Trophic Positions and Mercury Bioaccumulation in ... - Springer Link

ECOSYSTEMS

Ecosystems (2003) 6: 289 –299 DOI: 10.1007/s10021-002-0205-6

© 2003 Springer-Verlag

Trophic Positions and Mercury Bioaccumulation in Rainbow Smelt (Osmerus mordax) and Native Forage Fishes in Northwestern Ontario Lakes H. K. Swanson,1* T. A. Johnston,1 W. C. Leggett,1 R. A. Bodaly,2 R. R. Doucett,3 and R. A. Cunjak3 1

Department of Biology, Queen’s University, Kingston, Ontario K7L 3N6, Canada; 2Department of Fisheries and Oceans, Freshwater Institute, Winnipeg, Manitoba R3T 2N6, Canada; 3Department of Biology, University of New Brunswick, Bag Service 4511 Fredericton, New Brunswick E3B 6E1, Canada

ABSTRACT Hg concentrations to smelt invasion depends on both the species and size composition of their preversus post-invasion diet. At a standardized body mass of 10 g, rainbow smelt were significantly trophically elevated relative to most native forage species, but they did not have significantly higher muscle Hg concentrations. Relationships between Hg concentration and ␦15N were weak, both within and among forage fish species. This study shows that trophic elevation on a fine scale (within the forage fish community) may not result in increased contaminant bioaccumulation. It further challenges the general assumptions of food web theory and contaminant bioaccumulation.

Rainbow smelt (Osmerus mordax) is a recent invader to the lakes of the Hudson Bay drainage in northwestern Ontario, Canada. In some systems, the invasion has been linked to an increase in mercury (Hg) concentration in native predatory fish. This increase may be due to the fact that rainbow smelt are trophically elevated and thus accumulate more Hg than native forage fish species. To test this hypothesis, we compared the trophic positions and Hg concentrations of rainbow smelt and native forage fish in a series of smelt-invaded and reference lakes in northwestern Ontario. A comparison of forage fish ␦ 15N (an index of trophic position) between the smelt-invaded and reference lakes indicated that rainbow smelt moved into a trophic niche that was unoccupied prior to their arrival. Relationships between ␦15N and body size and between Hg concentration and body size differed among the forage species. This indicates that the response of predator

Key words: rainbow smelt (Osmerus mordax); biological invasions; stable isotope analysis; trophic positions; mercury concentrations; food webs; Ontario, Canada.

INTRODUCTION Rainbow smelt (Osmerus mordax) is an anadromous fish species that is native to coastal regions of North America and isolated lakes of the St. Lawrence River drainage (Scott and Crossman 1973; Evans and Loftus 1987). The species range has recently expanded, however, due to a series of both inten-

Received 30 January 2002; accepted 24 August 2002; published online April 3, 2003. *Corresponding author; e-mail: [email protected]

289

290

H. K. Swanson and others

tional and accidental introductions. Rainbow smelt were intentionally introduced into Crystal Lake, Michigan, in 1912 and subsequently spread throughout the Great Lakes during the 1920s (Evans and Loftus 1987). More recently, a series of accidental introductions allowed the species to invade the Rainy, English, Winnipeg, and Nelson river systems of northwestern Ontario and northcentral Manitoba (Evans and Loftus 1987; Campbell and others 1991; Franzin and others 1994; Remnant and others 1997). At least 79 lakes in northwestern Ontario have been successfully invaded (T. Marshall personal communication). In 1999, smelt were captured for the first time in the Nelson River estuary on Hudson’s Bay, thus completing the extension of their range from the Atlantic to Arctic oceans (R. A. Remnant personal communication). The ecological effects of smelt invasions are difficult to predict and variable among lakes. They have the potential to interact with a wide variety of prey and predator species because their life history is eurythermal and because they are opportunistic feeders (Scott and Crossman 1973; Evans and Loftus 1987; Franzin and others 1994; Hrabik and others 1998). Adult smelt have been implicated in recruitment failures of lake whitefish (Coregonus clupeaformis) and cisco (Coregonus artedi) (see Evans and Loftus 1987). In addition, lake whitefish, cisco, and yellow perch (Perca flavescens) may be negatively affected by food resource competition with rainbow smelt (Franzin and others 1994; Hrabik and others 1998). Rainbow smelt often become a preferred prey for predators such as walleye (Stizostedion vitreum), northern pike (Esox lucius), and lake trout (Salvelinus namaycush) in the lakes that they invade (Scott and Crossman 1973). This can lead to increased growth rates and higher concentrations of bioaccumulated contaminants in the predators concerned (MacCrimmon and others 1983; Mathers and Johansen 1985; Evans and Loftus 1987; Rasmussen and others 1990; Cabana and Rasmussen 1994). The increase in contaminant concentrations is believed to occur because smelt are more piscivorous, and thus feed at a higher trophic position, than native forage fish (Evans and Loftus 1987; Franzin and others 1994; Vander Zanden and Rasmussen 1996). Thus, their introduction into boreal lakes may lengthen the food chain to the top predators, resulting in higher predator mercury (Hg) concentrations in invaded lakes than in noninvaded ones (Rasmussen and others 1990; Cabana and others 1994). A temporal study on the effects of smelt invasion on predator fish was conducted on a series of Hud-

