A decade implementing ecosystem approach to fisheries management improves diversity of taxa and traits within a marine protected area in the UK

Ecosystem Approach to Fisheries Management has highlighted the importance of studying ecosystem functions and services, and the biological traits that drive them. Yet, ecosystem services and the associated benefits that they provide are rarely the motive for creating marine protected area (MPA). Therefore, many MPA monitoring projects do not explicitly study these functions and services or the underlying biological traits linked to them.


| INTRODUC TI ON
Coastal areas of the marine environment have historically been overexploited and subjected to high levels of pressure, such as commercial fishing, nutrient loading and noise from shipping (Brown et al., 2001;Letessier et al., 2019;Pine et al., 2016). Arguably one of the most destructive impacts to the marine environment have come from unsustainable commercial practices, such as fisheries trawling and dredging , and aggregate and maintenance dredging (Cooper et al., 2008). To negate the impacts of commercial fishing, one widely used management tool is the implementation of marine protected areas s (MPAs;Cleguer et al., 2015;Gallacher et al., 2016;Jones, 2008). There are many different types of MPA varying in level of protection, spatial extent and temporal coverage.
The spatial extent of a MPA can range from tens of square metres to thousands of square kilometres, while temporal extent can cover specific months, a season or all-year-round protection (Ferse et al., 2010). MPAs can be designated to protect overall biodiversity, specific species or habitat 'features' , with the level of protection ranging from the protection of one single species or 'feature', to the complete exclusion of potentially harmful activity in a whole area (Boonzaier & Pauly, 2016). By excluding the most destructive human activities to protect sensitive biogenic habitats that provide essential feeding and nursery ground for species of commercial importance, MPAs can help to achieve both conservation and fisheries goals. Consequently, MPAs have been advocated within an Ecosystem Approach to Fisheries Management (EAFM) approach (Halpern et al., 2010). To effectively manage the whole ecosystem, appropriate monitoring must be applied which can detect changes in ecosystem function, diversity and health over time-scales that can allow for adaptive management (Maxwell et al., 2015).
To monitor and assess whether ecosystems are recovering under the protection that a MPA has provided, taxonomy-based biodiversity metrics are often used, such as species richness or Shannon's diversity (Ferreira et al., 2017). Higher levels of taxonomy-based biodiversity are often linked to higher ecosystem functioning and increased productivity (Vackár et al., 2012). However, large changes in taxonomic based biodiversity do not necessarily imply equally large changes to ecosystem function (Törnroos & Bonsdorff, 2012;Wong & Kay, 2019), especially when an ecosystem contains high levels of functional redundancy: when many different species share the same trait modalities (such as predator, detritivore or filter feeder for an organisms feeding trait; Guillemot et al., 2011). Ecosystem function is a complex system of interactions, which combine to make up the whole system through multiple different processes (Jax, 2005). Thus, when assessing the change in the trait diversity, the combination of functional richness, functional evenness, functional divergence, functional distinctiveness and functional dispersion (Laliberté & Legendre, 2010;Murgier et al., 2021;Violle et al., 2017) of a system can indicate its functioning in relation to ecosystemwide processes, such as productivity and regulation of biogeochemical fluxes (Perović et al., 2018;Ricotta et al., 2016). Specific traits, modalities (e.g. filter feeding within the feeding trait) or groups of traits being heavily linked to specific processes and fluxes. For example, feeding habit has been linked to bentho-pelagic coupling and trophic linkage (Beauchard et al., 2017), environmental position (benthic, pelagic, etc.) with sensitivity to destructive fishing and longevity, maturation age and reproductive strategy with recovery from destructive fishing (Rijnsdorp et al., 2016). Functional richness measures the size of the trait space filled by the community, while functional evenness and functional divergence indicate how the abundance is distributed across this trait space (Table 1). Evenness increases as organism abundance is spread more evenly across the trait space, whereas divergence increases as specific extreme traits or combinations of traits become more abundant in relation to others (Laliberté & Legendre, 2010;Mason et al., 2007;Mason et al., 2008). Further, the study of functional redundancy, the similarity of traits within the community, can indicate the resilience of a MPA to perturbations, such as destructive fishing, biological invasions and storm events (McLean et al., 2019;Tillin et al., 2006). Hence, the study of functional traits has been suggested for both monitoring and management (Rijnsdorp et al., 2016;Tillin et al., 2006;Wiedmann et al., 2014).
