Nuanced influences of subtidal artificial shellfish structures on nekton communities in urbanised estuaries

Study system and restoration

This study took place in Port Adelaide, the urbanised upper section of the Port River–Barker Inlet Estuary, situated on the eastern side of the Gulf St Vincent, South Australia (34°50′43.8″S, 138°29′51.9″E). Owing to channel modification, the system forms an inverse estuary, with seawater inflows from the north and south via Outer Harbour and the West Lakes intake pipe respectively, and intermittent freshwater inputs from the Dry Creek wetland and stormwater drains. The ~0.28-ha oyster reef restoration site was established near Jervois Bridge, abutting a quay wall and a rock ‘rip-rap’ revetment (Fig. 1). This area once sustained flat oyster (O. angasi) reefs (historically termed ‘beds’) prior to the 1920s, on the basis of historical Indigenous and European Australian records (reviewed in Martin et al. 2025a). The site also hosted the Port Adelaide Fish Market (pre-1960s) with ~40 relict remnant wooden supporting beams present across the site. A bare control site near Birkenhead Bridge was established 400 m downstream, where habitat restoration has not historically occurred. This control site abuts a quay wall and is dominated by a bare mud benthos. Both sites are situated along the main Port River channel (5–6-m depth) and located within the Port Adelaide urban centre, and are therefore likely to experience similar environmental conditions. The quay walls at both sites support pre-existing sessile invertebrate communities, including Pacific oysters (Magallana gigas), black mussels (Mytilus spp.), Mediterranean fan worms (Sabella spallanzanii) and sea squirts, whereas the bare benthos is mainly characterised by turfing algae and terebellid bristle worms. As estuarine O. angasi reefs are functionally extinct locally (Alleway and Connell 2015), we were unable to include an established oyster reef as an additional reference control.

Fig. 1.

Map of Inner Harbour in the Port River–Barker Inlet Estuary, South Australia, depicting the oyster reef restoration site (red border) and bare control site surveyed from 2022 to 2024, with red circles representing remote underwater video station (RUVS) deployment locations. Above-water photographs of the restoration and control habitats are shown, along with a representative image of the RUVS field of view.

Subtidal oyster reef restoration in the Port River–Barker Inlet Estuary has occurred only at Jervois bridge, commencing in 2018, with efforts expanded in 2022, led by not-for profit organisations OzFish Unlimited and the Estuary Care Foundation. Some 80 modular gabion mesh baskets (40 × 40 × 30 cm) containing sterilised Pacific oyster shell were deployed by volunteers at the single restoration site, along with adult O. angasi broodstock (~50 individuals) during a single event in December 2022. This coincided with known O. angasi spawning periods when seawater temperatures were above 17°C (McAfee and Connell 2020). The units form a low-relief structural substrate for settlement by O. angasi larvae and other species. By 6 months post-restoration, each unit supported ~1000 O. angasi spat (n = 3 units; I. Kenny, unpubl. data, 2024). Successive settlement onto the units is expected to facilitate the re-establishment of flat oyster populations, and their associated functions.

Sampling approach

Nekton surveys began in November 2022 and occurred approximately every 3 weeks until February 2024. This included three survey periods prior to restoration, in November and December 2022, which provided a ‘before installation’ baseline. Post-restoration surveys commenced 3 weeks after installation of the modular oyster units between January 2023 and February 2024. Nekton were assessed from 20 survey timepoints across 14 months to evaluate temporal variability in the abiotic and biotic conditions associated with restoration.

The nekton communities at each site were surveyed using unbaited remote video stations (RUVS) during daylight hours (07:30 to 17:30 hours). Each RUVS consisted of a GoPro Hero7 camera attached to a weighted frame and positioned 15 cm above the benthos when deployed. The cameras were unbaited to minimise bait biases in interpreting the fish-habitat associations (Martinez-Baena et al. 2022; Martin et al. 2025b). RUVS were deployed for 35 min, because short-durations are suitable to rapidly quantify fish habitat use in complex and turbid estuary settings (Baker et al. 2022). Owing to the single oyster reef restoration site, and small restoration footprint available, three RUVS were deployed >15 m apart at each site along the adjacent wharf, with a replicate deployment occurring 1–5 h later for each timepoint (n = 12). Although this design was suboptimal for capturing broad spatial variability within the estuary, it provided a localised assessment of the monitored sites.

