Oyster abundance on subtidal reefs depends on predation, location, and experimental duration

ARTICLE Special Feature: Honoring Charles H. Peterson, Ecologist Oyster abundance on subtidal reefs depends on predation, location, and experimental duration Nathan R. Geraldi 1,2 | Maria L. Vozzo 3 | Stephen R. Fegley 1 | Andrea Anton 4 | Charles H. Peterson 1 1 Department of Marine Sciences, University of North Carolina at Chapel Hill, Institute of Marine Sciences, Morehead City, North Carolina, USA 2 Department of Bioscience, Aarhus University, Silkeborg, Denmark 3 Sydney Institute of Marine Science, Mosman, New South Wales, Australia 4 Global Change Research Group, IMEDEA (CSIC-UIB), Mediterranean Institute for Advanced Studies, Esporles, Illes Balears, Spain Correspondence Nathan R. Geraldi Email: [email protected] Funding information North Carolina Recreational Fishing License Grant; National Estuarine Research Reserve System; Juan de la Cierva, Grant/Award Number: IJC2018036527-I Handling Editor: Sean P. Powers Abstract Predation affects community structure and functioning within marine habitats. Predator–prey interactions can change through space and time. Documenting how these interactions change is essential to improve our understanding of food web dynamics and to enhance our ability to manage preferred species. In this study, our goals were to determine whether the density of subtidal oysters (Crassostrea virginica) differed spatially by looking at three separate restored oyster sanctuaries within Pamlico Sound (North Carolina, USA), whether oyster density changed over an interval of 16 months, and whether oyster density was related to the presence of different-sized predators by using an experimental approach. Multiple exclusion treatments were used in situ to exclude, selectively, different predator guilds from consuming oysters. Predator densities were also measured both within experimental treatments and on the restored oyster reefs by using multiple survey techniques. We found that oyster abundance differed among the four sample dates over the 16-month study and differed among the three sites. Mud crabs—one of the smallest predators measured—had the greatest predator biomass per unit of area, but the presence of other predators was largely site-dependent. Oyster abundance was affected by the exclusion of all predators, but this was dependent on sample date and location, which may suggest that mud crabs were the only predator to reduce oyster abundance in this study. In addition, large predators may have affected small predators, such as mud crabs and oyster drills, which were more abundant in treatments where large predators were excluded. The strongest evidence for top-down effects on oyster reefs occurred at one of the three field sites at the first and final sampling time, suggesting that predator effects are complex, as well as spatially and temporally variable. Field experiments that assess variables through time and at multiple locations are needed as this information could improve the success of oyster reef conservation and restoration efforts. Received: 2 March 2021 Revised: 24 June 2021 Accepted: 12 July 2021 DOI: 10.1002/ecs2.4087 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2022 The Authors. Ecosphere published by Wiley Periodicals LLC on behalf of The Ecological Society of America. Ecosphere. 2022;13:e4087. https://onlinelibrary.wiley.com/r/ecs2 1of13 https://doi.org/10.1002/ecs2.4087 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License KEYWORDS biogenic habitat, body size, environmental gradients, food web, in situ experiment, intraguild predation, oyster reef, Special Feature: Honoring Charles H. Peterson, Ecologist INTRODUCTION Predators can influence community structure (Carpenter & Kitchell, 1988;Hairstonetal.,1960;Paceetal.,1999), directly by reducing prey abundance through consumption or indirectly by altering species interactions that occur within the ecosystem (Denno et al., 2003;Paine,1969). However, predation does not occur in isolation and other drivers, such as competition and recruitment, may also alter communities with all these drivers varying in relative importance with location (Boyce et al., 2015; Navarrete & Manzur, 2008; Peterson, 1979), time (Menge & Olson, 1990), and environmental conditions (Menge & Olson, 1990; Navarrete et al., 2000). The many factors that affect predator–prey interactions make it inherently difficult to extrapolate results from controlled experiments to natural systems, mainly because of the complexity of natural food webs (Holt & Huxel, 2007). Community structure results from an interplay of many abiotic and biotic factors. For example, the sizes of organisms within a community influence the strength and number of interactions (Berlow et al., 2009; Emmerson & Raffaelli, 2004), as well as community structure (Petchey et al., 2008) and stability (Emmerson & Raffaelli, 2004). In addition, interaction strength of predation may be altered if smaller organisms are more abundant than larger individuals (Kalinkat et al., 2013). Therefore, thoughtfully designed in situ experiments coupled with surveys of natural abundances are needed to improve our understanding of how community structure changes through time and space. Oyster reefs can serve as a model system to study the effects of food web interactions and environmental context on a biogenic system. Focusing on biotic factors, oyster reefs are exposed to multiple ecological attributes that vary within estuaries including: (1) oyster recruitment and survival (Geraldi et al., 2013); (2) the type and abundance of oyster predators (Garton & Stickle, 1980); (3) individual oyster predator biomass, which can vary by multiple orders of magnitude (0.4–1800 g); and (4) intraguild predation (Geraldi, 2015; Johnson et al., 2014;O’Connor et al., 2008; Figure 1a). In addition, understanding ecosystem functioning and structure, and how these can vary spatially, is important for oyster reefs because they provide multiple essential ecosystem services (Grabowski & Peterson, 2007). Oyster reefs exist at only a fraction of their historic population (Ermgassen et al., 2012; Kirby, 2004) and, as such, are a focus of conservation and restoration efforts around the world (Beck et al., 2011). They can form both intertidal and subtidal reefs, and in recent years, there has been growing interest in creating oyster sanctuaries (b) (a) FIGURE 1 Representation of an oyster reef food web with the interactions among the predators of oysters (arrows point in direction of consumption), (a) and the design of the five different cage treatments and the respective excluded predators on the right (b). Animal illustrations courtesy of the Integration and Application Network, University of Maryland Center for Environmental Science (UMCES, Cambridge, MD, USA) 2of13 GERALDI ET AL. 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License for the purpose of habitat and oyster larval provision for adjacent habitats (e.g., Powers et al., 2009). In particular, subtidal oyster sanctuaries can create recreational fishing opportunities (Peterson et al., 2003). The success of subtidal sanctuaries may be influenced by a variety of environmental or anthropogenic factors such as water flow or illegal harvesting (Powers et al., 2009). As such, there is a growing need to understand whether the mechanisms that shape intertidal oyster communities, similarly influence those of subtidal reefs. We utilized a combination of mensurative surveys and field experiments to assess whether the abundance of eastern oysters (Crassostrea virginica) was affected by manipulating access of different predators within a natural subtidal marine food web at three different locations within an estuary. Although numerous studies have quantified the effect of multiple predators on intertidal oyster reefs across a range of environmental conditions (e.g., Grabowski et al., 2008, 2020; Johnson et al., 2014; Kimbro et al., 2020), predation of oysters can vary significantly between intertidal and subtidal locations (Johnson et al., 2014). Specifically, we monitored oyster abundance over 16 months within different cage designs to exclude predators based on their size and morphology, and surveyed predator abundance at all locations (Figure 1b). Specifically, the following null hypotheses were examined, whether: (1) oyster abundance would be similar through time; (2) oyster abundance would be similar at different locations; (3) oyster predator abundances would be similar at different locations; and (4) access to oysters by different predator guilds would have no effect on oyster abundance. This study builds on an extensive amount of research on oyster reefs and predator–prey interactions, but was novel in that it documents the effects of many predators in situ, including the assessment of abundance of both crab and fish predators, at multiple sites over multiple years. METHODS Study system This study was conducted at three oyster sanctuaries in Pamlico Sound, North Carolina, the largest sound on the east coast of the United States (Clam Shoal, Crab Hole, and Gibbs Shoal; Figure 2). We did not have prior knowledge of abiotic conditions or the assemblage of oyster predators at these three sanctuaries. Oyster harvest is illegal within the sanctuaries, although recreational take is allowed. These oyster sanctuaries were created from 2006 to 2010, and each consisted of 50–300 constructed mounds of riprap marl rock. Each mound was approximately 3-m high, in a mean water