Increasing urbanization and resource exploitation, along with global climate change, is reducing resilience and accelerating loss of important coastal habitats and ecosystems, such as mangroves, salt marsh, and seagrass (Waycott et al., 2009; Silliman, 2014; Spivak et al., 2019; Stafford et al., 2021). Typical of these, seagrasses underpin numerous ecosystem services, including blue carbon, coastal stabilization, and supporting biodiversity of intrinsic value as well as critical to numerous commercial fisheries (Fourqurean et al., 2012; Kerr, 2017; Röhr et al., 2018; Unsworth et al., 2019). Seagrass declines in recent decades have been substantial, with greater than 29% loss estimated globally since the 1980s (Waycott et al., 2009), and even great declines found in many regions; for example, over 60% loss estimated in Swedish waters (Baden et al., 2003) and 39% loss reported in United Kingdom waters (Green et al., 2021), over the same time period.

The causes for these losses come from many sources: local environmental changes such as eutrophication, pollution, habitat loss or disturbance, and changes in resource-consumer interactions (Short et al., 2006), as well as potentially from breakdown of regional connectivity (Kendrick et al., 2017). In addition, numerous feedbacks are known to be at play in seagrass ecosystems, e.g., through self-facilitation as a result of stabilizing sediment and increasing water clarity (Adams et al., 2016), inbreeding depression as declining populations suffer accelerated losses due to the resulting lack of genetic diversity (Maxwell et al., 2017), or metapopulation Allee effects (Amarasekare, 1998) as loss of local populations weakens connectivity at the broader scale, diminishing metapopulation resilience observed in numerous seagrasses (Rozenfeld et al., 2008; Lobelle et al., 2013; Grech et al., 2018; Jackson et al., 2021).

Since this range of pressures acts through different processes and mechanisms on different ecosystem properties, it is increasingly well understood that assessment of seagrass population status cannot rely on any single metric (Unsworth et al., 2015), and several studies have conducted multivariate assessments of resilience in seagrass (Jones and Unsworth, 2016; Jahnke et al., 2020; Bertelli et al., 2021; Krumhansl et al., 2021) and other coastal vegetative ecosystems (Battisti et al., 2020). However, there is now a need to advance this multivariate bioindicator approach, to understand the connections between components of seagrass ecosystems, linking environmental variables, seagrass genotype, seagrass phenotype, and associated biodiversity.

Our overall aim here was to understand the connections between key biotic and abiotic components in a seagrass ecosystem. While traditional species-habitat association models focus on correlations between presence of a target species and environmental covariates (Matthiopoulos et al., 2020), we wanted to understand the associations between environmental factors and multiple system components simultaneously. To achieve this, our specific objectives were to (1) quantify spatial population genetic structure in Zostera marina meadows across the southwest of the United Kingdom, based on hierarchical analysis of molecular variance (AMOVA), isolation by distance (IBD), and clustering methods, (2) use a mixed effects modeling approach to test hypotheses on environmental predictors of Z. marina population genetic variation as a first step toward understanding habitat suitability associated with genetic resilience in seagrass, and (3) develop a structural equation model (SEM) linking environmental drivers, seagrass genotype, seagrass phenotype, and major seagrass-associated faunal groups.

This study is correlative, rather than mechanistic, and focusses on genetic diversity measured through neutral genetic markers, accessible and affordable to a citizen science initiative. However, observations and analysis from this established seagrass ecosystem across a widespread natural environment complements and potentially can be used to validate more manipulative, experimental studies. Insight from this type of integrated analysis of system components is important to the development of habitat management, restoration, and policy; particularly as nature-based solutions are sought in coastal settings (Kumar et al., 2021; Seddon et al., 2021).

This study uses (1) novel seagrass population genetic data (Supplementary material), (2) seagrass phenotype and associated fauna data freely available as supplementary material associated with Smale et al. (2019), and (3) environmental data obtained from the European Marine Observation and Data Network (EMODnet) seabed habitats portal¹ and Copernicus Marine Environment Monitoring Service.²

Our study quantified variation and structure in seagrass genotype at 17 sampling locations across the south coast of the United Kingdom. We showed that this structure is associated with both environmental and geographic parameters. However, we present evidence that environment, more than genetics, predicts Zostera marina phenotypic variation and its associated faunal diversity. While our findings represent a substantial observational study of an important ecosystem in a natural setting, our analysis is correlative rather than mechanistic. In particular, the use of neutral loci (microsatellites) does not allow a test of whether phenotype in general is genetically versus environmentally determined. We go on to discuss our findings in that context.

