Introduction

Carbon flow is one of the main ecosystem functions, which is ultimately dependent on the availability of basal resources and the degree of their coupling to the upper trophic levels of the food web. Extensive changes in primary producers, such as those resulting from plant invasions, have the potential to disrupt food web interactions and affect the quality of carbon input. Only a few studies have examined this possibility (Gordon, 1998; Ehrenfeld, 2003), most of them conducted in terrestrial and estuarine habitats. Since aquatic plant additions were predicted to have far-reaching effects on aquatic ecosystems (Lodge et al., 1988), it is surprising that little attention has been given to food web impacts of invasive macrophytes in freshwater habitats (but see Kelly & Hawes, 2005). To address the possibility of ecosystem impacts of invasive macrophytes, we examined whether trophic structure in freshwater lakes is affected by a common invader, Eurasian watermilfoil (Myriophyllum spicatum L.).

Ecosystem-level changes are expected to be most pronounced when an invader is discretely different from native members of the community (D’Antonio & Hobbie, 2005). There are several reasons to expect that M. spicatum invasion may lead to changes in the trophic structure. First, M. spicatum possesses allelopathic polyphenolic compounds (Gross et al., 1996; but see Glomski et al., 2002), which may affect productivity and composition of the periphyton community as well as the quality of detritus after macrophyte senescence (reviewed in Ervin & Wetzel, 2003). Changes in epiphytic communities may involve changes in productivity and sloughing rates and changes in nutrient levels. Such changes affect stoichiometric ratios of carbon-to-nitrogen, which have been shown to have critical importance to carbon processing pathways (Elser et al., 2000; Wetzel, 2001). Macroinvertebrate communities could in turn be affected by the changes in the nutritional quality of epiphyton and detritus, more than by the invasive plant itself, because many resident macroinvertebrates are epiphyton grazers but few consume macrophyte tissue directly.

There is equivocal evidence for the effects of invasive watermilfoil of macroinvertebrate communities, with some studies reporting negative effects on abundance and changes in community composition (Keast, 1984), and others not finding noticeable negative effects (Cyr and Downing, 1988; Cheruvelil et al., 2000; Kovalenko et al., 2010). This discrepancy could be due to the interaction of several factors, including habitat complexity (Kovalenko et al., 2010) and edge effects (Sloey et al., 1997), in addition to the quality and amount of epiphyton. Fish and macroinvertebrate interactions and habitat use are sensitive to habitat complexity (Crowder & Cooper, 1982; Warfe & Barmuta, 2004; Kovalenko et al., 2010) and could be affected by the different structural complexity of M. spicatum. Previous studies show that fish foraging efficiency is often reduced in complex habitats (Diehl, 1988; Valley & Bremigan, 2002), yet fish are observed in M. spicatum stands, indicating a possibility that it may be used primarily as a structure, whereas foraging may be dependent on the remaining patches of native macrophytes (Kovalenko et al., 2009).

To understand the potential effects of invasive macrophytes on aquatic food webs, we compare trophic structure in lakes with and without dominant Eurasian watermilfoil by mapping stable isotope ratios of most common consumers and basal resources. Nitrogen stable isotope ratio, 14N/15N, increases with trophic position due to fractionation (Fry, 1991). Carbon isotope ratio, 12C/13C, is determined primarily by the type of photosynthesis in the case of terrestrial and emergent plants, and by the source of inorganic carbon and thickness of boundary layer in submerged macrophytes (Keeley, 1990; Boutton, 1991). Carbon isotopes undergo minimal fractionation (McCutchan et al., 2003) throughout the food web, and can be used to distinguish the proportional importance of different basal resources (Fry, 2006). The dual-isotope approach is frequently used to document the changes in carbon flows and food web structure (e.g., Herwig et al., 2004; Hoeinghaus et al., 2007). In a previous study, we examined the effects of invasive watermilfoil on the trophic niche width of secondary consumers (Kovalenko & Dibble, 2011), whereas here we report data on the entire littoral food webs in four lakes to trace potential importance of the invasive macrophyte to lacustrine carbon flows. We examine food web structure using basal resources, primary and secondary consumers (invertebrates and fish). Our hypothesis is that Eurasian watermilfoil-associated production is not incorporated into the macroscopic foodweb, resulting in differences between energy flows in native and invasive macrophyte stands. Alternatively, consistent incorporation of the dominant macrophyte-associated production would be detected at all levels of the food web in both native and invasive stands. In addition, we make a qualitative comparison of food webs in invaded and non-invaded lakes to see whether potential stand-related differences translate into food web changes on the lake level. To examine a possible mechanism for the hypothesized inferior quality of M. spicatum, we compare the carbon-to-nitrogen stoichiometric ratios of the basal resources.

