Dislodgement and mortality challenges when restoring shallow mussel beds (Mytilus edulis) in a Danish estuary

In recent decades, mussel beds in the northern Atlantic and Scandinavia have declined rapidly in extent due to anthropogenic impacts, similar to many other marine habitats. In this study, a large‐scale restoration experiment was conducted to identify major challenges that arise during restoration efforts on shallow subtidal mussel beds. Suspension‐grown mussels (Mytilus edulis) were relayed in two different treatments either directly on bare bottom sandy sediments, or on coir nets (Net), used as a proxy for suitable byssal attachment substrate. The treatments were monitored for 1.5 years and coverage (%), biomass (WW), and population dynamics were quantified. Two main challenges of shallow bed restoration were identified: (1) Lack of suitable attachment substrate resulting in dislodgment of individuals during storm events. The Net treatment had significantly higher coverage and biomass of Mytilus at the end of the monitoring period, clearly demonstrating the importance of suitable substrate at physically exposed locations. (2) High mortality of juvenile mussels. Population dynamics revealed a high mortality of juvenile Mytilus, which resulted in almost complete loss of relayed Mytilus individuals less than 30 mm within the first season. This was most likely due to high meso‐predator densities, as a result of declining top‐predator populations. The high mortality of juvenile Mytilus prevented successful annual recruitment, thereby making the population unsustainable long‐term. Both challenges need to be addressed to create stable beds during restoration. Additionally, the experiment demonstrated the viability of using suspension‐grown Mytilus as a seed‐source when restoring mussel beds.


Introduction
In recent decades, the extent of mussel beds has declined rapidly, with more than 50% in the littoral zone and more than 30% in the sublittoral zone in the North Atlantic coastal waters (Gubbay et al. 2016).Mytilus beds (Mytilus edulis species complex abbreviated as Mytilus) in Scandinavia are no exception and the disappearance of Mytilus is being reported in Norway, Sweden, and Denmark (Baden et al. 2021 and references within).These reports are consistent with local observations made in the estuary Vejle Fjord, Denmark, where Mytilus beds have been greatly reduced in extent.A video survey revealed that approximately 75% of Mytilus beds in the estuary only consisted of empty shells (Fig. 2).Similar reports have been made in the Gulf of Maine, United States, where shell beds are the testimony to declining Mytilus populations marking once stable populations (Sorte et al. 2017;Commito et al. 2018).
A multitude of factors (e.g.hypoxia and overfishing) have been ascribed as causes of decline in Mytilus populations (Sorte et al. 2017;Christie et al. 2020;Baden et al. 2021).Furthermore, high predation rates as a result of increased seabird (especially eider ducks, Somateria mollissima) and increased meso-predator densities in Scandinavian waters have received attention as a possible explanation for the disappearing Mytilus populations (Christie et al. 2020;Baden et al. 2021).Lack of top-predators and its cascading effects on lower trophic levels are well documented and can result in a dramatic density increase of smaller predators in a trophic interaction known as "meso-predator release" (Frank et al. 2005;Prugh et al. 2009;Eriksson et al. 2011).Reports of increasing meso-predator densities (e.g.green crabs, Carcinus maenas), presumably as a result of lacking predators like the cod (Gadus morhua), fits this theory (Svedang & Gwenola 2003;Eriksson et al. 2011).The ability and preference of C. maenas to predate recently settled Mytilus spat is well documented (Mascar o & Seed 2001;Murray et al. 2007;Christie et al. 2020).Increasing densities of C. maenas and other meso-predators can thereby drastically decrease the survival of juvenile Mytilus and prevent successful annual recruitment (van der Heide et al. 2014;Christie et al. 2020), eventually resulting in non-sustainable populations.
Mussel beds provide important ecosystem functions which are lost as the beds disappear.Beds of Mytilus have high water filtration rates, which directly counteract eutrophication-related effects on water clarity (Petersen et al. 2013;Timmermann et al. 2019).The derived increase in benthic light intensity can support recovery of important benthic vegetation (e.g.seagrasses) (Wall et al. 2011).Mytilus species also form biogenic reefs with dense aggregated three-dimensional structures that provide refuge for many species of fauna, combined with an increased food-supply from mussel fecal and pseudo-fecal matter (Seed & Suchanek 1992;Norling & Kautsky 2007;Orfanidis et al. 2021).The mussels are also an essential food source for many species of invertebrates, fish, and birds (Seed & Suchanek 1992) and hereby function as an important key element of the marine food web.
Restoration efforts have been made at multiple sites around the globe to palliate deteriorating bivalve populations (Lipcius & Burke 2018;Wilcox et al. 2018;Schotanus et al. 2020).However, restoration of mussel beds has proved challenging (de Paoli et al. 2015;Wilcox et al. 2018;Schotanus et al. 2020) with only few examples documenting successful long-term survival and sufficient annual recruitment to establish long-term stable populations (McDermott et al. 2008;Lipcius & Burke 2018).Identification of challenges preventing successful restoration is therefore of uttermost importance.
Restoration efforts of mussel beds in shallow and intertidal areas have revealed that dislodgement due to lack of suitable substrate is a stressor that prevents the persistence of relayed mussel beds in hydrodynamically stressed environments (de Paoli et al. 2015;Schotanus et al. 2020).Suitable substrate is essential to stabilize and increase byssal attachment strength of Mytilus beds (Christensen et al. 2015).Furthermore, Mytilus plantigrades have a clear preference during their primary settlement to settle on filamentous substrates (e.g.macroalgae and seagrasses) (Bayne 1964;Seed & Suchanek 1992) and later, during the secondary settlement, within beds of conspecifics followed by a preference for hard substrate types (Bayne 1964;Seed & Suchanek 1992;Commito et al. 2014).Lack of substrate, during both primary and secondary settlement, has been proposed as an additional possible explanation for diminishing Mytilus populations (Baden et al. 2021).In view of that, seagrass meadows and mussel beds have been severely reduced in areal extent as a result of eutrophication and other anthropogenic activities (e.g.fisheries) (Waycott et al. 2009;Baden et al. 2021), and locally in Denmark boulder extraction from the sea has resulted in a reduction of boulder reefs (Dahl et al. 2003).So, in this situation, settlement and recruitment of Mytilus could be even more severely impaired by this lack of substrate.
Dislodgement of individuals must be minimized in order for restoration of Mytilus beds to be successful in shallow environments and to allow the bed to mature and prevail long-term.Likewise, recruitment must be successful and equivalent to or higher than the mortality of the local population to ensure long-term persistence.In this paper, challenges affecting the restoration success of shallow subtidal mussel beds were investigated, thereby providing a framework for further studies to build upon.The effects of artificial substrate on byssal attachment and persistence of relayed M. edulis were investigated through a large-scale restoration experiment using suspension-grown mussels.The relay of mussels was carried out in the Danish estuary Vejle Fjord on shallow (1-1.5 m) bare bottom (BB) environments with and without coir nets added as a substrate for byssal attachment.Over a 1.5-year period, the coverage and area-specific biomass were monitored together with population dynamics that highlighted recruitment and mortality challenges.

