Ecosystem Approach to Fisheries Management works—How switching from mobile to static fishing gear improves populations of fished and non‐fished species inside a marine‐protected area

This is an open access article under the terms of the Creat ive Commo ns Attri bution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. Journal of Applied Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society School of Biological & Marine Science, University of Plymouth, Plymouth, UK


| INTRODUC TI ON
Globally, the implementation of marine-protected areas (MPAs) to conserve and protect marine biodiversity and aid fishery management has increased rapidly over the last 25 years (Da Silva et al., 2015;Halpern et al., 2010). By protecting vulnerable species and habitats, MPA management strategies have successfully increased the abundance and size of fisheries' target species and increased resilience to natural and anthropogenic disturbance . Thus, depending on how they are managed and enforced, MPAs have the potential to simultaneously benefit fisheries and conservation (Babcock et al., 2010). However, only 7.9% of the oceans are designated as MPAs (UNEP-WCMC, IUCN and NGS, 2018; ~17,000 MPAs covering 28.6 million km 2 ), over 2% short of the 10% target for 2020, set by the convention of Biological Diversity's Aichi Target 11 (Lubchenco & Grorud-Colvert, 2015). Furthermore, 'paper parks' (MPAs established without appropriate management and or resources to monitor, maintain or enforce protection) are prevalent despite increased global pressure to protect ecosystems using the MPA approach (Rife et al., 2013).
Permitted activities vary between different MPA designations and are typically zoned within MPAs, whereby only listed features such as specific habitats are afforded protection , particularly within European waters. Partial protection can also include seasonal closures, specific species protection and fishing practice restrictions (Dinmore et al., 2003;Hattam et al., 2014;Topor et al., 2019;Williams et al., 2006). This form of management limits recovery potential as the presence, extent and condition of features are required to be evidenced. For example, this 'feature-based' management, which is the most common approach employed within UK waters, means that out of 66,507 km 2 of sea- This approach only protects the evidenced extent of a 'feature' at a specific moment in time, potentially limiting any future growth or migration that may occur in other habitats through natural ecological processes . A more ambitious approach is the whole-site approach, a method for applying the Ecosystem Approach to Fisheries Management (EAFM: Serpetti et al., 2017) through consistent protection, across the whole seabed, acknowledging that habitats and species can recover beyond their current status when protected . Therefore, this approach protects a range of species and habitats across a larger area than the current evidenced extent of the 'feature' of interest , including species or habitats that are highly important to the 'feature' of interest. The most extreme example is no take zones (NTZs) that exclude all extractive or destructive practices (Harasti et al., 2018;Sale et al., 2005). However, partial protection that only excludes the most destructive fishing activities has also been shown to be highly effective at protecting conservation fea- tures, yet evidence of benefits to fisheries are rare (Beukers-Stewart et al., 2005;Sheehan, Stevens, et al., 2013). MPAs are often seen as a compromise between conservationists and groups with direct fishing interests (Denny & Babcock, 2004;Sciberras et al., 2015). This compromise can lead to less protection for partial MPAs and decreased spatial extent for NTZs (Hamel et al., 2013), and has led to a debate as to the effectiveness of these areas (Edgar, 2011;Turnbull et al., 2021). The level of protection and enforcement of an MPA, alongside the size, age and isolation, determines how species and habitats recover following designation, with greater protection generally causing a more positive response . Most studies to date considering the effectiveness of 'feature-based' partial MPAs have found them ineffectual at achieving the conservation or fisheries' goals that instigated their designation (Piet & Rijnsdorp, 1998;Shears et al., 2006;Turnbull et al., 2021) and, in some cases, even increased the human threats to the system inside the protected zone (Zupan et al., 2018). Thus, it has been suggested that the whole-site approach can more adequately achieve the goals of both fisheries and conservation management (Rees et al., 2020;Solandt et al., 2020).
Yet, due to the rarity of MPAs that have adopted the whole-site approach, few studies have assessed this style of marine management.
The Lyme Bay Statutory Instrument was established in 2008 (Mangi et al., 2011) to recover and protect reef habitats and species.
The most destructive fishing activities, trawling and scallop dredging, were excluded from a mosaic of habitats (~206 km 2 ) while static gear and diving were still permitted. This created both the Lyme Bay MPA and the rare opportunity to study the effect of the whole-site approach for the first time over such a large temporal (11 years) and spatial scale (>200 km 2 ).
Marine-protected area effectiveness is dependent on appropriate management and enforcement, and requires robust standardised monitoring to evidence ecological effectiveness, socio-economic benefit and justify the inherent costs . To evidence the ecological effectiveness and inform adaptive management, methods must be used which can quantify elements of the ecosystem of interest over appropriate temporal and spatial scales. Sessile and sedentary species were monitored in Lyme Bay using a flying towed This Ecosystem Approach to Fisheries Management can benefit and maintain sustainable fisheries and species of conservation importance.

