INTRODUCTION

Seagrasses and bivalves are ecosystem engineers in shallow marine waters, creating habitat structure and modifying abiotic conditions and resource availability (Gutiérrez et al. 2011). Yet their environmental effects are unlikely to be functionally equivalent because of differences in architecture, chemical composition, flexibility, and trophic level. Seagrasses frequently are managed for protection and restoration (van Katwijk et al. 2016), whereas bivalves additionally are targets for aquaculture and fishing (Beninger & Shumway 2018, Bersoza Hernández et al. 2018). Because the two taxonomic groups often co-occur on soft sediments, interventions on the part of one taxon potentially affect the other, such as facilitation during joint restoration (Gagnon et al. 2020). How do these two ecosystem-engineering taxa reciprocally interact? This quantitative review summarizes evidence from field experiments testing bivalve–seagrass interactions, in either direction, as well as surveys of bivalves inside and outside seagrass habitats.

Seagrass could affect bivalves negatively or positively through several pathways, including both structural attributes of seagrass and its biological activity as a primary producer (Fig. 1A). Furthermore, a single pathway may have different effects depending on functional traits of the bivalve. In general, epifaunal invertebrates respond positively to aboveground structure provided by seagrass, whereas infaunal invertebrates may be inhibited (Stoner 1980). Aboveground structure could be particularly beneficial for surface-dwelling or epifaunal bivalves under high-flow conditions where structure increases sediment stability, protects them from dislodgment, and does not result in long water residence times.

In some pathways, seagrass structure modifies trophic interactions of bivalves (Fig. 1A). Considering top-down effects, structure can exclude predators or reduce their foraging effectiveness both above- and belowground (e.g., Reise 1977, Peterson 1982). For instance, thin-shelled and surface-dwelling bivalves reached greater densities in seagrass than mud habitats (Glaspie & Seitz 2017), and vegetation shifted communities from endobenthic to epibenthic because of providing habitat and protecting from predation above sediment (Bouma et al. 2009). Alternatively, predators might be attracted to seagrass, enhancing the strength of top-down effects (Lowe et al. 2019). From the bottom-up effect, seagrass structure characteristically reduces water flow and food delivery (e.g., current speeds reduced by half; Fonseca et al. 2019), which could stunt bivalve growth or condition, although particulate matter may also be enhanced in seagrass beds (Judge et al. 1993, Ruesink et al. 2019). Through this bottom-up pathway, responses of bivalves could depend on the functional group, with vegetation harming filter feeders through reduced food delivery, but favoring deposit feeders through organic matter accretion (Bouma et al. 2009). In another study, deposit feeders were more abundant outside than in vegetation (Glaspie & Seitz 2017), emphasizing that functional group responses may be context-dependent. An emerging paradigm is for seagrass to foster bivalve survival but reduce growth (Carroll & Peterson 2013).

Another species interaction that may drive bivalve response to seagrass is through an intermediate bioturbator species. Bivalves could be facilitated not only by seagrass directly but also by seagrassÕ ability to exclude bioturbators. Accordingly, sandflats in the Knysna estuary (South Africa) without bioturbating shrimp had as many or more macrofauna relative to nearby seagrass (Barnes & Barnes 2014).

Figure 1.

Pathways of interactions among ecosystem engineers, distinguishing roles of physical structure, and biological activity. (A) Seagrass effects on bivalves. Explaining the grey arrows, predators and resource delivery are generally considered to be reduced in seagrass (inhibition), but counterexamples exist (facilitation). (B) Bivalve effects on seagrass. Explaining the grey arrow, bivalves may outcompete seagrass for space (inhibition), but structure could improve colonization of early life stages (facilitation). Indirect effects derive from two sequential arrows in a pathway of interaction.

