Impact of shorebird predation on intertidal macroinvertebrates in a key North African Atlantic wintering site: an experimental approach

Shorebirds, as migratory aquatic birds and top predators in intertidal ecosystems, can be affected by global environmental changes and escalations in local impacts on coastal lagoons and estuarine trophic networks. Many shorebirds winter in North African Atlantic coastal sites, most likely because these locations provide constant and reliable food supplies with less energy costs in comparison with the wintering sites of northern Europe. Although more information is available for other important southern coastal sites (e.g. Saharan Atlantic coastal desert and Guinean mangroves coast), very little information is available for the North African Atlantic coast. Here, we focus on the impact of shorebird predation on benthic macroinvertebrates in a major wintering site in this area—Sidi Moussa coastal lagoon, Morocco—using an exclosure experiment. For most of the macroinvertebrate species there was no significant effect of the exclusion of shorebird predation. Overall, our results do not show evidence that predation by shorebirds influenced the overall standing biomass of the benthic community. This may indicate that the benthic productivity is high enough to provide constant and reliable food supplies for non-breeding shorebirds.

Coastal systems all over the world are increasingly threatened by global environmental changes and increasing local anthropogenic impacts (Smit et al. 1987;Nicholls and Branson 1998;Anthony et al. 2009;Ducrotoy 2010;Berry et al. 2013). Migratory waterbirds, such as shorebirds, are among the groups most vulnerable to these changes (Catry et al. 2011), and as predators they may exert stronger predation pressure when the carrying capacity of these coastal systems is also under threat (Kraan et al. 2009;Ducrotoy 2010). Predation by shorebirds has proven to be a powerful trophic regulator of several communities of intertidal macroinvertebrates, especially during the wintering season and during both migration periods (Masero et al. 1999;Zharikov and Skilleter 2003).
To quantify the magnitude of the trophic link between shorebirds and benthic macroinvertebrates, several approaches have been implemented, isolated or integrated, including simulations, energy-budget analyses and experimental approaches (Kvist and Lindstrom 2003;Lopes et al. 2006;Gillings et al. 2007). To date, the main experimental methodology used to quantify predation in soft-sediment benthic communities has been the use of exclosures that prohibit predators from feeding in small patches of sediment. This technique has been performed in several intertidal systems to test a variety of factors, some of them simultaneously (e.g. different types of predation, removal of key benthic species, macroalgae cover) (Sewell 1996;Botto et al. 1998;Thrush 1999;Sutherland et al. 2000;Hiddink et al. 2002;Ferreira et al. 2005;Mendonça et al. 2007;Rosa et al. 2008;Cheverie et al. 2014).
Compared with intertidal systems to the north (e.g. Wadden Sea in the southeastern part of the North Sea, and Tagus Estuary in Portugal) and more southern Saharan coastal systems (e.g. Banc d'Arguin, Mauritania) and sub-Saharan mangrove coastal systems (e.g. Bijagós Archipelago, Guinea-Bissau), there is little information about food webs of the North African Atlantic coast and the impact of shorebirds as top predators (Kersten et al. 1983;Catry et al. 2015Catry et al. , 2016. Only very recently, the first experimental approach was performed at Merja Zerga lagoon, Morocco (Touhami et al. 2017). Sidi Moussa lagoon is part of the Sidi Moussa-Walidia wetland complex, Introduction classified as internationally important under the Ramsar Convention, and it is one of the most important staging and wintering areas for several shorebird species in Morocco (Delany et al. 2009). Its intertidal flats are used by a relatively dense community of shorebirds (21 species and a maximum of 6 000 individuals recorded) because of its strategic geographical position within the East Atlantic Flyway Qninba et al. 2006Qninba et al. , 2007El Hamoumi and Dakki 2010;Joulami et al. 2013). Despite the lagoon's ecological importance, the role of shorebird populations as potential top-down regulators on the trophic network is virtually unknown.