To minimise biases associated with sediment disturbance, only 30 min of footage was assessed, with analysis being delayed until the sediment plume had settled. Videos were processed in EventMeasure (ver. 6.45, see www.seagis.com.au), with all nekton (e.g. fishes and decapods) being identified to the lowest taxonomic level to quantify community composition and taxonomic richness (Gomon et al. 2008). Abundance was calculated as the maximum number of individuals within a single frame (MaxN) (Whitmarsh et al. 2016). Additionally, all species were categorised into water column guilds (e.g. Maus et al. 2024), consisting of (1) benthic species found on the benthos, (2) demersal species associated near the benthos, or (3) benthopelagic species that move and feed between the benthos and water column. We also calculated the number of ‘harvestable’ and ‘invasive’ nekton species on the basis of their socioecological categorisation as fisheries-targeted (i.e. FishBase) or as non-native species (i.e. Wiltshire et al. 2010; Hammer 2014).

All fieldwork was carried out according to permissions obtained from Flinders University Animal Ethics Committee (AEC:5317).

Environmental variables

Field of view composition percentages of each RUVS deployment was calculated using ImageJ (ver. 1.54j, W. S Rasband, US National Institutes of Health, Bethesda, MD, USA, see http://imagej.nih.gov/ij/) from a single screenshot to quantify the cover proportion of visible consolidated benthic structures, unstructured bare mud, wharf wall and the water column. Surface water temperature and salinity measurements were collected in situ, and a Secchi disc was used to estimate water clarity and maximum water depth. The maximum and minimum air temperature, and maximum wind speed (km h⁻¹) and direction were obtained from the Parafield Airport weather observation station for each deployment day (see http://www.bom.gov.au/climate/averages/tables/cw_023013.shtml).

Statistical analyses

The environmental and nekton community data were statistically analysed in RStudio (ver. 4.3.1, Posit Software, PBC, Boston, MA, USA, see https://posit.co/products/open-source/rstudio/). We first tested for potential site (two levels: restoration and control) differences in the measured water variables and field of view percentages using non-parametric Mann–Whitney U-tests. The environmental conditions were assumed to be independent of restoration activities and driven by other temporal variables (e.g. weather, hydrodynamics).

Spatio-temporal variation in nekton community structure was investigated using permutational multivariate analysis of variance (PERMANOVA) from the vegan package (ver. 2.6-4, J. Oksanen et al., see https://cran.r-project.org/package=vegan/). The PERMANOVA was conducted to assess for significant differences between site (fixed factor; two levels: restoration and control sites), and stage (fixed factor; two levels: before and after restoration), with season (four levels; spring, summer, autumn and winter), replicate (two levels), and stage:time (23 survey periods) as covariates in the PERMANOVA design to account for potential temporal autocorrelation in the data. Prior to analysis, potential skewness in the data was assessed through a series of transformations and visualised using shade plots to identify whether transformation to normalise the data was needed, and then checked for overdispersion. This showed that the multivariate data required fourth-root transformation to reduce the effects of overly abundant schooling fishes before calculating Bray–Curtis dissimilarly matrices. A dummy variable of +1 was added to the data because of surveys where nekton were absent (Clarke et al. 2006). Owing to the unbalanced number of surveys by stage, we focused the analyses on its interaction with other factors to mitigate potential biases. Post hoc pairwise PERMANOVAs were used to assess the sources of significant difference within levels of the factors.

For general comparisons, the community data were visualised using two separate non-metric multidimensional scaling (nMDS) plots depicting potential groupings by site, and the interaction of site and stage to support interpretation of possible effects of the main factors. A distance-based redundancy analysis (dbRDA) was used to ordinate the transformed-community data and identify potential relationships to the environmental variables. Using the ‘envit’ function of the vegan package, vectors from the environmental variables were tested for significant correlation to the ordination axes (9999 permutations) and fitted to the dbRDA plot, with the overlays depicting the direction and strength of the relationship.

