Round goby predation on Dreissena in coastal areas of eastern Lake Michigan

Abstract

Predator–prey interactions between round goby (Neogobius melanostomus) and Dreissena are an important component of the food web in their invaded range in North America and Europe. We conducted two experiments to test the effect of round goby predation on Dreissena in coastal lakes connected to eastern Lake Michigan, USA. First, we conducted a density gradient experiment using 1-m² cages stocked with 0 to 15 large round gobies (individuals > 7.5 cm total length (TL)). Presence of large round gobies significantly reduced Dreissena densities in all cages. Second, we conducted an exclosure experiment (treatments: none, 6-mm, and 24-mm mesh) in four coastal lakes in littoral habitats. Excluding large round gobies did not significantly affect Dreissena densities in any of the lakes, although we did find a significant effect of excluding large-bodied (i.e., animals larger than adult round goby that cannot pass through 24-mm mesh) predators in one lake. In combination, our two experiments showed that the strength of predator–prey interactions between round goby and Dreissena can vary with location and/or conditions. We hypothesize that the strength of predator–prey interactions between round goby and Dreissena vary spatially across the aquatic landscape depending on habitat (e.g., rocky vs. sandy substrates), which is critical for understanding food-web dynamics, especially in areas where the round goby and Dreissena are invasive.

Keywords:

• aquatic invasive species
• fish predation
• Laurentian Great Lakes
• predator–prey interaction
• quagga mussel
• zebra mussel

Introduction

Human activities associated with the introduction of non-indigenous species are a significant component of global environmental change (Vitousek et al. 1997a, 1997b; Rahel 2000). In the Laurentian Great Lakes, the introduction of the zebra mussel (Dreissena polymorpha), quagga mussel (Dreissena bugensis), and round goby (Neogobius melanostomus) during 1986–1991 via ballast water from transoceanic ships (Jude et al. 1992; May and Marsden 1992; Carlton 2008) has resulted in significant ecological change (e.g., Vanderploeg et al. 2002; Corkum et al. 2004; Strayer 2009). Introduction of zebra and quagga mussels (hereafter Dreissena) to the Great Lakes may have facilitated the introduction and spread of round goby because these species have similar native ranges in the Ponto-Caspian region of Eurasia (Ricciardi 2001; Vanderploeg et al. 2002). Thus, the round goby can exploit Dreissena as prey more efficiently than native Great Lakes fishes (Ricciardi 2001; Vanderploeg et al. 2002). Moreover, the ecological effects of these invasive species are not limited to the Great Lakes basin because they have and continue to spread across North America and Europe (Corkum et al. 2004; Strayer 2009).

Large round gobies (individuals > 7.5 cm TL) are known to feed extensively on Dreissena (e.g., Ghedotti et al. 1995; Ray and Corkum 1997; French and Jude 2001; Campbell et al. 2009). Dreissena tend to be a less important component of the diet of round gobies less than 7.5 cm TL compared to larger individuals (Jude et al. 1995; French and Jude 2001). There is mounting evidence that round goby predation can affect the size structure and densities of Dreissena populations (Kuhns and Berg 1999; Djuricich and Janssen 2001; Barton et al. 2005; Lederer et al. 2006, 2008). Nevertheless, the importance of Dreissena in the diets of round gobies can vary among habitats (e.g., shallower vs. deeper offshore areas) as well as over time (Schaeffer et al. 2005) and the round goby has invaded areas that lack Dreissena (Carman et al. 2006). Thus, the ecological importance of predator–prey interactions between round goby and Dreissena may vary spatially across the aquatic landscape.

