Introduction

Seagrasses are an important group of foundational marine species found in the shallow waters off the coasts of every inhabited continent (Orth et al. 2006). They provide numerous ecosystem services (e.g., shoreline protection, water filtration, and carbon sequestration) and serve as a vital habitat for many juvenile and adult marine organisms (Barbier et al. 2011; Lefcheck et al. 2019). Seagrass beds are home to not only many commercially important fish species (Unsworth et al. 2019), but also to a diverse group of small, epifaunal invertebrates vital to the maintenance of the ecosystem (Orth et al. 1984). These invertebrates both remove fouling algae which would otherwise overgrow the seagrass, and act as an important link in the food web between primary production, much of which is exported to other habitats (Heck et al. 2008), and higher trophic levels (Orth et al. 1984; Valentine and Duffy 2006). Understanding the factors that promote abundant and diverse epifaunal assemblages is therefore essential to the maintenance of both seagrass meadows and, more broadly, coastal food webs.

Seagrass epifaunal communities have been shown to respond to various aspects of both the habitat itself and the physical environment. For example, increased habitat complexity, either in the form of more seagrass biomass (e.g., Virnstein et al. 1984), increased shoot surface area (e.g., Sirota and Hovel 2006), or higher seagrass percent cover (e.g., Reed and Hovel 2006), has been shown to correlate with both increased epifaunal diversity and density. Several mechanisms have been proposed to explain these relationships. For example, complex canopies provide epifauna with greater protection from predators (Virnstein et al. 1984; Hovel et al. 2002; Reynolds et al. 2018) and additionally attract micro- and macroalgae that further increase habitat availability and complexity (Orth et al. 1984; Thomsen et al. 2018). The algae associated with seagrass blades (both epiphytes and drift macroalgae) can also act as the main food source for many epifaunal invertebrates (Orth et al. 1984; Duffy and Harvilicz 2001). In turn, studies have shown that increased biomass and diversity of algal resources can result in higher epifaunal densities and greater species richness (Hall and Bell 1988; Edgar 1990; Parker et al. 2001). In fact, the availability of algal resources appears to be the primary determinant of epifaunal production based on experimental manipulations (Edgar 1993; Edgar and Aoki 1993).

The abiotic environment can affect epifauna as well. For example, temperature and salinity variations have been shown to be among the best predictors of epifaunal community composition both in a single seagrass bed through time (Nelson and Virnstein 1997; Yamada et al. 2007; Micheli et al. 2008; Douglass et al. 2010) and across a region (Whippo et al. 2018; Namba et al. 2020). Low salinities can eliminate intolerant species, and extreme changes in salinity have been shown to significantly alter the community composition of epifaunal communities (Andrews 1973). Temperature can also alter composition and relative abundances, as seen in both seasonal studies (Momota and Nakaoka 2018) and experiments (Blake and Duffy 2010; Eklöf et al. 2012). Both temperature and salinity can also alter the density and biomass of seagrass itself, indirectly affecting the associated fauna by changing the amount (Baden and Boström 2001; Thomson et al. 2015) or species composition of available habitat (Richardson et al. 2018). Furthermore, regional temperatures drive additional ecological processes such as predation (Reynolds et al. 2018). Yet, while many studies have found a significant effect of these abiotic parameters on epifaunal community abundance or composition, several others have found no specific role for temperature or salinity (e.g., Walsh and Mitchell 1998; Nakaoka et al. 2001; Carr et al. 2011), suggesting that these effects are far from universal, at least at the local scale of these investigations.

Finally, geographic distance may also play a role in structuring these communities. Some seagrass epifaunal communities have been observed to vary greatly at very small scales (centimeters to meters; Virnstein 1990). Unlike many other marine organisms, many seagrass epifauna (e.g., amphipods, isopods, and some gastropods) are direct developers lacking a planktonic larval stage, meaning that juveniles emerge as miniature adults that remain the vicinity of the parents (Boström et al. 2010). This feature is somewhat rare among marine phyla, providing an interesting contrast to the extended dispersal period experienced by most larval invertebrates (e.g., 100 s–1000s km; Kinlan and Gaines 2003) that is well known to affect community structure over regional scales (Palmer et al. 1996; Chust et al. 2016). In an interesting parallel, dispersal principals derived for terrestrial plants, such as the Janzen-Connell hypothesis of seed survival (Janzen 1970), appear to also not apply to seagrass ecosystems (Manley et al. 2015). Thus, it appears general principles about the role of geographic distance in driving colonization and, ultimately, community structure, including well-established terrestrial paradigms, remain unresolved or unsupported in some marine ecosystems.