Dufrene–Legendre indicator species analysis from the labdsv package (ver. 2.1-0, D. W. Roberts, see https://CRAN.R-project.org/package=labdsv; Roberts and Roberts 2016) was used to determine indicator values for each species on the basis of their frequency and abundance, and identify species that differed across sites and restoration stages. We then used univariate linear mixed effects models from the packages glmmTMB (ver. 1.1.9, see https://cran.r-project.org/package=glmmTMB; Booth et al. 2003) and car (ver. 3.1-2, see https://CRAN.R-project.org/package=car; Fox and Weisberg 2018) to test for patterns in the (1) nekton total abundance and richness; (2) abundances of species identified from the indicator analysis, and abundance based on the (3) socioecological categories, and (4) water column position guilds. The model tested for interactions among the fixed factors of site and stage, with season, replicate and time as random factors to account for temporal variation. All models used a negative binomial distribution, and were checked for collinearity and overdispersion. For each model, we used the MuMIn package (ver. 1.48.4, see http://r-forge.r-project.org/projects/mumin/) to calculate the marginal and conditional R²-values to estimate the proportion of variance explained by the fixed factors, and in combination with the random factors to evaluate potential temporal effects.

Environmental conditions

Weather conditions recorded from the Parafield Weather Station were variable across surveys (Fig. 2). Maximum wind conditions were 20–67 km h⁻¹, typically in a westerly direction (Fig. 2a). Rainfall (mm) was relatively uncommon (17% of survey periods) and ranged from 0 to 2.6 mm (Fig. 2b). Air and water temperatures typically followed seasonal patterns, being warmer during the Austral summer (December–February), and cooler during winter (June–August; 20–40 weeks post-installation) (Fig. 2c, d). Air temperature ranged from daily minimums of 0.2–25.1°C to maximums of 14.1–35.9°C (Supplementary Table S1). Maximum depth was typically shallower at the restoration site (mean ± standard error, 4.47 ± 0.07 m) than at the control site (4.89 ± 0.05 m). Water clarity was also generally poorer at the restoration site (3.47 ± 0.07 m) than at the control sites (3.94 ± 0.06 m), regardless of tidal influence (Fig. 2e, f). Water temperature ranged from 9.3 to 32°C and showed seasonal cooling during the Austral winter period (Fig. 2d). Salinity, measured in parts per thousand (ppt) of dissolved salts, ranged from 20 to 40 ppt, with variation potentially caused by freshwater input from stormwater drains after large rainfall events (Fig. 2g), but did not vary seasonally. The Mann–Whitney U-test results found significant differences by site only for depth and clarity (P < 0.0001; Supplementary Table S1).

Fig. 2.

Environmental conditions associated with each remote underwater video station (RUVS) monitoring timepoint between November 2022 and February 2024, from the restoration (Res. site) and control (Con. site) sites in the Port River–Barker Inlet Estuary. This includes the (a) maximum (max.) wind speed and direction, (b) rainfall and (c) air temperature (temp.) recorded from the Parafield Airport; mean values of in situ measurements of (d) water temperature, (e) depth, (f) water clarity (Secchi disc depth), and (g) salinity; and (h) proportional bar graph of RUV field of view composition aggregated across sites. The vertical red dashed line delineates between the before and after restoration stages.

Video monitoring

Of the 276 RUVS deployments, 252 recordings were included in the analysis, consisting of 121 videos from the oyster reef restoration site, and 131 videos from the control site. Twenty-four deployments were excluded owing to recording issues (e.g. malfunctions, unsuccessful placement, tipping over).

The field of view composition was similar across sites on the basis of the Mann–Whitney U-test results for each category (P > 0.05, Table S1). The benthos was typically dominated by bare mud (mean ± standard error: 35.0 ± 1.5%) with some consolidated benthic structures (8.4 ± 0.8%) including pylons, debris (e.g. pipes, fishing gear, discarded shopping trollies) and biogenic structures, such as disarticulated Pacific oyster shells (M. gigas), spaghetti bryozoans (Amathia verticillata) and Mediterranean fan worms (S. spallanzanii). The vertical composition of the field of view consisted of the wharf wall (7.7 ± 1.2%), and the water column (48.9 ± 1.8%; Fig. 2h).