The goal of our study was to examine the predatory role of round goby on Dreissena densities in coastal areas of eastern Lake Michigan where predator–prey interactions between these organisms have not been previously studied. In situ enclosure and exclosure experiments are a practical way to identify strong interspecific interactions (Power 1992; Ruetz et al. 2002; Bartsch et al. 2005). First, we used in situ enclosures to examine the direct predatory role of large round gobies on Dreissena densities. We predicted a functional response of reduced Dreissena densities with increasing round goby densities. Second, we used in situ exclosures to examine the effect of round goby predation relative to larger predators (under natural densities) on Dreissena densities. The exclosure experiment was conducted at four study sites to determine whether the strength of the predator–prey interaction between round goby and Dreissena varied among study sites. We hypothesized that Dreissena densities would be reduced in treatments exposed to predation primarily by the round goby given their abundance in our study lakes (Cooper et al. 2007).

Materials and methods

Study sites

Field experiments were conducted in four coastal lakes along the eastern shore of Lake Michigan (Table 1). These lakes are all connected directly to Lake Michigan, and the round goby is present in each lake (Cooper et al. 2007, 2009).

Table 1. Site locations, lake surface area, water temperature, turbidity, Dreissena density and size, and round goby catch for the exclosure experiment. Temperature and turbidity (mean ± 1 SE) were measured during April (once), June (once in Kalamazoo and twice in other lakes), and September (once) 2005. Dreissena density (mean ± 1 SE, n = 3 bricks per lake) and size (mean and range) are based on bricks sampled at the beginning of the experiment. Round goby catch was in three fyke nets fished at each site during June 2005 (Cooper et al. 2007).

	Lake
Variable	Pentwater	White	Muskegon	Kalamazoo
Latitude (°)	+43.7710	+43.4102	+43.2431	+42.6489
Longitude (°)	−86.4156	−86.3575	−86.2822	−86.2015
Surface area (km^2)	1.96	10.26	16.96	0.65
Temperature (°C)	21.1 ± 2.3	20.1 ± 1.1	22.9 ± 2.3	22.6 ± 2.9
Turbidity (NTU)	8.1 ± 3.7	5.8 ± 1.8	4.4 ± 2.1	24.4 ± 7.3
Dreissena density (number/m^2)	13.3 ± 4.6	2.7 ± 0.7	9.7 ± 3.0	0 ± 0
Dreissena size (mm)	6.9 (1.4–25.6)	8.8 (4.3–15.5)	10.1 (2.9–16.6)	–
Round goby catch (number)	44	1	67	16

An enclosure experiment was conducted in Muskegon Lake at a site (N 43.244°, W 86.308°) with primarily sand substrate where submersed aquatic vegetation (SAV) was generally absent. Water depth at this site was about 60 cm, and mean daily water temperature (measured by a continuous recorder) during the experiment was 25.4°C (range = 22.2–28.2°C).

An exclosure experiment was conducted in four coastal lakes (Table 1). We selected a shallow (<1 m) site in each coastal lake that corresponded to the site Cooper et al. (2007) used as a SAV habitat. Previous work in these lakes showed round goby used areas with SAV preferentially over areas with bare sediment or emergent vegetation (Cooper et al. 2007). Among the four study sites where we conducted the exclosure experiment, round goby densities (measured as catch in fyke nets) were the highest in Muskegon and Pentwater lakes (Table 1). Daytime water temperatures were similar among sites, although turbidity was higher in Kalamazoo Lake. Additionally, surface area varied among lakes with Muskegon Lake, the largest, and Kalamazoo Lake, the smallest.

Enclosure experiment

We tested the effect of round goby predation on Dreissena abundance by conducting a density gradient experiment in Muskegon Lake. Twelve cages (1 × 1 × 0.5 m³) with 6-mm mesh on all sides were used to manipulate round goby density. The mesh size was sufficient for the movement of invertebrate prey in and out of cages (e.g., Ruetz et al. 2002), while acting as a barrier to round gobies ≥ 3.8 cm TL. Cages were secured to the lake bottom about 65 m from shore in a grid formation of three rows (four cages/row) with 0.5 m between cages. Each cage was stocked with three, randomly assigned cement bricks (19.5 × 9.3 × 5.8 cm³; surface area that could be colonized = 515.4 cm²) that were colonized by Dreissena. Bricks had been incubated in Muskegon Lake for 70 days prior to the start of the experiment. The initial density of Dreissena was 16.3 individuals/brick (SE = 2.9, n = 3 bricks), and the mean size of individuals was 13.0 mm (range = 6.9–18.6 mm). The experiment began on 28 June 2005 and was concluded after 71 days. Veliger densities of Dreissena often peak in July (Fraleigh et al. 1993; Garton and Haag 1993). Thus, the response of Dreissena densities on bricks reflects the effect of predation on adult Dreissena already on bricks as well as veliger colonization throughout the experiment.

