Mercury accumulation in the fish community of a sub-Arctic lake in relation to trophic position and carbon sources

Authors


M. Power, Department of Biology, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 (fax +519 746 0614; e-mail m3power@sciborg.uwaterloo.ca).

Summary

  • 1Stable isotope analysis has improved understanding of trophic relationships among biota. Coupled with contaminant analysis, stable isotope analysis has also been used for tracing the pattern and extent of biomagnification of contaminants in aquatic food webs.
  • 2Combined analysis of nitrogen (δ 15 N) and carbon (δ 13 C) isotopes from fish species in a sub-Arctic lake were related to tissue mercury (Hg) concentrations to assess whether carbon sources influenced Hg accumulation in fish, in addition to trophic position.
  • 3Statistical models were used to estimate Hg biomagnification and uptake, to elucidate Hg accumulation dynamics and to appraise the relative importance of Hg exposure routes for the fish species.
  • 4Species Hg contamination increased as a function of trophic position (δ 15 N) and was inversely related to the δ 13 C signature. Species connected to the benthic food chain had lower Hg concentrations than species connected to the pelagic food chain. Species undergoing ontogenetic dietary shifts with increasing size, e.g. lake trout Salvelinus namaycush , also showed increased Hg concentrations with increasing reliance on pelagic fish as prey.
  • 5The results indicate that both vertical (trophic) and horizontal (habitat) food web structure influence Hg concentrations in fish tissue.
  • 6The biomagnification and uptake models indicated that contamination at the base of the food chain in the lake exceeded estimates for more southerly environments, thereby demonstrating the importance of dietary and water column Hg exposure routes in the sub-Arctic for determining Hg concentrations in fish.
  • 7Overall, the data reported here demonstrate how a combination of ecological concepts (food webs), developing ecological methods (stable isotopes) and environmental geochemistry can combine profitably to indicate the risks of exposure to environmental contaminants. Additional studies of the dynamics of Hg accumulation in the food webs of sub-Arctic lakes are needed, particularly in the light of the estimated high biomagnification rates and the heavy reliance of Inuit communities on subsistence fish harvests.

Introduction

Recent studies have confirmed that elevated levels of mercury (Hg) occur in fish from lakes in northern Canada and elsewhere (Björklund, Borg & Johansson 1984; Håkanson, Andersson & Nilsson 1990; Bodaly et al. 1993). Some of the Hg found in aquatic systems in these regions originates from natural sources, but much of it is of anthropogenic origin, dispersed via atmospheric transport (Slemr & Langer 1992; Lockhart et al. 1995; Downs, Macleod & Lester 1998). The most important anthropogenic sources of Hg are the combustion of coal and municipal waste (Nriagu & Pacyna 1988; Jackson 1997). Once released, Hg is transported north, deposited on catchment soils, and moved via run-off to surface waters (Johannson et al. 1991; Mierle & Ingram 1991; Lucotte et al. 1995). Therefore, over-winter accumulations of Hg in the snow pack and flushing to aquatic systems during spring snow melt is a major pulse exposure route in the north (Scott 2001). In the eastern Canadian Arctic and northern Québec, sediment cores show that Hg from all sources has been increasing, particularly in recent years (Lockhart et al. 1995; Lucotte et al. 1995). Although inorganic Hg is released by natural weathering, most of the Hg accumulated in resident biota occurs as monomethyl mercury, MeHg (Huckabee, Elwood & Hildebrand 1979; Surma-Aho et al. 1986), which is biomagnified up food chains to produce concentrations in fish tissue exceeding food consumption guidelines in many jurisdictions (Shilts & Coker 1995; Braune et al. 1999).

Investigations of the potential for Hg biomagnification in lacustrine and marine food webs have yielded evidence for biomagnification from lower to higher trophic levels (Kidd et al. 1995; Jarman et al. 1996; Tremblay et al. 1996; Plourde, Lucotte & Pichet 1997; Atwell, Hobson & Welch 1998) and provided detailed descriptions of temporal changes in accumulation rates (Thompson, Hamer & Furness 1991; Thompson, Furness & Walsh 1992). However, the mechanisms regulating MeHg formation, its initial incorporation in pelagic and benthic food chains, and subsequent trophic transfer, remain controversial (Watras et al. 1998). Lake studies of variation in fish tissue Hg concentrations have implicated physical factors, such as catchment area, lake size and water chemistry (Evans 1986; Bodaly et al. 1993; Watras et al. 1998), and ecological factors, such as productivity, trophic position, growth rate, fish age and food web structure (Lindqvist et al. 1991; Wiener et al. 1990; Vander Zanden & Rasmussen 1996), as causes of the observed variation.

