Habitat‐setting affects biodiversity while predation determines oyster survival on experimental oyster reefs

Foundation species are being restored into inherently variable landscapes with multiple, interspersed habitats. However, understanding of the influence of different neighbouring habitats on community assembly and the survival of restored species is limited, despite their significant potential to affect restoration outcomes. We tested how habitat‐setting (being next to seagrass, seagrass and mangroves, or unvegetated sediments) and predation (by meso‐ and/or large predators) influenced macroinvertebrate community assembly and the survival of juvenile Sydney rock oysters (Saccostrea glomerata) on experimental oyster reef units in the Port Hacking estuary, New South Wales, Australia. Each habitat‐setting produced a distinct macroinvertebrate community on experimental reefs, whereas predation had limited effects on community structure. Juvenile oysters were instead highly predated everywhere, and oyster predation was dominated by the large, transient fish Acanthropagrus australis. Our findings allow practitioners to predict and tailor the communities which establish on restored oyster reefs by strategically placing them next to different habitats. If sites have a high predation risk but require seeding for reefs to establish, then caging or complex substrates must be used to increase seeded oyster survival.


Introduction
Foundation species are critical to ecosystem health because they support many ecosystem services and functions, including habitat provision, nutrient cycling, and shoreline protection (Arkema et al. 2013;Ellison 2019).They support diverse communities by reducing abiotic (e.g.temperature; McAfee & Bishop 2019; Uyà et al. 2020) and biotic stress (e.g.predation; Stachowicz 2001).However, numerous foundation species and their associated communities have experienced severe, widespread decline (Beck et al. 2011;Ellison 2019).In response, restoration efforts have increased to enhance the biodiversity, ecosystem functions, and services these species support (Bayraktarov et al. 2016).Restoration efforts have varied success rates and therefore further investigation is required of the ecological factors driving the establishment of and community assembly within foundation species to increase restoration success (Bayraktarov et al. 2016;Zhang et al. 2018).
Restoration is often planned at local scales without considering how landscape variation could affect the biodiversity supported by restored species (Gilby et al. 2018;Saunders et al. 2020).Foundation species often exist in heterogeneous landscapes which could determine the identity and relative abundance of species (source pools), which can reach and colonize restored foundation species.Communities could be supplemented by species immigrating from larger spatial scales, leading to a common set of species settling everywhere (Wilson 1992;Resetarits et al. 2005).Alternatively, different habitats could harbor distinct communities and therefore could influence the source pools available at local scales (e.g.Bloomfield & Gillanders 2005;Wong et al. 2011), which has been found in some terrestrial and coastal ecosystem mosaics (Sheaves 2009;Duflot et al. 2017).In addition to spatial variation, the available source pools can fluctuate over time, such as in response to seasonal immigration and recruitment of species (e.g.oysters; Lee et al. 2012), and therefore the impact of habitat-setting on an establishing community could vary through time.Understanding the impact of habitat configuration on the available source pools and the resulting community structure is particularly valuable in a restoration context because practitioners are relying on the existing source pools at a site to colonize the foundation species they restore.As a result, this understanding could improve the ability of practitioners to predict what communities will establish at a given site and better inform their restoration site selection.
Predation can also vary across landscapes (Esquivel-Muelbert et al. 2022) with implications for the survival of foundation species (Carroll et al. 2015) and the composition of their associated communities (Micheli & Peterson 1999).Across a landscape, transient predators may maintain consistent amounts and rates of predation, unlike predators which are limited to or more abundant in specific habitats.Consequently, proximity to different habitats can affect which predators access a habitat (e.g.Micheli & Peterson 1999;Schneider et al. 2013).Habitats can also vary in predator type (meso-vs.large) and their relative dominance in these and neighboring habitats.For example, high structural complexity in vegetated marine habitats or shellfish reefs can exclude larger predators and so promote higher abundance and/or diversity of mesopredators compared to unvegetated or less complex habitats (Grabowski et al. 2005;Carroll et al. 2015).Local habitatsetting could therefore interact with predation risk to affect communities of restored foundation species which has implications for managing predation risk in restoration efforts.
