Vrije Universiteit Brussel

Uranium aqueous speciation in the vicinity of the former uranium mining sites usingthe Diffusive Gradients in Thin Films and Ultrafiltration techniquesDrozdzak, Jagoda; Leermakers, Martine; Gao, Yue; Elskens, Marc; Descostes, Michael ;Phrommovanh, VannaphaPublished in:Analytica Chimica Acta

DOI:10.1016/j.aca.2016.01.052

Publication date:2016

Link to publication

Citation for published version (APA):Drozdzak, J., Leermakers, M., Gao, Y., Elskens, M., Descostes, M., & Phrommovanh, V. (2016). Uraniumaqueous speciation in the vicinity of the former uranium mining sites using the Diffusive Gradients in Thin Filmsand Ultrafiltration techniques. Analytica Chimica Acta, 913, 94-103. https://doi.org/10.1016/j.aca.2016.01.052

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 25. Mar. 2020

lable at ScienceDirect

Analytica Chimica Acta 913 (2016) 94e103

Contents lists avai

Analytica Chimica Acta

journal homepage: www.elsevier .com/locate/aca

Uranium aqueous speciation in the vicinity of the former uraniummining sites using the diffusive gradients in thin films andultrafiltration techniques

Jagoda Drozdzak a, *, Martine Leermakers a, Yue Gao a, Marc Elskens a,Vannapha Phrommavanh b, Michael Descostes b

a Analytical, Environmental and Geochemistry (AMGC), Vrije Universiteit Brussels (VUB), Pleinlaan 2, 1050 Brussels, Belgiumb AREVA Mines e R&D Dpt., Tour AREVA, 1 Place Jean Millier, 92084 Paris La D�efense, France

h i g h l i g h t s

* Corresponding author.E-mail address: jdrozdza@vub.ac.be (J. Drozdzak).

http://dx.doi.org/10.1016/j.aca.2016.01.0520003-2670/© 2016 Elsevier B.V. All rights reserved.

g r a p h i c a l a b s t r a c t

� The applicability of the DGT tech-nique in the vicinity of former ura-nium mining sites was evaluated.

� The binding selectivity order of thebinding phase is the imperative fac-tor influencing the performance ofthe DGT method.

� There is a good agreement between Uconcentration measured by the DGTtechnique and 10 kDa ultrafiltrate.

� DGT technique can be used as analternative to ultrafiltration methodto determine a potentially Ubioavailable pool.

a r t i c l e i n f o

Article history:Received 29 October 2015Received in revised form26 January 2016Accepted 31 January 2016Available online 3 February 2016

Keywords:UraniumDiffusive gradients in thin filmsUltrafiltrationDiphonix®

Former mining sitesBioavailable fraction

a b s t r a c t

The performance of the Diffusive Gradients in Thin films (DGT) technique with Chelex®-100, Metsorb™and Diphonix® as binding phases was evaluated in the vicinity of the former uranium mining sites ofChardon and L'Ecarpi�ere (Loire-Atlantique department in western France). This is the first time that theDGT technique with three different binding agents was employed for the aqueous U determination in thecontext of uranium mining environments. The fractionation and speciation of uraniumwere investigatedusing a multi-methodological approach using filtration (0.45 mm, 0.2 mm), ultrafiltration (500 kDa,100 kDa and 10 kDa) coupled to geochemical speciation modelling (PhreeQC) and the DGT technique. Theultrafiltration data showed that at each sampling point uranium was present mostly in the 10 kDa trulydissolved fraction and the geochemical modelling speciation calculations indicated that U speciation wasmarkedly predominated by CaUO2(CO3)32�. In natural waters, no significant difference was observed interms of U uptake between Chelex®-100 and Metsorb™, while similar or inferior U uptake was observedon Diphonix® resin. In turn, at mining influenced sampling spots, the U accumulation on DGT-Diphonix®

was higher than on DGT-Chelex®-100 and DGT-Metsorb™, probably because their performance wasdisturbed by the extreme composition of the mining waters. The use of Diphonix® resin leads to asignificant advance in the application and development of the DGT technique for determination of U inmining influenced environments. This investigation demonstrated that such multi-technique approach

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103 95

provides a better picture of U speciation and enables to assess more accurately the potentiallybioavailable U pool.

© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Uranium (U) is a primordial, naturally-occurring radioactiveelement that is present in aquatic environments at trace levels. Theaverage background concentration of uranium in surface water inEurope is around 0.5 mg L�1, while variations can extend over fourorders of magnitude (0.002e20 mg U L�1) [1,2]. Nevertheless,several anthropogenic activities such as U exploration, mining andmilling or industrial production of fertilizers increase uraniumconcentration in the environment. Uranium is known for its dualmode toxicity - radiological and chemical one, with the latter beingof a particular concern. Although the radiological impact of ura-nium is determined by its total concentration and the isotopecomposition, the U nature, its bioavailability and toxicity are asso-ciated with a chemical speciation [1,3]. The bioavailable fraction ofmetal refers to the metal species, which are immediately availablefor absorption by a (micro)organism [3,4]. The bioavailable Ufraction has not yet been conclusively determined, due to severalfactors such as inconsistencies in the operational definition of thebioavailable fraction, different methodologies employed as well asdifferences in the interpretation of the results.

Ultrafiltration (UF) technique is often employed for the esti-mation of the bioavailable metal fraction, as this method allowsprecise determination of particulate, colloidal and truly dissolvedfractions, using membranes with different cut-offs (e.g. 100 kDa,10 kDa). This information is vital as metal bioavailability isconsidered be linked to the truly dissolved fraction, which isoperationally defined as the phase with molecules smaller than10 kDa (~5 nm) [5]. The ultrafiltration technique is neither limitedby ionic strength nor by complexation equilibria and it has beenused for the uranium speciation measurements in freshwater [6],estuaries [7] and in coastal seawaters [8]. Nevertheless, a vigorousand intricate handling protocol needs to be followed and possiblecontamination during transport and storage cannot be ruled out.Moreover, the ultrafiltration technique is an expensive, time-consuming and not user-friendly procedure, which might notrepresent temporal variations in metal concentration, because it isbased on a grab-sampling. The alternative is to determine the dis-tribution of metal species present in an aqueous solution by thegeochemical speciation calculation codes (i.e. PhreeQC, MinteQ,JChess). Afterwards, the bioavailable metal fraction is deduced bytaking into account the toxicity of individual metal species.Generally, the uranyl ion and its complexes with hydroxides andcarbonates (e.g. UO2

2þ, UO2OHþ, UO2(CO3)20(aq)) are consideredbioavailable and can be interpreted as the potentially toxic U spe-cies [9e12]. The toxicity of U complexes with phosphates andnatural organic matter (humic and fulvic acids) is unclear, since theexisting scientific evidence both support [13,14] and reject [15,16]the toxic effects of uranyl phosphates and U-organic complexes.However, this computational approach needs to be performed inconjunction with filtration or ultrafiltration techniques and it isonly applicable in the environments close to pseudo-thermodynamic equilibrium.

An emerging tool for the estimation of the bioavailable metalfraction is the Diffusive Gradients in Thin Films (DGT) technique[17]. DGT is an in situ method that provides information about thetime-averaged concentration of labile metal species in a solution.