A density (0, 5, 10, or 15 fish/m²) of round gobies was randomly assigned to each cage in a row (n = 3 replicate cages). We stocked cages with round gobies > 7.5 cm TL that were collected by angling from the channel connecting Muskegon Lake to Lake Michigan. Fish of this size were selected because they often feed extensively on Dreissena (e.g., Ray and Corkum 1997; French and Jude 2001; Janssen and Jude 2001). Each round goby stocked in a cage was marked with a unique fin clip. We recorded TL (cm) and wet mass (g) of each fish at the time of stocking and conclusion of the experiment. Cages were monitored daily for mortality at the start of the study. Once mortalities ceased for 1 week, cages were checked weekly. Round gobies that died were replaced, and the date of stocking was recorded. At the conclusion of the experiment, Dreissena were removed from bricks in the field and transported to the laboratory for analysis.

Exclosure experiment

We tested the effect of round goby predation relative to other predators on Dreissena in four coastal lakes (Table 1). In each lake, a cement brick colonized by Dreissena (incubated for 48–55 days) was randomly assigned to a 6-mm-mesh exclosure, 24-mm-mesh exclosure, or no-mesh control (six replicates/lake = 18 bricks/lake). Exclosures had mesh on all sides (29.5 × 18.0 × 9.5 cm³), whereas there was no mesh associated with controls. Exclosures and controls were interspersed about 20 cm apart in a grid arrangement on the substrate. Mesh was cleaned every 3–5 weeks on exclosures to remove silt, algae, and macrophytes. The study was initiated on 6–13 June 2005 and concluded after 72–86 days. Mean initial densities of Dreissena in lakes ranged from 0 to 13.3 individuals/brick, and mean size of individuals in lakes ranged from 6.9 to 10.1 mm (Table 1). As noted for the enclosure experiment, the response of Dreissena densities on bricks reflects the effect of predation on adult Dreissena already on bricks as well as veliger colonization during the experiment.

We assumed that the 6-mm mesh excluded all predators of Dreissena because round gobies small enough to pass through the 6-mm exclosures (i.e., individuals < 3.8 cm TL) were likely too small to feed extensively on Dreissena (Ray and Corkum 1997). The 24-mm exclosures allowed round gobies (≤15.0 cm TL) access to bricks while excluding larger predators (i.e., fish, mammals, birds, and turtles that were unable to pass through the 24-mm mesh). Round gobies too large to physically pass through the 24-mm mesh are seldom encountered in and around our study sites (Clapp et al. 2001; Ruetz et al. 2007; Cooper et al. 2007, 2009). We assumed that any predation on Dreissena in the 24-mm-mesh exclosures was primarily due to the round goby because other small-bodied fishes, such as johnny darter (Etheostoma nigrum) or logperch (Percina caprodes), were unlikely to consume Dreissena (French and Jude 2001) and we encountered very few crayfish during our field work (CRR, personal observation). Based on those assumptions regarding our predator exclusions, we tested five ecological effects (Table 2; see ‘Statistical Analysis’ section for description of ecological effects).

Table 2. Ecological interpretations of differences in Dreissena densities in the exclosure experiment. Treatment levels were no-mesh controls (N), 24-mm exclosures (L), and 6-mm exclosures (S). We assumed that S excluded all predators of Dreissena and L excluded all predators of Dreissena except the round goby (see Materials and methods section for further discussion).