Neighboring habitats can also affect abiotic conditions with consequences for connected habitats and their communities (Van De Koppel et al. 2015).For example, temperature and light availability can vary in native forests depending on whether they are bordered by forest plantations or grazed pasture (Denyer et al. 2006).Marine foundation species often reduce wind/wave energy and increase sediment deposition (Jones et al. 1997;Stachowicz 2001), which can affect adjacent habitats.For example, vegetated habitats (e.g.mangroves and seagrass) are known to increase sediment deposition rates and reduce sediment resuspension (Granata et al. 2001;Lee et al. 2014;Barcelona et al. 2023) at rates higher than unvegetated sediments (Gacia & Duarte 2001).High sedimentation can limit the growth and survival of foundation species through smothering or obstructing filter-feeding, such as for reef-building molluscs (e.g. oysters, Colden & Lipcius 2015).Sedimentation can also determine where species occur and community assembly among foundation species (Thrush et al. 2003).Therefore, it is important to quantify how neighboring habitats alter key abiotic conditions, such as sedimentation, for connected habitats and whether these changes mediate any habitat-setting observed effects on the survival of the foundation species or their associated community.
Reef-building oysters support diverse communities and ecosystem functions, including nutrient cycling and water filtration in estuaries globally (McAfee & Bishop 2019).Oyster reefs have experienced severe decline driven by overharvesting and disease, with 85% lost worldwide (Beck et al. 2011).There are growing restoration efforts (Gillies et al. 2018), which involve restoring oyster reefs into complex estuarine seascapes, often with multiple habitats (e.g.seagrass, mangroves, and unvegetated sediments).Both predation and abiotic factors can strongly affect oyster recruitment and survival (Esquivel-Muelbert et al. 2022) and unsuitable conditions (e.g.high sedimentation) have led to restoration failures (Bayraktarov et al. 2016).Habitat-setting can also affect macroinvertebrate community structure, for example in Back Sound, North Carolina, United States, where restored reefs bordering mudflats, rather than seagrass or salt marsh, recorded higher densities of some taxonomic groups (e.g.bivalves and decapods; Grabowski et al. 2005Grabowski et al. , 2022)).Habitat-setting also affected macroinvertebrate predation because connected, vegetated habitats allowed blue crab (Callinectes sapidus) predator access (Micheli & Peterson 1999).In Australian estuaries, proximity to seagrass or mangroves can reduce predation rates on oyster reefs (Duncan et al. 2019), but the implications for macroinvertebrate communities and the role of meso-versus large predators are unknown.
In this study, we tested whether habitat-setting and predation interacted to influence juvenile oyster survival and macroinvertebrate community structure on experimental oyster reef units and whether this varied with deployment duration.We placed reef units on unvegetated sediment adjacent to three common habitats in Australian estuaries: unvegetated sediment only, seagrass, and both mangroves and seagrass.We predicted differences among habitat-settings in (1) juvenile oyster survival, (2) macroinvertebrate community structure, (3) amount and type (large vs. meso-) of predation on juvenile oysters and macroinvertebrate communities, and (4) higher sediment deposition rates in vegetated habitat-settings.We also expected that (5) the macroinvertebrate communities would vary among deployment durations.

Study Site
This study was conducted along approximately 1 km of the Port Hacking estuary at Grays Point (À34 03 0 S, 151 04 0 E) in Sydney, New South Wales, Australia (Fig. S1).Grays Point has multiple interspersed habitat types, including the seagrass Zostera muelleri in the shallow subtidal zone, the mangrove Avicennia marina in the upper intertidal zone, and unvegetated Restoration Ecology May 2024 sediments (Fig. S1).There are no remnant oyster reefs located at this site.Mangroves were always bordered by seagrasses on the seaward side, as is common in Australian estuaries, so were not tested independently but they were included to test whether mangroves alter the effects of seagrass.