depth of approximately 4 m, with a footprint diameter of 15 m. The mounds were placed in a uniform grid with individual mounds separated by approximately 25 m (Geraldi et al., 2013; Mroch et al., 2012). Experimental design To separate the effect of predator type (e.g., predator species, abundance, and size) on oyster abundance in different estuarine settings, we used five types of cage treatments (Figure 1b). All cage treatments contained juvenile oysters (C. virginica) that settled onto nine cleaned adult shells. Juvenile oysters were produced from hatcheryraised larvae and naturally settled onto the cleaned adult shells (dead harvested shell) in flow through tanks (see Geraldi et al., 2013, for details). Juvenile oysters on each shell were counted, and the shell height of five juvenile oysters per shell was measured for size frequencies (greatest width of shell). Adult shells of similar size were chosen to keep surface area available to recruitment consistent. To mimic natural variation in recruitment, juvenile oysters per shell were not standardized. The adult shells were attached to plastic mesh surrounding a cement block (34 20 8 cm) by a cable tie through a hole drilled in the adult shell prior to oyster settlement. Crab Hole Gibbs Shoal Clam Shoal 0 20 40 km Pamlico Sound N USA Atlantic Ocean FIGURE 2 The location of the study sites within Pamlico Sound, NC ECOSPHERE 3of13 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Individual cement blocks were labeled with an identification number, and the shells (with juvenile oysters) on each cement block were organized so that each shell could be uniquely identified. Each cement block was then haphazardly assigned to a cage treatment type (henceforth referred to as treatments). Five different treatments consisted of: open (allowed all predators); roof (roof made with a mesh of a hole size of 2.5 cm 2 to allow large crabs to feed on the oysters, but excluded large fish that feed by approaching the oysters from above, primarily sheepshead—Archosargus probatocephalus); large mesh (hole size of 2.5 cm 2 that excluded large fish and crabs, blue crab—Callinectes sapidus and stone crab—Menippe spp.); small mesh (hole size of 1 cm 2 that excluded most predators, including adult mud crabs—family Xanthidae—and oyster drills— Urosalpinx cinerea); and cage control (similar to small mesh but with only two walls and half the top covered in mesh). The cage control treatment allowed access to all predators but controlled for confounding effects of the cage, such as changes in water flow and/or animal recruitment (Figure 1b). The roof treatment was like the large mesh treatment, but a section of the side of the enclosure was removed so that there was a 5-cm gap in the mesh between the block and the roof (Figure 1b). Preliminary mesocosm experiments were conducted to ensure the effectiveness of predator exclusion and found no difference in oyster survival between open and roof treatments in the presence of blue crabs (large crabs fed on oysters in the roof treatment) and between roof and large mesh treatment in the presence of sheepshead (sheepshead were excluded from the roof treatment). It was not feasible to include controls for each cage type and the cage control with small mesh was chosen as this likely represents the mesh size to have the greatest effects on relevant parameters such as flow. One replicate of each cage treatment with oysters was deployed on top of 10–12 oyster mounds at each sanctuary (a total of 10 or 12 replicates per sanctuary), which were deployed at 2 separate times, during a single summer season, to mimic natural oyster recruitment. At each sanctuary, cage treatments were deployed on six or eight mounds on 30 June to 2 July 2010 and four additional mounds on 18–24 August 2010. This resulted in a total of 170 cage treatments that included 1530 adult oyster shells containing >12,000 juvenile oysters. Divers ensured that all cage treatments were on top of each oyster mound and positioned the cage treatment so that the oysters in the treatment were at approximately the same height as the surrounding substrate of the mound. Cage treatments were sampled four times over the 16-month study. Sample dates were 18–24 August 2010, 7–14 October 2010, 18 May to 3 June 2011, and 8–13 September 2011. Sampling consisted of divers placing the cage treatments in plastic crates lined with 1-cm plastic mesh that were then hauled onto a boat. Five out of nine adult shells within each cage treatment were haphazardly chosen, and all oysters on these adult shells were counted and the shell height (greatest length from umbo to other side of shell) was measured for five haphazardly selected oysters on each adult shell (25 oysters measured per cage treatment). To mimic