Surveyed nekton assemblages

In total, 4941 nektonic organisms were recorded from the RUVS, consisting of 31 fish species from 17 families, and 2 decapod species. From 6 weeks prior to oyster reef installation in December 2022, the RUVS detected a mean (±s.e.) of 4.6 ± 0.4 nekton species, and 22.5 ± 3.0 total abundances, across 36 deployments. In the 61 weeks following restoration activities to February 2024, the RUVS averaged 3.9 ± 0.1 nekton species, and 19.1 ± 1.16 total abundance, across 216 deployments. Most species (71%) were surveyed at both sites, with only three species exclusively observed from the restoration site, v. nine species uniquely sighted at the control site. However, these ‘site-specific’ species were rare, and were detected only in <3% of videos (Fig. 3, Table S2). Only 14 species (41%) were detected from the surveys across all four seasons (Fig. 3). The most ubiquitous species observed were black bream (Acanthopagrus butcheri, 83%), southern longfin gobies (Favonigobius lateralis, 48%) and western striped grunters (Pelates octolineatus, 42%).

Fig. 3.

Bubble plot representing the combined abundances of nekton species across sites detected during each sample period from remote underwater video station surveys, coloured by categorical values of season (spring, green; summer, orange; autumn, brown; winter, blue).

Of the observed nekton, eight were considered harvestable species, comprising 60.3% of total abundance (i.e. primarily A. butcheri and yelloweye mullet, Aldrichetta forsteri). Non-native Gobiidae species comprised 1.7% of the total abundance, and included crested oystergoby (Cryptocentroides gobioides), trident goby (Tridentiger trigonocephalus), largemouth goby (Redigobius macrostoma) and exquisite sandgoby (Favongobius exquisitious). Of the 34 nekton species, the majority were classified as demersal (38.2%) or benthic (35.3%), comprising 15.7% of total abundance across all sites and surveys. Benthopelagic species (nine species) comprised the majority of observed nekton (84.2% of total abundance).

Nekton community structure

Nekton assemblages differed among sites, and before and after oyster reef restoration, but these factors did not significantly interact (Table 1). The position of the ellipses showed some separation across surveys by both site and stage respectively, but there was considerable overlap in the combination of these factors (Fig. 4). Given that pre-restoration sampling occurred only in spring–summer, the greater variation of the post-restoration nekton community (Fig. 4b) can be attributed to the larger sample size, and seasonal nekton variability from autumn and winter surveys. The nekton assemblages were also temporally variable across survey timepoints on the basis of the significant interaction of restoration stage and time (Table 1). This was also identified with seasonal variation in the community, where peaks in the occurrence and abundances of most species were typically observed in summer and decreased into the winter period (Table 1, Fig. 3).

Fig. 4.

Non-metric multi-dimensional scaling (nMDS) plots of nekton community structure on the basis of Bray–Curtis dissimilarities. It depicts points categorised by (a) sites, comparing the restoration and control sites, and (b) site and restoration stage, depicting samples before and after oyster restoration, with overlaid 95% confidence ellipses. Stress value was 0.165.

Nekton communities were primarily clustered by season, rather than site-specific clustering, showing a distinct seasonal separation between summer–autumn and winter–spring samples (Fig. 5). These patterns were strongly associated with the measured environmental variables (Fig. 2), particularly temperature, weather patterns and water clarity.

Fig. 5.

Distance-based redundancy analysis (dbRDA) plot of transformed nekton community data, with points categories across the two sites, and four seasons. Points represent nekton surveys, and are presented by different colours and shapes to represent site and seasonal patterns. Vectors indicate the direction and strength of environmental variables found to have a significant (P < 0.05) correlation with the data.

The crested oystergoby (C. gobioides) was the sole indicator species that was positively associated with the restoration site, with greater detection in the months following the oyster reef restoration activities. By contrast, western striped grunters (P. octolineatus) were associated with the control site throughout the study period (Table 2). Both yelloweye mullet (A. forsteri) and wavy grubfish (Parapercis haackei) were identified as indicator species for the control site in the post-restoration period.

