Aquatic Geochemistry

, Volume 16, Issue 1, pp 151–172

Determination of Free Cd, Cu and Zn Concentrations in Lake Waters by In Situ Diffusion Followed by Column Equilibration Ion-exchange

Authors

    • INRS-Eau, Terre et EnvironnementUniversité du Québec
  • Y. Couillard
    • Ecological Assessment DivisionEnvironment Canada
  • B. Vigneault
    • Natural Resources Canada
  • P. G. C. Campbell
    • INRS-Eau, Terre et EnvironnementUniversité du Québec
Original Paper

DOI: 10.1007/s10498-009-9074-3

Cite this article as:
Fortin, C., Couillard, Y., Vigneault, B. et al. Aquat Geochem (2010) 16: 151. doi:10.1007/s10498-009-9074-3

Abstract

Combining in situ diffusion and column ion-exchange equilibration, we measured free metal ion concentrations (Cd, Cu and Zn) in water samples collected from the epilimnion of 14 lakes in the Rouyn-Noranda area (600 km north-west of Montreal, QC, Canada). Lakes were selected to represent a wide range of physico-chemical characteristics (hardness, pH, dissolved organic matter—DOM, degree of metal contamination), to determine the influence of these parameters on metal speciation. Total dissolved metal concentrations, as determined within the diffusion cells, varied over one to two orders of magnitude: [Cd] 0.19–2.9 nM; [Cu] 36–190 nM; [Zn] 7–2,800 nM. The proportion of total dissolved metal present as free Cd2+ and Zn2+ was relatively constant for the 14 selected lakes, despite the wide pH (4.5–8) and DOM (3–23 mg C/L) ranges, probably reflecting the inverse relationship observed between pH and DOM; this proportion did, however, vary with DOM and pH for Cu. Our experimental free metal ion concentrations were compared with those calculated with the thermodynamic models WHAM (Windermere Humic Aqueous Model VI) and ECOSAT 4.7 (incorporating the NICA-Donnan model). Measured and calculated values were in reasonable agreement for both Cd and Zn although measured values were generally slightly higher, i.e. less than one order of magnitude. For several lakes, measured free Cu concentrations were, however, much higher than the calculated values, suggesting that these models overestimate Cu complexation. The gap between measured and calculated free metal ion concentration becomes more important as the total metal concentration decreases and as pH increases.

Keywords

Metal speciationIn situ diffusion samplingLake waterWHAM modelNICA-Donnan modelDissolved organic matter

1 Introduction

One of the major challenges in environmental analytical chemistry is to develop methods for the determination of metal speciation in natural waters containing complex mixtures of heterogeneous dissolved organic matter. The ability to determine free metal ion concentrations ([Mz+]) in such matrices is of particular importance, given the compelling laboratory evidence that metal bioaccumulation and toxicity generally vary as a function of the free metal ion concentration in solution (Campbell 1995). However, many of the existing analytical techniques are not specific to the free metal ion, are not sensitive enough to measure environmentally relevant concentrations or suffer from interferences due to adsorption of surface-active organic matter. For example, anodic stripping voltammetry (Rozan et al. 1999) and the diffusion gradients in thin films (DGT) technique (Zhang and Davison 1995) respond not to the free metal ion alone but rather to labile metals. Potentiometry is selective for the free metal ion but with the possible exception of Cu2+, ion-selective electrodes are insufficiently sensitive to measure environmentally relevant Mz+ concentrations in unbuffered solutions. Ligand-exchange techniques, which use metal-specific strong ligands coupled with a voltammetric detection step, are elegant but are only applicable to well-buffered metal systems, and the method can be tedious and time-consuming especially when dealing with several metals (Xue and Sigg 1999). The recently developed hollow fibre permeation liquid membrane and Donnan membrane techniques are promising, but recent comparisons showed poor agreement when the two methods were used to determine free Cd2+ ion concentrations in the same water samples (Sigg et al. 2006; Unsworth et al. 2006).

An alternative to measuring free metal ion concentrations in natural waters would be to calculate the equilibrium speciation of the metal. Thermodynamic equilibrium calculations are now possible with computer speciation models that consider the polyelectrolytic behaviour of metal-complexing humic and fulvic acids (Keizer and Van Riemsdijk 1999; Tipping 2002). However, such calculations rely on thermodynamic data for metal–humic and metal–fulvic interactions that are often derived from laboratory titration experiments. Relatively few of these experiments have been performed at environmentally realistic metal:ligand ratios, and virtually all of them have been carried out on isolated humic and fulvic acids (isolation procedures typically involve acidification and neutralisation steps, and may well alter the properties of the isolated material). The extrapolation of the calculated results to natural waters is accordingly somewhat uncertain (Unsworth et al. 2006).

We have previously demonstrated that an ion-exchange column equilibration method could be used to determine free Cd2+ and Zn2+ in laboratory solutions containing various inorganic and organic ligands, including fulvic acid (Fortin and Campbell 1998; Fortin and Caron 2000; Néron et al. 2006; Vigneault and Campbell 2005). This technique has been used by others as well (Cantwell et al. 1982; Doig and Liber 2007; Ge et al. 2005a, b; Gopalapillai et al. 2008; Holm et al. 1995; Sweileh et al. 1987; Worms and Wilkinson 2008), and its capacity to measure the free ion species was demonstrated in complex solutions (some containing natural organic matter) either by comparing the results with thermodynamic equilibrium calculations (Cantwell et al. 1982; Fortin and Campbell 1998; Sweileh et al. 1987; Worms and Wilkinson 2008) or with values obtained with a different technique such as ion-selective electrodes when natural organic matter was present (Ge et al. 2005a, b; Sweileh et al. 1987).

The present study was designed to use this ion-exchange technique (IET) to determine free trace metal (Cd2+, Cu2+ and Zn2+) concentrations in natural lake waters with various levels of trace metal contamination. These measured values were then compared with those predicted by the most recent version of the popular WHAM1 (version 6.0) thermodynamic equilibrium model as well as to those of the ECOSAT2 model (version 4.7). Lakes were selected to represent a wide range of dissolved metal concentrations and the diverse physico-chemical characteristics (hardness, pH, dissolved organic matter—DOM) commonly encountered in the lakes of the Canadian Shield. In this manner, we tested equilibrium models using experimentally determined free metal ion concentrations as measured in natural water samples.

