in situ speciation measurements of trace metals in headwater streams

In Situ Speciation Measurements ofTrace Metals in Headwater StreamsK E N T W . W A R N K E N , † A L A N J . L A W L O R , ‡

S T E P H E N L O F T S , ‡ E D W A R D T I P P I N G , ‡

W I L L I A M D A V I S O N , * , † A N D H A O Z H A N G †

Lancaster Environment Centre, Lancaster University, Bailrigg,Lancaster LA1 4YQ, United Kingdom, and Centre for Ecologyand Hydrology, Lancaster Environment Centre,Bailrigg, Lancaster LA1 4AP

Received January 13, 2009. Revised manuscript receivedMay 6, 2009. Accepted May 8, 2009.

Concentrations of Al, Fe, Mn, Ni, Cu, Cd, Pb, and Zn weremeasured using DGT (diffusive gradients in thin-films) devicesdeployed in situ in 34 headwater streams in Northern England.Mean values of filtered samples analyzed by ICP-MS (inductivelycoupled plasma mass spectrometry) were used, along withDOC (dissolved organic carbon), pH and major ions, to calculatethe distribution of metal species using the speciation codeWHAM. DGT-measured concentrations, [Me]DGT, of Zn and Cdwere generally similar to concentrations in filtered samples,[Me]filt. For the other metals, [Me]DGT was similar to or lower than[Me]filt. Calculation of the maximum dynamic metal from thespeciation predicted using WHAM showed that most of the lowervalues of [Cu]DGT could be attributed to the dominance ofCu-fulvic acid complexes, which diffuse more slowly thansimple inorganic species. Similar calculations for Al, Pb, andMn were consistent with appreciable proportions of these metalsbeing present as colloids that are not simple complexes withhumic substances. Differences between WHAM predictions andthe measured [Ni]DGT indicated that WHAM used with thedefault binding parameters underestimates Ni binding to naturalorganic matter. Plots of [Me]DGT versus the ratio of boundmetal to DOC provided slight evidence of heterogeneous bindingof Pb and Cu, while results for Mn, Cd, and Zn were consistentwith weak binding and complete lability.

IntroductionThe technique of diffusive gradients in thin-films (DGT) isbeing used increasingly to measure trace metal species insitu (1, 2). The concentration of metal in solution is calculatedfrom the amount that accumulates on a Chelex binding layerafter it has diffused through a hydrogel, filter, and diffusiveboundary layer in solution (DBL). It is classed as a dynamictechnique because it detects a flux of metal species. Themeasured species must be mobile, that is, able to diffuseeasily through the hydrogel and filter, and labile, that is, ableto dissociate in the time scale associated with their transportthrough the diffusion layer (typically minutes) (3). With adynamic technique the derived concentrations of the col-lective kinetically available and mobile components requirefurther interpretation. If species in solution are labile, it ispossible to use the distribution of species predicted from

total dissolved concentrations using an equilibrium specia-tion model, and estimates of their diffusion coefficients, tocalculate the concentration of metal expected to be measuredby DGT (1, 4). Differences between prediction and measure-ment provide insight into the presence of relatively immobilespecies (colloids) or complexes which dissociate slowly, aswell as an assessment of model validity. When kineticlimitations apply, the species are said to be inert (notmeasured at all) or partially labile within the characteristictime scale of the measurement.

Two approaches have been used to extract kineticinformation. When DGT devices with a range of gel layerthicknesses are deployed in situ, information is gained onthe overall complex lability for each metal, providing a kineticsignature for the water being studied (5). Information ondissociation rates can be extracted, but the current treatmentis restricted to simple complexes which do not representwell the heterogeneous ligands of humic substances thatusually dominate freshwaters. Town et al. (6) have consideredthe role of heterogeneous ligands by examining how the DGTmeasurement depends on the ratio of the bound metal tothe ligand concentration. Their treatment of published DGTdata for a range of sites enabled estimation of the degree ofheterogeneity of the complexes and their dissociation rateconstants. The above measurements and interpretationsshow promise, but require further testing.

This study has examined the capability of DGT for routinemetal speciation measurements by performing in situmeasurements on 34 headwater streams in Northern En-gland. Measurement of the stream chemistry using traditionalsampling procedures allowed calculation of the distributionof metal species using an equilibrium speciation model.Comparison of measurements and predictions allowedconclusions to be drawn about metal complex lability, thepresence of inert species and the accuracy of modelparametisation. Measurements were not made at differentdiffusion layer thicknesses, as this approach to obtain kineticinformation is analytically challenging for such a large scalemonitoring study. Instead the opportunity was taken toexamine whether kinetic information could be derived fromthe ratio of the concentrations of bound metal and ligand.The results form part of a larger study to investigatebioaccumulation of metals and their toxic effects with respectto diverse biota (7).

