Journal of Contaminant Hydrology 119 (2011) 55–68

Contents lists available at ScienceDirect

Journal of Contaminant Hydrology

j ourna l homepage: www.e lsev ie r.com/ locate / jconhyd

Geochemical interactions between process-affected water from oil sandstailings ponds and North Alberta surficial sediments

A.A. Holden, R.B. Donahue 1, A.C. Ulrich⁎Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, T6G-2W2, Canada

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +1 780 492 8293; faxE-mail address: aulrich@ualberta.ca (A.C. Ulrich).

1 Present Address: Applied Geochemical Engineerin91 Street, Edmonton, Alberta, T6C 3P6, Canada.

0169-7722/$ – see front matter © 2010 Elsevier B.V.doi:10.1016/j.jconhyd.2010.09.008

a b s t r a c t

Article history:Received 5 March 2010Received in revised form 14 September 2010Accepted 21 September 2010Available online 29 September 2010

In Northern Alberta, the placement of out-of-pit oil sands tailings ponds atop natural buriedsand channels is becoming increasingly common. Preliminary modeling of such a site suggeststhat process-affected (PA) pond water will infiltrate through the underlying clay till aquitard,reaching the sand channel. However, the impact of seepage upon native sediments andgroundwater resources is not known. The goal of this study is to investigate the role ofadsorption and ion exchange reactions in the clay till and their effect on the attenuation orrelease of inorganic species. This was evaluated using batch sorption experiments (traditionaland a recent modification using less disturbed sediment samples) and geochemical modelingwith PHREEQC. The results show that clay till sediments have the capacity to mitigate the highconcentrations of ingressing sodium (600 mg L−1), with linear sorption partitioning coeffi-cients (Kd) of 0.45 L kg−1. Ion exchange theory was required to account for all other cationbehaviour, precluding the calculation of such coefficients for other species. Qualitativeevidence suggests that chloride will behave conservatively, with high concentrationsremaining in solution (375 mg L−1). As a whole, system behaviour was found to be controlledby a combination of competitive ion exchange, dissolution and precipitation reactions.Observations, supported by PHREEQC simulations, suggest that the influx of PA water willinduce the dissolution of pre-existing sulphate salts. Sodium present in the process-affectedwater will exchange with sediment-bound calcium and magnesium, increasing the divalentions' pore fluid concentrations, and leading to the precipitation of a calcium–magnesiumcarbonate mineral phase. Thus, in similar tailings pond settings, particularly if the glacial tillcoverage is thin or altogether absent, it is reasonable to expect that high concentrations ofsodium and chloride will remain in solution, while sulphate concentrations will exceed those ofthe ingressing plume (150 mg L−1).

© 2010 Elsevier B.V. All rights reserved.

Keywords:Process-affected waterIon exchangeGroundwaterSeepageOil sands tailings pond

1. Introduction

With an estimated 178 billion barrels of recoverable oil,reserves in Alberta's oil sands are second only to those of SaudiArabia. At present, only a fraction of the initially establishedreserves have been recovered (~3%) (Government of Alberta,

: +1 780 492 0249.

g Solutions Inc., 9535-

All rights reserved.

2006) and the industry is growing rapidly, with productionquadrupling in the past decade, and expected to further risefromover onemillionbarrels per day in2008 to 3 millionbarrelsper day by 2015 (Alberta Sustainable Resource Development,2009).

Every 1 m3 of synthetic crude oil produced generates 1.5 m3

of mature fine tailings (a mixture of fines and process-affected(PA) water) (Alberta Chamber of Resources, 2004). PA watertypically includes high concentrations of dissolved ions as wellas organic chemicals of concern. The latter includes naphthenicacids, BTEX, and polycyclic aromatic hydrocarbons, which arereleased during ore processing and become concentrated,

56 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

making the water toxic to many aquatic organisms. Dissolvedions are primarily comprised of sodium, chloride, sulphate andbicarbonate (Allen, 2008), which accumulate from ore extrac-tion chemicals (most notably caustic (NaOH) water), tailingstreatments (e.g. CaSO4 to enhance settling rates) (Chalaturnyket al., 2002; Allen, 2008) or may be released from the oil sandsores themselves — some of which are of marine origin andhighly saline (a detailed review of tailings composition isprovided by Kasperski, 1992). Because oil sands operationspresently adhere to a policy of zero water discharge to theenvironment, process-affected water cannot be released andmust instead be stored in large settling ponds— resulting in aninventory of PA water that is large and growing. As of 2008, itwas estimated that the province's inventory offluidfine tailingsrequiring long-term containment was already 720 million m3

(ERCB, 2009).In Northern Alberta, the South Tailings Pond is the first of

several tailings facilities to be sited atop buried sand channels—relicts of previous glacial rivers. Unchecked, these channelshave the capacity to act as potential migration pathways for PAwater, facilitating the release of tailings pond seepage intoadjacent surface or ground waters. Preliminary modeling of acase study site suggests that PA water from the pond willinfiltrate through the clay till and into the underlying sandchannel; however, the development and extent of this impactare not known (Hall et al., 2004).

As part of a larger study, the research presented hereinvestigates the geochemical impact of oil sands tailingsseepage on native surficial sediments and groundwaterresources. The specific objectives of this work are:

1. To investigate the effects of adsorptive and ion exchangereactions, in the clay till and sand channel sediments, onmitigation/mobilization of major inorganic cations. Thiswill be done using both traditional batch sorption experi-ments, and additionally, a recent modification of thetraditional method, featuring less disturbed sedimentsamples; and

2. To further understanding of system behaviour throughgeochemical simulations.

This is thefirst study to assess themitigating capacity of FortMcMurray surficial sediments, tomajor ionspresent in process-affected seepage waters, by batch sorption experimentation.Similarly, there exists only sparse information concerning theclay mineralogy, cation exchange capacity (CEC), and ex-changeable cation composition of surficial, Pleistocene clay tillsin Northern Alberta. However, in-house exchangeable cationexperiments (Husz, 2001; Sheldrick, 1984; USEPA, 1986) onthese glacial till sediments, suggest that in decreasing abun-dance, the easily displaceable cations consist of calcium,magnesium, sodium and potassium. Furthermore, PA water isknown to consist of high concentrations of sodium, chloride,sulphate and bicarbonate ions (Table 2; Marsh, 2006; Oifferet al., 2009; Price, 2005). It is therefore expected that theinteraction of PA water with the native sediments and porewater will result in the sorption of sodiumwith the subsequentdisplacement of pre-bound cations such as calcium andmagnesium. Depending on the saturation indices of mineralphases in the resultant solution, this may lead to precipitationof sulphate or carbonate minerals.

Research to date concerning the geoenvironmental impactof oil sands tailings seepage has largely focussed on naph-thenic acids. Janfada et al. (2006) and Peng et al. (2002) bothinvestigated the adsorption behaviour of naphthenic acidsthrough laboratory batch sorption experiments, using organ-ic-rich Fort McMurray soils. Oiffer et al. (2009) conducted afield investigation of the fate and transport of naphthenicacids, andof the potential for arsenicmobilization, by aweaklyreducing PA water plume, within a portion of a shallow sandaquifer, immediately adjacent to a Fort McMurray oil sandstailings pond. Tompkins (2009) recently evaluated thecapacity for natural attenuation of naphthenic acids, andalso for trace metal mobilization, through a series of aquiferinjection experiments into a sand channel underlying anotheroil sands tailings impoundment in Fort McMurray. Based onexisting literature, the present study offers the first detailedlaboratory characterization of the adsorption and ion ex-change processes to be expected when ingressing inorganiccomponents present in PA water interact with surficialsediments in the Athabasca Oil Sands region.