Effects of Predation Risk and Habitat-Setting on Macroinvertebrate Communities and Oyster Survival
We tested habitat-setting, predation risk, and deployment duration effects on macroinvertebrate community structure and juvenile Saccostrea glomerata survival using experimental reef units (Fig. S2).Reef units (20 Â 20 Â 18 cm) were constructed from plastic mesh cages filled 15 cm high with oyster shells (predominantly S. glomerata; valves disarticulated, half shells >50% intact) (Fig. S4) because they are commonly used for restoration (Bayraktarov et al. 2016).Oyster shells (sourced from Hornsby Shire Council) were dried on land for more than 5 years.Units were attached to cement paver bases (29 Â 29 Â 5 cm) to prevent displacement and subsidence.
We manipulated predation risk with three caging treatments: open, partial exclusion, and full exclusion (Fig. S2).The open treatment was constructed with a large sized mesh (2.5 Â 2.5 cm aperture) with no lid to allow all predators (e.g.fish, crabs, and octopi) access.The partial exclusion treatment was the same, but with a mesh lid added to exclude large predators but provide mesopredator access.The full exclusion treatment was constructed with a small-sized mesh (1 Â 1 cm aperture; Connell & Anderson 1999) to exclude the majority of predators.Caging controls were not included because mesh cages were used in all treatments.We monitored and gently scrubbed cages to clear any fouling every 2 weeks to minimize any associated artifacts and excluded excessively fouled cages (described below).If caging effects on flow and exposure were substantially different among treatments, we should have observed differences in oyster recruitment among predation treatments, with the lowest recruitment in the full exclusion treatment because oysters have planktonic larvae which will reach the substrate in higher abundances with increased flow.However, we did not observe differences in oyster recruitment among predation treatments and therefore suggest that caging artifacts among treatments were limited.Coronavirus disease restrictions on field work prevented cleaning in the last month.Due to damaged cages, significant fouling, and/or human disturbance, three seagrass blocks were excluded from the analysis, and therefore there were n = 7 replicate blocks for mangrove/ seagrass and n = 7 replicate blocks for unvegetated settings, and n = 4 replicate blocks for seagrass settings.
Units were deployed in a block design with seven blocks per habitat-setting.Block locations were assigned haphazardly within each habitat-setting, and habitat-settings were interspersed across the study site.Each block contained two replicates of each caging treatment.All blocks were deployed during low tide in the low intertidal zone (approximately 0.2-0.5 m above the lowest astronomical tide) with at least 20 m between blocks.We chose this spacing among blocks because it reflected the scale of variation among patches of different habitat types at this site.Each block was placed on unvegetated sediments within 1 m of the habitat being tested and at least 20 m from other habitat-settings.For the mangrove/seagrass habitat-setting, blocks were placed 1 m landward of seagrasses and less than 3 m from mangroves.Within blocks, caging treatments were randomly placed in a line parallel to the shoreline with approximately 2 m between units.Units were deployed during austral autumn between the 25th and 26th March, 2021.
To determine treatment effects on oyster predation, eight juvenile S. glomerata (2-4 cm in shell length; sourced from the Port Stephens Fisheries Institute breeding program) were placed on top of the oyster shell in each unit during unit deployment.Juvenile oysters were used as they are vulnerable to predation (Gribben et al. 2020;Esquivel-Muelbert et al. 2022) and are used to seed restored reefs (Geraldi et al. 2013).
After 2 months, one replicate of each caging treatment from each block was destructively sampled and the remaining units were sampled after approximately 5 months.We chose these durations as other NSW studies have demonstrated significant community establishment on oyster shells in approximately 2 months (e.g.Bishop et al. 2012;Hughes et al. 2014) and because 5 months was the longest possible duration within project limitations.