Effects	Predicted differences
No predation	N = L = S
Cage (no predation)	N > L > S
Round goby	N = L < S
Large predator and round goby	N < L < S
Large predator	N < L = S

We tested the assumption that round gobies could move through the 24-mm mesh by placing a round goby (range = 11.2–14.8 cm TL) in a 37.9-L aerated tank and visually observing the tank periodically to determine whether the fish had entered a 24-mm exclosure containing a brick (cage and brick were same as used in field experiment). Trials (n = 20) were run for a maximum of 21 h with a new fish each time and were concluded the first time a round goby was observed inside the 24-mm exclosure (i.e., fish had passed through mesh to enter exclosure).

Sample collection and processing

In the field, Dreissena were scraped from each brick onto a 500-µm sieve and preserved in 70% ethanol. In the laboratory, Dreissena were enumerated, measured, and weighed. A large proportion of Dreissena was small and could not be identified to species. For samples with >100 Dreissena, we sub-sampled to estimate density (number/brick) and TL. Sub-samples contained at least 30 Dreissena, which were used to estimate their TL in millimeter and ash-free dry mass (AFDM) in grams. We measured TL of Dreissena shells with dial calipers from the umbone to the posterior margin along the ventral hinge. To estimate AFDM, Dreissena were oven dried at 55°C for at least 24 h until a constant mass was reached and ashed in a muffle furnace at 550°C for about 20 h. All samples were cooled in desiccators prior to weighing after drying and ashing.

Statistical analysis

We used one-way analysis of variance (ANOVA) to assess the effect of round goby density (0, 5, 10, 15 fish/m²) on Dreissena abundance for the enclosure experiment. The response of Dreissena was measured as mean density (expressed as numbers and biomass) and length on three bricks in a cage (i.e., cage was the experimental unit). When a significant difference was detected, we made planned comparisons using orthogonal contrasts (Montgomery 1991) to test for an overall round goby effect (0 vs. 5 + 10 + 15 fish/cage) and to test whether the effect on Dreissena differed between the low and medium–high densities of round gobies (5 vs. 10 + 15 fish/cage) and between the medium and high densities of round gobies (10 vs. 15 fish/cage). Additionally, we examined whether the instantaneous growth rates (see Ruetz et al. 2006) of round gobies were a function of the initial round goby biomass stocked in a cage using linear regression. We expected that growth rates of round gobies would decrease with increasing fish biomass in a cage.

To examine the effects of excluding predators from Dreissena (Table 2), we used a nested, two-factor ANOVA to test whether Dreissena densities differed among lakes (Muskegon, Pentwater, White, and Kalamazoo) and predator exclusions (no-mesh control, 6-mm exclosure, and 24-mm exclosure) nested within lakes. Based on the findings of the enclosure experiment (i.e., statistical analyses based on Dreissena numbers and biomass did not change general findings, and length of Dreissena did not respond to round goby densities), we only examined the response of Dreissena numbers for this experiment. When a significant predator-exclusion effect was detected, we made planned comparisons (no-mesh controls vs. 6-mm exclosures, no-mesh controls vs. 24-mm exclosures, and 6-mm vs. 24-mm exclosures) within each lake using contrasts to test for differences in Dreissena density. We used these planned comparisons to test for five ecological effects (Table 2): no predation (predators did not affect Dreissena densities), cage (mesh size affected Dreissena colonization in the absence of a predator effect), round goby (predation by round gobies reduced Dreissena densities), large predator and round goby (predation by both large predators and round gobies reduced Dreissena densities), and large predator (predation by large predators – but not round gobies – reduced Dreissena densities). When using ANOVA in both experiments, we used log₁₀ transformations on Dreissena density to correct for heteroscedasticity.

