Northern Lake Huron marshes are among the most pristine wetlands in the Great Lakes. Almost 200 invertebrate taxa were collected from eight of these marshes from 1997 through 2001. Our objective was to explore relationships between wave exposure (fetch), plant community zones and invertebrate community composition using exploratory data analysis of invertebrate relative abundance. Effective fetch, an exposure measure which integrates fetch along three directions, ranged from 0.4 to 35.3 km. Invertebrates were collected with dip nets from wet meadow, Typha, and inner and outer Scirpus zones from 3 very protected (fetch <1 km), 3 protected (fetch 1–10 km) and 2 exposed (>10 km) marshes. Correspondence analyses of invertebrate relative abundance did not plot invertebrate communities of wet meadows along fetch gradients even though 7 of 30 common taxa were significantly (p < 0.05) correlated with fetch. After removing wet meadow data, correspondence analyses of data from remaining plant zones plotted marshes according to fetch with very protected and exposed sites at opposite ends of U-shaped gradients. Most taxa were generalists, occurred in marshes in all exposure categories, and plotted in the middle of correspondence analyses plots. Characteristic taxa plotting at the very protected end of the gradient included Gammarus, Crangonyx, Caecidotea, Chironomini, Tanytarsini, most Gastropoda and Sphaeriidae. Characteristic taxa plotting at the most exposed end included Sigara, Trichocorixa, Naididae (Stylaria), Tubificidae, and Bezzia. We present a conceptual model of potential changes in invertebrate community composition along gradients of wave exposure. In very protected marshes, organic sediments, detritus, and plant density are higher and dissolved oxygen is lower than in exposed marshes. Conditions are too harsh for some taxa found in very protected marshes.
Great Lakes coastal marshes are important feeding and nursery habitats for fish, amphibians, reptiles, birds, and mammals (Goodyear et al., 1982; Harris et al., 1983; Liston and Chubb, 1985; Jude and Pappas, 1992; Prince et al., 1992; Prince and Flegel, 1995; Whitt, 1996; Brazner, 1997; Maynard and Wilcox, 1997; Riffell, 2000; Weeber and Vallianatos, 2000). Invertebrates are important components of the diets of most of these vertebrates (Chow-Fraser, 1998). Even though knowledge about invertebrates and their response to abiotic factors in Great Lakes coastal marshes has increased substantially in the last decade (Krieger, 1992; Brady and Burton, 1995; Brady et al., 1995; Cardinale et al., 1997, 1998; Burton et al., 1999, 2002, Gathman et al., 1999; Gathman, 2000; Kashian and Burton, 2000; Stricker et al., 2001; Wilcox et al., 2002), the effects of wave exposure and plant zonation on the distribution of invertebrates in wetlands has only been examined for Saginaw Bay wetlands (Burton et al., 2002).
Hydrology and hydrogeomorphic setting are key factors in classification of Great Lakes coastal marshes (Dodge and Kavetsky, 1995; Minc, 1996, 1997; Maynard and Wilcox, 1997; Chow-Fraser and Albert, 1998; Minc and Albert, 1998; Keough et al., 1999; Albert and Minc, 2001). Water levels in the Great Lakes vary by more than 150 cm over periods of years to decades (Burton, 1985; Keough et al., 1999). These long term water level changes, along with changes of 20 to 40 cm from winter lows to summer highs, seiche-driven water level fluctuations of 10 to 20 cm over intervals of less than an hour to 14 hours, and storm surges of 1 to 2 m at infrequent intervals (Bedford, 1992) may all have major impacts on invertebrate and plant communities in wetlands (Burton, 1985; Keough et al., 1999). Relative exposure to wind and waves including storm surges and ice scour appears to be a major driving force in determining types of substrates and plant communities in coastal wetlands (Minc, 1996, 1997; Minc and Albert, 1998) and how rapidly plants respond to lake level changes (based on unpublished plant data from Northern Lake Huron marshes collected by D. Albert, T. Burton and D. Uzarski).
Our hypothesis is that invertebrate communities respond directly to wave exposure and lake level changes and indirectly to habitat changes that occur as plant communities respond to wave exposure and lake level changes. Separating direct effects of waves and lake level changes on invertebrates from indirect effects related to habitat changes will require an experimental approach, but some insight into the importance of each can be derived from existing data using exploratory data analysis such as correspondence analysis (CA) (Burton et al., 2002).
Our objective was to explore relationships between wave exposure, plant community zonation and invertebrate community composition for northern Lake Huron marshes and to compare relationships for these marshes with relationships described for Saginaw Bay marshes (Burton et al., 2002). We used invertebrate data collected from 1997–2001 from eight northern Lake Huron marshes to search for relationships between wave exposure and composition of invertebrate communities for four emergent plant zones. We used results to modify and apply the conceptual model developed for Saginaw Bay (Burton et al., 2002) to invertebrate communities in northern Lake Huron marshes.