The DGT technique is based on a simple device that accumulatessolutes on a binding agent (i.e a resin/adsorbent immobilized in athin layer of a hydrogel) after passage through a hydrogel. The keyrole of the hydrogel is discrimination of metal species based ontheir size, lability and mobility (the so-called the DGT-labile metalspecies). Afterwards, metal species are effectively immobilized andpre-concentrated on the binding phase gel, what in consequencefacilitates the detection of very low concentrations of metalsencountered under field conditions. The DGT technique has beenemployed for the purpose of the assessment of the metalbioavailability, because DGT imitates the diffusion limiting uptakeconditions that are characteristic of a metal biouptake [18,19].However, this is only applicable under some conditions, thereforethe relationship between the metal uptake by plant/biota and onthe DGT device should be evaluated taking into account the plant/biota species, the element of interest and its chemical form.

The most adequate methodology to estimate bioavailable metalfraction in the aqueous environment needs to involve not onlycomputational and in-situ experimental approaches, but shouldalso incorporate the results from naturalistic and ecological ap-proaches, such as the monitoring research on the bio-indicatorspecies. A considerable caution and profound knowledge of thelimitations of analytical speciation methods is required wheninterpreting the bioavailable metal fraction data based only onchemical speciation, therefore the data obtained in this study willbe operationally defined as a “potentially bioavailable pool”.

The primary aim of this studywas to evaluate the applicability ofDGT technique with Chelex®-100, Metsorb™ and Diphonix®

binding phases in both, natural and mining waters. The fraction-ation and speciation of U were carried out by filtration (0.45 mm,0.2 mm), ultrafiltration (500 kDa, 100 kDa and 10 kDa) assisted withgeochemical speciation modelling (PhreeQC) and the DGT tech-nique in the vicinity of the former uranium mining sites. Moreover,the feasibility of the DGT technique for the estimation of a poten-tially bioavailable U pool was investigated based on the comparisonof the U-DGT-labile and the 10 kDa truly dissolved U fractions.

2. Materials and methods

2.1. Field sites

The studied former uranium mines, Chardon and L'Ecarpi�ere,are located in Pays de la Loire (Loire-Atlantique department) inwestern France. The field sampling was performed at high flowregime of water systems in January 2015. The mining site of Char-don consists of a former open pit that has been filled up withdrainage waters from the underground mine and it is characterizedby high salinity (i.e. 1.8 � 10�2 M NaCl). In order to prevent the pitfrom overflowing, during the winter months a part of the water ispumped into the stream Margerie, which is the tributary of theriver Sevre Nantaise. Six sampling stations were located on 4 km ofthe Sevre Nantaise river's course, in order to investigate thedownstream changes in a river after the mining waters have beendischarged from the former open pit (Fig. 1). The average flow rateof the river Sevre Nantaise and the stream Margerie during thesampling period was 18.8 m3 s�1 and 10.1 m3 s�1, respectively (dataprovided by Department of Maine-et-Loire ANJOU).

Fig. 1. Sampling stations at former mining site of Chardon (www.maps.google.com).

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e10396

The site of L'Ecarpi�ere has a water treatment plant, which tar-gets particularly at the removal of radium and uranium by theprecipitation. This treatment involves the addition of BaCl2 toprecipitate radium by co-precipitation of Ba(Ra)SO4 and the addi-tion of lime to increase pH and to assist the precipitation of Fe(III)oxy-hydroxides, with co-precipitation of uranium and other min-ing related elements. The treated waters pass several precipitationbasins and afterwards they are discharged into the river Moine. Thequality of Moine river water was examined over 0.3 km course atthree sampling stations-the upstream of the Moine river, themining dischargewater and at the confluence zone after theminingdischarge (Fig. 2). The average flow rate of the river Moine duringthe sampling period was 7.1 m3 s�1 (data provided by Departmentof Maine-et-Loire ANJOU). The discharge flow rate at the samplingspot (7) was approximately 60 m3 h�1.

2.2. Filtration

The water samples were filtered in the field through a 0.45 mmand 0.22 mm pore-size membrane syringe filters (Chromafil). Thosesamples for trace metals analysis were acidified on-site with ul-trapure HNO3 and those for anions and major cations analysis werestored without treatment. Dissolved Organic Carbon (DOC) watersamples were acidified with ultrapure H3PO4, Dissolved InorganicCarbon (DIC) water samples were preserved using HgCl2 and no air

space was left in the bottles. The filtration field blanks werecollected at the start and at the end of the sampling mission bypassing MilliQ water through the membrane syringe filters.

The temperature, pH, Eh, dissolved O2 and conductivity weremeasured in situ with VWR MU 6100H and WTW 3430 portablemulti-parameter instruments and associated probes. Total alka-linity was measured titrimetrically using the Hach Digital Titratorand the measurements were done immediately on-site to preventloss or gain of carbon dioxide or other gases when exposed to theatmosphere or to turbulence during the transport.

2.3. Ultrafiltration

The ultrafiltration used in this study was a MilliPore Pellicon®2cassette filtering system with a Biomax MilliPore polyethersulfonemembranes cut-off of 500 kDa, 100 kDa and 10 kDa (0.5 m2

filterarea). The water samples were collected in 5 L acid-clean plasticbottles and then filtered immediately through a 0.45 mm celluloseacetate filter using Eijkelkamp peristaltic pump. Afterwards, thesamples were filtered first using a 500 kDa filter, and the collectedpermeate (<500 kDa) was ultrafiltered again, using a 100 kDa and10 kDa filters, respectively. In order to achieve a good recovery, ahigh cross-flow ratio was maintained at each step. Approximately0.5 L of water was passed through the filter and discarded beforethe filtrate was collected on-line in a 50 mL glass tubes. The

Fig. 2. Sampling stations at former mining site of L'Ecarpi�ere (www.maps.google.com).

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103 97

permeate samples for trace metals and DOC analysis were acidifiedwith ultrapure HNO3 and H3PO4, respectively. The system wassequentially rinsed with MilliQ water, 0.01 M HNO3 and 0.01 MNaOH solutions between each ultrafiltration cycle according to theprocedure described by Guo et al. (2007). The ultrafiltration fieldblanks were collected at the start, in the middle and at the end ofthe ultrafiltration handling.

2.4. Speciation modelling

The thermodynamic geochemical speciation software PhreeQCwas used to predict the U speciation and to assist with the inter-pretation of the results due to the lack of up to date approachabletechniques which are able to measure individual uranium chemicalspecies in solution [24]. A consistent database was based on theLLNL database, modified to include the set of uranium thermody-namic data selected by NEA (Table S1 in SI) [25,26].

