Many Great Lakes coastal wetlands that remain today have been heavily fragmented by anthropogenic activities. The rate of fragmentation tends to increase during periods of low lake levels, especially in areas of low-gradient bathymetry where wetland area expands substantially and prompts the desire to dredge channels and groom shorelines. We sampled fish and invertebrates, using fyke nets and dipnets respectively, from wetland fragments paired with either areas where wetland vegetation was mowed or removed completely. Our concurrent studies showed that removal of vegetation by beach grooming and channel dredging created conduits for pelagic water to infiltrate the marsh and disrupt the ambient chemical/physical conditions. Alterations to both fish and macroinvertebrate communities were also evident where a significant amount of vegetation was removed. However, where only enough vegetation was removed to allow for boat access, impacts on fish communities were generally non-detectable. Mowing seemed to impact fish, but not invertebrates. Our data suggest that wetland fragmentation may have substantial and long lasting effects on wetland biota, but the magnitude of the impact is likely associated with the area of vegetation removed coupled with the potential for pelagic water to penetrate remaining fragments.
More than 50% of Great Lakes coastal wetlands have been lost to anthropogenic disturbance since European settlement; loss in the lower lakes is nearly 95% in some areas (Krieger et al., 1992; Cwikiel, 1998). Wetlands remaining today are heavily fragmented. Large areas have been drained for agriculture and urbanization while boat launches and navigational channels cut through remaining segments. These systems continue to be fragmented by shoreline development and by navigation of small boats to and from docks.
Fragmentation increases during low lake level years as riparian owners and developers seek to deepen channels and create new ones. Lake levels dropped by more than one meter in Lakes Michigan and Huron from 1997 through 2006 and reached near record lows in 2003. Fragmentation accelerated during this period as riparian land owners not only maintained boat channels but also mowed and removed wetland vegetation from recently exposed bottomlands lake-ward of their properties. Michigan Public Act 14 facilitated these activities by allowing mechanical removal of vegetation from Saginaw Bay, Lake Huron and Grand Traverse Bay, Lake Michigan after obtaining a letter of approval from the director of the Michigan Department of Environmental Quality (MDEQ). A variety of techniques were employed, including mowing with both small lawn and large farm tractors and removal of roots and rhizomes with tractor-pulled rakes. In addition, sand was moved and raked to create and maintain beaches, particularly in public parks, but also on some private lands. While we had not documented such activities prior to legislative changes, it seems likely that this legislative action and approval of a general permit for such activities by the U.S. Army Corps of Engineers led to major increases in fragmentation and mowing of wetlands.
The effects of habitat fragmentation have been described for many terrestrial systems (e.g. Chen et al., 1992; Aizen and Feinsinger, 1994; Essen, 1994; Diffendorfer et al., 1995; Pasitschniak and Messier, 1995; Jokimaki et al., 1998; Jules, 1998; Dale et al., 2000; McKone et al., 2000; Laurance and Yensen, 1991; Groom and Grubb, 2002; Manolis et al., 2002;), but few studies have documented the effects of fragmentation on abiotic conditions, invertebrates, or fish in lacustrine wetlands. Studies of palustrine wetland fragmentation have focused primarily on amphibians (e.g. Lann and Verboom, 1990; Findley and Houlihan, 1997; Knutson et al., 1999; Gibbs, 2000), birds (e.g. Benoit and Askins, 2002), and plants (e.g. Hooftman and Diemer, 2002; Lienert and Fischer, 2003). We are aware of only one study of fragmentation in Great Lakes coastal wetlands (Hook et al., 2001); its focus was on fish in wetlands of northern Lake Huron.
Fringing coastal wetlands occur where protection from destructive forces of wind and waves allows emergent vegetation communities to establish (Heath, 1992; Keough et al., 1999; Burton et al., 2002, 2004). The degree of wave exposure is broadly predictive of the types of vegetation, invertebrate, and fish communities occurring in a particular system (Burton et al., 2002, 2004; Uzarski et al., 2005). Once established, vegetation inhibits direct infiltration of pelagic water into a wetland and also traps organic matter. Different growth forms and stem densities inhibit pelagic water infiltration and trap organic sediments to varying degrees. Accordingly, we have observed distinct chemical and physical conditions associated with specific vegetation types and densities. Biota (e.g. fish and invertebrate communities) respond to the degree that pelagic water penetrates a wetland resulting in specific shifts in community composition from open water to shore (e.g. Cardinale et al., 1998; Burton et al., 1999, 2002, 2004; Uzarski et al., 2004, 2005). Therefore, it seemed likely that removing wetland vegetation would influence both the chemical/physical environment and faunal community composition.