son Bay drainage lakes in 1999 and 2000 (Johnston and others forthcoming). The data showed that there were significant differences between smeltinvaded and reference lakes in the ␦15N signatures of predators. However, no significant differences in predator Hg concentrations were detected following smelt invasion. This was unexpected. Previous studies have indicated that predators utilizing smelt as forage have relatively higher Hg concentrations (MacCrimmon and others 1983; Mathers and Johansen 1985; Vander Zanden and Rasmussen 1996), and it is widely accepted that there is a positive relationship between trophic position and bioaccumulative contaminant load (Rasmussen and others 1990; Cabana and others 1994; Cabana and Rasmussen 1994; Kidd and others 1995). Therefore, the purpose of the present study was to examine the key assumptions of smelt-induced predator Hg increases: (a) that smelt are trophically elevated relative to native forage fish (analyzed using stable carbon (C) and nitrogen (N) isotope ratios), and (b) that this trophic elevation is indeed coupled to higher concentrations of Hg.

MATERIALS

AND

METHODS

Study Sites and Field Sampling Ten lakes in northwestern Ontario were sampled during August 2000 (Table 1). Lakes located near point sources of Hg were avoided (for example, lakes on the English–Wabigoon river system). Hg present in the study lakes is most likely derived from either natural sources or the long-range atmospheric transport of anthropogenic sources. The lakes ranged in trophic status from mesotrophic to oligotrophic, and all contained walleye as one of their top predators. Five of the lakes had established rainbow smelt populations, and five were uninvaded (reference) lakes. Sampled fish species included yellow perch, cisco, spottail shiner (Notropis hudsonius), rainbow smelt, pumpkinseed (Lepomis gibbosus), trout-perch (Percopsis omiscomaycus), rock bass (Ambloplites rupestris), and young-of-the-year walleye. All of these fish would be considered forage, or prey, for predator species. For this study, forage fish species are defined as having a fork length of less than 80 mm and are commonly found in predator fish stomachs. Forage fish were sampled with small-mesh monofilament gill nets (20 –50 mm stretched mesh) set in both the littoral and pelagic regions of the lakes. To establish a lake-specific baseline signature for stable isotope ratios, unionid clams (Pyganodon grandis) were also collected. Clams were collected by dip

Trophic Positions and Mercury Bioaccumulation

291

Table 1. Characteristics of Study Lakes Lake

Latitude

Longitude

Area (ha)

Max Depth (m)

Secchi (m)

Smelt-invaded?

Pickerel Mille-Lacs Sandbar Cloud Rainy Pickwick Cliff Red Coli Minnitaki

48°37' 48°50' 49°28' 48°08' 48°38' 49°00' 50°10' 51°03' 51°20' 49°58'

91°19' 90°30' 91°35' 89°33' 93°00' 93°06' 93°18' 93°49' 93°35' 92°00'

6058.8 24,112.8 1307.2 — 858.0 492.1 2663.4 17,676.7 2125.9 18,087.9

74.7 24.4 13.7 — 49.07 31.1 34.1 42.7 9.2 48.5

3.8 1.6 2.4 3.5 3.4 3.7 4.3 2.3 2.5 2.6

Yes No No Yes Yes No Yes Yes No Yes

netting and snorkelling and were generally found in sheltered depositional habitats with sand/silt substrate. Approximately 10 individuals of each fish species and clams were collected from each lake. Following capture, all forage fish and clam samples were frozen whole. In addition, frozen dorsal muscle samples and length data from approximately 10 adult walleye/lake were obtained from a previous fish collection (Johnston and others in press).

Stable isotope ratios were expressed as parts per thousand (‰ or per mil) delta values (␦15N and ␦13C) from an international standard. The standard for C is Pee-Dee Belomnite; that for N is atmospheric N gas (Peterson and Fry 1987). The delta values were calculated using the following formula:

Sample Preparation and Analyses

where R ⫽ 15N/14N or 13C/12C. A series of five internal standards were run every 24 samples for a total of 76 samples and 20 standards per run. Internal precision for standards run alongside the samples were usually less than ⫾ 0.2‰. Total Hg burdens in forage fish muscle were estimated from dry tissue using a modified method of Hendzel and Jamieson’s (1976) hot block method, followed by cold-vapor atomic absorption spectroscopy.

Forage fish and clams were thawed at room temperature and blotted to remove excess moisture; a series of body measurements was then made. Each clam was measured for shell length and width. Fork length, total length, wet mass, body depth, and gape were determined for each forage fish. Fish body depth was measured as the maximum dorsal– ventral distance, and gape was taken as the maximum distance between the upper and lower jaws at full maxillary distension. To minimize sampling error, all measurements were performed by the same person (H.K.S.). A dorsal sample of white muscle was dissected from each fish, and the foot muscle was dissected out of each clam. These tissues were chosen because of their low lipid and inorganic carbonate content (high concentrations of either can increase the variance in stable isotope results) (Pinnegar and Polunin 1999). All forage fish, clam, and adult walleye samples were then oven-dried (24 h at 60°C), ground to a fine powder using a mortar and pestle, and analyzed for stable isotope composition (all fish and clams) and Hg concentration (all fish except adult walleye). Analyses of stable C and N isotopes were performed at the University of New Brunswick on a Finnigan Mat Delta Plus continuous-flow isotoperatio mass spectrometer connected to a Thermoquest NC2500 elemental analyzer (EA-CFIRMS).