Materials and MethodsSampling and analysis. Thirty four streams located in threeareas of northern England, the Lake District (LD), Teesdale,Tynedale and Weardale (TTW) and Howgill Fells, Ribblesdale,Swaledale and Whernside (HRSW) were sampled during 2006(Supporting Information (SI) Table S1). They are numbered1-35 (25 missing) for consistency with Tipping et al. (7).Twenty-six of the streams were in the vicinity of abandonedmine workings and had appreciable concentrations of at leastone of Cu, Pb, Zn, Cd, and Ni, but eight sites were unaffectedby mines (SI Table S2-9). Total dissolved metal concentra-tions measured during 2005 were used to determine thedeployment time of DGT devices, to avoid problems as-sociated with saturation of the resin. Most DGT devices weredeployed for 14 days, but for five streams the sampling timewas 24 h (SI Table S1). They were deployed as a “cluster” ofsix devices with three having 0.8 mm thick, open-pore, (APA)diffusive gels. See refs 8, 9 for precise information on thesepolyacrylamide gels cross-linked with an agarose derivative(APA). Data from the other three devices, which had differentgels, were not used because of problems. The plastic devices

* Corresponding phone: +441524593935; fax: +441524593985;e-mail: [email protected].

† Lancaster Environment Centre, Lancaster University.‡ Centre for Ecology and Hydrology.

Environ. Sci. Technol. 2009, 43, 7230–7236

7230 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 19, 2009 10.1021/es900112w CCC: $40.75 2009 American Chemical SocietyPublished on Web 05/29/2009

were cleaned by soaking them overnight (∼20 °C) in (1) 2%Decon; (2) 1.5 M HNO3 (Anal-R); (3) 1.5 M HNO3 (Aristar),and (4) 4 M HCl (Aristar). They were completely rinsed (18.2MΩ deionized water, DIW) between each overnight step. Toremove all residual acid from the devices (pH < 7), they weresoaked in DIW over several days. All devices were assembledunder Class-100 clean room conditions.

The cluster, which was held together using Dyna CableTeflon coated fishing line, was suspended within a plasticpipe (22 cm length × 15 cm width), using plastic cable ties,and placed on the streambed. The pipe was anchored on thesides and top using rocks collected from the streams, beingcareful not to restrict flow through the pipe. Recovery of theDGT devices required only cutting of the cable ties andimmediately placing them into acid cleaned zip-lock bags.

On return to the laboratory, the DGT devices were rinsedwith DIW and opened on a sample cart in front of a class-100laminar flow cabinet, within a class-1,000 clean room. Theresin-gels were placed in a vial containing 1 mL of 1 M HNO3.Eluates from samples collected from sites 1-26 were analyzedusing a Perkin-Elmer Elan DRC II ICP-MS (inductivelycoupled plasma mass spectroscopy) and the other eluateswere analyzed using a Thermo XSeries2 ICP-MS.

Concentrations of metals measured by DGT, [Me]DGT, werecalculated using eq 1, where M is the mean accumulatedmass of metal, ∆g is the thickness of the gel (0.08 cm), ∆f isthe thickness of the filter (0.014 cm), and δ is the thicknessof the diffusive boundary layer (10). A representative valuefor a flowing stream of δ ) 0.029 cm was assumed (10). Thediffusion coefficients of metals in the APA gels used here,Dgel have been found to be related to the diffusion coefficientsin water, Dw, by Dgel ) 0.85Dw (8). Values appropriate to themean measured water temperature at deployment andretrieval were used. The diffusion coefficients of each metalin the filter (Df) were assumed to be the same as those in theAPA gel. The effective exposed area of the filter overlying thegel, Ae, was taken to be 3.8 cm2 and t was the exposure timein seconds. In practice the concentrations obtained usingthis complete approach were within 4% of those calculatedusing the regular DGT equation (eq 2) using a geometricarea, Ag, of 3.14 cm2, a total diffusion layer thickness, ∆dl, of0.094 cm and disregarding the diffusive boundary layer.Lateral diffusion in the gel causes Ae to be greater than Ag

(10).

Two procedures were used to determine the concentrationsof trace metals in filtered water, [Me]filt, collected at the timesof DGT deployment and retrieval. At 26 sites (SI Table S1)duplicate samples were collected directly into plastic syringesand filtered through precleaned, plastic filter holders loadedwith 0.4 µm Nuclepore membranes, into 15 mL precleanedpolyethylene bottles containing 0.3 mL of HNO3, whichyielded a final acid concentration of 2% (v/v). The bottleswere placed into double acid-cleaned zip-lock bags fortransport back to the laboratory where they were analyzedby ICP-MS (Thermo XSeries2). For all sites, close to the syringesampling where applicable, samples were collected by handimmersion of 500 mL acid-washed polyethylene bottles. Theywere sealed in polyethylene bags and stored in cool boxesfor transport to the laboratory. Within 24 h they were filteredusing 0.45 µm polypropylene filter devices (Whatman,Puradisc) and acidified with 1% HNO3 (Baker, Ultrex II).