To provide experimental data for analysis of the trophic importance of shorebirds in this system, we performed an exclosure experiment to test the predation effect of shorebirds on the overall benthic macroinvertebrate community. The data can be used to refine other, future approaches (e.g. energy-budget and prey-availability analyses). We hypothesised that there would be no strong evidence of a decrease in abundance and biomass of the main macroinvertebrate species between control and treatment plots. This would indicate that the lagoon's macroinvertebrate community as a whole is not affected strongly by shorebird predation.

Materials and methods
Sidi Moussa lagoon, Morocco, lies parallel to the Atlantic coast and is 5.5 km long and 0.5 km wide on average; its total area is estimated at 420 ha. The experiment was carried out during autumn and winter (August 2010 to January 2011) in the lagoon's largest intertidal flat ( Figure 1). The time of year was chosen to encompass the period with the largest number of waterbirds using this habitat during the southward post-breeding migration (El Hamoumi and Dakki 2010). The particular intertidal flat was chosen because it had already been assessed as one of the preferred foraging areas for shorebirds, according to monthly shorebird censuses in all the exposed low-tide areas (Kersten et al. 1983;Joulami et al. 2008). This was confirmed by a census that was carried out during the experiment, during low tide and in all intertidal areas (Joulami et al. 2013). The experiment was performed in an area with intermediate low-tide exposure (~3.5 h), representative of the exposure of most of this mudflat.

Experimental design
We set up the exclusion experiment with two treatments: plots where birds were excluded and control plots. Each treatment was replicated five times and the plots consisted of 2 × 2 m areas. The plots were located at the same tide level in two rows, and the replicates of each treatment were interspersed 10 m apart from each other ( Figure 1). Each exclosure plot was delimited on the top with a plastic net (mesh size 2 × 2 cm) and on the sides by horizontal plastic tapes. This kept out all birds during the low tide and allowed fishes to enter the exclosure. Four wooden stakes (50 cm high and driven 30 cm into the sediment) situated at the corners of each plot supported the net and the tapes, and one pole in the middle was used for additional support of the roof net. The control plots were delimited using small stakes at the corners. The experiment plots were monitored weekly during the low-tide period to remove any drifting weed or algae, which occasionally became trapped in the corners of the cages. On these occasions all plots were also inspected for bird footprints and droppings to assess the success of the exclusion plots and the usage of the control plots (Everett 1994;MacDonald et al. 2014). To assess usage of the flat in general, integrated in the monthly census of all intertidal and supratidal areas was a count of the number of shorebirds on the flat where the experiment was located (Joulami et al. 2013).

Sampling
The duration of the experiment was almost six months, and the plots were sampled for benthic macroinvertebrates at the beginning (August), in the middle (November) and at the end (January) of the period. On each sampling date, in a different quadrant each time, five 30-cm-deep circular core samples (diameter 12 cm, area 0.01 m 2 ) were taken randomly from the inner part of each plot (50 cm from the edge), to avoid any edge effect from the exclosures. Sediment cores were rinsed through a 1-mm sieve, fixed with pH-buffered 10% formalin in water, stained with rose bengal solution (to facilitate sorting and to increase sorting accuracy) and preserved in individual containers for later sorting and identification. Later, dry weight (DW) was estimated by drying individual species at 60 °C until constant weight was reached (at least 24 h). Samples were then incinerated at 500 °C in a muffle furnace until weight constancy was reached (~2 h, depending on sample size). Ash-free dry weight (AFDW) was calculated as the difference between DW and incinerated weight.