The development of stable isotope analysis has improved understanding of trophic relationships among biota (Peterson & Fry 1987). Stable isotope ratios of carbon (13C/12C) and nitrogen (15N/14N) have been used to provide information about feeding relationships in aquatic environments (Kling, Fry & O’Brien 1992; Vander Zanden et al. 1998; Bearhop et al. 1999). Unlike stomach content analyses, which provide only a limited snapshot of feeding habits and relationships at a particular point in time, stable isotope analysis provides a time-integrated view of diet and trophic position. More recently, stable nitrogen isotopes have been used to trace the pathways of contaminant biomagnification for top predators in marine and freshwater food webs (Spies et al. 1989; Broman et al. 1992; Kidd et al. 1995; Atwell, Hobson & Welch 1998; Bowles et al. 2001). These studies rely on the results of detailed laboratory experiments that have demonstrated consistent nitrogen enrichment as energy is transferred from prey to predator (DeNiro & Epstein 1981; Minagawa & Wada 1984). The consistency of nitrogen enrichment at each trophic transfer provides a convenient quantitative measure of relative trophic position within a food web (Cabana & Rasmussen 1994) that may be correlated with contaminant concentrations to estimate metal uptake (Sunda & Huntsman 1998) and biomagnification rates (Rolff et al. 1993). In contrast, carbon isotope ratios remain relatively unaffected by trophic transfer (DeNiro & Epstein 1981; Fry & Sherr 1984). Because pelagic and benthic algae in freshwater lakes exhibit distinctive carbon signatures as a result of differential fractionation during carbon fixation (Hecky & Hesslein 1995), the carbon signatures of consumers at multiple trophic levels may be used to differentiate between the relative reliance of organisms on food webs of pelagic and benthic origin (Hecky & Hesslein 1995). Accordingly, the combined analyses of carbon and nitrogen isotopes may prove important in further understanding the subtleties of contaminant impacts on fish communities as influenced by habitat use and/or food web dependencies.

Previous studies combining the use of stable isotope and contaminant analyses in temperate or sub-Arctic lacustrine environments have typically focused on a few dominant, or valued, species (Cabana & Rasmussen 1994; Kidd et al. 1995) and have used δ15N signature information alone. None has employed the cross-reference information available from the simultaneous analysis of multiple isotopes that eliminates many of the ambiguities associated with the use of a single isotope tracer (Peterson, Howarth & Garritt 1985). Here, carbon and nitrogen stable isotope analyses were used to examine the patterns of Hg biomagnification and bioconcentration in the fish community of a sub-Arctic lake and to test explicitly the hypothesis that Hg accumulation in fish is influenced by horizontal food web structure, e.g. by dietary shifts between prey items at a similar trophic level but living in different habitats (Lindqvist et al. 1991).

Methods and materials

SITE AND SAMPLING METHODS

Stewart Lake (58°11′N, 68°26′W) is a small (9·9 km2), shallow (≤ 12 m), oligotrophic lake located immediately north-west of Kuujjuaq, Nunavik, Canada, in the Ungava Bay drainage basin (Fig. 1). The lake is situated at the northern limit of the tree line within a climatic zone moderated by the influences of the Ungava Bay circulation and prevailing westerly winds. Precipitation averages 523·5 mm year−1 and is almost equally split between rain and snow. Mean daily temperatures range from the January low of −23·5 °C to the July high of 11 °C (Environment Canada 1993). The lake drains north, via a chain of similar lakes and the Nepihgee River, to Ungava Bay, and supports a simple biological community, including 10 species of fish within the lake watershed.

Figure 1.

Map of Stewart Lake showing depth contours (m) and position relative to Ungava Bay.

The fish community of Stewart Lake comprises eight species that are present in significant numbers throughout the year. Species include planktivorous cisco Coregonus artedii Lesueur, two omnivores (three-spine stickleback Gasterosteus aculeatus Linnaeus and northern lake chub Couesius plumbeus Agassiz), three benthivores (longnose sucker Catostomus catostomus Forster, slimy sculpin Cottus cognatus Richardson and round whitefish, Prosopium cylindraceum Pallas) and two piscivores (burbot Lota lota Linnaeus and lake trout Salvelinus namaycush Walbaum). Two additional species, lake whitefish Coregonus clupeaformis Mitchill and brook charr Salvelinus fontinalis Mitchill, are reported occasionally from local subsistence harvests (Makivik Research, Québec, Canada, unpublished data) and are considered to be rare in the lake itself. Lake whitefish were considered rare in initial surveys of lakes surrounding Kuujjuaq (Dunbar & Hildebrand 1952). Brook charr, however, are easily captured in the tributary streams and may use the lake in winter. Competitive exclusion by lake trout, predator avoidance or insufficient over-wintering habitat in streams may explain patterns of seasonal lake avoidance and/or use (Power 1980).

