INTRODUCTION

A third of freshwater mussel species in the province of Ontario (Canada) have been assessed as either federally threatened or endangered (COSEWIC 2012). Initial declines in native unionid populations have been related to the degradation of riverine habitats (Metcalfe-Smith et al., 1998). More recent and rapid declines followed the invasion of North America by driessenid mussels: Zebra Mussel (Dreissena polymorpha) and Quagga Mussel (D. bugensis) (Schloesser & Nalepa, 1994). By the early 1990s, native mussels were nearly extirpated from the offshore waters of Lakes Erie and St. Clair (McGoldrick et al., 2009). However, remnant mussel assemblages have persisted in nearshore and coastal wetland areas of Lakes Erie, Huron and St. Clair (Nichols & Amberg 1997; Zanatta et al., 2002; Bowers & Szalay, 2003; Crail et al., 2011; Sherman et al., 2013). Compared to adjacent open water habitats, wetlands are less suitable for driessenid colonization and survival ( Sherman et al., 2013), thereby providing a refuge for native mussels. Given that dreissenid mussel removal may not be practical and brood-stock is required for reintroductions, recovery depends on identifying and protecting remnant native mussel assemblages.

Actions undertaken to protect and recover Ontario's mussels at risk include the identification of protected habitats, and ongoing assessment of species status. To meet these commitments, the following information is required: (1) the locations of individuals and populations, (2) descriptions of the biophysical attributes of habitat for different life-stages, (3) the state of populations (i.e. density, size and age structure, sex-ratio), and (4) the presence of invasive species (Cudmore et al., 2006; DFO, 2011a). Outside of the Lake St. Clair delta, these activities have focused on populations in southwestern Ontario rivers. Riverine mussel assemblages are sampled with standardized time-search (Metcalfe et al., 2000) and quadrat methods (Metcalfe-Smith et al., 2007). In contrast to riverine mussels, little research has been undertaken on the design of wetland inventory or monitoring programs. The evaluation of alternative gear and sampling strategies is required for species found in inland lakes and coastal wetlands (e.g. Ligumia nasuta).

In this study, we compared timed-search (qualitative) collections from Lake Ontario coastal wetlands using clam rake and visual/tactile methods. Both methods have been used in previous wetland mussel surveys (Bowers & Szalay, 2003; Sherman et al., 2013). Other sampling approaches have included snorkeling surveys and opportunistic surveys of temporarily dewatered habitats (Nichols & Amberg, 1999; McGoldrick et al., 2009; Crail et al., 2011). The two methods were chosen based on the logistics of sampling coastal wetlands (soft sediments, poor water clarity and dense aquatic vegetation), and because they are not time-intensive (Bowers & Szalay, 2003). Also, it was expected that the clam-rake would be able to sample habitats too deep for visual/tactile methods. Comparisons were based on the: (1) number of species detected, (2) number and sizes of individuals collected, and (3) precision of mussel species and abundance estimates. We also assessed the concordance of mussel assemblage patterns described using clam-rake and visual/tactile data.

METHODS

Data Analysis

Differences between the two methods were tested using the following data: (1) number of species detected, (2) number of individuals collected, (3) number of sampling points containing mussels, (4) precision of parameter estimates (mussel abundance and species number), and (5) shell length (minimum, mean and maximum). For datasets 1–3, separate comparisons were done for live individuals and fresh shells. Precision (calculated for each wetland) was based on the coefficient of variation (CV = Standard Error/Mean) (Thompson, 2002). Except for precision and shell length, significant differences between the sampling methods were tested with the paired t-test. Due to differences in mussel data, an unequal number of CVs was calculated for each sampling method. Therefore, tests for significant differences were undertaken with the unpaired t-test. Differences in shell length were tested using the Sign-test (Zar, 1984). Species detection rates for each method were compared by calculating the mean (across wetlands): (1) time till the first species was detected at a wetland, and (2) cumulative number of species detected after each hour of searching.

Two approaches were applied to assess the concordance of variation in mussel assemblages among wetlands, as described using clam rake and tactile/visual data on live individuals. To assess whether estimates of live mussel abundance and species richness across wetlands agreed, the Spearman Rank Correlation was calculated. Secondly, distance matrices were constructed from site-by-species matrices of species presence/absence (Jaccard) and log-transformed species abundance (Bray-Curtis) data (Legendre & Legendre, 1998). The relationship between matrices constructed from clam rake and tactile/visual data was evaluated using the Mantel test. Significance was assessed with a Monte Carlo randomisation method, using 9999 permutations (Manly, 2007). Statistical analyses were completed using PAST version 1.94 (Hammer et al., 2001).