2 Methodology

2.1 Reagents and Glassware

All plasticware was thoroughly washed by soaking in 10% (w/v) nitric acid for at least 24 h followed by rinsing a minimum of seven times with ultrapure water (>18 Mohms cm) and dried under a Class 100 laminar flow hood. Ultrapure Ca(NO3)2 and Mg(NO3)2 (Puratronic; Alfa AESAR), analytical grade KNO3 and Chelex purified NaNO3 [see Fortin and Campbell (1998) for a detailed description] were used as background electrolytes, and trace metal grade nitric acid (Fisher Scientific) was used for acidification of samples and resin elution.

2.2 Sampling Sites

Fourteen lakes were sampled on one to two occasions between May and September 1998 within a 50 km range of Rouyn-Noranda (Quebec), a city located approximately 600 km north-west of Montreal, Canada (48.3°N, 79°W). Extensive mining and smelting activities have taken place in this area since 1926. The local lakes can be affected by acid and metal inputs from current mining operations, abandoned and decommissioned mine sites, and atmospheric deposition from a copper smelter. Past metal and SO2 emissions from the smelter led to the development of distinct gradients in metal concentrations and pH in lakes located down-wind from the smelter. Scattered mining activities also contribute to metal contamination notably through acid mine drainage (major input source for lakes Dufault and Dasserat). Dissolved metal levels in the lakes have been determined periodically since ~1990, and the responses of some of the indigenous aquatic organisms to the metal and pH gradients have been abundantly reported [e.g. (Couillard et al. 1993; Croteau et al. 2002; Giguère et al. 2004; Perceval et al. 2002, 2004; Sherwood et al. 2000; Wang et al. 1999)]. Lake locations and key sampling sites are illustrated in Fig. 1. Up to five sampling locations per lake were selected depending on the lake surface area. At stations CA-1, VA-1 and HE-1, three replicate diffusion cells were deployed to determine measurement variability. Two sampling sites from Lake Bousquet were in a lotic system (BO-4 and BO-5; Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10498-009-9074-3/MediaObjects/10498_2009_9074_Fig1_HTML.gif
Fig. 1

Map and locations of sampling stations

2.3 Sampling and Analysis

Equilibrium diffusion sampling was preferred over direct filtration for several reasons: (i) it involves less sample manipulation and thus less chance of contamination; (ii) possible metal contamination from, or adsorption to, the container and filter membrane is minimised by equilibrium with the virtually infinite volume of the water column; (iii) sample collection over several days dampens any daily fluctuations in concentration; and (iv) diffusion sampling minimises problems associated with changes in effective pore size caused by progressive clogging (Horowitz et al. 1996). Diffusion cells consisted of polypropylene jars (Nalgene; total volume ~300 ml) capped with custom-made plastic closures designed to hold a filter membrane (0.2 μm HT Tuffryn; Pall). The cells filled with ultrapure water were placed in the epilimnion of the lakes by SCUBA-equipped divers and fixed to a plastic rod approximately 15 cm from the sediment surface at depths between 3 and 5 m (alternatively, the cells could be put in place directly from the surface with a weighted line and a floater). They were left to equilibrate with the surrounding waters for 11–17 days (generally 14 days). When brought to the surface, the membrane of each diffusion cell was washed with ultrapure water to remove particulate material. The filter membrane was pierced with a clean syringe tip, and the custom-made filtering closure was then removed and replaced with a clean cap. Piercing of the membrane prior to its removal prevented contamination of the sample by suction of unfiltered lake water trapped in the threads of the cap. All samples were cleanly bagged and flown on ice to the central laboratory in Quebec City within 24 h of retrieval from the lakes. Lake water pH was measured in situ at the same sites upon diffusion cell retrieval. On one occasion, eight samples were collected and sent to the laboratory where half of them were analysed right away, and the other half were kept in the refrigerator for a week before analysis. We observed no significant variations (P > 0.05) between both groups for all measured parameters, i.e. there was no contamination or loss observed during this delay.

Once in the laboratory, the sample containers were opened in a clean Class 100 laminar flow hood and five sub-samples were taken for analysis. A first 250 ml aliquot was saved for immediate free metal ion determination. Two further sub-samples (~15 ml) were pipetted into high-density polyethylene containers for the determination of major cations (Ca, Mg, K, Na) and total dissolved metal concentrations (Al, Fe, Mn, Cd, Cu, Zn). Both 15 ml sub-samples were acidified with concentrated nitric acid (final concentration 0.2% v/v). Depending on the metal to be analysed and the concentration range, a variety of analytical techniques were used for these determinations: inductively coupled plasma atomic emission spectroscopy (ICP-AES; AtomScan 25 spectrophotometer, Thermo Jarrell Ash), flame atomic absorption spectroscopy (FAAS; SpectrAA-20 spectrophotometer, Varian) and graphite furnace atomic absorption spectroscopy (GFAAS; SIMAA 6000, Perkin-Elmer). A fourth sub-sample (~10 ml) was dispensed into polystyrene containers for the determination of anions (Cl, NO3, SO4, PO4) by ion chromatography (DX-300 Gradient Chromatography Systems, Dionex), whereas the last sub-sample (~4 ml) was pipetted into high-density polyethylene containers for dissolved organic and inorganic carbon analyses with a total organic carbon analyser (TOC-5000A, Shimadzu). Natural waters are often oversaturated with CO2 and can potentially re-equilibrate with atmospheric pCO2 once in the lab. The last sub-sample was thus collected after the free metal ion measurements had been performed in order to obtain the final dissolved inorganic carbon concentrations of the sample at the moment of these analyses. All sub-samples were kept in the dark at 4°C until analysis.

In addition to measurements of dissolved organic carbon, we also determined the absorbance of some of the water samples at 465 (E4) and 665 nm (E6) with a UV–visible spectrometer (DMS-100, Varian). The E4 to E6 ratios were determined as a qualitative indication of the molecular weight of the humic substances, in order to compare the composition of the natural DOM among lakes (Chen et al. 1977). Absorbance was determined after adjustment of the sample pH (to ~9) and ionic strength (0.05 M NaHCO3), since both variables can affect absorbance readings.