Fish used for this study were collected in the summer of 1999 at several sites using multifilament experimental gillnets (10–60 mm mesh) set at varying depths in the littoral, limnetic and profundal zones of the lake. To avoid minimum size-selective biases typically associated with netting alone, supplementary samples were obtained by setting trot lines in the profundal zone at maximal depths, deployment of standard minnow traps (5 mm mesh) in the littoral zone, and, electro-shocking of randomly selected littoral habitats. All captured fish were weighed (g), measured (cm), sexed and had stomach contents identified. Otoliths or opercular bones were removed for ageing, and dorsal muscle samples were obtained and preserved for stable isotope and Hg analyses. Ageing was completed by soaking otoliths and opercular bones in a 50% glycerol solution for 72 and 24 h, respectively, prior to reading. Soaking improved annuli clarity when viewed under reflected light at 25–40× magnification. As a result, there was almost 100% concordance between readings, with all discrepancies being resolved by a third reading. Water samples were obtained in acid-washed, rinsed 1-l bottles and analysed for nutrients at the Fisheries and Oceans Canada, Freshwater Institute Analytical Water Chemistry Laboratory, Winnipeg, Canada, following procedures described in Stainton, Capel & Armstrong (1977).

ISOTOPE METHODS

Stable isotope ratios are expressed as delta values (δ) and measured as parts per thousand differences (‰) between the isotope ratio of the sample and that of a defined international standard according to the formula:

δR = [(Rsample − Rstandard)/Rstandard] × 1000

where δR = the carbon (13C/12C) or nitrogen (15N/14N) isotope ratio of the sample or the standard. Samples depleted in the heavier isotope (13C or 15N) in comparison to the standard have lower delta values. Samples that are more enriched in the heavier isotope in comparison to the standard have higher delta values. All international standards are set at 0‰ by convention. Standards used to compute all values reported here included carbonate rock from the Pee Dee Belemnite formation (Craig 1957) and nitrogen gas in the atmosphere (Mariotti 1983).

Dorsal muscle tissue for stable isotope analysis was obtained posterior to the dorsal fin, above the lateral line, and dried at a constant temperature (60 °C) for 48 h. Dried samples were pulverized using a Retsch MM 2000 ball mill grinder (F. Kurt Retsch GmbH & Co., Haan, Germany) and the powder stored in glass desiccation vials until analysed. Approximately 1 mg of dried, ground muscle tissue was used in the simultaneous analysis of stable carbon and nitrogen isotopes. All analyses were performed on a Micromass VG Isochrom continuous-flow isotope ratio mass spectrometer connected to a Carlo Erba elemental analyser at the Environmental Isotope Laboratory, University of Waterloo, Ontario, Canada. The International Atomic Energy Agency CH6 and N1 standards, respectively, were used to determine the accuracy of δ13C and δ15N values measured as the mean difference ± one standard deviation of repeat measures of the standards [δ13C =−0·06 ± 0·11‰ (n = 19) and δ15N = −0·08 ± 0·14‰ (n = 17)]. Sample reproducibility was measured by repeat analysis of samples as the mean difference ± one standard deviation of the difference between duplicate analyses of randomly selected samples [δ13C = −0·020 ± 0·09‰ (n = 25) and δ15N = −0·01 ± 0·07‰ (n = 25)].

To facilitate comparisons between species and individuals with differing fat contents, δ13C values were normalized for lipid content using techniques developed by McConnaughey & McRoy (1979) and Kline, Wilson & Goering (1998). Normalization allows more accurate and valid δ13C comparisons. By removing the differential effects of lipid synthesis and storage (DeNiro & Epstein 1978), residual δ13C values directly reflect assimilated carbon rather than the combined effects of assimilation and synthesis.

To assess the relative significance of inputs from benthic and pelagic carbon to the top predator, lake trout, the isotopic mixing model of Kline, Wilson & Goering (1998) was applied. The model determines the percentage contribution of each source to the consumer signature, given two known carbon sources with significantly different isotope values:

image

where %A is the percentage contribution to the consumer signature from source A, δA and δB are the isotope signatures of known sources, and δSAMPLE is the consumer isotope signature. Longnose sucker are obligate benthivores, feeding exclusively on invertebrates taken from the substrate (Scott & Crossman 1973). Although cisco have been reported to consume aquatic insects, they are predominately planktivorous (Scott & Crossman 1973). Gut content analysis associated with this study confirmed the exclusive use of plankton as a food resource by cisco. On the combined basis of this evidence, planktivorous cisco and benthivorous longnose sucker were used, respectively, as representative proxies for basal pelagic and benthic food chain carbon signatures.