Unlike the inconsistent predicted effects of seagrass structure on bivalves, most pathways involving biological activity of seagrass are expected to be facilitative (Fig. 1A). Primary production can enhance secondary production, as indicated by the suggestion that flagellates consumed by oysters were supported by “the decomposition of a vast amount of eelgrass” (Imai et al. 1950). In daytime, oxygen released from seagrass photosynthesis is available for bivalve respiration, and drawdown of carbon dioxide could improve pH. Nevertheless, whether conditions for calcification are improved or primarily made more variable by seagrass remains an open avenue of research (Koweek et al. 2018). Lucinid and solemyid bivalves are particularly likely to show a trophic facilitation by seagrass because of organic matter enrichment and sediment sulfide that their symbiotic gill bacteria use, shunting some of the energetic benefit to their host (Reynolds et al. 2007, van der Heide et al. 2012).

The reciprocal interaction, in which bivalves are the effector and seagrass the responder species, likewise provides both facilitative and negative pathways that could govern the overall interaction (Fig. 1B). Space occupied by bivalves is unavailable to seagrass, a negative effect that may appear primarily at high bivalve densities (Wagner et al. 2012). Beyond spatial interference, structural aspects of bivalves could benefit seagrass. Structure could catch seeds or anchor seagrass at early life stages, especially in high-flow conditions, or harbor invertebrates that remove epiphytes from seagrass. Biological activities of bivalves are expected to modify water or sediment conditions in ways that benefit seagrass, for instance by particle filtration, thereby improving water clarity and light availability, or by nutrient release; however, biodeposits that fertilize plant growth may become detrimental to seagrass in excess, including reduced growth rate from sulfide toxicity or competitive epiphytes (Vinther & Holmer 2008).

In any particular bivalve–seagrass (or seagrass–bivalve) interaction, the outcome reflects the relative importance of a variety of pathways (Fig. 1). This meta-analysis intends to move beyond single cases to understand the weight of evidence across multiple bivalve–seagrass systems. The expectations are given as follows:

MATERIALS AND METHODS

Meta-Analysis of Manipulative Studies

Manipulative studies were required to have a treatment in which bivalves and seagrass were colocated, and a reference in which only the responding taxon was present. Studies in which recruitment was measured in and out of the effector taxon were included, although these were not strictly manipulative because neither the effector nor responder was added or removed by the researchers. Study designs were checked for true replication, which eliminated some studies where multiple plots were tracked in and out of one large patch of the effector taxon; if this design was established at multiple sites, sites were considered samples to calculate a grand study mean and SE. Studies that had artificial treatments (i.e., artificial seagrass and empty shells), involved bivalve aquaculture methods, or concerned habitat edges, instead of presence/absence of the effector, were excluded.

Figure 2.

Flow diagram for identification and screening of articles used in meta-analysis. Two publications included both seagrass and bivalves as responders (Reusch & Chapman 1995, Tsai et al. 2010). Three publications included both observational and manipulative components to study bivalve response to seagrass (Peterson 1982, Reusch & Williams 1999, Ruesink et al. 2014).

TABLE 1.

Responses measured in studies of reciprocal seagrass–bivalve interactions.

Many different response metrics were recorded across studies, which were categorized as growth, size, density, survival, and recruitment. Diet, isotope, physiological, or transcriptome responses were not included. Studies often reported more than one measurement for each metric (i.e., growth per day and size corrected growth per day), but only one measurement per metric was retained. For duplicate metrics, the most prevalent measure based on the hierarchy in Table 1 was retained. Treatment conditions were considered to include both seagrass and bivalve, colocated, whereas reference conditions contained only the responder taxon. From each article, mean, measure of dispersion [SD, SE, or confidence interval (CI)], and sample size were extracted for the responder taxon in both treatment and reference conditions. Duration since the beginning of the experiment was recorded, including any repeated measures during the experiment. All studies included information about site location and effector/responder species identity, and for some, effector density was quantified. These values were subsequently used to calculate standardized effects as HedgesÕ d with a small sample size correction (sometimes referred to as HedgesÕ g, Gurevitch et al. 1992, Viechtbauer 2010).

Here, and are treatment and reference means. S is a pooled SD including treatment (n_e) and reference (n_c) sample sizes and treatment (s_e) and reference (s_c) SD. J is a correction factor that removes small sample size bias. Values overlapping zero are nonsignificant, negative values indicate inhibition, and positive values indicate facilitation.