2.4 Diffusion Kinetics

When in situ diffusion is used to collect natural water samples, it is obviously important to ensure that near-equilibrium conditions are established before the diffusion cells are recovered. To determine the time required to reach equilibrium in our diffusion cells, we installed 15 replicate cells in each of lakes Bousquet and Dufault. Three cells were retrieved after 3, 7, 10, 15 and 21 days. Sampling and analyses of Cd, Cu, Zn and the major cations, as well as determinations of the E4/E6 ratios, were performed as mentioned earlier. These results provided time-course concentration profiles for lake water solutes diffusing through the filter membrane into the diffusion cell. Based on the measured concentrations inside the retrieved diffusion cells, [C]diff, and the deployment time in seconds, t, the water column concentrations, [C]ext, and the effective coefficients of diffusion, Deff, were estimated using non-linear regression with (Eq. 1),
$$ [C]_{\text{diff}} = [C]_{\text{ext}} \left( {1 - e^{{{\frac{{AD_{\text{eff}} }}{V \times l}} \times t}} } \right) $$
(1)
where A is the surface area of the membrane, 21.5 ± 0.2 cm2, V is the diffusion cell volume, 321 ± 3 cm3 and finally l is the nominal membrane thickness, 0.0145 cm.

2.5 Ion-exchange Analytical Measurements

Free Cd2+, Cu2+ and Zn2+ were determined using an ion-exchange technique [see Fortin and Campbell (1998) for a detailed description]. Briefly, the method consisted of equilibrating a calibrated sulphonic acid-type resin with a solution of an unknown concentration of free divalent metal (M2+; Eq. 2).
$$ M^{2 + } + (3 - x) \cdot R_{x} \,{\text{Cat}}\overset {K_{IE}^{c} } \longleftrightarrow R_{2} M + (3 - x) \cdot {\text{Cat}}^{x + } $$
(2)
The conditional equilibrium constant \( K_{IE}^{C} \) corresponding to this reaction is (Eq. 3),
$$ K_{IE}^{c} = {\frac{{[R_{2} M] \cdot [{\text{Cat}}^{x + } ]^{3 - x} }}{{[M^{2 + } ] \cdot [R_{x} \,{\text{Cat}}]^{3 - x} }}} $$
(3)
where R = resin, RxCat = resin-binding sites occupied by a cation (Na+, K+, Mg2+ or Ca2+), x = charge of the cation (thus 1 or 2), and R2M = resin-binding sites occupied by metal M2+. When the test solution is swamped with “inert” electrolytes, the concentration of the electrolyte in the test solution ([Catx+]) and associated with the resin ([RxCat]) should not change appreciably with the addition of minute amounts of a trace metal. If [Catx+] and [RxCat] are assumed to remain constant, Eq. 3 can be rearranged to yield a distribution coefficient λo,i,pH (l g−1) valid at fixed ionic strength and pH:
$$ \lambda_{{o,i,{\text{pH}}}} = K_{IE}^{c} \cdot {\frac{{[R_{x} \,{\text{Cat}}]^{3 - x} }}{{[{\text{Cat}}^{x + } ]^{3 - x} }}} = {\frac{{[R_{2} M]}}{{[M^{2 + } ]}}} $$
(4)
The metal bound to the resin (RzM) can be measured experimentally by eluting the resin with a volume V (2.2 ml) of strong acid (1.5 M HNO3). According to the quantity of resin used (mr; ca. 7 mg) and the concentration of metal measured in the eluate, [R2M] can be calculated by Eq. 5:
$$ [R_{2} M] = {\frac{{[M_{\text{Eluate}} ] \cdot V}}{{m_{r} }}} $$
(5)
Combining Eqs. 4 and 5 gives:
$$ [M^{2 + } ] = {\frac{{[M_{\text{Eluate}} ] \cdot V}}{{\lambda_{{o,i,{\text{pH}}}} \cdot m_{r} }}} $$
(6)
Once a distribution coefficient specific to the metal of interest has been determined, for the relevant concentration and nature of the electrolyte and for the pH of solution, one can easily calculate [M2+].

In our original IET work, we used a solution of NaNO3 (200 meq l−1) to pre-equilibrate the ion-exchange resin (Fortin and Campbell 1998). In an effort to minimise perturbation of the natural water samples, we used a lower ionic strength matrix (~10 meq l−1). The composition of the electrolyte solution was as follows: 0.50 mM Ca, 0.20 mM Mg, 0.03 mM K and 7.87 mM Na (all added as nitrate salts). This modification in ionic strength, and the resultant decrease in competition for binding sites on the resin, led to an increase in the amount of trace metal binding to the resin before equilibrium was reached and to an increase in the sample volume needed to achieve equilibrium. Breakthrough curves showed that equilibrium between the sample solution and the ion-exchange resin column was reached after ~150 ml (compared to 20 ml in Fortin and Campbell (1998) for the miniaturised method). At this point, the metal speciation within the sample solution going through the resin column is no longer altered by the resin. As a precaution, 250 ml of sample solution was passed through the column to ensure that equilibrium had been reached. In this matrix, calcium was found to be the major counter-ion on the resin sites after equilibrium was achieved (i.e. Ca was the major cation present in the eluant). Calibration of the IET for Cu2+, Cd2+ and Zn2+ in the standardised matrix was performed at four pH values (4.5, 6.0, 7.0 and 8.0) in the concentration range of 0.4–4.8 nM Cd2+, 4–80 nM Cu2+ and 18–180 nM Zn2+ (n = 48; Appendix). No pH buffers were used to avoid both contamination and possible metal complexation by the buffers.

To adjust the ionic strength of the lake water samples to that of the standard IET matrix, we added a small volume of a concentrated stock solution of Ca(NO3)2 (250 mM), Mg(NO3)2 (200 mM), KNO3 (50 mM) and NaNO3 (2,000 mM) to each sample; dilution never exceeded 0.75% of the sample volume. The volume added to each sample was adjusted according to the concentrations of each cation already present in the lake water. Analytical blanks were run along with the natural samples to evaluate background contamination of samples from the nitric acid used for elution of the resin column and were subtracted from the sample values.

2.6 Metal Speciation Calculations

For each lake, results obtained with the IET were compared with mean computed values that were calculated using the Windermere Humic Aqueous Model [WHAM; model version 6.0; (Tipping 2002)] and the ECOSAT model [version 4.7 which incorporates the NICA-Donnan model; (Keizer and Van Riemsdijk 1999)]. The WHAM and ECOSAT models require the concentrations of humic and fulvic acids as input data. To apply the models to the content of our diffusion cells, we assumed that (i) the ratio of DOM:DOC is 2:1 (Buffle 1988), and (ii) on average, 60% of DOM is composed of humic and fulvic acids with a ratio of 1:3 (Perdue and Ritchie 2003). Other input parameters were pH, diffusible Cd, Cu and Zn, anions (CO32−, NO3, SO42−, PO43−, Cl) and major cations (Ca2+, Mg2+, Na+ and K+ as defined in the standard matrix) as well as other trace metals potentially in competition for DOM-binding sites (Mn2+, Al3+ and Fe3+; entered in WHAM as dissolved solution components). For both models, a few pertinent inorganic formation constants were updated using a reliable source of thermodynamic data (Martell et al. 2004) (e.g. the two most significant changes concern the copper and zinc bicarbonate complexes, with log K CuHCO3+ = 1014.62 and log K ZnHCO3+ = 1013.12 in the default WHAM data base, to 1012.13 and 1011.83, respectively). These changes had, however, little influence on the predicted metal complexation with an average increase in Cu and Zn complexation of 7 and 3%, respectively.