Granulometry and organic-matter content were monitored throughout the experiment to verify whether the exclosures themselves could have had an indirect effect on the densities of macroinvertebrates, through changes in these key parameters. On each sampling date, we collected one core sample from each plot. To assess the granulometry, samples were dried in an oven at 60 °C, after which 150 g (if the sediment was sandy) or 250 g (if the sediment was muddy) of material was washed in running water in two sieves (63 μm and 2 000 μm). The remaining material in each sieve was returned to the oven until completely dehydrated and reweighed. The sedimentary facies of each station was determined as the percentage of the fraction greater than 63 μm (Switzer 2013). This percentage was used to classify sediments according to the scale adopted by Shepard (1954) and modified by Schlee (1973). The organicmatter content was estimated by drying 2-5 g of sediment, previously ground in a mortar, in porcelain crucibles and then weighing (DW). The AFDW was calculated as the difference between the DW and the incinerated weight obtained after 6 h cremation in a muffle furnace at 500 °C (Touch et al. 2017).

Data analysis
All analyses were conducted in R 3.2 (R Core Team 2015), using specific packages as required. To quantify the potential predatory pressure of shorebirds in the area during the experiment, the variation in instantaneous monthly shorebird abundance and biomass was derived from the number of birds that used the flat where the experiment was performed, obtained from the monthly low-tide census using a nonparametric regression method (LOESS). Biomass estimates were extrapolated by multiplying the number of birds by average winter shorebird biomass values, estimated from birds captured at Sidi Moussa, either by the authors or as recorded in the literature (see Supplementary Tables S1  and S2). To test for changes in conditions potentially caused by the experimental apparatus, sediment granulometry and estimates of organic-matter content were compared within a general linear mixed model to test for treatment effects.
Testing for differences in macroinvertebrate abundance and biomass between treatments was performed for each date separately to avoid confounding temporal correlations and because it was assumed a priori that temporal variation would occur (Underwood 1997). The differences in community structure (species diversity and abundance) between treatments were first visualised in two dimensions for each sampling date and treatment (control vs exclosure) using nonmetric multidimensional scaling (nMDS) on average data per plot. This was run using the metaMDS procedure in the 'vegan' package (Oksanen et al. 2015), using 100 random starts. Data were first square-root transformed and the Bray-Curtis dissimilarity measure was used. To test for observed differences, distance matrices were used in permutational multivariate analysis of variances (PERMANOVA; Anderson 2001) after checking for dispersion through analysis of the multivariate homogeneity of group dispersions, using PERMDISP2, with the functions 'adonis' and 'betadisper' in the 'vegan' package, and 999 permutations. The contribution of individual species to the overall Bray-Curtis dissimilarity was assessed with similarity percentage analysis (SIMPER; Clarke 1993) in the 'vegan' package.
Generalised linear mixed-effects models (GLMMs), as implemented with the 'glmer' procedure of the 'lmer4' package (Bates et al. 2015), were used to analyse variations of abundance (Poisson family) and biomass (Gaussian family) of the main benthic macroinvertebrate species. Treatment (exclosure vs control) was inserted as a fixed factor in the model, while plots were treated as a nested random factor, using data from all replicates.

Results
Shorebirds were abundant throughout the experimental period, with a peak in their number and biomass in the middle of the experiment (November) (Figure 2). There were no effects of the exclosure structures on the sediment, since no significant differences were shown for the percentage of clay and organic matter between treatment plots (Supplementary Figure S1). The control plots were used regularly by shorebirds, as revealed by continuous visual observations and the physical indicators observed during all visits (e.g. faeces, footprints), whereas there was no evidence that the exclosures were ineffective in deterring shorebirds.
At the beginning (August) and end (January) of the experiment we did not find differences in community structure between treatments (Figure 3). In November, the treatment plots differed significantly, as shown by nMDS ( Figure 3) and PERMANOVA (Table 1). Additionally, SIMPER showed that the gastropods Hydrobia ulvae and Nassarius pfeifferi were the main contributors to the differences in community structure (Table 2). On all dates, multivariate dispersion did not differ among treatments. Although 54 macroinvertebrate species were identified in the benthos (Supplementary Table S3), a small number of species were numerically dominant. The gastropods H. ulvae (54.4% of the total number of individuals) and N. pfeifferi (13.4%) and the bivalves Abra tenuis (6.2%) and Cerastoderma edule (5.4%) dominated the community; thus, further analyses focused on these species. We found no differences in the density and biomass of the main macroinvertebrate species between control and treatment plots at the beginning of the experiment (Figure 4; Table 3). During the experiment, only for N. pfeifferi was there a significant positive impact of exclosures on abundance and biomass (Table 3). Both metrics increased for the control plots from August to November, and then decreased in January, but always with higher estimates than for the exclosures (Figure 4).