Initially applied within terrestrial and freshwater environments, the use of biological trait analysis has increased in recent years in the marine environment (Berthelsen et al., 2015;Coleman et al., 2015) and has mostly been used to assess either fish or benthic invertebrate communities (Beauchard et al., 2017). Benthic invertebrates are well known as bio-indicators of ecosystem health, disturbance or biogeochemical processes (Belley & Snelgrove, 2016;De-La-Ossa-Carretero et al., 2012;Munroe et al., 2018;Parmar et al., 2016) and fish assemblages have been used to assess the impacts of fishing and climate change (Benoıt & Swain, 2008;Benoit et al., 2013). Particular groups of organisms can be used as indicators for different elements of ecosystem health, and the assessment of multiple groups of organisms will provide a more comprehensive image of the function of the whole ecosystem and allow for adaptive management, as set out by EAFM (Long et al., 2015).  . This management means that the most destructive activities are only excluded in designated areas within MPAs, where specific features of conservation importance have been evidenced , leaving the majority of MPAs effectively open to damage and further degradation . These "feature-based" MPAs have been considered ineffective for both fisheries and conservation management (Pikitch et al., 2004), meaning MPAs that provide consistent protection across their whole area, known as whole-site approach, are being advocated . One example of a MPA managed through a whole-site approach is in Lyme Bay, SW England. Due to high levels of mobile demersal fishing in the bay, which were shown to be damaging the rocky reef system characterized by sessile fauna species, such as pink sea fans Eunicella verrucosa and ross corals Pentapora foliacea, stakeholders created a voluntary closure agreement (Mangi et al., 2011;Rees et al., 2010 Sessile and sedentary epifauna were monitored using a towed underwater video system ("towed flying array" henceforth), while mobile species were monitored with Baited Remote Underwater Video systems (BRUVs; Davies et al., 2020;Davies, Holmes, Rees, et al., 2021;Sheehan et al., 2010;Sheehan, Stevens, et al., 2013;Sheehan et al., 2016;Stevens et al., 2014). The combination of these data collection methods, which estimate the abundance of species ranging from sessile branching sponges to highly mobile elasmobranchs, allow the assessment of a large proportion of the benthic ecosystem. In the present study, trait analysis was carried out on this surveyed proportion of the benthic ecosystem, analysing change in trait diversity metrics captured by the combination of the towed video and BRUVs data. As protection of the benthic ecosystem was expected to allow increases in biodiversity as well as prevalence and abundance of regionally rare trait combinations, trait diversity metrics were expected to increase over time in the MPA relative to the nearby areas open to mobile demersal fishing. Furthermore, as mobile demersal fishing gear is known to significantly impact the relative proportion of specific trait modalities present within an ecosystem (Howarth et al., 2018; van Denderen et al., 2015), causing decreases in sessile, filter feeding and long-lived organisms and an increase in mobile and scavenging organisms (Tillin et al., 2006), the relative proportions of trait modalities were assessed over time. Here, a unique assessment ranging across a large portion of the benthic ecosystem is provided, with the intent to best inform adaptive management of Lyme Bay and improve MPA management elsewhere.
To assess how the health and overall ecosystem of the functional reef habitats inside Lyme Bay MPA changed over time, the diversity (taxonomic and trait) and traits of epibenthic and demersal communities were assessed inside the MPA and outside the MPA, in unprotected Open Controls over 10 years.
The following hypotheses were tested:  (Sheehan, Stevens, et al., 2013). Since designation, a 'feature'-based Special Area of Conservation (SAC) has been designated in the surrounding areas of the MPA. However, the focus of the current study is on the whole-site approach designated in 2008, which prohibited mobile demersal fishing across all habitats within the 206 km 2 .