Description of study area
Eight fringing, littoral marshes along the northern shore of Lake Huron between St. Ignace and DeTour Village, Michigan were sampled from 1997 through 2001 (Figure 1). Only Duck and Mackinac Bays were sampled all five years; Mismer Bay was sampled in all years except 2001 (Table 1). The other five marshes (Figure 1) were sampled one or more of the five years (Table 1). All marshes except McKay Bay were sampled in 1998. The eight marshes are among the most pristine wetlands in Lake Huron (Burton et al., 1999; Uzarski et al., this issue). All marshes are part of the Les Cheneaux Islands region with the exception of St. Martin's Bay marsh. St Martin's Bay is a large bay west of the Les Cheneaux Islands (Figure 1).
Only emergent zones were sampled. Emergent zones in most marshes included wet meadow, cattail, and inner and outer Scirpus zones respectively from upland or adjacent swamp to open water (Burton et al., 1999; Uzarski et al., this issue). The wet meadow was dominated by Carex stricta and/or Carex lasiocarpa and Calamagrostis canadensis intermixed with a high diversity of other herbaceous plants including other species of Carex, Juncus and Eleocharis and scattered shrubs, particularly Potentilla fructicosa, Salix and Myrica gale. The wet meadow transitioned into a narrow (25–75 m wide), dense cattail zone dominated by Typha angustifolia in most marshes. In Mismer and Mackinac Bays, the cattail zone consisted of scattered patches of cattail in the transitional zone between the wet meadow and Scirpus zones rather than as a distinct zone as in other marshes. An inner Scirpus zone dominated by Scirpus acutus and a high diversity of submersed plants extended into deeper water from the outer edge of the cattail or mixed cattail/wet meadow zones. The inner Scirpus zone was protected from open wave exposure by a slightly deeper, outer 50 to 100 m wide Scirpusacutus zone. The wave-swept, outer Scirpus zone was characterized by fewer stems of Scirpus per m2 and higher interspersion of open water/bare substrate between Scirpus clumps compared to the inner Scirpus zone. Only scattered patches of submersed plants were present in the outer zone.
During 1997, all four zones were inundated. As lake levels fell from 1998 through 2001, progressively fewer zones were inundated. By 1999, only depressions in the inner Scirpus zone were inundated during low points of the seiche cycle. As water levels peaked during seiches, rising water fully inundated the inner Scirpus zone with standing water extending into the outer edge of the cattail or mixed cattail/wet meadow zone. The cattail and wet meadow zones were fully inundated only occasionally during large storm surges from 1999 through 2001.
St. Martin's Bay (Figure 1) marsh was located between two parallel sand ridges on an unprotected shoreline in a large bay. The inner sand ridge supported upland vegetation along the top of the ridge. A wet meadow was located between the inner sand ridge and adjacent forest. A dense inner Scirpus zone between the inner and outer sand ridges was partially protected from waves by the outer sand ridge, although an opening in this ridge allowed dampened waves to penetrate into the zone. A Typha zone occurred at the inner edge of the outer sand ridge. The outer ridge also supported a narrow upland zone at its top. At the outer edge of the outer sand ridge, a very narrow and sparse outer Scirpus acutus patch was present. Only the inner and outer Scirpus zones were included in the analyses, since the wet meadow and cattail zones were not comparable to other sites.
Wave exposure calculation
The degree of protection from waves and storm surges is a function of fetch. Effective and maximum fetch were calculated using procedures recommended by the British Columbia Estuary Mapping System (Resource Inventory Committee 1999). Using GIS software (ArcView GIS 3.2, ESRI, Inc.), fetch distances at each site were measured along three angles relative to the shoreline: 90° (perpendicular), 45° to the left of perpendicular, and 45° to the right of perpendicular. These measurements were then used to calculate effective fetch (Fe) as follows:
A modification of the effective and maximum fetch wave exposure matrix (Resource Inventory Committee, 1999) was used to determine the exposure category for each site (Table 1). This modification reflected the smaller range of fetch distances in the Great Lakes compared to British Columbia estuaries. The eight wetlands (Figure 1) represented a gradient of exposure from very protected to exposed with effective fetch varying from 0.4 to 35.3 km (Table 1).