Separate meta-analyses were performed for each category of response metric in seagrasses and bivalves, using the metafor package in R version 3.5.2 (Viechtbauer 2010, R Core Team, 2017). The meta-analysis included 25 bivalve response studies and 11 seagrass response studies (Fig. 2, Appendix 1), with two studies of reciprocal effects germane to both. No publication bias was found via funnel plot asymmetry in EggerÕs regression tests in bivalve responses (t₈₁ = 1.05, P = 0.30) or seagrass responses (t₄₄ = –1.88, P = 0.07). Many studies reported repeated measures, but only the final duration was used in the main analysis, which is a standard practice in ecological meta-analyses (Gurevitch et al. 1992). Based only on final samples, overall effect size was estimated for each response metric category in a linear mixed-effects model with study as a random effect. In addition, all time points and all metrics were included in a linear mixed-effects model with duration as a fixed effect and study as a random effect. A separate meta-analysis was carried out for seagrass density response, including only the five studies with information about bivalve density, which was considered a fixed effect. Studies that reported densities of bivalves as individuals per area were converted to percent cover (density multiplied by individual shell area) to make all densities comparable.

RESULTS

Surveys of Bivalves In and Out of Seagrass

Based on the grand mean log response ratio, bivalves were 1.6 times more abundant in than out of seagrass (log10 (in/ out) = 0.215 ± 0.096 SE, t_1,104 = 2.24, P = 0.03). Bivalve response to seagrass did not show any significant interaction between body size and functional group (LRT F_3,97 = 0.15, P = 0.93), nor a significant main effect of body size (LRT F_1,100 = 0.94, P = 0.33). Response to seagrass differed by functional group (LRT F_3,101 = 6.0, P = 0.0008), primarily because almost all cases of mutualists (lucinid or solemyid clams) showed substantially greater densities in than out of seagrass (Fig. 7). Seagrass clearly facilitated mutualists [log10 (in/out) = 1.376 ± 0.353 SE, t_1,8 = 3.9, P = 0.005] and nonsignificantly augmented epibenthic bivalves [log10 (in/out) = 0.350 ± 0.176 SE, t_1,24 = 1.99, P = 0.059]. Shallow burrowers were overall not responsive to seagrass [log10 (in/out) = 0.019 ± 0.120 SE, t_1,53 = 0.16, P = 0.88], nor were deep burrowers [log10 (in/out) = 0.024 ± 0.246 SE, t_1,16 = 0.10, P = 0.92].

Figure 3.

Map of study sites, (A) global, (B) United States. Experiments used in the meta-analysis are represented by open circles, and surveys are represented by crosses. Experimental studies are concentrated in North America. Surveys are more widespread than experiments.

DISCUSSION

The present review detected minimal evidence of consistent reciprocal interactions between bivalves and seagrasses with some notable exceptions. Bivalves had higher survival in seagrasses in experimental studies (hypothesis 1) and greater abundances of some functional groups in seagrass habitats (hypothesis 3), and seagrass showed some evidence of negative effects of spatial competition from bivalves at high cover (hypothesis 2) and appeared more negatively affected by bivalves over time; however, generalizations about the outcome of interaction between these two ecosystem engineers when they are colocated cannot be made, but rather require case-specific information. This conclusion follows from the weak overall effect sizes, although particular studies and metrics can reveal strong facilitative (Irlandi & Peterson 1991, Irlandi 1996, Goshima & Peterson 2012) or negative interactions (Booth & Heck 2009, Wagner et al. 2012). The results of the present review are reassuringly similar to another recent review that casts a wider net to evaluate plant–bivalve interactions (Gagnon et al. 2020). They, 225 publications covering not just seagrass but also mangroves, salt marsh, and freshwater plants, were evaluated in a vote-counting approach to determine the percentage of studies with positive, negative, mixed, or only nonsignificant responses. These 225 publications provided 491 studies, of which 316 (64%) involved seagrasses. Overall, positive interactions occurred in 51% of studies and negative in 25%, with pathways similar to those suggested in Figure 1. Among seagrass–bivalve studies, those with epibenthic and sulfide-oxidizing bivalves were especially prone to positive interactions (Gagnon et al. 2020). As Gagnon et al. (2020) noted, they were restricted to a vote-counting approach because of the wide variety of methods applied in the studies in their review, for instance, both colocated and spatially distinct arrangements of plants and bivalves, modern and paleontological time frames, and laboratory and field studies. Of the 34 publications suitable for the quantitative meta-analysis reported here, 20 overlapped with Gagnon et al. (2020), and for surveys, the number overlapping was 19 of 39. Therefore, only about half of the articles in the present review were previously covered by Gagnon et al. (2020), with the present review also smaller because of stricter inclusion criteria for meta-analysis. The present review expands on Gagnon et al. (2020) by using a quantitative meta-analytic approach and highlights the different responses (survival, growth, reproduction, and density) that could occur in a reciprocal interaction.

