Identifying how restoration measures influence the presence of shorebirds: a case study on the use of artificial structures for restoring mussel reefs

Artificial structures are often used as a tool for habitat restoration and the recreation of degraded coastal ecosystems and their associated food webs. However, it is often unknown if and how these artificial structures may influence the habitat use of target species, thereby hampering restoration goals. In this study, we test how artificial barriers, deployed to enhance the creation of an intertidal mussel bed, influenced the mussel‐habitat use by two bird species under pressure, Eurasian oystercatchers and Eurasian curlews. Average bird presence was monitored using time‐lapse camera's from the start of the mussel bed restoration in August 2018 until April 2019. We found that in the first few months of the experiment, both oystercatchers and curlews avoided the restored mussel beds containing artificial structures in the form of metal barriers that act as traps for dislodged mussels. Thereafter, the presence of barriers had no negative influence on the average presence of oystercatchers or curlews. In fact, curlews were significantly attracted to the mussel beds with barriers in January and February. In addition, we found a negative effect of the presence of European herring gulls on the presence of oystercatchers and curlews. The higher herring gull densities in the first 2 months of the experiment might explain the lower‐than‐expected curlew and oystercatcher densities observed in August. To our knowledge, no other studies have investigated the effect of artificial restoration structures on the abundance or habitat use of specific shorebirds.