3 Results

3.1 Calibration of the Ion-Exchange Technique

Metal concentrations used for calibrating IET covered the ranges 0.4–4.8 nM Cd2+, 4–80 nM Cu2+ and 18–180 nM Zn2+ (Appendix). As anticipated, mean distribution coefficients (l g−1 ± SD) for Cd (2.6 ± 0.3) and Zn (2.1 ± 0.4) were considerably higher than those determined previously at higher ionic strength (Fortin and Campbell 1998). The coefficients for Cd and Zn also showed no significant trends over the entire pH (4.5–8.0) and metal concentration ranges (slope of linear regression not significantly different from zero; t-test, P > 0.05). In contrast, the Cu distribution coefficient increased significantly with pH; it was described by λo = (0.8 ± 0.1) · pH−(1.0 ± 0.7) (r² = 0.64; linear pH range = 4.5–7.2; slope of linear regression significantly different from zero; t-test, P < 0.001), and variability was more important at pH > 7.5.

3.2 Diffusion Rates into the Diffusion Cells

For both lakes Bousquet and Dufault, Cd, Cu and Zn diffused into the cells very rapidly over the first 3 days (e.g. after 3 days, total concentrations within the cells represented 36–86% of those reached after 21 days). After 7 days, influx was slow, and after 15 days, metal levels had reached 93–104% of those determined for the last sampling time (21 days; data not shown). For all cations including Cd, Cu and Zn, 90% of the water column concentrations, estimated using Eq. 1, were reached inside the cells within 15 days (Table 1). As expected, diffusion of dissolved organic matter was slower, the estimated time needed to reach 90% of the water column concentrations being between 19 and 24 days.
Table 1

Diffusion kinetics for major cations, trace metals and dissolved organic matter (UV absorbance) for the diffusion cells deployed in two lakes: Estimated water column concentrations, Cext, and effective diffusion coefficients, Deff. Based on the latter, the deployment times in days required to reach 90% of the estimated water column concentrations (T90) were also calculated

Parameter

Lake Bousquet

Lake Dufault

Cext

Deff (cm2 s−1)

T90 (days)

Cext

Deff (cm2 s−1)

T90 (days)

Na (μM)

60 ± 1

6.4 × 10−7

9

178 ± 2

1.3 × 10−6

4

K (μM)

10.2 ± 0.1

8.4 × 10−7

7

17.6 ± 0.1

1.7 × 10−6

3

Ca (μM)

112 ± 4

4.1 × 10−7

14

420 ± 7

8.8 × 10−7

7

Mg (μM)

47 ± 2

4.3 × 10−7

13

129 ± 3

7.8 × 10−7

7

Cd (nM)

0.85 ± 0.06

1.2 × 10−6

5

7.3 ± 0.2

7.0 × 10−7

8

Cu (nM)

45 ± 1

1.1 × 10−6

5

200 ± 10

3.9 × 10−7

15

Zn (nM)

79 ± 1

1.0 × 10−6

6

1,400 ± 40

6.7 × 10−7

9

Abs. 465 nm

15 ± 1

3.0 × 10−7

19

ND

ND

ND

Abs. 665 nm

2.1 ± 0.6

2.5 × 10−7

23

1.6 ± 0.3

2.4 × 10−7

24

ND indicates that values could not be determined as estimated parameters were not statistically significant (P > 0.05)

3.3 Spatial Distribution of Metals

Total dissolved metal concentrations varied appreciably among the sampled lakes: from 0.06 to 9 nM Cd; from 36 to 608 nM Cu; and from 7 to 3,700 nM Zn (Table 2). Lakes Dufault and Turcotte along with station 6 in Lake Dasserat had the highest metal concentrations ([Cd] > 2 nM; [Cu] > 100 nM; [Zn] > 300 nM), presumably reflecting nearby acid mine drainage. Lakes located upwind of the Rouyn-Noranda smelter (Opasatica, Dufay, Evain and Ollier; Fig. 1) had the lowest metal concentrations ([Cd] < 0.2 nM; [Cu] < 65 nM; [Zn] < 10 nM).
Table 2

Average (±SD) measured pH, dissolved concentrations of CaT, MgT, CdT, CuT, ZnT, Cd2+, Cu2+, Zn2+ and dissolved organic carbon in samples from all lakes studied (N = number of samples within a lake)

Lake (N)

pH

[Ca] (μM)

[Mg] (μM)

[Al] (μM)

[Fe] (μM)

[DOC] (mg l−1)

[Cd] (nM)

[Cd2+] (nM)

[Cu] (nM)

[Cu2+] (nM)

[Zn] (nM)

[Zn2+] (nM)

Vaudray (10)

7.24 ± 0.05

76 ± 8

37 ± 3

1.7 ± 0.2

0.33 ± 0.09

6.8 ± 0.5

0.57 ± 0.05

0.44 ± 0.05

46 ± 7

ND

28 ± 2

28 ± 4

Caron (10)

7.45 ± 0.07

275 ± 49

100 ± 6

1.7 ± 0.6

0.9 ± 0.6

10 ± 2

0.52 ± 0.07

0.43 ± 0.05

68 ± 6

ND

30 ± 7

24 ± 4a

Bousquet (15)

6.9 ± 0.2

107 ± 8

49 ± 3

3.1 ± 0.5

1.6 ± 0.3

11.3 ± 0.8

0.55 ± 0.06

0.38 ± 0.03

62 ± 12

4 ± 1

40 ± 10

31 ± 9

Stn BO-5 (2)

6.1 ± 0.7

78 ± 8

42.9 ± 0.1

20 ± 10

15 ± 7

20 ± 4

1.4 ± 0.2

0.9 ± 0.4

71 ± 12

2.1 ± 0.6

99 ± 35

68 ± 34

Turcotte (2)