Several recent studies have attempted to resolve the drivers of epifaunal communities by integrating biotic and abiotic drivers across large distances. For example, Whippo et al. (2018) comprehensively sampled local epifauna in Barkley Sound, British Columbia, and found that while the communities could be grouped by high- and low-salinity environments, they exhibited no other predictable spatial patterns across a portion of an estuary spanning approximately 40 km2. Similarly, Stark et al. (2018) also found no evidence of spatial structuring among epifaunal communities at a scale of 1000 km along the coastline of British Columbia and found that environmental variables were only able to explain 22% of the variation in community structure between sites. Farther south along the Northwest Pacific coast, Hayduk et al. (2019) sampled seagrass beds across three different estuaries and found that upwelling played a significant role in structuring the communities across estuaries, but did not find any significant environmental drivers of community structure within a single estuary. Finally, Yeager et al. (2019) sampled 21 meadows around Back Sound, North Carolina, across approximately 20 km2 and found that measures of available habitat (i.e., fragmentation level, seagrass biomass) were the best predictors of community structure at both landscape and microhabitat scales. Thus, as with local variability, there remains little consensus on what individual factors structure epifaunal communities across regional scales (100–1000 km), despite well-resolved regional drivers for other marine communities (e.g., phytoplankton; Rutherford et al. 1999).

In the current study, we surveyed epifauna across the lower Chesapeake Bay (Fig. 1) and related patterns in community properties to geographic distance, habitat characteristics, resource availability, and the abiotic environment. The Chesapeake Bay is the largest estuary in the United States and among the most environmentally and ecologically dynamic on Earth (Najjar et al. 2010), especially with respect to seagrass habitat (Lefcheck et al. 2017). It also has a long history of foundational research on seagrass invertebrates (Marsh 1973; Neckles et al. 1993; Duffy and Harvilicz 2001; Douglass et al. 2010; Lefcheck et al. 2016). We predicted that increased habitat complexity, provided by higher levels of seagrass cover, and increased food availability, provided by algal resources, would result in overall higher levels of epifaunal abundance, biomass, and species richness, as found in some previous studies (Yeager et al. 2019). Furthermore, we predicted that due to the limited dispersal potential of direct developers, sites located closer to each other would harbor more similar epifaunal communities. Finally, we predicted that sites with more similar abiotic environments, measured by water temperature and salinity, would have more similar communities (Douglass et al. 2010).

Fig. 1
figure 1

A map of study sites in the lower Chesapeake Bay estuary. Each symbol marks the location of one site, while the color and shape of the symbols correspond to the geographic region each site is located within. The inset denotes the location of the Chesapeake Bay along the Atlantic coast of the USA

Full size image

Methods

Study Sites

We collected epifaunal, seagrass, and environmental data at 16 seagrass beds dominated by eelgrass (Zostera marina) located throughout the lower Chesapeake Bay (Fig. 1). Seagrass cover has been regularly surveyed at our sampling locations since 2006 (Richardson et al. 2018) and our sampling design corresponded with the major areas of eelgrass cover throughout the region, meaning that certain areas were more densely sampled due to more extensive cover, while some areas were unsampled due to large unvegetated expanses. Sampling occurred over 3.5 weeks in June and July of 2018, coinciding with the peak summer eelgrass biomass. To account for the distance between study sites in our analysis, we calculated the minimum overwater distance between each pair of sites.

Seagrass Cover Data

We conducted a single transect survey at each site to quantify the cover of each species of seagrass throughout the bed. Transects ran perpendicularly out from the shoreline to the offshore edge of the bed and ranged in length from 260 to 1330 m. Every 10 m along each transect, snorkelers estimated the percent cover of seagrass within a 1.0-m2 PVC quadrat by using 10% bins ranging from 5 to 95%. For our analysis, we calculated the median percent cover along each transect.

Epifaunal and Macrophyte Data

We collected epifaunal samples at haphazardly chosen points along the transect at each site. We used a flexible mesh bag (500-μm mesh size, 75 cm deep with a 20-cm-diameter round opening capturing approximately 0.03 m2 of bottom area) to enclose a cluster of eelgrass blades underwater, then cut the blades at the sediment-water interface, and washed all the seagrass and associated algae and animals into the bag by gently inverting it while holding the opening at the top of the bag closed. The bag was then tied shut, placed on ice, and taken back to the lab. Four of the seagrass beds sampled were very extensive and therefore contained two transects. At these locations, we took n = 8 mesh bag samples at each transect. At all other transects, we collected n = 16 samples.