Location
The experimental setup was made in the West-East facing estuary Vejle Fjord, Denmark.The estuary has a mean depth of 8.3 m and a length of 22 km with a surface area of 109 km2 and a catchment area of 727 km 2 (Fig. 1).The water in the estuary has an estimated residence time of 35-45 days (Miljoeministeriet 2011).Vejle Fjord is heavily affected by eutrophication, but considerable reductions in N-load have the recent decades been implemented resulting in a decrease from a mean annual load in 1989-1995 of 2303 t N/year (Skop & Sorensen 1998) to a present load of 930 t N/year (Miljoeministeriet 2023).Even so, the system is heavily affected by nutrient loads and is classified as poor ecological status in both the inner and outer part of the estuary according to EU's Water Frame Directive (Miljoegis, Danish Environmental Protection Agency, Miljoeministeriet n.d.).
Based on landing records of Mytilus edulis the estuary historically had a large population with yearly landings between 640 and 7240 t/year in the period 1994-1999 (data from the Danish Fisheries Agency).Landings have since decreased due to new regulations (Nielsen et al. 2015) and presumably a decrease in the population of Mytilus.Since 2017, the commercial fishery after Mytilus has not been conducted within the estuary.Mapping of Mytilus beds in Vejle Fjord in 2020 (Fig. 1) revealed that Mytilus beds had decreased greatly in extend with large areas consisting of shell beds.Contradictory is a large yearly settlement of juvenile Mytilus covering benthic macrophytes and sediment surfaces of the estuary observed each spring or early summer.Recruited individuals, however, disappear within weeks of the first initial sightings (T.L. Banke 2022, University of Southern Denmark, Odense, Denmark, personal observation).In the inner part of the estuary, a suspended mussel farm was set up in 2020 with high quantities of Mytilus likewise settling each year, however, in contradiction to benthic settlement, the mussels on the farm persist.
The experimental setup was located on the southern coastline in the inner part of the estuary at Sellerup Strand (55.68908828N, 9.67691766 E, Fig. 2) at a distance of 75-175 m from the shore.The location was subtidal with depths of 1-1.5 m (AE0.5 m mean sea level).The area is characterized by large areas of BB sandy sediment with patches of eelgrass.