Materials and methods

Sampling sites

We studied the effects of invasive watermilfoil in four lakes in Minnesota, USA, located between 44°49′-53′N and 93°22′-41′W. The lakes were selected based on similar trophic status (moderately eutrophic) and bathymetry (Skogerboe & Getsinger, 2006; 66–106 ha, approximately 50% littoral area to overall lake area, 9–18 m maximum depth), but differed in abundance of invasive watermilfoil. Watershed land-use was primarily medium-density residential (see SI Table 1 for additional information about the study lakes). Two of the lakes (Auburn, Pierson) had large stands of invasive Eurasian watermilfoil, which had overtaken approximately 50% of the littoral area (Kovalenko et al., 2010), with nearly 100% cover in some areas (hereafter, M. spicatum stands). These two lakes also had uninvaded native plant stands, which allowed within-lake comparison of the contribution from different plant sources. Following selective M. spicatum control in 2004–2006 (Skogerboe & Getsinger, 2006), lakes Bush and Zumbra were dominated by native plants, including Nymphaea odorata, Ceratophyllum demersum, Schoenoplectus sp., and Vallisneria americana. Fish community structure was similar in all study lakes, with predominance of centrarchids, primarily bluegill (Lepomis macrochirus), and largemouth bass (Micropterus salmoides, the most abundant top predator, Kovalenko et al., 2010).

In each of the invaded lakes, we randomly selected several sites within invasive watermilfoil and mixed native macrophyte stands. A site was defined as a general area of approximately 20-m diameter, and sites were far removed from one another (hundreds of meters). Invasive M. spicatum stands were large (minimum 50 m in shoreline length) monospecific beds, not containing any native macrophytes. Mixed native macrophyte stands in invaded lakes contained the same plants listed above for the uninvaded lakes. In the other two lakes, all samples were collected from randomly selected sites in native macrophyte stands. All sites were located at the same distance from the shoreline and the same depth (0.4–0.6 m). At each site, samples were taken to represent basal resources as well as primary and secondary consumers, as detailed below.

Basal resource, macroinvertebrate, and fish sampling for isotope analysis

Only young macrophyte tissue (i.e., apical shoots) was sampled for macrophyte source analysis to standardize potential age effect and reduce interference from epiphytes (e.g., Fry, 2006). Plants were rinsed to ensure the removal of attached biota and inorganic deposits, and examination under a dissecting microscope confirmed absence of marl or epiphyton on young shoots. For epiphytic source, loose epiphyton was gently scraped off the surface of older plant leaves and stems. Particulate detritus was sampled by resuspending sediment samples to allow settling of coarse inorganic particles while retaining the lighter fraction, and then removing living organisms and large particles (visible pieces of leaves, stems) by hand. All components of the food web were sampled at the same time, except for fish which were collected at night at the same sites, and a minimum of three samples was taken for each compartment-by-site combination in each treatment.

In each macrophyte stand, invertebrates were collected using a dipnet and sorted in the field. Of several types of primary consumers, Amphipoda (Hyallella azteca Saussure) and Chironomidae were collected in each lake and in sufficient quantities to allow replicated analysis (several to several dozens of individuals were pooled for each sample). We sampled the most common predatory invertebrates: nymphs of damselflies (Enallagma sp., Coenagrionidae; Lestes sp., Lestidae) and dragonflies (Erythemis and Leucorrhinia spp., Libellulidae). Macroinvertebrates were placed in individual vials containing lake water and left in the shade to purge for 2–3 h after which they were rinsed, blotted, and frozen on dry ice.