Once collected, the contents of individual cages were washed on a 200 μm sieve in order to capture as many macroinvertebrates as possible (invertebrates visible to the naked eye), preserved in 7% formalin and then stored in 70% ethanol for identification.In the lab, we counted the number of live and remaining dead juvenile oysters.All mobile invertebrates retained were identified and counted under a dissecting microscope.Species that could not be identified to species level were described as morphospecies at the finest taxonomic level possible.Oyster shells were randomly subsampled (approximately 50% of the sample biovolume) to count attached sessile macroinvertebrates (>1 mm) due to their high abundance, and then counted and identified in the same way.The total sample and sub-sample displacement volume (biovolume) of oyster shells in each unit was measured because biovolume can influence invertebrate abundance (Callaway 2018) and provides an estimate of the available surface area, interstitial space, shape, and size of oyster shells in each unit.
Predation on oysters appeared to occur quickly, so we conducted a second experiment to investigate the predation rate over shorter time scales.We deployed eight juvenile oysters (as described above) in open cages only, as most predation was recorded in open cages compared with exclusion cages.An open cage was placed in each block, across each of the three habitat-settings.We counted juvenile oysters in situ to calculate survival on each of the 3 days following oyster deployment.

Differences in Predator Assemblages With Habitat-Setting
To test if differences in predator communities among habitatsettings explained differences in macroinvertebrate communities, we quantified predators in each habitat-setting using unbaited remote underwater video cameras (Lanham et al. 2020;Martínez-Baena et al. 2022).During the first experiment, the reef units were filmed twice over 2 days in May-June 2021.One camera (GoPro Hero 4; GoPro Inc., San Mateo, CA, U.S.A.) was deployed shoreward of each block facing toward the units at approximately 0.5 m away.All cameras were filmed for 1 hour before the peak high tide to ensure all units were fully submerged to allow predator access.
Seventeen minutes of footage (between the 18-and 34-minute marks) was analyzed from the middle of each video to eliminate potential deployment or boat disturbance effects on predator foraging on the reef units (following Gribben et al. 2017).Videos were analyzed using EventMeasure software (SeaGIS Pty Ltd., Bacchus March, Victoria, Australia) to determine the total abundance and species richness of all predators observed.Only fish were observed in videos, so potential fish predators of macroinvertebrates (invertivorous) were identified from FishBase (https://www.fishbase.se/),Fishes of Australia (https://fishesofaustralia.net.au/)and/or the literature.Non-invertivorous fish species were removed from the analyses.MaxN, the maximum number of fish of a species observed within a single video frame per video, was used as a conservative measure of relative fish abundance and summed across species to calculate total abundance.
During the short-term predation experiment, we deployed video cameras on each unit to identify which species were predators of juvenile oysters.Video cameras were deployed in the same way following oyster deployment, but with cameras directly facing units at approximately 0.2 m away.Thirty minutes of footage were analyzed from the middle of each video (excluding the first and last 15 minutes), and MaxN along with the number of bites and entries into the cage were recorded by species.

Differences in Abiotic Conditions Among Habitat-Settings
We compared temperature, dissolved oxygen (DO), salinity, and sediment deposition rates in oyster reef habitats among the three habitat-settings.Temperature ( C), salinity (ppt), and DO (%) were measured at high tide using a YSI water quality meter at each block on 12th June, 2021.Sediment deposition rates were sampled twice (May and June 2021) using sediment traps constructed from polyvinyl chloride pipes (10 cm diameter; 2 cm deep) and deployed for 2 weeks.In each block, traps were attached side-by-side to stakes and deployed in the center of each block with the trap base resting on the sediment surface.
Traps were collected at low tide and stored at À18 C. Sediment was dried at 60 C for 48 hours and weighed to the nearest 0.01 g to obtain the total dry mass.Mean sediment mass was calculated for replicate samples, and then divided by the sampling time to calculate the mean sediment deposition rate (g/day).