natural settings, oysters were not standardized after each sampling and natural recruitment did occur. Mud crabs and oyster drills within each cage treatment were counted after thoroughly searching the cement block, oysters, and the crate used to bring the cage treatment to the boat during the fall 2010 and spring 2011 samplings. To determine the abundance of small predators within each of the treatments, the size of all mud crabs (carapace width) and oyster drills (shell height as the longest distance from lip to apex) were measured during the spring 2011 sampling. The treatment, with predators removed, was then returned to each respective oyster mound. Predator surveys Natural predator abundance at each of the three oyster sanctuaries was quantified using four different sampling techniques. Collection bins, traps, and visual surveys were needed because of the wide range in predator sizes and life history. Mud crab and oyster drill abundance were measured by deploying plastic bins (31.5 16.5 10 cm) onto oyster mounds. The bins contained 3 cm of cement at the base and were filled with the surrounding oyster reef material (i.e., marl rock and oyster shells). A bin was deployed atthecrestandbottomedgeofeachmoundonwhichwe had deployed cage treatments. Bins were deployed twice for 6–8 weeks in early (May–July) and late summer (July– September) of 2010 to quantify the abundance of mud crabs and oyster drills. Bins were retrieved by divers and placed directly into 5-mm mesh cloth bags underwater to reduce the chance for fauna to escape. Contents of the bins were thoroughly searched, and all fauna >5 mm were retained. Samples were brought back to the laboratory, and all individuals were counted and massed (wet mass). The carapace width of up to 20 haphazardly selected mud crabs was measured from each bin to obtain size frequencies of this oyster predator. The relative abundance of large crabs (blue and stone crabs) was quantified using two techniques. First, wire traps (71 71 43 cm) covered in plastic mesh (1-cm square openings) and baited with frozen fish were used to catch large crabs. Six traps were deployed on different oyster mounds at each oyster sanctuary for 4 h during 4of13 GERALDI ET AL. 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License the day on three dates from May to October in 2010. The abundance, size (as carapace width), and mass of trapped individuals were recorded. Second, divers conducted visual surveys of large crabs and fish. Large crabs were counted within 50 50 cm quadrats haphazardly placed by divers six times on each reef mound with a treatment. The number and species of crabs (C. sapidus or Menippe spp.) were recorded within each quadrat. Quadrat sampling was conducted on five different occasions from May to October in 2010 and 2011 when visibility was >0.5 m. The abundance of large fish (primarily A. probatocephalus) was visually quantified by divers on the same day crabs were counted in quadrats, but only when visibility was>2m.Adiverswam10malongatransecttapefrom the base of the mound to over the top of the mound. All sheepsheadthatoccurredwithin1monbothsidesand2m ahead of the diver (20-m 2 transects) were counted. The total length of each fish was visually estimated within 5-cm increments. Three transects were conducted consecutively on a mound to attain a robust average number of sheepshead per 20 m 2 . Visual surveys were conducted on 1–4mounds(2–4 times throughout the experiment) at each oyster sanctuary during cage sampling and at least once per season. Predator biomass The biomass of each predator was determined differently depending on the predator. The biomass of individual mud crabs ≥10mmwasdeterminedfromawell-fittedregression (R 2 =0.98) between biomass and carapace width of mud crabs that was calculated from 22 crabs (B=0.00091 L 2.7 ). The mean weight of blue crabs caught by traps was used to estimate the biomass for each individual detected visually in quadrats. The biomass of individual stone crabs was calculated by multiplying the abundance by the mean individual biomass (28 g) of similar-sized crabs collected on oyster reefs in North Carolina (O’Connor et al., 2008). The biomass of sheepshead was calculated from diver-based size estimates and a length to biomass conversion for sheepshead (W=0.0296 L 3.05 ;Schwartz,1990). Statistical analysis Given that many of our dependent variables (e.g., counts of oyster abundance) were skewed toward 0, we tested multiple generalized linear mixed-effects models (family =Poisson, nbinom1, or nbinom2; which all had log links), both with and without zero inflation (glmmTMB function within the glmmTMB package; Brooks et al., 2017). The most appropriate model was chosen based on the lowest Akaike information criterion, and the