␦ 15N or ␦ 13C ⫽ [(R sample ⫺ R standard)/R standard] ⫻ 1000

(1)

Data Analyses Statistical analyses included linear regression, analysis of covariance (ANCOVA), and analysis of variance (ANOVA) (GLM procedure; V. 6; SAS Institute, Cary, NC, USA). We used ␦15N as an index of trophic position (DeNiro and Epstein 1981; Minigawa and Wada 1984). However, differences in ␦15N values at the base of the food web make absolute comparisons among lakes difficult (Cabana and Rasmussen 1996; Vander Zanden and Rasmussen 1999; Post 2002). Baseline (clam) ␦15N values differed significantly among the lakes of this study (ANOVA, F ⫽ 54.6, P ⬍ 0.0001, df ⫽ 9, 84). There are a number of different models available to account for this variation. We could not use the two-member mixing model proposed by Post (2002) because we were not able to collect snails from all lakes. We

292


chose not to use the method suggested by Vander Zanden and Rasmussen (1999) because there is no consistent relationship between baseline ␦13C and baseline ␦15N in our set of lakes (a necessary assumption for their model) (Post 2002). We therefore used a modified version of the model proposed by Cabana and Rasmussen (1996) where lake-specific clam ␦ 15N values are subtracted from fish ␦15N values. We did not divide by the conventional 3.4‰, however, since this estimate of ␦15N fractionation is only valid when applied to entire food webs or multiple trophic interactions (Post 2002). Adjusted ␦15N values were therefore calculated as follows: adjusted ␦ 15N ⫽ ␦ 15N fish–␦ 15N baseline

(2)

where ␦15Nbaseline is the arithmetic mean of clam ␦15 N values within each lake. Adjusted ␦15N values and ␦13C values were then compared among forage fish species and between treatment groups (smelt-invaded versus reference lakes) using an ANCOVA model with fish wet mass as the covariate and lake as the class variable. Fish wet mass was loge-transformed prior to analyses. Least-squared means for ␦15N and ␦13C were calculated for each forage species in each lake at a standardized body mass of 10 g. A value of 10 g was chosen because it fell within the size ranges of captured fish for all species. Once calculated, the least-squared mean adjusted ␦15N and ␦13C values were compared among species using a one-way ANOVA followed by Tukey’s HSD comparison of means. The same least-squared means were compared between treatment groups (smelt-invaded versus reference) using a t-test. Comparison of adult walleye stable isotope ratios between smeltinvaded and reference lakes followed a similar procedure, with the exception that fork length (in mm) rather than weight was used as the body size covariate. Muscle Hg concentrations of forage fish species were analyzed in a similar manner to adjusted ␦15N and ␦13C. An ANCOVA model was developed with loge wet mass as the covariate and lake as the class variable. Least-squared mean Hg concentrations were calculated for each species at a standard mass of 10 g. These values were then compared among species using a one-way ANOVA and Tukey’s test. As well, the relationships between adjusted ␦15N and muscle Hg concentrations were examined using linear regression analysis. Because heterogeneity of slopes was observed among species for ␦15N versus body size and Hg concentration versus body size relationships, simple

regression analyses were performed to compare the species-specific slopes for these relationships. This analysis was performed to determine how changing the standard body size at which least-squared means were estimated may have influenced the observed trends. In addition to ␦15N and Hg concentration, two body–morphology ratios were analyzed: (a) fish gape– body mass, and (b) fish body depth– body mass. Again, an ANCOVA model was developed to account for variability in fish body size and among lakes. The gape– body mass ratio was analyzed to determine the ability of smelt to eat relatively large prey items; the depth– body mass ratio was analyzed to determine the availability of smelt to gapelimited predators.

RESULTS Trophic Position A comparison of adjusted ␦15N among forage fish species indicated that at a body mass of 10 g, rainbow smelt had a significantly higher ␦15N signature than all native forage species except trout-perch and juvenile walleye (Figure 1) (ANOVA, F ⫽ 13.59, P ⬍ 0.0001, df ⫽ 7, 30) (Tukey’s test, P ⬍ 0.05). The ␦13C signature of rainbow smelt appeared to be most similar to that of cisco and trout-perch (Figure 1). Rainbow smelt, cisco, and trout-perch were 13C depleted relative to juvenile walleye, yellow perch, spottail shiners, trout-perch, and pumpkinseed. Pumpkinseed had a significantly higher (that is, less negative) signature than all other species (Figure 1) (ANOVA, F ⫽ 3.43, P ⫽ 0.018, df ⫽ 5, 37) (Tukey’s test, P ⬍ 0.05). Adjusted ␦15N of 10-g native forage fish did not differ significantly between smelt-invaded and reference lakes for either yellow perch (t-test, t ⫽ ⫺0.34, df ⫽ 7.4, P ⫽ 0.74) or cisco (t-test, t ⫽ 0.29, df ⫽ 4.0, P ⫽ 0.78) (Figure 2). It was not possible to perform this test with the less common native forage species because the necessary sample sizes were not available. Following invasion, rainbow smelt appear to move into a previously unoccupied trophic position (Figure 2); their ␦15N value was approximately 0.8‰ greater than that of yellow perch, the closest native forage species. Based on the elevated ␦15N signature of rainbow smelt relative to native forage species, we predicted that the ␦15N of predators such as adult walleye should be elevated in smelt-invaded lakes relative to reference lakes. For 500-mm walleye, mean ␦15N values were 0.57‰ higher in smelt-invaded lakes


293

Figure 2. Vertical dot plot illustrating the trophic position occupied by yellow perch, cisco, and walleye (adults) in the two different treatment groups (smelt-invaded and reference). Rainbow smelt trophic position is also indicated. The trophic positions of the native forage species do not differ significantly between treatments (t-tests, t ⱕ 0.29, P ⬎ 0.05, df ⱖ 4.0); rainbow smelt appear to move into a previously unoccupied trophic position. Error bars represent SE.