Our objectives in this study were to document the effects of fragmentation and mowing on invertebrate and fish communities as well as chemical and physical conditions in fringing wetlands of Lakes Huron and Michigan. While community composition along natural gradients perpendicular to shore are predictable, it was unclear how communities would respond to partial or complete remove of parcels of vegetation. We explored these effects by making comparisons between manipulated habitats and adjacent reference habitats. The pair-wise comparisons studied and included in this manuscript represent one of several studies that we conducted on the effects of coastal wetland fragmentation.
General approach for determining effects of wetland fragmentation
We worked with MDEQ staff to identify areas where property owners had removed, or proposed to remove, vegetation from exposed bottomlands that had supported emergent plant communities when lake levels were near-normal or above average. We selected additional sites from local observations and after conversations with local land owners in an effort to represent all types of wetland manipulation that had occurred in the study regions. When evidence of wetland manipulation was found, we sought access to the site from property owners or from appropriate governmental employees in the case of public lands. If access was granted, we selected adjacent or nearby intact wetland habitat as a reference and sought permission to access that area as well. Though access was denied in some cases, we were able to include a representative sample of groomed or mowed and reference sites in the study.
At each sampling area, we compared biotic communities and chemical/physical parameters of disturbed shoreline areas (e.g. tilled, mowed, or raked) with biotic communities and chemical/physical parameters in adjacent or nearby non-disturbed reference areas. Reference sites were chosen based on their proximity to groomed sites and whether they had the same vegetation type as the associated groomed site. Both of these criteria were necessary in order to isolate chemical/physical and biotic factors that were due to the grooming practice alone. In 2004, in Saginaw Bay, we sampled eight paired sites (groomed and reference) for juvenile and adult fish using fyke nets and 11 paired sites for invertebrates using dip nets (Table 1). We also sampled three paired sites in Grand Traverse Bay with both fyke nets and dip nets (Table 1). In 2005 we sampled 12 open beaches along Saginaw Bay (5 raked and 7 unraked) paired with adjacent intact vegetation using timed dip net sampling (Table 2).
Chemical and physical measurements
Basic chemical/physical parameters were collected each time biological samples were taken. Temperature, dissolved oxygen (DO), percent saturation of dissolved oxygen (% DO), chlorophyll a, oxidation-reduction potential (redox), pH, turbidity, total dissolved solids (TDS), and specific conductance (SpC) were measured in situ using a Hydrolab Datasonde 4a. Dissolved nutrients, including soluble reactive P (SRP), NO3-N, and NH4-N, as well as Cl−, SO4-S, and total and phenolphthalein alkalinity were measured in the laboratory. Immediately after collection, water samples were filtered through 0.45 μ m Millipore filters. Analytical protocols and quality control measures followed those recommended by U.S. EPA and Standard Methods (APHA, 1998). Chemical and physical data were used to explain ecosystem function and to describe the mechanisms that were most likely responsible for structuring biological communities.
One of the most common ‘beach maintenance’ activities was mowing exposed wetland plants. Mowing either extended into shallow water or took place early in the year when Great Lakes water levels were at seasonal lows before increasing to inundate mowed areas. In 2004 we subdivided mowed sites into those that had been mowed recently (plant height 30 cm or less) and those where mowing had occurred but plants were greater than 30 cm tall at the time of sampling. Mowed sites with plants greater than 30 cm included a combination of those mowed in 2003 and those mowed early enough in 2004 for plants to have re-grown.
In 2004, macroinvertebrate samples were collected with standard 0.5 mm mesh D-frame dip nets from late July through August. Macroinvertebrates were sampled in all available inundated plant zones at each site. Dip net sweeps for each replicate sample were made through the water column at the surface, mid-depth and just above the sediment surface. Net contents were emptied into white pans and organisms were collected on site by picking all specimens from one area of the pan before moving on to the next area. As a means of semi-quantifying samples and to standardize effort, picking of specimens was timed. Individual replicates were picked for one-half person-hour or until 150 specimens per replicate had been collected. If 150 specimens had not been collected after one-half person-hour of picking, picking continued to the next multiple of 50. Three replicates were collected from each major plant zone present. This sampling method was used effectively in previous research on Great Lakes coastal wetlands (Burton et al., 1999, 2002, 2004; Uzarski et al., 2004).