We observed population genetic structuring at the regional scale in the form of isolation by distance (IBD). Isolation by distance has been observed in Z. marina previously across many areas of the northern hemisphere (Olsen et al., 2004; Muñiz-Salazar et al., 2005). At scales below c. 150 km, IBD was shown to be highly variable between location but displayed a clear “break point” at c. 150 km, above which IBD was consistently observable (Olsen et al., 2004). Our study was somewhat at odds, with a linear, or even decelerating pattern of IBD, up to a maximum of 170 km separation. However, the statistical evidence for nonlinearity in our study was weak, and the core finding of IBD is consistent with other studies and populations. Our study area consists of a series of suitable locations, separated by stretches of coastline where seagrass cannot survive, which would be expected to maximize potential for genetic differentiation with distance.

In addition, we observed population genetic structure at the local scale, with up to three genetic clusters at many sampling locations. In particular, through cluster analysis we found a genetic cluster strongly associated with estuarine environments. We are not aware of other reports of this type of population differentiation, but our findings are statistical and based on neutral markers. It would be interesting to follow up these findings with a trait-based approach, attempting to validate this pattern and infer the causal mechanism if upheld. Elsewhere, it has been found that Z. marina population genetics differ along a salinity gradient (Martínez-García et al., 2021) and that isolated locations such as fjords can harbor distinct genotypes (Olsen et al., 2013). It is also known from reciprocal transplant and common garden experiments that local Z. marina populations can show a “home-site advantage” at scales of a few kilometers (Hämmerli and Reusch, 2002; DuBois et al., 2022). Therefore, it may not be surprising to find seagrass genotypes selected for estuarine environments, and studies like ours, based on microsatellites, will hopefully provide support for whole genome approaches designed to investigate adaptation and selection.

To investigate species-habitat associations in our coastal ecosystem, we selected uncorrelated environmental variables likely to impact on Z. marina morphology, diversity, and resilience (Salo et al., 2015; Bertelli and Unsworth, 2018; Bertelli et al., 2021; Martínez-García et al., 2021). These broadly fell into three categories: light-associated, kinetic energy-associated, and sea surface temperature (SST). Kinetic energy may have direct impacts on seagrass through mechanical damage or act indirectly by increasing turbidity. In our data, we found low correlation between light and kinetic energy, so infer that any energy-mediated effects are mechanical here. We did not include some other environmental variables known to limit seagrass distribution, such as salinity (Martínez-García et al., 2021) and organic matter (Krause-Jensen et al., 2011), primarily because reliable data were not available at the locations and scales we needed. Both are known to affect Z. marina genetics (Martínez-García et al., 2021) and distribution (Krause-Jensen et al., 2011) and the potential for interactions between these and other environmental variables would be an interesting future research direction. We also included sea temperature in our analysis, as it is well known to affect seagrass growth (Blok et al., 2018; Hammer et al., 2018) and other regulatory processes of Z. marina such as host-pathogen interactions (Bull et al., 2012). However, sea surface temperature was strongly correlated with wave energy in our study, so we did not find evidence of direct correlations between temperature and other variables. Given the narrow latitudinal range of our survey locations, the main driver of differences in temperature is likely to be location within estuaries versus open coastline. This would account for the correlation with wave energy and might also be expected to be correlated with salinity. Finally, it should be noted that seagrass and associated fauna were sampled over August of two consecutive years and pooled. This survey design limitation was addressed by considering whether environmental variables were likely to vary substantially between 2 years. With the exception of SST, all the environmental data we used were obtained from databases that remain unchanged from year to year, based on the assumption that these variables are slow moving or static. In the case of SST, we demonstrated a strong positive correlation between years and calculated the variance inflation factor associated with this correlation to support retaining 2016 over 2017 SST. This is also the more appropriate year to retain on mechanistic grounds, being at the start of our survey period.

