The natural architecture of oyster reefs maximizes recruit survival

Abstract

The three-dimensional architecture of natural habitats is a key determinant of species biodiversity, harvestable biomass and resilience to disturbance^1,2. Indeed, some species, including trees, corals and oysters, alter resource availability and modify biotic and abiotic pressures through their own three-dimensional structures—thereby enhancing their own survival^3,4. However, which aspects of the three-dimensional architecture of these ecosystem engineers shape ecosystem dynamics and species survival are rarely examined by empirical studies, leaving much of the broader ecological and conservation impact of ecosystem engineering underexplored^4–6. Here we show that oyster reefs have combinations of geometric variables that maximize recruit survival, which is a key factor influencing oyster reef growth and persistence. Using three-dimensional habitat designs that capture the full spectrum of natural oyster reef architectures, as well as a geometric theory⁷ that links habitat surface area, fractal dimension and height, we show that oyster settlement and survival are greatest at particular combinations of fractal dimension and height that minimize predation. Our study provides a template for understanding optimal three-dimensional habitat configurations for habitat restoration projects that are proliferating globally, without targeting key architectural features of habitat space that maximize restoration success^8,9.

Survival of reef-building oysters peaks at specific combinations of fractal dimension and height that reduce predation, showing how optimal three-dimensional habitat geometry can guide more effective ecosystem restoration.

Main

Habitat structural complexity—the architecture of habitat space—is a key driver of species abundance and biodiversity throughout Earth’s ecosystems^1,2. Highly complex, three-dimensional habitats tend to support greater numbers of individuals than simple, planar habitats because abundance scales with area¹⁰ and surface area increases with habitat complexity^2,11. Habitat structural complexity also confers further benefits to species that extend beyond an increase in area, including the mediation of biotic and abiotic stressors². For example, along intertidal shorelines, areas of high structural complexity create shaded microhabitats that protect inhabitants against heat stress at low tide^12,13 and mitigate predation¹⁴. By mediating biotic and abiotic interactions, habitat structural complexity allows the establishment and persistence of species in areas that would otherwise be uninhabitable^2,15.

Many of the planet’s most iconic structurally complex habitats arise from the activities of ecosystem engineers—organisms that manipulate biotic or abiotic materials and by doing so, alter resource and habitat availability^3,4. Habitat modifications by ecosystem engineers can have profound effects on the distribution and abundance of species, as well as on their trophic relationships^16–18. Particularly in marine environments, consumptive interactions involving predators or grazers exert strong top-down influence on populations and their dynamics^19–21. In environments with high densities of consumers, habitat structural complexity can provide protection against predators by decoupling or weakening predator–prey interactions^2,22. For reef-building invertebrates with planktonic larvae, such as corals and oysters, years of successive settlement and growth of new individuals generate a highly complex reef structure composed of live and dead animals^23–25 (Fig. 1a,b). These biogenic reef structures present a high level of habitat structural complexity that can provide young recruits with both suitable settlement substrate and protection from the effects of environmental stressors and predators such as fish^26,27. However, the establishment of clear links between the structural attributes of invertebrate reefs and the key ecological processes influencing their population dynamics has historically been limited by our ability to measure habitat structural complexity with precision^6,28.

Fig. 1. Oyster reef habitat structural complexity underpinned the experimental design and hypothesis of the study.

a,b, Successive settlement and growth of new oyster individuals generate a highly complex reef structure, shown from afar (a) and close up (b), comprising live and dead animals. c, Here we compared oyster recruitment among 16 artificial habitat units (triangles), which spanned the full range of fractal dimension, height range and surface area displayed by 28 three-dimensional models of oyster reef surfaces (crosses), obtained from natural oyster reefs. Artificial habitat units and natural oyster reef surface models were of a standardized 15 × 15 cm planar area. Oyster reef surface models were obtained from four naturally occurring S. glomerata reefs in New South Wales, Australia, using photogrammetry together with structure from motion⁵². d, The 16 artificial habitat units used in the study resulted from crossing four levels of fractal dimension with four levels of height range, giving 11 different values of surface area across the 16 units. The experimental design explores a geometric relationship between surface area, fractal dimension and height range. Increasing the fractal dimension or height range (or both) while holding planar area constant results in an increase in surface area⁷. Image in a reproduced with permission from ref. ⁵³, John Wiley & Sons. Image in b reproduced from ref. ⁵⁴ under a CC BY 4.0 licence.