2.5. Diffusive gradients in thin films

Milli-Q (ultra-pure) water (>18.2 MU cm, Millipore, USA) wasused for the preparation of the solutions, gels, cleaning glasswareand containers. All plastic equipment were pre-cleaned in 10% (v/v)HNO3 (pro analysi, Merck, Germany) for at least 24 h and rinsedthoroughly with Milli-Q water. The polyacrylamide (PAM) diffusive

gel (Dg ¼ 0.4; 0.8; 1.2 mm) and Chelex®-100 resin gel(Dg ¼ 0.4 mm; Biorad) were prepared according to the proceduredescribed by Zhang and Davison (1995) [17]. Metsorb™ adsorbentgel (Dg ¼ 0.4 mm; Graver Technologies) was prepared based on theprotocol of Bennett et al. (2010) [20]. Diphonix® resin gel(Dg ¼ 0.8 mm; Eichrom Technologies) was prepared according toDrozdzak et al. (2015) [21]. The preparation of various gels wascarried out under laminar flow hood (class-100) in a clean room.The DGT samplers were supplied by DGT Research and assembledaccording to the protocol from Lancaster (www.dgtresearch.com).Concisely, the binding phase gel disc was placed on the bottom of aDGT device, with the binding phase side faced up. The bindingphase gel was covered by a diffusive gel disc and then a MilliporeDurapore membrane filter (0.45 mm, HVLP) was placed on a top ofthe gels. Assembled DGT samplers were stored at 4 �C in doublezippered plastic bags, which contained several drops of 0.01 MNaNO3.

The DGT device with Chelex®-100, Metsorb™, Diphonix®

binding phase gels were deployed in 4 replicates at each samplingpoint for approximately 24e48 h, with the shorter DGT deploy-ment time at sampling points with elevated U concentration. Toassess the presence of the Diffusive Boundary Layer (DBL) at theformer open pit (3), the DGT devices containing Chelex®-100 resingel with different thicknesses of the diffusive layer (0.055, 0.095and 0.135 cm) were deployed. The DGT devices were affixed to

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e10398

several Perspex plates, which were specifically designed fordeploying up to nine DGT devices. Then the DGT-Perspex units withthe DGT devices facing towards the direction of the stream wereattached to a rope and weighted to the river bed. Special cautionwas taken to prevent the DGT- Perspex units to be settled on theriverbed.

After deployment, the DGT devices were rinsed with MilliQwater and the DGT binding phase gels were peeled off from the DGTdevices. The U accumulated on Chelex®-100, Metsorb™, Diphonix®

resin gels were eluted with 1 M HNO3, 1 M HNO3 and 1 M HEDPA,respectively [17,21,22]. The DGT field blanks (in triplicate) under-went all processes except the DGT deployment.

The DGT concentration of U accumulated on resin gel (CDGT, mgL�1) was calculated using Eq. (1) [17]:

CDGT ¼ ðMðDgþ sÞÞ=ðDtAÞ (1)

where M is accumulated mass of U on the binding phase gel disc(ng), t is the deployment time (s) and A is the exposure area of theDGT sampler (cm2), assumed to be equal to 3.8 cm2, to account forlateral diffusion in the DGT device [23]. Dg is the thickness of thediffusion layer (cm) and s is Diffusive Boundary Layer, that isdetermined experimentally for each set of deployment conditionsor is assumed to be 0.23mm inmoderate towell mixed waters [23].D is the temperature-corrected diffusion coefficient (cm2 s�1) usingthe Stokes-Einstein equation [17]. The uranium diffusion co-efficients determined previously were applied to the DGT calcula-tions in this study [21].

2.6. Analysis

Trace metals analysis was performed using inductively coupledplasma sector field mass spectrometry (ICP-SF-MS, Element II,Thermo Fisher Scientific Bremen GmbH, Germany), equipped withESI fast autosampler. Calibration was performed by external cali-bration, using Indium (2.5 mg L�1 in 2% HNO3) as an internal stan-dard and 2% HNO3 as a carrier solution. Analytical standardsolutions were prepared from a 1000 mg L�1 in 2% HNO3 U stocksolution (Johnson Matthey Materials Technology, UK) and a1000mg L�1 in 2% HNO3multi-element solution (Merck, Germany).Certified river water standard reference materials- 1640a (NationalInstitute of Standards and Technology) and SLRS-5 (NationalResearch Council Canada) were used to validate the precision andaccuracy of the instrumental analysis. Measured concentrations ofall elements in 1640a and SLRS-5 werewithin the range of providedcertified uncertainties. An average instrumental blank of0.6 ng U L�1 was obtained, resulting in a detection limit of1.3 ng U L�1 (determined as the average blank þ 3 times standarddeviation).

The concentration of major cations and anions were measuredby inductively coupled plasma atomic emission spectrometry (ICP-AES, Optek Iris Advantage) and by ion chromatography (DIONEX DX500), respectively. The concentrations of ammonium (NH4

þ), nitriteand nitrate (NO2

� þNO3�), silicate (SiO4

�) and phosphate (PO43�) were

determined by automatic colorimetric methods (QuAAtro, Seal,Analytical). The concentrations of DOC were measured with a totalorganic carbon analyzer (HiPerTOC, Thermo) and DIC measure-ments were carried out using stable isotope ratio mass spectrom-etry (IRMS, Nu Perspective Instrument).

3. Results and discussion

3.1. Filtration and ultrafiltration

River waters were characterized as well oxygenated (dissolved

O2 ~12.3 mg L�1), neutral (pH ~7.5) and oxic (Eh ~330 mV ESH�1)waters with low conductivity (~336 mS cm�1). On the other handthe mining waters at the former open pit of Chardon (3) and at themining discharge point of L'Ecarpi�ere (7) were alkaline (pH ¼ 8e9),characterized by higher major cations concentrations and thus veryhigh conductivity (~2000 mS cm�1 at (7) and ~3600 mS cm�1 at (3)).However, the conductivity and the major ion composition at thedownstream sampling stations of Moine (4) and Sevre Nantaise (5)rivers, were at the same level as at the upstream points, (2) and (1),respectively (Table S3 and S4). This indicates that the waters fromthe former open pit (3) and the mining discharge water (7) did notcontribute appreciably to the chemistry of the receiving rivers.

The U concentrations in 0.45 mm, 0.2 mm, 500 kDa, 100 kDa,10 kDa fractions at each sampling station of Chardon andL'Ecarpi�ere mining sites are displayed in Fig. 3 and Fig. 4. Thedetailed spatial distribution and fractionation of major cations,anions, trace metals and DOC at each sampling point is also pre-sented in Supplementary Information (Table S3 and Table S4). Thetotal dissolved U concentration in control blank samples and at theupstream sampling points was below 1.5 mg L�1, what indicates thatthe upstream sampling stations were not influenced by the miningactivity. It also implies that the geochemical background U con-centration is low and comparable to other natural waters [6,8]. Theinput from the former open pit (3) significantly increased the Uconcentration in the stream water of the Margerie (4), but no sig-nificant difference between the U concentration encountered at thedownstream (5) and the upstream (1) of Sevre Nantaise wasobserved (Fig. 4). Likewise, the same pattern could be noticed at theL'Ecarpi�eremining site, where the U content at the confluence zoneafter the mining discharge (8) was relatively comparable to the oneat the upstream sampling station of Moine river (6) (Fig. 3). Giventhe nature of U, this U decrease at two mining sites is due to adilution and/or sorption predominantly with Fe(III)oxy-hydroxidesin a less extent [27,28]. The precipitates containing Fe(III)oxy-hydroxides were visible in the discharge water flume at (7) andon the DGT devices.

The total dissolved Fe concentration did not differ significantlybetween the sampling stations and between each former miningsite, except at the mining discharge point of L'Ecarpi�ere (7). Thelowest total dissolved Fe concentration was encountered in themining waters at (7), what was probably due to the removal of Fe inthe precipitation ponds. Themajority of total dissolved Fe (80e90%)was present in >500 kDa fraction, therefore the presence of large Fecolloids of both origin, organic and inorganic, is evident [29e31].