| Sample design
To identify suitable sites for monitoring, spatial analyses were conducted combining historical fishing effort, benthic substrate and depth (Sheehan, Stevens, et al., 2013;Stevens et al., 2014). Sites were selected inside and outside the MPA and were located on comparable depths, bathymetry and substrate Sheehan, Stevens, et al., 2013;Stevens et al., 2014).

| Video collection
The towed flying array is used to record: 200 m by 0.5-m-wide high definition (HD) video transects over heterogeneous and fragile benthic ecosystems (Sheehan et al., 2010(Sheehan et al., , 2016. The array was a bespoke aluminium frame mounted with: a HD video camera (Surveyor-HD-J12 colour zoom titanium, 720p); LED lights (Bowtech Products limited, LED-1600-13); two green lasers (Z-bolt Scuba-1); and a mini CTD profiler (Valeport Ltd.). The camera was connected to a Bowtech System power supply/control unit by an umbilical cable, which allowed video to be monitored in real time to ensure control of the lights, camera aperture and camera focus. The camera and the parallel lasers were positioned at an oblique angle to the seabed, with the lasers set 300 mm apart, to allow the quantification of the field of view.   (Cappo et al., 2004;Willis et al., 2000).

| Combining BRUV and Towed data
Baited remote underwater video systems and towed video data were combined by converting all abundance and MaxN values, averaged by area, into relative values using the 'make_relative'

F I G U R E 2
Steps taken to collect, combine and convert baited remote underwater video system and towed video data into trait diversity metrics and composition function within the 'funrar' package in R (Grenié et al., 2017;Grenié et al., 2020), as relative values can be used in functional metric creation. Relative data sets were then joined by year, treatment and similar areas (based on habitat, substrate and depth), creating a combined site by species matrix of relative abundances.
For all analyses, BRUV and towed video data were combined to assess the trait diversity and how it changes over time inside and outside the MPA.

| Trait acquisition
In total, 11 traits were used with a cumulative 58 modalities (Table 2).
They were selected for importance for the benthic environment and its coupling with other components in the ecosystem (e.g. Pelagic/ Neritic), as well as availability of information. Trait data were taken F I G U R E 3 Temporal changes in diversity metrics from combined towed underwater video and baited remote underwater video systems data inside the MPA (blue diamonds) and OC (grey triangles). An example screenshot from the baited remote underwater video systems (a),  (Chevenet et al., 1994), all other factors were considered ordinal (Podani, 2005). For categorical traits, all modalities within each individual trait sum to equal 1 so that a trait with more modalities would not be weighted higher than another (Laliberté & Legendre, 2010).

This created a Species x Trait matrix ([T]) for the species sampled by
the towed flying array and the BRUVs (Appendix S1).

| Community-level weighted means
The CWM, which represents the relative proportions of all traits at each site, was calculated from the 'FD' package in R. Specific traits, known to be affected by mobile demersal fishing, were selected a priori for analysis. The traits selected were as follows: longevity; filter and scavenger feeding types; and sessile, crawler and swimmer motility types.
(1) O = 1 − U TA B L E 2 Traits and their constituent modalities used to assess trait diversity. Bold denotes traits above their constituent modalities

| Feeding type
The relative proportion of the filter feeding modality, within the   Figure 4b).

| Longevity
Only longevity ranges 1-2 years, 2-10 years and more than 20 years were the dominant community longevities across all years and treatments. These longevity ranges showed no significant change over time between treatments. 1-2 years had the highest probability (the MPA was 0.818, while the OC was 0.688 throughout:  Figure 4c). Similarly, 2-10 years longevity probability did not change over time (0.169 in the MPA and 0.306 in the OC: Table 4 & Figure 4c). There was no difference in the probability of taxa of more than 20 years longevity between treatments or through time (0.0133 in the both treatments: Table 4 & Figure 4c).