MERCURY METHODS

Dorsal muscle tissue for Hg analysis was obtained posterior to the dorsal fin opposite to the point from where the isotope sample was obtained. Tissue samples were rinsed in ultra-pure water and excess water was blotted away with clean-room grade tissue (Durx 670TM; Berkshire Corporation, Great Barrington, Massachusetts, USA). Samples were transferred to polyethylene bags and stored at −75 °C until analysed. Prior to digestion, samples were thawed for 1 h and 2 g (±0·1 g) were measured into a 60-ml Teflon® digestion vessel with 5 ml of 70% nitric acid (trace metal grade; J. T. Baker, Phillipsburg, New Jersey, USA) and the cap fastened. Vessels were transferred to a stainless steel digestion oven (125 ± 4 °C) for overnight digestion. After digestion, vessel contents were removed and placed in a graduated polypropylene test tube. Vessels were rinsed with ultra-pure rinse water and the rinsing solution was used to bring each sample to a constant volume. Quality control tests of the digestion procedure using spiked controls indicated Hg recovery was consistently greater than 98%. Cold vapour atomic absorption spectrometry (Zeeman model 4110ZL Albertville, Minnesota, USA; equipped with an autosampler) was used to determine total Hg concentrations in digested samples.

Certified reference materials from the National Research Council of Canada, Canada (dogfish muscle, DORM-2, and dogfish liver, DOLT-2) were digested and analysed with each batch of samples. Analyte recovery to within 10% of the certified value was used as the batch validation criterion. Duplicate analytical blanks were included in each sample batch to monitor contamination during digestion and sample preparation. Blank signals exceeding the absorbance of the analyte detection limit invalidated the batch. The mean of two measurements performed for each sample were accepted only when the relative standard deviation was less than 10%. For additional methodological and quality control detail see Hendzel & Jamieson (1976) and (Kwan 1999).

In this study, all results are reported as total Hg per gram of wet weight. Organic and inorganic species of Hg, however, have different dynamics within aquatic food webs, with the bioaccumulation of MeHg of greatest concern. The main discrimination between MeHg and inorganic Hg in aquatic systems occurs during trophic transfer between biota at lower trophic levels, with the result that the proportion of MeHg in total Hg increases from 24% in phytoplankton to 96% in fish (Becker & Bigham 1995). Similar results have been reported by Watras & Bloom (1992), Plourde, Lucotte & Pichet (1997), Watras et al. (1998) and Bowles et al. (2001). The proportion of MeHg in total Hg varies between studies but is usually within the range of 80–99% for muscle tissue (Downs, Macleod & Lester 1998), suggesting that measures of total Hg provide adequate surrogates for bioaccumulative MeHg where studies of fish communities are concerned.

Hg, biological (e.g. length, weight) and isotope data were used to develop single and multiple variable regression models explaining observed variation in Hg concentrations among conspecifics and between species. To estimate biomagnification potential, the theoretical model developed by Broman et al. (1992) and Rolff et al. (1993) was applied:

image

where Cbiota is the Hg concentration in the biota, A and B are parameters estimated by linear regression after logarithmic transformation and δ15N is the corresponding biota nitrogen isotope signature. The model constant, A, is a scaling factor that will depend on the concentration of Hg at the base of the food chain (Rolff et al. 1993). The parameter B estimates the biomagnification potential of Hg (Broman et al. 1992). If B > 0, transfer of Hg between trophic levels is more efficient than biomass transfer and Hg is biomagnified.

To estimate differences in species uptake rates as a function of trophic position the model of Sunda & Huntsman (1998) was applied:

Cbiota= Vnet

where Cbiota is the Hg concentration in the biota, Vnet is the net Hg uptake rate from all sources and µ is the growth rate. Growth rate was estimated as the instantaneous growth rate from available mean size at age data following standard practice (Wootton 1990). Vnet was estimated using the above relationship, and net differences in Cbiota for the mean ages used to compute µ. Although developed for phytoplankton, the model is a general paradigm for trace metal accumulation in aquatic systems that holds for phytoplankton, zooplankton, benthic invertebrates and fish despite substantial differences in uptake pathways and growth rates (Watras et al. 1998).

STATISTICAL METHODS

Data normality and variance homogeneity were assessed prior to statistical analysis. Hg data were log-transformed to stabilize the variance (Draper & Smith 1981). Linear regression was used to examine relationships between fish characteristics (e.g. length and age), Hg and isotope data. Statistical significance was judged using coefficient t-tests and the F-statistic calculated for the regression as a whole. All regression residuals were examined to ensure statistical adequacy following procedures outlined in Draper & Smith (1981). Where multiple regression models were used, analysis of variance (F-test) was used to establish the significance of additional model variables (Koutsoyiannis 1977) and standardized regression coefficients (Cox 1987) were used to infer the relative importance of each explanatory variable for explaining variations in the dependent variable. Maximal type I error in all statistical test procedures was set at α= 0·05.