K E Y W O R D S
baited remote underwater video systems, biodiversity, conservation, marine management, marine-protected areas, monitoring, whole-site approach video array Sheehan et al., 2010), and recovery of certain benthic species was only detectable 3 years after bottom towed fishing was excluded. Monitoring mobile, often shy, species with highly variably temporal and spatial distributions in the marine environment is challenging and in the past has been limited to destructive trawl surveys (Murphy & Jenkins, 2010) and fisheries' landings (Coleman et al., 2004). Increasingly, less destructive methods are now used, such as underwater visual census (Kough et al., 2017), underwater video survey Sheehan Stevens & Attrill, 2010) and fisheries' acoustic surveys (Erisman & Rowell, 2017).
Trawl surveys are destructive and so could compromise the recovery of the MPA that is being monitored (Murphy & Jenkins, 2010) while fisheries' landing assessments are restricted to commercially desirable species (Murphy & Jenkins, 2010). Underwater visual censusing, in the form of diver surveys , is restricted by diver ability (Harvey et al., 2004), depth range and number of dives in a day while acoustic surveys struggle to reliably identify fish species (Gannon, 2008). Underwater video survey is restricted by water clarity, light levels, camera specification and organism behaviour (Cappo et al., 2004). However, it is non-extractive and non-invasive, and is capable of sampling extreme depths for long periods of time while creating a permanent record of the survey, which can allow subsequent reanalysis and quality control (Stevens et al., 2014). Baited remote underwater video systems (BRUVs) sample the mobile fauna of a large area, unconstrained by depth, to provide cost-effective data on fish diversity and relative abundance (Harasti et al., 2018;Whitmarsh et al., 2017). Frequently used to monitor MPAs, BRUVs provide a conservative estimate of relative abundance of predatory species that are attracted to the bait, as well as non-predatory species that pass through the field of view (Cappo et al., 2004;Whitmarsh et al., 2017).  (Stevens et al., 2014). Despite the continued fishing pressure on many mobile species within the MPA, it was considered that the recovery of the biogenic reefs, which are essential fish habitats (Rabaut et al., 2010), would lead to increases in both Exploited and Non-Exploited mobile species .
To assess this prediction, the following hypotheses were tested:

| Survey location and design
Lyme Bay MPA (Figure 1), located on the southwest coast of England, covers 206 km 2 of nationally important rocky reef habitat (Hiscock & Breckels, 2007). For site selection, suitably comparable rocky reef regions comprising bedrock, boulders and cobbles were identified by utilising fishing effort and habitat data (Stevens et al., 2014). Within these broadly defined regions, sites were spread across each treatment (MPA and open controls: OC) to ensure that sites were spatially interspersed as much as possible ( Figure 1). BRUVs were deployed each summer from 2009 to 2019. Sites of three replicate BRUVs, spaced ~100 m apart, were deployed, to depths ranging from 14 to 29 m (see Figure S1), for 45 min before being recovered. In all, 12 sites were inside the MPA (36 BRUVs) and 6 were in the OC (18 BRUVs). Annually, the same latitude and longitude of sites were used as targets, yet each replicate is considered independent as location will be influenced by the prevalent tidal and atmospheric conditions during deployment.