Fish isotope samples were obtained from the dorsal muscle of fish collected using night boat electrofishing using pulsed DC at 250–350 V and 6–10 A. Fish were measured to the nearest mm and size-selected to minimize the effects of ontogenetic niche shifts (see Kovalenko & Dibble, 2011). Total length of largemouth bass was 13.3 ± 3.1 cm (SD); this size group was chosen because preliminary assessment showed that it was possible to capture sufficient numbers (n ≥ 3) of similar-sized fish at each site, in all four lakes. Only adult bluegill (>8 cm) was kept for the analysis; isotopic variation within this size group was minimal. Fish stomachs were preserved in 99% ethanol and stomach contents were later analyzed under a dissecting microscope to confirm validity of inferred carbon flow pathways (see SI Table 2 for details). All samples were collected in early September 2008 and were kept frozen until analysis.

Samples were dried in the oven at 60°C to a constant weight (usually around 24 h) and homogenized using a mortar and pestle. Epiphyton samples were dried on tinfoil and homogenized in small centrifuge tubes using glass probes. Isotope mass spectroscopy analysis was done by the Colorado Plateau Stable Isotope Laboratory. Isotope ratios were expressed using standard notations as a ratio between the sample and a standard (Vienna Pee Dee Belemnite standard for 13C and air for 15N). Lack of carbonate deposit interference in detritus and macrophyte tissue was confirmed by analyzing acid-treated subsamples, which were not different from untreated tissues (paired t test t = −0.69, n = 3, P = 0.56). All samples included in further analyses were not acid-treated. Analytical precision was calculated by averaging standard deviation of 30 samples analyzed in duplicate, and was 0.07% for δ13C and 0.06% for δ15N. Data were not lipid-corrected as C:N ratios were low and consistent across compartments (fish 3.19 ± 0.04 SD; invertebrates 4.18 ± 0.41).

Data analyses

Stable isotope ratio data were plotted in δ13C–δ15N space for comparison of trophic pathways between stands with and without invasive macrophytes and among lakes. Differences among basal resource isotopic values were analyzed using one-way ANOVAs with Bonferroni correction for multiple tests. Differences between corresponding consumer compartments in native and invasive stands were analyzed using directional statistics (Schmidt et al., 2007). Change between native and invasive consumer isotope ratios was expressed in directional degrees in the biplot space for each of the consumer compartments and analyzed using Rao’s spacing test (Batschelet, 1981). Relative contribution of native and invasive macrophytes and their epiphyton and detritus to the diet of bluegill and predatory invertebrates was analyzed using a Bayesian mixing model MixSIR (Moore & Semmens, 2008; Semmens et al., 2009), with 108–109 iterations as needed to obtain a robust solution. The standard mixing model determines isotopic composition of the mixture according to the following formula:

$$ {{\updelta}}_{M} = \mathop \sum \nolimits f_{i} \times \left( {{{\updelta}}_{i} + {{\upgamma}}_{i} } \right), $$

where f i is the proportional contribution of each considered basal resource, δ i is the isotope value of each source, and γ i is the isotope-specific fractionation (Moore & Semmens, 2008). The MixSIR models integrated across sources of uncertainty by incorporating consumer individual variability as well as fractionation uncertainty and standard deviation of source signatures (see Moore & Semmens, 2008 for details). M. spicatum was merged with its epiphyton in lake Auburn to decrease the number of sources, because the isotopic composition of the two was not statistically different. A trophic shift of 0.5 ± 0.17‰ per trophic level was used to adjust isotopic values for carbon (McCutchan et al., 2003). For nitrogen adjustment, we used 1.8 ± 0.92‰ per trophic level for invertebrate predators, which is an average between consumers raised on invertebrate and plant and algal diets, (McCutchan et al., 2003). For the fish, we used 2.8 ± 0.85‰, based on values reported in literature and trophic shifts for consumers feeding on invertebrate, high protein, and algal diets (Vander Zanden et al., 1997; McCutchan et al., 2003). We used basal resources rather than first-order consumers as model sources to directly address the question of proportional contribution of the different plant sources. Using basal resources is particularly sensitive to our ability to accurately account for the trophic position; therefore, we checked for stability of solution by incorporating a wide range of fractionation uncertainty and by removing δ15N altogether. Because sources were not significantly different in δ15N, its contribution to the final solution was negligible. For stoichiometric analysis, C:N mass ratios of respective basal resources were compared using t tests.