TABLE 2.

Meta-analysis of bivalve responses to seagrass and seagrass responses to bivalves.

Comparison of the present review with others covering seagrass and marine animals also helps focus on lessons learned here. One widely examined ecological effect of biogenic habitats is their provision of nursery habitat, specifically defined when juveniles, rather than adults, are found in the habitat (Beck et al. 2001). Seagrasses provide nursery habitat for a wide variety of fish and invertebrates through aboveground structure (Heck et al. 2003, McDevitt-Irwin et al. 2016), but no examples of bivalves appeared in these reviews, probably because of the sedentary lifestyles of bivalves that preclude moving out of seagrass as adults. Because these reviews of nursery habitat included no bivalve responses, there is no overlap with the present evaluation of bivalve–seagrass interactions. Whereas juvenile fish and invertebrates generally were enhanced in seagrass nursery habitat, the results here showed bivalve enhancement by seagrass was primarily driven by species with sulfide-oxidizing bacteria (Fig. 7). Bivalves may not be consistently facilitated because many live infaunally or create their own structure and are less dependent on aboveground structure than mobile species in the water column. Another prior meta-analysis found that density and demography of seagrasses were generally negatively affected by bivalve shellfish aquaculture (Ferriss et al. 2019). Of 28 publications reviewed by Ferriss et al. (2019), five included field experiments without aquaculture gear and in small planted plots, and these experiments also appear in the present review addressing bivalve effects on seagrass. The generally negative effects of farming bivalves in seagrass (Ferriss et al. 2019) were not evident in the present review addressing the species interaction alone, without aquaculture gear or disturbance from farming.

The wide variety of potential mechanisms involved in reciprocal interactions between seagrasses and bivalves underpins a lack of consistency in whether these interactions are facilitative, negative, or neutral. For example, some pathways of interaction may be positive and other pathways may be negative creating an overall net neutral effect (see Fig. 1). In principle, biological activity tends to exert positive effects reciprocally, but these magnitudes could be quite small (Bruno et al. 2003). In addition, the relative importance of negative and positive interactions could shift with spatial scale. For example, endangered pen shells in the Gulf of Oristano (Italy) reached highest density in the region of the gulf without seagrass (Addis et al. 2009), but at smaller scales, were denser in seagrass than in interspersed sand patches (Coppa et al. 2019). Meanwhile, space competition occurs at small spatial scales as well as recruitment facilitation (Mikkelsen et al. 2004). At larger scales, facilitation of seagrass by bivalves may be mediated through filtration or wave attenuation and improved water clarity, with evidence primarily from models (Newell & Koch 2004, Smith et al. 2009). Thus, although the current review deliberately focuses on consequences of colocation (i.e., plot or quadrat scale), seascape-level considerations of distance to edge, patch size (Bowden et al. 2001, Mills & Berkenbusch 2009), and habitat composition of the neighborhood are needed. Gagnon et al. (2020) found more positive effects between bivalves and plants when they were in adjacent areas than when colocated.

Density of the effector species can play a large role in how another species responds to their presence. Effector density was often not provided in published studies so could not be applied as a predictor in the full meta-analysis, except for the effect of bivalve density on seagrass density, but it is reasonable to expect that interaction strength could vary with density. In some studies, low-density outplants were a method that was deliberately used to avoid space competition (Valdez et al. 2017). In other studies, density treatments were explicitly tested; Booth and Heck (2009) showed that seagrass density and growth responded more negatively to higher oyster density treatments. In the future, studies on seagrasses and bivalves should consider a range of densities, rather than comparing just presence/absence of effector species.