| Equipment
Baited remote underwater video systems consisted of a metal frame, lead weights (~30 kg), underwater wide-angle camera housing with horizontal facing camera (Panasonic HDC-SD60 and HDC-SD80), LED lights and a fixed bait pole (Bicknell et al., 2019). Metal bait boxes were fixed on the pole one metre from the camera filled with ~100 g of Atlantic mackerel Scomber scombrus cut into segments. Fresh bait was replenished for each deployment. Videos from BRUVs were assessed in situ to ensure that the camera had landed and recorded a viable sample. Failed attempts were repeated to ensure that all samples were suitable.

| Video analysis
Videos were subject to quality control checks according to the (c-f) videos). All criteria must be maintained for a minimum of 30 min across the recording. Videos which did not meet these requirements were omitted from analysis. Videos which did meet the requirements were watched at normal speed for 30 min, after a preliminary 5 min settling period. For every minute, all mobile fauna were identified to the highest taxonomic resolution possible, and counted.
Mobile species were categorised as taxa that were deemed able to continuously move, either in response to the bait or in response to other taxa, which are themselves reacting to the bait. Thus, benthic taxa such as Pecten maximus, Aequipecten opercularis and Ophiothrix fragilis were not included. For every 1-minute segment of the video, the MaxN (maximum number of individuals on screen) for each taxon was recorded. Relative abundance of each taxa was recorded as the greatest MaxN value in any 1 minute, within the 30 min analysed.
MaxN is considered a conservative estimate of relative abundance of mobile species attracted to the bait, which decreases the chance of an individual being repeatedly recorded (Cappo et al., 2004).

| Statistical analysis
The univariate metrics, number of taxa and total abundance, were calculated in 'dplyr' and 'vegan' in r using BRUVs MaxN values (Oksanen et al., 2019;Wickham et al., 2019b). Unless stated otherwise, total abundances were fourth root transformed to meet assumptions of normality. Exploited taxa were defined as taxa which are either landed by fishers or caught and used as bait to catch other species in Lyme Bay (Personal Communication with Lyme Bay fishers, Table 1). As the BRUVs enumerated a wide range of species (Table 1) Euclidean distances. The statistical significance of the variance components was tested using 9,999 permutations under a reduced model (Anderson, 2001). PERMANOVA was selected as it is robust to unbalanced designs (Sheehan, Stevens, et al., 2013). Visualisation of multivariate data was carried out by a non-metric multidimensional scaling (MDS) ordination. Percentage contribution of taxa to dissimilarity between sites was assessed using the SIMPER (similarity percentages) method within each year and treatment (Clarke & Gorley, 2015).
Due to a high proportion (~60%) of zero values when the data were split into Exploited invertebrates, zero-inflated Poisson (ZIP) regression models were used from the 'pscl' package in r to assess the data (Zeileis et al., 2008;Zuur & Ieno, 2016). Model selection utilised Akaike information criteria (AIC) for both the Poisson 'count' and binomial (Bernoulli) 'zero' portions of the model.

F I G U R E 1 Baited Remote Underwater Video system locations within Lyme Bay marine-protected area (blue circles) and open controls (grey triangles)
To assess long-term linear trends in univariate metrics, significant (p < 0.05) temporal terms (Year and Year × Treatment) were further analysed and visualised, using linear regression analyses.
Linear regression analyses were carried out utilising the 'tidyverse' and 'stats' packages within r (R Core Team, 2019; Wickham, Averick, et al., 2019). Sample versus fitted residuals, quartile-quartile and autocorrelation of temporally sequential samples were assessed visually, to fit assumptions of the models used.

| Assemblage composition
Assemblages at MPA sites were always different from those in open controls (Figure 3; Table 2), but over time the assemblage composition of the two treatments also shifted in discordant ways, shown by a significant year:treatment interaction ( Table 2). The MPA showed large shifts in assemblage in the first years, then after 5 years proceeded to become consistent over time, unlike the OC, which showed random annual assemblage shifts with little to no consistency over time (Figure 3). Assemblage similarities, within sites across years and treatments, were driven primarily by the Small-Spotted catshark Scyliorhinus canicula and Gobiidae spp ( Figure 4). Most of the remainder of the similarity within the MPA sites was driven by reef-associated wrasse species (dark blues, Figure 4), whereas in the OC this was driven by scavenging crustaceans, echinoderms and gastropods (yellows, oranges and dark browns, Figure 4). Excluding Scyliorhinus canicula, the vast majority of the similarity within the OC sites was driven by the scavenging crustacean, Pagurus spp. (Figure 4).