Three replicate invertebrate samples were collected from each inundated emergent marsh zone with 0.5 mm mesh, D-frame dip nets in late July or early August from 1998 through 2001. In 1997, the marshes were sampled in June and August. Invertebrate samples in June included a predominance of early instars that were more difficult to identify than were the late instars which were present by mid-July. Thus, samples were collected in July or August after 1997. Dip net samples were taken by sweeping the net though the water in each plant zone at the surface, at mid depth and just above the sediments. Dip nets were emptied into a white pan, and a representative sample of 150 invertebrates were picked from the pan in the field by picking all invertebrates in one area of the pan before moving on to another area until 150 specimens were collected. Additional dip net sweeps were taken if the first set of sweeps did not yield 150 specimens. Beginning in 1999, individual replicates were picked for 1/2 person-hour, organisms were then tallied and picking continued to the next multiple of 50. Thus, each replicate contained 150 invertebrates prior to 1999 and either 50, 100, or 150 organisms from 1999–2001. Most 1999–2001 replicates contained 150 organisms except those from the outer Scirpus zone where fewer specimens were collected per sweep. Each replicate was preserved in 90% ethanol in the field and processed individually in the laboratory so that a measure of sampling variance could be calculated.
Specimens were sorted to operational taxonomic unit; usually genus or species for most insects, crustaceans and gastropods. Taxa difficult to identify, such as Chironomidae, were identified to tribe or subfamily. Oligochaeta, Hirudinea, Turbellaria, Hydracarina, and Sphaeriidae were identified to family or order. Taxonomic keys such as Thorp and Covich (1991), Merritt and Cummins (1996), and mainstream literature were used for identification. Identification of representative samples of Gastropoda were confirmed by Dr. Richard Snider at Michigan State University and/or Dr. Brian Keas of Ohio Northern University. A reference collection of Trichoptera and Hemiptera taxa were confirmed by Dr. Brian Armitage of the Ohio Biological Survey. Additional reference specimens were confirmed by Dr. Patrick Hudson of the Great Lakes Science Center (U.S.G.S).
We used correspondence analyses (CA) (SAS version 8, SAS Institute Inc., Cary, NC, USA) of invertebrate community composition for each plant zone to determine if sites clustered in relation to fetch and/or plant zone for each year from 1997 through 2001. Taxa were included in the CA if they represented at least 1% of mean total relative abundance of the invertebrate community for any plant zone in a given year. A separate CA was conducted for June and August in 1997 when all four zones were inundated in Duck, Mismer and Mackinac marshes. When sites separated according to fetch, groups of individual taxa containing the most inertia responsible for the separation were identified. Those taxa that contributed to separation of sites based on exposure over multiple years were plotted in relation to fetch and those with significant correlation coefficients were identified. Taxa responsible for most inertia separating plant zones in each year were also identified. Significant differences (alpha set at p < 0.05) in abundance of these taxa between zones were established using Kruskal-Wallis or Mann-Whitney U tests.
To assess response of taxa to declining water levels, a repeated measures ANOVA (Systat 8.0, SPSS, Inc.) was used to determine whether significant differences (p < 0.05) in taxa relative abundance occurred from 1997 to 2000 (lake level dropped each year from above average in 1997 to substantially below average in 2000, a 1.08 m overall drop. While significant differences may be a response to lake level decline, responses to other differences among years cannot be ruled out). Various taxonomic levels were examined to identify broad-scale community shifts at the order, family, or genus level. Since Duck, Mackinac, and Mismer Bay wetlands were the only sites sampled all four years (Table 1), the analysis was limited to these wetlands. Analysis was limited to the inner Scirpus zone, since it was the only plant zone sampled all four years in all three wetlands. Since time is the ‘within-subject’ factor in this repeated measures design, it is likely that data collected in adjacent years are more correlated than are data from separated years, violating the assumption of circularity. Therefore, adjustment of F-test degrees of freedom using the Huynh-Feldt Epsilon correction (von Ende 1993) was necessary to test the null hypothesis that taxa relative abundance did not change over time as lake levels dropped.
Plant zones and wave exposure
Almost 200 invertebrate taxa were collected from the eight marshes from 1997 through 2001 (Table 2). Seventy-six percent were insects. Mollusca, the second most taxa rich group, included 23 genera of snails and four bivalves (Table 2). Other important groups included Crustaceans, especially Amphipoda and Isopoda; Annelida, especially Naididae and Tubificidae; Nematoda and Cnidarians (Hydra). Lake levels dropped each year from 1997 through 2001 from a mean annual lake level of 176.97 m in 1997 to 175.93 m in 2001 (based on NOAA data for DeTour Village, MI; Station 9075099 available from: http://www.co-ops.nos.noaa.gov/cgi-bin). Different numbers of wetlands were sampled from year to year with two to seven wetlands sampled in any particular year (Table 1). Varying numbers of wetlands sampled and changes in lake level from year to year made it necessary to analyze data for each year separately in order to determine relationships between fetch (intensity of wave exposure) and invertebrate community composition for each plant zone.