Introduction
Habitat-forming organisms, or so-called ecosystem engineers, facilitate complex food webs and have inordinately large effects on the abundance of other organisms (Jones et al. 1997;van der Zee et al. 2016;Borst et al. 2018).In addition to being part of the food web as predator or prey, ecosystem engineers also have non-trophic effects on associated organisms by creating new habitats, altering resource availability, and modifying physical environmental conditions (Jones et al. 1994).This is especially true for coastal ecosystem engineers that provide structural complexity in otherwise predominantly flat soft sediment landscapes, such as seagrasses (Attrill et al. 2000), mangroves (Nagelkerken et al. 2010), oysters (Karp et al. 2018), and mussels (Arribas et al. 2014).Because ecosystem engineers strongly affect communities through environmentally mediated interactions (Crain & Bertness 2006), it has been suggested that these kinds of systems should be a key priority for conservation and restoration (Bouma et al. 2009).Understanding how the restoration of ecosystem engineers and the use of specific restoration methods, such as engineering measures, influence the rehabilitation of target species should, therefore, be a priority in restoration ecology.
Engineering measures are increasingly used to diminish or counteract limiting abiotic (e.g.wave action, erosion) and/or biotic (e.g.predation) factors (Schotanus et al. 2020;Temmink et al. 2020) or to stimulate the establishment of self-facilitating feedback processes (Schotanus et al. 2020).Diminishing environmental stressors can potentially be used to create a window of opportunity for establishment (Balke et al. 2014) and constitute a valuable tool for the restoration of ecosystem engineers in highly dynamic ecosystems.However, there has been limited scientific research on the ecological impacts associated with the implementation of engineering measures in restoration projects (Bilkovic & Mitchell 2013).Food webs of restored systems can differ from natural systems, especially if artificial structures have been used (Mossman et al. 2012).If rehabilitation of biodiversity is a primary goal of a restoration project, it is important to identify how the use of artificial structures can influence the presence and distribution of target species.
Restoring intertidal mussel beds may serve as an important measure in rehabilitating shorebird populations by halting the continued erosion of mudflats and providing essential foraging and resting grounds.Intertidal blue mussels (Mytilus edulis) are coastal ecosystem engineers that form the foundation of complex food webs on mudflat ecosystems (Commito et al. 2008;Donadi et al. 2013).Loss of intertidal areas and foraging grounds, such as mussel beds, due to coastal development and sea-level rise, are considered the leading cause of shorebird population declines worldwide (Galbraith et al. 2002;van Roomen et al. 2012;Studds et al. 2017).In the 1970s, Zwarts et al. (2011) determined that a quarter of all birds in the Dutch Wadden Sea were concentrated on intertidal mussel beds, although mussel beds covered only 3% of the intertidal flats.Previous research showed that artificial structures, in the form of metal barriers, 3-m long and 0.25-m high, placed in strategic locations between transplanted mussels, can help to kick-start the establishment of a mussel bed (Schotanus et al. 2020).By acting as mussel traps, the barriers prevented dislodged mussels from getting washed out by currents and waves, and stimulated the formation of a large-scale banded pattern comparable to the patterns found in natural intertidal mudflat mussel beds (van de Koppel et al. 2005).Many restoration sites tend to be small (Atkinson 2003) and the use of artificial structures may greatly improve the scalability of restoration efforts (Schotanus et al. 2020;Temmink et al. 2020).
Engineering measures used for restoration goals, such as mussel trapping barriers, may have unforeseen effects on shorebird presence.For example, the artificial barriers increase structural complexity, which may attract organisms (Borst et al. 2018), such as crabs, which are an important food source for bird species such as the curlew (Goss-Custard & Jones 1976;van der Zee et al. 2012).On the other hand, the artificial barriers may also reduce the line of sight and thereby increase the level of perceived risk (Metcalfe 1984).Eurasian curlews and Eurasian oystercatchers are both species known to be sensitive to human activity (Stillman & Goss-Custard 2002) and have been recorded to stop feeding or fly away when disturbed (Fitzpatrick & Bouchez 1998).Reduced visibility may ultimately cause foraging birds, which are cautious by nature, to avoid mussel beds that have been restored with the help of artificial structures, and thereby not comply with the original aims set for intertidal mussel bed restoration.While the effect of structures on bird behavior has not been studied much from a restoration perspective, studies have been conducted on the effect of intertidal aquaculture operations, such as the use of oyster culture racks or oyster longlines, on shorebirds (Connolly & Colwell 2005;Maslo et al. 2020).Some of these studies have shown that shorebird species increased foraging around these aquaculture structures (Hilgerloh et al. 2001;Connolly & Colwell 2005), while other studies showed that the presence of oyster culture structures did not influence habitat use and feeding rates (Maslo et al. 2020).This contrast in findings emphasizes the need for similar studies in restoration.
In this study, we investigated how a restoration design that uses artificial structures for the restoration of intertidal blue mussel (M.edulis) beds (Schotanus et al. 2020), influenced the presence and distribution of shorebirds.We evaluated the use of mussel trapping barriers in low and high density and investigated how this restoration method influences the distribution and presence of Eurasian oystercatchers and Eurasian curlews.A particular focus was put on oystercatchers and curlews as their populations have been declining (Roodbergen et al. 2012).In addition, each of these species differs in its degree of association with mussel beds and may, therefore, be influenced differently by the created mussel bed and the barriers.For instance, mussels are an important food source for oystercatchers (Goss-Custard & dit Durell 1983;Ens & Goss-Custard 1984), whereas associated organisms, such as crabs and shrimps, are key prey for curlews (Goss-Custard & Jones 1976;van der Zee et al. 2012).The average presence of herring gulls was included in the analysis because herring gulls were observed to chase away oystercatchers and curlews, which can significantly influence the distribution and presence of oystercatchers and curlews.
As mussels are an important food source for oystercatchers (Hilgerloh et al. 1997), we expect that their presence will be positively correlated to mussel coverage, but also that presence of barriers will have a negative effect on their presence.Since curlews mainly prey on species associated with mussel beds (Ens & Alting 1996;Zwarts et al. 2011), we expect that their presence will be positively correlated with mussel coverage, but we also hypothesize that they will avoid the plots with barriers because they barriers reduce the line of sight and foraging areas without any barriers are close by.To test whether a higher density of barriers in a mussel bed would increase the perceived risk of shorebirds, which would likely reduce the mean presence of shorebirds, barriers were placed in two densities.