Results

Site-level comparisons in invaded lakes

Myriophyllum spicatum δ13C values were significantly different from those of native plants (F 3, 13 = 64.8, P < 0.001; post hoc comparisons: P < 0.001 for all between-plant comparisons except coontail, which was marginally different at P = 0.056; for Auburn lake; F 2, 11 = 137.3, P < 0.001 for Pierson lake; see SI Table 3 for actual isotope values and SD). M. spicatum detritus also tended to be less depleted than detritus of native plant stands; these differences were significant for Pierson (t = 10.3, n = 7, P < 0.001), but could not be tested for Auburn due to a lost sample, resulting in low power. M. spicatum epiphyton was significantly less δ13C depleted than epiphyton of native plants (t = 148, n = 7, P < 0.001 for Auburn and t = 2.47, n = 8, P = 0.043 for Pierson, the latter not significant after the Bonferroni correction).

In both invaded lakes, there was a clear separation between food webs in native and invasive macrophyte stands (Fig. 1a, b), supporting our expectation of stand-related differences in the energy pathways. Consumer isotopic ratios exhibited a significant directional change (Rao’s U = 250, n = 6, P = 0.004 and U = 238, n = 5, P = 0.006, for Auburn and Pierson, respectively). In both M. spicatum-dominated lakes, δ13C values of macroinvertebrate consumers were similar to δ13C values of epiphyton or detritus in the respective plant stands (Fig. 1a, b). However, fish collected both in M. spicatum and native stands had δ13C values that matched those of native plant-associated production. Although native coontail was marginally different from M. spicatum in δ13C, this plant was not observed in M. spicatum stands and had negligible contribution to food webs in the native stands (see below).

Fig. 1
figure 1

Food web structure mapped using stable isotopes δ13C and δ15N in the invaded lakes a Pierson and b Auburn, and native plant-dominated lakes (c, d). Basal resources are represented by squares (epi epiphyton, det detritus, coon coontail, mil milfoil, bul bulrush, eelgrass Vallisneria americana), invertebrate consumers by triangles (chi chironomids, amphi amphipods, zyg Zygoptera, damselfly nymphs, Ani Anisoptera, dragonfly nymphs), and fish by diamonds. Filled symbols represent compartments sampled in invasive plant-dominated stands. The hypothetical dashed line delineates approximate extent of the food web in the invasive stands. Asterisk due to lost sample, the inter-lake average value was used. Values are mean ± SE, N ≥ 3 for all organisms except Chironomidae, for which samples had to be pooled further to get sufficient weight for isotope analysis. See Supplemental Table 3 for values

Full size image

Bayesian mixing model analysis demonstrated low contribution of M. spicatum, its epiphyton and detritus to carbon flows to the upper trophic level (Fig. 2a, b; 1–99th percentile ranges 0–20% and 0–9.8% for Auburn and Pierson, respectively). The low ranking of milfoil-derived sources was unchanged in fish from invasive stands (Fig. 2 a, b lower left panels). In both native and invasive stands, emergent and floating macrophytes and their epiphyton and detritus made the greatest contribution to bluegill diets (1–99th percentile ranges 46–84% for native epiphyton and 3–35% for native emergent plants, for Pierson; 35–60% for native floating macrophyte for the other invaded lake, Fig. 2a). Contrary to the situation with fish, for predatory invertebrates, ranking of sources was altered in the invasive stands with significant contribution of milfoil-associated production.