In the lab, we emptied the mesh bags, scraped any epiphytes and sessile invertebrates off the seagrass blades and then identified them to the lowest taxonomic level possible. Any free macroalgae collected in the sample were identified as well. We then dried the seagrass, macroalgae, total epiphytes, and all sessile invertebrates at 60 °C for 5 to 7 days and weighed them all to the nearest mg. We used the sum of the dry weights of the macroalgae, sessile invertebrates, and epiphytes as an indicator of total resource biomass, as epifauna are known to consume all three.

Epifauna were preserved in 70% ethanol and later passed through a series of nested sieves of increasingly smaller mesh size, identified to the lowest taxonomic level possible, and enumerated. Empirical equations relating size class to mass were used to estimate biomass (mg ash-free dry weight) of each species (Edgar 1990). In our analysis, we removed any exceptionally rare species (fewer than 5 individuals across all samples), unidentified animals, or species that due to their natural history were likely not actually epifaunal (e.g., infauna). Finally, we scaled both epifaunal biomass and abundance by the total seagrass biomass in each sample (mg ash-free dry weight mg−1 seagrass) to account for differences in the amount of seagrass sampled in each bag.

Environmental Data

To collect temperature data, we set out HOBO data loggers (ONSET model no. UA-002-64) adjacent to the transects at seven sites, which recorded water temperature every 30 min. For sites without loggers, we used either the data from the nearest site or the nearest biweekly water quality monitoring station from the US Environmental Protection Agency’s (EPA) Chesapeake Bay Program (CBP) database (www.chesapeakebay.net/data/downloads/cbp_water_quality_database_1984_present), whichever was closer (following Richardson et al. 2018). We also collected salinity data from the EPA CBP database. In our analysis, we used the average water temperature and salinity from the 30 days prior to each sampling date.

Statistical Analyses

Before running our analyses, we calculated several indicators of epifaunal community structure: total abundance, total biomass, species richness, evenness, and multivariate composition. We calculated richness as the number of species present across all samples at a site. To quantify evenness, we first transformed Shannon diversity into Hill numbers by taking the natural logarithm. We then divided this number by the natural logarithm of species richness, following the formula in Jost (2010) for calculating Pielou’s evenness. To capture differences in multivariate community composition (as Bray-Curtis dissimilarities), we used the vegan package (Oksanen et al. 2018) in R (version 3.6.2, R Core Team 2019). To visualize these quantities in two dimensions, we performed nonmetric multidimensional scaling (NMDS) on square root–transformed biomass using the vegan package (Oksanen et al. 2018). For our main analyses, we calculated differences in community composition using the average biomass and abundance (Bray-Curtis dissimilarity) and presence-absence (Jaccard dissimilarity) of species at each site. We emphasize results based on biomass in our presentation, as seagrass epifaunal species are primarily valued for their contributions to secondary production and trophic transfer, of which biomass is a more relevant measure.

To disentangle the roles of different environmental factors in structuring epifaunal communities, we performed multiple regressions on distance matrices (MRMs) using the ecodist package in R (Goslee and Urban 2007). First, we created site-by-site distance matrices for all of our community-level response variables (epifaunal species abundance, biomass, richness, Pielou’s evenness, and community composition) and our environmental explanatory variables (overwater distance, temperature, salinity, seagrass cover, macrophyte biomass, and sampling date). In most cases, these were Euclidean distances (i.e., the difference in temperature, salinity, etc. between each pair of sites) except for community composition (see above). We then regressed each response matrix against all explanatory matrices, conducting one MRM for each dependent variable.

In MRMs, relationships are assessed by holding covarying matrices constant and computing the partial matrix correlation between the explanatory and response matrices (i.e., the Mantel correlation). Significance is assessed by randomly permuting the explanatory matrix, recomputing the matrix correlation with the response, and repeating 10,000 times to create a distribution of correlations. If the observed correlation is greater than expected by chance, as determined by a pseudo-t test (α < 0.05), we interpreted that explanatory variable as being significantly correlated with the response. The benefit of a matrix-based approach is that inherently multivariate properties can be analyzed without first being collapsed into a single vector (e.g., an axis from the NMDS), which results in a greater loss of information (Legendre et al. 1994; Lichstein 2007). This property makes MRMs particularly relevant to our application, as we had both pairwise multivariate explanatory (intersite distances) and response (community composition) variables. However, the use of distances means that care must be taken in the interpretation of the correlation, namely as follows: as two sites become more dissimilar in terms of the predictor matrix (temperature, salinity, etc.), then they are expected to become more divergent in terms of their richness, composition, etc.