Statistical Analysis
Effects of predation risk, habitat-setting, and deployment duration on oyster survivorship and univariate community data (total abundance-with/without oysters, species richness, oyster recruit abundance) were analyzed with linear mixed models fitted with maximum likelihood (R package lme4; Bates et al. 2015).Habitat-setting, predation treatment, and deployment duration were fixed factors, and block was a random factor nested in habitat-setting.Log-transformed biovolume was included as an offset term in macroinvertebrate analyses.Statistical inference was conducted using likelihood ratio tests to test the main and interaction effects.Tukey's post hoc tests were conducted with R package emmeans (Lenth et al. 2021).All statistical analyses were conducted using R 4.1.1(R Core Team 2021), unless specified.
Macroinvertebrate community composition was analyzed in Primer V6 (Clarke & Gorley 2006) and visualized using canonical analysis of principal coordinates (CAP) plots fitted with Bray-Curtis dissimilarity.Species vectors were added to CAP plots (Pearson correlation coefficients >0.5).Community composition was analyzed using permutational analysis of variance (PERMANOVA; Anderson et al. 2008) with Bray-Curtis dissimilarity.PERMANOVA was conducted with the same factors as above.We used square-root and presence-absence transformations, to distinguish variation in relative abundance while downweighing highly abundant species and/or occurrence of species, respectively.Post hoc tests were conducted with PERMANOVA pairwise tests where there were significant interactions between factors.
Total abundance (MaxN) and species richness of invertivorous fish were analyzed using generalized linear models with a Poisson distribution and habitat-setting and sampling time as fixed factors.Community composition was analyzed using multivariate generalized linear models with a negative binomial distribution and the same factors as above (R package mvabund; Wang et al. 2021).Statistical inference was conducted using probability integral transform (PIT-trap) bootstrapping.MaxN, number of entries, and number of bites from the short-term experiment were analyzed using analyses of variance, with habitat-setting as a fixed factor.Community composition was analyzed with multivariate generalized linear models (R package mvabund; Wang et al. 2021) fitted with a negative binomial distribution.Multivariate data for both video sets were visualized in Primer V 6 as for macroinvertebrates.
Temperature, salinity, and DO were contrasted among habitat-settings using a one-factor analysis of variance (ANOVA).Sediment deposition rates were analyzed using a two-factor ANOVA with habitat-setting and sampling time as fixed factors, and Tukey's post hoc tests.

Effects of Predation Risk and Habitat-Setting on Macroinvertebrate Communities and Oyster Survival
Overall, 38 macroinvertebrate species were identified on reef units, with seven unique species in seagrass, three in mangrove/seagrass, and five in unvegetated settings (Table S1).Total macroinvertebrate abundance varied among habitatsettings (df = 12, χ 2 = 23.421,p = 0.024), but not among predation treatments or deployment durations (Table S2; Fig. 1A & 1B).Total abundance was higher in unvegetated compared with Restoration Ecology May 2024 both seagrass and mangrove/seagrass habitat-settings, which did not differ from each other (Table S3; Fig. 1A & 1B).When the dominant species, oyster recruits (82.4% total individuals), was excluded from abundance, there was a significant interaction between habitat-setting and deployment duration (df = 6, χ 2 = 21.966,p = 0.001; Fig. 1C & 1D).Macroinvertebrate abundance (excluding oysters) was higher in unvegetated compared with mangrove/seagrass and seagrass settings for the 5-month but not 2-month duration (Table S3; Fig. 1C & 1D).Total oyster recruit abundance did not vary among habitatsettings, predation treatments, or deployment durations (Table S2; Fig. 1E & 1F).Species richness was higher in partial Figure 1.Macroinvertebrate total abundance (A, B), macroinvertebrate total abundance (excluding oysters) (C, D), abundance of oyster recruits (E, F), macroinvertebrate species richness (G, H), and juvenile oyster counts (I, J) were compared among habitat-settings, predation treatments, and deployment durations.Boxes represent the interquartile range, the line in the center of the box represents the median, whiskers represent the minimum and maximum value of that data (that is, within 1.5Â the interquartile range), and circles represent values greater or less than 1.5Â the interquartile range.