The distribution of U species among different fractions (0.45 mm,0.2 mm, 500 kDa, 100 kDa, 10 kDa) at different sampling points didnot vary substantially except at the mining influenced samplingstations (Figs. 3 and 4). Precisely, a membrane filtrated (<0.45 mmand <0.2 mm) U concentration at sampling spots directly influencedby the mining activity ((3) and (4) at Chardon and (7) and (8) atL'Ecarpi�ere) markedly exceeded the ultrafiltrate U concentration. Apaired t-test showed no statistically significant difference between500 kDa, 100 kDa and 10 kDa fractions/ultrafiltrate samples interms of U concentration at each sampling point at a specified alphalevel of 0.05. At the upstream sampling point of L'Ecarpi�ere (6),there was no significant difference between the U concentrationsmeasured by 0.45 mm, 0.2 mm, 500 kDa, 100 kDa and 10 kDa frac-tions. In turn, at (7) and (8) approximately 40% of the total dissolvedU was bound to large colloids (i.e. >500 kDa), which were probablya mixture of Fe(III)oxy-hydroxides, and aluminosilicates. Similarly,at all sampling spots of the mining site of Chardon, the majority oftotal dissolved U concentrations (70e90%) was observed in the10 kDa truly dissolved fraction. Only in the alkaline waters of theformer open pit (3), approximately 30% of U was in form of largecolloids (>500 kDa).

Fig. 3. The U concentration in 0.45 mm, 0.2 mm, 500 kDa, 100 kDa, 10 kDa and DGT labile concentration estimated by Chelex®-100, Metsorb™ and Diphonix® binding phasesat mining site of Chardon. Uncertainty bars of DGT concentration represent standard deviation of 4 replicates. Uncertainty bars of 0.45 mm total concentration represent standarddeviation of duplicate measurements at the beginning and at the end of the DGT deployment.

Fig. 4. The U concentration in 0.45 mm, 0.2 mm, 500 kDa, 100 kDa, 10 kDa and DGT labile concentration estimated by Chelex®-100, Metsorb™ and Diphonix® binding phasesat mining site of L'Ecarpi�ere. Uncertainty bars of DGT concentration represent standard deviation of 4 replicates. Uncertainty bars of 0.45 mm total concentration representstandard deviation of duplicate measurements at the beginning and at the end of the DGT deployment.

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103 99

The relative dominance of small organic molecules (<10 kDa) atall sampling points has been verified byDOCmeasurements (Fig. 5).Only at the upstream of Sevre Nantaise river (1), the colloidalorganic carbon comprises a dominant fraction of the DOC pool(62%). It is well known that the U distribution is strongly influencedby the presence of DOC [7,32,33], however no significant correla-tion was found in this study. Along the river Sevre Nantaise,Margerie stream and Moine river, the DOC concentrations wererather moderate and constant, besides very low DOC content at themining discharge point of mining site of L'Ecarpi�ere (7). The DOCconcentration value ranged between 533.8 mM and 868.4 mM (withvery low concentration of 198.4 mM at (7)).

Hence, uranium is dominantly transported in the 10 kDa trulydissolved fraction in the vicinity of the former uraniummining sitesof Chardon and L'Ecarpi�ere. The geochemical modelling speciationindicated that the inorganic U species (i.e. CaUO2(CO3)32�)predominates.

3.2. U- DGT labile speciation

The U concentration measured by DGT-Chelex®-100, DGT-Met-sorb™ and DGT- Diphonix® is presented in Figs. 3 and 4. The Uconcentration estimated by a single DGT measurement and theaverage CDGT at each sampling spot used for the statistical evalua-tion is displayed in Table S2 in the Supplementary Information.

The repeatability of the DGT method was expressed as therelative standard deviation of the CDGT values at each samplingpoint (4 replicates of each binding phase per sampling point,n ¼ 32). The repeatability of DGT-Chelex®-100 was estimated at9.1% (with 3.9% min, and 24.5% max), DGT-Metsorb™ at 7.9% (with0.7% min, and 17.6% max) and DGT-Diphonix® at 8.1% (with 0.5%min, and 16.1% max). These results show a considerable variation inthe repeatability between the binding phases and between thesampling points. Poorer repeatability was observed at the samplingpoints influenced by themining activity, probably due to the spatial

Fig. 5. Distribution of dissolved organic carbon in 0.45 mm, 0.2 mm, 500 kDa, 100 kDa, 10 kDa fractions in the vicinity of the former uraniummining sites of Chardon and L'Ecarpi�ere.(1) the upstream of Sevre Nantaise; (2) the upstream of Margerie; (3) the former open pit water; (4) the downstream of Margerie; (5) the downstream of Sevre Nantaise; (6) theupstream Moine; (7) the mining discharge water; (8) the confluence zone after the mining discharge.

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103100

and temporal heterogeneity of the mining influenced waters. Therepeatability of the U-DGT measurements reported by other au-thors for the comparability is rather scarce, however in general, theprecision of 10e15% is considered satisfactory [34,35].

The analysis of variance using Holm-Sidak method for multiplepairwise comparisons was applied to assess binding phases' per-formance. The Grubbs test was used to detect the presence of theoutliers. At sampling points not influenced by the mining activity,no significant difference was obtained in terms of U uptake onChelex®-100 and Metsorb™, while similar or inferior U uptake onDiphonix® in comparisonwas observed. It was found that there is astatistically significant difference between different DGT devices atthe former open pit (3), where the U accumulation on DGT-Diphonix® was markedly higher than on DGT-Chelex®-100 andDGT-Metsorb™. In turn, at the downstream of the stream Margerie(4) and at the mining discharge water point of L'Ecarpi�ere (7), the Uuptake on Chelex®-100 was significantly higher than on Diphonix®

and Metsorb™, which produced similar results. Surprisingly, Ulabile concentration estimated by Chelex®-100 at (7) of L'Ecarpi�erewas notably greater than total dissolved (0.45 mm) U concentration.

The equivocal performance of DGT-Chelex®-100, DGT-Met-sorb™ and DGT-Diphonix® is strongly interrelated to the complexcomposition of natural waters and to the specific chemistry of themining environments. Several studies correlated the discrepanciesin metal uptake on different DGT devices to the distribution ofmetal species and/or the functional properties of the DGT bindingphases [36e38]. However, the extensive laboratory investigation[21] demonstrated that CaUO2(CO3)32�, which constitutes approxi-mately 100% of all U species at all sampling points, is quantitativelyaccumulated on the DGT devices irrespectively of the functionalproperties of Chelex®-100, Metsorb™, Diphonix®. Other factorssuch as pH, ionic strength or the U capacity of the DGT bindinglayers had probably a minor influence on the observed discrep-ancies in the U uptake on different DGT devices. The pH at allsampling spots was neutral and ionic strength was of 0.003 M,except at the former open pit (3) and at the mining discharge point(7), where surface waters were slightly alkaline with ionic strengthof 0.04 M. Neither effect of pH nor ionic strength of this range, onthe U uptake on DGT-Chelex®-100, DGT-Metsorb™ and DGT-Diphonix® was observed under laboratory conditions [21]. It is also

unlikely that the saturation of either DGT binding layer was ach-ieved and negatively affects the U accumulation on DGT-Chelex®-100, DGT- Metsorb™ and DGT-Diphonix® under field conditions.The maximum U capacities exhibited by Chelex®-100, Metsorb™,Diphonix® binding phase gel discs are 1.05 mmol, 1.05 mmol and10.5 mmol, respectively [21]. The mass of U accumulated on Che-lex®-100, Metsorb™, Diphonix® at the former open pit ([U0.45

mm] ¼ 1.1 mg L�1), were 1.3 � 10�2 mmol U, 1.4 � 10�2 mmol U and1.7 � 10�2 mmol, respectively.