Here we test whether the reef-building oyster Saccostrea glomerata constructs reefs with habitat structural complexity levels that maximize recruit survival by mitigating predation pressure on young oyster recruits. Specifically, we examined how two geometric descriptors of reef structural complexity^7,29—fractal dimension and height range—influence oyster recruitment through predator mitigation. Fractal dimension is one of the most commonly used habitat structural complexity metrics^6,7,30–33; it captures how reef surfaces fill and fold within three-dimensional space^7,29. Fractal dimension measures space-filling at different scales and can be used to quantify the availability of protective spaces (for example, nooks and crannies), which can influence predator–prey interactions^31,33. Height range, along with planar area, determine the three-dimensional space occupied by reef structures⁷. Fractal dimension and height range can be combined, using a mathematical relationship, to describe the configuration of reef surface area within a planar area (also known as rugosity; Fig. 1c)⁷, providing a simple description of habitat structural complexity on the basis of two continuous variables.

We engineered 16 artificial larval settlement habitats that encapsulate the fractal dimension and height range of oyster reefs (Fig. 1c,d); we then deployed 480 replicates of habitat units across three different estuaries, half of which were caged to prevent oyster predation following larval settlement. We hypothesized that increasing habitat structural complexity would confer benefits to oysters that extend beyond increasing surface area for larval settlement, and that most of these extra benefits would arise from predator mitigation. We also expected that in the presence of predators, oyster recruitment would peak within the range of fractal dimension and height range of oyster reefs, following an ecosystem engineering hypothesis that reef-building oysters construct reefs with levels of habitat structural complexity that maximize oyster recruit survival by mitigating predation pressure.

Complexity mediates predation on oysters

Effects of habitat structural complexity on species abundances are often assumed to arise through effects on habitat area². As expected, based on area–abundance relationships^2,10, oyster counts increased monotonically with surface area in the caged treatment (Fig. 2a). In the uncaged treatment, by contrast, we found a hump-shaped pattern (Fig. 2a), suggesting that the effects of habitat structural complexity on oyster recruitment were not driven only by the area–abundance relationship but also by the mediation of predator–prey interactions (Fig. 2a, Supplementary Tables 1 and 2).

Fig. 2. Habitat structural complexity mediates predation on oyster recruits.

a, Relationships between oyster recruit counts and surface area (over a standard planar area) for caged and uncaged treatments. Oyster recruit counts varied significantly according to the interaction between surface area and treatment (P < 0.0001 based on two-sided Wald χ² test). b, Relationships between the density of oyster recruits (per centimetre squared of habitat surface, square-root-transformed) and fractal dimension for caged and uncaged treatments. Oyster recruit densities varied significantly according to the interaction between fractal dimension and treatment (P < 0.0001 based on a two-sided Wald F-test). c, Relationships between the density of oyster recruits (per centimetre squared of habitat surface, square-root-transformed) and height range for caged and uncaged treatments. Oyster recruit densities varied significantly according to the interaction between height range and treatment (P = 0.03795 based on a two-sided Wald F-test). Points represent raw (a) and square-root-transformed (b,c) data at individual habitat units (n = 240 caged and 240 uncaged habitat units deployed within three research sites: Hawkesbury, Port Hacking and Brisbane Waters). Solid lines denote the mean fit from a generalized linear mixed model (GLMM) (a) and from linear mixed models (LMMs) (b,c). Shaded areas denote 95% confidence intervals. Statistical parameters and significance of each model are found in Supplementary Tables 1–4.

Comparisons of the effects of fractal dimension and height range on oyster densities (oyster counts per unit area) between caged and uncaged habitat units showed that each of these geometric descriptors has a role in mediating predation on oyster recruits (Fig. 2). In the caged treatment, we found flat relationships between oyster densities and both fractal dimension and height range (Fig. 2b,c and Supplementary Table 3), suggesting that oyster larval settlement is most probably an area-dependent process with little larval preference for specific levels of fractal dimension and height range. Post-settlement survival, however, is highly influenced by predation^21,34. At low levels of fractal dimension and height range, caging enhanced the density of oyster recruits through predator mitigation; however, as fractal dimension and height range increased, the effects of caging diminished and were closer to (but still significantly different from; see Supplementary Table 3) the uncaged densities. This mechanism arises because increasing habitat structural complexity can generate habitat compartmentalization, which in turn creates a greater range of predator-free microhabitats^2,7.