TABLE 1

Comparison of relative abundance (% of total collection) and frequency of occurrence (% of wetlands sampled) of wetland mussel species collected using clam-rake and visual/tactile methods. Absolute numbers are provided in parentheses. Summary statistics are based on live individuals.

RESULTS

A total of 756 live mussels (representing nine species) were collected from Lake Ontario coastal wetlands (Table 1). Between one and five species were detected as live individuals from each wetland. At least one live mussel was collected from most wetlands (except Big Island Marsh, Bay of Quinte). Pyganodon grandis was the most widespread (collected from all wetlands where live mussels were found) and abundant species (65% of visual/tactile and 40% of clam rake collections). Other species were encountered at five or fewer wetlands, and typically represented <10% of the total collection. Three at-risk mussel species (COSEWIC 2012) were collected: L. nasuta, Quadrula quadrula, and Toxolasma parvum, of which L. nasuta was the most widespread. For both methods, live individuals were more often collected than fresh shells (Figure 1). In total, 870 shells (either halves or whole) were collected. Shells of all species except T. parvum were found. At half of the wetlands, some species (range: one to three species) were detected only as fresh shells (Table 1). At four wetlands, the presence of shells was the only indicator of the occurrence of Elliptio complanata, Leptodea fragilis and Utterbackia imbecillis. Shells were also the only evidence of L. nasuta within Presqu'ile Bay.

Abundance and Number of Mussel Species

Live mussels were collected with visual/tactile searches from 90% (19 of 21) of wetlands sampled, and from 71% (15 of 21) with the clam-rake. Visual/tactile searches collected three times as many live mussels (t = 2.35; p = 0.02) and twice as many fresh shells (t = 2.81; p = 0.01) as clam-raking (Figure 1). They also produced significantly more (>35%) sampling points at each wetland with live individuals (t = 4.1; p < 0.001), and more species from live individuals than clam-raking (t = 2.95; p = 0.008). Alternately, there was no difference between methods in the number of sampling points with shells at each wetland or species detected with shells (t-test; p > 0.10). Clam-raking only detected species not present in visual/tactile collections at five wetlands. There were no significant differences between search methods in the precision of live mussel abundance and species number estimates (t-test, p > 0.25). Overall, estimates were typically imprecise (CV>0.35).

FIGURE 1

Comparison of mussel catch rates in Lake Ontario coastal wetlands using visual/tactile (VT) and clam-rake (CR) sampling methods. “Points” refers to the number of sampling points within a wetland where a fresh shell or live individual was collected.

Over the sampling period, detection of new species within wetlands was more rapid with visual/tactile searches than clam-raking (Figure 2). Although mean (±SE) time spent searching until the first mussel was collected at a wetland was slightly longer (VT: 3.5 hr ± 0.9. CR: 2.7 hr ± 0.7), there was little improvement in the detection of new species after 10 hours of visual/tactile searching or six hours of clam-raking. When compared to combined species lists (both methods) for each wetland, the mean percentage of mussel species detected using the visual/tactile method was greater than 80%. By contrast, clam-raking detected (on average) 23% fewer species known from each wetland (mean = 57.5%).

Variation Among Coastal Wetland Mussel Assemblages

The number of live individuals (r_s = 0.81, p < 0.001) and species (r_s = 0.78, p < 0.001) collected by each method was strongly correlated across wetlands. However, distance matrices constructed from species presenceabsence data (Jaccard) were not correlated (r = -0.04, p = 0.60). At 10 of the 21 wetlands, there was no overlap in the species composition of visual-tactile and clam-rake collections. Most of these cases reflect the failure of a sampling method to collect any mussels. Using species abundance data (Bray-Curtis), there was a weak correlation (r = -0.23, p = 0.007) between distance matrices associated with each method. The relative abundances of individual species were equal at only 10% (2 of 21) of wetlands sampled.

FIGURE 2

Comparison of increase in mussel species richness with increase in effort, during visual/tactile (○) and clam-rake (▪) surveys of Lake Ontario coastal wetlands. Mean species richness represents the average calculated across all wetlands sampled.