fit of the model was checked by comparing figures of fitted values versus residuals and fitted values versus actual data. For the continuous dependent variable (oyster length), we used a similar process but with Gaussian family models with either log, identity, or inverse links. Independent fixed categorical factors were sanctuary, treatment (open, cage control, roof, large mesh, and small mesh cages), and sample date (spring 2010, fall 2010, spring 2011, and fall 2011). Changes in oyster abundance on shells deployed on mounds primarily resulted from natural recruitment and location differences, and not deployment date (Geraldi et al., 2013), so we pooled all data regardless of deployment date. All interactions were included except for oyster drill abundance for which the model would not converge and only main effects were modeled. A random variable was also included in the model to account for the nested design of blocks on mounds within each sanctuary (1jsanctuary/mound/treatment). Although there was a potential for nonindependence among sample dates to affect the statistical results, part of our goal was to determine the difference among sample dates, and nonindependence was minimized by randomly measuring five of nine adult oyster shells during each sampling and removing predators within cages. Five separate statistical analyses were performed with the following dependent variables: total number of oysters per shell, number of juvenile oysters per shell (<2.5-cm shell height), oyster length, number of mud crabs per treatment, and number of oyster drills per treatment. Only one oyster sanctuary, Clam Shoal, had oyster drills on the treatments, and the model only tested for the difference among treatments (open, cage control, roof, large mesh, and small mesh) at that location. Results of statistical models were presented using ANOVA summaries using the Anova function within the car package (Fox & Weisberg, 2011),whichisbasedontheTypeIIWaldchisquared tests, to make findings clear and concise. Differences among factor-level combinations were determined by estimated marginal means adjusted for multiple comparisons using the emmeans function (adjust =“tukey”) within the emmeans package (Lenth, 2020). All statistics were conducted using R software version 3.5.1 (R Core Team, 2017). Difference in predator abundance and biomass among sanctuaries was tested for all mud crabs, large mud crabs (crabs with carapace width greater than 10-mm readily consume juvenile oysters; personal observation, Nathan R. Geraldi), oyster drills, blue crabs, stone crabs, and sheepshead. The abundance and biomass of oyster predators, other than mud crabs, had a high frequency of zeros (previously described linear models did not converge for multiple predators), and a nonparametric Kruskal–Wallis test was used to determine significance among sanctuaries and between seasons for these predators. This test cannot include random factors, and ECOSPHERE 5of13 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License the mean predator densities were calculated to obtain one value per mound at each sanctuary. The mean density measure of predators taken at each mound was used as a replicate to minimize the potential for pseudoreplication (nonindependence of samples not accounted for with a random variable) and to reduce differences in the number of samplings among sanctuaries and sampling techniques. Abundance data from underwater quadrats and traps for large crabs had the same order when ranked by abundance; consequently, quadrat data were used for the analysis to estimate density. RESULTS Effects of predator exclusion on oyster and predator abundance The number of oysters on shells within treatments was influenced by the three-way interaction among sanctuary, treatment, and sample date (Table 1,Figure3). The small mesh exclusion treatments were the only cages to have more oysters than the open treatments, but this was only significant based on the multiple comparison test among treatments at the same sanctuary and sample date at the first and last sample date at Crab Hole (August 2010 [Figure 3b] and September 2011 [Figure 3e]; Appendix S1: Figure S1). However, there was evidence of this effect with open treatments having, on average, approximately half of the oysters compared to the small mesh treatments at Gibbs Shoal at the first and final sampling dates (Figure 3b,e) and at Crab Hole at the fall (October) 2010 sample date (Figure 3c; Appendix S1: Figure S1). There were four other instances of significant differences among treatments at the same site and sample date, including more oysters within large mesh than open treatments at the first sampling at Crab Hole (Figure 3b); fewer oysters within both the large mesh and roof compared