Results

Water analysis (Table 1) indicated low levels of dissolved minerals and nutrients in the lake. The waters were slightly acidic, with low calcium and elevated levels of humic materials typical of oligotrophic, low productivity northern lakes. There was little or no difference in the analytical results from duplicate samples obtained from the north and south ends of the lake.

Table 1.  Summary of Stewart Lake water chemistry data. Values based on averages of duplicate samples collected mid-lake at the north and south ends of the lake
ParameterValue
pH  6·7
Conductivity (µS cm−1) 59·5
Alkalinity (µeq l−1)270
Total dissolved N (µg l−1)115
Total dissolved P (µg l−1)  1
Total suspended solids (mg l −1)  1
Dissolved inorganic C (µg l−1)270
Dissolved organic C (µg l−1)250
Na (mg l−1)  1·79
K (mg l−1)  0·71
Mg (mg l−1)  1·28
Ca (mg l−1)  3·46

Longnose sucker and cisco, respectively, were the most δ13C enriched and depleted fish species (Table 2 and Fig. 2). Lake trout and longnose sucker, respectively, were the most δ15N enriched and depleted. There was clear separation of planktivorous- and profundal-feeding coregonid species from benthic-feeding longnose sucker and a wide variation in δ13C signatures among forage fish occupying a similar trophic level. The top predator in the lake, lake trout, had a much narrower range of δ13C values, mid-range between the extremes of forage fish. Except for burbot, the variability in δ15N signatures for all species was small.

Table 2.  Summary of fish species captured in Stewart Lake. Numbers captured ( N ) and subsampled for isotope and Hg analysis ( n ) are given in column 1. Other columns give the mean ± 1 standard deviation of the stated parameter
SpeciesN / nLength (cm)Age (year)Hg (µg g−1)δ13C (‰)δ15N (‰)
Burbot11/1019·8 ± 17·6 3·6 ± 2·30·08 ± 0·08−23·1 ± 1·4 8·5 ± 2·0
Cisco12/1217·3 ± 3·0 3·8 ± 0·90·24 ± 0·09−27·9 ± 0·4 8·8 ± 0·2
Lake trout24/1739·4 ± 9·910·1 ± 4·20·59 ± 0·41−24·3 ± 1·111·3 ± 0·4
Longnose sucker52/1820·7 ± 7·9 7·7 ± 5·70·07 ± 0·03−22·0 ± 1·6 7·1 ± 0·5
Northern lake chub33/1410·3 ± 2·0 4·9 ± 1·80·13 ± 0·05−24·3 ± 1·6 8·0 ± 0·6
Round whitefish54/1525·2 ± 3·0 8·9 ± 3·20·09 ± 0·03−24·0 ± 0·7 9·4 ± 0·7
Slimy sculpin14/13 5·3 ± 0·9 3·6 ± 1·00·07 ± 0·04−23·3 ± 1·6 7·9 ± 0·6
Three-spine stickleback44/20 5·0 ± 0·5 2·9 ± 0·30·11 ± 0·04−26·4 ± 1·4 8·6 ± 0·5
Figure 2.

Mean δ 15 N and δ 13 C signatures ± one standard deviation of fish from multiple length classes captured in Stewart Lake in August, 1999.

The significance of the relationships between relative trophic position (δ15N) or carbon sourcing (δ13C) and Hg accumulations in fish on a species by species basis was tested using linear regression (Table 3). Changes in trophic position were significantly, and positively, related to changes in Hg concentrations in slimy sculpin, northern lake chub and burbot. Changes in Hg concentrations were significantly, and negatively, related to δ13C signatures for longnose sucker and lake trout, and significantly, and positively, related to δ13C signatures for three-spine sticklebacks.

Table 3.  Coefficient of determination ( r2 ) values measuring the proportion of variation in Hg concentrations explained by changes in length, weight, age and isotope signatures of fish species sampled in Stewart Lake. All regressions were significant at the 0·05 level except where r2 -values are underlined. Significant isotope regressions are marked with a single asterisk ( * ) when isotope values are positively related to Hg, and a double asterisk ( ** )when isotope values are negatively related to Hg
SpeciesLength (cm)Weight (g)Age (year)δ15N (‰)δ13C (‰)
Burbot0·8080·5610·7050·726*0·001
Cisco0·7330·7620·6050·0130·273
Lake trout0·7360·7220·7920·1910·571**
Longnose sucker0·7230·7580·8650·1350·581**
Northern lake chub0·6730·6270·7850·548*0·287
Round whitefish0·5530·5150·5530·1330·025
Slimy sculpin0·4840·4680·3150·464*0·269
Three-spine stickleback0·0880·5230·0010·0080·231*

Mixing model analysis indicated that benthic algal and phytoplankton carbon contributed in varying degrees to measured lake trout δ13C signatures as fish increased in fork-length. In the 20–40 cm fork-length range, pelagic source carbon obtained from three-spine stickleback predation contributed 33·7% to the mean δ13C lake trout signature, and benthic carbon from chironomid larvae and pupae consumption contributed the remaining 66·3%. Above 40 cm fork-length, lake trout increasingly relied on carbon obtained from pelagic sources (55%) as a result of predation on planktivorous fishes (e.g. cisco).