Presumably, the U uptake is hampered by the presence of otherco-adsorbing analytes, which can compete individually and/orcollectively for the binding sites of the adsorbent and in conse-quence rearrange the selectivity order of the DGT binding layertowards uranium [36,39,40]. Furthermore, Wazne et al. [41] andJaffrezic-Renault et al. [42] highlighted the twofold competitioneffect of the aqueous carbonate on the U uptake on the titaniumdioxide based adsorbent (i.e. Metsorb™). Under some conditions,the aqueous carbonate will compete with U species for the bindingsites of the adsorbent, but it can also competewith the binding sitesof the adsorbent to complex uranium.

There are several publications that tackle the issue of theselectivity order of Chelex®-100 and Metsorb™ as adsorbents andas the DGT binding layers. According to our knowledge, there is noinvestigation on the selectivity order of Chelex®-100, Metsorb™and Diphonix® that includes uranium as the analyte of interest. Theextensive laboratory investigations and in situ field studiesdemonstrated the complexity and non-applicability of a selectivityorder of a specific binding phase under every environmental con-dition [43e45]. Commonly, only the effect of a few co-adsorbinganalytes at particular pH on the uptake of the metal of interest onthe DGT binding phase is assessed, consequently those findingsmight not be adequate under environmental conditions, especiallyin mining waters. The actual selectivity order of the DGT bindingphase depends on several parameters and can vary from one sys-tem to another, thus possible competition of the metal of interestwith other co-adsorbing analytes should be tested under conditionsof a particular mining system [46,47].

It is generally accepted, that the selection of the efficient eluentfor specific DGT binding phase should be analyte-oriented. There-fore, the elution procedures applied here for the U extraction might

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103 101

not be applicable for other analytes, but they still might be accu-mulated on Chelex®-100, Metsorb™ and Diphonix®. For instant,Bennett et al. (2010), Panther et al. (2011) and Price et al. (2013)used 1 M NaOH to elute As, Se, V, P from DGT-Metsorb™, while inthis study 1 M HNO3 as an eluent was utilized [20,44,48]. Chiariziaet al. (1994) employed 0.33 M H2O2/1 M NaOH mixture to elute Crfrom Diphonix® resin, while in this study 1 M HEDPA as efficienteluent for DGT- Diphonix® was applied [49]. Consequently, theestimated As, Se, V, Cr and other metals uptake on Chelex®-100,Metsorb™ and Diphonix® might not reflect accurately their accu-mulation on the DGT devices.

In all probability, the reason behind a diminished U uptake onDGT-Diphonix® is a synergistic effect of a near neutral pH andhighly complex composition of aquatic systems. Chiarizia andHorwitz (1994) and Chiarizia et al. (1997) highlighted that at cir-cumneutral pH, the gem-diphosphonate groups of Diphonix® resincan dissociate the uranyl species; hence the neutral U speciesshould be still sorbed, although less efficiently than the positivelycharged complexes [49,50]. It has been also pointed out that theuptake of negatively charged uranyl carbonates species into thecationic exchange resin is likely to be significantly suppressedbecause of the Donnan potential. Drozdzak et al. (2015) demon-strated no effect of U species distribution on U uptake on Diphonix®

over wide pH range of 3e9 under laboratory conditions; howeverthose findings in case of highly complex matrix of natural watersmight not be fully pertained. The performance of Chelex®-100 andMetsorb™ binding phases is clearly affected by the extreme con-ditions encountered in mining influenced environments. Mitigatedor impossibly aggrandized U uptake on Chelex®-100 andMetsorb™is most likely associated with very high concentrations of majorcations, anions and other metals typical for mining environments,such as Zn, Mn, Ba or Sr. A small variance between the replicates ofDiphonix® was observed regardless of the milieu of a samplingpoint, while the repeatability of the Chelex®-100 and Metsorb™was negatively affected at sampling stations influenced by theformer uranium mining activity. Precisely, at least one of the rep-licates of Chelex®-100 and Metsorb™ has been identified as anoutlier and excluded from the DGT data interpretation at mininginfluenced sampling points. Diphonix® resin was designed andused mainly for the preconcentration of actinides and the removalof uranium from the radioactive wastewater [51]. Therefore, it canbe presumed, that given all the above mentioned aspects, theUeDGT labile concentration measured by the DGT-Diphonix®

method provides the most reliable results at sampling pointsdirectly influenced by the mining activity.

3.3. Ultrafiltration and DGT

The comparison of the U concentration in the 10 kDa ultrafil-tration permeates and the DGT-labile U concentration is presentedin Figs. 3 and 4. There is a good agreement in respect of the Uconcentration measured by those two techniques. The observeddiscrepancies are probably due to the differences between theoperationally-defined fraction measured by each method and theartifacts influencing each measurement. Ultrafiltration is alaboratory-based technique that discriminates metal species onlyby their size. The water samples for the analysis are collectedusually by a single grab sampling and the relevant fractionation isperformed immediately to minimize the speciation changes asso-ciatedwith sample collection and storage. The pore size (i.e. cut-off)of the ultrafiltration membranes is specified by the manufacturer;however the actual rejection rate depends not only on a nominalmolecular weight, but also on the structure and the properties ofmetal species. Moreover, the real cut-off of themembrane differs byapproximately 10e20% from the manufacturer-defined one.

Therefore, a certain percentage of metal species with higher mo-lecular weight might breakthrough the ultrafiltration membrane,and in turn, it is also possible that some metal species with lowermolecular weight might be retained by the membrane [5]. Favre-R�eguillon et al. (2005) and Guo et al. (2007) also emphasized a highpossibility of an artifactual rejection or retention of large anionicspecies, such as, UO2(CO3)22� or CaUO2(CO3)32�, due to the negativelycharged ultrafiltration membranes [5,52]. In consequence, thismight lead to a significant underestimation of the 10 kDa trulydissolved U fraction.

The imperative feature of the DGT technique is its ability toprovide an in situ, time-averaged estimation of labile metal con-centration. Metal species are discriminated by their size, lability,mobility. Although, a straightforward comparison of the molecularsizes expressed in a hydrodynamic diameter (i.e. nm) and a mo-lecular weight (i.e. kDa) is rather cumbersome, it can be assumedthat the open pore of a diffusive gel (i.e. 5 nm) used in this study,corresponds to the 10e13 kDa cut-off membrane [5]. Generallyspeaking, only the metal species smaller than the pore size of ahydrogel, labile enough to dissociate over the deployment time andable to bind effectively on the DGT binding phase, will be retainedby the DGT device.