Ecosystem engineering in oyster reefs

To test whether reef-building oysters construct reefs with combinations of geometric descriptors that maximize oyster recruit survival through predator mitigation, we evaluated the interactive effects of fractal dimension and height range on the density of oyster recruits in the presence of predators. When the optimal fractal dimension and height ranges for oyster density generated from our model were compared with observed values for natural oyster reefs (Fig. 3), we found that most reefs occupy the region in which oyster densities are in the top 10% of those found in the experiment. This striking concordance supports our hypothesis that reef-building oysters construct reefs with levels of habitat structural complexity that maximize recruit survival by mitigating predation pressure.

Fig. 3. Ecosystem engineering maximizes oyster recruit survival.

Predicted oyster recruit densities (per square centimetre, square-root-transformed) on the basis of the artificial habitat unit experiment plotted as the surface descriptor plane given by height range and fractal dimension. Oyster recruit density was significantly influenced by the interaction between fractal dimension and height range (P = 0.002875, two-sided Wald F-test; n = 240 uncaged habitat units). Prediction contours are from the LMM summarized in Supplementary Table 5 and represent 5% increments in the predicted density of oysters. Statistical parameters and significance of the model are found in Supplementary Tables 5 and 6. The black circles represent the mean height range and fractal dimension of four naturally occurring oyster reefs. Error bars represent the s.d. from 15 × 15 cm samples taken within four reef patches.

For biogenic habitats formed by aggregations of individuals (for example, coral and oyster reefs), persistence can be dependent on self-facilitation, whereby the habitat complexity conferred by groups of individuals provides new recruits with protection from predators and environmental stressors^15,35,36. By showing that oysters can build a habitat structure that counteracts predation pressure—one of the key natural processes that affect their population stability—our work provides empirical evidence that helps explain why oysters modify the seascape by aggregating to form reefs^17,18. According to the niche construction perspective, ecosystem engineers can modify selective pressures for subsequent generations and, consequently, the evolution of traits that benefit their fitness^5,37. Although this study was not designed to assess whether oyster reef formation arises out of an evolutionary process, our results warrant further investigation to evaluate whether oysters have evolved to build habitat structures that maximize the survival of young recruits.

Habitat geometry in restoration practices

Oyster reefs are critically endangered ecosystems that have suffered a greater than 85% decline since industrialization^38–40. Restoration of oyster reefs is rapidly scaling up, with investments generally made under the premise that restored reefs improve water quality, enhance fish production and protect coastal areas from erosion⁴⁰. Although there have been several examples of the successful re-establishment of self-sustaining oyster reefs, by 2016 nearly half of oyster reef restoration projects had failed⁴¹. Oyster reef restoration generally involves deployment of hard substrate (for example, rock, shells, concrete) to create a base on which oyster larvae can settle and grow to form a reef^42–44. Substrate deployments are generally necessary because historic overharvest using destructive fishing methods removed not only live oysters but the structurally complex foundation of dead shells on which oyster reefs build³⁸. In many instances, unsuccessful restoration has been attributed to recruitment failure related to low habitat structural complexity of the substrate provided^43,45,46. Still, there is no consensus on the levels of substrate complexity that can maximize oyster recruitment and, consequently, increase the likelihood of restoration success.

Our study provides a template for developing targets of habitat structural complexity metrics that can be incorporated into oyster reef restoration guidelines to improve restoration success. Our results also highlight that fractal dimension and height should be used in combination rather than independently, as each of these metrics is important for maximizing oyster recruitment. Here, optimal values of fractal dimension (2.41) and height range (7.96 cm) yielded oyster densities that were 35% higher than densities obtained with the maximum fractal dimension (2.52) and height range (12 cm), whereas low values of fractal dimension and height range resulted in little to no recruitment (Fig. 3). Moreover, the maximum height and minimum fractal dimension yielded 90% fewer oysters and the minimum height and maximum fractal dimension yielded 53% fewer oysters than optimum levels of these two descriptors of habitat structural complexity. It is important, however, to highlight that although our results are readily applicable to restoration of intertidal S. glomerata oyster reefs, further tests and different habitat configurations may be required to maximize the survival of other oyster species living under different environmental conditions^34,47.