Regressions of Hg concentrations against length (cm), weight (g) and age yielded significant and statistically adequate models explaining > 50% of Hg variation for all but two species (Table 3), three-spine sticklebacks and slimy sculpin. In the case of three-spine sticklebacks, weight provided the only adequate explanation of Hg variability. For slimy sculpin statistically adequate models were obtained for length, weight and age. However, none of the estimated models explained more than 48% of the Hg variation. For burbot, cisco, round whitefish and northern lake chub, single factor length, weight or age regression models best explained Hg variations. Multiple regression models better explained variations in Hg concentrations for lake trout (weight, age, δ13C, r2 = 0·934), slimy sculpin (length, age, δ15N, r2 = 0·805) and longnose sucker (weight, age, r2 = 0·899). Age was the only factor common to all three models and, on the basis of standardized regression coefficients, the most important variable for explaining variations in Hg concentrations in two of the three models (lake trout and longnose sucker). Weight occurred in two models and was always the second most important variable for explaining Hg variations. Length was used in a single model and was most important for explaining Hg differences among slimy sculpin.

Considered as a community, Hg levels in fish varied among conspecifics and species (Fig. 3). There was a general tendency for Hg to increase as a function of individual trophic position, indicative of a biomagnification within the community.

Figure 3.

Estimated linear relationship between relative trophic position (δ 15 N) of fish in Stewart Lake and the logarithm of Hg concentrations measured in dorsal muscle tissue (µg g wet weight −1 ). Associated 95% confidence intervals plotted as solid lines. Data for individual fish plotted by species. Individual values for fish are plotted as follows: burbot (open squares), cisco (inverted triangles) lake trout (solid squares), longnose sucker (solid circles), prosopium (open triangles), northern lake chub (solid triangles), sculpin (open circles) and three-spine stickleback (crosses).

log Hg (µg g wet weight−1) = −2·615 + 0·192 δ15N

r2  = 0·560, all P < 0·001, n = 117

Mean Hg concentrations also varied systematically with carbon sources (Fig. 4) as reflected by δ13C (lipid normalized) signatures for lake trout and other species (forage fish), indicating an increase in Hg as the percentage importance of pelagic carbon in the diet increased.

Figure 4.

Estimated linear relationships between mean δ 13 C of forage fish and individual δ 13 C of lake trout in Stewart Lake, and the corresponding logarithm of Hg concentrations measured in dorsal muscle tissue (µg g wet weight −1 ). Mean lake trout δ 13 C plotted among individual lake trout δ 13 C values as a solid circle. The anomalous lake trout outlier represents a juvenile lake trout whose stomach contents indicated exclusive benthic invertebrate feeding. Data used to estimate each regression plotted as symbols. The r2 for forage fish and lake trout regressions were 0·928 and 0·594, respectively.

Lake trout:

log Hg (µg g wet weight −1) = −4·294 − 0·158 δ13C

r2  = 0·594, all P < 0·001, n = 16

All other species (forage fish):

log Hg (µg/g wet weight) = −3·240 − 0·090 δ13C

r2  = 0·928, all P < 0·001, n = 7

Correlation of species Hg uptake rates with trophic position (Fig. 5) yielded significant results as follows:

Figure 5.

Relationship between trophic position (δ 15 N) and Hg uptake rate based on means for Stewart Lake fish species. Estimated model plotted as a solid line and associated 95% confidence intervals as dashed lines. Data used to estimate the model are plotted as solid circles.

log (mean species uptake) = −4·498 + 0·503 (mean species δ15N)

r2  = 0·922, all P < 0·001, n = 8

with lake trout, at the highest trophic position, having the highest Hg uptake rate and longnose sucker, at the lowest trophic position, having the lowest uptake rate.

When combined in a multiple regression model, δ15N, δ13C and weight explained a high proportion of Hg variation among conspecifics and species as follows:

log Hg (µg g wet weight −1) = −3·536 + 2·658 × 10−4 weight (g) + 0·115 δ15N − 0·062 δ13C

r2  = 0·695, all P < 0·001, N = 117

Standardized regression coefficients indicated weight, as a proxy for uptake (Pearson correlation coefficient for mean weight and Hg uptake = 0·935), was the most important explanation of Hg variation among individuals in the Stewart Lake fish community. Measures of vertical (δ15N) and horizontal (δ13C) food web structure, respectively, were the second and third most important explanatory variables.