Concentrations were determined using a Perkin-Elmer ElanDRC II ICP-MS for laboratory filtered samples. For the fivesites with 24 h DGT deployments, samples were only collectedon retrieval.

Water samples for major cations and anions (laboratoryfiltered), pH, and dissolved organic carbon (DOC) were alsocollected at all sites within four time periods during 2006(March 6-8, March 20-22, April 3-5, April 17-19). Samplingand analysis procedures have been described (7). Samplesfor trace metals (laboratory filtered) were also collected atthese times, with the April dates coinciding with DGTdeployment and retrieval.

Modeling. Calculations of streamwater chemical spe-ciation were performed using WHAM (11) incorporatingHumic Ion-Binding Model VI (12), as described in detailfor the same study sites (7). Input concentration and pHdata for the model were averages of the measured valuesat the start and finish of the DGT deployments, except incases where the deployments were for one day only, forwhich analytical data from a single water sample wereused. The concentrations of Na, Mg, K, Ca, Cl, NO3, SO4

and trace metals Ni, Cu, Zn, Cd, and Pb in filtrates wereassumed to represent truly dissolved components, that isinorganic species and metal bound to dissolved organicmatter (DOM). Any metal associated with mineral colloidswas assumed to be negligible. We assumed that aluminumin filtrates was in true solution, unless the preliminaryspeciation calculation based on this total dissolved valueshowed Al3+ activity exceeded that expected from asolubility product of 108.5 (at 25 °C) for the reaction Al(OH)3

+ 3H+ ) Al3+ + 3H2O. In that case the activity of Al3+ wasassumed to be controlled by Al(OH)3 solubility, correctedfor temperature using a reaction enthalpy of -107 kJ mol-1

(13, 14). The activity of Fe3+ was estimated from theempirical equation of Lofts et al. (15), which is animprovement on estimation from a single solubilityproduct, used by Tipping et al. (7). The “binding activity”of DOM was estimated to be equivalent to that of averageisolated fulvic acid at a concentration of 65% of the DOMconcentration, assumed to be twice the DOC concentration.

Results and DiscussionGeneral Water Chemistry. The chemistry of the streams waswide-ranging, with pH varying from 4.1 to 8.3, DOC from 0.6to 13.2 mg L-1and alkalinity by Gran titration from near zeroto 2.33 milli-equivalents L-1 (SI Table S1). Site 11 was stronglyinfluenced by a local disused Pb mine, and its high sulfateconcentration and low pH (4.1) suggest that pyrite oxidationis occurring. Site 21 also had high sulfate, again probablyfrom pyrite, but the high pH (7.8) is consistent withweathering of Mg, K, and Ca, which were all high.

The relative standard deviations (RSDs) for the duplicatesamples of total dissolved metals collected at the start andend of DGT deployment were usually less than 5% (SI TablesS2-S9). Due to its very low concentrations, the RSDs for Cdwere higher, but they were all less than 18%. Generally themean measured concentrations at the start and end of DGTdeployment were well within 50% of one another. However,for Al, Mn, and Ni the difference could be as high as 100%for sites 13, 14, and 32, indicating that at these sitesconcentrations changed appreciably during deployment,presumably due to changing flows, as indicated by differentwater levels on retrieval and deployment.

Concentrations of Pb, Zn, and Cd spanned approximately4 orders of magnitude; Al, Mn, and Ni 3 orders of magnitude;and Cu less than 2 orders of magnitude. Zn concentrationswere in some cases very high and greatly exceeded those ofPb and Cu, even though there is a history of Cu and Pb miningin parts of the region. The highest concentrations of all metalsexcept Fe occurred at site 11.

CDGT )M( ∆g

Dgel+

∆f

Df+ δ

Dw)

Aet(1)

CDGT )M∆dl

DgelAgt(2)

VOL. 43, NO. 19, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 7231

Dependence of DGT Measurements on Filterable Metal.All DGT devices were successfully retrieved and analyzed.DGT measurements were generally reproducible, withaverage RSDs varying between 6% for Cd and 12% for Cuand Pb.