1.1. Site description

The Suncor Energy Inc. oil sands lease is situated approx-imately 35 km North of Fort McMurray, Alberta. Irregularlyshaped, the South Tailings Pond is approximately 23 km2 inarea: 4 km North to South, and 5 km East to West. Approxi-mately 8 km of the South and West dyke walls, and 50% of theSouth Tailings Pond footprint, have been constructed atop aPleistocene glaciofluvial outwash channel, the Wood CreekSand Channel (WCSC) (Klohn Crippen Consultants Ltd., 2004).

To date, the South Tailings Pond has been designated forstorage of fine tailings and process-affected water from thePlant 8C process stream, with placement having commencedJune 2006 (Stephens et al., 2006; personal communicationwith Suncor Energy Inc.). The South Tailings Pond has anapproved, current tailings holding capacity of 230 Mm3, butwith application under review to increase the dyke elevationto allow for an ultimate tailings holding capacity of 350 Mm3.

In descending order, the surficial geology consists of 1–2 mof surficial muskeg deposits, 8–15 m of Pleistocene glacial till,and approximately 30 m of WCSC sands. Although thethickness of the glacial till deposit is generally uniform, thin(b5 m) or the altogether absence of coverage has beenobserved at several locations across the site, which couldpotentially create a direct hydraulic link between the Pondand the WCSC (Klohn Crippen Consultants Ltd., 2004). TheWCSC lies atop clay shales of the Cretaceous ClearwaterFormation, although in some places it cuts below into theunderlying McMurray Formation. Physical connection to thislatter bedrock formation, which contains the oil sandsdeposits, may influence the groundwater geochemistrywithin the sand channel.

2. Materials and methods

2.1. Sediment samples

Soil cores were collected during the construction of agroundwater monitoring well research transect, between April11 andMay7, 2006. The transect is located across the dykewall

2 Synthetic process-affected water was comprised of Fisher Scientific(USA) reagents: NaCl (certified A.C.S. crystal), Na2SO4 (certified A.C.S.anhydrous), NaHCO3 (certified A.C.S), KCl (certified A.C.S) and ACROSOrganics (ES) CaSO4–2H2O (A.C.S.) and BDH Chemicals (Canada) MgSO4–

7H2O (A.C.S.).

57A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

at the Northwest edge of the South Tailings Pond. Boreholeswere drilled using a 0.11 m casing on a 1503 Nodwell SONICDrill Rig operated by SDS Drilling. Sleeves made from Lexan, apolycarbonate resin, were used to line the drill stem such thatcores were directly encapsulated as drilling progressed. At thesurface, lined cores were further capped to preserve thesubsurface conditions. Samples were then shipped to theUniversity of Alberta Applied Geoenvironmental ResearchFacility and stored at −20 °C. A detailed site characterizationis presented in Holden et al. (2009).

Core sections for laboratory use were cut using a radial armsaw, then wrapped in cling wrap and aluminum foil, sealed inair-tight plastic bags, and finally placed in cold storage (4 °C) tothaw for 24 h. Moisture loss was monitored by mass measure-ments made before and after thawing. Laboratory experimentsutilized sediments in their native moisture content in order toavoid dissolution complications with sparingly soluble salts.Such salts, for example gypsum, are thought present basedupon previously conducted saturated paste extraction experi-ments using these same sediments. Initially dissolved withinthe pore fluid, these salts may not re-dissolve sufficientlyquickly, or to their initial extent, if the sediment is first driedand then re-hydrated. Gravel, although rarely encountered,wasremoved by hand from the small experimental samples toavoid underestimating interactions or properties. To investi-gate variations in sediment responsewith depth, sampleswereselected from the upper, middle and lower glacial till, from thelower portion of the clay till-WCSC transition zone, and foradditional interest, from deep in the WCSC. The respectivedepths are: 3.4–4.6 m below ground surface (mbgs) (uppertill), 6.1–6.4 mbgs (mid-till), 8.2–9.1 mbgs (low till), 12.2–13.7 mbgs (till-WCSC transition), and 37.7 and 38.7 mbgs(deep WCSC).

Select samples were analyzed for fraction of organiccontent by collaborative researchers at the University ofWaterloo. During preliminary grain size analysis, it becameapparent that small silt/clay clusters remained followingdrying and gentle grinding, which would have been incor-rectly retained by sieves in the sand-sized fraction. Thereforesamples were first washed using a 0.075 mm sieve, therebyremoving the silt/clay coatings on sand particles, or silt/clayclusters themselves, without subjecting true sand particles tocrushing by grinding. Two of the sediment samples whichwere qualitatively observed to have a high clay content, werealso analyzed using a hydrometer, to further resolve the grainsize of the fines fraction. Sediment properties are presented inTable 1.

2.2. Ion chromatography

Aqueous lithium, sodium, ammonium, potassium, mag-nesium and calcium concentrations were quantified using aDionex ICS-2000 Ion Chromatography system. 25.0μL ofliquid sample is first injected into the sample loop (DionexAS50 Autosampler) and thereafter merged with an isocraticeluent stream (20.00 mM Methanesulfonic Acid). The eluentand sample are pumped through guard (IonPac® CG12A) andseparator columns at a flow rate of 1.00 mL min−1. A DionexCSRS300 4 mm suppressor preferentially enhances thesample ions' signal relative to the eluent conductivity, withanalyte quantification ultimately performed by Chromeleon

Client v6.50 software. The mean detection limit is less than0.2 mg L−1 for each of these six cation species.

Aqueous fluoride, chloride, nitrite, bromide, nitrate,phosphate and sulphate concentrations were characterizedusing a Dionex 2500 Ion Chromatography system. 25.0μL ofliquid sample is first injected into the sample loop (DionexAS50 Autosampler) and thereafter merged with an isocraticeluent stream (8.0 mM Na2CO3+1.0 mM NaHCO3). A DionexGP50 Gradient Pump forces the eluent and sample throughguard (IonPac® NG1 and IonPac® AG14A) and separator(IonPac® AS14A) columns at a flow rate of 1.00 mL min−1. ADionex ASRS300 4 mm suppressor preferentially enhancesthe sample ions' signal relative to the eluent conductivity,with signal measurement and analyte quantification per-formed by a Dionex CD25 Conductivity Detector andChromeleon Client v6.50 software respectively. The meandetection limit is less than 0.3 mg L−1 for each of these sevenanion species.

Five calibration standard solutions analyzed at the outsetof each sequence enable Chromeleon characterization of theliquid sample (Dionex, Seven Anion Standard and Six CationStandard II). A quality control check standard and blank arerun after these calibration standards and after approximatelyevery 10 to 15 experiment samples.

2.3. Batch sorption experiments

In order to investigate the utility of a recently proposedmethod of laboratory sorption experiments, and in order toexamine thegeochemical systembehaviour at an increasing levelof complexity, a total of four different treatments were applied:

□ standard batch sorption tests using a simplified, simulatedprocess-affected water;

□ standard batch sorption tests using process-affected watercollected on site;

□ a new method of sediment cube sorption tests (modifiedfrom Zhang et al., 1998) using simulated process-affectedwater; and

□ sediment cube sorption tests using process-affected watercollected on site.

Process-affected water was collected from a discharge linefeeding directly into the South Tailings Pond. Samples werecollected August 25, 2008 and stored sealed, in the dark, at 4 °C.In addition, a simplified synthetic or simulated process-affectedwater was created, based upon the chemical composition of STPprocess-affected water averaged across 6 sampling eventsbetween September 2006 and August 2008. This water wascomprised only of major ions, and in order to evaluate systemresponse at future, potentially higher discharge concentrations,the concentration was conservatively selected to be 1.2 timesthat of Suncor PA water averaged across these dates2 (Table 2).The simulated process water possesses similar ionic character-istics to the actual process water but does not contain the

3 Samples were preserved in 1% nitric acid, and analysed via PerkinElmerSCIEX ELAN 9000 — results not shown.

4 Plastic containers were deemed suitable given our focus on inorganicsand the short-term 24 hour duration.

Table 1Physical properties of the tested sediments.