Figure 1. Box plots illustrating A adjusted ␦15N (per mil) and B ␦13C (per mil) of the forage species. Adjusted ␦15N is corrected for baseline differences across all lakes and calculated at a standard body size of 10 g. The box plots show, from bottom to top, the 10th, 25th, 50th (center line), 75th, and 90th percentiles. Rainbow smelt show significant trophic elevation compared to all other native forage species except trout-perch and juvenile walleye (ANOVA, F ⫽ 20.43, P ⬍ 0.0001, df ⫽ 3, 2). Pumpkinseed have a significantly less negative ␦13C signature than all other species (ANOVA, F ⫽ 3.43, P ⫽ 0.018, df ⫽ 5, 37) (Tukey’s test, P ⬍ 0.05). Sample sizes refer to the number of lakes sampled for each species. CIS, cisco; PS, pumpkinseed; SS, spottail shiner; YP, yellow perch; RS, rainbow smelt; WALL-J, juvenile walleye; TP, trout-perch.

than reference lakes; this difference was statistically significant (t-test, t ⫽ 2.51, P ⫽ 0.038, df ⫽ 1,8). In addition, walleye of smelt-invaded lakes were depleted in 13C relative to the reference lakes by an average of 1.1‰ (t-test, t ⫽ 4.99, P ⫽ 0.0012, df ⫽ 1, 8).

Hg Concentrations Muscle Hg concentrations (adjusted to 10-g body mass) differed significantly among the forage fish species (ANOVA, F ⫽ 7.27, P ⬍ 0.0001, df ⫽ 5, 28). Spottail shiners had the highest muscle Hg concen-

tration of all forage species. The differences between spottail shiner and most other species were significant (Tukey’s test, P ⬍ 0.05), with the exception of juvenile walleye and pumpkinseed (Figure 3). Smelt had the second lowest mean muscle Hg concentration; only cisco was lower. The differences between smelt and other forage species were largely nonsignificant (Tukey’s test, P ⬎ 0.05) (Figure 3). Following adjustment for among lake variation and pooling of all species, we found no significant interspecific relationship between muscle Hg concentration and ␦15N (regression analysis, R2 ⫽ 0.0007, n ⫽ 260, P ⫽ 0.68). The relationships within the four most common species of forage fish (yellow perch, cisco, spottail shiner, and smelt) were variable. The relationships were positive and significant for cisco and smelt (regression analysis, R2 ⬎ 0.21, n ⬎ 60, P ⬍ 0.001), but they were nonsignificant for yellow perch and spottail shiner (regression analysis, R2 ⬍ 0.007, n ⬎ 40, P ⬎ 0.51) (Figure 4).

␦15N versus Body Size and Hg Concentration versus Body Size Relationships ␦15N–logemass relationships were heterogeneous among the four most common forage species (ANCOVA, F ⫽ 6.54, P ⬍ 0.0003, df ⫽ 3, 292). Follow-

294


Figure 3. Box plots illustrating muscle Hg concentrations (ppm) for cisco (CIS), rainbow smelt (RS), yellow perch (YP), juvenile walleye (WALL-J), trout-perch (TP), pumpkinseed (PS), and spottail shiner (SS). Flesh Hg concentrations were compared at a standardized body mass of 10 g. The box plots show, from bottom to top, the 10th, 25th, 50th (center line), 75th, and 90th percentiles. Rainbow smelt have lower Hg concentrations than all native species except cisco. It is significantly lower than spottail shiner (ANOVA, F ⫽ 7.267, P ⬍ 0.0001, df ⫽ 5, 28).

Figure 4. Regression analysis quantifying the speciesspecific relationships between muscle Hg concentration (ppm) and adjusted ␦15N (per mil) for the four most common forage species. Cisco and smelt showed significant relationships between Hg concentration and ␦15Ndetermined trophic position (regression analysis, r2 ⬎ 0.21, P ⬍ 0.0001, n ⫽ 63, 63), whereas yellow perch and spottail shiner did not (regression analysis, r2 ⬍ 0.007, P ⬎ 0.51, n ⫽ 84, 47).

ing adjustment for among lake variation and pooling of data across lakes, simple regression analysis of these relationships revealed that both the slope

Figure 5. Regression analysis quantifying the speciesspecific relationships between trophic position (␦15N per mil) and loge(mass). Rainbow smelt have the steepest slope and the highest intercept. Cisco and spottail shiner were intermediate for slope; yellow perch again showed almost no relationship. There is one point of intersection between yellow perch and spottail shiner; it occurs at a mass of approximately 9 g.