Plots of log [Me]DGT versus log[Me]filt were used tocompare the paired data sets (Figure 1). [Me]filt was themean for samples collected at DGT deployment andretrieval, except for the 24 h deployments when only themeans of the duplicates at retrieval were used. For bothZn and Cd, [Me]DGT and [Me]filt were generally very similar,as indicated by the close adherence to the 1:1 line. Thisgood correspondence is consistent with predictions usingWHAM that in most cases less than 20% of the metal iscomplexed with DOC (SI Table S9). For Zn, there was 1very clear outlier (site 16) and two less marked (sites 22,30) above the line, and one outlier below (site 29). Site 22is one of the three Nenthead streams (21-23) leaving a

mine area. Its zinc concentration is noticeably lower thanthe very high values of the other two streams, where highsulfate reflected pyrite oxidation. As the deployment timewas only one day, and only one sample for total dissolvedmetals was taken, meaningful comparison of DGT andspot samples is questionable. For site 16, [Zn]filt measuredon samples collected at the start and end of the 14 dayDGT deployment were within 10%. There is no obviousrationale for the discrepancy between [Zn]DGT and [Zn]filt

for sites 16 and 30. Mean [Zn]filt at site 29 decreased slightlyduring deployment from 1.58 to 1.22 µmol L-1, but theDGT measurement was much lower at 0.2 µmol L-1.Explanation in terms of complexation is unlikely, as theWHAM calculation suggests that only 8% is organicallycomplexed (SI Table S9). Except for site 16, the Cd plothad the same obvious outliers as identified for Zn. Thisagreement between these two metals with similar chem-istries is consistent with systematic rather than random

FIGURE 1. Log of the DGT measured concentration of each metal plotted against the log mean concentration in filtrates for eachdeployment. The ideal 1:1 line is shown. Data are grouped according to pH: triangles >7, circles 6-7, squares <6; and DOC: opensymbols e3.4 mg L-1, filled symbols >3.4 mg L-1.

7232 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 19, 2009

causes for the observed discrepancies between [Me]DGT

and [Me]filt, as discussed above.Apart from two obvious outliers above the line, corre-

sponding, as for Zn and Cd, to sites 21, 22, and 30, most Mndata lie on or below the line. The reasons for the systematicspread of data below the 1:1 line, which is more marked inthe cases of Ni, Cu, Pb, and Al, are discussed fully in the nextsection. For Ni and Cu there are no outliers above the lineand for Pb there is only one, corresponding to site 16. Thetwo outliers above the line for Al (sites 1 and 21) do notcorrespond to those observed for other metals.

When [Fe]filt < 3.10-6 M, [Fe]DGT was fairly constant at5-16 10-8 M, only approaching [Fe]filt at the lowest measuredvalues. For this group of data, DOC was less than 3.4. When[Fe]filt > 3.10-6 M, [Fe]DGT spanned the range 0.067-7.2.10-6

M, approaching the 1:1 line. DOC was greater than 3.4 forthese data except for the point actually on the 1:1 line, whichwas for site 11, with the exceptionally low pH of 4.1. It appearsthat at very low pH or Fe concentrations, Fe is mainly insolution and therefore fully measured by DGT. As pH and/orFe concentration increases, colloidal forms dominate, withthe proportion that can be maintained in solution increasingin the presence of DOC, due to the formation of dissolvedcomplexes with fulvic acid.

Predicting the DGT Measured Concentrations. Thecalculation of the DGT concentration using eqs 1 or 2 usesonly the diffusion coefficient of the free metal ion. However,for a fully labile system, where all species are assumed tobe in equilibrium, the amount of metal accumulated byDGT is proportional to the sum of the concentration ofeach species times their diffusion coefficients (16). Infreshwaters the dominant ligand is usually fulvic acid.Experiments with diffusion cells have shown that, to agood approximation, the mean diffusion coefficients ofmetal-fulvic-acid complexes in the APA gel are typically20% of the diffusion coefficients of the free metal ions (8).Consequently, in solutions where a high proportion of themetal is bound to fulvic acid, the DGT measured con-centration would be expected to be less than the totaldissolved metal. The DGT-measured metal has beenreferred to as the dynamic metal because it reflects thetransport rate of species (3). The dynamic metal has itsmaximum value, termed [Me]max

dyn , when all species are fully

labile (able to dissociate rapidly and be measured (4). Ifthere are only simple inorganic species, Minorg, and metal boundto fulvic acid, MFA, it can be calculated using eq 3.

The predicted distribution of species available from the WHAMcalculation provides concentrations of Minorg and MFA andhence an estimate of [Me]max

dyn , which can be compared to[Me]DGT (2). All inorganic species are reasonably assumed (17)to have the same diffusion coefficient as the free metal ion. Thesimple approach of eq 3 does not accommodate the differencein the ratio of the diffusion coefficients in the DBL, but as thisaccounts for only 24% of the total thickness any associatederror will be less than 10%.