Discussion

There is no direct input of Hg into Stewart Lake, and Hg reaches the lake via long-distance atmospheric transport (Jackson 1997). Many remote lakes receive most of their Hg input from the atmosphere and there is evidence that the atmospheric transport of Hg has caused appreciable increases in Hg concentrations of lacustrine biota and sediments. Lucotte et al. (1995) demonstrated on average 2·3-fold increases in the Hg concentrations from anthropogenic sources in boreal forest lake sediments south (49–56°N) of Stewart Lake. With the exception of three-spine sticklebacks, measured Hg concentrations for Stewart Lake fish are comparable to, or higher than, values reported for other northern Québec lakes (Langlois & Langis 1995; Verdon & Tremblay 1999). For example, Hg concentrations in Stewart Lake longnose sucker measuring 40 cm varied between 0·137 and 0·141 mg kg−1, within the range of mean values (0·12–0·19 mg kg−1) reported for Québec lakes in the region 49–56°N (Schetagne & Verdon 1999). Lake trout (60 cm) from the same region, however, had Hg concentrations (0·18–1·02 mg kg−1) exceeded by similarly sized Stewart Lake fish (1·1–1·7 mg kg−1). Similarly, mean Hg concentrations (0·07 mg kg−1) for 5·3-cm slimy sculpins in Stewart Lake exceeded measurements for larger (7·6 cm) sculpins in Lac Detcheverry (45°N, 77°W). However, 5·0-cm sized three-spine sticklebacks from Lake Stewart had lower Hg concentrations (0·11 mg kg−1) than 4·0 cm individuals from Lac Detcheverry (Verdon & Tremblay 1999). Previous studies have provided no evidence of a latitudinal gradient in Hg concentrations in northern Québec, and results from this study would support that conclusion (Lucotte et al. 1995).

Differences in lake physical and chemical factors, however, complicate comparisons between lakes (Watras et al. 1998) and are often invoked as explanations of among-lake Hg variability for common taxa. Hg accumulation rates in lake biota depend on a variety of environmental factors that combine to determine measured Hg concentrations at any point in time. These factors include temperature (Lindberg et al. 1991; Bodaly et al. 1993), dissolved oxygen (Henry et al. 1995), dissolved organic carbon (Håkanson, Nilsson & Andersson 1988; Richardson, Egyed & Currie 1995) and pH (Wiener et al. 1990; Watras et al. 1998; Rose et al. 1999). Comparative studies of similarly sized fish within a lake have also shown that Hg variability may be related to biological factors, including differences in physiology, growth rate or feeding habits (Schetagne & Verdon 1999), length (Grieb et al. 1990), weight (Berninger & Pennanen 1995), age (Kim 1995) and trophic position of the fish (Cabana & Rasmussen 1994).

This study focused on differences in biological factors within a lake and revealed that trophic position influenced Hg concentrations through differences in uptake rates. A significant linear relationship between δ15N values and the logarithm of total Hg concentrations denoted biomagnification within the fish community (Broman et al. 1992; Kidd et al. 1995; Atwell, Hobson & Welch 1998). The estimated biomagnification rate obtained from regression analysis may differ from direct field-based estimates because the slope of the biomagnification regression model estimates relative increases in Hg per unit of δ15N based on contamination at the base of the food chain, rather than in the water column (Broman et al. 1992). Hg water concentrations are variable, and depend on numerous biogeochemical factors (e.g. dissolved organic carbon and pH) requiring specific measurements that make estimation of valid biomagnification rates difficult. However, the biomagnification coefficient can easily be estimated from isotope data that integrate both ecological (feeding) and biota-specific (e.g. physiological) factors that facilitate direct comparison between study sites.

The basal food chain contamination level estimated from the biomagnification model for Stewart Lake was intermediate between values estimated by Kidd et al. (1995) and Atwell, Hobson & Welch (1998), respectively, for boreal forest lake and Arctic marine food chains. Basal food chain contamination in Stewart Lake (0·0024 µg g−1) was 4·8-fold greater than the value reported by Atwell, Hobson & Welch (1998). Basal food chain contamination (0·006–0·575 µg g−1) estimated for lakes in north-western Ontario (Kidd et al. 1995), however, exceeded the estimate for Stewart Lake. Differences in local geochemistry and/or atmospheric deposition rates will partially account for the differences, with differences in temperature-driven methylation rates probably accounting for much of the remaining differences (Bodaly et al. 1993; Kelly et al. 1997).