The occurrences of the bivalve (X.securis) and the decapods (P.laevis and Enigmaplax littoralis) were more strongly correlated with unvegetated and mangrove/seagrass settings (Fig. 2B).The occurrences of P. micans and the decapod (Pilumnopeus serratifrons) were instead more strongly correlated with seagrass settings (Fig. 2B).The occurrences of Amphipod 1 (Melitidae) and E. littoralis were more strongly correlated with the 5-month duration compared to the 2-month duration (Fig. 2B).

Differences in Predator Assemblages Among Habitat-Settings
Overall, 14 invertivorous fish species were observed (Table S6).Acanthropagrus australis was the only consumer of oysters, although other species (Silago cilliata and Myxus elongatus) can prey on bivalves.Fish total abundance (MaxN) did not vary among habitat-settings or sampling times (Table S7; Fig. 3).Species richness was higher in May compared to June but did not vary among habitat-settings (Table S7; Fig. 3).There was a significant interaction between habitat-setting and sampling time for community composition, with differences between Restoration Ecology May 2024 unvegetated and mangrove/seagrass settings in June only and between May and June in unvegetated settings only (Table S7; Fig. S3).There were no differences among habitat-settings for individual species (Table S8).For the short-term predation experiment, A. australis was the only predator species observed entering or feeding in the cages and 21 bites were recorded across 14 cage entries.

Differences in Abiotic Conditions Among Habitat-Settings
Temperature, DO, and salinity did not vary among habitatsettings (Table S9; Fig. S4a-c).Sediment deposition rates differed among habitat-settings but there was no difference between deployment durations or interaction (Table S9; Fig. S4d).Sediment deposition was lower in mangrove/seagrass compared with unvegetated ( p = 0.017) and seagrass settings ( p = 0.014), which were not different ( p = 0.996).

Discussion
Variation in different habitat-settings across a landscape, and related differences in biotic and abiotic processes, can strongly affect connected foundation species and their associated communities (Bayraktarov et al. 2016;Gilby et al. 2020;Lester et al. 2020;Saunders et al. 2020).We found evidence that habitat-setting does affect oyster reef macroinvertebrate communities, likely by affecting the available source pools and habitat selection processes of colonizing species.Juvenile oyster survival was instead consistent across the site suggesting that large, transient predators can strongly determine oyster survival at this scale.Our study suggests local habitat-setting is an important consideration for restoring oyster reefs as well as biodiversity and predation on reefs.
Differences in macroinvertebrate composition and total abundance (with and without oysters) among habitat-settings were not mediated by variation in predation risk, because predator exclusion did not strongly affect macroinvertebrate communities and predator communities did not vary among habitat-settings.Instead, neighboring habitats likely mediated the available source pools of certain macroinvertebrate species.Estuarine habitats support different communities and levels of biodiversity (Bloomfield & Gillanders 2005;Hosack et al. 2006;Shervette & Gelwick 2008;Wong et al. 2011) and so could influence the identity and abundance of species migrating from connected habitats to reef units.Although we did not sample neighboring habitat communities directly, patterns reflect known habitat distributions for some species and other studies demonstrate migration between physically connected habitats (Micheli & Peterson 1999;Hanke et al. 2017).For example, the crab Paragrapsus laevis had higher occurrences in reef units placed in mangrove/seagrass settings and likely was colonizing from mangroves where it is common in temperate Australia (Mazumder & Saintilan 2010).The gastropod Pseudoliotia micans was more abundant in seagrass settings and has recorded higher abundances when seagrass is present (Beard 2012).Placement next to different habitats could therefore influence restored communities by mediating the available source pools of colonizing species.