To better understand which species were driving associations with seagrass cover in our MRM analysis, we conducted an indicator species analysis. To do this, we first collapsed our seagrass percent cover measurements at each site into 5 bins (< 5%, 5–25%, 26–50%, 51–75%, 76–100%) following the methods in Braun-Blanquet (1965). We then used multilevel pattern analysis as implemented in the indicspecies package (De Caceres and Legendre 2009) to explore which species were most associated with different levels of cover. This analysis calculates an indicator value as the proportion of the total individuals of a given species that occur in a given cover level across all sites, multiplied by the proportion of the total sites in each cover bin that the given species occurred at (Dufrêne and Legendre 1997). A permutation test (with 999 permutations) is then run to calculate the significance of each observed indicator value. We then reran our MRM analysis excluding the species significantly associated with specific cover bins, to examine which factors may be structuring the remainder of the community when the strong signal of these species was removed. Finally, to test how distance may influence the populations of species with different developmental modes, we reran all analyses after subsetting the species present into direct and planktonic developers.

Results

Community Patterns

In total, we analyzed n = 190 samples across 16 sites, containing 19,719 individuals from 18 taxa. Species richness at each site ranged from 9 to 16 species, with 6 species present at every location (the polychaete Alitta succinea, gastropod Bittiolum varium, caprellid amphipod Caprella penantis, isopod Erichsonella attenuata, and gammaridean amphipods Elasmopus levis and Gammarus mucronatus). Three species were responsible for more than half (60%) of the total biomass collected: E. attenuata, C. penantis, and the isopod Idotea baltica. The NMDS analysis revealed no consistent spatial patterns in community composition (calculated as Bray-Curtis dissimilarities using biomass; Fig. 2).

Fig. 2
figure 2

Results from nonmetric multidimensional scaling (NMDS) on average biomass at each sampling site showed no consistent geographic patterns. Each round, filled-in circle corresponds to a single sample, while the larger symbols correspond with the centroid (or mean) of each site. The color and shape of these symbols corresponds with the geographic region (see Fig. 1)

Full size image

Environmental Drivers

The results of the MRMs revealed several significant correlations between biotic and abiotic factors and different aspects of the epifaunal communities (Table 1). We found that differences in average seagrass cover and macrophyte biomass were significantly correlated with dissimilarity in community composition (as measured by Bray-Curtis dissimilarities calculated using biomass; P < 0.001 and P = 0.039 respectively), while temperature differences were the only significant predictor of differences in species richness (P = 0.023). Finally, sampling date differences were a significant predictor of abundance, biomass, and community composition differences (P = 0.0013, P < 0.001, P = 0.033). Beyond date and temperature, none of the predictors was significantly correlated with abundance, biomass, or richness, reflected in their relatively low proportions of variance explained (Table 1).

Table 1 Results of multiple regressions on distance matrices (MRMs) for various aspects of the epifaunal community. Values are partial Mantel correlations with respective P values based on a permutation procedure in parentheses. Results significant at the α < 0.05 level are italicized
Full size table

In contrast to our initial expectations, geographic distance did not explain trends in any community property when biomass, abundance, or presence/absence data were used to calculate community composition (P > 0.05 in all cases; Tables 1 and 2), as previously established in our NMDS analysis (Fig. 2). Overall, our different models using biomass, abundance, and presence/absence data to calculate community composition explained 35%, 26%, and 11% of the variation in the data, respectively (Table 2).

Table 2 Results of multiple regressions on distance matrices (MRMs) for different measurements of community composition. Interpretation is the same as Table 1
Full size table

Species-Specific Drivers

The results of our indicator species analysis revealed that one species, the caprellid amphipod Paracaprella tenuis, was significantly associated with the intermediate 26–50% cover of seagrass (P = 0.001), while two species, E. attenuata and E. levis, were significantly associated with the highest observed 51–75% cover bin (P = 0.004, P = 0.045). When we removed these species and repeated the analysis to test how much of the associations we recovered were due to their presence, the effect of seagrass cover was no longer significant (P = 0.35), suggesting that the significant effect of cover on regional differences in community composition was due to these three species. Finally, when we subset our data by development type and reran our analyses, we found a significant effect of overwater distance on community composition of direct developers (Table 3; P = 0.036), but not on community composition of planktonic developers (Table 3; P = 0.87).