Study Site
A large mussel restoration experiment was conducted from 6 August 6 2018 to 16 June 2019, on a wave-exposed intertidal mudflat (Viane), in the Oosterschelde estuary located Southwest of the Netherlands (51.616211, 3.992755).This intertidal mudflat is part of the Natura 2000 network, meaning that free access without a special permit is prohibited.The experimental mussel beds were positioned approximately 400 m from the shore.The experiment was located at an elevation conducive to mussels, with an average inundation time of approximately 60%.The Viane mudflat is characterized by sandy sediment and experiences erosion with an average net rate of 15 mm/year (Salvador de Paiva et al. 2018).Since the construction of a stormsurge barrier in 1986, the tidal flats in the estuary have slowly been eroding (Santinelli & de Ronde 2012), which has subsequently resulted in the reduction of suitable foraging grounds for shorebird species (van Roomen et al. 2012).Natural intertidal mussel beds have since disappeared and intertidal mussels can only be found incidentally on commercial mussel culture plots located at sheltered sites, in natural intertidal oyster reefs or attached to other hard substrates, such as wooden poles and stones.

Experimental Setup
To investigate the effect of the restored mussels and the barriers on the presence of oystercatchers, curlews and herring gulls, a total of 12 experimental mussel beds (30 Â 40 m) were constructed in three blocks of four plots (Fig. 1A).Approximately 32,000 kg of mussel seed (juvenile mussels), originating from spat mussel collectors situated in the Oosterschelde, and with an average length of 1.7 AE 0.04 cm and a condition index of 4.5 mg/cm 3 AE 0.08 SE (n = 64), was transplanted on the plots by a commercial mussel farmer no more than 12 hours after harvest.Mussel seed was transplanted on the experimental plots through shafts below the waterline while the vessel moved in circular patterns over a block of four mussel plots at once.Although this seeding method is common in commercial seeding of mussels, it is difficult to be very precise.As a result, mussel densities were heterogeneously distributed over the plots, in concentric patterns similar to those found on on-bottom mussel culture plots (Fig. 1B & 1C).Despite these shortcomings, this method did give us the opportunity to create mussel beds on a large scale.
Mussel seed was transplanted in four configurations (30 Â 40 m), each with three replicates (12 plots in total): (1) low mussel coverage, no barriers; (2) high mussel coverage, no barriers; (3) high mussel coverage, 45 barriers; and (4) high mussel coverage, 91 barriers (Table 1; Fig. 1).In the first and second mussel bed configuration, mussel seed was transplanted without any engineering measures, with a low average initial coverage (151 m 2 AE 40 SE, 12.6%, n = 3) in the first configuration, and a high average initial coverage (521 m 2 AE 74 SE, 43.4%, n = 3) in the second configuration.In half of the plots, barriers were placed between the transplanted mussel seed.The barriers were constructed using concrete mesh overlaid with 8 mm chicken wire, formed into an elongated pyramid shape measuring 3 m in length, with equivalent sides of 0.25 m (Fig. 1D).These barriers were similar to those successfully tested in Schotanus et al. (2020), with the difference that they did not have to be partially buried in the sediment to prevent dislodgement by hydrodynamics, but could be easily anchored with 30-cm long metal pins.The barriers were placed parallel, in rows of 30 m, to the incoming waves coming mostly from the Southwest.In the 45-barrier configuration (i.e.configuration 3), mussels (294 m 2 AE 67 SE, 24.5%, n = 3) were transplanted between barriers, which were evenly distributed in 9 rows of 5 barriers spaced 3 m apart.In the 91-barrier configuration (i.e.configuration 4), mussels (509 m 2 AE 76 SE, 42.4%, n = 3) were transplanted between barriers, which were evenly distributed in 13 rows of 7 barriers spaced 1 m apart.The relatively low initial mussel density in the low-coverage barriers configuration was not intentional, but rather the result of the large-scale seeding method.