Experimental Setup
The experiment was initiated in July 2020.Two treatments with relayed M. edulis were established, one on coir nets (Net) and one directly on BB.The experiment was constrained by the logistical difficulties of relaying a large amount of mussels and likewise allows regular monitoring.Despite these constraints, the two experimental treatments were spatially dispersed by the alternation of two subplots within each treatment.Subplots are denoted as Net1, Net2, BB1, and BB2 when relevant (Fig. 2).These subplots were dispersed with a minimum and maximum distance of 100 and 500 m, respectively (see Fig. 2).
The layout of the experimental plots was somewhat limited by the size constraints of the available coir nets (coconut fiber nets).Coir nets of the type BG ERO 400 with a mesh size of 20 Â 25 mm were used as substrate in the treatment Net.The nets were delivered in a size of 2 Â 50 m and were manually cut to a size of 2 Â 25 m.A week before the relay of mussels, the nets were soaked in sea water on site, to make them negatively buoyant, and afterwards placed on the bottom in rows.The nets were secured to the bottom using 30 cm iron pegs in 2 m intervals along the perimeter of each net.
Each subplot had the same layout consisting of three rows equivalent to the size of the coir nets (size: 2 Â 25 m) with relayed M. edulis with a spacing of 5 m between rows.Each treatment consisted of two subplots with a total of six rows (see Fig. 1).The rows in all subplots were placed perpendicular to the coast to ensure similar depth variation between rows within the individual subplots.
A total of 3950 kg of M. edulis with a size distribution of 0.5-60 mm were provided from a suspended mussel farms situated in Limfjorden, Denmark.Before delivery, the mussels were mixed to artificially represent as many size classes as possible to allow more precise tracking of individual size classes and performance.The mussels were delivered in 25 kg bags and evenly distributed within each row manually by free-divers with a biomass of 6.6 kg WW/m 2 .
A HOBO U26 Dissolved Oxygen Data logger was placed in the subplot Net1 inside a row of relayed Mytilus during the summer and fall of 2020 approximately 15 cm above the sea floor.The logger captured data in 10 minutes intervals and the aim was to monitor potential anoxic events at high-temporal resolution to be able to explain specific oxygen related loss rates.

Coverage and Biomass Monitoring
The relayed mussels in each row were regularly monitored by throwing a circular quadrat (d: 98 cm, A: 0.75 m 2 ) 12 times down the full length of the row at uniform distance.This covered ≈18% of the area in each row and was sufficient to estimate the general coverage.Each quadrat was treated as a replicate resulting in a total of 36 replicates for each subplot combined yielding 72 replicates for each treatment.Monitoring in the first period after relay was carried out in 2-4 weeks intervals and gradually reduced to 2-3 months intervals as coverage became more stable between measurements.
Monitoring at each sampling date was completed in all instances within 5 days for all subplots and aggregated for each treatment at the first sampling date.Individual subplots were monitored at all samplings within the same day.At the first sampling date (24 September 2020) only subplots Net1 and BB1 were monitored (n = 36) due to harsh weather conditions preventing further data collection in the following days.
Monitoring was conducted by free diving.In each quadrat (replicate), the coverage of live mussels (%) was visually estimated.The shells of dead mussels were usually cracked or opened and easily separated from live individuals in the coverage estimate.Additionally, the number of top-down visible European green crabs (Carcinus maenas) and common sea stars (Asterias rubens) was counted within each replicate.The number of A. rubens were divided into two categories "small (<5 cm)" and "large (>5 cm)" to get a better estimate of the potential exerted predation pressure.While the number of predators estimated using this approach certainly is an underestimation due to predators hiding underneath or in crevices of the mussels, it was deemed to be a relative measure of the potential predation pressure exerted on each treatment without using a destructive sampling technique.
Predator densities were calculated as the average density at each sampling date (individuals/m 2 ).Further, to better allow comparison between treatments, a weighted average over the entire monitoring period was calculated for each treatment.The weighted average was calculated using days between sampling dates to weigh the measured predator density (n/m 2 ) at each sampling date.
Biomass (WW) of the relayed mussels was estimated in September and October 2020, and in April and September 2021.The biomass was estimated by collecting all mussels inside a circular quadrat (same as for coverage) at six different sites with varying degrees of coverage (%) within each subplot.The collected mussels were put into nets (1 mm mesh size) and weighed in the field with a spring weight.The weight of the nets (WW) was measured separately and subtracted from the biomass measurements.Afterwards, the mussels were carefully re-laid within the same quadrat to minimize any disturbance to the performance of the treatment.This was done for all four subplots Net1, Net2, BB1, and BB2 at each sampling date.Based on the estimated coverage and associated biomass measurements, biomass-coverage regressions were made for each subplot (Fig. S1).Based on the regressions and measured coverage (%), the biomass was estimated for each replicate within the treatments and a mean (n = 72) was calculated.Biomass sampling is inherently a destructive sampling technique and therefore a lower sampling intensity was chosen.As the mussels are removed from the substrate during weighing, the byssal attachment is broken, and the affected mussels will therefore be more prone to dislodgment in the period following the sampling, as byssal attachment strength increases over time (Christensen et al. 2015).The effects on the experiment were minimized by sampling biomass in all subplots, thereby affecting all plots equally.