1997 correspondence analyses
Duck, Mackinac and Mismer Bay marshes were sampled in 1997 when wet meadows were inundated. Wet meadows were not inundated at most sites from 1998 through 2001. The 1997 data were used to search for relationships between fetch and wet meadow invertebrate communities. Correspondence analyses of 1997 data from all plant zones (wet meadow, Typha, inner Scirpus, outer Scirpus) separated wet meadow zones from other plant zones in June and August (See Figure 2 for an example of how CA was used to identify taxa responsible for inertia pulling some sites or zones away from others). Grouping of wet meadow communities from the three sites away from other plant zones suggested that the wet meadow invertebrate community was substantially different from deeper water communities. Taxa with significantly (Kruskal-Wallis, p < 0.05) greater relative abundance in the wet meadow zone than in other plant zones in June and August included: Gerridae, Pisidium, Planorbula armigera and Physa gyrina. The relative abundances of Ceratopogonidae, Tanytarsini, Dytiscidae, and Sympetrum were significantly greater (Kruskal-Wallis, p < 0.05) in wet meadows in June than in other plant zones, while Hesperocorixa and Libellula were significantly greater in wet meadows in August than in other zones.
The wet meadow invertebrate community of Duck Bay, the most protected of the three sites, plotted apart from wet meadow communities of the other two sites in both months. Taxa responsible for most inertia separating Duck from the other two sites included: Planorbula armigera, Pisidium, and Dytiscidae in both months. In June, Tanytarsini and Libellula were also important in separating the Duck Bay wet meadow from wet meadow communities at the other sites, while in August, Hesperocorixa michiganensis was one of the taxa separating Duck Bay from the other two sites.
Removal of wet meadow zone invertebrate data from correspondence analyses resulted in a clear separation of Typha zones from Scirpus zones in June but only partial separation in August. The grouping of Typha zones regardless of site suggested that wave exposure was relatively unimportant in structuring invertebrate communities in the Typha zone. However, Scirpus zones dampened the wave exposure each Typha zone experienced resulting in a relatively narrow gradient of exposure. Taxa with a significantly (Kruskal-Wallis, p < 0.05) greater relative abundance in the Typha zone compared to other plant zones in June included Crangonyx pseudogracilis, Caenis, Tipulidae and Nehalenia irene.
With only three marshes sampled, differences in invertebrate community composition in relation to fetch should be viewed as suggestive rather than conclusive. Once the wet meadow zones were removed from the 1997 data set, CA arranged sites according to a gradient of exposure to waves as determined by fetch calculations (Table 1), in both June and August with the first two dimensions explaining at least 50% of the variance in each. In both June and August, the Scirpus zones at the most protected site (Duck Bay) were plotted together in the top-right corner of the plot with the perceived exposure gradient proceeding towards the bottom-left corner, where the Scirpus zone of the most exposed site (Mismer Bay) was located. The taxa responsible for the separation of Mismer Bay Scirpus zones from the more protected sites in either June or August were Baetidae (Callibaetis and Procloeon), Corixidae (Sigara and Trichocorixa), Orthocladiinae, Phryganeidae (Agrypnia), Oligochaeta, Lymnaeidae, and Pyrgulopsis lustricus (Hydrobiidae). Taxa that had a greater relative abundance in the most protected site (Duck Bay) Scirpus zones in either June or August included Amnicola limosa, Bithynia tentaculata, Laevopex fuscus, Musculium securis, Tanytarsini, Chironomini, Tanypodinae, Caecidotea, Gammarus fasciatus and Crangonyx pseudogracilis. Only Amnicola limosa, Bithynia tentaculata, Musculium securis and Gammarus fasciatus had greater relative abundances at Duck Bay in both months.
1998 correspondence analyses
In 1998, the inner Scirpus zones of seven marshes were sampled (all except McKay Bay; Table 1). Correspondence analysis of inner Scirpus data separated very protected sites (Duck, Prentiss and Peck Bays) from more exposed sites (Voight and St. Martins Bays) with very protected sites grouping at the top-right and the most exposed sites at the top-left of the graph (Figure 2, also see Table 1 for fetch and exposure categories of the seven sites). Grouping of sites with intermediate levels of fetch exposure (Mackinac and Mismer Bays) in the middle of the graph and the most exposed sites toward upper left, and the most protected sites towards the upper right of the graph is consistent with groupings along a U-shaped gradient even though the presence of only 3 exposure classes in Figure 2 makes detection of a U-shaped gradient problematic. U-shaped gradients are common in CAs when detrended techniques are not used because extreme ends of the continuum tend to lack many organisms found toward the middle of the continuum making the extremes more similar to each other as in Figure 2.