Monitoring Mussel Bed Coverage over Time
The mussel coverage development over time was monitored by taking monthly top-view pictures with a drone of every plot.Because mussel coverage variation over time is expected to be an important explanatory factor for the average presence of birds, we briefly explain how mussel coverage was monitored, while the results of this monitoring can be found in Figure S1.The drone pictures were converted to black and white images in ImageJ, with white pixels indicating bare sediments and black pixels mussels and other hard substrates, such as barriers and marking poles.Marking poles and barriers within mussel plots were converted from black to white pixels by hand.The drone pictures were then converted into pixel-density maps using the package Imager of R Studio.Subsequently, a grid with a sample of eight-pixel intervals (number of pixels and the distance between those pixels in the x and y direction) was placed over each image.Grid squares with 25% or fewer black pixels were annotated as background noise and, therefore, not included in the determination of the total mussel coverage.Grid squares with more than 25% black pixels were converted to proportion mussel coverage within a plot.

Monitoring Average Presence of Shorebirds
Five meters in front of each experimental mussel plot, a camera was placed on a 5-m-high pole (Fig. 1E).For 10 months, these wild cameras made three photos in quick succession of the front part of a plot every 15 minutes.Taking three consecutive photos allowed birds to be identified more quickly by their movement between the three photos.The number of Eurasian oystercatchers (Haematopus ostralegus) and Eurasian curlews (Numenius arquata) were counted every 15 minutes during four low tides every month from August to April.In each month, we specifically chose the four longest low tides occurring during daytime for our observations.The monitoring results of the average presence of herring gulls, including monthly variations and the correlation with mussel coverage, can be found in Figure S2.The month of March is missing from the dataset as during this month, the plots did not emerge four times due to stormy weather.The counting of birds was stopped after April because of a mass mortality event throughout the Oosterschelde (Capelle et al. 2021).
As birds were counted every 15 minutes, individual birds may have been counted multiple times on consecutive photos, which cannot be distinguished from individuals moving between locations.Hence, we use the term "average bird presence" as an outcome of our camera counting.The counting of birds started when approximately 50% of the mussel plot on the photo was above the water and stopped when 50% of the mussel plot was submerged again.The average count duration during one low tide was approximately 2.89 hours AE 0.03 SE.Besides the count duration, the count area of a plot could slightly differ between low tides because the camera direction sometimes shifted a little due to remounting at a slightly different angle or displacement of the camera due to stormy weather.Therefore, the plot area visible on a picture was recalculated for each counted low tide and was, on average, around 651.25 m 2 AE 2.70 SE (n = 12), which is slightly more than 50% of a plot.To account for the differences in count duration and count area between plots and low tides, the average number of birds was calculated as the number of birds per hectare per hour (nr ha À1 hour À1 ).To see if the monthly fluctuations in the average presence of oystercatchers and curlews that occurred in this experiment somewhat corresponded to natural fluctuations, we compared our data with data Table 1.Overview of the four mussel bed configurations.The high variance in initial mussel coverage between plots was not intentional but a result of the seeding method, where a vessel releasing mussels through a shaft moved in circular patterns over three mussel plots at once.

Statistical Analysis
All statistical analyses were carried out in R. All models were visually validated for normality (Q-Q plot) and homogeneity of residuals.Models were simplified according to Akaike's information criterion scores, and nonsignificant factors were removed.When terms were significant, post hoc comparisons were carried out to test for significant differences between specific configurations.

Statistical Analysis on the Effect of Barriers and Mussel Coverage on Average Bird Presence
For each month, the average presence of oystercatchers and curlews on the various plots was analyzed with a generalized linear mixed model with a Poisson family for count data using the lme4 package.The full models contained the fixed factors: the presence of barriers (without barriers, low-density barriers, high-density barriers), mussel coverage, average presence of herring gulls on plots, and an interaction between the presence of barriers and mussel coverage.The mussel coverage and the average presence of herring gulls were rescaled to prevent model instability.All full models contained plots nested in blocks as random factors.To account for overdispersion, an observationlevel random effect was included.