and intercept were greatest for rainbow smelt. This indicates that smelt ␦15N was higher than that of the native forage species over a broad range of body sizes, and that the rate of ␦ 15N increase with body size was greatest for smelt. Slopes were intermediate for cisco and spottail shiner, whereas the slope for yellow perch approached zero, indicating a very weak ␦15N–logemass relationship (Figure 5). There was one point of intersection between yellow perch and spottail shiner at a mass of approximately 9 g; however, due to the high amount of intraspecific variation, this intersection point should be interpreted with caution (Figure 5). The relationships between Hg and logemass were also heterogeneous among species (ANCOVA, F ⫽ 6.48, P ⬍ 0.0003, df ⫽ 3, 292). Following adjustment for between-lake variation and pooling of the data, a simple regression analysis revealed that spottail shiner and rainbow smelt had the steepest slopes and thus the strongest Hg–logemass relationships. The slope for yellow perch again approached zero (Figure 6). Rainbow smelt had a lower intercept and a steeper slope than either cisco or yellow perch. This produced two points of intersection. The cisco–smelt intersection occurred at a logemass of 2 (mass ⫽ 7.4 g); the yellow perch–smelt intersection occurred at a logemass of 3.17 (mass ⫽ 23.6 g) (Figure 6). Again, however, these intersection points should be viewed with caution, particularly the cisco–smelt intersection.


Figure 6. Regression analysis quantifying the speciesspecific relationships between muscle Hg concentration (ppm) and body mass. Spottail shiner and rainbow smelt have the steepest slopes. The slope for yellow perch is almost zero; the Hg–logemass relationship is very weak for this species. Because rainbow smelt have a lower intercept but a steeper slope than either cisco or yellow perch, there are two points of intersection. The cisco– smelt intersection occurs at a logemass of 2 (mass ⫽ 7.4 g); the yellow perch–smelt intersection occurs at a logemass of 3.17 (mass ⫽ 23.6 g).

Gape–Body Mass and Depth–Body Mass Analyses At a common body size of 10 g, we found significant differences among species for both the depth– body mass (ANOVA, F ⫽ 234, P ⬍ 0.0001, df ⫽ 9, 208) and gape– body mass ratios (ANOVA, F ⫽ 279, P ⬍ 0.0001, df ⫽ 6, 119) (Tukey’s test, P ⬍ 0.05). The depth– body mass ratio of rainbow smelt was lower than that of all other native forage species, indicating that they are more streamlined than the native forage species. Most species by species comparisons of gape– body mass were significantly different from one another. Rainbow smelt had a significantly higher gape– body mass ratio (indicating a larger relative gape) than all other species except rock bass (which were not included in stable isotope or Hg analyses because they were only present in two lakes). Pumpkinseed and trout-perch gape– body mass ratios were also not significantly different from each other (Tukey’s test, P ⬎ 0.05).

DISCUSSION Trophic Position The observed trophic elevation of rainbow smelt relative to cisco, spottail shiner, yellow perch, and

295

pumpkinseed is consistent with the conclusions of earlier dietary (gut contents) analyses (Vander Zanden and Rasmussen 1996). This is the first study, however, to quantify the trophic positions of these species using stable isotope analyses. Previous authors have either inferred trophic elevation indirectly from contaminant evidence or used stomach contents data. Our results, based on N isotope analyses, indicate that smelt show the greatest trophic elevation relative to cisco (approximately 2.75‰) and a lesser elevation relative to yellow perch and spottail shiner (approximately 1.2‰). Rainbow smelt were not trophically elevated relative to trout-perch or either juvenile or adult walleye. This result is not surprising for walleye given that they are a predatory rather than a forage species. The trout-perch result was unexpected. The feeding ecology of trout-perch is not well documented, but they are believed to feed on chironomid larvae, ephemeroptera larvae, amphipods, and small fish (Scott and Crossman 1973). Data are sparse, however, and it is possible that fish may contribute more to the diet of trout-perch than was previously thought. In addition, the majority of trout-perch sampled ranged in size from 5 to 7 g. Thus, in making a limited number of ␦15N comparisons at 10 g, it is possible that an extrapolation bias was introduced. This bias was not expected to affect the results of any of the other study species since adequate numbers of 10-g fish were caught and sampled. Previous investigations of the relationship between trophic position and body size have produced mixed results. For example, lake trout of southern Ontario and Quebec lakes show no significant relationship between trophic position and body size (Vander Zanden and others 2000). There is also a very weak relationship between predator and prey size and no relationship between prey body size and prey trophic position in these lakes. In contrast, positive body size versus ␦15N relationships have been reported for several northern salmonid populations (Hobson and Welch 1995; Kline and others 1998), possibly because of ontogenetic shifts in feeding. Ontogenetic shifts in feeding do not seem to account for changes in trophic position in Lake Champlain walleye, however, where ␦15N increases more strongly with age than body size (Overman and Parrish 2001). We observed significant trophic position (␦15N)– body size relationships in all forage species except yellow perch. These relationships differed among lakes and among species (that is, heterogeneity of slopes was found in the ANCOVA model), so comparisons of trophic position had to be made at a

296


standardized body mass of 10 g. However, examination of the species-specific ␦15N– body size relationships showed that rainbow smelt were trophically elevated relative to the three major native forage species (cisco, spottail shiner, and yellow perch) over a wide range of body sizes. Thus, the body size chosen for comparison did not seem to affect the trends observed.

Stable C Isotope Analyses The ␦13C analysis indicates that smelt are relatively depleted in 13C. This suggests that either they are part of the pelagic food web of the study lakes (France 1995), or that their muscle tissue is relatively lipid rich (Focken and Becker 1998). It is consistent with the results of previous studies showing that measurable impacts of smelt on native fish (for example, competition, heavy predation) are most commonly experienced by pelagic species (Evans and Loftus 1987). It also suggests that predators that switch from native forage to smelt should show a downward shift in ␦13C. Our results agree with this conclusion; adult walleye ␦13C signatures were significantly more negative in smelt-invaded lakes than in reference lakes.