Remote Sensing and Camera Monitoring
Drone-based monitoring (remote sensing) was used to validate in-field coverage measurements and visually inspect treatments when conditions allowed it.A commercially available drone, DJI Phantom 4 advanced, with a 20MP RGB-camera and a field of view of 84 was used.Pictures were taken at a height of 40 m yielding an approximate ground sample resolution of 1.1 cm/px.Pictures were stitched into a georeferenced orthomosaic using the software Agisoft Metashape Professional ver.1.7.5.Object-based Image analysis of coverage was analyzed in a likewise manner as described by Svane et al. (2021) using the software Trimble eCognition Developer ver.9.5.
Drone-based monitoring was initially intended to be the primary method to estimate M. edulis coverage; however, local weather and turbidity conditions prevented a high temporal resolution and the method previously described was used instead.All flights were performed according to local legislation and in consideration of local wildlife and persons near the flight area.
Additionally, in order to gain further insight into loss processes of Mytilus, stationary underwater cameras were set up.Cameras were initially set up in August through the fall, but all cameras flooded due to faulty housings.The cameras were replaced from December 2020 to March 2021.This yielded a high-temporal visual dataset, providing further insight into the development of the relayed Mytilus.One camera of the type Garmin VIRB Ultra 30 was set up at each subplot facing the length of a row, taking one picture each day at 12:00 hours.

Sediment Analysis
Sediment acrylic core-liners (diameter: 5.2 cm) were taken in November 2020 and used for the analysis of median grain size and fraction of particles less than 63 μm (silt, clay, and organic particles).Triplicates of core-liners were taken at each subplot Net1, Net2, BB1, and BB2 in close vicinity to the relayed Mytilus.The cores were sliced in the sections 0-1, 1-2, 2-5, 5-10, and 10-15 cm and analyzed using a Malvern Mastersizer 3000 Particle Size Analyzer.

Cohort Analysis
In situ population dynamics of the relayed M. edulis were monitored using a cohort-based analysis.The cohorts were identified using the graphical analysis of polymodal frequency distributions described by Harding (1949).All analyses were done in SigmaPlot 12.5.
Cohort samples of the relayed Mytilus were collected from all subplots when coverage was monitored.An additional cohort sample was collected in January 2022.The samples were collected randomly by using a circular quadrat (d: 30 cm).The quadrat was used to mark areas of collection, where all Mytilus individuals were collected in nets with a mesh size of 0.5 mm.This approach was used to ensure that all size classes of the population were collected.From each subplot, a sample size of greater than 250 individuals was collected ensuring a large sampling size and adequate representation of the population.Every individual had its shell length measured in the lab using a digital caliper of the type MarCal 16EWRi.The measurements were aggregated into the nearest 0.5 mm size classes.
The relayed population of Mytilus was, before delivery, artificially mixed to represent many size classes and therefore did not represent a natural cohort and size-composition.Through the technique of Harding (1949), multiple discrete cohorts were identified of which the temporal development was tracked.Natural populations of Mytilus are typically polymodal as a result of yearly spawning and often multiple yearly spawnings during the summer period.Each spawning can potentially yield a different cohort.Recruitment of new juvenile cohorts was detected and tracked during the monitoring period through the same technique by Harding (1949).
The mean (AE1 SD) shell length for each cohort and cohort composition was calculated at each sampling date.Mean length, growth rate, and recruitment rates were compared between subplots and treatments, and no difference was found.The samples were therefore pooled (n = 1083-2058) at each sampling date, which greatly increased the strength of the analysis.

Applied Statistics
Statistics were performed in the software SigmaPlot 12.5.All statistics were tested with an α level of 0.05.
All datasets failed Shapiro-Wilk's normality test.As a result, a non-parametric Mann-Whitney U test was used to test for significant differences between the two treatments Net and BB.Potential difference in coverage of M. edulis and biomass was tested on the final sampling date.Potential differences in predator densities were tested on the weighted average density for each of the monitored predators C. maenas, large A. rubens, and small A. rubens.
Spearman rank order correlation was used to test if predator densities (C.maenas, large A. rubens, and small A. rubens) were significantly correlated with coverage (%) of M. edulis within each replicate across treatments (n = 72-144) at each sampling date.

Sediment Conditions
All subplots had sandy sediment with a median grain size of 216-292 μm in the surface layers (0-2 cm) and low silt and clay fractions of sediment particles.This fraction generally accounted for less than 9% of total particles in all subplots (Fig. S2).

Oxygen Concentration
During the monitoring period from 10 July 2020 to 2 December 2020, a combined 40.8 hours of moderate oxygen depletion (<4 mg O 2 /L) and 1 hour of severe oxygen depletion (<2 mg O 2 /L) were measured.The longest consecutive period of sustained moderate oxygen depletion was 10.7 hours (Fig. S3).