Preliminary observations in 2004 indicated that groomed, open beach areas supported a very low-density faunal community compared to nearby areas where vegetation was still intact. In order to document and quantify these differences in 2005, we focused our macroinvertebrate sampling efforts on open beach areas (raked and unraked) paired with adjacent vegetated areas to serve as reference habitats. To be extremely conservative in our conclusions, we chose adjacent reference areas with very sparse vegetation since it was often argued that the groomed open beaches never supported dense vegetation. We also used a more quantitative sampling technique so that macroinvertebrate density differences were better represented and to ensure that the same unit effort was expended for each replicate sample.
Timed dip net sampling was used to sample the upper layer of substrate by pushing the net back and forth through the upper 3-5 cm of substrate for 3 minutes per replicate. Fine silt and detritus was removed from each sample by repeatedly washing the net contents with water. The entire rinsed sample was preserved in 95% ethanol and returned to the laboratory where specimens were removed from sediment and detritus using a dissection microscope. Three replicate 3-minute long sweeps were made in each zone.
Organisms were identified to lowest operational taxonomic unit, usually genus or species, and enumerated. Taxonomic keys such as Thorp and Covich (1991) and Merritt and Cummins (1996), and comparison with specimens in reference collections (previously identified by taxonomic experts and maintained in T.M. Burton's laboratory) were used for identification.
Fish were sampled in 2004 using six fyke nets per site with 3 nets set in disturbed and 3 set in reference zones for one net-night. Fyke nets are an effective fish sampling gear in Great Lakes coastal wetlands (Brazner, 1997; Uzarski et al., 2005). Site selection for fish sampling was coordinated with site selection for macroinvertebrate sampling. A number of the sites sampled for macroinvertebrates were too shallow for effective sampling of the juvenile and adult fish communities. Thus, fewer sites were included in fish community analyses in 2004 than were included for macroinvertebrates (Table 1). We also included four sites from northern Lakes Huron and Michigan in which non-vegetated boat channels were paired with zones of intact vegetation to predict impacts of these activities following the rise of Great Lakes water levels.
Two sizes of fyke nets (91 × 46 cm and 122 × 91 cm) with 5 mm mesh were used. Smaller nets were set in water < 0.50 m deep; the larger nets were set in water 0.50 to 1.0 m deep. Nets were set perpendicular to or within habitats of interest with 7.3 m leads extending from the middle of each net into the zone to be sampled. Two 1.8 m wings, connected to each side of the net opening, were set at 45° angles to the lead (Uzarski et al., 2005). Nets were set in the afternoon and retrieved the following morning. Fish captured in the nets were identified to species and enumerated on site. Voucher specimens were returned to the laboratory to confirm identification if a positive I.D. could not be determined in the field. Catches per net per night were recorded.
Only those taxa that composed greater than 1% and 5% of invertebrates caught per zone were included in analyses of community composition. Average catch (among triplicate samples per zone) was used as a measure of central tendency for each taxon in each zone. For the 2004 analyses, we used relative abundances (percent of total catch per zone). Since a more quantitative sampling technique was used in 2005, we used true mean abundance of each taxon per habitat. This raised our power of detection because differences in macroinvertebrate density among habitats were more accurately represented when the quantitative sampling technique was used.
All taxa captured were included in measures of Shannon diversity and taxon richness. All means reported include standard errors. To detect differences in Shannon diversity, taxon richness, and taxon abundances between reference and disturbed (groomed) habitats, data were subjected to Kruskal-Wallis and Wilcoxon signed-rank tests. Analyses of 2004 data were kept separate from 2005. Where appropriate, a matched pair statistical test (Wilcoxon signed-rank test) was used so that among-wetland variability was minimized and the effect of the treatment (grooming) was isolated. We decided a priori to compare the invertebrate communities among treatments by retaining α = 0.05 for all statistical analyses without correcting for the potentially increased type I error rate resulting from multiple comparisons. A Bonferroni correction could have been performed, but we chose not to do so because this very conservative technique would have greatly diminished the power of our tests and increased the potential for type II errors. Therefore, we acknowledge an increased probability that marginally significant p-values may be due to chance alone.
Macroinvertebrate community data from 2004 were also subjected to correspondence analysis (CA) to relate community composition to mowing treatment. When wetlands separated according to their degree of mowing, groups of individual taxa responsible for the separation were identified. Macroinvertebrate data from 2005 were also analyzed using CA to identify gradients in community composition that could be explained by the raking treatment. When these gradients were found, individual taxa or groups of taxa responsible for the separation were identified.