Trophic Analyses across Treatment Groups Previous analyses of the effects of smelt on energy transfer within lakes have focused primarily on pathways leading to the top predator (Evans and Loftus 1987; Vander Zanden and Rasmussen 1996). It is thought that smelt may (a) divert more energy to the top predators, resulting in increased growth rates (Evans and Loftus 1987); and (b) lengthen trophic pathways, resulting in trophic elevation of terminal predators (Vander Zanden and Rasmussen 1996). However, it has not yet been established whether changes in trophic structure feed down the food chain as well as up. That is, it is not yet known whether the trophic position of native forage species changes after the invasion of smelt. We found the trophic positions of cisco and yellow perch in smelt-invaded and reference lakes to be virtually identical. This suggests that invading rainbow smelt do not disrupt the trophic structure of the native forage fish community. Rather, they appear to move into a previously unoccupied trophic position. The effects of smelt on native forage fish therefore appear to be limited to those caused by interspecific competition and predation (Evans and Loftus 1987; Hrabik and others 1998) rather than by trophic restructuring. In agreement with the dietary trophic position study of Vander Zanden and Rasmussen (1996),

however, we found that walleye (a predatory fish) do experience a trophic shift following smelt invasion. We detected significantly higher ␦15N and significantly lower ␦13C in adult walleye of smeltinvaded lakes relative to reference lakes. However, the mean difference in walleye ␦15N between smelt-invaded and reference lakes was only 0.57‰, which equates to an increase of only around onesixth of a trophic level. This weak response is probably due to the fact that walleye in these lakes continued to consume a high proportion of native forage fish after smelt invasion. Pelagic predators, such as lake trout, would be more likely to switch completely from low ␦15N prey (for example, cisco) to smelt and thus may show a greater trophic shift than walleye.

Gape–Body Mass and Depth–Body Mass Ratios It has been hypothesized that the trophic elevation of smelt relative to that of native forage species results from higher levels of piscivory and cannibalism in smelt (Scott and Crossman 1973; Evans and Loftus 1987; Franzin and others 1994). Anecdotal evidence suggests that this situation arises because smelt have a larger gape than native forage species. We confirmed this assumption quantitatively. Rainbow smelt had a higher gape– body mass ratio than all other native forage species except rock bass. Since there is often a positive correlation between body size and trophic position in aquatic systems (Hobson and Welch 1995; France and others 1998), the ability to eat relatively larger prey may help to explain the elevated trophic position of smelt. In addition to a high gape– body mass ratio, we found that smelt have a relatively low body depth– body mass ratio. This fact, along with their soft body and absence of spiny rays, suggests that smelt may be relatively easier to eat (at a given mass) than native forage species, particularly for gape-limited predators. The combination of a high gape– body mass ratio and a low body depth– body mass ratio appears to allow smelt to consume a wide range of prey sizes and, in turn, to be readily eaten by a wide range of predators. As a result, they may be able to transfer energy through aquatic food webs more efficiently than native forage species. It may also explain the observation that smelt show a higher incidence of intercohort cannibalism than native forage species.

Hg Concentrations Based on the elevated trophic position occupied by smelt in invaded lakes, we predicted that Hg con-

Trophic Positions and Mercury Bioaccumulation centrations would be higher in smelt than in native forage species (MacCrimmon and others 1983; Mathers and Johansen 1985; Rasmussen and others 1990; Kidd and others 1995; Vander Zanden and Rasmussen 1996). But contrary to our expectation, there was no consistent relationship between trophic position and Hg bioaccumulation at the reference body size (10 g). Spottail shiner, which had the highest Hg concentration, occupied an intermediate trophic position whereas rainbow smelt, one of the most trophically elevated species, had the second-lowest muscle Hg concentration. This decoupling of trophic position from muscle Hg concentration was unexpected, but the absolute Hg values we obtained seem to be reasonable. Muscle Hg concentrations reported for smelt in Cayuga Lake, New York, averaged 0.285 ppm (dry weight) (Gutenmann and others 1998), in close agreement with the average of 0.30 ppm (dry weight) we reported. Caution must be exercised in making comparisons, however, because our data are corrected for body size, whereas Gutenmann and others (1998) reported a simple mean. Wren and MacCrimmon (1986) analyzed muscle Hg concentrations in a number of forage fish species, including yellow perch, rock bass, and rainbow smelt—three of the species in the current study. The muscle Hg concentrations they reported were within 0.07 ppm of those found by us. We therefore conclude that the observed decoupling between trophic position and muscle Hg concentration at 10 g is not due to laboratory or sampling error. One possible explanation for the unexpected decoupling may be differences in growth and/or metabolic rates among species. Hg concentrations at a given fish size are often inversely related to prior growth rate (the so-called growth-dilution effect) (Huckabee and others 1979; Doyon and others 1998; Harris and Bodaly 1998). Fish with fast growth rates are thought to experience lower cumulative Hg uptake than those with slow growth rates, presumably because less time is required to reach a given size (Harris and Bodaly 1998). Additionally, fish that are relatively smaller at a given age tend to have higher metabolic rates (Doyon and others 1998). This means that at a given food intake, smaller fish direct more energy toward maintenance and less to muscle production than do large fish. At a given age, therefore, contaminants will be more concentrated in smaller fish (Huckabee and others 1979; Doyon and others 1998; Harris and Bodaly 1998), and it is possible that smelt are growing more quickly than the native forage species. Literature values for the von Bertalanffy k parameter (rate of fish growth to maximum size) are