Results

Enclosure experiment

The presence of round goby significantly reduced the total number (F _3,8 = 15.84, p = 0.001; Figure 1a) and biomass (F _3,8 = 4.71, p = 0.035; Figure 1b) of Dreissena. Cages without round goby had significantly higher numbers and biomass of Dreissena than cages with round goby (contrasts, p < 0.008), whereas total numbers and biomass of Dreissena did not significantly differ among enclosures with 5, 10, and 15 round gobies (contrasts, p > 0.150). The mean size of Dreissena did not differ significantly within the gradient of round goby densities (F _3,8 = 2.75, p = 0.113; Figure 1c). Finally, instantaneous growth rates of round gobies were negative and declined with the initial total biomass of round gobies in cages (slope: p = 0.022, R ² = 0.549; Figure 2), suggesting that resources (e.g., food or shelters) were limited and intraspecific competition increased at higher round goby densities. In total, 40 round gobies died during the experiment (and were replaced), with 85% of the mortalities occurring during the first 7 days of the experiment.

Figure 1. Response of mean (a) number, (b) biomass, and (c) length of Dreissena to a gradient of round goby densities in enclosures. Error bars represent ± 1 SE. Means with different letters indicate significant differences (p < 0.05).

Figure 2. Relationship between mean instantaneous growth rates (±1 SE) of round gobies and initial biomass of round gobies stocked in an enclosure. Each point represents one cage.

Exclosure experiment

Excluding round goby from bricks did not reduce the numbers of Dreissena in any of the study lakes. Total Dreissena densities were greatest in Muskegon Lake and lowest in White Lake (Figure 3). Bricks incubated in Kalamazoo Lake were not colonized by Dreissena, which could be related to differences in water quality among lakes (Table 1), and were excluded from this analysis. Dreissena densities differed significantly among lakes (F _2,45 = 735.02, p < 0.001) and across exclosure treatments nested within lakes (F _6,45 = 7.56, p < 0.001). Muskegon Lake was the only lake where we observed evidence of a predator effect (Table 2); large predators appeared to decrease Dreissena densities (Figure 3a). Dreissena densities in no-mesh controls were significantly lower than in 24-mm (contrast, F _1,45 = 24.95, p < 0.001) and 6-mm exclosures (contrast, F _1,45 = 17.58, p < 0.001) and Dreissena densities did not differ significantly between the 24-mm and 6-mm exclosures (contrast, F _1,45 = 0.64, p = 0.427). In Pentwater Lake, Dreissena densities in no-mesh controls did not differ significantly from 24-mm (contrast, F _1,45 = 0.97, p = 0.330) or 6-mm exclosures (contrast, F _1,45 = 2.36, p = 0.131), although Dreissena densities were significantly lower in 6-mm compared with 24-mm exclosures (contrast, F _1,45 = 6.36, p = 0.015; Figure 3b). However, the direction of this effect was not clearly associated with any of our ecological hypotheses (Table 2). In White Lake, Dreissena densities in no-mesh controls did not differ significantly from 24-mm exclosures (contrast, F _1,45 = 0.40, p = 0.531) but were significantly higher than 6-mm exclosures (contrast, F _1,45 = 5.70, p = 0.021; Figure 3c). Similarly, Dreissena densities in 24-mm exclosures were significantly higher than 6-mm exclosures (contrast, F _1,45 = 9.12, p = 0.004; Figure 3c). The direction of this effect was not clearly associated with any of our ecological hypotheses (Table 2) but could be due to a weak cage effect in which 6-mm exclosures reduced the colonization of Dreissena.

Figure 3. Mean density (±1 SE) of Dreissena in (a) Muskegon, (b) Pentwater, and (c) White lakes. All predators had access to Dreissena in no-mesh (none) controls, whereas all predators were excluded from the 6-mm (small) exclosures. Only large predators were excluded from 24-mm (large) exclosures (i.e., round goby had access to Dreissena). Means with different letters indicate significant differences (p < 0.05). Bricks in Kalamazoo Lake (not shown) were not colonized by Dreissena.

We found that round gobies ≤ 14.3 cm TL were able to pass through the 24-mm mesh in laboratory trials, demonstrating that round goby behavior should not prevent them from entering the 24-mm exclosures used in field experiments. Only three round gobies (among 20 individuals) did not pass through the mesh during a trial, and they were ≥ 14.6 cm TL. For round gobies that passed through the mesh, the fish was observed inside the 24-mm exclosure 2.01 h (SE = 1.13) after the trial began on average.