Coverage and Biomass Changes
At the end of the experiment, a significantly higher coverage (Table 1) of Mytilus was found in the treatment Net compared to the treatment BB with 39.5 AE 3.1 and 9.9 AE 1.7%, respectively (Fig. 3).Decreases in coverage were measured during fall and winter 2020-2021 in both treatments, but with a distinctly smaller decrease in the treatment Net (Fig. 3).During spring 2021, an increase in coverage was observed in the treatment Net, while no increase was evident in the treatment BB.The nets in the subplot Net2 became partially buried during the duration of the experiment due to sand mobility, thereby increasing dislodgement and disappearance of relayed Mytilus from this subplot and slightly decreasing the performance of the treatment (Fig. 3).
The biomass of the relayed Mytilus were at the final sampling date likewise significantly higher (Table 1) in the treatment Net and consistently increased during the experimental period from the initial biomass of 6.6-11.0AE 0.9 kg WW/m 2 .Oppositely, the treatment BB decreased to an average biomass of 2.7 AE 0.5 kg WW/m 2 at the end of the experiment (Fig. 4).
Camera monitoring revealed that the mussel coverage mainly decreased during weather events with high physical impact, this indicates that losses were mainly controlled by dislodgement.Losses were especially large during high wind speeds from an eastern or north-eastern direction (data available at DMI.dk/vejrarkiv/, Danish Meteorological Institute) that had the longest possible fetch distance in relation to the placement of the treatments.Multiple weather events with dislodgement of mussels were identified during the period December 2020 and February 2021 (Fig. 3) with mean windspeeds and gusts exceeding greater than 4.8 m/s and greater than 11.3 m/s, respectively.One storm event in December 2020 even dislodged mussels with wind from a southern direction and consequently small fetch area, but with gusts of up to 21.4 m/ s.The largest disappearance of mussels visually confirmed happened during a storm in February 2021 (Fig. 5) with mean windspeeds and gusts from northeast up to 8 and 18 m/s, respectively.Additionally, a decrease in coverage in the BB treatment was measured in the period from 24 September 2020 to 8 October 2020, which most likely could be attributed to a storm event in this period with mean windspeeds of 8.1 m/s and gusts of 18.5 m/s.However, this was not possible to confirm due to the flooding of the cameras.
Drone footage validated the described trends of changes in coverage, with a clear visual difference between treatments after dislodgement had occurred during fall and winter in the BB treatment.Most marked were the losses of Mytilus within the subplot BB1, with near complete losses of relayed individuals (Fig. 6).
The average density of Carcinus maenas (Fig. S4A), large Asterias rubens (Fig. S5C), and small A. rubens (Fig. S5A) over the monitoring period were all significantly higher (Table 1) in the treatment Net.Predator abundances within the relayed Mytilus beds were greatly reduced during winter and C. maenas abundance increased again in spring 2021.The increase in abundance was much higher in treatment Net than treatment BB (Fig. S4B).The abundance of A. rubens remained low for the rest of the experiment duration (Fig. S5B & S5D).
Correlation analysis between Mytilus coverage and predator densities across treatments showed that the presence of predators in the relayed mussels at most sampling dates was proportional to the prevalence of Mytilus (Table 2).At the first sampling date, no   significant correlation between predator densities and coverage of Mytilus could be found.In the following monitoring in October, a significant positive correlation could be found between the densities of C. maenas and the coverage of Mytilus.No significant correlation was found in the following period until the period July-September 2021, where C. maenas again was positively significantly correlated with Mytilus coverage.Densities of small and large A. rubens were at all times positively correlated with the coverage of Mytilus except at the first sampling date.

Growth and Recruitment
No differences in growth rate or recruitment could be found between the two treatments and subplots.Samples used for cohort analysis were therefore pooled and used to describe general tendencies in the population dynamics.
Eight different cohorts of Mytilus edulis could be identified at the time of the relay.The cohorts grew consistently, with an increase in the largest cohort (C1) from 45.7 to 65.1 mm during the monitoring period (July 2020-January 2022), equivalent to a mean growth rate of 13.1 mm/year (Fig. 7).Two new cohorts were found in 2021, one in June, denoted as C9 and the second in September 2021, denoted as C10.These new cohorts had in September 2021 a mean length of 10.7 and 5.8 mm, respectively (Fig. 7).It should be noted that the calculated mean length of cohorts C4-C8 (Fig. 7) contains some uncertainty after the end of the relay year (2020) as mortality in these had drastically decreased the number of individuals present within each sample (Fig. 8).
At the time of the relay in July 2020, the population of M. edulis was dominated by small individuals in cohorts C8 and C7 with a mean length less than 6 mm of which the individuals constituted 53% of the population.The slightly larger cohorts C6 to C4 with a mean size less than 30 mm comprised 6% of the population, while the largest cohorts C3 to C1 represented 3.6, 4.9, and 32.2% of the entire population, respectively.The proportion of individuals in cohorts C8 and C7 greatly decreased during the fall of 2020 and represented only approximately 1% of the total population in January 2021.The number of individuals in cohorts C6 to C4 decreased, but to a lesser extent, to 1.5% of the population.Cohort C3 decreased from 3.5 to 0.5%, while cohorts C2 and C1 had greatly increased proportionally at this point consisting of 8 and 87% of the entire population, respectively (Fig. 8).A similar loss of juvenile individuals was observed in 2021, when the cohorts C9 and C10 (mean length <11 mm) decreased from 3.3% of the population in September 2021 to 0.5% in January 2022.C1 comprised 94.3% of the total population at this point (Fig. 8).