Sample ID Soil coresamplinglocation d

Cation exchangecapacity c

(meq(100 g)−1)

Specificgravity(20 °C)

Grain sizesummary

Gravimetricwater content

% FOC byweight

Description from borehole logs

MW3-4 Monitoring well3B, 3.4 to 4.6 mbgs

11.32 2.614 0.5%N0.475 mm 0.142 1.530±0.057 Clay (till), silty, sandy, low-medium plastic, brown, moist67.5%b0.475 mm

N0.075 mm32.0%b0.075 mm19.2%b0.0023 mm

MW3-6 Monitoring well 3B,6.1 to 6.4 mbgs

4.02 2.633 4.0%N0.475 mm 0.121 3.690±0.028 See above (experimentally notedto be sandy, red-brown comparedto other till samples)

70.3%b0.475 mmN0.075 mm

25.6%b0.075 mmMW3-9 Monitoring well

3B, 8.2 to 9.1 mbgs15.50 2.643 0.6%N0.475 mm 0.124 0.986±0.018 a See above

45.3%b0.475 mmN0.075 mm

54.0%b0.075 mm30.0%b0.0022 mm

MW4-13 Monitoring well 4B,12.2 to 13.7 mbgs

3.51 2.660 3.7%N0.475 mm 0.056 0.978±0.028 b Transition zone, clay till to sand(silty)74.4%b0.475 mm

N0.075 mm22.0%b0.075 mm

MW2-38 Monitoring well2A, 37.7 mbgs

6.50 2.669 9.0%N0.475 mm 0.169 not tested Sand, dark grey, fine-grained, verysilty, wet, some fine gravel, tracecoarse gravel, occasional pebblesub-angular, occasional CL clay lens4–6″ thick with high amount of gravel(till-like)

42.5%b0.475 mmN0.075 mm

48.6%b0.075 mm

MW2-39 Monitoring well2A, 38.7 mbgs

5.59 2.733 19.1%N0.475 mm 0.103 not tested Sand/gravel, dark grey, fine tomedium-grained, clean, gravelmedium to coarse-grained,occasional small cobble

56.9%b0.475 mmN0.075 mm

24.0%b0.075 mm

FOC=fraction of organic content, mbgs=meters below ground surface.Grain size distribution curves are numerically summarized to show the major features.

a Tested sample came from a marginally lower elevation of 9.1 to 9.8 mbgs.b Tested sample came from a marginally higher elevation of 11.0 to 12.2 mbgs.c Sodium Acetate method (USEPA, 1986).d Nests of monitoring wells are situated linearly, with approximately 100 m spacing from 2A to 3B and again from 3B to 4B.

58 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

complex organic compounds. A comparison of the simulatedprocess water results to the actual process water results shouldtherefore provide some insight into the role of organiccompounds in the attenuation and release of inorganic species.

2.3.1. Traditional methodSorption and ion exchange behaviour were investigated as a

function of process-affected water concentration. Six differentaqueous solutions were used: deionized water as well as 5different concentrations of PAwater. A concentration rangewasused as a means of modeling the geochemical response to anadvancing plume of PA water. For simulated PA water, themaximum concentration was conservatively selected to be 1.2times the average concentration of Suncor PAwater observed todate. The concentrations of the 4 intermediate stock solutionswere selected to be 20%, 40%, 60%, and 80%, by volume, of thisproposed maximum. Similarly, South Tailings Pond-destined orhereafter known as real PAwater stock solutionswere preparedto 20%, 40%, 60%, 80% and no dilution or 100% strength of thewater collected on site. Immediately after preparation, a portionof each stock solution was retained for pH (Thermo ElectronCorporation, pH electrode 9107BN) and alkalinity analysis(potentiometric titration), while additional samples werefiltered (0.2 μm) and major ion and trace metal speciation

ascertained by Ion Chromatography (see 2.2.) and InductivelyCoupled Plasma Mass Spectrometry,3 respectively.

Based upon preliminary experiments to optimize theprocedure, batch samples were created at a soil to water ratioof 1:2. Fifteen grams of undried sedimentwere added to 30 g ofstock solution, within a 50 mL plastic centrifuge vial,4 andmixed into a slurry using a glass stirring rod. Order ofpreparation of batch vials was randomized across solutionconcentrations. Samples were prepared quickly to minimizesediment desiccation. Vials were agitated at room temperature(20.0 °C) for 24 h (ASTM Standard D4646, 2003) using a wrist-action shaker (BurrellWrist-Action®Laboratory Shaker,ModelBB, 10° amplitude, 400±20 Osc.min−1). Subsequently, sam-ples were centrifuged at 4150 rpm for 15 min (HeraeusMultifuge® 3 L-R). Any changes to the physical characteristicsof either the solid phase or solution such as colour, opacity, etc.were noted. A portion of the resultant supernatant wasanalyzed for pHand alkalinity,while the remainderwasfiltered

Table 2Chemical composition of full strength process-affected waters used in batchsorption experiments (average of source solutions used for each of the fivesediment regions' samples) [units: mg L−1].

Species Simulated PA water Real PA water

Calcium 16.07 7.62Magnesium 12.14 4.17Potassium 12.59 10.05Sodium 720.62 591.35Ammonium 0.47 2.02Lithium Not added 0.16Chloride 398.00 374.51Fluoride 0.01 2.21Bromide Not added 0.93Nitrite Not added Not detectedNitrate Not added 4.65Phosphate Not added Not detectedSulphate 228.03 150.39Alkalinity (as CaCO3) 709 491pH 8.3 8.6Total organic carbon Not added 83

59A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

(0.2 μm) and major ion and trace metal speciation ascertainedby Ion Chromatography and Inductively Coupled Plasma MassSpectrometry, respectively. Tests were conducted in triplicate.

A sample of the soil core was placed in a drying oven forgravimetric water content analysis before the first batch vialwas prepared, and another after the final vial was created. Twofluid-only blanks were prepared at each concentration — onebefore batch vials were created, one after — and subject to thesame treatment as the sediment–water vials to measure non-adsorptive losses to the system. Mass measurements weretaken at all stages of the experiment to further record any suchlosses.

Following final decantation and conclusion of the sorptionexperiment, a simple desorption procedurewas applied to theremaining solid phase. The procedure reflects the latterportion of the Sodium Acetate cation exchange capacitymethod (USEPA, 1986); in short, three 33 mL rinses with2-propanol Optima (Fisher Scientific) followed by three rinseswith a prepared 1 N ammonium acetate solution (pH 7.0).

2.3.2. Cube batch sorption methodA modification of a recently proposed method was

critically evaluated, as a means to investigate sorption andion exchange behaviour as a function of process-affectedwater concentration, by diffusive equilibration of an intactsediment sample (Zhang et al., 1998). The details of theapplied stock solution concentrations are the same as above(see 2.3.1.).