DISCUSSION

Sampling gear and strategy evaluations for North American freshwater mussels have largely focused on riverine habitats. Given the catastrophic losses of freshwater mussel diversity across the Great Lakes, the inventory and population monitoring of remnant populations in coastal wetlands are priority recovery actions. Our study shows that visual/tactile surveys are more efficient at collecting mussels and detecting species than clam-raking. This result is consistent with Sherman et al. (2013) who reported that visual searches of a Lake St. Clair site collected four times more mussels than clamraking. We also found a broader range of shell lengths to be associated with visual/tactile collections. It is not known whether this result reflects differences in the likelihood of capture between methods, or that the probability of detecting the smallest and largest sizes increases as one collects more mussels.

Compared to visual/tactile sampling, the clam-rake permits sampling at deeper water depths and further into soft sediments. However, it was less effective at collecting mussels and often labour-intensive. When sampling heavily vegetated habitats or soft sediments, the basket required continuous cleaning to remove plant and/or organic material and careful searching to find mussels. Unionids in Lake Erie wetlands have been observed to burrow 2–40 cm into the substrate for at least part of the day (Nichols & Wilcox, 1997). In rivers, the excavation of bed material during quadrat sampling improves the likelihood of collecting juveniles and small-bodied species (Obermeyer ,1998). In our study, there was no evidence that dragging the clam-rake through sediments improved the detection of small individuals. This may reflect the loss of small individuals through the basket mesh, or that tactile sampling was effective at detecting burrowed individuals by probing the sediments. Certain wetland characteristics (deep water or dense aquatic vegetation) may prevent visual/tactile searches. In these cases, the completeness of species lists may be improved by increasing the time spent clam-raking and/or shoreline searches for fresh shells.

FIGURE 3

Length-frequency distributions of mussels collected from Lake Ontario coastal wetlands using visual/tactile (white) and clam-rake sampling methods (black).

While the qualitative sampling approach applied in this study was appropriate for gear comparison and species inventory, quantitative sampling strategies are recommended for population monitoring (Strayer & Smith, 2003). Both methods evaluated in this study provided imprecise parameter estimates and were unsuitable for long-term monitoring. Future research could test whether stratified sampling designs, and/or large increases in search effort would improve precision. Alternatively, the ability of quantitative approaches developed for riverine mussel populations (e.g. systematic quadrat sampling with random starts) to provide abundance estimates could be evaluated for these low-density populations. If it is not necessary to track the number of individuals (or it is deemed impractical), repeat survey designs could be implemented across lower Great Lakes coastal wetlands to monitor species distributions instead (MacKenzie et al., 2012).

The overall objective of our study was to inform the design of native freshwater mussel collections in Great Lakes coastal wetlands. However, recovery plans for mussels at risk also identify the need to monitor dreissenid distribution and abundance. This information is used to interpret threat risks for individual populations (DFO 2011b). Live dreissenids were collected from nine of the wetlands we sampled. At these sites, dreissenid shells were present at 30% of sampling points. Across a variety of European and North American waterbodies, zebra mussel infestation rates (number or mass attached to native unionids) are correlated with zebra mussel densities (Lucy et al., 2013). We found the mass ratio of attached dreissenids to live unionids ranged from 0.0006 to 2.0 (mean = 0.15). Counts or weights of zebra mussels (and presence of byssal threads) on live mussels may therefore provide a surrogate abundance index for monitoring and risk assessments.

Great Lakes coastal wetlands are important habitat for amphibians and reptiles, birds, fishes and mammals (Sierzen et al., 2012). Over the past 15 years, an increasing number of studies have demonstrated that, as refuge habitats, coastal wetlands are also important for unionid conservation throughout the Great Lakes basin. We found Lake Ontario wetland mussel assemblages to be less diverse than Lake Erie and Lake St. Clair wetland assemblages (Bowers & Szalay, 2003; Zanatta et al., 2002; Crail et al., 2011) but more diverse than those recently sampled in Lake Huron and Lake Michigan (Sherman et al., 2013). Ligumia nasuta (formerly considered one of the most common species of the lower Great Lakes) was believed extirpated from the Canadian waters of Lake Ontario (COSEWIC 2007). We detected small remnant populations of this endangered species at five wetlands. Additionally, undocumented populations of Q. quadrula (Threatened) and T. parvum (Endangered) were identified at another (Jordan Harbour). These findings highlight the need for additional inventories of coastal wetlands in the lower Great Lakes and upper St. Lawrence River to properly delineate critical habitats and identify provincially significant wetlands (OMNR 2013). We recommend that these surveys be implemented using visual/tactile methods.