The articles included in the meta-analysis of bivalve–seagrass interactions included many different treatments (e.g., patch size, density, tidal height, and cages), which likely increases variability in outcome both within and among studies. Laboratory and mesocosm studies were not included in this meta-analysis but were less common than field studies. The overall findings of the meta-analysis are most applicable to the United States, especially Washington state and the Carolinas, where most of the experiments were performed. Further experimental research is needed, especially in the tropics. Some metrics were much better represented than others, for example, there were no experimental data on the survival of seagrasses and very little on seagrass reproduction when colocated with bivalves; these metrics warrant further investigation. This review did not cover chemical or genetic metrics, where pathways mediated by the trophic facilitation of consumers by primary producers might be revealed, for instance in tissue nutrient or stable isotope ratios (Hori et al. 2019, Lowe et al. 2019).

Figure 4.

Responses of bivalves to seagrasses (left) and /seagrasses to bivalves (right), with studies ordered within each panel from facilitative to inhibitory interactions. Studies with multiple points reflect multiple experimental treatments (e.g., density, cages, etc.). Responses are given as mean effect sizes (HedgesÕ d) and 95% CI for density, growth, recruitment, size, and survival. Seagrasses and bivalves have little overall directional effect on each other in experimental studies.

Figure 5.

Responses in seagrass-bivalve experiments according to study duration. Responses are given as mean effect sizes (HedgesÕ d) and 95% CI, points are slightly jittered. Few studies included measurements after more than a year. For studies with multiple time points, all were included in this figure and analysis. (A) Bivalves show no change in effect sizes over time. (B) Seagrass shows a slight negative effect of bivalves on seagrass responses over time.

Figure 6.

Seagrass density across treatments differing in bivalve density, showing a slight decrease in seagrass density with increased bivalve density. Responses are given as mean effect sizes (HedgesÕ d) and 95% CI.

Substantial sampling effort has gone into reporting bivalves across seagrass-vegetated and unvegetated habitats because surveys were more prevalent than experiments (Fig. 2). Yet the survey dataset only accommodated studies with species-specific data, so missed cases reporting bivalve diversity or a multivariate response, as well as publications of macrofauna that did not distinguish the bivalve component. Habitat specialists have been defined when species are at least 2.5 times more abundant in one habitat than another (Barnes & Hamylton 2013). Based on this criterion, of the 105 bivalve species in a seagrass-bare comparison at a site, 39 were generalists, 40 were seagrass specialists, and 26 were bare specialists. Yet particular species often showed different results in different sites, for instance Limecola balthica (L.) was generalist (four of six cases) or bare-specialist (two of six cases), and Cerastoderma edule (L.) was generalist (two of five cases), seagrass-specialist (two of five cases), or bare-specialist (one of five cases). Most species (48 of 66, 73%) were recorded at only a single site so were not possible to evaluate for consistent patterns of habitat association within a species. Nevertheless, disparate patterns within species emphasize the value of case-specific exploration of the magnitude and direction of seagrass–bivalve effects.

Figure 7.

Density of bivalves in seagrass relative to bare sand or mud, as log10 (abundance in/abundance out). Bivalves differed in functional traits of body size, position in sediment, and presence of mutualistic sulfide-oxidizing bacteria in gills. Analysis included 105 data points (unique site and species combination) from 39 publications.

Weak overall directional effects of colocation of seagrass and bivalves likely stem from the variety of mechanisms by which these biogenic habitat formers can affect each other, or from low densities in soft-sediment systems. Differences in species responses to the effector may also change with shared evolutionary history. In estuaries, many species of bivalves and some seagrasses have been moved beyond their native range creating new interactions between species pairs (Ruesink 2018), but nonnative species were considred in four of 39 surveys, insufficient to use as a predictor. Meanwhile, lucinid bivalves are considered seagrass mutualists with strong shared evolutionary history (Reynolds et al. 2007) and showed the strongest positive association. Further work is needed to explore the specific mechanisms and spatiotemporal scales enabling seagrasses and bivalves to coexist because there are many pathways at work.