As in 1997, Corixidae was an important taxon responsible for inertia separating the most exposed sites (Voight and St. Martin's) from the most protected sites (Cor (unidentified corixids) and Tri (Trichocorixa borealis) in the upper left hand corner of Figure 2) as were the snails, Gyraulus (Gyr) and Valvata (Val) (Figure 2). The relative abundance of Naididae, including unidentified Naididae (Nai) and Stylaria (Sty), was higher in the two most exposed sites compared to protected and very protected ones (Figure 3) and were partially responsible for pulling these two sites away from the other sites (Figure 2). This may mirror results from 1997 when Oligochaeta were higher in exposed sites. However, Oligochaeta were not identified to family in 1997, so we cannot be sure that the greater oligochaete relative abundance in 1997 was due to Naididae as in 1998. Bezzia (=Be which is partially obscured by the Voight Bay label in Figure 2) and Hyalellaazteca (Hya) were also more abundant in the two most exposed sites than in the rest of the protected and very protected sites in 1998 (Figures 2, 3). Ancylidae (Ferrissia parallelus (Fer) and Laevapex fuscus (Lae)), Chironomini (Chi), Gammarus (Gam), Crangonyx pseudogracilis (Cra), Oxyloma retusa (Oxy), Amnicola limosa (Amn)and flatworms (Turbellaria, Tur) were more abundant in 1998 in very protected sites than in more exposed ones (Figures 2, 3). Caenis (Cae) was more common in protected sites (Mackinac and Mismer) and exposed sites in 1998 than in very protected sites (Duck, Prentiss and Peck Bays, Figures 2, 3).
1999 correspondence analyses
Due to decreased water levels in Lake Huron in 1999, invertebrate sampling was limited to inner and outer Scirpus zones at most sites. Correspondence analysis of 1999 data revealed two distinct groups representing inner Scirpus and outer Scirpus zones (See Figure 2 for an example of how CA was used to identify taxa responsible for inertia pulling some sites or zones away from others). Taxa that exhibited significantly (Mann-Whitney U, p < 0.05) greater relative abundances in the inner Scirpus zone included Caecidotea, Hyalella azteca, Caenis, Aeshnidae, Libellulidae, Gerridae, Belostoma, Mesovelia, Hydrophilidae, and Pseudosuccinea columella. Taxa that exhibited significantly (Mann-Whitney U, p < 0.05) greater relative abundances in the outer Scirpus zone included Hydracarina, Hexagenia, Sialis, Tanytarsini and Amnicola. The outer Scirpus zone at St. Martin's was the outlier of these two groups, perhaps due to the more intense wind and wave exposure this site received in comparison to other sites sampled in 1999. Taxa responsible for most inertia pulling this site away from the others included Corixidae (Sigara, Trichocorixa), Stagnicola and Helicopsyche.
Correspondence analysis of 1999 inner Scirpus data again resulted in an apparent U-shaped exposure gradient with Prentiss and St Martins Bays at either end. Taxa responsible for inertia pulling the very protected sites with highest relative abundances in very protected sites included Orthocladiinae, Mystacides (Leptoceridae) and Gammarus. Trichocorixa, Ishnura verticalis, Tubificidae worms and Physa gyrina all had highest relative abundances in the most exposed site sampled in 1999 (St. Martins). Those taxa that made up 1% or more of invertebrates at any site are graphed in Figure 4. As in 1998, the relative abundance of Caenis was greater in protected sites than in very protected or exposed sites. A CA plot of outer Scirpus data failed to arrange sites according to exposure.
2000 and 2001 correspondence analyses
In 2000 and 2001, only the inner and outer Scirpus zones were sampled, since the other zones were not inundated. Correspondence analyses of these data did not reveal any distinct groupings based on plant zones in either year (See Figure 2 for an example of how CA was used to identify taxa responsible for inertia pulling some sites or zones away from others). However, 2000 Duck Bay inner Scirpus data plotted well away from inner Scirpus data for the other two sites. The fetch for the Duck Bay marsh was lower than for the other two marshes. The taxa responsible for inertia pulling Duck Bay apart from other sites included Caecidotea, Bithynia tentaculata, Mystacides, and Coenagrionidae. In 2000, sites were grouped by correspondence analysis in a manner similar to CA grouping in 1997 when the same sites were sampled. Some of the same taxa responsible for pulling the sites apart in 1997 also pulled them apart in 2000. For example, Corixidae (Sigara and Trichocorixa) and Procloeon were important in separating Mismer from the other two sites in both 1997 and 2000. However, taxa that were not important in separating Mismer from other sites in 1997 (e.g., Sphaeriidae, Fossaria, Physa gyrina, and Amnicola) were important in 2000.
Correspondence analysis of 2001 data plotted sites in an arrangement similar to 1997 and 2000, except that Mismer Bay was replaced by McKay Bay in 2001 as the most exposed site sampled. The corixids, Sigara and Trichocorixa, separated McKay Bay from the other sites in 2001 as they had for Mismer Bay in 1997 and 2000. Caecidotea and Bithynia tentaculata contributed to the separation of Duck Bay from the other sites in 2001 just as they had in 2000 when Mismer Bay was sampled instead of McKay Bay.