Samples were prepared by carefully cutting a 15 g cube ofundried sediment, with approximate dimensions of2×2×2 cm, from the soil core section, using a stainless steelpalette knife. The procedure was designed so that both thetraditional and cubemethods would utilize the same sedimentsample size and soil to water ratio, to facilitate latercomparisons between results from the two methods. Pre-tests were conducted at 5 day, 6 day, 7 day, 8 day and 11 daydurations andapreliminary simulationwas runusingPHREEQC(Parkhurst and Appelo, 1999) in order to establish the length oftime needed for the system to equilibrate, using a cube of these

dimensions. The simulation considered 1D diffusive, non-reactive transport, for sodium and chloride, at their concentra-tions within full strength PA water (0.026 mol kgs−1 and0.010 mol kgs−1 respectively). The diffusion coefficient wasestimated to be 10−10 m2s−1 and the diffusion path length for1D transport was assumed to be half the length of the cube (i.e.assuming diffusionwould occur from each side of the cube intoits centre). Results from both laboratory analyses and thesimulation predicted that equilibration would be achieved in5 days for a cube of these dimensions.. Sediment cube locationwithin the core section and order of preparation of the batchsamples were randomized across solution concentrations.Given the greater duration of this experiment, 5 days and not24 h, glass jars (Fisher Scientific, I-Chem 200series 4 oz, clear)were used to minimize the sorption of organics onto containerwalls, which could have altered the inorganic system response.The undried sediment cube was placed in a jar and 30 g ofsolution carefully added using a syringe (B-D 20 cc luer lok,bulk, non-sterile — Fisher Scientific). The jar was then sealedand left in the dark, at 20.0 °C, for 5 days. After this period, anychanges to the physical characteristics of either the solid phaseor solutionwere noted. The equilibrated solutionwas decantedand analyzed for pH, alkalinity, and major ions by IonChromatography and trace metals by Inductively CoupledPlasma Mass Spectrometry. A portion of the soil core wasdried at 100 °C for 24 h for gravimetric water content analysis.A fluid-only control was prepared at each concentration. Massmeasurements were conducted throughout the experiment torecord evaporative losses to the system. Tests were conductedin triplicate.

2.3.3. Batch sorption analysisLinear and Langmuirian sorptionmodels are commonly used

to analyze batch sorption data. The linear sorption model isspecified as:

Kd = Cs= Cw ð1Þ

with Kd representing the non-equilibrium 24 hour distributioncoefficient (ASTM Standard D4646, 2003), Cs the solute masssorbed per unit mass of dry soil (mg kg−1) and Cw, the soluteconcentration remaining in solution (mg L−1).

The Langmuirian model in turn, is given by:

Cs = bNmaxCwð Þ= 1 + bCwð Þ ð2Þ

where b is a constant, and Nmax the maximum possiblesorption by the solid (Langmuir, 1997).

Mass balance calculations accounting for non-adsorptivelosses, and initial pore fluid and added PA solution chemistrieswere used to provide context for final, observed, equilibriumconcentrations. Analysis was conducted as a two-step process:

1) Is sorption behaviour sufficiently represented by either thelinear or Langmuirianmodels, and overwhat concentrationrange of PA solutions?

2) If neither model is suitable, do other models exist that areable to better account for the observed data?

60 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

3. Results

3.1. Batch sorption experiments — control case

Fluid-only controls were used to quantify non-adsorptivelosses to the system. In cube batch experiments, the differencebetween the initial stock solution and final control concentra-tions was less than 10% for each of the principal cations:potassium, sodiumandmagnesium, but not for calcium. In thetraditional (slurried) batch experiments, the differencebetween the initial stock solution concentration and thefinal concentration, averaged from the before and aftercontrols, was typically less than 5% for potassium, and sodium.The difference was less than 5% for magnesium in simulatedPA solutions but less than 10% for magnesium in real PAsolutions, and the difference was greater than 10% for calciumin both solutions. Together, this suggests that linear partition-ing coefficients calculated for calcium may be unreliable(ASTM Standard D4646, 2003). Non-adsorptive losses wereaccounted for within the mass balance calculations.

3.2. Batch sorption experiments — deionized case

Batch sorption samples created using deionized water areconceptually similar to saturated paste extractions (Janzen,1993), and serve to identify the distribution of chemical speciesavailable to interact with ingressing PA seepage. Each of thesediments tested showed that interactionwith deionizedwaterled to the dissolution of calcium and magnesium sulphate andcarbonate salts. The concentration of released ions wasgreatest in the sand channel, with the most abundant cationand anion species being: Ca2+ 9.09 meq kgdry sediment

−1 and SO42−

10.15 meq kgdry sediment−1 (averaged across cube and traditional

trials, from MW2-38 and MW2-39 combined). Conversely, theconcentration of released ions was smallest in the lower clay tilland adjacent till/sand interface region. For example, in thelatter, the most abundant cation and anion species were: Ca2+

2.12 meq kgdry sediment−1 and alkalinity 2.71 meq kg−1

dry sediment (av-eraged across cube and traditional trials).

These results support earlier, unpublished findings fromsaturated paste extractions performed at the University ofAlberta on these same sediments (2006). Those studies notedthat with increasing depth from the upper clay till through tothe base of the Wood Creek Sand Channel: a) there was ageneral increase in the concentrations of salts released intosolution, attributed primarily to the dissolution of calcium andmagnesium sulphates, although calcium and magnesiumcarbonate dissolution was also proposed in some sections ofthe till; and b) anions became increasingly dominated bysulphate with an increasingly marginal relative contributionattributable to alkalinity.

For the Deionized case, cube batch samples tended to showan identical, albeit subdued response when compared totraditional (disturbed) batch samples; approximately 70% ofthe magnitude of the traditional samples' response. This likelyreflects the fact that sample stirring and shaking inherent to thetraditional experiment procedure exposes a greater sedimentsurface area available for reaction, and may also introduceparticle grinding and milling effects (Burgisser et al., 1993),which cause an increased amount of dissolution to occur.

3.3. Batch sorption experiments — zero to full strength PA cases

Cation sorption was best represented by a linear isothermonly for sodium, and selectively for ammonium and lithium.The sodium responsewas linear from 0 to full strength (100%)PA water, with possible evidence of approaching a Langmuir-ian plateau at full strength. Sediment partitioning coefficientsfor sodium are tabulated below (Table 3). Lithium was wellcharacterized by simple linear desorption for all solutionstrengths, although only for simulated PA experiments.Samples mixed with real PA water, which includes aqueouslithium, displayed more complicated sorption behaviour thansamples mixed with lithium-devoid, simulated PA water. Theammonium response was variable, but showed evidence oflinear desorption from 0 to 100% for simulated PA samplesand also occasionally for real PA, traditional batch samples.

For the majority of cation species, including calcium,magnesium and potassium, neither the linear nor Langmuirianmodels suitably captured the system behaviour. For example,although calcium displayed a relatively linear desorptionisotherm, it was most often with non-zero y-intercept so thatthe simple linear model no longer applied. Calcium concentra-tions from 0 to 100% PA strengths were often clustered about aregion apart from the origin, including those from lowconcentration input stock solutions, reaffirming the presence ofsoluble salts, which complicate system behaviour beyond simpledesorption. Fig. 1 illustrates an example dataset showing cationsbothwell and poorly characterized by the linear sorptionmodel.

The experimentally-determined, final, equilibrium solu-tions were modeled using PHREEQC (Parkhurst and Appelo,1999), to ascertain their degree of saturation with respect tocarbonate and sulphatemineral phases. The results consistentlyfound equilibriumsolutions to be oversaturatedwith respect tocalcite and dolomite, at all PA strengths except, usually, theDeionized case. Solutions from all PA strengths tended to bevery near the saturation point of magnesite, typically crossinginto oversaturationwhen higher PA strengthsweremixedwiththe sediment. All solutions were well undersaturated withrespect to gypsum. These results suggest that regardless of PAstrength, in reaching the equilibrium state, carbonate saltswould be precipitated out of solution, while sulphate salts, ifpresent, would dissolve, being released into solution.

The trends of equilibrium solution ion concentrations withincreasing PA strength were examined for each batch samplecreated. Mass balance calculations permitted a comparison ofa predicted final concentration, based upon non-adsorptivelosses and initial porefluid and added PA solution chemistries,with theobserved,final equilibriumconcentrations. A positivevalue for the difference between the predicted and observedvalues is indicative of uptake of that ionwhile a negative valuerepresents release, for example, by dissolution or desorption.