Average Presence of Oystercatchers
When the average oystercatcher presence per month in the current study is compared to the average number of oystercatchers per month estimated by Sovon (Dutch Centre for Field Ornithology) in 2018 (Fig. 2), the seasonal trends are broadly similar.
Only in August and April, the average presence of oystercatchers seems to be below the average monthly trend.
In August, September, and January, the presence of barriers had a significant negative effect on the average presence of oystercatchers inside a plot (resp.p < 0.001; p < 0.001; p = 0.04).In August, more oystercatchers were counted in plots without any barriers than in plots with 45 barriers (post hoc, p = 0.007) or in plots with 90 barriers (p < 0.001).In addition, more Oystercatchers were counted in plots with 45 barriers than in plots with 90 barriers (p < 0.001).In September, a similar pattern was shown.Significantly more oystercatchers were counted in plots without any barriers than in plots with 45 barriers (post hoc, p = 0.03) or in plots with 90 barriers (post hoc, p = 0.02).And just like in August, more Oystercatchers were counted in plots with 45 barriers than in plots with 90 barriers (post hoc, p < 0.001).
In August and September, there was a significant positive interaction between mussel coverage and the number of barriers inside a plot (resp.p < 0.001; p = 0.009).In September and January, mussel coverage within a plot had a significant influence on the average presence of oystercatchers (resp.p = 0.007; p = 0.05).The number of herring gulls counted inside a plot did have a significant negative effect on the presence of oystercatchers in August (p < 0.001) and September (p = 0.02).The monthly average presence of herring gulls can be found in Figure S2.

Average Presence of Curlews
When the average presence of curlews per month is compared to the average number of curlews per month estimated by Sovon (Dutch Centre for Field Ornithology) in 2018 (Fig. 3), the seasonal trends are broadly similar.That is to say, only at the start of the experiment, in the months of August and September, the average presence of curlews seems to be below the average monthly trend estimated by Sovon.
In August, the overall presence of curlews was relatively low and the degree of mussel coverage or the presence of barriers within a plot had no significant effect on the average presence of curlews.In September and November, the presence of barriers did have a significant negative effect on the average presence of curlews (resp.p = 0.02; p = 0.05).However, in January and February, there was a positive significant effect of the presence of barriers on the presence of curlews (resp.p = 0.005; p = 0.05).Mussel coverage had a significant positive effect on the average presence of curlews in December ( p = 0.003) and January ( p < 0.001).The average presence of herring gulls (Fig. S2) had no significant effect on the presence of curlews in any of the monitored months.

Discussion
Intertidal mudflats and mussel beds are important foraging habitats for many shorebird species (Galbraith et al. 2002), and loss or degradation of these areas is a significant cause of migratory shorebird declines globally (van Roomen et al. 2012;Waser et al. 2016;Studds et al. 2017).In this study, we examined the effect of artificial restoration structures employed to restore intertidal mussel beds on the abundance or habitat use of specific shorebirds.Our findings reveal negative effects in terms of areaavoidance to be short-term (i.e.first few months), whereafter the presence of barriers had no longer a negative influence on the average presence of oystercatchers or curlews.