The dependence of the DGT measured concentrationon[Me]max

dyn is shown in Figure 2 for Al, Mn, Cu, and Pb. ForNi, Cd, Zn, and Fe there was very little difference betweenthese plots and the plots of [Me]DGT versus [Me]filt. Outliersidentified for Zn and Cd appeared in both data sets. Thenegligible difference is not surprising, as the proportion ofmetal bound to fulvic acid has to be high before there willbe a noticeable effect. For example when 50% of the metalis bound to FA, [Me]max

dyn is 60% of the value when there is nobinding to FA. Therefore use of the [Me]max

dyn plot would onlybe expected to move the data point 0.22 log units closer tothe 1:1 line. If there is only 25% bound, the shift would be0.1 log units. Less than 3% of the total Fe present in thefiltrate is predicted to be bound for all sites (SI Table S5),with most of the remaining 97% being colloidal iron oxide.Only at site 13 is Ni more than 25% bound (SI Table S7). Znand Cd are also predicted to be more than 50% bound at site13, but for most of the remaining sites less than 25% is bound(SI Tables S3 and S9).

The DGT measurement may be less than [Me]maxdyn for four

main reasons: (1) there may be colloidal metal with sub-stantially lower diffusion coefficients than the metal fulvicacid complexes (note that here we define colloidal metal asfilterable, but not in true solution or bound to FA), (2) therelease of metal ions from the fulvic acid binding sites maybe insufficiently quick to sustain the demand for metal byDGT, (3) there may be some inert metal species that do not

FIGURE 2. Log of the DGT measured concentration of Cu, Pb, Mn, and Al plotted against the log dynamic maximum concentration.The ideal 1:1 line is shown. Data are grouped according to pH: triangles >7, circles 6-7, squares <6; and DOC: open symbols e3.4mg L-1, filled symbols >3.4 mg L-1.

[Me]maxdyn ) ∑ (Minorg + 0.2MFA) (3)


dissociate within the measurement time, or (4) WHAM mayunderestimate metal binding to FA.

A substantial proportion of [Ni]DGT is less than[Ni]maxdyn .

Except for one site (34), all the substantial deviations belowthe line are for sites with DOC > 3.4 mg L-1. Recalculationof [Ni]max

dyn , assuming 100% complexation, moves most pointsvery close to the line. Moreover, dissociation of Ni from itscomplexes is known to be slow compared to other transitionmetals (19). If a very small proportion is bound, as predicted,this would make little difference, but if there is more boundthan predicted, this kinetic limitation would lower [Ni]DGT

compared to [Ni]maxdyn

..Compared to Ni, a slightly greater proportion of Mn is

predicted, as previously suggested (18), to be bound (>40%for nine of the sites (SI Table S6)). This has the effect ofslightly improving the agreement between [Mn]DGT and[Mn]max

dyn over that between [Mn]DGT and [Mn]filt. Other thansamples not being comparable, there is no reasonable reasonfor [Mn]DGT to be greater than [Mn]filt and the two sitescorresponding to the data points that lie above the 1:1 linehave already been identified as outliers. A substantialproportion of the data lie below the 1:1 line. Inert species arenot expected for Mn and complexes with fulvic acid dissociaterapidly (5). The likely candidate is colloidal oxides, assuggested by the bryophyte studies of these waters (7).

The high proportion of Cu that is bound to FA has theeffect of bringing the points for the [Cu]DGT versus [Cu]max

dyn

plot much closer to the 1:1 line than for the [Cu]DGT versus[Cu]filt plot. However, for a high proportion of the measure-ments, [Cu]DGT was less than[Cu]max

dyn . Kinetic limitationassociated with a proportion of Cu occupying strong bindingsites has been reported previously (5). Most of the data belowthe line are for sites with DOC > 3.4 mg L-1 (SI Table S4). Theexceptions are at very low Cu concentrations. It would beexpected that kinetic limitation would be greatest when Cuconcentrations are low, favoring a greater proportion oc-cupying strong binding sites of FA.

While there is generally a high proportion of Pb boundto FA, it is not predicted to be as dominant as for Cu (SITables S4 and S8). Compared to the plot of [Pb]DGT versus[Pb]filt, a higher proportion of the points on the plot of [Pb]DGT

versus [Pb]maxdyn are close to the 1:1 line, However, there is a

set of data where [Pb]DGT is approximately an order of

magnitude less than [Pb]maxdyn . These correspond to most of

the HRSW sites and some of the TTW sites. For all sites inthe outlying set of data DOC is fairly high (>3.4) and pH isquite high, suggesting that there may be more organicallybound Pb than predicted. However, it is most probable thatthese lower values of [Pb]DGT are due to the presence ofcolloidal Pb. A similar large difference between [Pb]DGT and[Pb]max

dyn was attributed previously to colloidal Pb (4). Pb bindsstrongly to particles generally, but particularly to MnO2. Thedifference between [Me]DGT and [Me]max

dyn was used to esti-mate colloidal Fe and Mn. Calculations using WHAM showedthat all the colloidal Pb could be accounted for by adsorptionto MnO2, with adsorption to iron oxides being relativelyunimportant.