The estimated biomagnification coefficient in Stewart Lake (0·19) was greater, and statistically different (P < 0·05), from the 0·09–0·16 values reported for fish lakes in north-western Ontario (Kidd et al. 1995), and equivalent to (P > 0·05) the 0·20 value reported for Arctic marine food webs (Atwell, Hobson & Welch 1998). In trophic studies, Hg concentrations typically increase by a factor of three between trophic levels (Meili 1991; Cabana & Rasmussen 1994). Comparisons between predatory and forage fish in Stewart Lake (Table 1), however, indicate that Hg concentrations increase by a factor of 5·4 (weighted average) between trophic levels, and that the Stewart Lake biomagnification rate exceeds rates in more southerly ecosystems (Kidd et al. 1995). Although longer food chains have been correlated with increased contaminant concentrations (Rasmussen et al. 1990; Vander Zanden & Rasmussen 1996), differences in food chain lengths between boreal forest sites and Stewart Lake cannot explain the elevated biomagnification rate in the simpler sub-Arctic system. Differences, however, may reflect other ecological and physical factors, such as differential feeding ecology and differences in individual metabolic rates leading to slower growth in the colder climate (Schindler et al. 1995).

Mean species Hg uptake rates describe net Hg uptake from all sources, whereas biomagnification considers only Hg uptake from prey ingestion. Differences in the correlation between mean species Hg uptake rates and trophic position (r2 = 0·922) and mean species Hg tissue concentrations and trophic position (r2 = 0·676) indicate that, while dietary exposure is the dominant Hg accumulation pathway (Klaverkamp et al. 1983; Harris & Bodaly 1998), uptake of water-borne Hg is also an important route of exposure. The importance of uptake from the water column and increases in the relative efficiency of MeHg uptake as water hardness declines (Rodgers & Beamish 1983) are consistent with differences in dietary-based biomagnification estimates from Stewart Lake and boreal forest system lakes. Differential Hg uptake from the water column associated with differences in water hardness would lead to site-specific differences in Hg body burdens for taxa at lower trophic levels, and cause differential biomagnification effects as measured in predators at higher trophic levels.

The primary source of Hg in fish is their food (Klaverkamp et al. 1983; Kelly et al. 1997). Diet (Mathers & Johansen 1985; Lindqvist et al. 1991) is therefore critical in determining observed Hg levels. Studies of natural lakes in northern Québec have demonstrated that variations in measured Hg concentrations depended on both Hg bioavailability at lower trophic levels and individual feeding behaviour (Tremblay 1999). Bioavailability will depend on environmental factors that vary among lakes and modify microbial activity and methylation rates, e.g. temperature, humic substances and the organic content of sediments. Within lakes, bioavailability will be constant in the absence of marked spatial heterogeneity or point source emissions. Although feeding behaviours may vary between species within a lake, regional among-lake species differences are minimal, even in perturbed environments (Verdon & Tremblay 1999). Accordingly, Lindqvist et al. (1991) suggested that horizontal food web structure within a lake would be as important for understanding the dynamics of species Hg accumulation as differences in biogeochemistry between lakes. For example, fish feeding on benthos ingest organisms at the base of the food chain (e.g. dipterans and trichopterans) typically lower in Hg, and will contain less Hg than fish feeding on zooplankton from the pelagic food chain (Tremblay 1999). The resulting pattern of dietary differences between members of the fish community will yield distinctive differences in the relative trophic position (δ15N) and carbon source (δ13C) signatures of fish. As was found in this study, isotope measures should then correlate with measured Hg levels. In addition, the δ15N and δ13C signatures of species undertaking ontogenetic dietary shifts may be expected to correlate with measured Hg levels as a result of changes in diet reflective of occupancy of a specific dietary niche or habitat type.

The importance of both trophic position and horizontal food web structure as explanations of Hg variability suggested models including δ15N and δ13C data, in addition to biological data reflecting net differences in uptake dynamics (e.g. weight), would more accurately describe Hg variation within a lacustrine fish community. Stewart Lake multiple regression results demonstrate that horizontal and vertical food web structure information improve the explanation of Hg variability in a lacustrine fish community. Analysis of variance, conducted as F-tests of the improved fit from the inclusion of δ15N and δ13C information, indicated significantly improved model fits (F-statistic P-value < 0·05). In addition, when controlling for variations in weight with the use of partial correlation coefficients, δ15N and δ13C correlated equally well with Hg (0·368 and −0·376, respectively).

As results here show, additional studies of within-lake dynamics of Hg accumulation by species that employ multiple stable isotope methods will aid in improving scientific understanding of Hg in lacustrine food webs. Furthermore, the heavy reliance of local Inuit communities on subsistence harvests drawn from sub-Arctic lakes that may be similarly characterized by rapid biomagnification, suggests immediate human health concerns and an imperative need for supplementary studies of Hg biomagnification in Ungava region lakes.

Acknowledgements

The field assistance of G. Power, B. Doidge, P. May and A. Saunders aided in obtaining required samples. Important support for the study was provided by Allan Gordon as part of the Nepihgee River fishway construction project. Funds to support the work came from the research budget of Makivik Corporation and the NSERC operating grant of M. Power.

Ancillary