Nekton species richness and abundances

Variation in the multivariate community structure could be, in part, attributed to site- and seasonal-effects on species presence and abundance. Overall, the linear mixed-effect models indicated elevated total nekton abundances at the control site (F = 27.62, P < 0.001), but similar nekton richness across sites (F = 2.70, P = 0.100; Table 3). The RUVS at the restoration site detected 3.8 ± 0.2 nekton species and 14.5 ± 1.2 total abundance, which was lower than at the control site with 4.2 ± 0.2 species and 24.3 ± 1.7 total abundance (Fig. 6, Supplementary Fig. S1). Similarly, the benthopelagic guild abundances were greater at the control site (F = 27.72, P < 0.001), but both sites supported similar demersal and benthic nekton guild abundances (Fig. S1). The random effects (season, time and replicate) approximately quadrupled the amount of variation explained by the models compared with the fixed effects only (site and stage; Table 3), reflecting the influence of temporal variation in the data (Fig. 3–5). Although surveys in the weeks following restoration showed elevated nekton richness and total abundances at the restoration site, these values generally decreased during the cooler winter period and stayed below abundances at the control site over the proceeding spring–summer period (Fig. 6). Significant interactions of site and stage were identified from both yelloweye mullet (A. forsteri) and western striped grunters (P. octolineatus) (Table 3). Aldricetta forsteri was not detected prior to restoration, but was increasingly common post-restoration, with high abundances at the control site over the winter–spring period (Fig. 6). By contrast, P. octolineatus was detected at the restoration site only post-restoration and was generally detected at higher abundances at the control site throughout the study period (Fig. 6, S1).

Fig. 6.

Time series plot of mean ± s.e. of the nekton (a) richness, (b) total abundance, (c) western striped grunter, and (d) yelloweye mullet observed per RUV survey timepoint at the restoration and control sites between November 2022 and February 2024. The vertical dashed line denotes restoration activities related to the installation of modular oyster units.

Nekton community responses

Oyster reef restoration projects are expected to benefit nekton assemblages by increasing beneficial habitat functions, including foraging opportunities, refugia and reducing environmental stressors (Grabowski et al. 2012; Gilby et al. 2018). These functional enhancements may increase fish survival and recruitment, resulting in net gains to nekton biodiversity and productivity (e.g. Smith et al. 2023; Connolly et al. 2024). For example, restored eastern oyster (Crassostrea virginica) reefs in the United States support 34% greater species richness and 99% greater total community biomass compared with degraded (control) sites (Smith et al. 2023). This is further supported by a global meta-analysis of 160 oyster restoration studies, finding that restoration generally enhanced nekton abundance and diversity, albeit not across all studies (Sievers et al. 2024). In regions without remnant reefs or with understudied oyster species, expected nekton responses may be identified only following restoration trials (Zu Ermgassen et al. 2020; Martin et al. 2025b).

Estuarine shellfish restoration trials elsewhere in Australia (e.g. Gilby et al. 2021; Maus et al. 2024) have detected post-restoration enhancements in species richness and fish abundances 1–2 years post-restoration. For example, Gilby et al. (2021) found that by 30 months post-restoration, a 1.4-ha modular Sydney rock oyster (Saccostrea glomerata) reef enhanced fish richness and abundance by 3.8 and 10.7 times respectively, without depleting nekton from the surrounding seascape. Furthermore, at the ‘unit’ scale, oyster garden trials in Morton Bay using caged shellfish found higher fish and invertebrate diversity than in both control cages and non-caged areas (Boström-Einarsson et al. 2022). Contrastingly, our dataset found little observable ecological responses of the nekton community following oyster reef restoration activities. Instead, temporal and interannual variation in environmental conditions, particularly weather and water parameters, were key predictors for interpreting variation in the nekton assemblages. Trends were consistent with established seasonal changes typical of estuarine nekton assemblages (e.g. Chuwen et al. 2009; Potter et al. 2016; Harrison and Whitfield 2021). During the summer and autumn months, when temperature and salinity are generally elevated, nekton diversity was typically greater owing to migrations and spawning pulses of marine species, and lower in winter and spring (Harrison and Whitfield 2021). Similar seasonal patterns have also been recorded from estuarine fish monitoring associated with remnant oyster reefs from New South Wales (Martínez-García et al. 2022) and from 100-m² mussel restoration sites in the Swan–Canning estuary (Maus et al. 2024). Most shellfish reef fish studies conduct monitoring for ~12 months (Martin et al. 2025b); thus, multi-year or subsequent studies would be required to determine whether some of the observed differences were caused by seasonal or annual variations or whether real ecological changes are occurring from restoration.