Significant correlations within plant zones between fetch and invertebrate relative abundance
For each plant zone, Pearson correlation coefficients were calculated between effective and maximum fetch and the relative abundance of each taxon that made up 5% or more of relative abundance in either Scirpus zone for any marsh in any year from 1997 through 2001 (a total of 30 taxa, Table 2). Correlation coefficients between effective fetch and relative abundance (Table 3) were similar to correlation coefficients between relative abundance and maximum fetch. Thus, the results for effective fetch also apply to results for maximum fetch (see Table 1 for effective and maximum fetch values for each marsh). Only 1.5 significant results per zone would have been expected by chance alone (alpha set at p<0.05); thus, fetch was correlated significantly more often with relative abundance of taxa than would be expected by chance alone with 7 to 8 significant correlations per zone (Table 3).
The majority of correlations between fetch and relative abundance were positive suggesting that most taxa were more abundant in exposed sites than in protected ones (Table 3). Taxa characteristic of exposed sites, as identified by correspondence analyses, also tended to increase in relative abundance with increasing fetch (e.g., Naididae, Corixidae, Sigara, Ceratopogonidae) when data across all years were examined (Table 3). Conversely, taxa associated with more protected sites (e.g., Chironomini, Gammarus) decreased in relative abundance as fetch increased.
Even though correspondence analyses did not plot wet meadows according to fetch, there were as many significant correlations between fetch and individual taxa in this zone as in the other three zones (Table 3). Caenis was positively correlated with fetch in wet meadow and inner Scirpus zones even though it was associated with protected sites in most years (Table 3). Enallagma was also positively correlated with fetch in wet meadows as were all oligochaete taxa (Table 3).
Temporal variation associated with declining water levels
Only 7 (10%) of the taxa were significantly (p<0.05, ANOVA) influenced by declining water levels (or some other time related factor) in Lake Huron over the period from 1997 to 2000. Half of these would have been expected by chance alone since alpha was p<0.05. Most of the significant responses involved individual Odonata taxa and were not independent of each other. The dragonfly, Epitheca, exhibited the strongest decline in relative abundance (p=0.001, ANOVA). Its decline contributed significantly to the similar trend exhibited by the suborder Anisoptera. The damselfly, Enallagma, was also significantly influenced by the changing water levels, but, instead of a decline each year, its relative abundance was significantly greater in 1998 than in any other year. The decline of Enallagma after 1998 influenced the four year decline of the family, Coenagrionidae, the suborder, Zygoptera, and, along with declines in Epitheca, may have accounted for much of the four year decline in Odonata at the order level. The caddisfly genus Oecetis displayed a similar pattern to Enallagma, with a significantly elevated relative abundance in 1998 and declines after that as lake levels continued to fall. Given that the number of significant correlations was only 3 to 4 higher than expected by chance alone and that some of the 7 significant correlations were for the same taxa at different levels of taxonomic resolution, these data provide little evidence that significant changes in community composition were related to lake level decline.
Synthesis of results
Many taxa tended to be more commonly collected in one of the three exposure categories than in the other two (Table 4). Ten taxa always occurred in highest relative abundance in the very protected marshes, while another eight taxa reached their highest relative abundance in either very protected or protected marshes. Nineteen taxa were more abundant in protected marshes than in either very protected or exposed marshes. Fewer taxa were found in highest relative abundance in exposed marshes (3) or in a combination of protected and exposed sites (5) (Table 4). Only 7 to 8 of these were significantly (p<0.05) correlated with fetch within a particular plant zone (Table 3).
Taxa consistently important in separating very protected sites from protected and exposed sites in correspondence analyses included the amphipods, Gammarus fasciatus and Crangonyx pseudogracilis, midges in the tribes Chironomini and Tanytarsini, and Leptoceridae caddisflies, especially Mystacides interjecta (Table 4). As a group, snails were also much more commonly found in very protected or protected sites (Table 4), although the importance of individual snail species varied from marsh to marsh and year to year. For example, Bithynia tentaculata was found almost exclusively in Duck Bay, one of the very protected marshes. Other species of snails that were important in separating very protected marshes from protected or exposed sites included Amnicola limosa and Oxyloma retusa. Other mollusks found more commonly in very protected sites included the limpets (Ferissia parallela and Laevopex fuscus) and a species of fingernail clam (Musculium securis). At the family level, however, Sphaeriidae were found more commonly in protected sites than in very protected or exposed marshes, while another genus, Pisidium, was important in separating wet meadows from other habitats. Several snails were found in either very protected or protected marshes (e.g., Lymnaeidae including Fossaria parva, Physa gyrina, Pyrgulopsis lustricus). The only snail taxon that reached its highest relative abundance in exposed marshes was Valvata (Table 4).