Identified Challenges
Two main challenges were identified during the relay experiment of Mytilus edulis: (1) Dislodgement of individuals during intense weather events due to unsuitable substrate and (2) high mortality of juvenile Mytilus.Both challenges are crucial to overcome to assure long-team survival of relayed shallow Mytilus beds.High dislodgement rates make restoration attempts unreliable, as mussels are transported to areas outside the selected restoration site.This transport can result in individuals being subjected to suboptimal or even fatal growth conditions, like deeper waters with prevalent hypoxia.The high mortality of juvenile Mytilus has two mayor implications for restoration efforts.It prevents successful annual recruitment, thereby making the population unsustainable long-term, and affects size selection of mussels that can initially be used for restoration.

Dislodgement
Results in this study clearly demonstrate the need for a suitable substrate for byssal attachment to prevent dislodgement when restoring shallow Mytilus beds, coinciding with results found

C. maenas
Small A. in Schotanus et al. (2020).Camera-based monitoring confirmed that dislodgement of Mytilus happened during events with high wind speeds and high physical impact, resulting in a higher loss rate in the BB treatment.
Coir nets was an effective measure to reduce dislodgement of Mytilus in this investigation.The effect of the coir nets at subplot Net2 was however limited as these were partially buried by local sand mobility.Dislodgement increased after the partial burial of the nets at Net2, as the relayed Mytilus lost the ability to sufficiently attach to the substrate.These effects indicate a limited application of coir nets at locations with high sand mobility.Similar findings were reported by de Paoli et al. (2015), where extensive burial of the coir nets resulted in almost total loss of relayed Mytilus.However, the nets act as a proxy and demonstrate the need for suitable substrate for byssal attachment as a prerequisite to successfully restore shallow Mytilus beds in  hydrodynamically stressed environments.Other suitable substrates include mussel shells (Christensen et al. 2015;Schotanus et al. 2020) or hard substrate types like stone and gravel (Seed & Suchanek 1992;Commito et al. 2014).Visual inspections of the BB treatment in this study confirmed that any Mytilus remaining after storms were usually firmly attached to stone or shell material embedded in the sediment.Restoration of shallow Mytilus beds should be optimized through site-selection that target areas with high density of stones or shells.Shell material indicates the presence of a former mussel bed and thus enhances the probability of the area being able to support mussels.Furthermore, targeting areas with suitable substrate would reduce the workload associated with anchoring coir nets or add other suitable substrates.Site-selection through the use of small-scale pilot relays has further potential to increase restoration success by revealing location-specific differences that could affect restoration outcomes (Benjamin et al. 2023).
It should be noted that even though the general consensus is that Mytilus prefers to settle and thrive on hard or filamentous substrates, it is also common in Denmark to find Mytilus beds on soft bottom substrates in sheltered or deep areas with low physical exposure (Nielsen et al. 2015).The findings in this paper therefore reflect the challenges when establishing beds at shallow locations with exposure to wave impacts, which are prevalent in many coastal areas.
In this study, the mussels were relayed in rows perpendicular to the coast.This approach made treatments comparable in terms of water depth and exposure.However, it is unknown how or if the relay pattern in the specified rows affects the self-organization of the mussels and the formation of resilient patterns when deployed on a landscape level (de Paoli et al. 2017;Bertolini et al. 2019).Further studies with a focus on large-scale relay patterns and densities would be beneficial to increase insight into the processes increasing resilience toward dislodgement during the initial phase after relay.
Predation by Asterias rubens and Carcinus maenas of the relayed Mytilus was observed in all plots but could not be determined to have any effect on the relative difference in coverage and biomass of Mytilus between the two treatments.All monitored predator densities were significantly higher in the Net treatment, not coinciding with the BB treatment having the highest loss rate and lowest Mytilus coverage.For most monitoring periods, a significant positive correlation between Mytilus coverage and densities of C. maenas and A. rubens was found across treatments.However, a significant correlation between C. maenas and Mytilus could not be demonstrated in the period November-April, most likely because of the low activity levels of C. maenas during the winter with low water temperatures (Aagaard et al. 1995).This trend suggests that predator densities were mostly governed by the availability of prey, Mytilus, and that each treatment were affected by predation proportionally equally.