Our results suggest that as marine and terrestrial habitat restoration continues to scale up globally⁸, projects should look to nature to guide the selection and design of restoration methods. In instances in which habitat destruction and degradation have removed the structural, three-dimensional foundation for habitat formation, our approach provides a framework for increasing the success of future habitat restoration efforts. First, we collected information from our natural study system to understand the variability in habitat complexity metrics between remnant patches. Second, we used three-dimensional modelling techniques to create surface designs that spanned and extended beyond the level of surface complexity found in nature. Finally, we used a robust field experiment to test predictions on habitat geometries that maximize recruitment of the foundational ecosystem engineer, providing proof-of-concept for application to scaled restoration projects. Our approach—which decomposes habitat structural complexity into two descriptors of surface geometry—greatly elevates the current approaches to biomimicry of emergent structural traits of ecosystems that are critical to their development¹⁵. It does so by identifying those specific geometric combinations that maximize recruitment and, hence, habitat formation. Although our experimentation focused on small concrete units, the geometric understanding generated by this study can be fed into the industrial design frameworks that are increasingly being used to develop scaled solutions for restoration⁴⁸, often from biodegradable materials.

Conclusion

Here we show that habitat structural complexity mediates predation on oyster recruits and that these effects are in addition to the effects of habitat structural complexity on species abundances arising from increasing surface area. We also show through experimentation with new habitat units that the values of height range and fractal dimension observed on remnant oyster reefs match those that are optimal for oyster recruitment—suggesting that oysters construct reefs that maximize the survival of new recruits by providing refuge from predation, which, consequently, increases their persistence in the marine seascape^25,35. Positive feedbacks have important roles in the population dynamics of many ecosystem engineers^17,18 and have important implications for the conservation of threatened biogenic habitats^5,49. By showing that three-dimensional ecosystem engineering maximizes recruit survival, our study highlights the importance of conserving not only habitat area, but the integrity of the natural architecture of biogenic habitats; it also helps explain why the greatest success of oyster reef restoration occurs in areas in which the underlying habitat structure of oyster reefs, provided by dead oyster shells, has been preserved^50,51. Combined, our results provide a baseline for designing habitat restoration projects that use artificial substrates to facilitate habitat formation by reef-forming organisms. Our results also highlight the importance of looking to nature when designing restoration projects.

Methods

Experimental design

Reef structural complexity metrics—fractal dimension and height range—were manipulated using artificial habitat units, of standardized 15 × 15 cm planar area, but of varying three-dimensional geometry and surface area. There were 16 distinct designs (Fig. 1c), produced by crossing four levels of each fractal dimension and height range to produce 11 resultant levels of surface area. The range of levels of fractal dimension (2.02−2.52), height range (3−12 cm) and surface area (244.75−1,467.79 cm²) were centred around, but extended beyond, values observed for S. glomerata oyster reefs in Sydney, New South Wales, Australia (mean ± s.e.m., fractal dimension = 2.31 ± 0.009, height range = 7.67 ± 0.40 cm, surface area = 571.01 ± 17.22; Fig. 3).

Values of fractal dimension, height range and surface area were calculated from digital elevation models (DEMs) using the habtools package of R (ref. ⁵⁵); DEMs for all artificial and natural surfaces were generated from three-dimensional models using the ‘mesh_to_dem’ function of habtools. All DEMs had the same resolution (1 mm), to allow for comparisons between surfaces. Fractal dimension was estimated using the fd() function and the variation method, based on measurements taken across six spatial scales obtained by subdividing each DEM into squares with side lengths of 15, 7.5, 3.75, 1.875, 0.9375 and 0.46875 cm. The variation method was used because it allows estimating the fractal dimension of real surfaces⁵⁶, which do not necessarily behave like pure fractals (for example, they are rarely self-similar across scales). Height range was calculated using the hr() function, and surface area was calculated using the surface_area() function. Three-dimensional models of oyster reefs (15 × 15 cm, n = 28), were obtained using photogrammetry together with structure from motion⁵² (Supplementary Methods) from four naturally occurring reefs located at Towra Point Nature Reserve, New South Wales, Australia (34°0′56.21″ S, 151°10′ 44.57″ E). Reefs from Towra Point were selected because, at the time of the study, this was the only known location in the Greater Sydney region that contained multiple large undisturbed S. glomerata reefs³⁹.