Data in the plot of [Al]DGT versus [Al]maxdyn are only slightly

more regularly grouped about the 1:1 line than observed forthe [Al]DGT versus [Al]filt plot. There are a set of data for highpH and DOC sites where [Al]DGT is much less than[Al]max

dyn .Where the WHAM calculation indicated supersaturation withrespect to Al(OH)3, the Al concentrations were calculatedassuming equilibrium with Al(OH)3. However, the consis-tently lower [Al]DGT than [Al]max

dyn for the set of outlying datasuggest that the solubility of Al(OH)3 in these streams maybe less than assumed, which is not unreasonable given thatreported values of logKso for Al(OH)3 range ( 1 of the valueused (14). Alternatively, fine clay colloids may be present.Only for some of the sites is the binding of Al to FA sufficientto account for the difference between [Al]DGT and [Al]max

dyn , butit is possible that the WHAM calculation may have under-estimated the binding.

Kinetic Limitations and Heterogeneity Effects. As arepresentative constituent of natural organic matter, FA canbe regarded as a heterogeneous ligand with a distribution ofbinding sites. The occupancy of the sites and therefore theeffective binding strength depends on the metal to ligandratio (20). The heterogeneity can be described by a hetero-geneity parameter, Γ, which provides a measure of thevariation in the affinities of binding sites that can be availableto a metal (21, 22). Town et al. (6) have shown that it ispossible to obtain Γ from DGT measurements. They assumedthat metal, Me, is distributed between i sites of natural organicmatter, Si, of total concentration ∑i[Me]i so that the totalbound species are ∑i[MeS]i. The average dissociation rate

FIGURE 3. Log of the DGT measured concentration of Cu, [Cu]DGT, and [Cu]DGT minus the concentration of the free ion, [Cu′]DGT, plottedagainst log([Cu]filt/DOC) and log([Cu]org/DOC). Linear regression lines are shown. Data are grouped according to pH: triangles >7,circles 6-7, squares <6; and DOC: open symbols e3.4 mg L-1, filled symbols >3.4 mg L-1.


constant for the sites, jkd, is then given by eq 5, where ka isthe characteristic association rate constant for Me and B isa constant related to complex stability.

When most of the metal is complexed and there is slowdissociation, the mass accumulated by DGT is primarily underkinetic control. Then the DGT measured flux and interpretedconcentration are directly proportional to jkd[MeS]i provided(∑i[S]i)(1-Γ)/Γ does not vary (3). Town et al. (6) have shownthat for Cu, which is strongly complexed, the filterable metal(in mol L-1) can be used as a reasonable approximation for∑i[CuS]i. They also found pragmatically that the concentrationof DOC (in g L-1) can be substituted for (∑i[S]i)(1-Γ)/Γ. As theonly concern is the scatter of the data and the slope, it isacceptable to mix units. The slope of this plot effectivelyprovides 1/Γ because the DOC term varies much less thanfilterable metal.

A straight line could be reasonably fitted to the plot oflog[Cu]DGT versus log([Cu]filt/DOC) (Figure 3), with the slope(SI Table S10) corresponding to Γ)1.03. For a true kineticallylimited situation, a value of Γ approaching 1 indicates ahomogeneous ligand. Progressively lower values indicateincreasing heterogeneity, with 0.5 or less being typical ofestablished heterogeneous binding. When the data for theother metals were treated similarly, the fits were generallyas good as for Cu (SI Figure S1) and the derived values of Γwere scattered about 1. These results provide little evidencefor kinetic limitation.

By assuming that the speciation using WHAM is correct,the approximated estimate of ∑i[MeS]i can be improved.Rather than use [Me]filt, we can use [Me]org, the totalconcentration of the metal fulvic species. With correlationcoefficients (R2) of 0.15 and 0.13 for the Fe and Ni plotsrespectively, use of the regression coefficient was not justified(SI Table S10). For the other metals there was no evidenceof heterogeneity. To exclude the metal which is not com-plexed, the concentration of the free ion was subtracted fromthe DGT-measured concentration. This procedure had littleeffect on Γ values when [Me]filt/DOC was used. When thefree ion correction and [Me]org/DOC were used together (SITable S10) the lowest values of Γ were obtained for Pb andCu. They would be expected to show a degree of heterogeneityas there is generally substantial bound metal. However, theΓ values of 0.85 (Cu) and 0.84 (Pb) were not substantiallydifferent from the values for the other metals. They are alittle higher than those obtained previously using DGT of 0.7(Cu) and 0.8 (Pb) (Town et al., 2009) and substantially higherthan estimates from voltammetry (22-24). The Γ values forZn, Cd, and Mn of 0.92-1.39 are consistent with weakcomplexation and high lability.