High Mortality of Juvenile Mytilus
Cohort analysis revealed a high mortality of juvenile Mytilus in the months following the relay.Cohort samples were pooled across treatments and therefore represent the general population dynamics of the relayed mussels.Cohorts less than 6 mm at the time of relay (C8 and C7) were substantially affected and only a few individuals remained at the end of the relay year.Cohorts ranging 10-30 mm (C6 to C4) were affected to a lesser but still substantial extent, while cohorts greater than 30 mm were least affected.To reduce mortality following relay, we propose that a large proportion of the relayed Mytilus should be greater than 30 mm in shell length, especially as the Mytilus are particularly vulnerable to predation at the initial stages after relay (Capelle et al. 2016).A proportionally higher number of large individuals will minimize the initial mortality, thereby increasing the likelihood of a successful establishment.
Recruitment of Mytilus was observed during the following season after relay.The newly settled mussels on the number of individuals represented 3.3% of the entire population in September 2021, but quickly decreased to 0.5% by the end of the year.At the time of relay in July 2020, Mytilus of a similar size as newly recruited individuals (<6 mm) in C8 and C7 likewise within the same season exhibited high mortality with a reduction from constituting 53% of the population at relay to approximately 1% of the population in January the following year.This juvenile mortality is extreme and results in a negligible annual recruitment that, if continued, will result in the population not being self-sustaining and potentially total loss of the population within a few years as the largest and oldest individuals succumb to senescence.
Coverage and biomass were in these populations largely governed by adult individuals and the high mortality of juveniles was therefore negligible compared to the effect of dislodgment during storm events.Even so does the loss of juveniles poses an important challenge even if not immediately apparent on the coverage of the bed.Lack of sufficient recruitment is reported for multiple bivalve restoration efforts (van der Heide et al. 2014;Wilcox et al. 2018;Benjamin et al. 2023) and is one of the main challenges preventing long-term sustainability of the populations.Paradoxically, this is reported even in instances with high larval supply (van der Heide et al. 2014;Benjamin et al. 2023) suggesting settlement processes or post-settlement juvenile mortality in these instances are the primary limiting factor of successful recruitment.In this study, the experimental site had a large annual larval supply as observed by large settlement and survival on suspended structures within the estuary.This observation, combined with the high mortality of juvenile individuals after initial relay, indicate that post-settlement mortality is the most likely explanation for the lack of annual recruitment of new individuals in Vejle Fjord.Furthermore, the high survival of Mytilus on a suspended mussel farm 4 km from the experimental site suggests water parameters like salinity and temperature, which are similar at both locations, are not the main mortality factors for juvenile Mytilus either.
Oxygen depletion above the sediment was limited in this study, with almost no occurrence of severe oxygen depletion (<2 mg/L) and only short consecutive periods of moderate oxygen depletion.Mytilus species are very tolerant against prolonged periods of hypoxia (Wang & Widdows 1993;Babarro & de Zwaan 2002), making oxygen depletion an unlikely factor for the high mortality of juvenile Mytilus.
The high mortality was most likely due to high local densities of meso-predators.Juvenile Mytilus are predated by numerous invertebrates and vertebrates, but C. maenas have received particular attention as a heavy top-down control on Mytilus (Mascar o & Seed 2001;van der Heide et al. 2014;Christie et al. 2020).Especially, as this meso-predator seems to thrive in ecosystems affected by anthropogenic disturbance where larger predators are scarce (Eriksson et al. 2011;Støttrup et al. 2020).Increasing C. maenas densities have generally been observed in Danish waters but are especially prevalent in Vejle Fjord, which exhibits some of the highest C. maenas densities in Denmark (Støttrup et al. 2020).Especially, high top-down control on Mytilus populations by this meso-predator can therefore be expected.The common starfish (A.rubens) is also a common predator on Mytilus populations.However, A. rubens have been reported to have a 23 times lower consumption rate of juvenile Mytilus compared to that of C. maenas (Kamermans et al. 2009).Even so, large quantities of juvenile A. rubens were observed shortly after relay in this investigation, potentially adding a substantial predation pressure on juvenile Mytilus.Other species, that were present at the study site, likewise predate on juvenile Mytilus, such as green sea urchins (Strongylocentrotus droebachiensis) (Seed & Suchanek 1992) and the goldsinnywrasse (Ctenolabrus rupestris) (Christie et al. 2020) further contributing to heavy top-down control on the Mytilus population.
Predation pressure is highly size-dependent in Mytilus populations (Seed & Suchanek 1992;Kamermans et al. 2009) and predation dramatically decreases with the increasing size of Mytilus, when crabs are the primary predators (Mascar o & Seed 2001;Murray et al. 2007;Kamermans et al. 2009).In this investigation, the smallest cohorts had a drastically higher mortality, which decreased with increasing size of the cohorts, indicating that C. maenas could be one of the primary mortality factors for juveniles.
The high mortality of juvenile Mytilus found in this study support the hypothesis proposed by Christie et al. (2020), that increased predation pressure of juveniles could be one explanation for the recent observed declines in Mytilus populations in Scandinavia.However, multiple potential pressures affecting the observed decline in Mytilus should not be disregarded (Baden et al. 2021) with dominant pressures potentially being diverse between systems.In this study location, a prerequisite for successful restoration of Mytilus populations is thus a significant reduction in predation pressure by meso-predators.Exclusion of predators by cages or fences has proven to drastically increase the survival of new recruits (van der Heide et al. 2014), however, this method is not suitable for large-scale implementation (Schotanus et al. 2020).Human top-down control of specific meso-predators by concentrated fishing could be another possible approach.Strategic fishing of the predator A. rubens is already conducted in another Danish estuary, Limfjorden, to reduce predation pressure on mussel populations (Aguera et al. 2021).To our knowledge, very little literature is available on this topic (Calderwood et al. 2016;Aguera et al. 2021) and further research is needed.Fundamentally, a broader ecosystem-restoration approach could be required to help reduce trophic cascades and meso-predator densities.Such an approach would involve requirements to further reduce anthropogenic impact (Eriksson et al. 2011) and concurrent restoration of other lacking marine habitats like seagrasses (Orth et al. 2020) and boulder reefs (Wilms et al. 2021).