Taxa that were consistently more important in separating exposed marshes from protected or very protected marshes in correspondence analyses included Corixidae, especially Sigara and Trichocorixa borealis (Table 4). However, one corixid, Hesperocorixa michiganensis, was responsible for separating wet meadow invertebrate communities from Typha and Scirpus communities. Oligochaeta, especially Naididae (Stylaria) and, to a lesser extent, Tubificidae, were also important in separating exposed from protected marshes in CA plots (Table 4). Oligochaeta relative abundance correlated with fetch for three of the four plant zones (Table 3).
We developed a conceptual model to integrate invertebrate results along exposure gradients (Figure 5). As fetch and wave exposure increase, the outer plant community becomes increasingly dominated by widely spaced clumps of Scirpus acutus interspersed with sandy substrates containing little detritus, and invertebrate densities decrease (although we did not take quantitative samples, it took much longer to collect 150 specimens in outer Scirpus zones than it did in more protected zones). Characteristic taxa were determined by analyzing CA graphs for each year of observations (as the example shown in Figure 2) as well as by compiling data on which zones had the highest relative abundances of individual taxa (Table 4). Characteristic taxa in marshes at the exposed end of the exposure gradient differ substantially from those in marshes at the more protected end (Figure 5). The very protected marshes have more plant stems per unit area and richer organic substrates than do marshes at the exposed end. ‘Characteristic’ invertebrate taxa listed for the exposed and very protected ends of the exposure gradient are more common at one end or the other of the gradient but also occur in marshes with intermediate fetch exposure. This results in highest diversity in the middle of the exposure gradient, since characteristic taxa listed for mid-gradient marshes (Figure 5) mix with characteristic taxa from both ends of the gradient resulting in highest diversity in mid-exposure gradient marshes (Figure 5).
Within each wetland, there is also a gradient of exposure to waves as depths increase and plant communities change from wet meadow to outer Scirpus zones. Wet meadows do not plot along fetch gradients in correspondence analyses and are not included in Figure 5. Along exposure gradients in each wetland, predictable changes in dissolved oxygen and dissolved ions occur as resistance from plant stems damps out wave energy and limits penetration of pelagic water into wetlands (Cardinale et al., 1997, 1998). The more exposed the wetland is, the more interspersion between stems there tends to be in the outer Scirpus zones, so that waves mix pelagic water farther into wetlands than in protected sites. This results in predictable changes in communities at the opposite ends of the exposure gradient (Figure 5).
Recently, Burton et al. (2002) published a conceptual model of the effects of wave exposure and plant community zonation on Saginaw Bay wetland invertebrate communities. They based their conceptual model on comparisons of invertebrates of similar plant communities in inland, protected wetlands to littoral wetlands (e.g., cattail zones in each complex) and on comparisons between wave exposed and protected wetlands (e.g., Scirpus communities) on windward and lee sides of islands. They found that most invertebrate communities of littoral wetlands were likely established along a gradient of exposure with differences between plant zones being less important for exposed sites than for protected sites. Thus, their results from Saginaw Bay generally agree with our findings for northern Lake Huron marshes. However, direct comparisons for specific taxa in Saginaw Bay and northern Lake Huron marshes did not always agree even though several of the same taxa were present in both wetland complexes. Since Burton et al. (2002) did not quantify fetch for the Saginaw Bay sites, it is difficult to know where their sites fit along the fetch gradient calculated for northern Lake Huron wetlands. Therefore, only general trends in taxa relative abundance relative to exposure can be compared between the two regions. Some trends described for taxa in Saginaw Bay wetlands agreed with our results from northern Lake Huron wetlands. For example, Burton et al. (2002) found that Oligochaeta (Naididae, Stylaria), were more common in littoral than in inland marshes, and this parallels our finding that Naididae and Stylaria were more common in exposed than in very protected marshes (Figure 5). They found that Asellidae (their Asellidae = our Caecidotea; we checked their samples from Saginaw Bay to confirm this) occurred in large numbers in inland and protected wetlands. Similarly, we found that Caecidotea tended to be more abundant in the most protected marshes (Figure 5). They found that Hydracarina (water mites) were much more common in inland or protected sites than in exposed ones, we found that Hydracarina were more likely to achieve highest relative abundances in protected rather than in very protected or exposed marshes (Figure 5). Other findings are not as comparable. For example, Orthocladiinae midges were associated with exposed sites in Saginaw Bay. While this was true for the inner Scirpus zone in 1997 in NLH marshes, Orthocladiinae did not exhibit any trends relative to exposure in most years and were associated with more protected sites in the outer Scirpus zone in 1999. The more exposed sites in Saginaw Bay may not be as exposed as the most exposed NLH marshes. The outer Scirpus zone of the most exposed NLH marshes may be too exposed to allow this subfamily of midges to thrive. Another possibility is that different species with different requirements were present in Saginaw Bay and NLH marshes. Corixidae (water boatmen) were associated with protected Typha and wet meadow zones in Saginaw Bay, but were consistently more abundant in the most exposed NLH marshes. This apparent disagreement may represent lack of taxonomic resolution. We found that one corixid, Hesperocorixa, was associated with protected Typha and wet meadows but that two others, Sigara and Trichocorixa, were more abundant in the most exposed marshes.