It is also of a paramount importance to emphasize several at-tributes that can influence the precision of the DGT measurement,thus might influence an accurate comparison of DGT and ultrafil-tration. The extensive laboratory investigation on the uncertainty ofthe DGT measurement has been performed by Knutsson et al.(2014) and Kreuzeder et al. (2015) [34,53]. Several parameters suchas elution factor, diffusion layer thickness, DGT blank, analyte signalintensity and so forth, were taken into account. It is noteworthy,that other contributors might also deteriorate the accuracy of theDGT measurement under field conditions. The U concentrationmeasured by the DGT technique is a sum of both, labile inorganicand organic species, however CDGT in this study was calculatedusing mean value of the diffusion coefficient of inorganic U species(Eq. (1)). Drozdzak et al. (2015) have shown no effect of a distri-bution of inorganic U species on the effective U diffusion coeffi-cient. However U-organic bearing complexes that might alsocontribute to the mass of U accumulated on the DGT binding layer,diffuse much slower than the inorganic U species, hence Dinorg willnot correlate well with Dorg. Quantitative discrimination betweeninorganic and organic metal complexes can be achieved by a par-allel deployment of the DGT devices with different pore sizes of thediffusive gel (Zhang and Davison, 2000). Up to date, no investiga-tion on the diffusion coefficient of organic U complexes has beenconducted. Nevertheless, in this study neither tremendous differ-ence between the DGT and the ultrafiltration data nor any trend indiminished U-DGT labile concentration was observed, which in-dicates that the DGT calculation strategy (i.e. using only Dinorg) waslegitimate. Indeed, the geochemical modelling speciation indicatedthat the inorganic U species (i.e. CaUO2(CO3)32�) prevail.

There is a good agreement between the DGT-measured and the10 kDa-ultrafiltrate U concentrations, given all the differences be-tween the fractions and the artifacts and limitations of eachmethod. At all sampling spots in the vicinity of the former miningsites of Chardon, the U concentration estimated by the 10 kDa ul-trafiltrate was higher or equal to the DGT-measured one. Thissuggests that the uranyl species with low molecular weight, whichpredominate at each sampling spot, can pass through the 10 kDamembrane and the open pore of the DGT diffusive layer, but they donot have sufficient dissociation rate to be accumulated on the DGTdevice.

Likewise, similar behaviour was observed in the vicinity of themining site of L'Ecarpi�ere, with the exception at the miningdischarge station (7), where the 10 kDa-ultrafiltrate U

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103102

concentrationwas lower than the DGT labile U concentration. Mostlikely, the observed discrepancy is correlated to the highly complexmatrix of the mining discharge water and temporal variation in theU concentration over the DGT deployment time, neverthelessdiurnal variations in streams with similar pH ranges might bepossible as well [54].

DGT has considerable benefits over ultrafiltration technique as itprovides an information about the dynamic supply of metal speciesfrom the medium [18]. Other features, such as robustness, pre-concentration and in situ abilities are worth-mentioning as well.Furthermore, the application of the DGT technique allows obtainingmore precise estimation of the potentially bioavailable U pool andthe nature of the uranyl complexes in natural and mining influ-enced water systems. This study accentuates a high potential of theDGT technique as an alternative to ultrafiltration method for thepurposes of the determination of the potentially bioavailable metalfraction.

4. Conclusions and perspectives

This is the first time that the feasibility of the DGT techniquewith Chelex®-100, Metsorb™ and Diphonix® as binding agents forthe determination of labile U species was evaluated under fieldconditions. The application of Diphonix® resin leads to a significantadvance in the development of the DGT technique for the mea-surement of U in uranium mining environments. The performanceof DGT-Diphonix® is not disturbed by highly complex matrix withelevated levels of major cations, anions and othermetals such as Zn,Mn or Sr.

This work disclosed that it is a prerequisite to perform anexhaustive laboratory characterization of the DGT binding phaselayer prior to undertaking an in situ fieldwork, what was performedpreviously by Drozdzak et al. (2015). The comprehension of theoptimal working parameters of the DGT technique under specificfield conditions ensures a reliable and precise metal speciationmeasurement. Moreover, such a combination of the comprehensiveDGT laboratory investigation and in situ field studies promote abetter understanding of the applicability of the DGT technique as awater speciation tool in aqueous environments.

It is likely that this research has environmental significance as itdemonstrates the capability of DGT to predict the potentiallybioavailable U fraction. Furthermore, it shows that the multi-technique approach, which integrates ultrafiltration combinedwith the computational modelling speciation and the DGT tech-nique provides a deep insight into the U speciation in mininginfluenced environments and the measurement of the potentiallybioavailable U fraction. Finally, this study contributes to the overallunderstanding of the feasibility of the DGT technique as a moni-toring tool for the environmental research within the scope of theWater Framework Directive.

The future work should assess the applicability of DGT-Dipho-nix® technique for the measurement of other co-adsorbing analytesusually encountered under field conditions. The study on thevariability of the DGT technique with different DGT binding phasesis required as well. Furthermore, more research should be devotedto study the relationship between the DGTmeasured concentrationand the response of the biota in the aquatic environments,including mining influenced waters.

Acknowledgements

We acknowledge AREVA Mines France for funding this project.We thank the Hercules Foundation for financing the ICP-SF-MSinstrument (UABR/11/010). Furthermore the authors thank GraverTechnologies (www.gravertech.com) for providing the Metsorb™

adsorbent.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.aca.2016.01.052.

References

[1] H. Vandenhove, C. Hurtgen, T.E. Payne, Uranium, in: D.A. Atwood (Ed.),Radionuclides Environ, John Wiley & Sons, Ltd, 2010, pp. 261e287.

[2] W. de Vos, T. Tarvainen, Geochemical Atlas of Europe. Part 2-Interpretation ofGeochemical Maps, Additional Tables, Figures, Maps, and Related Publications,Geological Survey of Finland, 2006.

[3] P.M. Chapman, Environmental risks of inorganic metals and metalloids: acontinuing, evolving scientific odyssey, Hum. Ecol. Risk Assess. An Int. J. 14(2008) 5e40.

[4] G.M. Rand, P.G. Wells, L.S. Mc Carthy, Introduction to Aquatic Toxicology.Effects, Environmental Fate and Risk Assesment, Taylor and Francis, London,1995.

[5] L. Guo, P.H. Santschi, Ultrafiltration and its applications to sampling andcharacterisation of aquatic colloids, in: Environ. Colloids Part. Behav. Sep.Characterisation, Wiley, 2006, pp. 159e221, http://dx.doi.org/10.1002/9780470024539.ch4.

[6] L. Guo, K.W. Warnken, P.H. Santschi, Retention behavior of dissolved uraniumduring ultrafiltration: Implications for colloidal U in surface waters, Mar. Chem.107 (2007) 156e166, http://dx.doi.org/10.1016/j.marchem.2007.06.017.

[7] P.S. Andersson, D. Porcelli, O. Gustafsson, J. Ingri, G.J. Wasserburg, Theimportance of colloids for the behavior of uranium isotopes in the low-salinityzone of a stable estuary, Geochim. Cosmochim. Acta 65 (2001) 13e25, http://dx.doi.org/10.1016/S0016-7037(00)00514-7.