| Number of taxa and total abundance
In the MPA, the mean number of taxa and mean total abundance, de- This change over time was significant in both the number of taxa and total abundance, yet neither metric showed a significant year:treatment interaction (Table 2). However, the total abundance was significantly different between treatments ( Table 2). The number of taxa showed a significant linear increase over time inside the MPA (Figure 5a) while the total abundance showed a significant linear increase in both treatments over time (Figure 5b).

| Number of taxa and total abundance
In the MPA, the mean number of taxa and mean total abundance of

TA B L E 2 (Continued)
decrease in the number of taxa and total abundance, respectively).
In the OC, the mean number of taxa and mean total abundance of  Table 2). The MPA showed a much greater increase over time (gradient of 0.14: Figure 6a) than that of the OC (gradient of 0.062: Figure 6a). The number of taxa of Non-Exploited fish was significantly different across years and treatments but, like the Exploited fish, showed no year:treatment interaction. However, the change over time inside the MPA, unlike that of the Exploited fish, was expressed as a significant linear decrease (Figure 6b). The total abundance of Exploited fish showed a significant difference between years but not between treatments (Table 2). Both treatments showed significant linear increases over time (Figure 6c). There was a significant year:treatment interaction for the total abundance of Non-Exploited fish (Table 2), which was expressed as a significant linear decrease over time inside the MPA (Figure 6d).

| Number of taxa and total abundance
In the MPA, the mean number of taxa and mean total abundance increase in the number of taxa and total abundance, respectively).
Neither year nor treatment could be fitted to model the number of Exploited invertebrate taxa, with the 'best' ZIP model utlising only the intercept for both the count and zero parts of the model (Table 2).
However, there was a significant year:treatment interaction for the number of taxa for Non-Exploited invertebrates (Table 2), with a significant linear decrease with time in the OC (Figure 7b). The total abundance of Exploited invertebrates was significantly lower in the MPA compared to the OC (Table 2; Figure 7c). The total abundance of Non-Exploited invertebrates did show a significant year:treatment interaction (Table 2) but there was no significant linear trend over time (Figure 7d). The protection has shown to positively benefit sessile reef fauna Sheehan, Stevens, et al., 2013) and

| D ISCUSS I ON
the effects of this protection have now led to positive increases to the mobile fauna over time, with increases in the number of taxa in the MPA. This is likely to be due to a combination of direct displacement of species, from areas subject to bottom towed fishing to areas not subject to bottom towed fishing (Dinmore et al., 2003), and through indirect protection and proliferation of the sessile reef habitat, which, in turn, increases survivorship of mobile taxa (Howarth et al., 2015;Wilson et al., 2010).
Fish assemblages are dependent on depth, habitat complexity and availability, competition/predation and larval/recruitment variability (Harasti et al., 2018;Meekan et al., 2018), and, as such, can be highly variable (Stige et al., 2019).  Stevens, et al., 2013). The increase in Exploited fish will likely have been driven by this increase in functional reef area, which is known to be an Essential Fish Habitat (Rabaut et al., 2010).
The increase seen in the OC was found to a be at a slower rate than the MPA and may have been due to 'spillover' effects, likely driven by a combination of increased larval export and direct adult movement from the MPA to the surrounding area (Berkeley et al., 2004;Garcá-Rubies et al., 2013). Thus, the simultaneous increase in EFH and reduction in collateral damage to habitat complexity associated with seabed dredging and trawling may have contributed to this general increase in taxa and abundance of around 400%. This co-occurred with a decrease in the number of taxa and total abundance of Non-Exploited fish over time, potentially indicating competitive exclusion by the commercially Exploited fish, which are more likely to be larger higher trophic predators (Baudron et al., 2019). For example, the Exploited shark and ray species Scyliorhinus canicula, Scyliorhinus stellaris and Raja clavata are known to predate on small bony fish (Ellis et al., 1996), As an indirect effect of exclusion of towed bottom fishing within Lyme Bay, decreases in conflict between towed fishers and potters led to increases in potting levels within the MPA (Mangi et al., 2011).
Although less destructive than bottom towed fishing, potting at high densities can have impacts to sensitive habitats (Gall et al., 2020) and target species have harvest-associated selection applied to them, which could lead to alterations in population size and behavioural selection (Madin et al., 2010;Meekan et al., 2018