Consistent findingswere observed in the ion trends— acrosssediment treatment (cube vs. traditional disturbed), PA solution(real vs. simulated) and sediment type (clay vs. sand). Chlorideshowed evidence of a constant, near-zero response across PAstrengths, suggesting neither its uptake nor further chloriderelease. Sulphate showed evidence of a constant to slightlyincreasing extent of release with PA strength. The amount ofsulphate released wasmuch greater than the chloride response,but smaller than the calcium or magnesium release, except inthe Sand Channel. Magnesium typically showed a constant

Table 3Linear sorption partitioning coefficients (Kd) for sodium [units: L kg−1].

Sample ID Geology Real PA solution, traditionalbatch experiment

Simulated PA solution,traditional batch experiment

Real PA solution,cube batch experiment

Simulated PA solution,cube batch experiment

MW3-4 Upper clay till 0.51±0.01 0.57±0.08 0.46±0.03 0.44±0.04MW3-6 Mid clay till 0.26±0.06 0.15±0.04 0.25±0.03 a

MW3-9 Lower clay till 0.58±0.07 0.60±0.03 0.63±0.01 0.73±0.01MW4-13 Clay/sand interface 0.12±0.01 0.17±0.07 0.16 b 0.17±0.15MW2-38 Deep sand channel 0.22±0.01 d 0.27±0.02 d c c

MW2-39 Deep sand channel Not measured 0.14±0.04 d c c

a Data was non-linear.b Only one sample so no standard deviation possible.c Sediment too cohesionless for cube batch tests.d Isotherm possessed non-zero intercepts.

61A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

release across all PA strengths. Calcium behaviour was similar,with a constant to slightly increasing extent of release with PAstrength, although at approximately twice the magnitude ofmagnesium. By contrast, sodium showed evidence of increas-ingly large uptake with PA strength, beginning with a near-zeroresponse or very slight release in the Deionized case. Thealkalinity response tended to begin with release at low PAstrengths, quickly switching into often large uptake by high PAstrengths. Here, again the Sand Channel response was slightlydifferent, showing a more significant initial release andsubsequent reduced uptake of alkalinity-based species withincreasing PA strength (Fig. 2).

4. Discussion

4.1. Linear sorption

In descending geological order, the linear sorption parti-tioning coefficients for sodium range from about 0.45 L kg−1 inthe till, to0.15 in the interface regionand0.19deep in theWoodCreek Sand Channel (WCSC). This suggests that significantsodium attenuation may be possible in the till. However, if, assuspected, theWCSC acts as a preferential pathway for seepagefrom the South Tailings Pond, most incoming ions in fullstrength process-water seepage, where sodium concentrationsnear 600 mg L−1, will remain in solution, and unchecked, mayimpact neighbouring aquatic environments.

Neither linear nor Langmuirian sorption was able toaccount for the behaviour of most ions in the laboratoryexperiments; however, important clues to their behaviour canbe inferred from simple plots of the equilibrium solutionresponse versus input PA strength (Fig. 3). Across all samples,plots revealed strikingly similar behaviour between the trendsof sulphate and the more abundant calcium — and also to alesser extent, magnesium. Plots further show that all threespecies are released and that their concentrations remainrelatively constant despite interaction with increasinglyconcentrated PA water. Together, these behaviours suggestthe dissolution of pre-existing salts within the sediments:calcium, and to a lesser extent, magnesium sulphates.Saturation indices of resultant, equilibrated solutions supportthis interpretation as gypsum was undersaturated at all PAstrengths. Finally, the minor increases in the amount ofcalcium and sulphate released with PA strength that wereobserved could reflect a greater extent of salt dissolution athigher ionic strength.

The chloride responsewas constant and near zero across PAstrengths. In other words, the predicted final concentration,based only upon mixing the pore fluid with added stocksolution chemistries, suitably predicted the experimentalequilibrium solution composition. This suggests that chlorideions are neither uptaken, nor released through interactionwithnative pore water or sediments and hence that incoming ions,present within a contaminant plume, will remain in solution.Given the high concentrations of chloride in the PA water(375 mg L−1), unchecked, conservative chloride, like sodium,may therefore impact surrounding aquatic environments.

The plotted alkalinity trends, corroborated by the saturationindices of calcite in the equilibrated solutions, suggest thatexcept at low PA strengths, alkalinity is lost from the system—

believed to be by the precipitation of carbonates. At lowstrengths, for example in theDeionized case, both the plots andthe saturation indices instead suggest alkalinity release— likelythe dissolution of pre-existing carbonate salts.

A proper understanding of the divalent cations' behaviouris not immediately obvious from the graphs. Comparing iontrends with increasing PA strength, calcium behaviour, mildrelease, is opposite to the strong sorption of sodium, and theamount of calcium and magnesium released is greater thancan be attributed to sulphate salt dissolution alone — both, asion exchange would require. Confusion arises though, as thetrends suggest calcium release, while saturation indicesinstead propose calcite precipitation. However, if the divalentions' responses are combined, correcting for the quantitybelieved released by sulphate salt dissolution, and usingprecipitated carbonate (alkalinity) to estimate additionalcalcium and magnesium that may have been exchanged,then precipitated, the resultant curve (Ca+Mg–SO4+Alk) isalmost identical to the sodium response. Therefore, it ishypothesized that calcium and magnesium are released intosolution both by dissolution of pre-existing sulphate salts andby sodium-induced desorption, followed by their limitedprecipitation as a carbonate mineral phase.

4.2. Broader implications — is ion exchange theory a sufficientmodel to predict system behaviour?

Ion exchange between sodium and calcium may berepresented as follows:

Naþ + 1=2Ca−X2 = Na−X + 1=2Ca2+ ð3Þ

Fig. 1. Applying linear sorption isotherms to experimental data forMW3-4 (upper till) batch cube samplesmixedwith simulated process-affectedwater (sodium— topcalcium — middle, and magnesium — bottom). The series shown are method triplicates.

62 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

with the equilibrium ion exchange coefficient characterizingthe reaction,

KNa=Ca =Na−X½ � Ca2þ

h i0:5

Naþ½ � Ca−X2½ �0:5 ð4Þ

,

The square brackets represent activities, or effective concentra-tions, either in solution or bound to exchange sites (X).

Solution activities at the low ionic strengths relevant tothese experiments (I~0.04) may be approximated using theDebye–Huckel equation,whereas the activities of exchangeablecations may be determined by following the Gaines Thomasconvention. The latter involves calculating the equivalent

Fig. 2. Example trends of equilibrium ion concentrations with applied solution strength for MW3-9 (lower till) sediments: (A) simulated process-affected water,traditional batch sample, sample #1 in triplicate series; and (B) real process-affected water, cube sample, sample #1 in triplicate series).

Applied Solution Strength (as % of Full Strength Process-Affected Water)

0 20 40 60 80 100Pre

dict

ed M

inus

Obs

erve

d Q

uant

ity (

mill

iequ

ival

ents

)

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20ClSO4AlkalinityCaMgNaCa+Mg-SO4+Alkalinity

Fig. 3. Sample dataset depicting ion behaviour versus applied solution strength (MW3-4 (upper till), simulated process-affected water, traditional batch sample,sample #1 in triplicate series).

63A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

64 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

fraction of each exchangeable cation, compared to the overallexchange capacity.