Fish community data from 2004 were subjected to correspondence analyses. To minimize ecoregional effects, separate analyses were conducted for Saginaw Bay, Grand Traverse Bay, and Northern Lakes Michigan and Huron. When sites separated based on maintenance activities, taxa responsible for the separation were identified.
Chemical and physical data
A total of 13 water quality parameters were measured in 2004. Because a number of these variables were highly correlated, we used principal components analysis (PCA) to reduce the dimensionality of the abiotic dataset and to explain differences in overall condition among the treatments. Results from these analyses were used to identify potential mechanisms structuring the biotic communities. That is, when biotic communities showed a response to a grooming treatment, the abiotic data were used to identify potential drivers of the community shift.
Results and discussion
Chemical and physical measurements
Comparison of water quality parameters among mowing treatments in 2004
Principal components analysis returned three principal components (PCs) that explained 52% of the variation in water quality. Principal component 1 loaded with SpC, alkalinity, and Cl− representing the most variation among sites based on the eigenvectors for these variables. Principal component 2 loaded with temperature, DO, and NH4-N and PC 3 loaded with TDS, SRP, and SO4-S. Plotting the three PCs suggested that abiotic conditions at less recently mowed wetlands and most recently mowed wetlands were within the range of water quality measurements found at reference wetlands (Figure 1). However, the most recently mowed wetlands slightly separated from reference and less recently mowed wetlands in PC 1 based on SpC, alkalinity and Cl−. This separation could reflect the tendency of mowed wetlands to be near roads and/or septic systems with more input from road salt and well water. The three treatments also separated slightly in PC 2, which was loaded with temperature, DO, and NH4-N (Figure 1). This separation could reflect higher DO concentrations resulting from wave action and/or higher rates of algal photosynthesis in the most recently mowed wetlands due to reduced vegetation cover and greater sunlight penetration.
Comparison of water quality parameters between open water reference, raked beaches, and adjacent intact vegetation in 2005
A total of nine water quality parameters were included in the PCA for open water reference and open water raked wetlands sampled in 2005. The first three PCs explained 68% of the variance in the water quality data (Figure 2). Principal component 1 loaded with temperature, DO, %DO, and depth, explaining the most variation among zones. Principal component 2 loaded with pH and alkalinity. Principal component 3 loaded with SpC, TDS, and turbidity being most important. Water quality at raked sites was within the range of water quality measurements found at reference sites. However, abiotic conditions among raked sites appeared more similar to one another than at the reference sites, since raked sites plotted in close proximity to one another (e.g. BTw, Cw, Tw) (Figure 2).
Raked and unraked open beach areas were also compared to adjacent areas where the vegetation had not been removed. A total of 9 water quality parameters were used in these comparisons. Three principal components explained 78.7% of the variance in water quality among these sites. Principal component 1 loaded with temperature, DO, and %DO being most important. Principal component 2 loaded with pH and alkalinity. Principal component 3 loaded with SpC, TDS, turbidity, and depth (Figure 3). With the exception of Pinconning, the main differences in water quality were not between open water and outer Schoenoplectus (bulrush) zones but were among site pairs (i.e. among site variability masked differences between vegetated and open water zones).
Effects of Mowing on Invertebrates in 2004
A total of 135 invertebrate taxa were collected from 44 wetland zones in 2004. Shannon diversity (H′) ranged from 1.78 to 2.62 per zone (overall mean = 2.20 ± 0.15) in the 16 reference wetlands, while mean taxon richness ranged from 11.7 to 25.0 (overall mean = 19.4 ± 2.0) (Table 1). In the 15 less recently mowed wetlands H′ varied from 1.47 to 2.75 (overall mean = 2.16 ± 0.22) and mean taxon richness ranged from 9.0 to 27.7 (overall mean = 18.7 ± 2.7) (Table 1). In the 13 most recently mowed wetlands H′ ranged from 1.45 to 2.71 (overall mean = 2.23 ± 0.20) and mean taxon richness ranged from 8.0 to 28.7 (overall mean = 20.7 ± 3.6) (Table 1). No significant differences in Shannon diversity (Kruskal Wallis: p = 0.44) or taxon richness (p = 0.65) were found among reference wetlands, less recently mowed wetlands, and most recently mowed wetlands.