to the small meshtreatmentatthefinalsamplingatCrabHole (Figure 3e); and fewer oysters within roof treatments compared to small mesh treatments at the final sampling at Gibbs Shoal (Figure 3e; Appendix S1: Figure S1). One sanctuary (Clam Shoal) had a high abundance of oysters during initial samplings but few oysters during the later samplings, which likely caused a significant interaction between sanctuary and sampling date (Figure 3a–e). The spring sampling period (May 2011; Figure 3d)hadfewer oysters than the other sample dates that occurred after oyster recruitment periods during the summer. The number of juvenile oysters on shells in treatments displayed similar patterns to the total number of oysters on shells and depended on the interacting effects of sanctuary, treatment, and sample date (Table 1, Figure 3). The lowest number of juvenile oysters per shell occurred in spring 2011 (Figure 3i). The difference in juvenile oysters was significant among treatments within the same sanctuary and sample date at the final sample date with small mesh treatments having more juvenile oysters than the open, roof, and large mesh treatments at Crab Hole and more juvenile oysters in the small mesh treatments compared to the roof treatments at Gibbs Shoal (Figure 3j; Appendix S1: Figure S2). The number of mud crabs within treatments was influenced by the main effects of sanctuary, treatment, and sample date (Table 1, Figure 4a,b). Overall, the small mesh treatments had more crabs than the roof, control, or open treatments, but this was dependent on sanctuary and sample date (Figure 4a,b; Appendix S1: Table S1). Additionally, the large mesh often had an intermediate number of crabs compared to the small mesh, and the other three treatments (Figure 4a,b; Appendix S1: Figure S3). In contrast to mud crabs in treatments, oyster drills were only observed in the treatments at Clam Shoal and were influenced by treatment type and sample date (Table 1). In general, oyster drill abundance was higher in small, large, and roof treatments compared to open treatments, and there were more oyster drills present in roof, small, and large treatments in spring 2011 than fall 2010 (Figure 3c,d; Appendix S1: Table S1 and Figure S4). Oyster length There was no measured effect of treatment on oyster length (Table 1). However, there was an effect of the interaction between sanctuary and sample date with oysters generally increasing in size through time, except for the final sampling when oyster length decreased at Crab Hole and Gibbs Shoal (Appendix S1: Figure S5). The decrease in average size likely resulted from summer recruitment at these two sanctuaries, which resulted in an increase of smaller, juvenile oysters on our experimental blocks, but recruitment was almost absent at Clam Shoal where oyster length continuously increased through time. Oyster predators across sanctuaries The density of predators on the oyster reefs differed among sanctuaries for some of the taxa. The densities of all mud crabs (χ 2 =22.2, df =2, p< 0.001) and large mud crabs (χ 2 =10.5, df =2, p=0.005) were lower at Clam Shoal sanctuary than the other two sanctuaries 6of13 GERALDI ET AL. 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License (Figure 5a,b). Oyster drills were only observed at the Clam Shoal sanctuary, where there was significantly greater abundance and density than the other two sanctuaries (abundance: χ 2 =9.48, df =2, p=0.009; Figure 5c; density: χ 2 =9.48, df =2, p=0.009; Figure 5i). Of the 384 quadrats sampled during the entire study for large crabs, only 1 quadrat had a blue crab (at Gibbs Shoal sanctuary) and 5 quadrats had stone crabs (4 of 41 quadrats at Clam Shoal). There was no statistical difference among sanctuaries for the density of blue crabs (χ 2 =22.2, df =2, p=0.223; Figure 5d)orstone crabs (χ 2 =1.92, df =2, p=0.383; Figure 5e). The density of sheepshead, as measured by visual transects, did not vary among sanctuaries (χ 2 =1.92, df =2, p=0.383; Figure 5f). The biomass of predators across sanctuaries did not vary for any predator: all mud crabs (χ 2 =3.83, df =2, p=0.148), large mud crabs (χ 2 =2.82, df =2, p=0.244), blue crabs (χ 2 =3, df =2, p=0.221), stone crabs (χ 2 =1.92, df =2, p=0.383), or sheepshead (χ 2 =3.00, df =2, p=0.224; Figure 4g–l). Even though there were large differences in means, high variance obscured significant differences, particularly for stone crabs and sheepshead. TABLE 1 Summary of statistical results of five separate generalized linear mixed models Dependent variable Independent variable c 2 df p All oysters Sanctuary 0.9 2 0.635 nbinom1, zi (log) Treatment 25.9 4 <0.001 Sample date 353.4 3 <0.001 Sanctuary:Treatment 41.2 8 <0.001 Sanctuary:Sample date 834.7 6 <0.001 Treatment:Sample date 13.6 12 