5.4 ± 0.2

81 ± 5

29.1 ± 0.1

3.12 ± 0.04

0.20 ± 0.07

3.3 ± 0.5

9 ± 3

8 ± 3

193 ± 56

58 ± 12

2,800 ± 1,300

2,000 ± 860

Moore (2)

7.2 ± 0.2

232 ± 20

93.5 ± 0.7

0.4 ± 0.2

0.5 ± 0.3

8 ± 2

0.69 ± 0.04

0.46 ± 0.09

63 ± 29

ND

53 ± 20

42 ± 9

Joannès (6)

7.4 ± 0.2

174 ± 8

69 ± 3

1.3 ± 0.4

0.5 ± 0.1

8.4 ± 0.5

0.30 ± 0.06

0.22 ± 0.06

52 ± 7

ND

13 ± 3

10 ± 3

Héva (6)

6.89 ± 0.08

50 ± 3

32.7 ± 0.9

2.2 ± 0.7

0.76 ± 0.09

7.6 ± 0.9

0.34 ± 0.10

0.24 ± 0.04

43 ± 2

5 ± 3

13 ± 1

13 ± 3

Dufay (6)

6.9 ± 0.2

71 ± 4

51 ± 3

1.3 ± 0.4

0.48 ± 0.06

8.4 ± 0.3

0.19 ± 0.06

0.102 ± 0.006

36 ± 4

2 ± 1

9 ± 4

10 ± 6

Évain (4)

7.58 ± 0.03

165.9 ± 0.7

93 ± 4

0.5 ± 0.2

0.2 ± 0.1

6.4 ± 0.3

0.06 ± 0.01

0.03 ± 0.01a

54 ± 6

ND

8 ± 5

12 ± 5

Despériers (1)

5.65

31

13

2.0

0.20

2.9

1.93

1.86

46

12

168

149

Ollier (4)

7.6 ± 0.1

314 ± 18

108 ± 2

0.2 ± 0.1

0.07 ± 0.03

5.6 ± 0.3

0.07 ± 0.04

0.05 ± 0.04

54 ± 11

ND

7 ± 3

12 ± 3

Opasatica (8)

7.8 ± 0.2

203 ± 9

100 ± 2

0.3 ± 0.2

0.13 ± 0.04

6.4 ± 0.9

0.07 ± 0.04

0.05 ± 0.04

64 ± 16

ND

7 ± 3

12 ± 8

Dufault (4)

7.61 ± 0.07

460 ± 80

127 ± 6

0.4 ± 0.1

0.17 ± 0.07

4.8 ± 0.7a

2.7 ± 0.3

1.6 ± 0.3

127 ± 19

ND

430 ± 90

240 ± 40

Dasserat (3)

7.4 ± 0.3

195 ± 13

83 ± 3

0.9 ± 0.9

0.19 ± 0.06

6.3 ± 0.7

0.5 ± 0.3

0.29 ± 0.07

68 ± 14

ND

170 ± 80

130 ± 40

Stn DS-6 (1)

4.52

224

113

8.3

0.55

3.8

6.8

4.8

608

307

3,700

2,860

ND Not determined, pH outside of linear working range of IET Cu calibration

aOne data point rejected due to contamination

Several lakes were sampled at more than one location (Fig. 1) to evaluate the variability of both metal contamination and metal complexation within a lake. In most cases, standard deviations were reasonably low for all parameters measured (Table 2), reflecting a low spatial variability within each lake and good analytical precision. However, two sites located in lakes Bousquet (BO-5) and Dasserat (DS-6) exhibited greater spatial variability and are treated separately (Table 2; Figs. 2, 3, 4). Station BO-5 is located within a stream that flows into Lake Bousquet; the stream water composition differs markedly from that prevailing in the lake itself (mainly lower pH and higher DOC content). Station DS-6 is located at the point where the water from Baie Arnoux enters into Lake Dasserat; Baie Arnoux has a high metal content and very low pH due to acid mine drainage from nearby abandoned mine tailings. Metal concentrations at station DS-6 were approximately one order of magnitude greater than at the other sites in Lake Dasserat; these other sites reflected a declining concentration gradient as a function of their distance from the high metal concentrations in this bay.
https://static-content.springer.com/image/art%3A10.1007%2Fs10498-009-9074-3/MediaObjects/10498_2009_9074_Fig2_HTML.gif
Fig. 2

Comparison of mean (±SD) measured free metal ion concentrations as a function of mean (±SD) total dissolved metal concentration. a Cd; b Cu; c Zn. The 1:1 lines (full) and the linear regressions (dashed lines) are shown

https://static-content.springer.com/image/art%3A10.1007%2Fs10498-009-9074-3/MediaObjects/10498_2009_9074_Fig3_HTML.gif
Fig. 3

Comparison of the mean (±SD) free metal ion concentrations calculated with WHAM 6 with those measured with the ion-exchange technique. a Cd; b Cu; c Zn. The 1:1 line (full) and the linear regression (dashed lines) are shown

https://static-content.springer.com/image/art%3A10.1007%2Fs10498-009-9074-3/MediaObjects/10498_2009_9074_Fig4_HTML.gif
Fig. 4

Ratio of free metal concentrations measured by IET to those calculated with WHAM 6 as a function of pH. Error bars are means (±SD) calculated using propagation of error theory. a Cd; b Cu; c Zn

3.4 Free Metal Ion Concentrations

Concentrations of free metal ions measured with the IET are correlated with total metal concentrations (r2 of 0.975 and 0.995 for Cd and Zn, respectively on a linear scale), indicating that the proportions of free Cd2+, and to a lesser extent Zn2+, are reasonably constant across the entire range of pH, total metal and DOC concentrations (Fig. 2). The mean proportions (±SD) of total Cd and Zn in the form of Cd2+ and Zn2+ are 71 ± 16 and 84 ± 21%, respectively. This latter average proportion of free metal was calculated only for the lakes that had dissolved [Zn] > 20 nM; below this concentration, the measured free Zn2+ concentrations values were tainted with low but erratic background contamination.3 As expected, in contrast with Cd and Zn, the proportion of free Cu2+ was not constant (Fig. 2b), which reflects the influence of water quality parameters (such as pH and DOC) on the speciation of dissolved Cu.