[8] R.K. Singhal, S.N. Joshi, A.G. Hegde, Association of uranium with colloidal andsuspended particulate matter in Arabian sea near the west coast of Mahara-shtra (India), J. Radioanal. Nucl. Chem. 261 (2004) 263e267, http://dx.doi.org/10.1023/B: JRNC.0000034857.44889.e2.

[9] S.C. Sheppard, M.I. Sheppard, M.-O. Gallerand, B. Sanipelli, Derivation ofecotoxicity thresholds for uranium, J. Environ. Radioact. 79 (2005) 55e83,http://dx.doi.org/10.1016/j.jenvrad.2004.05.015.

[10] S.V. Bhat, J.S. Melo, B.B. Chaugule, S.F. D'Souza, Biosorption characteristics ofuranium(VI) from aqueous medium onto Catenella repens, a red alga,J. Hazard. Mater 158 (2008) 628e635, http://dx.doi.org/10.1016/j.jhazmat.2008.02.042.

[11] C. Fortin, P.G.C. Campbell, An ion-exchange technique for free-metal ionmeasurements (Cd 2þ Zn 2þ ): applications to complex aqueous media, Int. J.Environ. Anal. Chem. 72 (1998) 173e194, http://dx.doi.org/10.1080/03067319808035889.

[12] C. Fortin, F.H. Denison, J. Garnier-Laplace, Metalephytoplanktoninteractions: modeling the effect of competing ions (Hþ, Ca2þ, and Mg2þ)on uranium uptake, Environ. Toxicol. Chem. 26 (2007) 242, http://dx.doi.org/10.1897/06-298R.1.

[13] J.W. Huang, M.J. Blaylock, Y. Kapulnik, B.D. Ensley, Phytoremediation ofuranium-contaminated soils: role of organic acids in triggering uraniumhyperaccumulation in plants, Environ. Sci. Technol. 32 (1998) 2004e2008,http://dx.doi.org/10.1021/es971027u.

[14] L. Laroche, P. Henner, V. Camilleri, M. Morello, J. Garnier-Laplace, Root uptakeof uranium by a higher plant model (Phaseolus vulgaris) e bioavailabilityfrom soil solution, Radioprotection 40 (2005) S33eS39, http://dx.doi.org/10.1051/radiopro:2005s1-006.

[15] S.D. Ebbs, D.J. Brady, L.V. Kochian, Role of uranium speciation in the uptakeand translocation of uranium by plants, J. Exp. Bot. 49 (1998) 1183e1190,http://dx.doi.org/10.1093/jxb/49.324.1183.

[16] M. a. Trenfield, S. McDonald, K. Kovacs, E.K. Lesher, J.M. Pringle, S.J. Markich, etal., Dissolved organic carbon reduces uranium bioavailability and toxicity. 1.characterization of an aquatic fulvic acid and its complexation with uranium[VI], Environ. Sci. Technol. 45 (2011) 3075e3081, http://dx.doi.org/10.1021/es103330w.

[17] H. Zhang, W. Davison, Performance-Characteristics of diffusion gradients inthin- films for the in-situ measurement of trace-metals in aqueous- solution,Anal. Chem. 67 (1995) 3391e3400.

[18] H. Zhang, W. Davison, Use of diffusive gradients in thin-films for studies ofchemical speciation and bioavailability, Environ. Chem. 12 (2015) 85, http://dx.doi.org/10.1071/EN14105.

[19] F. Degryse, E. Smolders, H. Zhang, W. Davison, Predicting availability ofmineral elements to plants with the DGT technique: a review of experimentaldata and interpretation by modelling, Environ. Chem. 6 (2009) 198e218,http://dx.doi.org/10.1071/EN09010.

[20] W.W. Bennett, P.R. Teasdale, J.G. Panther, D.T. Welsh, D.F. Jolley, New diffusivegradients in a thin film technique for measuring inorganic arsenic and sele-nium(IV) using a titanium dioxide based adsorbent, Anal. Chem. 82 (2010)7401e7407, http://dx.doi.org/10.1021/ac101543p.

[21] J. Drozdzak, M. Leermakers, Y. Gao, V. Phrommavanh, M. Descostes, Evalua-tion and application of diffusive gradients in thin films (DGT) technique using

J. Drozdzak et al. / Analytica Chimica Acta 913 (2016) 94e103 103

Chelex®-100, Metsorb™ and Diphonix® binding phases in uranium miningenvironments, Anal. Chim. Acta 889 (2015) 71e81.

[22] C.M. Hutchins, J.G. Panther, P.R. Teasdale, F. Wang, R.R. Stewart, W.W. Bennett,et al., Evaluation of a titanium dioxide-based DGT technique for measuringinorganic uranium species in fresh and marine waters, Talanta 97 (2012)550e556, http://dx.doi.org/10.1016/j.talanta.2012.05.012.

[23] K.W. Warnken, H. Zhang, W. Davison, Accuracy of the diffusive gradients inthin-films technique: diffusive boundary layer and effective sampling areaconsiderations, Anal. Chem. 78 (2006) 3780e3787, http://dx.doi.org/10.1021/ac060139d.

[24] D. Parkhurst, C. Appelo, User's Guide to (Version 2)- a Computer Program forSpeciation and Inverse Geochemical Calculations, 1999.

[25] I. Grenthe, J. Fuger, R. Konings, R. Lemire, A. Muller, C. Nguyen-Trung, et al.,Chemical Thermodynamics 1: Chemical Thermodynamics of Uranium, Elsev-ier Ltd, Amsterdam, 1992.

[26] R. Guillaumont, T. Fangh€anel, J. Fuger, I. Grenthe, V. Neck, D. Palmer, et al.,Update on the Chemical Thermodynamics of Uranium, Neptunium, Pluto-nium, Americium and Technetium, Paris, 2003.

[27] M.C. Duff, J. Urbanik Coughlin, D.B. Hunter, Uranium co-precipitation withiron oxide minerals, Geochim. Cosmochim. Acta 66 (2002) 3533e3547, http://dx.doi.org/10.1016/S0016-7037(02)00953-5.

[28] L.S. Balistrieri, R.R. Seal, N.M. Piatak, B. Paul, Assessing the concentration,speciation, and toxicity of dissolved metals during mixing of acid-minedrainage and ambient river water downstream of the Elizabeth CopperMine, Vermont, USA, Appl. Geochem. 22 (2007) 930e952, http://dx.doi.org/10.1016/j.apgeochem.2007.02.005.

[29] R. Liu, J.R. Lead, H. Zhang, Combining cross flow ultrafiltration and diffusiongradients in thin-films approaches to determine trace metal speciation infreshwaters, Geochim. Cosmochim. Acta 109 (2013) 14e26, http://dx.doi.org/10.1016/j.gca.2013.01.030.

[30] R. Dahlqvist, K. Andersson, J. Ingri, T. Larsson, B. Stolpe, D. Turner, Temporalvariations of colloidal carrier phases and associated trace elements in a borealriver, Geochim. Cosmochim. Acta 71 (2007) 5339e5354, http://dx.doi.org/10.1016/j.gca.2007.09.016.

[31] J. Ingri, S. Nordling, J. Larsson, J. R€onnegård, N. Nilsson, I. Rodushkin, et al., Sizedistribution of colloidal trace metals and organic carbon during a coastalbloom in the Baltic Sea, Mar. Chem. 91 (2004) 117e130, http://dx.doi.org/10.1016/j.marchem.2004.06.009.