Several factors are likely to have contributed to the subdued nekton responses to restoration in our study, including habitat characteristics, environmental stressors and species-specific responses. The modular steel gabion baskets comprise a low-profile reef that cover a small spatial extent (~0.28 ha) across a single location. This small spatial scale of restoration may have limited enhancements in large, mobile nekton (i.e. >20 cm TL; Gilby et al. 2018; Fraschetti et al. 2021), whereas cryptobenthic species (e.g. Gobiidae spp.) that are more sensitive to structural differences (Ushiama et al. 2019) are less detectable by using RUVS (Martin et al. 2025b). Furthermore, our study occurred during the early stages of restoration when nekton are responding to the reef substrate, rather than the biological role of oysters in environmental modification. Successive flat oyster (O. angasi) recruitment combined with upscaled restoration within the estuary may be required before functional modifications (e.g. enhanced biofiltration, bio-deposits, habitat creation) lead to major restoration outcomes on nekton assemblages (Smith et al. 2023). Additional restoration efforts would also increase available shellfish habitat for monitoring, which could lessen the effects of confounding anthropogenic and environmental variables in understanding nekton responses.

In the context of the broader estuary seascape, both the control and restoration sites, although lacking remnant biogenic habitat, contain artificial vertical and benthic structures, including cement wharfs, wooden pylons and debris. These pre-existing consolidated structures may already provide valuable nekton habitat (Chapman et al. 2018), and could make the additional structures from the oyster reef restoration functionally redundant. Similar outcomes have been documented in studies comparing fish communities of oyster reefs and vegetated habitats (e.g. saltmarsh, mangroves), with the lack of differences in communities being attributed to shared functional and structural attributes (Geraldi et al. 2009; Martinez-Baena et al. 2022). Subsequently, more pronounced nekton responses may be observable from large-scale restoration efforts, particularly in seascapes dominated by unstructured habitats (Gilby et al. 2018).

As oyster reef restoration within the Port River–Barker Inlet Estuary has been confined to a single site within the estuary, localised characteristics and scale-dependent limitations of our survey design may have limited the detection of nekton community enhancements at our restoration site. The pre-existing structural habitat (remnant pylons) at the restoration site, and its selection for oyster reef restoration, may have inversely affected nekton by attracting natural predators, or functioning as socio-ecological traps, whereby installed structures result in actual or expected catchability enhancements, leading to greater recreational fisheries pressure (Chong et al. 2024). Although not quantified in our study, predators, including dolphins, cormorants, and recreational fishers, were observed at both sites (B. Martin, pers. obs.). Their presence may have influenced species detections by modifying nekton behaviour and population dynamics, including non-lethal predation effects (Kimbro et al. 2014; Pauli and Sih 2017). Similarly, the higher turbidity (poorer water clarity) at the restoration site may have influenced nekton distribution, as turbidity can disrupt predator–prey interactions, modify habitat characteristics, and increase physiological stress and mortality (Kjelland et al. 2015). High turbidity can also influence fish detectability by visual monitoring methods, with sonar or extractive sampling as potential alternatives (Martin et al. 2025b). Within the Port Adelaide catchment, stormwater drainage has been identified as a primary source of suspended particles, with an estimated 26,900 kg year⁻¹ of pollutants being discharged annually (Southfront 2019). Monitoring additional restoration sites, should further oyster reef restoration occur within the estuary, may reduce some of these localised confounding influences and improve the detectability of potential nekton enhancements.