0.330 Sanctuary:Treatment:Sample date 50.1 24 0.001 Juvenile oysters Sanctuary 0.1 2 0.959 nbinom, zi (log) Treatment 18.2 4 0.001 Sample date 1037.1 3 <0.001 Sanctuary:Treatment 13.9 8 0.085 Sanctuary:Sample date 467.3 6 <0.001 Treatment:Sample date 31.2 12 0.002 Sanctuary:Treatment:Sample date 37.1 24 0.043 Oyster length Sanctuary 1.5 2 0.468 gaussian (log) Treatment 8.1 4 0.088 Sample date 1115.1 2 <0.001 Sanctuary:Treatment 14.5 8 0.070 Sanctuary:Sample date 185.6 4 <0.001 Treatment:Sample date 4.9 8 0.772 Sanctuary:Treatment:Sample date 23.8 16 0.094 Mud crabs Sanctuary 19.6 2 <0.001 nbinom1, zi (log) Treatment 158.4 4 <0.001 Sample date 29.0 1 <0.001 Sanctuary:Treatment 5.3 8 0.729 Sanctuary:Sample date 5.8 2 0.056 Treatment:Sample date 6.6 4 0.160 Sanctuary:Treatment:Sample date 8.6 8 0.381 Oyster drills Treatment 17.2 4 0.002 nbinom1, zi (log) Sample date 29.8 1 <0.001 Note: Each model included a random variable of oyster mound nested within each sanctuary. The dependent variable family and link used in the model are listed below the dependent variable. ECOSPHERE 7of13 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DISCUSSION We found that oyster abundance was dependent on predator access, location, and sample date of the experiment. Oyster abundance was sometimes enhanced when all focal predators were excluded, while allowing access by larger predators (e.g., fishes and large crabs) did not further reduce oyster abundance compared to when only small predators had access (e.g., mud crabs and oyster drills). This effect of predator exclusion was statistically significant at one of three sanctuaries at the first and final sampling event, while a similar but statistically insignificant pattern was seen at a second sanctuary. The sanctuary closest to an ocean inlet had the greatest oyster growth but a boom-and-bust pattern in oyster abundance. Our results agree with past findings from intertidal systems that mud crabs may be the predator with the greatest effect on oyster abundance (e.g., Grabowski et al., 2008,2020; Hill & Weissburg, 2013; Rindone & Eggleston, 2011); however, we found that this effect is spatially and temporally mediated on subtidal oyster reefs. Oyster mean abundance decreased by nearly half during the third sampling event (May 2011, 10 months after deployment) across all predator exclusion treatments, suggesting unmeasured factors depressed abundances at all sites. The number of oysters was similar at the end of the second summer compared to the end of the first FIGURE 3 The mean (SE) total number of oysters (left column) and juvenile oysters (right column) per adult shell for each predator treatment before deployment (a and f) and at the four sampling dates over 14 months, August 2010 (b and g), October 2010 (c and h), May 2011 (d and i), and September 2011 (e and j). The three different sanctuaries are indicated by bar shade. 8of13 GERALDI ET AL. 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License summer in all treatments except for the oysters in the small mesh treatment that excluded all measured predators. This exclusion treatment at the first and final sampling had almost twice the number of oysters (juvenile and total) than the other treatments at two of the three sites. Although this treatment difference was statistically significant at only one sanctuary, we suggest that the density of oysters was mediated by small predators, but this effect was time and location specific. Mud crabs were more abundant inside the cage treatment with small and large mesh that excluded larger predators, suggesting that cages could have been used by mud crabs as shelter from large predators (Geraldi, 2015; Geraldi & Macreadie, 2013; Grabowski, 2004; Johnson et al., 2014). This seemingly contradictory finding that oysters were enhanced in treatments that also had small predators may result from three factors. First, the majority of the mud crabs in these cages were too small to readily consume oysters (<10 mm carapace width) and instead possibly facilitated oyster recruitment by consuming sessile species that compete with settling oysters for suitable substrate (McDonald, 1982). Second, the abundance of mud crabs we measured was likely the maximum number and size present in cages because we FIGURE 4 The mean number of mud crabs within a treatment (SE) during the fall (October) 2010 (a) and spring (May) 2011 (b) sampling and oyster drills within a treatment during the fall 2010 (c) and spring 2011 (d) at the three sanctuaries (denoted by bar shades) ECOSPHERE 9of13 21508925, 2022, 6, Downloaded from https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecs2.4087 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [09/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License