Different approaches exist for implementing ion exchangetheory. By measuring the equilibrated concentrations of majorcations, both in solution andbound to the solid phase (the lattervia desorption experiments), an equilibrium ion exchangecoefficient can be calculated. This coefficient can then beevaluated against established values for similar sediments. Inthe present experiments, this approach was not possible sincebothdesorption andpreviously conducted exchangeable cationexperiments (Sheldrick, 1984; USEPA, 1986) were ammoniumacetate-based, and found to be confounded by carbonate andsulphate mineral dissolution acting concurrently with ex-change reactions. Alternately, ion exchange coefficients (e.g.KNa/Ca) may instead be assigned previously established,empirical values, and used to back-calculate estimates ofbound exchangeable cation activities. Constrained by theseestimated activities, the equilibrated aqueous and solid phasesystemspredictedby simulations can thenbe comparedagainstlaboratory observations. PHREEQC (Parkhurst and Appelo,1999) may be used to facilitate this latter approach.

4.2.1. Geochemical modeling of batch sorption experiments usingPHREEQC

The United States Geological Survey software programPHREEQC (Parkhurst and Appelo, 1999), in conjunction withtheWATEQ4F database (Ball and Nordstrom, 1991), was usedto evaluatewhether or not ion exchange theory could accountfor the observed response of the batch sorption experiments.Several scenarios were modeled, each sharing the same basicstructure. First, a surface of exchange sites was equilibratedwith the adjacent groundwater, to derive the initial, assumedin-situ, composition of the exchange assemblage. Previouslydetermined CEC values (Table 1) were used to quantify thenumber of exchange sites, while groundwater chemistry wasdetermined by in-house analysis of field samples collected atthe same time and physical location as the sediment samplesbeing modeled. This modeled assemblage was then mixedwith 30 g of PA water, mimicking the treatment appliedduring the batch experiments. At this point, the modelsdeviate from one another. One model included only the stepsdescribed above (“PHREEQC no mineral” in graphs). Howev-er, earlier saturation index simulations had found carbonatesolids to be oversaturated. In order to improve the fit of themodeled calcium and magnesium responses to the experi-mental data, mineral phase equilibrationwas therefore addedto different stages of the conceptual experiment. First,mineral precipitation of the most oversaturated component,calcite, was simulated by equilibrating the resultant, mixedsolution, by itself, with calcite (“PHREEQC Calcite at End”).However, this fails to account for oversaturated magnesiumcarbonates and furthermore, empirical evidence suggests thatnaturally occurring minerals exist more often as solidsolutions than as pure phases (Appelo and Postma, 2005).Therefore, another approach instead equilibrated the finalsolution not with the pure phase calcite, but with an idealsolid solution comprised of calcite and magnesite. A potentialconceptual shortcoming of these two models, is that mineralequilibration is applied only to the final solution, once it hasequilibrated with the exchange sites. A more complicatedmodel was therefore run, in which calcite precipitation was

permitted during the mixing reaction of the 30 g PA waterwith the exchange surfaces, thereby impacting both thesolution and exchange site compositions (“PHREEQC Calcitewith Mix”).

Further improvement to the fit of the modeled response,particularly at low PA strengths, required that sulphate saltdissolution, predicted by earlier analysis but hitherto ignored,be included in the model. In an attempt to achieve the sameextent of dissolution as observed in the batch experiments,equilibration with gypsumwas included in the models, at theprecise saturation indices noted in the laboratory solutions.Gypsum equilibration was included either during the mixingreaction of the 30 g PA water with the exchange surfaces(“PHREEQC CalciteGypsumwMix”), again, thereby impactingboth the aqueous and bound chemistries, or alternativelyapplied to the final solution in isolation (“PHREEQC SolidSln-Gypsum at End”).

4.2.2. Simulation resultsIon exchange models including both calcite and gypsum

mineral phase equilibration simulated the observed batchexperiment data very closely for the major cation species(Fig. 4). The sodium response was predicted very well, by allof the models attempted.

For calcium, accounting for calcite precipitation improvedthefit of themodel to the observed data, particularly in termsofshape but often in terms of the magnitude as well (either“Calcite with Mix” or “Calcite at End” vs. “no mineral” model).The refinementof including gypsumequilibration led to furtherimprovement, especially at low PA strengths, so that “Calcite-GypsumwMix” and “SolidSlnGypsumat End”both very closelypredicted the observed behaviour (owing to the solutionchemistry, the resulting solid solution mineral phase ispredominantly calcite-based. Therefore, it is not surprisingthat these two models' results appear similar, since essentiallyboth equilibrated with calcite and gypsum). The model bestfitting the observed data was the “SolidSlnGypsum at End.”

For magnesium, concentrations were often under-pre-dicted, regardless of the model. Once again, the combinationof shape and magnitude of response were typically bestrepresented by the “CalciteGypsum with Mix” model, andnext by the “Calcite with Mix” or “SolidSlnGypsum at End”models.

The strong similarity between the best models' output andthe observed experimental data suggests that particularly fordivalent species, ion exchange theory, supplemented by carbon-ate mineral precipitation and sulphate mineral dissolution offersa suitable explanation of the observed calcium and magnesiumsystem behaviour.

4.2.3. Further validating the PHREEQC modelsBoth hand calculations (not shown — Appelo and Postma,

2005 p256) and experimental data were used to validate theoutput from PHREEQC (Parkhurst and Appelo, 1999) modelsduring their development.

For calcium and magnesium, the ammonium acetate-based batch desorption data were not of use, as theyinfluenced by the same carbonate and sulphate salt dissolu-tion confounds as observedduring recent exchangeable cationexperiments on these same sediments. However, therewas nosuch evidence to discredit the sodium desorption data.

Fig. 4. Comparison of observed concentrations (average of triplicates) with results predicted by PHREEQC [12]models (for traditional batch experiments, sedimentMW3-9 (lower till), using real process-affected water; sodium — top, calcium — middle, and magnesium — bottom).

65A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

Experimentally-derived exchangeable sodium fractions werecalculated as the milliequivalent-based fraction of boundsodium relative to the cation exchange capacity, using

traditional samples from region MW3-4. These observedexchangeable sodium fractions were then evaluated againstthe corresponding values calculated by the two best PHREEQC

66 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

simulations (Parkhurst and Appelo, 1999): one with thecalcite, magnesite solid solution and gypsum equilibrationafter mixing PA water with the exchange assemblage, theotherwith calcite and gypsumequilibrationduring themixingof PA water with exchange sites. The results show excellentsimilarity between the modeled and experimental-derivedsodium fractions, supporting the assertion that themodels areaccurately representing system behaviour (Table 4).

4.3. Integration of results

Laboratory results and corroborative geochemical simula-tions establish a unified conceptual model elucidating thecomplex geochemical interactions expected to take placewhen migrating PA water encounters surficial sediments. Inlight of observed equilibrium ion trends, exchangeable cationand saturation index data, and PHREEQC ion exchange model-ing, it is expected that ingressing PA water will cause thedissolution of pre-existing calcium and magnesium sulphatesalts. Therefore high concentrations of sulphate, in addition toconservative chloride, are expected to persist. Furthermore,evidence suggests that sodium present in the PA water, bycompetitive exchange, releases pre-bound calcium and magne-sium into solution. However, aqueous concentrations of thesedivalent cations are subsequently mitigated by limited precip-itation as a carbonate mineral phase. These findings aresupported by observations made during recent field studies. Ata road salt storage facility in Massachusetts, Ostendorf et al.(2009) noted that an advancing de-icing agent plume withdissolved sodium fraction of 0.975, desorbed calcium andmagnesium that were initially bound to glacial till drumlinsediments. Oiffer et al. (2009) investigated the evolution of a PAseepage plume in a shallow sand aquifer adjacent to a FortMcMurray oil sands tailings pond. In that study, researchersproposed that observed retardation of bicarbonate migration,relative to conservative chloride,was due to calcite precipitationin the calcite-oversaturated plume. In further affirmation of thepresent findings, ingressing ammonium and sodium wereretarded — believed to be attributable to cation exchangeprocesses.