[32] D. Porcelli, P.S. Andersson, G.J. Wasserburg, J. Ingri, M. Baskaran, The impor-tance of colloids and mires for the transport of uranium isotopes through theKalix River watershed and Baltic Sea, Geochim. Cosmochim. Acta 61 (1997)4095e4113, http://dx.doi.org/10.1016/S0016-7037(97)00235-4.

[33] B. Lyv�en, M. Hassell€ov, D.R. Turner, C. Haraldsson, K. Andersson, Competitionbetween iron- and carbon-based colloidal carriers for trace metals in afreshwater assessed using flow field-flow fractionation coupled to ICPMS,Geochim. Cosmochim. Acta 67 (2003) 3791e3802, http://dx.doi.org/10.1016/S0016-7037(03)00087-5.

[34] A. Kreuzeder, J. Santner, H. Zhang, T. Prohaska, W.W. Wenzel, Uncertaintyevaluation of the diffusive gradients in thin films technique, Environ. Sci.Technol. 49 (2015) 1594e1602, http://dx.doi.org/10.1021/es504533e.

[35] R. Buzier, A. Charriau, D. Corona, J.F. Lenain, P. Fondan�eche, E. Joussein, et al.,DGT-labile As, Cd, Cu and Ni monitoring in freshwater: toward a frameworkfor interpretation of in situ deployment, Environ. Pollut. 192 (2014) 52e58,http://dx.doi.org/10.1016/j.envpol.2014.05.017.

[36] W. Li, H. Zhao, P.R. Teasdale, R. John, F. Wang, Metal speciation measurementby diffusive gradients in thin films technique with different binding phases,Anal. Chim. Acta 533 (2005) 193e202, http://dx.doi.org/10.1016/j.aca.2004.11.019.

[37] W. Li, J. Zhao, C. Li, S. Kiser, R.J. Cornett, Speciation measurements of uraniumin alkaline waters using diffusive gradients in thin films technique, Anal.

Chim. Acta 575 (2006) 274e280, http://dx.doi.org/10.1016/j.aca.2006.05.092.[38] J.G. Panther, W.W. Bennett, P.R. Teasdale, D.T. Welsh, H. Zhao, DGT mea-

surement of dissolved aluminum species in waters: comparing chelex-100and titanium dioxide-based adsorbents, Environ. Sci. Technol. 46 (2012)2267e2275, http://dx.doi.org/10.1021/es203674n.

[39] W. Li, C. Li, J. Zhao, R.J. Cornett, Diffusive gradients in thin films technique foruranium measurements in river water, Anal. Chim. Acta 592 (2007) 106e113,http://dx.doi.org/10.1016/j.aca.2007.04.012.

[40] M. Pesavento, R. Biesuz, G. Alberti, M. Sturini, Characterization of the sorptionof uranium(VI) on different complexing resins, Anal. Bioanal. Chem. 376(2003) 1023e1029, http://dx.doi.org/10.1007/s00216-003-1951-9.

[41] M. Wazne, X. Meng, G.P. Korfiatis, C. Christodoulatos, Carbonate effects onhexavalent uranium removal from water by nanocrystalline titanium dioxide,J. Hazard. Mater. 136 (2006) 47e52, http://dx.doi.org/10.1016/j.jhazmat.2005.11.010.

[42] N. Jaffrezic-Renault, H. Poirier-Andrade, D.H. Trang, Models for the adsorptionof uranium on titanium dioxide, J. Chromatogr. A 201 (1980) 187e192, http://dx.doi.org/10.1016/S0021-9673(00)83873-X.

[43] Bio-rad Laboratories, Chelex 100 and Chelex 20 Chelating Ion Exchange Resins- Manual Instruction, 2000.

[44] H.L. Price, P.R. Teasdale, D.F. Jolley, An evaluation of ferrihydrite- and Met-sorb™-DGT techniques for measuring oxyanion species (As, Se, V, P): Effectivecapacity, competition and diffusion coefficients, Anal. Chim. Acta 803 (2013)56e65, http://dx.doi.org/10.1016/j.aca.2013.07.001.

[45] H. Osterlund, S. Chlot, M. Faarinen, A. Widerlund, I. Rodushkin, J. Ingri, et al.,Simultaneous measurements of As, Mo, Sb, V and W using a ferrihydritediffusive gradients in thin films (DGT) device, Anal. Chim. Acta 682 (2010)59e65, http://dx.doi.org/10.1016/j.aca.2010.09.049.

[46] H.M. Conesa, R. Schulin, B. Nowack, Suitability of using diffusive gradients inthin films (DGT) to study metal bioavailability in mine tailings: possibilitiesand constraints, Environ. Sci. Pollut. Res. Int. 17 (2010) 657e664, http://dx.doi.org/10.1007/s11356-009-0254-x.

[47] R.L.F. de Oliveira, J.H. Pedrobom, A.A. Meneg�ario, R.N. Domingos, D.A. Py Jú-nior, C.H. Kiang, Determination of in situ speciation of manganese in treatedacid mine drainage water by using multiple diffusive gradients in thin filmsdevices, Anal. Chim. Acta 799 (2013) 23e28, http://dx.doi.org/10.1016/j.aca.2013.09.022.

[48] J.G. Panther, P.R. Teasdale, W.W. Bennett, D.T. Welsh, H. Zhao, Comparingdissolved reactive phosphorus measured by DGT with ferrihydrite and tita-nium dioxide adsorbents: anionic interferences, adsorbent capacity anddeployment time, Anal. Chim. Acta 698 (2011) 20e26, http://dx.doi.org/10.1016/j.aca.2011.04.049.

[49] R. Chiarizia, E. Horwitz, S.D. Alexandratos, Uptake of metal ions by a newchelating ion-exchange resin. Part 4. Kinetics, Solvent Extr. Ion. Exch. 12(1994) 211e237.

[50] R. Chiarizia, P. Horwitz, S.D. Alexandratos, M. Gula, Diphonix resin: a review ofits properties and applications, Sep. Sci. Technol. 32 (1997) 1e35.

[51] R. Chiarizia, E. Horwitz, Uptake of metal ions by a new chelating ion-exchangeresin. Part 6. Calculations on the effect of complexing anions on actinides,Solvent Extr. Ion. Exch. 12 (1994) 847e871.

[52] A. Favre-R�eguillon, G. Lebuzit, J. Foos, A. Guy, A. Sorin, M. Lemaire, et al., Se-lective rejection of dissolved uranium carbonate from seawater using crossflow filtration technology, Sep. Sci. Technol. 40 (2005) 623e631, http://dx.doi.org/10.1081/SS-200042529.

[53] J. Knutsson, S. Rauch, G.M. Morrison, Estimation of measurement un-certainties for the DGT passive sampler used for determination of copper inwater, Int. J. Anal. Chem. 2014 (2014).

[54] D.A. Nimick, C.H. Gammons, S.R. Parker, Diel biogeochemical processes andtheir effect on the aqueous chemistry of streams: a review,, Chem. Geol. 283(2011) 3e17, http://dx.doi.org/10.1016/j.chemgeo.2010.08.017.