| Motility
The

| DISCUSS ION
The taxonomic and trait changes in the Lyme Bay MPA were as-  Although it has had many definitions, functional richness has consistently been shown to be an important driver for ecosystem stability, resilience and services (Canning-Clode et al., 2010;Törnroos et al., 2015;Wahl et al., 2011). Functional and taxonomic richnesses are closely related, regardless of how functional richness is defined: the trait space filled by a community, used here (Boyé, et al., 2019;Villeger et al., 2008); the species richness within functional groupings (Canning-Clode et al., 2010;Wahl et al., 2011); the species richness of functionally unique species (Canning-Clode et al., 2009;McLean et al., 2021); or the number of functional traits (de Juan et al., 2015).
Here, the exclusion of mobile demersal fishing inside the MPA has allowed the increase in both the number of taxa and the trait space.
It is likely the increase will have altered ecosystem-wide processes, with positive effects to productivity and regulation of biogeochemical fluxes (Perović et al., 2018;Ricotta et al., 2016;Vackár et al., 2012). However, further research is required to explicitly assess how changes in trait diversity and composition can alter or influence the functioning of ecosystems, especially when the identity and traits of dominant species can be more influential than overall diversity (Gammal et al., 2020;Strong et al., 2015).
Pressures imposed by high levels of demersal towed fishing can impact communities in varying ways (Fragkopoulou et al., 2021;Olsgard et al., 2008;Wang et al., 2021). One example, strong disturbance regimes, will continually reset communities to small initial successional stage assemblages (Song & Saavedra, 2018). This may introduce high levels of competitive interactions and, as such, will increase the functional divergence and decrease trait redundancy (Perronne et al., 2017). Cumulatively, this will restrict the increase in abundance of novel traits into the community and can lead to dominance of a few species with unique traits, such as high fecundity and short life span that allow them to persist (Boyé, et al., 2019). This was likely the driver for the observed minor decrease over time in functional divergence across both treatments and decrease in functional redundancy seen in the unprotected areas (Table 3). Furthermore, trawling and other destructive fishing practices can significantly alter the proportion of traits present within a benthic community (Howarth et al., 2018). Specifically, chronic trawling can cause a decrease in sessile filter feeding organisms and an increase in mobile scavenger species (Tillin et al., 2006). This effect of chronic impact can produce a shift in the baseline to measure change or recovery from in heavily trawled systems (Brown & Trebilco, 2014;Ulman & Pauly, 2016). As shown here, the area protected from demersal mobile fishing showed increases in filter feeding organisms, with the unprotected areas showing decreases in the proportion of sessile organisms and increases in the proportion of swimmers and crawlers. The Lyme Bay MPA was designated to protect the rocky reef habitat, which in turn is characterized by sessile fauna species, such as pink sea fans Eunicella verrucosa and ross corals Pentapora foliacea (Sheehan, Stevens, et al., 2013). Yet in 2008, the first surveys of the MPA showed limited sessile life growing on the boulders and cobbles (Sheehan, Stevens, et al., 2013). Thus, the marginal increase in the MPA in the proportion of sessile modality alongside the strong decrease in the unprotected areas show the protection is being effective and is protecting the sessile rocky reef species, while the unprotected areas continue to be dispossessed of species with these trait.
In this study, functional evenness remained the same. Assuming that resource availability, such as substrate, prey or mate availability, was even across the system, this implies that as the trait space (functional richness) increased, the dominance pattern within this trait space stayed the same. Hence, the community was maintaining its effectiveness at utilizing the available resources (Mason et al., 2005). This lack of change in evenness in the MPA, even though the functional richness (trait space) increased over time, may be due to increases in abundance of species with locally rarer combinations of traits. Yet, these abundances stay minimal in comparison with the traits of the more dominant species. This continued dominance of common combinations of traits across both treatments may have led to the observed functional divergence (Figure 3d).
Communities containing high levels of trait overlap (functional redundancy) have been found to provide a higher resilience to environmental impacts, such as fishing, storms or biological invasions (Mason et al., 2005;McLean et al., 2019;Tillin et al., 2006).
When a species becomes regionally extinct, its suite of traits is less likely to become lost when there is high trait overlap. The higher levels of functional redundancy witnessed in the MPA compared to the unprotected areas show potential higher resilience to perturbations (Rincón-Díaz et al., 2018), and the effect increased over time. However, the resilience provided by redundancy will be heavily linked to the traits used (Boyé, et al., 2019).
When seen alongside the increases in taxa and the functional richness inside the MPA, the increase in redundancy would imply that the increases in richness are across and within a wide range of niches and trophic levels (Rincón-Díaz et al., 2018). Many studies focusing on fish assemblages have found an opposing pattern, with an increase in richness simultaneous to a decrease in redundancy (Rincón-Díaz et al., 2018;Stuart-Smith et al., 2013). This may be due to the probability of overlap in traits decreasing when there is an increase in functional richness, and the concomitant growth in trait space. The relationship between diversity and the buffer created by trait redundancy is of high importance to managers and conservationists for setting goals and priorities (Micheli et al., 2014), and as such needs to be fully understood at both regional and global scales. With perturbations likely to continue to increase through direct and indirect anthropogenic impacts, so the importance of community resilience increases. Severe storms in early 2014 across the south-west of the UK have been shown to impact the marine benthos in Lyme Bay .
