersaturated for the duration of the experiment (Figures 4A and 4B); thus, theoretical equilibrium with Al(OH)3(a) overpredicted dissolved Al concentration compared to the observed

values (Figure 4C) Gibbsite and kaolinite were initially oversaturated but approached equilibrium and then exhibited oversaturation late in the experiment, as pH increased (Figures 4A and 4B) Likewise, basaluminite was initially oversaturated, but became undersaturated later

in the experiment Thus, kaolinite, gibbsite, and basaluminite equilibrium values for dissolved Al concentration initially underpredicted the dissolved Al concentration compared to the observed

the initial 0.5 hours of the experiment, Al coprecipitation underpredicted the Al concentration at 2.5 hours and later times when the Al concentration began to increase with pH (Figure 4D) The inclusion of Al adsorption on HFO combined with the potential for precipitation of Al(OH)3(a)

(to maintain SI < 0) produced the best fit to the observed Al concentrations (Figure 4E), including both the initial decrease and the later increase in dissolved Al, and therefore were included in the calibrated model

of Al(OH)3(a), gibbsite, kaolinite, and basaluminite across pH in Aer 3 samples; and B,

Al(OH)3(a), gibbsite, kaolinite, and basaluminite over time in Aer 3 samples Fine dotted lines in

A and B indicate where SI was not calculated due to non-detectible concentration of dissolved Al

in samples C-E, changes in dissolved Al relative to initial concentration (47 µg/L) (Ct/C0)

considering: C, Al(OH)3(a), gibbsite, kaolinite, or basaluminite solubility control; D, Al

coprecipitation with HFO; E, Al(OH)3(a), gibbsite, kaolinite, or basaluminite solubility control

with Al adsorption on HFO (strong sites log K = 1.0, weak sites log K = -2.1)

The kinetics-adsorption model initially specified the HFO surface characteristics and adsorption equilibrium constants for Zn from Dzombak and Morel (1990), and the estimated adsorption equilibrium constants for Al (this study) A sensitivity analysis of these parameters was conducted with comparison to the observed data for the Aer 3 experiment (Figure 5) For Al, changes in the HFO surface parameters, such as increasing surface area or the number of strong

or weak sites, only slightly improved the simulation of the initial decrease in Al concentration, but diminished the model fit of increased Al concentrations (desorption) at high pH values after 2.5 hours (Figures 5A-C) Furthermore, the addition of coprecipitation, such as 0.09% Al with HFO [Fe0.9991Al0.0009(OH)3] or combined Al and Zn coprecipitation with HFO

[Fe0.9985Al0.0009Zn0.0009(OH)3], resulted in a predicted Al concentration slightly lower than the

observed values after more than 3 hours of aeration (reaction), indicating that coprecipitation could decrease potential for adsorption or desorption of Al (Figure 5E) and was not actively occurring during the experiment Therefore, coprecipitation of Al with HFO was not included, and no changes were made to the HFO surface characteristics or adsorption equilibrium constants for Al in the calibrated model

For dissolved Zn, increasing the surface area or the number of strong or weak sites resulted in a greater removal of dissolved Zn compared to the Dzombak and Morel (1990) adsorption equilibrium parameters, and improved simulation (closer fit) of the initial decrease in

Zn2+ By decreasing the log K values for adsorption, a greater fraction of the Zn forms these

mobile carbonate complexes and remains dissolved (desorbed) Equivalent magnitude adjustments to the log K values for weak and strong sites indicated that weak sites had a larger effect on the total Zn adsorbed than the strong sites The addition of coprecipitation of 0.09% Zn with HFO [Fe0.9994Zn0.0009(OH)3] substantially improved the model fit to the observed values for

Zn concentration (Figure 5J); therefore, no changes were made to the HFO surface characteristics or adsorption equilibrium constants for Zn, but Zn coprecipitation [Fe0.9994Zn0.0009(OH)3] was included in the calibrated model (included as Supplementary Data S.7

and S.8)