Fig. 2
figure 2figure 2

Bayesian mixing model analysis of the relative contribution of native and invasive macrophyte-associated carbon sources to diet of bluegill and Odonata in M. spicatum-dominated lakes a Auburn and b Pierson. Increasing proportion of the “Source contribution to the mix” indicates more likely/more significant contribution of a given source. Note that the primary role of native plants and their epiphyton, and similarity of source contributions between the invertebrate part of the food web (integrated by Odonata) and the most abundant fish (bluegill) in the native stands and uninvaded lakes, but lack of this correspondence in the invasive stands

Full size image

Myriophyllum spicatum epiphyton C:N ratio was 40% greater than that of epiphyton of native plants (F 1, 11 = 58.8, P < 0.001, Fig. 3). In general, epiphyton of native plants was more nitrogen-enriched (lower C:N ratios) than the plant material itself, whereas M. spicatum epiphyton was more nitrogen-depleted. No significant differences were detected in the detritus from native versus invaded lakes (F 1, 9 = 1.3, P = 0.293).

Fig. 3
figure 3

Carbon-to-nitrogen ratios of epiphyton and detritus in invasive- and native plant stands (see SI Table 4 for raw data). Values are mean ± SE, asterisk indicates a significant difference (P < 0.05)

Full size image

Among-lake comparisons

Food webs in lakes with invasive M. spicatum were different from lakes dominated by native plants, because uninvaded lakes had no equivalent of δ13C-enriched invertebrate communities (Fig. 1c, d). In lakes without invasive plants, both invertebrate and fish predators were supported primarily by native floating macrophyte-associated production and detritus (SI Fig. 1). Fish dietary analysis for both within- and across-lake comparisons confirmed validity of the benthivorous pathway, because stomachs of both bluegill and largemouth bass contained primarily Chironomidae, Amphipoda, and Odonata and only 7% of bluegill contained non-negligible amount of zooplankton (SI Table 2).

Discussion

Effects of invasive macrophyte

Our expectation of the different pathways of energy flow in the native and invaded stands was supported as diets of fish collected in invasive plant stands were derived from native macrophyte-associated production, whereas macroinvertebrates had δ13C-enriched isotopic values in M. spicatum stands. This indicates that M. spicatum-associated production is partially incorporated into the food web, but its contribution is not detected at upper trophic levels. In other words, we observed a replacement of native with non-native carbon sources with decreasing abundance of native macrophytes for invertebrate consumers but not for the fish, indicating potential uncoupling of energy flows (i.e., M. spicatum signature was detected in lower but not higher trophic levels). This finding also implies that upper trophic levels in the littoral area may be dependent on continuing presence of native macrophytes in nearby habitats. Two previous studies on M. spicatum-invaded lakes had arrived at disparate conclusions: one showed that M. spicatum was not incorporated into a reservoir food web (Toetz, 1997), whereas the other (Munyon, pers. comm., 2006) indicated its potential contribution to the food web. However, no other littoral sources were considered in either of the studies. Our study demonstrated that in natural lakes with diverse macrophyte assemblages, energy pathways are different in the invasive milfoil stands and milfoil-associated production is not incorporated into the upper trophic levels.

Negligible contribution of M. spicatum-associated carbon to upper trophic levels of the food web was supported by mixing models. It was possible to trace the importance of individual sources because they had relatively distinct carbon values. Epiphyton values were more similar to those of their respective macrophytes, which is not always the case (e.g., Keough et al., 1996). Nevertheless, a relatively high number of sources, which is to be expected in most littoral areas with diverse macrophyte communities, resulted in diffuse mixing model outcomes. However, mixing models are more powerful in disproving significance of certain sources rather than confirming high importance (Benstead et al., 2006). This quality of mixing models increases our confidence in negligible contribution of M. spicatum and its epiphyton and detritus to higher trophic levels. The minor contribution occurs despite the high potential productivity of M. spicatum and associated epiphyton (e.g., Cattaneo & Kalff, 1980; Cronin et al., 2006). In our previous work, we reported increased isotopic niche width in fish from the invaded lakes (Kovalenko & Dibble, 2011); however, as we show here, this increased trophic variability is not associated with incorporation of the 13C-enriched M. spicatum-derived production. In invertebrate predators, on the other hand, previously reported increased isotope niche width appears to be associated with at least partial incorporation of M. spicatum-associated production. When considering negligible M. spicatum contribution in light of the similar low contribution of the native submerged macrophytes, it is important to realize that M. spicatum does not simply displace native submergent beds, but also native floating and to some extent emergent macrophytes as well. Native macrophyte beds sampled in the current study were diverse heterogeneous assemblages, with floating and emergent vegetation.