In the above treatment, data for sites 21-23, where thecomparability of [Me]DGT and [Me]filt data is questionable,were omitted, but this had little effect on Γ values. For thefive sites where Al(OH)3 was supersaturated, [Al]filt used thecalculated total Al. This approach to investigate heterogeneityshould only be used for a set of data with a narrow pH range(6). However, including all data in the plots against [Me]filt/DOC did not affect Γ appreciably. As the low pH site (11) wasa clear outlier on the plots against Corg/DOC, it was omitted.Eliminating data for sites with DOC < 3.4 to favor a greaterproportion of complexed and potentially kinetically limitedmetal did not appreciably change Γ.

Interpretation of this substantial data set in terms of Cmaxdyn

showed there was a high degree of consistency between thespeciation measured using DGT and predicted by WHAM

from the total filtered concentrations. Exceptions for Al, Pb,and Mn were consistent with the presence of colloidal (notMFA) forms, whereas for Ni the evidence suggests that, whenusing the default binding parameters, WHAM underestimatesorganic complexation. Except for Cu, there was generallylittle evidence that complex dissociation kinetics affectedthe DGT measurement in these waters. As noted by others(25), the combined use of these approaches can greatlyincrease confidence in assessing water quality.

AcknowledgmentsWe are grateful to S. A. Thacker and C. D. Vincent for fieldassistance, the CEH Lancaster Environmental AnalyticalGroup for their assistance with streamwater chemicalanalyses, J. Hamilton-Taylor for helpful discussions, and R.Town for input on heterogeneity aspects. This work wasundertaken within the project “Environmental QualityStandards for trace metals in the aquatic environment” jointlyfunded by the Environment Agency of England and Wales,the European Copper Institute, European Nickel IndustryAssociation, International Cadmium Association, Interna-tional Zinc Association (Europe), Rio Tinto, and the ScottishEnvironment Protection Agency.

Supporting Information AvailableSite locations, pH, DOC, major ion and metal concentrations,DGT measured concentrations with deployment times anddates, values of the heterogeneity factor, plots relating tobinding heterogeneity for each metal and derived values ofΓ. This material is available free of charge via the Internetat http://pubs.acs.org.

Literature Cited(1) Meylan, S.; Odzak, N.; Behra, R.; Sigg, L. Speciation of copper

and zinc in natural freshwater: comparison of voltammetricmeasurements, diffusion gradients in thin-films (DGT) andchemical equilibrium models. Anal, Chim. Acta 2004, 510, 91–100.

(2) Gimpel, J.; Zhang, H.; Davison, W.; Edwards, A. C. In situ tracemetal speciation in lake surface waters using DGT, dialysis andfiltration. Environ. Sci. Technol. 2003, 37, 138–146.

(3) van Leeuwen, H. P.; Town, R. M.; Buffle, J.; Cleven, R. F. M. J.;Davison, W.; Puy, J.; van Riemsdijk, W. H.; Sigg, L. Dynamicspeciation analysis and bioavailability of metals in aquaticsystems. Environ. Sci. Technol. 2005, 39, 8545–8556.

(4) Unsworth, E.; Warnken, K. W.; Zhang, H.; Davison, W.; Black,F.; Buffle, J.; Cao, J.; Cleven, R. F. M. J.; Galceran, J.; Gunkel, P.;Kalis, E. J. J.; Kistler, D.; Van Leeuwen, H. P.; Martin, M.; Noel,S.; Nur, Y.; Odzak, N.; Puy, J.; Van Riemsdijk, W. H.; Sigg, L.;Temminghoff, E. J. M.; Tercier-Waeber, M.-L.; Toepperwien, S.;Town, R. M.; Weng, L. P.; Xue, H. Model predictions of metalspeciation in freshwaters compared to measurements by in situtechniques. Environ. Sci. Technol. 2006, 40, 1942–1949.

(5) Warnken, K. W.; Davison, W.; Zhang, H.; Galceran, J.; Puy, J. Insitu measurements of metal complex exchange kinetics infreshwater. Environ. Sci. Technol. 2007, 41, 3179–3185.