Discussion
In the present study, no significant effect of the exclosure cages was found on the sediment composition and organicmatter content, indicating that there was no alteration in sediment structure as a result of the exclosures. In theory, the impact of exclosure experiments in an area of soft sediment may be influenced by several physical factors, including hydrodynamics, that can change the sediment structure within the exclosures (Sewell 1996). This can be prevented by using an exclosure design that minimises changes in these factors, either by increasing the size and height of each plot or by minimising the structure complexity (e.g. using tapes instead of grids). Similar results to our study have been reported in several experiments, proving that exclosures can be deployed with small impact on the sediment structure (Sewell 1996;Lopes et al. 2000;Hiddink et al. 2002;Mendonça et al. 2007). Also, there was no observational evidence that shorebirds avoided the area where the experiment was deployed and therefore we assume that the predation effect on control plots depicts actual levels of predation. The regular presence of footprints and faeces in these plots, similar to the surrounding area, indicates that the whole experimental apparatus did not influence the use of the area by shorebirds. There were no significant effects of shorebird predation on the whole macroinvertebrate community or on most of the most-abundant species. The response of the gastropod Nassarius pfeifferi was unexpected, in that we found a marked increase in the control plots. This might be explained as a response of a highly mobile species to the release of spatial competition due to predation on the other species, since gastropods of this genus are not considered as food items of small shorebirds, even when they are the dominant genus (Placyk and Harrington 2004), mainly because of their large size as adults and their shell thickness, although smaller age classes may be eaten (Skagen and Oman 1996). Moreover, we never observed shorebirds feeding on the visible, large-sized individuals of this species. This result may be likewise due   Table 3: Statistical comparison of generalised linear mixed models, between a model with the factor treatments (control and exclosure) and the null model, for the variation in abundance (Poisson link function) and biomass (linear function) of the most-abundant benthic macroinvertebrate species. Treatment (exclosure vs control) was inserted as a fixed factor in the model, while plots were treated as a nested random factor. The statistical significance of Wald chi-square tests is shown: **p < 0.01; ***p < 0.001 to other indirect causes, such as the high mobility of this species associated with the clustering behaviour of many gastropods, which may have induced a larger density of this species in the control plots. We did not find any significant effect of predation for the main prey species in this coastal lagoon, namely the gastropod Hydrobia ulvae and the bivalves Abra tenuis and Cerastoderma edule. This was also observed in the only previous study performed in North African systems (Touhami et al. 2017), as well as in a study in a European temperate estuary (Lopes et al. 2000), whereas an experiment conducted in winter in a northern European estuary found a negative effect of predation on the gastropod H. ulvae and the bivalves Macoma balthica and C. edule (Mendonça et al. 2007). Rosa et al. (2008) found a depletion effect on polychaetes, especially Hediste diversicolor, in European temperate estuaries. In other flyways, other studies have also found negative effects of predation on some species of molluscs, polychaetes and crustaceans (Sewell 1996;Botto et al. 1998). The lack of significant effects in our data might be explained by low predation pressure, below the carrying capacity of this coastal lagoon, either due to high macroinvertebrate production, even in winter, or to the use of alternative or supplementary supratidal habitats, such as the saltpans located in the vicinity of this coastal lagoon, which would increase the carrying capacity of the whole system (Kalejta 1992). Recently, a major analysis of the structure and functioning of food webs in this coastal lagoon showed, on average, large niche widths and highly overlapped isotopic niches among the main shorebird species (Catry et al. 2015), indicating that predation pressure is distributed among the invertebrate community. Additionally, analysis of the isotopic niches also indicated that supratidal saltpans can contribute to over 30% of biomass ingested by several shorebird species (Lourenço et al. 2017).