In the CA of 2004 invertebrate data 10.27 and 9.61% of the total variance in the dataset was explained by the first and second dimensions, respectively (Figure 4). This analysis did not reveal any distinct separation of wetlands based on the degree of mowing, and coupled with the low percentage of variance explained, suggested that invertebrate communities were similar to each other in the three wetland treatments in 2004. Three reference zones from two sites, Bay Port swale, Bay Port Schoenoplectus and Pt. Au Gres Eleocharis, were identified as ‘outliers’ based on their high dimension 1 scores. To explore the data further, these sites were removed and a second CA was performed. Results again showed no distinct separation of wetlands based on degree of mowing. The mean relative abundance of dominant taxa from each treatment category are listed in Table 3.
When the 2004 invertebrate dataset was reduced to the taxa comprising > 1% mean relative abundance, 41 taxa were retained. When the dataset was reduced to taxa comprising > 5% mean relative abundance, 26 taxa were retained. Only two taxa, Hyalella (p = 0.017) and Sminthuridae (p = 0.022), were significantly different among the three treatments at the > 1% level and Hyalella was the only taxon that was significantly different among treatments at the > 5% level. Since this is fewer significant differences than would be expected by chance alone at an alpha of 0.05, we conclude that there were no detectable differences among treatments. Nevertheless, we examine the taxa that were different below in more detail.
A multiple pairwise comparison revealed that the relative abundance of Sminthuridae was statistically different among reference and most recently mowed wetlands (p = 0.013) with a higher relative abundance in the most recently mowed wetlands (overall mean = 1.37%). Specific sites, such as Pt. Au Gres Park Eleocharis/swale and Surfwood Eleocharis/Schoenoplectus had the greatest numbers of this taxon in the most recently mowed wetland zones (3.97% and 11.59% respectively). Both sites included very shallow water that may have been inundated only at high points in the seiche cycle of Saginaw Bay. Water level can vary by as much as 20 cm throughout the seiche cycle based on our previous observations.
The family Sminthuridae (common name: springtails), typically feed by shredding or collecting and gathering dead plant material and microflora. They are semi-aquatic and are typically found along margins of aquatic systems on surface film (Merritt and Cummins, 1996). According to property owners, the most recently mowed wetlands were usually dry earlier in the season, when the mowing was generally conducted. Our reference and less recently mowed wetlands included some that may not have been dry earlier in the season (water level was higher in 2004 than it had been in 2003 when some of the less recently mowed wetlands were mowed). Thus, differences in springtails among treatments may reflect a greater likelihood that more recently mowed sites had been inundated for less time on average than the wetlands included in the reference and less recently mowed treatments.
The other taxon that differed significantly among the three wetland treatments was Hyallela (scud or side-swimmer). Its mean relative abundance was significantly different between most recently mowed and less recently mowed wetlands (p = 0.006), but not between wetlands in either mowing treatment and reference wetlands. Mean relative abundance of Hyallela was greatest in the most recently mowed wetlands (overall mean = 13.54%), intermediate in reference wetlands (6.64 %), and lowest in less recently mowed wetlands (1.63 %). This crustacean is common in aquatic systems and tends to occur in relatively high densities where aquatic vegetation is present (De March, 1981). Hyallela is usually classified as a general detritivore, but collects food primarily by gathering plant and animal debris, especially diatoms and bacteria according to Hargrave (1970), and is an important food source for a variety of fish species. The lack of significant difference between less recently mowed and reference wetlands may be due to the Hyallela population recovering from mowing quickly after plants begin to grow back in the mowed area. However, there was also a lack of statistical difference between most recently mowed (overall mean = 1.63%) and reference wetlands (overall mean = 6.64%). This lack of statistical difference may be due to the large standard error associated with the mean relative abundance of Hyallela in reference wetlands (6.64 ± 2.68). The most recently mowed site, Port Austin Rd- wave exposed Schoenoplectus, exhibited a mean relative abundance of 9.61% and was an outlier in the most recently mowed dataset. This habitat may have been misclassified as most recently mowed, since it was in an area that was inundated by about 30 cm of water, perhaps suggesting that it might not have been mowed as recently as other habitats in this category.