Discussion

Our results in conjunction with evidence on round goby diets (Cooper et al. 2009; Ruetz et al. 2009) suggest that round goby predation effects on Dreissena density are not uniform across coastal areas of eastern Lake Michigan. In our enclosure experiment, the presence of large round gobies (i.e., 5, 10, and 15 individuals/m²) significantly decreased Dreissena abundance in Muskegon Lake. However, based on an exclosure experiment, round goby predation did not significantly affect Dreissena density at any of the three coastal lakes we investigated, although we did find evidence that large predators reduced Dreissena density in Muskegon Lake. There was no evidence of a large predator effect in either Pentwater or White lakes.

Our enclosure experiment clearly demonstrated that large round gobies are capable of reducing Dreissena densities, which is consistent with previous investigations in southern Lake Michigan (Kuhns and Berg 1999; Djuricich and Janssen 2001), western Lake Michigan (Lederer et al. 2006, 2008), and eastern Lake Erie (Barton et al. 2005). Similarly, previous investigations on fish predation in areas without the round goby showed negative effects on Dreissena density (Thorp et al. 1998; Bartsch et al. 2005). In our experiment, round goby growth rates were negatively associated with initial stocking biomass in enclosures (Figure 2), suggesting that food was a limiting resource in cages. This, in turn, may indicate why we did not detect differences in the strength of predation at low (5 individuals/m²) and high (15 individuals/m²) densities of round gobies (Figure 1). We used round goby densities of 5–15 individuals/m² in enclosures, which are not uncommon in the Great Lakes (Ray and Corkum 2001; Barton et al. 2005; Johnson et al. 2005) with densities as high as 133 individuals/m² reported in areas of Lake Michigan (Chotkowski and Marsden 1999). However, the size and density of round gobies in cages were likely greater than round gobies inhabiting littoral habitats in Muskegon Lake given the size and number of round gobies collected in fyke nets (Cooper et al. 2007, 2009; Ruetz et al. 2007).

In the exclosure experiment, there was not a significant round goby effect on Dreissena density in any of the coastal lakes we studied. The results from our exclosure experiment contrast with other studies, including our enclosure experiment, which showed a strong effect of round goby predation on Dreissena (Kuhns and Berg 1999; Barton et al. 2005; Lederer et al. 2006, 2008). There are several possible factors that may contribute to this lack of effect, including statistical power, round goby size, habitat type, as well as predator and prey densities. We discount the likelihood that the lack of a predation effect by round goby on Dreissena was caused by low statistical power because the trends in Dreissena density are not in the direction expected (i.e., higher Dreissena densities in small-mesh relative to large-mesh exclosures; Figure 3) if the round goby had a strong effect on Dreissena density.

The size structure of round goby populations in littoral areas where we conducted exclosure experiments may be an important mechanism underlying the lack of a strong predation effect of round goby on Dreissena. The round goby is an aggressive, territorial fish (Jude and DeBoe 1996; Johnson et al. 2005; Stammler and Corkum 2005). Large adult round gobies prefer an environment with cover; therefore, areas with rocks and riprap are ideal habitat (Jude and DeBoe 1996; Ray and Corkum 2001). Younger, smaller round gobies often are forced out of these ideal habitats to live in less complex habitats (Ray and Corkum 2001; Johnson et al. 2005; Stammler and Corkum 2005). Diets of small round gobies are comprised of mostly non-mussel invertebrates, whereas diets of larger (>7.5 cm) round gobies often contain a large proportion of Dreissena (Ray and Corkum 1997; Skora and Rzeznik 2001; Barton et al. 2005; Campbell et al. 2009). This shift in diet with size is important because round gobies found near our exclosure sites had mean standard lengths of 5–6 cm (Cooper et al. 2007, 2009), whereas round gobies used in the enclosure experiment were >7.5 cm TL. Therefore, many of the round gobies living in the littoral habitats associated with the exclosure experiment may have not yet shifted their diets from non-mussel invertebrates to Dreissena, which could explain the lack of a strong round goby predation effect on Dreissena density in our exclosure experiment.