Variability in free and total metal concentrations of replicate samples was relatively low, i.e. with coefficients of variation below 10% for Cu, 20% for Cd and 25% for Zn, except on one occasion, were contamination had obviously occurred (Zn2+ at station CA-1) in one of the sub-samples (Table 3).
Table 3

Measured (±SD) total and free metal concentrations at the three stations where three replicate diffusion cells were deployed

Station

[CdT]

[Cd2+]

[CuT]

[Cu2+]

[ZnT]

[Zn2+]

CA-1

0.58 ± 0.02

0.43 ± 0.02

63 ± 4

ND

34 ± 5

46 ± 33

VA-1

0.53 ± 0.01

0.40 ± 0.07

41 ± 3

ND

29 ± 2

26 ± 6

HE-1

0.25 ± 0.04

0.21 ± 0.02

42 ± 1

7.6 ± 0.7

11.8 ± 0.6

11 ± 1

Concentrations are expressed in nM

ND Not determined because pH was outside the linear working range of IET Cu calibration

3.5 Comparison Using the WHAM and ECOSAT Models

Mean values of [Cd2+] and [Zn2+] for each lake calculated with WHAM 6 were within a factor of two and seven, respectively, of the mean values measured by IET (Fig. 3a, c). On the other hand, the calculated and measured [Cu2+] mean values markedly diverged with differences of up to 40-fold (Fig. 3b). Differences between calculated and measured mean free metal ion concentrations became more marked as the metal concentration decreased and as the sample pH increased, especially for Cu. In order to illustrate this latter effect, the ratio of IET values to those calculated with WHAM 6.0 was plotted as a function of pH (Fig. 4). The ratio of measured free metal ion concentrations relative to the calculated values increased as pH increased. Note that a ratio of one indicates perfect agreement between measured and calculated values. Finally, [Zn2+] calculated with ECOSAT agreed just as well with those obtained with WHAM 6 (not shown). However, ECOSAT predicted much lower free Cu2+ or Cd2+ than WHAM, roughly one order of magnitude in difference; the agreement between the two models improved at higher metal concentrations. Since WHAM provided estimated free metal ion concentrations closer to our measured values, we focused on the former model.

4 Discussion

4.1 Free-metal Ion Measurement

4.1.1 Sampling

The diffusion kinetics parameters obtained from lakes Bousquet and Dufault indicate that equilibrium was approached well within the 15-day deployment period for Cu, Zn and Cd, whereas a somewhat longer deployment period would be required for dissolved organic carbon to reach full equilibrium. This possibly resulted in slightly lower DOC in the equilibrium cells than in the lake. However, the water column absorbance measurements were low and had high uncertainty values such that after 15 and 21 days equilibration, values were not significantly different (data not shown; t-test, P > 0.05) suggesting that 15 days were sufficient. Also, note that Beneš and Steinnes (1974) showed that longer sampling times resulted in increased colonisation of the outer surface of the diffusion cells by periphyton and other aquatic organisms, without yielding additional information on the chemical composition of the waters. Therefore, we conclude that a 2-week period was appropriate to obtain a representative water sample.

The calculated effective diffusion coefficients for Cd2+, Cu2+ and Zn2+ were ~7 times lower than their self-diffusion coefficients at infinite dilution (Li and Gregory 1974). The reduction in effective diffusion coefficient is likely related to a number of factors; the effectiveness of mixing, the non-ideal diffusion in the filtration membrane and also complexation of the metals by dissolved organic carbon. Deff values are similar for both lakes for Cd and Zn. However, Deff is much lower for Cu, and the estimated time to reach 90% of the water column concentrations is longer for Cu in Lake Dufault than in Lake Bousquet. Since Lake Dufault has a higher pH, 7.6 compared to 6.9 for Lake Bousquet, this slower diffusion for Cu may be associated with a higher degree of complexation.

Considering that ion concentrations can vary to some extent over a short lapse of time (e.g. due to a major rain event), the contents of our diffusion cells were considered representative of the ambient conditions and, over a 2-week period, attenuated any fluctuations in the water column of the sampled lakes. With the exception of station BO-5, all the samples were taken in lakes that have long residence times (well over the 15 days of equilibration). The variability observed at station BO-5 was thus slightly higher than at the other sites, where metal concentrations and other measured parameters were relatively stable.

4.1.2 Ion-exchange

In our previous work with the IET, the distribution coefficients for Cd and Zn proved to be sensitive to changes in pH (Fortin and Campbell 1998; Néron et al. 2006; Vigneault and Campbell 2005). In these earlier IET applications, we used sodium as the major counter-ion, with which protons could effectively compete for binding on the resin, resulting in a pH dependence of the distribution coefficients. When a stronger binding cation such as calcium is used as the counter-ion, the influence of pH is attenuated. In the present work, calcium was found to be the dominant counter-ion on the resin (through the elemental analysis of the acid eluate; data not shown), and the resulting Cd and Zn distribution coefficients were insensitive to pH changes.

In contrast, the Cu distribution coefficient did not remain constant over the pH scale of interest. This could be due to binding of CuOH+ to the resin as the proportion of CuOH+ increases with pH (Sweileh et al. 1987). However, at pH 7.0, the species CuOH+ represents only 17% of the total Cu in the absence of organic ligands compared to 75% for the free Cu2+ ions (the other species being CuCO30(aq) and CuNO3+, representing 6 and 2% of the total Cu, respectively). Intuitively, it might be assumed that the affinity of the monovalent CuOH+ complex for the resin would be less than that of the divalent free Cu2+ and that the contribution of CuOH+ to the overall resin copper pool would be negligible. In fact, the affinity of CuOH+ for the resin would have to be much greater than that of free Cu2+ to explain the observed increase in the Cu distribution coefficient at pH values as low as 6.0, where the calculated proportion of aqueous CuOH+ is only 2%. This reasoning raises the fundamental question as to why would ion-exchange resins be selective. The resin used here contains sulphonate functional groups, which do not form inner-sphere complexes. Since the strength of outer-sphere complexes with a given ligand is primarily governed by the charge of the cation and the ionic strength (Morel and Hering 1993), it follows that the distribution coefficient should be constant for all ions of equal charge. This is obviously an oversimplification since sulphonate ion-exchange resins are known empirically to exhibit some selectivity among equally charged ions (e.g. the relative selectivity of 50W-X8 resin for Ba2+ is more than three times higher than for Mg2+) and selectivity decreases with ionic radii (resin affinity sequence: Ba2+ > Sr2+ > Ca2+ > Mg2+; Bio-Rad 1988).