The effects can be seen in a reduction in taxa and functional richness and an increase in the scavenging 2014 in both the MPA and unprotected areas. There was also a reduction in the functional redundancy and an increase in the functional divergence in the OC, but the only discernible reduction in the MPA was a year later in 2015, potentially showing a buffering effect in the MPA from the higher levels of functional redundancy. Here, the Lyme Bay MPA has shown increases in not only diversity but also its potential resilience, displaying the possible functional benefits provided by the whole-site approach to MPA management, which has been employed at the site.
The two survey methods (towed flying array and BRUVs) used were originally conducted to perform taxonomic-based assessments, evaluating the change in the species or taxa, and communities present (Davies et al., 2020;Davies, Holmes, Rees, et al., 2021;Sheehan et al., 2010).
Combining the two survey methods data into a single analysis allows an assessment of a broader range of the benthic ecosystem, with benefits of providing more comprehensive understanding of how the monitored ecosystem is changing in relation to anthropogenic and natural perturbations. Functional assessment is already being advocated to inform management (Rijnsdorp et al., 2016;Tillin et al., 2006;Wiedmann et al., 2014), and enabling the assessment of functional change across a large proportion of the benthic ecosystem could aid adaptive management of MPAs, yet caution is needed when comparing between different systems or locations. As mobile demersal fishing impacts specific traits, so other fishing methods will effect other specific traits or trait combinations (Tillin et al., 2006). Therefore, utilizing trait diversity and trait composition to detect effects of different management strategies must be targeted to the expected consequence of the management (Trindade-Santos et al., 2020). For example, the current work is specific to Lyme Bay, where a comparatively rare management regime, especially in Europe, has been in place. This management has protected the whole-site, covering a mosaic of different habitats, from mobile demersal fishing . However, the vast majority of marine protection in Europe utilizes what has been termed 'feature'-based protection . Therefore, the effects seen here in this "whole-site" MPA, increases in trait diversity and specific traits, will be specific to this management approach, as well as the assessed spatial and temporal scales. Magnitudes of functional metric values are linked to the number of different traits, these need to be kept consistent to allow comparison (Villeger et al., 2008). Furthermore, the maximum number of potential species assessed requires consistency, as this will heavily influence the potential number of different traits found within a community. Therefore, increases in repositories of trait information for a wide range of species and the standard reporting practices are highly important to allow comparison between locations, nationally and internationally.
Diversity metrics (taxonomic and trait) showed significant trends of increase over the 10 years surveyed, yet without data from pre-fishing, defining recovery is problematic (Ulman & Pauly, 2016). It is expected that a recovering system will eventually reach a plateau of diversity. However, the time-scales for this to occur can be large, more than 75 years in some instances (Anderson et al., 2014). The current work shows diversity increases over 10 years with no clear sign of plateau, indicating the potential for continued recovery. Yet, climate change will inevitably impact marine systems over the coming decades. Thus, emphasizing the need for long-term monitoring to fully appreciate trajectories of recovery both in this specific system and others globally, especially as climate change impacts intensify.
In conclusion, the ecosystem function of the benthic community in Lyme Bay has significantly changed over 10 years following the exclusion of mobile demersal fishing, with increases in number of taxa, the functional richness and the functional redundancy in the MPA. The MPA in comparison with unprotected areas has become more diverse both taxonomically and functionally, which will likely have lead to greater levels of ecosystem service. Sessile organisms, fundamental to the health and development of rocky reef habitats, decreased outside the protected area over time, showing that this MPA is protecting the rocky reefs in areas that were previously damaged by destructive fishing practices. It is difficult to suggest whether the trends of increasing number of taxa, functional richness, functional redundancy and filter feeding traits are a recovery to a pre-fishing state, due to the unknown level of shifting baselines, which before-fishing data could reveal. However, it does show a trend towards a more diverse and potentially resilient rocky reef habitat, providing further evidence of the benefits of employing the whole-site approach to MPA management.

ACK N OWLED G EM ENTS
To carry out fieldwork, thanks are given to Lyme Bay Fishers John Walker, Robert King and Keiran Perree and University of Plymouth staff and student volunteers.

CO N FLI C T O F I NTE R E S T S
The authors declare that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available via the Archive for Marine Secies and Habitats Data (DASSH) http://www.dassh.ac.uk/doito ol/data/1718 (Davies, Holmes, Bicknell et al., 2021).