An alternative hypothesis would be that the experimental design lacked power to detect impacts smaller than a certain threshold. This could happen if one or multiple factors occurred during the experiment. For example, the exclosures might induce avoidance behaviour by shorebirds over the experimental area, including the control plots. Also, the large spatial variability typical of benthic communities and/or the high mobility of certain species could contribute to high variances, decreasing the statistical power. Additionally, green algae growth, a phenomenon very common in this estuary due to eutrophication, could increase this variability if conditions of anoxia occur (Lopes et al. 2000;Greenberg 2013). In addition, the small spatial and temporal scale of most experiments of this type might not provide enough statistical power to discern the impact of predation on benthic species unless the effect is truly important (van der Meer et al. 2001;Hiddink et al. 2002). Additionally, a bias might occur if shorebirds do not feed on all macroinvertebrates because of differing accessibility. Potentially, it would have been possible to separate each core into two or more layers (for example, at 5 cm depth) to filter bigger macroinvertebrates that live deeper in the substrate. However, even smaller shorebirds can forage on the siphons of bivalves or on polychaetes that reach depths lower than 5 cm but that can be exposed to predation when they reach for the surface (De Vlas 1985). Therefore, as a precautionary measure, we preferred to sample all the benthos. Finally, the age structure of each prey population was not taken into consideration, despite our prior knowledge that waders might select optimal foraging sizes (Oken and Essington 2015).
Although our experiment design was intended to minimise these factors, it is not possible to be certain whether they were negligible, although we can infer that the magnitude of their effect was not particularly strong. In the future, our understanding would be refined by using an integrated approach that combines experimental evidence with spatial and temporal data on benthic productivity and trophic relations. The impact of local predation can vary according to several factors, such as habitat heterogeneity (longer tidal exposure of some areas) or the phenology and cohort distribution of each macroinvertebrate species (selection  of prey sizes or preference for a prey species even at low abundances). However, this would be feasible only with a large investment of time and resources; the experimental approach adopted here provided baseline data with a smaller investment of field-based research.
In conclusion, we found no evidence that during the autumn migration and during winter there is a significant negative impact on the macroinvertebrate community due to shorebird predation. We could not determine, however, whether shorebirds would affect a particular macroinvertebrate group or species. Our primary goal was to understand whether this system was strongly affected by shorebird predation, and our results indicate that shorebirds might not be a top-down regulator in this system, a pattern common to many coastal systems in the world (Raffaelli and Hawkins 1996;Day et al. 2012;Greenberg 2013). The other potential candidate for regulation would be nekton. However, more detailed research on specific trophic relations is needed to understand predation effects better, as well as an assessment of the importance of alternative/supplementary habitats, such as the saltpans, in reducing predation pressure. It is known that the North African Atlantic coast is important as a wintering and stopover ground for many migratory shorebirds (Wetlands International 2012) and one major reason for this could be the provision of a constant and reliable food supply, with a lower energy cost in comparison with north European wintering sites (Alves et al. 2013). Our results are consistent with this understanding.
Maintenance of current levels of macroinvertebrate productivity in the Sidi-Moussa system is dependent on the implementation of measures to fight major anthropogenic pressures. This type of coastal system is highly susceptible to environmental change due to pollution from agricultural activities, which has led to increasing nutrient concentrations and degradation of water quality (El Himer et al. 2013;Stigter et al. 2014). As a result of these trends, eutrophication and macroalgal blooms are recurrent in the mudflats having a lower tidal influence, and this could have a bottom-up impact on the viability of the habitat for use by shorebirds. There is an urgent need to counteract these pressures with specific interventions to maintain the current levels of biodiversity in North African Atlantic coastal systems.