The paucity of statistically significant differences among treatments for other invertebrate taxa may be due to functional habitat group affiliation. Invertebrates with morpho-behavioral adaptations for utilizing vegetation, such as snails, which cling to vegetation, may not have been significantly affected by the mowing treatments. Attachment sites and stable vegetation was still available for taxa in these habitat groups during 2004, even in recently mowed sites. However, repeatedly mowing (for multiple years) could potentially lead to a build up of detritus and cause a shift in invertebrate community composition. Repeated mowing could also lead to the demise of perennial emergent vegetation with a resultant shift in invertebrate communities. Therefore, since our experimental design only compares invertebrate communities in wetlands mowed within the last year to reference wetlands, we do not view our findings as conclusive on the effects of longer-term mowing practices.
Effects of Raking on Invertebrates in 2005
A total of 4,730 invertebrates were collected from 7 open water reference zones, while only 118 invertebrates were collected from 5 open water raked zones along Saginaw Bay. Thus, mean catch per zone was significantly greater (p = 0.028) in open water reference zones (677.0 ± 499.73) than open water raked zones (23.6 ± 9.24). Dunn Road (total abundance = 3,665) contributed to the large standard error for the open water reference zones. If we treat this site as an outlier and remove it from analysis, total invertebrate density would still be significantly greater (p = 0.044) in open water reference (179.0 ± 45.56) than raked (23.6 ± 9.24) zones. Mean Shannon diversity (Table 2) differed substantially between open water reference (1.59 ± 0.13) and open water raked zones (0.83 ± 0.36). Yet these differences were not significant (p = 0.104). Mean species richness (Table 2) for reference zones (9.5 ± 1.1), however, was significantly greater (p = 0.023) than raked zones (3.60 ± 1.30).
Invertebrate abundances were compared between open water reference and raked zones by comparing mean catch of taxa comprising > 1% (37 taxa) and > 5% (19 taxa) mean relative abundance. One taxon, Oecetis (Trichoptera) was significantly different between the zones at > 1% (p = 0.030). Oecetis had a mean relative abundance of 3.49 ± 1.20 in open water reference zones while none were captured in open water raked wetlands.
Comparison of Outer Schoenoplectus versus Open Water Wetland Zones in 2005
Based on observations of reference areas, it seems likely that open beaches maintained by raking, tilling, and pulling of vegetation originally supported wetland plants. Thus, we decided to compare unraked, open reference areas with reference areas that had an outer Schoenoplectus plant zone still intact. Any differences between these two reference zone habitats would imply that differences between raked and open water reference zones were potentially underestimates of differences resulting from conversion of wetland plant zones to open beach habitat. There was no significant difference (p = 0.078) in Shannon diversity (Table 2) between the outer Schoenoplectus zones (2.01 ± 0.14) and the open water zones (1.59 ± 0.13). Taxon richness (Table 2) in the outer Schoenoplectus zones (17.9 ± 1.0), however, was significantly higher (p = 0.031) than in the open water zones (9.5 ± 1.1). A total of 3,771 invertebrates (538.7 ± 142.1 per zone) were collected from 7 reference outer Schoenoplectus zones compared to 4,730 (677.0 ± 499.7 per zone) collected from 7 reference open water zones in Saginaw Bay. This difference was not statistically significant (p = 0.297). An unusually large number of invertebrates (3,665), mostly Caenis spp.(mayfly), were collected at the Dunn Road open water zone. If we were to treat this zone as an outlier and remove it from the analysis, total number of individuals would be statistically greater (p = 0.031) in the outer Schoenoplectus (572.83 ± 151.11) than in the open water (179 ± 45.56) zones.
Correspondence analysis was used to determine if aquatic invertebrate community composition in outer Schoenoplectus and open water wetland zones differed. The first and second dimensions of the analysis explained 25.4 and 18.4% of the total variance, respectively (Figure 5). There was a slight separation of the outer Schoenoplectus and open water zones in the second dimension. Taxa most responsible for pulling outer Schoenoplectus zones away from open water zones included a number of Mollusca (mostly snails) and Amphipod taxa while several dipteran taxa were responsible for separating open water zones from the outer Schoenoplectus zones.
The invertebrate community was compared between outer Schoenoplectus and open water zones by analyzing taxa comprising > 1% and > 5% mean relative abundance. Rare taxa were eliminated from the analysis, leaving a total of 47 taxa compared at > 1% mean relative abundance and 19 taxa analyzed at > 5% mean relative abundance. Only one taxon, Chironomini, was significantly different between the zones at > 1% and 5% (p = 0.0001), while the snail, Stagnicola (p = 0.090) and the amphipod, Hyallela (p = 0.078) were marginal. Chironomini was found in greatest abundance in open water zones (25.6 ± 5.4) compared to outer Schoenoplectus (2.9 ± 0.7) zones. Stagnicola had a mean relative abundance of 14.5 ± 4.9 in outer Schoenoplectus versus 3.0 ± 1.4 in open water. Hyallela had a mean relative abundance of 23.2 ± 4.1 in outer Schoenoplectus zones and 6.4 ± 1.5 in open water zones.