Differences in our findings may also be due to ecoregional differences, and these are detailed in the companion paper (Uzarksi et al., this issue). In addition, gradients in northern Lake Huron marshes (Figure 5) may differ from those described for Saginaw Bay (Burton et al., 2002), especially within individual marshes. The greatest differences are related to water quality. Saginaw Bay drains a large agricultural watershed, and pelagic water is highly turbid, nutrient enriched, and exposed to more agricultural chemicals than is pelagic water of northern Lake Huron (Burton et al., 1999; Uzarski et al., this issue). Thus, exposure gradients within individual marshes in Saginaw Bay include an increase in turbidity with exposure due to mixing of turbid, pelagic water into the outer marsh (Cardinale et al., 1997, 1998). Northern Lake Huron wetlands drain primarily forested watersheds and pelagic waters of Lake Huron are much less turbid than pelagic waters of Saginaw Bay. Thus, risk of fish predation increases from wet meadow to outer Scirpus zones in northern Lake Huron marshes (Gathman, 2000), but decreases with exposure in Saginaw Bay marshes due to turbidity limiting detection of prey by visual predators (Cardinale et al., 1998). This may account for some differences in the two areas.
Gathman (2000) conducted research in 1996 and 1997 in Mackinac Bay marsh, one of the eight marshes included in our analyses (Figure 1) and found that depth was more important than plant zonation in determining invertebrate community composition. This too parallels our finding that exposure (which would correlate with depth within a marsh) was more important than plant zonation in determining community composition after wet meadow data were removed from analyses.
Our findings and those of others suggest that invertebrate communities in marshes are made up of many generalists which occur across all plant zones regardless of fetch (wave exposure) and a smaller number of specialists that are found on either end of the exposure gradient (Figure 5). Wet meadow communities are well protected from waves by outer plant zones and do not tend to relate to fetch in correspondence analyses. Wet meadows contain several taxa that are more abundant there than elsewhere (e.g., Dytiscidae, Gerridae, Pisidium, Planorbula armigera, Hesperocorixa michiganensis). There are enough of these taxa to make this zone plot away from other plant zones in correspondence analyses. Gastropoda are important taxa in the wet meadow zone, and they appear to be affected by differences in water chemistry, structural habitat complexity, and plant dominance in northern Lake Huron wetlands (Keas, 2002). Even though the wet meadow zone is protected, a few taxa are correlated with fetch. These may be taxa that migrate into or out of the zone from deeper water where they are periodically exposed to wave action. Gathman (2000) described migrants from deeper water as being important in wet meadows in late season for Mackinac Bay, one of our study sites.
The Typha zone invertebrate community is also well protected from waves by the two Scirpus zones, but a few taxa in this zone also correlate with fetch (exposure to waves). The two Scirpus invertebrate communities are much more exposed to waves, and invertebrate communities in them tend to plot in relation to fetch and away from the more protected zones in correspondence analyses. Even so, the same number of taxa correlate with fetch in each of the four plant zones (7–8 per zone, Table 3). We conclude that fetch and plant community composition are important parameters in understanding habitat requirements of coastal wetland invertebrates with fetch being important for comparisons among several wetlands while plant community composition is more important in determining invertebrate species composition along exposure gradients within individual marshes.
The next step in understanding these relationships should involve experiments with individual species groups in order to establish how each species is affected by wave exposure and the concomitant changes in plant community composition and structure. It is also likely that biotic interactions such as predation pressure and competition will shift as habitats change in relation to fetch and lake levels (Gathman, 2000), and these factors will also have to be examined before a true understanding of community dynamics in coastal wetlands can be achieved.
Funding provided by: The Nature Conservancy, Michigan Department of Environmental Quality (MDEQ), The Michigan Great Lakes Protection Fund (MDEQ), Region V, U.S. Environmental Protection Agency (USEPA), U.S. Geological Survey's Great Lakes Science Center (GLSC, USGS), and Great Lakes Fishery Commission. Taxonomic assistance provided by Brian Armitage, Patrick Hudson and Brian Keas. Field and/or laboratory assistance provided by: Melissa Asher, Beau Braymer, Matthew Cooper, Angie Conklin, Kari Divine, Joe Gathman, Kristen Genet, Rochelle Heyboer, Donna Kashian, Brian Keas, Katie Kiehl, Todd Losee, Scott Mueller, Ryan Otter, Sam Riffell, Mark Scalabrino, Rebekah Serbin, Christy Stricker, Craig A. Stricker, Shawn Wessell, and Jamie Zbytowski.