(6) Town, R. M.; Chakraborty, P.; van Leeuwen, H. P. DynamicDGT speciation analysis and application to natural heteroge-neous complexes. Environ. Chem., in press.

(7) Tipping, E.; Vincent, C. D.; Lawlor, A. J.; Lofts, S. Metalaccumulation by stream bryophytes, related to chemicalspeciation. Env. Poll. 2008, 156, 936–947.

(8) Scally, S.; Davison, W.; Zhang, H. Diffusion coefficients of metalsand metal complexes in hydrogels used in diffusive gradientsin thin films. Anal. Chim. Acta 2006, 558, 222–229.

(9) Warnken, K. W.; Davison, W.; Zhang, H.; Galceran, J.; Puy, J.Interpretation of in situ speciation measurements of inorganicand organically complexed trace metals in freshwater by DGT.Environ. Sci. Technol. 2008, 42, 6903–6909.

(10) Warnken, K. W.; Zhang, H.; Davison, W. Accuracy of the diffusivegradients in thin-films technique: diffusive boundary layer andeffective sampling area considerations. Anal. Chem. 2006, 78,3780–3787.

kd[MeS]i ) kaB-1/Γ(∑

i

[MeS]i)1/Γ

(∑i

[S]i)(1-Γ)/Γ(5)


(11) Tipping, E. WHAM-a chemical equilibrium model and computercode for waters, sediments and soils incorporating a discretesite/electrostatic model of ion-binding by humic substances.Comput. Geosci. 1994, 20, 973–1023.

(12) Tipping, E. Humic ion-binding Model VI: an improved descrip-tion of the interactions of protons and metal ions with humicsubstances. Aquat. Geochem. 1998, 4, 3–48.

(13) Tipping, E.; Rey-Castro, C.; Bryan, S. E.; Hamilton-Taylor, J.Al(III) and Fe(III) binding by humic substances in freshwaters,and implications for trace metal speciation. Geochim. Cosmo-chim. Acta 2002, 66, 3211–3224.

(14) Tipping, E. Modelling Al competition for heavy metal bindingby dissolved organic matter in soil and surface waters of acidand neutral pH. Geoderma 2005, 127, 293–304.

(15) Lofts, S.; Tipping, E.; Hamilton-Taylor, J. The Chemical Spe-ciation of Fe(III) in Freshwaters. Aquat. Geochem. 2008, 14, 337–358.

(16) Zhang, H. In situ speciation of Ni and Zn in freshwaters:comparison between DGT and speciation models. Environ. Sci.Technol. 2004, 38, 1421–1427.

(17) Buffle, J.; Zhang, Z.; Startchev, K. Metal flux and dynamicspeciation at (bio)interfaces. Part I: Critical evaluation andcompilation of physico-chemical parameters for complexes withsimple ligands and fulvic/humic substances. Environ. Sci.Technol. 2007, 41, 7609–7620.

(18) Van Laer, L.; Smolders, E.; Degryse, F.; Janssen, C.; Scham-phelaere, K. A. C. Speciation of nickel in surface waters measured

with the Donnan membrane technique. Anal. Chim. Acta 2006,578, 195–202.

(19) Morel, F. M. M.; Hering, J. G. Principles and Applications ofAquatic Chemistry; Wiley: New York, 1993; p 400.

(20) Tipping, E. Cation Binding by Humic Substances; CambridgeUniversity Press: Cambridge, 2002.

(21) Buffle, J. Complexation Reactions in Aquatic Systems: anAnalytical Approach; Ellis Horwood: Chichester, 1998.

(22) Town, R. M.; Filella, M. Dispelling the myths: is the existenceof L1 and L2 ligands necessary to explain metal ion speciationin natural waters. Limnol. Oceanogr. 2000, 45, 1341–1357.

(23) Filella, M.; Town, R. M. Determination of metal ion bindingparameters for humic substances. Part 1. Application of a simplecalculation method for extraction of meaningful parametersfrom reverse pulse polarograms. J. Electroanal. Chem. 2000,485, 21–33.

(24) Chakraborty, P.; Fasfous, I. I.; Murimboh, J.; Chakarabarti, C. L.Simultaneous determination of speciation parameters of Cu,Pb, Cd, and Zn in model solutions of Suwannee River fulvicacid by pseudopolarography. Anal. Bioanal. Chem. 2007, 388,463–474.

(25) Balistrieri, L. S.; Blank, R. G. Dissolved and labile concentrationsof Cd, Cu, Pb, and Zn in the South Fork Coeur d’Alene River,Idaho: Comparisons among chemical equilibrium models andimplications for biotic ligand models. Appl. Geochem. 2008, 23,3355–3371.

ES900112W


in situ speciation measurements of trace metals in headwater streams

Documents