297

higher for smelt (0.31) than for either cisco (0.07– 0.29) or yellow perch (0.17– 0.19) (Froese and Pauly 2002). In addition, anecdotal and limited experimental evidence suggest that smelt experience high growth rates and reach high abundances in recently invaded lakes. It is thought that this initially high growth rate subsequently slows until an equilibrium is reached (Copeman and McAllister 1978). The majority of the “smelt” lakes in this study were invaded within the last 15 years. It is thus possible that the unexpectedly low Hg concentrations observed in the smelt of this study resulted from growth dilution/age difference effects between species. As with trophic position, heterogeneity of slopes was observed among species in the Hg concentration– body size relationships. Comparisons of muscle Hg concentrations were therefore made at a standardized body mass of 10 g. As discussed, this was not a problem in the trophic position analysis (which also standardized for body mass), because smelt had a higher trophic position than native forage species across the full range of body sizes. However, this consistency was not seen in the Hg concentration– body size relationships. The regression lines for smelt and yellow perch intersected at approximately 20 g, whereas those for smelt and cisco intersected at 7 g. This may reflect speciesspecific differences in growth trajectories. Species that grow fast initially but then show declining growth rates as they mature should have stronger Hg concentration– body size relationships than species that sustain a rapid growth rate. Detailed age data and a larger size range of sampled fish would be needed to determine if this explains the differing Hg concentration– body size relationships for our study species. The heterogeneity of Hg concentration– body size slopes also has important implications for predicting the potential effects of smelt invasions on piscivore Hg concentrations. An accurate prediction would require data on both the species composition and the size distribution of the preinvasion and postinvasion prey diet. In some cases, the ability to make accurate predictions may be further complicated by ontogenetic shifts in predator feeding. For example, Colby and others (1987) observed that in boreal lakes walleye that were smaller than 350 mm fed primarily on yellow perch whereas those that were larger than 350 mm fed mostly on cisco.

CONCLUSIONS We predicted that trophic position and Hg concentration would be tightly coupled, and that both

298


would be elevated in smelt relative to native forage fish. Instead, it appears that invading smelt move into a previously unoccupied and elevated trophic position relative to native forage species, but that this trophic elevation does not necessarily lead to higher Hg concentrations. We suggest that in boreal lakes the general rules of food web theory and bioaccumulation may not be applicable at finer scales of trophic position (for example, comparing species that differ only by one-fourth to one trophic level). We also suggest, therefore, that the use of contaminants as tracers of trophic relationships (see Vander Zanden and Rasmussen 1996) may be inappropriate in some study systems.

ACKNOWLEDGMENTS Field assistance was provided by the Ontario Ministry of Natural Resources. Hg analyses were performed by Neil Strange, stable isotope analyses were performed by Anne McGeachy, and additional lab assistance was provided by Tom Herra and Alison Mudge. Irene Gregory-Eaves, Peter Hodson, Karen Kidd, and two anonymous reviewers gave constructive criticisms on earlier drafts of this manuscript. This research was supported by the Department of Fisheries and Oceans (R.A.B.) and a Natural Sciences and Engineering Research Council of Canada (NSERC) operating grant (W.C.L.). H.K.S. was supported by a NSERC Undergraduate Research Award, and T.A.J. was supported by an NSERC Postdoctoral Fellowship.

REFERENCES Cabana G, Rasmussen JB. 1996. Comparison of aquatic food chains using nitrogen isotopes. Proc Natl Acad Sci USA 93: 10844–7. Cabana G, Rasmussen JB. 1994. Modelling food chain structure and contaminant bioaccumulation using stable nitrogen isotopes. Nature 372:255–7. Cabana G, Tremblay A, Kalff J, Rasmussen JB. 1994. Pelagic food chain structure in Ontario lakes: a determinant of mercury levels in lake trout (Salvelinus namaycush). Can J Fish Aquat Sci 51:381–9. Campbell KB, Derksen AJ, Remnant RA, Stewart KW. 1991. First specimens of rainbow smelt, Osmerus mordax, from Lake Winnipeg, Manitoba. Can Field Nat 105:568–70. Colby PJ, Ryan PA, Schupp DE, Serns SL. 1987. Interactions in north-temperate lake fish communities. Can J Fish Aquat Sci 44:104–28. Copeman DG, McAllister DE. 1978. Analysis of the effect of transplantation on morphometric and meristic characteristics in lake populations of the rainbow smelt, Osmerus mordax (Mitchill). Environ Biol Fish 3:253–9. DeNiro MJ, Epstein S. 1981. Influence of diet on the distribution of nitrogen isotopes in animals. Geochim Cosmochim Acta 45:341–51. Doyon JF, Schetegne R, Verdon R. 1998. Different mercury