Variations in indicator species abundances underscores the complex responses of nekton communities to oyster reef restoration and broader environmental conditions. For example, following restoration activities, the restoration site in our study was strongly associated with crested oystergobies (C. gobioides), which was likely in response to the addition of recycled shell material used in the restoration substrate. Whereas this goby is non-native to South Australia (Wiltshire et al. 2010), recent studies have found strong habitat associations of that goby species to oyster reefs (Martinez-Baena et al. 2022; Martin et al. 2025c). Many Gobiidae species are reef-residents that use disarticulated oyster shells for nesting sites and refugia, and therefore respond positively to oyster reef restoration (Davenport et al. 2021; Lewis et al. 2021). In comparison, yelloweye mullet (A. forsteri) and western striped grunter (P. octolineatus) are highly mobile and transient fishes, and may benefit from the restoration site only for short durations associated with temporal shifts (i.e. diel, seasonal, ontogenetic) in their movement (Davenport et al. 2021). For A. forsteri, they are known to have undertaken seasonal movement into the upper section of the Port River in winter (Jones et al. 1996). The species A. forsteri is also a detrital feeder and may prefer unstructured habitats for feeding (Martin et al. 2025c), leading to higher abundances at the control site. These examples demonstrate that oyster restoration can have mixed and context-dependent outcomes for nekton species. Determining whether the observed species composition and responses change as the reef develops in future years would provide useful information of long-term outcomes, both in Port Adelaide and other urban waterways.

Considerations for shellfish reef restoration

Urbanised estuaries are complex systems for oyster reef restoration, with site suitability and reef resilience being limited by anthropogenic modification and spatial use (Howie and Bishop 2021). Although restoration initiatives are typically aimed at facilitating habitat-forming species, they may not optimise nekton habitat value (e.g. structure, prey), nor address stressors influencing their populations (e.g. fishing, pollutants, invasive species), which should be assessed as a part of developing restoration baselines (Gilby et al. 2018; Sievers et al. 2024). For example, at least 50 non-native species are established within the Port River–Barker Inlet Estuary (Wiltshire et al. 2010), including sessile marine pests that may compete with native O. angasi and challenge efforts to restore ‘natural’ shellfish reef baselines (Maus et al. 2024). In our study, restoration was associated with non-native Gobiidae presence, particularly C. gobioides. Although they are likely to have benign impacts, their co-occurrence with native gobies including F. lateralis and Arenigobius bifrenatus may result in increasing competition for benthic resources alongside the development of restoring shellfish reef systems (Hammer 2014). Other environmental stressors including sedimentation, stormwater inputs and pollution may similarly affect the capacity of restoration sites to sustain populations of habitat-forming shellfish and associated nekton communities (Gilby et al. 2018; Howie and Bishop 2021). Thus, practitioners should consider a range of abiotic and biotic variables when considering potential restoration suitability, particularly for urbanised estuary systems.

Whether shellfish reefs enhance nekton communities may depend on their capacity to provision ecological functions that modify structural and resource heterogeneity (Gilby et al. 2018). Our results indicated that in urbanised estuaries with artificial structural habitats, small-scale reef restoration can be functionally redundant for nekton communities, at least in the early stages and while reefs occupy a limited spatial footprint. Although not quantified in this study, the reef restoration may already support other beneficial functions such as enhanced biofiltration or macroinvertebrate productivity, which could indirectly benefit nekton communities; however, that would require further evaluation at smaller spatial scales (e.g. Andersson et al. 2023). The strong influence of seasonal and interannual environmental variation on the nekton community demonstrated why practitioners should also monitor abiotic variables when examining restoration outcomes (Baggett et al. 2015). Given the short duration of this study, longer-term monitoring and the installation of additional oyster reef sites is recommended to better understand nekton community dynamics and the role of abiotic variables. As shellfish reef restoration is increasingly being considered for degraded coastal habitats in urban areas, this study highlighted some of the challenges associated with the ecology of modified estuarine settings. Thus, in urbanised coastal settings, multiple, small-scale pilot shellfish reefs should be trialled to evaluate restoration risks and benefits so as to better inform upscaled habitat restoration aimed at long-term nekton enhancements.