Table 3 Results of multiple regressions on distance matrices (MRMs) for the community composition (calculated using epifaunal biomass and Bray-Curtis dissimilarities) of different aspects of the epifaunal community split by development type. Interpretation is the same as Table 1
Full size table

Discussion

In our study of seagrass invertebrates across the lower Chesapeake Bay, we found that habitat complexity, resource availability, and, for a subset of species (i.e., direct developers), geographic distance were all significantly associated with variation in biomass-weighted epifaunal community composition. Our significant results and relatively high proportion of variance explained for community composition (Table 2; R2 = 0.35) stand in contrast to other studies, which did not find any significant biotic or abiotic predictors of epifaunal community properties at this scale (e.g., Stark et al. 2018; Yeager et al. 2019). We believe our results are due to three main reasons: (1) we focused our study on a relatively low-diversity, semi-enclosed estuary, the Chesapeake Bay, which provided a more tractable seascape in which to work and analyze; (2) the Bay is highly heterogenous with respect to abiotic variables and seagrass habitat; and (3) we comprehensively and simultaneously sampled many variables, allowing us to account for multiple covariates at once in our analyses. Ultimately, it appears that the composition of these assemblages is largely a factor of seagrass cover, highlighting the vital role of seagrass as a habitat in coastal regions (Beck et al. 2001; Lefcheck et al. 2019).

In contrast, we found no significant effect of seagrass cover on epifaunal abundance, biomass, or richness, suggesting that cover’s effects on community structure are not a consequence of increasing the density or diversity of individuals. Rather, cover appears to directly alter the composition independently of these other properties, in contrast to past studies (Stoner 1980; Virnstein et al. 1984; Hovel et al. 2002). We have shown that, in the present study, this effect is a function of just three species of peracarid crustaceans, the isopod Erichsonella attenuata, the caprellid amphipod Paracaprella tenuis, and the gammaridean amphipod Elasmopus levis, which may respond uniquely to changes in their seagrass habitat.

Initially, we hypothesized that two communities that were closer to each other in space would also have more similar communities, due to the fact that many of the taxa we sampled are direct developers, meaning they do not have the planktonic larval stage key to the dispersal of many marine invertebrates (Cowen and Sponaugle 2009). However, certain members of the epifaunal community do have planktonic larval stages (e.g., gastropods, shrimp, polychaetes), so, to determine the degree to which larval development affected this finding, we reran our analysis subsetting the community into direct and planktonic dispersers. We found that sites that were closer together in geographic space were also more similar in their community composition of direct developers. The mechanism behind this result seems clear: as direct developers produce juveniles that emerge as miniature adults, they have a reduced capacity to disperse widely and therefore stay in the same vicinity of the adults, reinforcing local community composition.

Beyond cover, we found that algal biomass also significantly affected epifaunal community composition (when calculated with biomass). Micro- and macroalgae are the main source of food for most of the taxa we surveyed; thus, algal biomass can be equated to the available food resources of a bed (Orth and Van Montfrans 1984). Furthermore, similar amphipod taxa in terms of body size and resource limitation can have very different feeding preferences (Duffy and Harvilicz 2001), perhaps influencing where they are found within a seagrass bed or region. Other studies of epifaunal communities have found similar results to ours, linking algal resources to community composition at smaller scales (Hall and Bell 1988; Morales-Núñez and Chigbu 2019).

Further, we found that temperature was significantly and negatively associated with community richness, suggesting that sites with higher temperatures had fewer species. In temperate regions, models have predicted that warmer temperatures, in combination with an increase in temperature variance, will significantly decrease performance (oftentimes measured in terms of fitness) of invertebrates (Vasseur et al. 2014). While the temperature gradient we observed was relatively small (16.1 to 17.7 °C), perhaps the minor changes from site-to-site were just at the threshold of a few species’ preferred levels of heat tolerance, leading to their exclusion in the hotter areas. Alternately, temperature may be interacting with other unknown factors, such as water clarity (which we did not measure), to indirectly influence habitat characteristics. A recent study has shown that higher temperature in combination with low clarity lead to greater dieback of eelgrass in Chesapeake Bay than expected by either alone (Lefcheck et al. 2017).