The selectivity of ion-exchangers is not well understood. Factors affecting ion-exchange include the non-uniformity of charge distribution within the resin beads, the degree of cross-linking, electrostatic attraction/repulsion and specific adsorption (Diamond and Whitney 1966; Rhee and Dzombak 1999). Nevertheless, early work on ion-exchangers suggests that selectivity is closely related to the hydration requirement of the ion (Diamond and Whitney 1966). Indeed, there is very little room for water within the resin beads, especially within the highly cross-linked matrices. This results in a “preference” for ions having less affinity for water. For example, the binding strength between water molecules and cations of group 2A of the periodic table decreases with ionic radius (i.e. Mg2+ > Ca2+ > Sr2+ > Ba2+). Barium should thus have a lower hydration requirement than magnesium, and as mentioned earlier, Ba2+ has a greater affinity for the resin functional groups than Mg2+. We thus speculate that the decrease in the number of water molecules present within the secondary solvent shell of the CuOH+ species results in a higher affinity for the sulphonic acid-type resin than is the case for the free Cu2+ ions, notwithstanding the decrease in ionic charge. This interpretation is consistent with previous observations suggesting interference by hydroxo-complexes with the IET (Sweileh et al. 1987). Should our hypothesised preferred binding of hydroxo-complexes be correct, the applicability of the IET to determine free metal ion concentrations in natural waters becomes more challenging under conditions where the presence of hydroxo-complexes is not negligible (e.g. copper) and would require calibration for multiple species of a given metal.

In addition to the pH dependence of λCu, we also experienced calibration difficulties for Cu at pH > 7.5 (Appendix). We suspect that the poor reproducibility in this pH range was due to the poor buffering capacity of our standard solutions (no pH buffers were used). For an initial pH of 8.0, the final pH usually decreased a few tenths of a pH unit, but the decrease was variable from one run to another. In this pH range (7.5–8.0), the concentration of Cu2+ is highly sensitive to complexation by carbonates and hydroxides. Actual values of [Cu2+] in our standard solutions in this pH range appear difficult to estimate reliably; accordingly, Cu determinations were rejected for samples with pH values greater than 7.5.

To adjust the lake water samples to the standard IET matrix, we added a small volume of a concentrated stock solution of Ca(NO3)2 (250 mM), Mg(NO3)2 (200 mM), KNO3 (50 mM) and NaNO3 (2,000 mM) to each sample; the most substantive changes were Na additions to a final concentration of 7.87 mM. Can we assume that these additions did not appreciably modify the speciation of Cd, Cu and Zn in our samples? The effects of adding cations could in principle be of a competitive nature (i.e. direct competition among cations—Na, K, Ca and Mg—for DOM-binding sites) or could arise from an overall decrease in effective binding affinity caused by the change in ionic strength (from a range of 0.1–1.3 meq l−1 in the sampled lakes, up to a final adjusted ionic strength of 10 meq l−1 in the standardised IET matrix). Our review of the influence of major cations on metal complexation by humic and fulvic acids suggests that competitive effects are normally prevalent only at high cation and metal concentrations. Some researchers have reported an important competitive effect between background cations and “trace” metals, but these studies have involved disproportionately high concentrations of metals [usually in the μM to mM range; (Brown et al. 1999; Kinniburgh et al. 1996; Oste et al. 2002)] or seawater conditions (Hamilton-Taylor et al. 2002). In the few studies identified where sub-micromolar metal concentrations were used, the researchers reported very little to no effect of Ca additions on Cu and Cd complexation when added Ca was less than 10 mM (Cao et al. 2006; Hering and Morel 1988; Rozan and Benoit 1999). We thus presume that the relatively small additions of Ca, Mg, Na and K used in the present study would not have affected metal complexation other than by increasing the ionic strength. Nevertheless, all WHAM and ECOSAT calculations were performed using the standardised IET matrix, so that ionic strength effects were taken into account in the speciation calculations.

4.2 Lake-to-Lake Variations in Metal Speciation

The linear relationships between measured free Cd2+ and Zn2+ concentrations and their total concentrations (Fig. 2) were somewhat unexpected. Since both metals are known to have a relatively low affinity for natural DOM (Mantoura et al. 1978), we anticipated that the uncomplexed form, M(H2O)x2+, would predominate in our lake water samples, as is the case. However, considering the wide distributions of pH (4.5–8.0) and DOM concentrations (2.9–23 mg l−1 dissolved organic carbon), we originally expected appreciable lake-to-lake variations in Cd and Zn speciation—such variations are not apparent. The relative constancy of Cd and Zn speciation probably reflects the inverse relationship observed between pH and DOC (for lakes in the pH range 6–8 and DOC > 4 mg l−1; r2 = 0.49; P < 0.0001, thus resulting in counteracting influences on metal speciation. Also, due to the weak affinity of these metals for DOM, variations in DOC or pH only affect a relatively small fraction of the dissolved metal and result in minor changes in free Cd2+ and Zn2+. In addition, lakes with low total dissolved metal concentrations ([Cd] < 1 nM; [Zn] < 50 nM) tended to show positive relationships between DOC and total dissolved [Cd] (r2 = 0.26; P < 0.0001) or [Zn] (r2 = 0.47; P < 0.0001), i.e. the higher the DOC concentration, the greater the dissolved Cd and Zn concentrations. As a result, the [DOC]:[M] ratio does not vary greatly between lakes, and accordingly the proportion of bound metal remains relatively constant.

In the IET working range for Cu (pH 4.5–7.5), measured free Cu2+ concentrations were consistently much lower than values for total diffusible copper (Fig. 2b), as expected on the basis of organic and inorganic complexation. Inorganic Cu complexes contribute significantly to total dissolved copper in the pH range studied, with both CuOH+ and CuCO3 (aq) species present in proportions >6% at pH > 6.5 and up to 60% at pH 7.5.

At low ZnT concentrations, free Zn2+ concentrations measured with the IET were found to be close to but still, occasionally, greater than the total concentrations (Fig. 2c). This contradictory situation presumably arises from low but highly variable contamination. Indeed, as mentioned earlier, the blank values for Zn2+ were substantial and highly variable (6.6 ± 3.1 nM Zn2+; N = 6). Such variability in laboratory contamination during the free Zn2+ determination would explain the occasional free Zn2+ values that exceeded ZnT.

For those lakes for which we have historical metal data (notably lakes Bousquet, Dufay, Héva, Joannès, Opasatica and Vaudray), the total dissolved metal concentrations measured in the present study are consistently lower than those reported earlier (Couillard et al. 1993; Croteau et al. 2002); this is true especially for dissolved Zn concentrations, which have decreased up to tenfold over the past decades. These decreases presumably are due to lower atmospheric emissions from the local smelter (Croteau et al. 2002; Perceval et al. 2006).