Points with errors bars showing mean values and standard errors
Temporal trends of Non-Exploited groups showed decreases in number of taxa and total abundance of fish inside the MPA and total abundance of invertebrates in the OC. As mentioned, fish population dynamics are highly linked to the available habitats, as well as predation and competition (Harasti et al., 2018;Meekan et al., 2018). Thus, as the functional reef extent has increased, this may have simultaneously increased predation and competition, and decreased the area of the favourable habitat to Non-Exploited fish within the MPA. The decrease in the number of Non-Exploited invertebrates outside of the MPA may be linked to displacement, either of species (Dinmore et al., 2003) or fishing effort (Agardy et al., 2011).
Previous studies of the ecological response to MPAs with partial protection have had varying results (Sciberras et al., 2013), with some, like the current study, finding increases in Exploited taxa (Beukers-Stewart et al., 2005;Pipitone et al., 2000), and others finding no difference between MPAs with partial protection and control sites (Denny & Babcock, 2004;Piet & Rijnsdorp, 1998). This variability in effects of MPAs with partial protection could be attributed to many factors, such as pre designation fishing pressure, enforcement/adherence level, age of protection, size of protected area, the level of protection, as well as the sensitivity/appropriateness of the monitoring effort to detect protection effects (Babcock et al., 2010;Claudet et al., 2008;. Utilising a whole-site approach, such as in Lyme Bay, is being advocated to better protect the whole ecosystem and, by extension, lead to fisheries' increases , particularly for larger (>100 km 2 ) MPAs .
The number of sites assessed inside and outside the MPA here was not fully balanced with 36 BRUV deployments inside the MPA and 18 outside, so potentially this could be seen as a weakness in terms of the comparability of data in and out of the MPA. However, methods that are robust to uneven survey design (PERMANOVA) were used to assess difference between treatments while temporal trends were assessed by regression analyses separately for each treatment, minimising any effects of uneven survey design. This gives high confidence in the reported results.
As many taxa are used as bait by fishers, often extensively, and thus not landed (Davies et al., 2009), the separation between Exploited and target taxa is difficult to define. This creates difficulties in assessing fishing pressure on taxa that are not locally targeted or landed but are used within the fishery. Exploited taxa were defined by landings data, expert commentary and local fisher knowledge. However, the majority of the Exploited invertebrate taxa were the main target taxa of the fishers in Lyme Bay and showed lower total abundance inside the MPA as a result.
Yet, long-term increases and decreases in abundances of target species, which were only found for the Exploited fish and not the invertebrates, will be highly dependent on temporal fishing pressures (Mumby et al., 2012). Thus, to fully assess the effects of the protection to the local fishery, comparison of landings alongside abundance data could more adequately quantify any benefits or losses.
In conclusion, after 11 years of BRUVs monitoring and 12 years of protection, Lyme Bay MPA is showing a positive response in the number and total abundance of Exploited fish taxa. Increases in the number of taxa and total abundance of Exploited fish (~400% increase over 11 years) inside the MPA, which happened at the same time as an increase in static fishing, show that the protection and enforcement of the area provide benefits to both conservation and fisheries alike. Yet, inconclusive results regarding the main targeted taxa by value, namely Whelks, Brown Crab and Lobster, require further assessment, alongside fisheries landings data, to fully quantify any benefits the protection has granted the local fishery.
Regardless, this study provides further evidence of the capabilities of well enforced and monitored partial protection, which follow an Ecosystem Approach to Fisheries Management, and how the compromise between conservation and fisheries management can benefit benthic ecosystems when the whole-site approach is employed, as opposed to individual feature protection. Furthermore, it illustrates the importance and necessity of monitoring MPAs over appropriate temporal and spatial scales to aid management.

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