We found that our first environmental variable category (photosynthetically active radiation (PAR) and depth) had significant effects on all three ecosystem components: seagrass genotype, seagrass phenotype, and associated biodiversity. Our mixed effects modeling indicated increasing depth is associated with increasing allelic richness (Ar) but plateauing or possibly reversing at the deepest locations. In our study area, depths ranged between 1 and 4 m below chart datum, which represents the shallower end of the natural range in this region, with seagrass found down to around 10 m at some locations (Jackson et al., 2011). This may explain why we only identified weak evidence for the reversal of allele richness at the deeper end of our range. Elsewhere, Z. marina allelic richness has found to be maximized at intermediate depths along sub-tidal gradients (Hays et al., 2021), broadly consistent with our findings. Some other studies have appeared to show the opposite trend, with highest diversity at the shallowest depths, but these compared inter-tidal and sub-tidal meadows, so qualitatively different environments (Ruckelshaus, 1998; Kamel et al., 2012; Reynolds et al., 2017). While we do not probe mechanisms in our observational field study, the relationship between flowering versus clonal reproduction and depth could potentially explain our findings. However, the relationship between environment, flowering, and genetic diversity in Z. marina is complex and remains a key challenge to understand, as discussed by Hays et al. (2021).

Structural equation modeling confirmed our finding of increasing allelic richness with depth and showed that increasing light (PAR) was associated with less fragmented (more continuous) seagrass vegetation, as well as increased abundance of associated biodiversity (crustaceans and bryozoans). Overall, the positive effect of light on seagrass spatial continuity was the strongest ecosystem link identified in our study. Light is well known to be limiting to seagrass growth, but this is typically measured in terms of growth rate, shoot morphology, or shoot density (Grice et al., 1996; Collier et al., 2016; Bertelli and Unsworth, 2018) rather than heterogeneity in cover, or fragmentation. In fact, light was not associated with variation in shoot density in our study, suggesting reduced fragmentation in higher light conditions may not be simply through increased growth rate. However, light is known to promote flowering shoot production (van Lent et al., 1995), reproduction by seeds (as opposed to vegetative reproduction by rhizome extension) is associated with infilling of gaps in seagrass populations (Potouroglou et al., 2014), and vegetation fragmentation dynamics are best explained by infilling of gaps rather than localized losses (Irvine et al., 2016), providing a mechanism to explain our findings.

While kinetic energy due to waves (KeW) was weakly positively associated with increasing allele richness in our mixed effects modeling, this was not recapitulated in our SEM. We conclude that there is little evidence for kinetic energy-associated effects in our data, although all our study locations were relatively sheltered. As a further confounding issue, it is well known that seagrasses act as ecosystem engineers, attenuating wave energy and reducing turbidity (increasing ambient light), so self-facilitating its own growth (van der Heide et al., 2007; Adams et al., 2016; Reidenbach and Thomas, 2018). Since our sampling was typically conducted toward the center of seagrass meadows, mechanical stress from wave and current action may have been dampened below levels where variation drives observable effects on seagrass ecosystem properties.

Despite the population genetics structure evident in our study through analysis of molecular variance (AMOVA), local clustering, and IBD, we found little evidence to suggest that genetic and genotypic diversity had any association with the abundance of the seagrass or its associated fauna. The exception to this was a statistically significant positive association between clonal richness and variance in shoot densities within sampling locations. Genotypic diversity is well known to increase variability in traits such as productivity or biomass in animal and plant species [reviewed in Hughes et al. (2008)], including Z. marina (Hughes et al., 2009). However, in our study this association between seagrass genotype and phenotype had no connections to other ecosystem components.