Possible Negative and Neutral Effects of Artificial Structures
One of the main findings of the current study was that during the first 2 months of the experiment, August and September, both oystercatchers and curlews were counted significantly less in plots containing barriers than in plots without barriers.The barriers also had a significant negative effect on the average presence of oystercatchers in January and on curlews in November, although these effects were less pronounced.These findings indicate that both curlews and oystercatchers initially avoided the barriers, but, over time, became more accustomed to the new structures or that the new habitat provided a more suitable food source through fouling on the barriers or changing sediment characteristics that affected the infaunal community.This largely corresponds with similar observations in studies done on the effect of intertidal aquaculture structures, such as oyster culture racks or longlines, on the habitat use of shorebirds.In some cases, shorebirds, including oystercatchers and curlews, occurred in significantly lower numbers in areas with oyster racks when other suitable foraging habitats were nearby (Hilgerloh et al. 2001;Burger & Niles 2017).However, there are also studies that found no significant effect of the presence of untended aquaculture structures, while tending the oyster cultivation structures by oyster farmers reduced the probability of shorebird presence (Maslo et al. 2020).
The experimental site was visited once a month by researchers to monitor the mussel coverage development and to collect pictures of the wild cameras.These visiting days were excluded from the analysis.Apart from the researchers, no people entered the research area because free access without a special permit was prohibited.Thus, the lower number of birds in plots with barriers at the beginning of the experiment cannot be attributed to the presence of humans, but is probably a direct effect of the presence of the barriers.This is supported by our finding that the negative effect on average bird presence in the first 2 months of the experiment was stronger when more barriers were present in the mussel plot.The reason why the birds avoided the barriers at the beginning of the experiment is difficult to determine.The barriers did not appear to increase the perceived risk by reducing the line of sight to approaching predators, humans, or neighboring birds (Metcalfe 1984), as the birds did not continue to evade the barriers for an extended period of time.A more likely explanation is that they, by definition, initially avoid new structures.

Possible Positive Effects of Artificial Structures
Surprisingly, significantly more curlews were counted in the plots with barriers than in plots without barriers during the winter months of January and February.Previous studies have shown some shorebird species increased foraging around intertidal aquaculture operations (Hilgerloh et al. 2001;Connolly & Colwell 2005).This increase could be caused by an increase in the availability of prey species in the sediment surrounding the structures or in prey species that use the oyster racks or longlines as a substrate (Hilgerloh et al. 2001;Connolly & Colwell 2005).The main food source of curlews on the mudflats is marine worms and crabs (Zwarts et al. 2011;van der Zee et al. 2012).An estuarine birds survey (Zwarts et al. 2011), conducted on the intertidal flats in the Oosterschelde in 2011, reported that 39% of the winter diet of curlews consisted of shore crabs.Although most shore crabs migrate to subtidal grounds during the winter period (Zwarts et al. 2011), the barriers may have provided refuge for juvenile crabs, creating a food hotspot for curlews in the winter.

Bird Observations in Relation to Mussel Coverage
The average presence of curlews was positively correlated to mussel coverage in the months of December and January.This corresponds with findings in previous studies, in which curlew abundance was found to be unaffected (Ens & Alting 1996) or positively affected by the construction of a new intertidal mussel bed (Caldow et al. 2003).Bird species that do not feed on mussels, like curlews, but instead feed on the animals living within the mussel bed, are expected to be less responsive to mussel beds in comparison with molluscivorous shorebirds such as oystercatchers (Ens & Alting 1996).Hence, it was unexpected that we observed a positive correlation between the average presence of oystercatchers and mussel cover in our experimental mussel beds solely during the months of September and January.In a previous study, an experimental intertidal mussel bed attracted many oystercatchers, who came to feed on the transplanted mussels (Ens & Alting 1996).A likely explanation for the relatively weak response of oystercatchers to the experimental mussel beds could be the size of the mussels.The mussels used in this experiment had an average length of 17 mm, while oystercatchers tend to select mussels between 30 and 45 mm in length (Meire & Ervynck 1986).Another possible explanation could also be the lack of a reference area without mussels in our experimental design.

Bird Observations in Relation to Interspecific and Intraspecific Interactions
Herring gulls were present were present in large numbers right at the start of the experiment, where they probably fed on damaged mussels, starfish, and crabs that had been transported inadvertently with the mussels (Ens & Alting 1996).When this food supply had apparently been depleted, the number of gulls dropped again in October.The average presence of oystercatchers was significantly influenced by the average presence of herring gulls in August and September.Herring gulls have been observed to use oyster culture racks as perch, from which they kleptoparasitized, steal food, from oystercatchers feeding underneath (Hilgerloh et al. 2001).Kleptoparasitism also occurs between oystercatchers, when densities reach more than 50 birds/ha (Ens & Goss-Custard 1984;Triplet et al. 1999).Intraspecific and interspecific competition for food can play an important role in determining habitat use (Triplet et al. 1999).This may also explain why, relative to expected seasonal population trends in the Oosterschelde, low curlew and oystercatcher densities were counted at the experimental site in the first 2 months when the herring gull densities were very high.