Neighboring habitats could also have affected habitat selection of certain colonizing species, leading to differences in community structure.Higher total abundance in unvegetated settings may result from habitat preference for isolated patches, as has been observed for mobile invertebrates in other shallow subtidal estuarine systems (Roberts & Poore 2006;Lanham et al. 2020).In addition, the mussel Xenostrobus securis had higher Figure 3. Box plots for invertivorous fish total abundance (MaxN) and species richness compared among habitat-settings and sampling times.Boxes represent the interquartile range, the line in the center of the box represents the median, whiskers represent the minimum and maximum value of that data (that is, within 1.5Â the interquartile range) and circles represent values greater or less than 1.5Â the interquartile range.
Habitat-setting affects oyster reef biodiversity occurrences on reefs adjacent to mangrove/seagrass, despite having planktonic larvae (Wilson 1969).X. securis settles on hard substrata (e.g.mussels and mangrove pneumatophores; Wilson 1969), and so this pattern may reflect a preference for settlement near mangrove habitats and thereby reef units in mangrove/seagrass settings.However, oyster larvae are also long-distance dispersers but recruitment did not vary among habitat-settings likely because oyster shell was used in all units.Together, our findings indicate that numerous species vary among habitat-settings, contributing to community differences, and that habitat configuration may influence immigration and habitat selection depending on the species.These findings are significant given the small scale of our study and indicate that even patch-scale variation in habitat types can affect community assembly.We recommend continued study of habitat-setting effects at a range of spatial scales within and across different estuary systems to determine whether similar patterns hold at multiple spatial scales and under different environmental conditions.For restoration, this information improves our ability to predict which species and communities will establish at different sites, depending on the habitat-setting, and so we can select certain sites depending on the community characteristics that we want to restore.
Across all habitat-settings, macroinvertebrate species richness was reduced only when large predators had access (i.e. in open cages) to invertebrate communities with 18 species absent in open cages in at least one habitat-setting.Reflecting differences in community composition, and possibly prey availability, among habitat-settings, large predators excluded different species in different habitats.However, predation did not affect total abundance or community composition, which we suggest results from the low abundance of almost all prey species that were excluded.In contrast, studies on intertidal U.S. oyster reefs have shown that predation can influence macroinvertebrate abundance (Micheli & Peterson 1999;Knights et al. 2012;Carroll et al. 2015).High predation risk has been recorded on Australian restored reefs using standardized prey items ("squidpops"; Duncan et al. 2019), but the effects on macroinvertebrate communities have not been determined.Predation effects could therefore vary depending on the predator communities at a site or could be mediated by other factors like patch-size, as observed in other systems (e.g.Atlantic forest; Mendes et al. 2016), and could explain differences between our study and those on larger reefs (Micheli & Peterson 1999;Duncan et al. 2019), which is an important avenue to consider in future research.However, we did find that predation consistently reduced species richness and is significant for certain macroinvertebrate species that colonized reefs.
Where oyster recruitment limits establishment, restored reefs are often seeded with broodstock and/or spat (Geraldi et al. 2013) however we found low survival of our 'seeded' juvenile oysters regardless of habitat-setting, due to large predators.Acanthropagrus australis likely dominated predation because they were the only species observed known to consume juvenile Saccostrea glomerata (Esquivel-Muelbert et al. 2022;Martínez-Baena et al. 2022) and directly predating oysters and dominate predation on other Australian restored reefs (Duncan et al. 2019).Predation risk was likely consistent across the site because fish communities, including A. australis abundance, did not vary among habitat-settings, unlike in other studies (Ziegler et al. 2017;Duncan et al. 2019) and A. australis often forages widely across seascapes (Pollock 1982).Factors acting on larger spatial scales (e.g.vegetated habitat configuration) may affect predator distribution more strongly (Micheli & Peterson 1999;Carroll et al. 2015;Duncan et al. 2019;Martínez-Baena et al. 2022).