This study demonstrates that invasive plants can have conspicuous effects on food web structure, as energy pathways differed between native and invasive macrophyte stands within a lake. Although, additional evidence from other systems is needed to unequivocally make this conclusion, the finding that there is limited littoral predator feeding in the invasive stands implies that invasive aquatic plants can have significant effects on littoral food webs. Furthermore, given the importance of macrophyte sources for littoral-pelagic coupling (Schindler & Scheuerell, 2002), this effect may propagate to affect entire lacustrine ecosystems. Previous research showed that invasive plants can alter ecosystem processes, such as nutrient dynamics (e.g., soil nitrogen enrichment due to invasion of terrestrial plants: Vitousek et al., 1987; Ehrenfeld, 2003), but few studies examined how these effects are transmitted to upper trophic levels. Where such studies were done, effects were often significant: for example, invasion of Phragmites australis in estuarine habitats resulted in significant changes in arthropod food webs, with shift from native plant herbivory to utilization of detrital and algal resources (Gratton & Denno, 2006). In another example, invasive macrophytes in a New Zealand lake changed energy flow to the dominant primary consumers (from fine organic matter benthivores to grazers in invasive beds) and from primary consumers to fish (Kelly & Hawes, 2005). This study demonstrated that fish inhabited invasive beds but their foraging was limited in that habitat (Kelly & Hawes, 2005). Another study demonstrated that mangroves made a substantial contribution to detrital food webs in their native, but not introduced range, where they caused a shift from algal-based food webs to the ones relying on multiple sources (Demopoulos et al., 2007). Similar to the latter study, we have previously found a wider range of basal resources utilized by consumers in invaded habitats (Kovalenko & Dibble, 2011).

Several possible mechanisms could be responsible for the lack of incorporation of M. spicatum-associated production into the upper trophic levels of the littoral area. Of the reasons discussed earlier, allelopathy appears an unlikely mechanism, since macroinvertebrates seemed to incorporate some of the invasive plant-associated production. In support of our inferior stoichiometric ratios hypothesis, we detected a small but significant difference in the relative nitrogen content of M. spicatum epiphyton in comparison with epiphyton of native plants. Due to stoichiometric imbalance of primary consumer–resource interactions (Elser et al., 2000), C:N ratios of basal resources are critical for carbon processing, and resources with high carbon-to-nitrogen content are increasingly recalcitrant (Wetzel, 2001). However, this difference was negligible in comparison with the difference between nutrient content of submerged and emergent macrophytes. Emergent macrophytes have much higher C:N ratios due to greater investment in carbon-rich supportive structures, and yet apparently contribute significantly to the overall carbon flow (see also Otsuki & Wetzel, 1974). Therefore, another possibility—interference with fish foraging due to dense, homogeneous structure—appears to be a more plausible mechanism. It was previously shown that fish foraging efficiency was reduced in dense stands of invasive plants (Valley & Bremigan, 2002; Kelly & Hawes, 2005; Rennie & Jackson, 2005; Theel & Dibble, 2008). Capture of fish inside the M. spicatum beds in this and the previous studies (Weaver et al. 1997; Kovalenko et al. 2010) suggests that watermilfoil beds are used for structure and/or refuge, but not as foraging grounds. Our previous 4-year study did not detect differences in fish abundance, biomass, or community composition related to invasive plant abundance (Kovalenko et al., 2010). Thus, due to consumer foraging adaptability (feeding in nearby habitats), it is possible that few structural changes in higher trophic levels would be observed, until native macrophytes are reduced to a low density.