Meaningful predictions regarding the potential ecologicalconsequences of PA seepage reaction products, such as sodium,chloride or sulphate release cannot be made at this time. Thepresent experiments offer qualitative understanding and notquantitative estimates of concentrations of ions that couldultimately appear in downstream waterways. Furthermore, ingeneral, the ecological effects of moderately saline oil sands

Table 4Batch desorption experiment validation of PHREEQC (Parkhurst and Appelo1999) models (MW3-4).

Case βNa, desorb,experimental

βNa, Phreeqc, SolidSlnGypsumat End

βNa, Phreeqc, CalciteGypsumw Mix

Deionized 0.001 0.013 0.00520 0.032 0.051 0.04840 0.061 0.083 0.08060 0.092 0.110 0.11280 0.121 0.133 0.138100 0.157 0.158 0.172

Values in each column were averaged across real PA and simulated PAsamples.

,

process-affected water are not well understood. However,research to date suggests that such impacts could includealtering the composition of phytoplankton communities(Leung et al, 2001, 2003), mortality and reduced numbers ofoffspring of daphnid species (Harmon et al., 2003), and osmoticstress (Crowe et al., 2001) and inhibited germination char-acteristics of plant species (Crowe et al., 2002).

4.4. Traditional batch method versus cube diffusion batchmethod response

Overall, the results of the cube batch sorption experimentswere very similar to those of the traditional set-up, featuringhighly disturbed slurried sediment samples, affirming the utilityof this newer approach. The magnitude of ion concentrations inthe equilibrium solutions was noticeably smaller for cubesamples; however, this did not impact the analysis in anyapparent manner. For example, the calculated sodium linearsorption partitioning coefficients, the shape and magnitude ofammonium, sodium and lithium linear isotherms, and the shapeand magnitude of batch sorption ion trends versus PA strengthwere all essentially the same across both methods. Similarly,Zhang et al. (1998) reported that the adsorption coefficients forbenzene using Regina clay were quite similar between thetraditional and diffusion sorption methods, at 5.1±1.6 mg L−1

and 4.2±0.9 mg L−1 respectively. Thus the cube diffusionmethod appears to offer a viable alternative to the standardbatch method and a conceptual improvement in simulatingsorption behaviour within unperturbed geologic settings.

4.5. Influence of organics: comparing the simulated PA to realPA response

The results from the batch sorption experiments usingsimulated PA water were comparable to those using (real) PAwater collected on site, with observed differences more likelyreflecting parent solution concentration differences than theimpact of organics. Sodium linear sorption partitioning coeffi-cients were not statistically different between simulated andreal PA samples (paired sample t-test, alpha=0.05). However,equilibrium concentrations tended to be larger within simu-lated PA systems, particularly for sodium, magnesium, chlorideand sulphate. Plots of ion trends versus PA strength weresimilar in shape and magnitude across both cases, althoughsediments mixed with simulated PA water tended to showgreater uptake of sodium and alkalinity and a marginallysmaller release of calcium.

These effects are believed to reflect concentrationdifferencesbetween simulated and real PA solutions. Simulated stocksolutions were created to a maximum of 1.2 times the strengthof real PA water in order to conservatively investigate systemresponse at a greater PA concentration than observed to date.Additionally, the simulated solution attempted to broadlyrecreate real PA water using only major ions; therefore, certainspecies' concentrations, such as calcium, were significantlygreater in the simulated stock solutions (Table 2). The greatersodium concentration in simulated PA mixing solutions notsurprisingly resulted in greater sodium uptake. Similarly,simulated solutions, with much higher calcium and alkalinityconcentrations, were inherently more saturated with respect tocalcite (confirmed by saturation index analysis), resulting in

67A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

greater precipitation of that mineral phase, and hence alkalinityuptake and reduced aqueous calcium concentrations. Togetherthese results suggest a minimal impact of organics as complex-ing agents upon system behaviour.

4.6. Next steps in the research

Batch sorption experiments allow for easy, rapid determi-nation of ion exchange tendencies and sorption capacities ofsediments. However, despite its widespread use, the conven-tional procedure suffers from several limitations including thatsediments are disturbed greatly, reaction times are very shortand redox state is not controlled for — all of which limit therepresentativeness to on-site conditions. These limitations willbe minimized by the next stage in this research project, radialdiffusion cell experiments (van der Kamp et al., 1996), whichwill build upon the present findings, evaluating the mitigationormobilization ofmajor ionsand tracemetals in the subsurface,in response to the diffusion-driven ingress of PA water. Radialdiffusion cells permit geochemical investigations over longerdurations, featuring minimally disturbed sediments and acontrolled redox environment. These experiments will there-fore help to address questions such as, in the long run, willadsorption of organics coat binding sites or mineral phasesinhibiting ion exchange or mineral dissolution processes? Theresults from the radial diffusion cells will be additionallyinterpreted in light of collaborative anaerobic microcosmexperiments featuring these same sediments. This will enablebetter understanding of the biogeochemical processes and endproducts expected on site — for example, identifying whetheror not released sulphate will remain in its present form orinstead perhaps immobilized as a result of microbially-mediated reduction reactions.

5. Conclusions

This study presents the first detailed laboratory charac-terization of the adsorption and ion exchange processes to beexpected when ingressing inorganic species present withinoil sands' PA water interact with surficial sediments inNorthern Alberta. The batch sorption experiments revealthat where amply present, clay till sediments have thecapacity to significantly mitigate the high concentrations ofingressing sodium (600 mg L−1), with linear sorption parti-tioning coefficients of about 0.45 L kg−1. However, it shouldbe noted that in localized areas on site, the glacial till coverageis thin (b5 m) or altogether absent. The results also suggestthat chloride will behave conservatively, with high concen-trations remaining in solution (375 mg L−1). In areas oflimited glacial till coverage, high concentrations of ingressingsodium and chloride ions are expected to remain in solution.Unchecked these species are a potential risk to neighbouringaquatic environments.

In addition, this study offers new and important insightinto the complex nature of geochemical reactions in clay tilland sand channel sediments in this setting. The analysisshows that the prevalent cations' behaviour is not suitablyrepresented by the simple linear sorption isotherm model.Instead, system behaviour is shown to be controlled througha combination of competitive ion exchange, dissolution andprecipitation reactions. Specifically, the observed ion trends

in equilibrated solutions, in conjunction with geochemicalsimulations, suggest that the influx of process-affected waterwill induce the dissolution of pre-existing sulphate salts.Furthermore, sodium present in the process-affected waterwill exchangewith sediment-bound calcium andmagnesium,increasing the divalent ions' pore fluid concentrations, andleading to the precipitation of a calcium-magnesium carbon-ate mineral phase.

By addressing how reactions in the clay till affect theultimate composition of seepage prior to interaction withadjacent water resources, the present research has importantimplications upon future environmental impact assessments,seepage mitigation efforts and remediation strategies. Be-cause buried sand channels will be encountered at a numberof future oil sands tailings facilities, this research is expectedto have industry-wide benefit.

Acknowledgements

The authors wish to thank Suncor Energy Inc. for thepermission to publish this work. The authors also kindlythank Suncor Energy Inc. and the Natural Sciences andEngineering Research Council of Canada for providingfinancial support for this project. The authors would like toacknowledge Shirley Chatten of the University of Waterloofor conducting the fraction of organic carbon analysis of theclay till samples.

References

Alberta Chamber of Resources, 2004. Oil Sands Technology Roadmap:Unlocking the Potential [online]. Available from http://www.acr-alberta.com/Portals/0/projects/OSTR_report.pdf [verified 14 January2010].

Alberta Sustainable Resource Development (a ministry of the Government ofAlberta), 2009. Oil Sands [online]. Available from http://srd.alberta.ca/ManagingPrograms/OilSands/Default.aspx [verified 14 January 2010].