Since some differences in macroinvertebrate community composition were found between open water reference and outer Schoenoplectus zones, we hypothesize that differences between open water raked and Schoenoplectus zones would be greater than the differences we report between raked and unraked open water habitats. Since we hypothesize that many of the maintained open beach areas would support emergent macrophytes in the absence of maintenance, a comparison of open water raked habitats to vegetated habitats seems warranted. These comparisons represent an area of necessary future research. Furthermore, if maintained open beaches once supported emergent macrophyte communities, as we believe they did, comparisons of vegetation zones immediately adjacent to the shoreline (i.e. wet meadow zones) to maintained open water habitats would potentially reveal great differences in community composition resulting from the maintenance activity. Inner, more shoreward vegetation zones have been shown to support a much different invertebrate community than outer Schoenoplectus zones (Burton et al., 2002, 2004), and we hypothesize that they are also much different from the open, raked beach habitats based on our community analyses for this project and our past research on coastal wetlands of Saginaw Bay.
Effects of grooming on fish communities
Twenty-five fish taxa were collected from 16 zones on Saginaw Bay and 14 were collected from seven zones on Grand Traverse Bay. Correspondence analysis revealed a dichotomy between Bayport Reference and the other Saginaw Bay zones in the first dimension (Figure 6). This dichotomy resulted from collecting a large school of gizzard shad (Dorosoma cepedianum) at the Bayport Reference zone and nowhere else. The second most variation was represented in dimension 2 and reflected variation in fish community composition due to maintenance activities (Figure 6). Reference zones plotted much higher than groomed sites in this dimension. Taxa found more often and in higher densities at reference sites are listed in Table 4.
The first two dimensions of the correspondence analysis of Grand Traverse Bay fish data appears distorted since only two fish were collected at one of the zones (Figure 7). Accordingly, this zone appears to overwhelm the variation in both dimensions. However, the first dimension is better represented by maintenance activities and its ordination is very similar to that of Saginaw Bay with distinct differences between reference and maintained zones. In dimension 1, reference zones plotted on the right side of the axis, whereas maintained zones were plotted on the left. The Acme Township Park groomed zone plotted among the reference zones most likely because inundated vegetation was present at this zone. The most important characteristic setting the Acme Township Park groomed zone apart from the other groomed zone was its lack of sand shiners (Notropis stramineus). The other outlier was the ungroomed zone at the Waterfront Inn. We actually sampled three areas at this site (reference, ungroomed, and groomed). The ungroomed area had relatively sparse vegetation compared to the Waterfront Reference zone, but was not directly groomed. This site had a very similar fish community to the groomed zones (Figure 7). This was likely due to the lack of vegetation found there, making the fish community in this area more characteristic of a groomed zone. Fish taxa that were more common and found in higher abundance at reference zones and those more closely associated with maintained zones in Grand Traverse Bay, as revealed by the correspondence analysis, are listed in Table 4.
Comparisons of fish communities between boat channels and adjacent vegetated reference habitats were made for wetlands of northern Lakes Michigan and Huron. These comparisons were made since grooming activities were not permissible in these regions and boat channel maintenance represents one of the most significant forms of fragmentation in many of the wetlands in this area.
Correspondence analysis revealed no apparent differences in fish communities between boat channels and adjacent vegetated areas in the first 2 dimensions of the ordination. The majority of variation in both dimensions was due to inter-site variability. The boat channel fish community only separated from the adjacent intact wetland community for one site. This separation was due to the increased abundance of brown bullheads (Ameiurus nebulosus) and blacknose shiners (Notropis heterolepis) in the intact vegetation compared to their abundance in adjacent boat channels and resulted in a lower dimension 1 score for the zone.
The boat channels sampled in the study may have been too narrow to serve as distinct habitats for most fish. Fish are mobile, and many taxa almost certainly had home ranges that include both vegetated and associated boat channel habitat. Even so, fish communities in these fragmented wetlands may have responded to the fragmentation caused by boat channel maintenance. However, juvenile and adult fish data were too variable to detect such a response.