Sampling date was a significant predictor of abundance, biomass, and community composition (calculated using biomass) across both analyses. While our samples were all taken over a relatively short period of time (3.5 weeks from mid-June to early July), epifaunal reproduction peaks in the summer (Douglass et al. 2010), around the same time that fish predators reach their highest levels of abundance (Orth and Heck 1980). This confluence of events can result in very dynamic epifaunal population sizes in the span of weeks to months (Fredette and Diaz 1986). However, by including sample date in our models, we were able to control for at least some of its potentially confounding effect on epifaunal community properties.

The three species identified in our indicator species analysis—E. attenuata, E. levis, and P. tenuis—may possess adaptations that allow them to be influential in determining the relative composition of epifaunal communities across different habitat conditions in the Chesapeake Bay. In many coastal systems, mesopredatory fishes and crabs have been found in higher densities in seagrass beds with higher levels of cover (Ralph et al. 2013). E. attenuata has previously been shown to be effective at avoiding fish predation, potentially due to its larger body size and distinct morphology, which may make it a challenge to be physically manipulated and consumed by predators (Douglass et al. 2010; Moore and Duffy 2016). Moreover, its slender shape allows it to easily blend in with thick grass as well as be almost undetectable when attached vertically to seagrass blades (Main 1985; Douglass et al. 2007), potentially explaining why E. attenuata was most prevalent in denser beds. When given the choice between a habitat with more food, or one with more structure, both E. attenuata and Idotea baltica (the second most common isopod we encountered in our study) chose the food (Boström and Mattila 1999). However, when presented the same choice in the presence of a predator, I. baltica still preferred the greater food resources, while E. attenuata switches its preference to the more complex habitats (Boström and Mattila 1999). This predator avoidance behavior may also explain why E. attenuata was so strongly associated with high cover areas, while other isopods with similar body shapes were not.

P. tenuis has a long cylindrical body, which, like E. attenuata, may also allow it to blend in better in dense beds, avoiding potential predators. Interestingly, however, the other caprellid amphipod observed in our survey, Caprella penantis, was not significantly associated with high cover in our indicator analysis. P. tenuis, which typically prefers to live near bryozoans, sponges, and hydroids, has been described as more of a specialist than C. penantis, which has historically been widespread throughout the Chesapeake Bay (McCain 1965). Perhaps beds with higher levels of cover also support higher levels of the sessile invertebrates that P. tenuis prefers due to the increase in total leaf surface area. Although, in our samples, total fouling biomass on the blades was only weakly correlated with cover (r = 0.25). However, for ease of processing, we did not distinguish between sessile invertebrates and other periphyton when calculating fouling biomass.

P. tenuis is also an ambush predator and has been observed defending itself from potential attacks (Caine 1989; Caine 1998), while C. penantis is a less aggressive filter feeder (Caine 1974). This distinction may have allowed P. tenuis to avoid the higher levels of predation in dense beds better than C. penantis could. This trait may also help explain why P. tenuis is associated with beds characterized by intermediate levels of cover (26–50%) while E. attenuata and E. levis are found in higher cover areas (51–75%). While being defensive may relieve some predation pressure, ultimately, small amphipods and isopods likely will not be able to fight off every large fish. Thus, camouflage may be a better antipredation strategy in this system as an increase in seagrass cover will only serve to decrease detectability, allowing for cryptic species to thrive in the areas with the densest seagrass cover (and by extension, likely the heaviest levels of predation pressure).

Unlike E. attenuata and P. tenuis, E. levis is more similar to other amphipods in our dataset that were not significantly associated with very high levels of cover. A study of predation rates on three different species of gammarid amphipods found that E. levis was attacked at significantly lower rates than the other two species, seemingly due to its lack of pigment, making it relatively more difficult for the visual fish predators to find (Clements and Livingston 1984; Ryer 1988). This coloration of E. levis may have allowed it to camouflage better than other gammarid amphipods in the areas with the densest seagrass cover and, thus, potentially the highest levels of predation. In summary, the natural history of these three particular species may have significant implications for how they are distributed through space, and ultimately their contributions to epifaunal community structure across the region.

Overall, we explained one-third of the regional-scale variation in epifaunal community composition with a model including seagrass cover, food resources (macroalgal and epiphyte biomass), and, for a subset of species determined by development mode, spatial structure (overwater distance). No predictors other than date explained epifaunal community abundance or biomass, suggesting that these community changes were not a function increasing numbers of individuals/biomass. Rather, we propose that certain species thrive in areas of high seagrass cover, largely due to their ability to avoid predators and exploit resources. Incorporating information about relative dispersal abilities and top-down control into future investigations may explain even more of the variance between seagrass epifaunal communities.