We suspect that the littoral habitats where we conducted exclosure experiments and feeding preferences of round goby may be additional factors explaining our finding that excluding the round goby did not affect Dreissena densities. The site in each coastal lake where we conducted the exclosure experiment had SAV and was found to have high numbers of round gobies compared with other types of littoral habitats with bare substrate or mono-dominant stands of Nuphar (Cooper et al. 2007, 2009). However, the site in each coastal lake where we conducted the exclosure experiment lacked preferred rocky substrate where both round goby and Dreissena are likely more abundant (Jude and DeBoe 1996). Although round goby diets were dominated by non-mussel invertebrates (mainly dipteran larvae and zooplankton) in coastal lakes connecting to Lake Michigan (Cooper et al. 2009), Dreissena are likely an important component of the diets of round gobies inhabiting areas with hard substrates in coastal lakes (e.g., riprap in channels connecting coastal lakes to Lake Michigan; Ruetz et al. 2009). Round goby often prefer non-mussel invertebrates over Dreissena in laboratory experiments (Diggins et al. 2002; Bauer et al. 2007; Polačik et al. 2009; Coulter et al. 2011), suggesting that densities of non-mussel invertebrates in the littoral habitats also could be an important predictor of the strength of predation by the round goby on Dreissena.

In Muskegon Lake, there was a large predator effect on Dreissena density. There are several fishes in Muskegon Lake (Ruetz et al. 2007; Bhagat and Ruetz 2011) that have been known to eat Dreissena such as the freshwater drum (Aplodinotus grunniens), pumpkinseed (Lepomis gibbosus), redhorse suckers (Moxostoma spp.), and common carp (Cyprinus carpio) (Tucker et al. 1996; Thorp et al. 1998; Magoulick and Lewis 2002; Andraso 2005; Bartsch et al. 2005). Large-bodied molluscivorous fishes were implicated in limiting Dreissena densities in a Lake Erie costal wetland (Bowers and de Szalay 2007) and in the Upper Mississippi River (Bartsch et al. 2005), although there was spatial variation in the magnitude of the effect in the Upper Mississippi River. In addition, there are several non-fish predators of Dreissena in Muskegon Lake, including the map turtle (Graptemys geographica; Lindeman 2006; Bulté and Blouin-Demers 2008), muskrat (Ondatra zibenthicus; Sietman et al. 2003), crayfish (Orconectes spp.; Martin and Corkum 1994; Perry et al. 1997), and possibly waterfowl (Hamilton et al. 1994; Badzinski and Petrie 2006). Although all these large-bodied molluscivores had access to the no-mesh exposed brick in Muskegon Lake and could have consumed Dreissena, our personal observations suggest that predation effects were caused by large-bodied fishes.

We conclude that the predation effect by the round goby on Dreissena density is spatially heterogeneous across the aquatic landscape. We hypothesize that habitat type is critical for predicting the effect of round goby predation on Dreissena density. For example, in areas of SAV in coastal lakes of eastern Lake Michigan, round goby are not likely to strongly reduce Dreissena densities because of the size structure of round gobies (i.e., more small individuals) and round gobies are likely consuming energy rich non-mussel invertebrates rather than Dreissena as their dominant food source (Cooper et al. 2007, 2009; Ruetz et al. 2007, 2009). Our hypothesis should apply beyond eastern Lake Michigan and should be tested in other areas in North America and Europe where the round goby is invading.

Acknowledgments

The authors thank Matt Cooper, Matt Breen, Aaron Parker, and Dana Strouse for assistance with fieldwork. Kristin Thomas shared ideas that helped shape the discussion. Matt Altenritter and Yakuta Bhagat provided helpful suggestions on an earlier version of this manuscript. MRR was partially supported by the Herbert L. VanderMey Internship. This research was funded by the generous support of Mr. Allen D. Hunting.