Unlike studies on other oyster reefs (e.g.NC, U.S.A.; Grabowski & Kimbro 2005;Carroll et al. 2015), mesopredators had weak effects on oyster survival.Potential mesopredators, primarily juvenile A. australis, were observed but may have prioritized other food sources.In contrast, predation did not affect oyster recruit abundance, indicating that complex shell matrices can offer predation refuge and could benefit juvenile oysters (McAfee & Connell 2020).Predation nevertheless can strongly determine oyster survival (Carroll et al. 2015;Esquivel-Muelbert et al. 2022), and warrants consideration in restoration site selection and approaches, including seeding.
There were minor differences in sediment deposition among habitat-settings, which was expected because different habitats can both establish and promote different sedimentation regimes.However, this variation does not relate to the differences we observed in oyster survival or macroinvertebrate communities across habitat-settings.The lack of juvenile oyster mortality in full exclusion treatments across any habitat-setting (where there was no mortality due to predation) also suggests that even the elevated sediment deposition rates in unvegetated and seagrass settings were not high enough to impact oyster survival.Habitat-setting differences in sediment deposition may be more significant in estuaries with higher sediment loads or lower flow velocities where sediment deposition is substantial enough to limit recruitment and survival of oysters (Grabowski et al. 2005;Colden & Lipcius 2015) or macroinvertebrates (Thrush et al. 2003;Callaway 2018).
Habitat-setting and predation effects on juvenile oyster mortality and macroinvertebrate communities remained consistent over the experiment.As expected, reef units deployed for 5 months had the highest macroinvertebrate species richness and distinct community compositions.However, despite additional colonization by species and community shifts between the 2-and 5-month durations, the habitat-setting had a prevailing effect on the establishing communities.The communities may develop further with seasonal recruitment of certain species or predator visitation (e.g.oysters with peak recruitment in January-March; Robertson & Duke 1990;Hindell & Jenkins 2004;Lee et al. 2012), which could further shift the proportion of individuals immigrating from nearby connected habitats compared to the wider estuary.Habitatsetting nevertheless had a consistent effect on early-stage communities and may persist long-term as found in other studies over 13-years post-restoration (Ziegler et al. 2017).We also observed rapid mortality of juvenile oysters.Initial decision-making both regarding restoration site selection and design (e.g.predator exclusion) can therefore have longterm effects.
Our study reinforces the importance of considering habitat-setting in initial decision-making (Bayraktarov et al. 2016;Gilby et al. 2020;Saunders et al. 2020), including site selection, in restoration efforts because it significantly affects the communities which establish.Using findings like ours, practitioners can begin to predict which communities will establish based on where a restoration site is in relation to different habitats, and therefore predict restoration outcomes for biodiversity at a certain site.This could provide a more effective, targeted approach compared to the predominant "built it and they will come" approach.The influence of different habitats on oyster reef communities also provides opportunities for co-restoration with other foundation species (seagrass and mangroves; e.g.McAfee et al. 2022) and landscape-scale management approaches to further tailor the communities that establish.
To optimize oyster survival, and therefore reef persistence, it is also critical to consider predation risk in restoration efforts, particularly where similar predator communities to those at this site are present.Practitioners can quantify predation risk by conducting pilot predator surveys, and considering known factors that influence predation risk (e.g.depth; Esquivel-Muelbert et al. 2022).Oyster survival can be enhanced by mitigating predation, including through caging, placing oysters into complex substrates (McAfee & Connell 2020) and/or using less vulnerable, larger size classes where seeding is required (Esquivel-Muelbert et al. 2022) to enhance reef persistence and restoration of associated ecosystem services and functions.

Figure 2 .
Figure 2. Canonical analysis of principal coordinates (CAP) plots of (A) square-root and (B) presence-absence transformed macroinvertebrate community composition compared among habitat-settings and deployment durations.CAP plots were created using Bray-Curtis dissimilarity and display species vectors for species with a Pearson correlation coefficient greater than 0.5.Vectors indicate direction in which abundance and/or occurrence of each species is most strongly correlated with the ordination configuration.Circle indicates a correlation of 1 of that species to the 2-d CAP x and y coordinates.