Our finding that seagrass phenotype had no direct association with fauna is at odds with the positive association between Z. marina shoot density and faunal diversity found by McCloskey and Unsworth (2015) and various earlier studies, summarized in Attrill et al. (2000). As discussed by Attrill et al. (2000), this association could simply be explained as a sampling effect, as “more” seagrass offers greater overall opportunity for researchers to encounter additional species, particularly epifauna. Looking beyond shoot density (Jackson et al., 2006) specifically assessed the relationship between mobile fauna and seagrass habitat heterogeneity and patchiness (so niche diversity), with positive results. Again, this result is at odds with our results for shoot density variance and fragmentation. We see two possible explanations for our differing findings. A substantive difference between our study and those is that we assembled relatively coarse taxonomic data, while others have focused on higher taxonomic resolution but within restricted groups, such as fish or arthropods, and reduced spatial extent. Therefore, it could be that by accepting the trade-off between the amount versus resolution of data available through a large-scale citizen science project, we missed the appropriate taxonomic scale at which habitat-biodiversity relationships exist. Alternatively, our study differs from those in testing links between environment, seagrass genotype, seagrass phenotype, and associated biodiversity using an integrated SEM approach. Our model suggests that associations between seagrass complexity and faunal diversity may simply be correlative, driven by shared environmental drivers. These would otherwise be hard to disentangle either statistically or experimentally due to the positive feedback relationship between seagrass structure and turbidity, as well as sedimentation rates (Potouroglou et al., 2017). To partially resolve this, we note that the previously published analysis of the seagrass phenotype and associated fauna data in our study did report a positive relationship between fauna and both Z. marina density and continuity (Smale et al., 2019), suggesting that these faunal data are of sufficient resolution, and that analyses suited to quantifying linkages within high-dimensional ecosystems can play an important part in understanding complex field studies. Elsewhere, experimental approaches have been conducted, with a threshold effect observed such that faunal diversity is only strongly affected by substantial (e.g., > 50%) seagrass removal (Reed and Hovel, 2006) and that more specific aspects of seagrass fragmentation, such as distance to patch edges, explains epifauna diversity patterns (Moore and Hovel, 2010), suggesting that edge effects more than niche diversity per se might be the underlying driver of associated biodiversity.

Overall, we found relatively low levels of genetic diversity across our study, compared with (Oetjen et al., 2010), who developed the microsatellite panel we used. In our study, mean allele richness was 2.1, compared to 5.0 (calculated from the subset of their loci used in our study—see their Table 1) based on a comparable number of samples (284) from six locations at a maximal distance of 53 km apart across the Wadden Sea. This lack of genetic diversity may be the reason behind failing to find an association between genotype and other ecosystem components. However, genetic diversity is predicted to be lower at the leading edge of an expansion, e.g., post-glacial, compared to locations at the center of a species’ range or in ancient refugia, and this pattern has been documented in terrestrial [reviewed in Hewitt (2000)] and marine [reviewed in Maggs et al. (2008)] species across the north Atlantic, including in Z. marina (Diekmann and Serrao, 2012). Therefore, our finding that genetic diversity is poorly connected to other ecosystem components (seagrass phenotype and associated biodiversity) may be commonplace in other areas at the front of range expansion, or where genetic diversity is otherwise reduced, e.g., as part of a restoration program (Jahnke et al., 2015).

Alternatively, contemporary population genetic diversity and structure may be a recent historical legacy. This may not be surprising, given that much of current distribution of seagrass is a result of intense pressures and declines over the last hundred years (Short et al., 2006; Waycott et al., 2009), but that many of those pressures (e.g., wasting disease, industrial pollution) are considerably less severe now. Indeed, despite continuing seagrass losses globally, there is evidence that some populations in Europe may no longer be declining (de los Santos et al., 2019).

To conclude, we found clear evidence of local environmental conditions being associated with variation in seagrass genotype, seagrass phenotype, and associated fauna, but little evidence of links between contemporary population genetic structure and ecosystem state: the environment, particularly ambient light, more than genetics predicts natural variation in other components of this temperate seagrass ecosystem. Many of our individual findings (genetic isolation by distance, an association between genotypic and phenotypic diversity, and the key role of light on seagrass distribution) are largely in agreement with other studies. However, by assembling these elements into a single analysis, we provide new insights into the relative roles of associations between ecosystem components. While it is undoubtedly true that even apparently simple ecosystems, such as seagrasses, comprise interesting and important complexities in their dynamics, understanding the dominant role played by predictable environmental drivers can help to focus conservation and restoration efforts.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary material.

JB and NA contributed to study design and data analysis. MP and JM coordinated sample collection. NA, EK, and RA-Q contributed to molecular data acquisition. CB and NA contributed to environmental data acquisition. All authors contributed to the article and approved the submitted version.

This work was supported by the UK Natural Environment Research Council (NE/V016385/1) as part of the Sustainable Management of Marine Resources (SMMR) initiative awarded to JB, as well as a PhD scholarship awarded to NA by the Cultural Bureau of Saudi Arabia, with additional funding to NA from Princess Nourah bint Abdulrahman University, KSA. The Community Seagrass Initiative (CSI) project was supported by the UK Heritage Lottery Fund.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.