Three-dimensional models of the 16 artificial habitat units were designed using the three-dimensional graphics software Blender. Each of the 16 artificial habitat designs was three-dimensionally printed in polylactic acid filament and used as a master to create silicon moulds that allowed casting replicate units using concrete. A total of 500 concrete habitat units were produced (480 habitat units for hypothesis testing and 20 habitat units for assessing caging artefacts; see the next section). Concrete releases positive settlement cues for oysters, is easily moulded into a range of geometries and is used as a substrate in restoration projects^44,57.

Field experiment

To assess the effects of habitat complexity descriptors on oyster recruitment, we used habitat units (n = 10 for each of the 16 designs) at three estuarine sites in the greater Sydney region, New South Wales, Australia, in October and November 2022: Brisbane Waters (33°30′40.29″ S, 151°10′18.75″ E), the Hawkesbury River (33°33′46.16″ S, 151°13′46.70″ E) and Port Hacking (34°04′41″ S, 151°06′38.03″ E). These sites are adjacent to S. glomerata oyster reefs, indicative of larval supply and support a diverse range of oyster predators⁵⁸ (Supplementary Table 7). At each of the sites, five units of each design were caged (1 cm mesh; to exclude finfish predators) and five were uncaged (so they could be accessed by predators). At each of the three sites, we deployed units at a mid-intertidal elevation (0.6–0.9 m above the lowest astronomical tide; daily inundation period of about 56%), with units haphazardly interspersed with respect to treatment along the shoreline and separated by at least 15 cm from neighbouring units or natural habitat features (for example, rocks). After twelve months—a period long enough to allow for both settlement and post-settlement mortality to occur—we counted the number of oyster recruits on the surface of each experimental unit. This required scraping oysters out of habitat features to accurately enumerate them, preventing repeated sampling. To test for caging artefacts, partial cages that allow predator access were established for a subset of four designs with five replicates each at one randomly selected site (Brisbane Waters; Supplementary Table 8). We found no effects of caging artefacts in oyster recruitment (Supplementary Table 8). Research activities were performed under the New South Wales Department of Primary Industries scientific collection permit P07/0047-7.0 and approved by Macquarie University Animal Ethics Authority (2019/036).

Statistical analyses

Surface area and height range were base-10-log-transformed to improve model residuals. Oyster densities were square-root-transformed to improve model residuals. Oyster densities were calculated by dividing oyster counts found at each experimental unit by the surface area of the experimental units. We quantified habitat structural complexity–abundance relationships using a GLMM (for count data) and LMMs (for density data), with second-order polynomials to allow for non-linear relationships between predictor and response variables. Study sites were included as random effects. To test our hypothesis that increasing habitat structural complexity would confer benefits to oysters that extend beyond increasing surface area for larval settlement, and that most of these additional benefits would arise from predator mitigation, we first compared the effects of surface area on oyster counts (Fig. 2a) between habitat units with and without cages using a GLMM with a negative binomial distribution. We then compared the effects of both fractal dimension and height range on oyster densities at habitat units with and without cages (Fig. 2b,c) using LMMs to highlight the influence of predation in driving habitat structural complexity–abundance relationships. To test our hypothesis that ecosystem engineering maximizes recruit survival per unit area, we quantified the interactive effect of fractal dimension and height range (for uncaged units only) on oyster densities using a LMM (Fig. 3 and Supplementary Table 5). Residuals for all these models were approximately normal and were homogeneous when plotted against predictor variables. To test for caging artefacts, a zero-inflated generalized linear model was run using a negative binomial distribution to assess treatment effects (n = 2; uncaged and partially caged) on oyster counts. All analyses were conducted in R (ref. ⁵⁹ ; v.3.6.1) and can be accessed at https://github.com/juan-muelbert/Oyster-complexity.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-026-10103-8.

Author contributions

J.R.E.M, L.F., K.Z., M.J.B. and J.S.M conceptualized the study. J.R.E.M designed and produced the artificial habitat units used in the study. J.R.E.M, L.F. and K.Z conducted the field experiment, ran the analyses and produced the figures. K.E. and W.F. generated the three-dimensional models of naturally occurring oyster reefs. J.R.E.M led the project. J.R.E.M and M.J.B funded the project. J.R.E.M wrote the first draft of the paper and all authors reviewed at least one draft.

Peer review

Peer review information

Nature thanks Jonathan Shurin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

Source data for statistical analyses and figures are available at https://github.com/juan-muelbert/Oyster-complexity.

Code availability

Code for statistical analyses and figures is available at https://github.com/juan-muelbert/Oyster-complexity.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-026-10103-8.