Outlook
For the construction of mussel beds, we deliberately chose to use small mussels because they are easy to obtain and because they can easily adapt to harsh intertidal conditions, unlike mature subtidal mussels (de Paoli et al. 2015;Schotanus et al. 2019).However, this implies that the mussels were not a favorable size for oystercatchers to feed on (Meire & Ervynck 1986).Had the mussel beds continued to develop further, instead of dying off, the presence of appropriately sized mussels would have increased, increasing the habitat's suitability for oystercatchers.Furthermore, the behavioral response of the oystercatchers and curlews to the barriers and mussels changed over time and might have changed again.For instance, curlews preferred the mussel beds containing barriers during the winter months but may have moved out of these more complex habitats to the bare mudflats during the summer when shore crab abundances are very high and easy to spot and catch.In other words, long-term research is needed to establish the restoration impact of specific methods on a more temporal scale.
Despite shorebirds being an important component of coastal ecosystems, few intertidal habitat restoration schemes have specifically targeted this group (Atkinson 2003).Many restoration sites tend to be small and the factors controlling the restoration need to be better understood to create high-quality habitats.In summary, our findings demonstrate that artificial structures can play a crucial role in restoring foraging habitats for shorebirds, particularly in estuaries undergoing degradation due to human alterations.However, it is important to note that habitat utilization and interspecies interactions can be affected differently by the presence of barriers for different species and foraging strategies.Understanding how artificial structures influence the habitat use of target organisms over a longer period, is essential to adapt restoration methods to target species.Thus, for the creation of a new habitat, it is important to evaluate how these techniques will eventually influence target species and how these effects may change over time and cascade through the ecosystem.

Figure 1 .
Figure 1.(A) Map of the 12 experimental mussel beds divided into three blocks on the mudflat.(B) Top-view stitched picture taken with a drone a few days after transplantation of the mussel seed.(C) Top-view stitched picture taken with a drone 3 weeks after transplantation of the mussel seed.(D) Mussel trapping barrier in the field.(E) Wild camera was placed on a 5-m-high pole in front of every experimental mussel bed.

Figure 2 .
Figure 2. (A) Bar graph: Number of Eurasian oystercatchers (Haematopus ostralegus) per hectare per hour counted over time in plots without barriers (n = 6) with 45 (n = 3) and 90 barriers (n = 3).The error bars represent AE SE.Line graph: The average number of oystercatchers in 2018 estimated by Sovon divided by 100 to give an indication of seasonal population dynamics (https://stats.sovon.nl/stats/gebied/1000118). (B) Correlation between the number of Eurasian oystercatchers counted per hectare per hour and the percentage of mussel coverage, for every monitoring month, for experimental plots without barriers (n = 6) with 45 (n = 3) or 90 barriers (n = 3).Significant factors and interactions ( p < 0.05) are denoted for every month (B = barriers, M = mussel coverage, B:M = interaction barriers: mussel coverage, and H = herring gulls).

Figure 3 .
Figure 3. (A) Bar graph: Number of Eurasian curlews (Numenius arquata) per hectare per hour counted over time in plots without barriers (n = 6) with 45 barriers (n = 3) and 90 barriers (n = 3).The error bars represent AE SE.Line graph: The average number of curlews estimated by Sovon divided by 600 to give an indication of seasonal population dynamics (https://stats.sovon.nl/stats/gebied/1000118). (B) Correlation between the number of Eurasian curlews counted per hectare per hour and the percentage mussel coverage, for every monitoring month, for experimental plots without barriers (n = 6) with 45 (n = 3) or 90 barriers (n = 3).Significant factors and interactions ( p < 0.05) are denoted for every month (B = barriers, M = mussel coverage, B:M = interaction barriers: mussel coverage, and H = herring gulls).