4.3 Model Predictions (WHAM; ECOSAT)

Both models predicted free Cd2+ and Zn2+ concentrations that were lower than our measurements, but within one order of magnitude (e.g. Fig. 3a, c); WHAM predictions for Cd2+ were somewhat closer to our measured values than were those of ECOSAT. However, the situation for Cu was different; free Cu2+ concentrations as measured by IET were markedly higher than both the WHAM (up to 40-fold; Figs. 3b, 4b) and the ECOSAT predictions (up to ~1,000-fold; data not shown).

For all three metals, the divergence between measured free metal ion concentrations and WHAM predicted values was worse at higher pH (Fig. 4); the calculated values at pH > 7 are consistently much lower than the measured values. This divergence could be related to the choice of proton dissociation constants within the WHAM model [WHAM 6.0 incorporates “model VI” (Tipping 1998)]. As the pH increases, competition between protons and trace metal decreases and more metal complexation should be observed; proton dissociation constants define the extent of this competition. As mentioned by Tipping (Tipping 1998), independent proton dissociation data sets are only rarely available for the metal dissociation data sets used to calibrate these models. Tipping demonstrated how modifying parameter values (metal-binding constants and proton dissociation constants) resulted in a better fit to experimental data found in the literature. Such changes in the WHAM-binding parameter values might help to bridge the gap between measured and modeled free metal ion concentrations values.

More recently, the importance of taking into account possible competition by Fe(III) and Al(III) when simulating trace metal binding to humics in natural waters has been demonstrated (Tipping et al. 2002). In the simulations reported here, we have included the measured dissolved Fe(III) and Al concentrations. Simulations performed without Al and/or Fe indicated much greater complexation of Cd and Zn; free Cd2+ and Zn2+ decreased by factors up to ~10 and 30, respectively, and the agreement with the measured free ion concentrations was markedly poorer than with the original simulations that included Fe and Al interactions. In contrast, estimates of free Cu2+ were virtually insensitive to the presence or absence of Al and Fe, reflecting the higher affinity of Cu for humics (i.e. Cu better able to compete than Cd or Zn). Manganese did not have an appreciable effect on the simulations of Cd, Cu and Zn complexation, presumably due to its low ambient concentrations (<1 μM Mn) and its low affinity for DOM.

Several recent studies have compared the predictions of the WHAM and/or NICA-Donnan models to field measurements. However, to our knowledge, these models have never previously been challenged with such a large set of data coming from natural surface waters (84 samples for Cd and Zn; 33 samples for Cu). As Unsworth et al. have pointed out (Unsworth et al. 2006), for metals such as Cu, for which the concentrations of free ions represent a very small fraction of the total dissolved metal, these models may overestimate complexation by several orders of magnitude. Indeed, other studies dealing with Cu, including the current one, have concluded that the free Cu2+ ion concentrations are greatly underestimated by these models (Christensen et al. 1999; Guthrie et al. 2005; Kalis et al. 2006; Nolan et al. 2003). For metals that have a weaker affinity for DOM, like Cd and Zn, the agreement with model predictions is better but not perfect (Cheng et al. 2005; Cheng and Allen 2006; Christensen and Christensen 1999; Ge et al. 2005a, b; Kalis et al. 2006; Meylan et al. 2004; Nolan et al. 2003; Unsworth et al. 2005; Zhang 2004). Unlike our results, some marked underestimations of metal complexation by the models have been reported, but in these cases, the origin of the organic matter (derived from anthropogenic activities or from an autochthonous origin) (Cao et al. 2006; Meylan et al. 2004; Xue and Sigg 1999; Xue et al. 2005) or even the technique used may explain such a trend (Guthrie et al. 2005).

5 Conclusions

The method presented here allowed us to determine free Cd2+ concentrations down to the pM level. Background Zn contamination in the low nM range limited the applicability of the method to lakes with ambient Zn levels above ~20 nM. Applications below this level would require access to a complete trace metal clean facility (or the development of in situ techniques). In the case of copper, application of the IET method to natural waters is limited to samples of pH < 7.5 due to poor reproducibility in the calibration at higher pHs and interference presumably caused by the binding of the CuOH+ species. Free Cd2+ and Zn2+ concentrations were found to be closely correlated to their respective total concentrations, indicating that changes in Cd and Zn speciation due to counteracting variations in DOC concentrations and pH in the lakes studied are not critical and may lie within the experimental error of measurement. On the other hand, the speciation of Cu proved to be very sensitive to increases in pH and DOC concentrations, as anticipated. The WHAM 6.0 model predictions of metal speciation in the presence of DOM compared reasonably well with our measured free metal ion concentrations for Cd and Zn but much less so for Cu. The ECOSAT model prediction of free Zn2+ concentrations were similar to those of the WHAM 6.0 model, but for Cd2+ and Cu2+, it predicted greater complexation than WHAM 6.0 and IET measurements. Overall, both models predicted more complexation than what we observed in our field data set. These results indicate the need for further validation of DOM complexation models. This can be achieved only by using analytical speciation techniques such as the IET. These measurements provide valuable information to develop and validate complex models that will be able to predict trace metal speciation more accurately.

Footnotes
3

Although we subtracted the mean free metal ion concentration measured using analytical blanks (Cd: 0.030 ± 0.017 nM; Cu: 0.14 ± 0.03 nM; Zn: 6.6 ± 3.1 nM; N = 6), the high variability in Zn contamination, presumably due to sample handling despite all precautions taken, strongly affected some of the low Zn samples.

 

Acknowledgments

The authors acknowledge the technical assistance provided by M.G. Bordeleau, F. Beauchamp, R. Néron, M. Guillot and M. Arsenault in the laboratory and R. Rodrigue, J. Orvoine, O. Perceval, I. Louis, A. Giguère, A. Dumoulin, and A. van den Abeele in the field. Comments provided by A. Tessier and K.K. Mueller on earlier versions of the MS were greatly appreciated. Financial support was provided by the Natural Sciences and Engineering Research Council of Canada (NSERC strategic research grant STP 0192937) and by the Fonds québécois de la recherche sur la nature et les technologies (FQRNT team grant). P.G.C. Campbell is supported by the Canada Research Chair Program.

Copyright information

© Springer Science+Business Media B.V. 2009