Relay Using Suspension-Grown Mytilus
In this investigation, the relay was conducted using suspensiongrown M. edulis.Mussels grown on suspension or other structures have previously for Mytilus (Capelle et al. 2014;Schotanus et al. 2020) and other species of bivalves, e.g.green-lipped mussels (Perna canaliculus) (Wilcox et al. 2018;Benjamin et al. 2023;Toone et al. 2023) been successfully used for relay and restoration purposes.This study adds another example of a successful relay using seed-mussels from suspension and highlights its viability for restoration purposes.
Mussels grown on suspension have similar mortality rates due to predation following relay, compared to wild mussel seed (Kamermans et al. 2009;Christensen et al. 2012).Furthermore, use of suspension-grown mussels for restoration purposes is advantageous, as this removes the need for dredging or harvesting from natural populations, thereby preventing potential negative impact of these.Mussel farms have a high area-specific production (Lindahl et al. 2005) limiting the area needed for obtaining large quantities.Sedimentation is however increased beneath such facilities enriching the sediment below with organic content and potentially increasing local oxygen demand (Timmermann et al. 2019).Conversely, mussel farms can also have a positive effect on water quality by lowering phytoplankton concentrations and sedimentation rates downstream (Timmermann et al. 2019). in the ecology group at SDU for help with sediment analyses and student helper L. Wagnersen for help in relaying the mussels.A special thanks to Professor M. Pardal from University of Coimbra, Department of Life Sciences, Portugal, for his useful feedback on the handling of data and on the contents of the paper in the initial stages of writing.

Figure 2 .
Figure 2. Coverage (%) of Mytilus beds and mussel shells in Vejle Fjord at depths greater than 3 m.Areas with a coverage less than 5% have been omitted to precisely represent the extent of the beds.Lower left corner shows the position of Vejle Fjord within Denmark.Mapped in 2020 using video-sledge transects with a spacing of approximately 200 m in the length of the estuary totaling 256.3 km of continual transects.Data from a pre-study of the mussel population within the estuary (Banke et al. unpublished data).

Table 1 .
Results of Mann-Whitney U test comparing parameters of the two treatments Net and BB.The ranked sum for the treatment Net (R Net ), Mann-Whitney U Statistic for the treatment Net (U Net ) and p value shown for each measured parameter.Significant differences (p < 0.05) highlighted with bold.Sample size for all parameters n 1 = n 2 = 72.Differences in coverage and biomass was tested on the final sampling date, while differences in predator densities were tested on the weighted average density over the monitoring period within each treatment.C. maenas = Carcinus maenas, A. rubens = Asterias rubens.

Figure 3 .
Figure 3. Coverage (% AE SE) of Mytilus edulis over the monitoring period (month/year) within the two treatments Net and Bare Bottom (BB).Vertical dashed lines (red) indicate storm events with dislodgement of Mytilus that could be confirmed by camera footage.

Figure 4 .
Figure 4.Estimated wet biomass of Mytilus edulis (kg/m 2 AE SE) over the monitoring period (month/year) within the two treatments Net and Bare Bottom (BB).Vertical dashed lines (red) indicate storm events with dislodgement of Mytilus that could be confirmed by camera footage.

Figure 5 .
Figure 5. Timelapse of relayed Mytilus edulis before, during and after a short period of pronounced wind gusts and physical impact (4 February 2021-12 February 2021) within all subplots with net (Net1 and Net2) and bare bottom (BB1 and BB2).

Figure 6 .
Figure 6.Drone footage of all subplots with net (Net1 and Net2) and bare bottom (BB1 and BB2) at the start of the experiment (5 August 2020) and after winter with weather events imposing high physical impact (14 April 2021).

Table 2 .
Results of Spearman rank order correlation analysis between Mytilus coverage (%) and predator densities (n/m 2 ) across treatments.Correlation analysis shown for each monitoring date (dd-mm-yy).Spearman's correlation coefficient (r s ) and p value shown, significant correlations (p < 0.05) highlighted with bold.Number of samples n = 144 except for the first sampling where n = 72.C. maenas = Carcinus maenas, A. rubens = Asterias rubens.

Figure 7 .
Figure 7. Development in mean length (mm AE SD) of Mytilus edulis individuals in each identified cohort from 2020 (C1-C8) and from 2021 (C9 and C10) over the monitoring period (month/year).Dashed lines indicate sampling intervals where mean length of specific cohorts could not be estimated due to a low number or lack of individuals in the cohort.

Figure 8 .
Figure 8. Development in the percentagewise contribution of each identified cohort to the total population of relayed Mytilus edulis from 2020 (C1-C8) and from 2021 (C9 and C10) over the monitoring period (month/year).Legend and columns are stacked in the same order.