bioaccumulation rates between sympatric populations of dwarf and normal lake whitefish (Coregonus clupeaformis) in the La Grande complex watershed, James Bay, Quebec. Biogeochemistry 40:203–16. Evans DO, Loftus DH. 1987. Colonization of inland lakes in the Great Lakes region by Rainbow Smelt, Osmerus mordax: their freshwater niche and effects on indigenous fishes. Can J Fish Aquat Sci 44:249–66. Focken U, Becker K. 1998. Metabolic fractionation of stable carbon isotopes: implications of different proximate compositions for studies of the aquatic food webs using ␦13C data. Oecologia 115:337–43. France RL. 1995. Differentiation between littoral and pelagic food webs in lakes using stable carbon isotopes. Limnol Oceanogr 40:1310–3. France RL, Chandler M, Peters R. 1998. Mapping trophic continua of benthic foodwebs: body size-␦15N relationships. Mar Ecol Prog Ser 174:301–6. Franzin WG, Barton BA, Remnant RA, Wain DB, Pagel SJ. 1994. Range extension, present and potential distribution, and possible effects of rainbow smelt in Hudson Bay drainage waters of northwestern Ontario, Manitoba, and Minnesota. North Am J Fish Manage 14:65–76. Froese R, Pauly D, editors. 2002. FishBase. Available online at: www.fishbase.org, 19 August 2002. Gutenmann WH, Reid CM, Lisk DJ. 1998. Mercury content of smelt in Cayuga lake in central New York state. Chemosphere 37:391–2. Harris RC, Bodaly RA. 1998. Temperature, growth, and dietary effects on fish mercury dynamics in two Ontario lakes. Biogeochemistry 40:175–87. Hendzel MR, Jamieson DM. 1976. Determination of mercury in fish. Analyt Chem 48:926–8. Hobson KA, Welch HE. 1995. Cannibalism and trophic structure in a high arctic lake: insights from stable-isotope analysis. Can J Fish Aquat Sci 52:1195–201. Hrabik TR, Magnuson JJ, McLain AS. 1998. Predicting the effects of rainbow smelt on native fishes in small lakes: evidence from long-term research on two lakes. Can J Fish Aquat Sci 55: 1364–71. Huckabee JW, Elwood JW, Hildebrand SG. 1979. Accumulation of mercury in freshwater biota. In: Nriagu JO, editor. The biogeochemistry of mercury in the environment. Amsterdam: Elsevier/North-Holland Biomedical Press. p 277–302. Johnston TA, Leggett WC, Bodaly RA, Swanson HK. Temporal changes in mercury bioaccumulation by predatory fishes of boreal lakes following the invasion of an exotic forage fish. Environ Toxicol Chem. Forthcoming Kidd KA, Hesslein RH, Fudge RJP, Hallard KA. 1995. The influence of trophic level as measured by ␦15N on mercury concentrations in freshwater organisms. Water Air Soil Pollut 80:1011–5. Kline TCJ, Wilson WJ, Goering JJ. 1998. Natural isotope indicators of fish migration at Prudhoe Bay, Alaska. Can J Fish Aquat Sci 55:1494–502. MacCrimmon HR, Wren CD, Gots BL. 1983. Mercury uptake by lake trout, Salvelinus namaycush, relative to age, growth, and diet in Tadenac Lake with comparative data from other precambrian shield lakes. Can J Fish Aquat Sci 40:114–20. Mathers RA, Johansen PH. 1985. The effects of feeding ecology on mercury accumulation in walleye (Stizostedion vitreum) and pike (Esox lucius) in Lake Simcoe. Can J Zool 63:2006–12.

Trophic Positions and Mercury Bioaccumulation Minagawa M, Wada E. 1984. Stepwise enrichment of 15N along food chains: further evidence and the relation between ␦15N and animal age. Geochim Cosmochim Acta 48:1135–40. Overman NC, Parrish DL. 2001. Stable isotope composition of walleye: 15N accumulation with age and area-specific differences in ␦13C. Can J Fish Aquat Sci 58:1253–60. Peterson BJ, Fry B. 1987. Stable isotopes in ecosystem studies. Annu Rev Ecol Syst 18:293–320. Pinnegar JK, Polunin NVC. 1999. Differential fractionation of ␦13 C and ␦15N among fish tissues: implications for the study of trophic interactions. Funct Ecol 13:225–31. Post DM. 2002. Using stable isotopes to estimate trophic position: models, methods, and assumptions. Ecology 83:703–18. Rasmussen JB, Rowan DJ, Lean DR, Carey JH. 1990. Food chain structure in Ontario lakes determines PCB levels in lake trout (Salvelinus namaycush) and other pelagic fish. Can J Fish Aquat Sci 47:2030–8. Remnant RA, Graveline PG, Bretecher RL. 1997. Range exten-

299

sion of the rainbow smelt, Osmerus mordax, in the Hudson Bay drainage of Manitoba. Can Field Nat 111:660–2. Scott WB, Crossman EJ. 1973. Freshwater fishes of Canada. Bulletin of the Fisheries Research Board of Canada no. 184: 966 p. Vander Zanden MJ, Rasmussen JB. 1999. Primary consumer ␦13C and␦ 15N and the trophic position of aquatic consumers. Ecology 80:1395–404. Vander Zanden MJ, Rasmussen JB. 1996. A trophic position model of pelagic food webs: impact on contaminant bioaccumulation in lake trout. Ecol Monogr 66:451–77. Vander Zanden MJ, Shuter BJ, Lester NP, Rasmussen JB. 2000. Within- and among-population variation in the trophic position of a pelagic predator, lake trout (Salvelinus namaycush). Can J Fish Aquat Sci 57:725–31. Wren CD, MacCrimmon HR. 1986. Comparative bioaccumulation of mercury in two adjacent freshwater ecosystems. Water Res 20:763–9.