Allen, E.W., 2008. Process water treatment in Canada's oil sands industry: I.Target pollutants and treatment objectives. J. Environ. Eng. Sci. 7,123–138.

Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution,Second Ed. A.A. Balkema Publishers, Leiden.

ASTM Standard D4646, 2003. Standard Test Method for 24-h Batch-TypeMeasurement of Contaminant Sorption by Soils and Sediments. ASTMInternational, West Conshohocken, PA.

Ball, J.W., Nordstrom, D.K., 1991. User's Manual for WATEQ4F, with RevisedThermodynamic Database and Test Cases for Calculating Speciation ofMajor, Trace and Redox Elements in Natural Water: U.S. GeologicalSurvey, Open-file Report, pp. 91–183.

Burgisser, C.S., Cernik, M., Borkovec, M., Sticher, H., 1993. Determination ofnonlinear adsorption isotherms from column experiments: an alterna-tive to batch studies. Environ. Sci. Technol. 27, 943–948.

Chalaturnyk, R.J., Scott, J.D., Ozum, B., 2002. Management of oil sands tailings.Petrol. Sci. Technol. 20, 1025–1046.

Crowe, A.U., Han, B., Kermode, A.R., Bendell-Young, L.I., Plant, A.L., 2001.Effects of oil sands effluent on cattail and clover: photosynthesis and thelevel of stress proteins. Environ. Pollut. 113, 311–322.

Crowe, A.U., Plant, A.L., Kermode, A.R., 2002. Effects of an industrial effluenton plant colonization and on the germination and post-germinativegrowth of seeds of terrestrial and aquatic plant species. Environ. Pollut.117, 179–189.

Energy Resources Conservation Board (ERCB), 2009. Alberta's Oil Sands:Strengthening Regulatory Requirements. 2008 Year in Review 2008[online]. http://yearinreview.ercb.ca/pdfs/Albertas_Oil_Sands.pdf [verified14 January 2010].

Government of Alberta, 2006. Investing in Our Future: Responding to theRapid Growth of Oil Sands Development — Final Report [online].Available from http://alberta.ca/home/395.cfm [verified 14 January2010].

68 A.A. Holden et al. / Journal of Contaminant Hydrology 119 (2011) 55–68

Hall, T., Mugo, R., Digel, M., Gibson, C., Gulley, J.R., 2004. Water QualityModelling Report for the Suncor South Tailings Pond Project. GolderAssociates Ltd, Calgary, pp. 23–24. Contact: 102, 2535 — 3rd Avenue S.E.Calgary, AB, Canada T2A 7 W5.

Harmon, S.M., Specht, W.L., Chandler, G.T., 2003. A comparison of thedaphnids ceriodaphnia dubia and daphnia ambigua for their utilizationin routine toxicity testing in the southeastern united states. Arch.Environ. Contam. Toxicol. 45, 79–85.

Holden, A.A., Perez, L., Martin, J., Mendoza, C., Sego, D., Ulrich, A., Tompkins,T., Barker, J.F., Haque, S., Mayer, K., Sutherland, H., Bowron, M., Biggar, K.,Donahue, R., 2009. Fate and Transport of Process-Affected Water in Out-of-Pit Tailings Ponds in the Oil Sands Industry in Canada. In: Sego, D.,Alostaz, M., Beier, N. (Eds.), Proceeding of the Thirteenth InternationalConference on Tailings and Mine Waste, Banff, Canada. 1–4 November2009, pp. 527–539.

Husz, G., 2001. Lithium chloride solution as an extraction agent for soils. J.Plant Nutr. Soil Sci. 164, 71–75.

Janfada, A., Headley, J.V., Peru, K.M., Barbour, S.L., 2006. A laboratoryevaluation of the sorption of oil sands naphthenic acids on organic richsoils. J. Environ. Sci. Health. A 41, 985–997.

Janzen, H.H., 1993. Soluble salts. In: Carter, M.R. (Ed.), For Canadian Society ofSoil Science, Soil Sampling and Methods of Analysis. Lewis Publishers,Boca Raton, pp. 161–165.

Kasperski, K.L., 1992. A review of properties and treatment of oil sandstailings. AOSTRA J. Res. 8, 11–53.

Klohn Crippen Consultants Ltd., 2004. Millennium mine, design of the southtailings pond, prepared for Suncor Energy Inc. Contact: 500 — 2618Hopewell Place N.E., Calgary, AB, Canada, T1Y 7 J7.

Langmuir, D., 1997. Aqueous Environmental Geochemistry. Prentice-HallInc., Upper Saddle River.

Leung, S.S., MacKinnon, M.D., Smith, R.E.H., 2001. Aquatic reclamation in theAthabasca, Canada, oil sands: naphthenate and salt effects on phytoplanktoncommunities. Environ. Toxicol. Chem. 20, 1532–1543.

Leung, S.S., MacKinnon, M.D., Smith, R.E.H., 2003. The ecological effects ofnaphthenic acids and salts on phytoplankton from the athabasca oilsands region. Aquat. Toxicol. 62, 11–26.

Marsh, W.P., 2006. Sorption of naphthenic acids to soil minerals. Master ofScience Thesis, University of Alberta, Canada.

Oiffer, A.A.L., Barker, J.F., Gervais, F.M., Mayer, K.U., Ptacek, C.J., Rudolph, D.L.,2009. A detailed field-based evaluation of naphthenic acid mobility ingroundwater. J. Contam. Hydrol. 108, 89–106.

Ostendorf, D.W., Xing, B., Kallergis, N., 2009. Cation exchange in a glacial tilldrumlin at a road salt storage facility. J. Contam. Hydrol. 106, 118–130.

Parkhurst, D.L., Appelo, C.A.J., 1999. User's Guide to PHREEQC (version 2)—AComputer Program For Speciation, Batch-reaction, One-dimensionalTransport, and Inverse Geochemical Calculations: U.S. Geological SurveyWater-resources Investigations Report 99-4259, p. 312.

Peng, J., Headley, J.V., Barbour, S.L., 2002. Adsorption of single-ring modelnaphthenic acids on soils. Can. Geotech. J. 39, 1419–1426.

Price, A.C.R., 2005. Evaluation of groundwater flow and salt transport withinan undrained tailings sand dam. Master of Science Thesis, University ofAlberta, Canada.

Method 84-005 — 1 N Ammonium Acetate Extractable Ca, Mg and K. In:Sheldrick, B.H. (Ed.), Analytical Methods Manual 1984 [online]. Availablefrom http://sis2.agr.gc.ca/cansis/publications/manuals/analytical.html[verified 14 January 2010].

Stephens, B., Langton, C., Bowron, M., 2006. Design of Tailings Dams on LargePleistocene Channel Deposits a Case Study — Suncor's South TailingsPond [online]. Available from http://www.klohn.com/news/technical-papers%5CCDA_2006-Paper_No.074.pdf [verified 14 January 2010].

Tompkins, T.G., 2009. Natural gradient tracer tests to investigate the fate andmigration of oil sands process-affected water in the wood creek sandchannel. Master of Science Thesis, University of Waterloo, Canada.

United States Environmental Protection Agency (USEPA), 1986. Method9081 Cation-Exchange Capacity of Soils (Sodium Acetate) [online].Available from http://www.epa.gov/solidwaste/hazard/testmethods/sw846/pdfs/9081.pdf [verified 14 January 2010].

van der Kamp, G., Van Stempvoort, D.R., Wassenaar, L.I., 1996. The radialdiffusion method 1. Using intact cores to determine isotopic composition,chemistry, and effective porosities for groundwater in Aquitards. WaterResour. Res. 32 (6), 1815–1822.

Zhang, X., Barbour, S.L., Headley, J.V., 1998. A diffusion batch method fordetermination of the adsorption coefficient of benzene on clay soils. Can.Geotech. J. 35, 622–629.