Macrophyte production can be incorporated into the food web through particulate organic matter in detritus or dissolved organic matter (DOM) (e.g., Demarty & Prairie, 2009). Although, there is evidence that herbivory is at least as common in aquatic as in terrestrial systems (Lodge et al., 1997), in our study, direct contribution of macrophytes to carbon flow is unlikely, since few true herbivores were present in our samples, and contribution indicated by mixing models was most likely mediated through DOM. Dissolved organic carbon released from emergent macrophytes by autolysis accounts for larger part of lake carbon metabolism/DOM inputs (Otsuki & Wetzel, 1974). Most of this carbon is highly labile (Wetzel & Manny, 1972) and is incorporated into the epiphyton (Wetzel & Allen cited from Wetzel 2001) and can thereby be a major contributor to the macroscopic food web through grazing consumers. This DOM release would also explain why epiphyton was similar in δ13C value to its macrophyte substrate. DOM isotopic composition would be an average of several combined sources, and difficulty in attributing its source may be at the root of disagreement about the role of macrophytes in lake food webs (see studies providing support for the importance of littoral sources: e.g., Karlsson & Byström, 2005; Vander Zanden et al., 2006; Hoeinghaus et al., 2007; Mendonça et al., 2013 vs. studies demonstrating that their contribution is relatively minor: Keough et al., 1996; James et al., 2000; Hart & Lovvorn, 2003; Herwig et al., 2004). More refined techniques, such as macrophyte labeling using tracers, may be necessary to characterize DOM dynamics and energy cycling in the detrital loop. For example, Oakes et al. (2010) successfully resolved importance of microphytobenthos and mangrove detritus using in situ labeling and compartment modeling. This approach could help address the problem of continuity among epiphyton, detritus, and living macrophytes, as most macrophytes continuously senesce (Wetzel & Howe, 1999), and DOM is cycled within the macrophyte–epiphyton complex (e.g., Burkholder & Wetzel, 1990).

Sources of uncertainty

Our study focused on the littoral zone and did not take into account the pelagic component of lacustrine food webs, which we believe to be a realistic picture of energy flows because fish stomach analysis confirmed minor zooplankton contribution. In addition, our previous study of more than 600 fish collected over 4 years and two different seasons, found a similar small contribution of zooplankton to the diets of fish inhabiting the littoral zone (Kovalenko et al., 2009). The contribution of pelagic component could be different in larger lakes, with more extensive pelagic areas, and the effects of M. spicatum invasion could therefore be different. Another potential pathway for planktonic–benthic coupling, through the consumption of settled phytoplankton by benthic invertebrates, was most likely accounted for through collection of epiphytic samples. Second, concerns could be raised over the use of previously invaded lakes, as few lakes in the area did not have a history of M. spicatum invasions. However, we argue that it is unlikely that the observed differences are due to M. spicatum eradication 2 years prior to this study, because within-lake contrasts in the untreated lakes found similar food web structure in the native stands. As with all mixing model approaches, the present models accounted only for the proportional representation of the sampled portions of the food web. It is possible that we adequately portrayed the dominant energy flows in littoral macroconsumers because the two species of fish selected for this study comprised approximately 85% of the littoral fish community. However, other fish, which may use a mix of vegetated littoral and open water habitat (e.g., yellow perch, crappie, and northern pike) were not studied and the effects of invasive plants on those parts of the food web are not known. It is important to note that full-grown largemouth bass were not included in this study because of difficulties in obtaining sufficient numbers of similar-sized fish. Finally, this study was limited in its account of seasonal variation: although some temporal integration was provided by the detritus, full consideration of seasonal effects was not possible for logistic reasons. Despite these limitations, we argue that lack of higher trophic-level littoral predator foraging in invasive stands may have important implications for overall food web structure as invasive plants become increasingly dominant.

Conclusions

This study demonstrates that invasive plants can affect the food web structure in the littoral portion of the lake, possibly by uncoupling energy flow to secondary consumers. Since a previous study in the same system showed no differences in fish abundance and community composition with changes in invader abundance (Kovalenko et al., 2010); in light of present findings, it is clear that the assessment of trophic structure is necessary to understand the full scale of impacts.