Summary and conclusions
Chemical and physical measurements
Comparison of water quality parameters among mowing treatments in 2004
Mowing, either recently or after approximately one year, had relatively little impact on chemical and physical parameters. However, some differences may have been masked by variability among sites. Those sites that were most recently mowed did show differences in specific conductance, alkalinity, and chloride. These differences may have reflected a tendency for the recently mowed wetlands to be adjacent to development.
Comparison of water quality parameters between open water reference and raked wetlands
When grab samples were collected from the centers of ‘beach’ areas where vegetation had been cleared, chemical and physical parameters fell within the ranges of vegetated reference areas, but some differences were apparent. Specifically, the greatest differences were found in DO and temperature, likely from the intrusion of pelagic water at the open sites, and depth likely from excess erosion where vegetation was lacking. Removal of vegetation disrupted the ambient chemical and physical conditions of the wetlands.
Effects of Mowing or Raking on Macroinvertebrates
To our knowledge, no other studies of effects of mowing or raking on invertebrates have been conducted in Great Lakes coastal wetlands. A few studies conducted elsewhere have documented limited effects of short-term mowing on invertebrate communities. These include studies of the effects of mowing on invertebrate colonization of salt grass in Suisun Marsh, California (de Szalay and Resh, 1997, 2000) and cattail wetlands in Kansas (Kostecke et al., 2005). These studies were in systems very different from Great Lakes coastal wetlands. Even so, the impacts they documented were limited in scope to a few taxa and agreed with our results in finding limited overall impacts of short term mowing on invertebrate communities. While our data do not cover long term effects of continued mowing on invertebrates, it seems likely that continued mowing would result in conversion of wetlands to open beach areas. Based on our field observations and conversations with local citizens, past beach grooming activities have converted many areas of Great Lakes coastline from wetlands to open beaches. Despite this, few data on the effects of such conversion exist. Our results document major impacts of conversion on invertebrate density and species richness. Since invertebrates are relied upon as a food resource by numerous fishes, amphibians, reptiles, and waterfowl, such decreases could have long-term, negative effects on fish and wildlife uses of the Great Lakes.
Effects of Mowing on Invertebrates
Only two taxa were significantly different among reference, less recently mowed, and most recently mowed wetlands (of two sets of analyses with 41 and 26 tested respectively). More differences would have been expected by chance alone at α = 0.05. The two taxa that did differ significantly among treatments did not differ consistently between reference wetlands and the two mowing treatments. This coupled with the lack of any separation that correlated with mowing treatment in correspondence analyses strongly supports the conclusion that no consistent effect of mowing on invertebrate community composition was detected.
Effects of Raking on Macroinvertebrates
The effects of raking and conversion of wetlands to open beach areas on macroinvertebrates included: (1) an 8 to 29 fold significant decrease in the number of invertebrates present, (2) a significant, nearly 3 fold decrease in taxa richness (number of species present), and (3) a marginal (p = 0.11) decrease in Shannon diversity from 1.59 to 0.83. Few differences in community composition between raked and open reference areas were detected, but this may reflect the low numbers collected from the raked sites coupled with high variance in the data.
There was a clear difference in fish communities between reference sites and beach maintenance activities at both Saginaw and Grand Traverse Bays. Reference sites at both locations tended to have higher fish diversity and a greater number of certain taxa. Boat channels located in Northern Lakes Michigan and Huron did not have detectably different fish communities from the adjacent intact vegetation. The channels sampled were likely too small to detect great changes in fish community composition since fish are extremely mobile. The fish found in the channel could be considered associated with vegetation since the channels themselves were only approximately 10 m across. The communities may actually be responding to this impact, but our power to detect the change is quite small since juvenile and adult fish data are often extremely variable.
Additional wetland fragmentation research
The data discussed in this paper were a small portion of the overall studies conducted by Uzarski, Burton, and Albert to address the ecosystem impacts of wetland grooming and fragmentation. These additional data are beyond the scope of this manuscript but can be viewed at: www.michigan.gov/deqwetlands.
We would like to thank MDEQ and NOAA for funding this research. We thank Adam Bosch, Nathan Coady, Kenneth Davenport, Nicolas Fiore, Keto Gyekis, Mary Ogdahl Aaron Parker, Pam Parker, Cole Provence, and Jamie M. Zbytowski for assistance in the field and laboratory. We also want to thank Dr. Dennis Albert for collaboration on this research, and the MDEQ field staff and Joseph E. McBride for assistance in gaining access to study sites. Finally, we would like to thank two anonymous reviewers for improving the manuscript.