Alkalinity-enhanced artificial substrates modulate local pH and increase survivorship of early-stage coral recruits

Coastal cities and communities are susceptible to the impacts of waves, flooding, storm surge, and sea-level rise. In response to these threats, coastal jurisdictions typically invest in engineered shoreline defenses such as breakwaters and seawalls that are expensive to build and maintain. To date, built infrastructure, including sea walls, levees, culverts, bulkheads, and other hardened structures have dominated thinking about coastal protection¹. However, there is increasing interest in using natural habitats, including wetlands, dunes, barrier islands, sea grasses, coral and oyster reefs, and mangroves to reduce the risk of coastal flooding and erosion while providing other social and economic benefits². Many co-benefits associated with natural infrastructure draw people to live and work in these vulnerable regions. The coastal ecosystems that enhance resilience by providing protective services also contribute raw goods and materials, plant and animal habitat, water and air quality regulation, carbon sequestration, nutrient cycling, and opportunities for tourism, recreation, and education³.

There is growing global interest in combining the use of built and natural infrastructure to help coastal communities become more resilient to extreme events and reduce the risk of coastal flooding through the creation of hybrid coastal resilience structures. One such approach is the deployment of an artificial reef base structure consisting of a submerged breakwater onto which reef-building corals are actively planted, creating a hybrid reef. These hybrid reefs, which incorporate an overlay of living corals, could offer the immediate benefits of wave attenuation while providing ecosystem benefits and promoting long-term ecosystem resilience⁴. Coral reefs are naturally formed, low-crested breakwaters that aid in wave energy dissipation and flood reduction and represent biodiversity hot spots that provide valuable ecological and economic benefits^5,6. Healthy coral reefs can attenuate wave energy by up to 97%, providing tremendous natural protection from coastal hazards⁷. Considering this, hybrid reefs serve as opportunities to offer communities the immediate benefits of shoreline protection and the long-term benefits of restoring coral reefs, and important economic benefits and ecological services. In fulfillment of this vision, hybrid coral reef projects would become testbeds for new methods of coral restoration and increasing coral resilience to climate change.

While the concept of hybrid coral reefs sounds attractive, natural coral reefs around the world are rapidly diminishing. According to the latest Status of the World’s Coral Reefs report, between 2009 and 2018, there was a progressive loss of approximately 14% coral cover from the world’s coral reefs⁸. This was primarily due to recurring large-scale coral bleaching events, combined with other local pressures such as coastal development, land-based and marine pollution, unsustainable fishing, disease, and tropical storms.

The global decline in coral populations creates a snowballing problem of decreased larval supply and genetic diversity, suppressing recovery following disturbance. Larval availability, settlement success, post-settlement survivorship, and growth have long been recognized as key factors driving the recovery of coral reefs following disturbances⁹. Settlement rates of corals in Caribbean reefs are much lower than in reefs in the Pacific and Indian Oceans¹⁰. This difference may result from much lower coral cover on Caribbean reefs and hence, a smaller supply of larvae compared to larger reefs¹¹. Additionally, slow-growing coral recruits remain vulnerable to predation and competitive exclusion by macroalgae for longer than faster-growing corals, further reducing survivorship¹².

Ocean acidification is a known coral stressor, causing corals to grow more slowly or maintain a constant growth rate by diverting energy from other essential life processes. This led us to hypothesize that alkalinity enhancement (AE) might be a way of relieving physiological stress of the coral spat. Adding alkaline compounds to the environment has been shown to increase calcification of adult corals^13,14,15. However, these studies involved adding chemicals to the seawater around corals. While the additions had the desired effect, this approach is not sustainable long-term due to the large amounts of chemicals that would need to be added given the volume of seawater that flows over a coral reef every day. Alternatively, compounds increasing alkalinity can be added to the substrate where they leach into the water column, modulate pH, and provide an alkalinity-enhanced (AE) boundary layer for coral spat (Fig. 1a).

Substrate-level manipulation, particularly to aid in coral settlement, has been studied. Laboratory coral settlement is often aided by the presence of glycoglycerolipids and polysaccharides cues from crustose coralline algae (CCA) and associated bacteria¹⁶. Tetrabromopyrrole (TBP) is another chemical compound that can affect settlement¹⁷. Coral settlement may also be positively affected by soluble inorganic compounds such as silica and strontianite¹⁸. Substrates made with magnesium additions, such as magnesium carbonate or magnesium sulfate, can also increase coral settlement^19,20. These studies establish a precedent that modifying the chemical composition of the substrate and, the local chemical environment directly above the substrate can aid in coral settlement and increase restoration success. We seek to add to these substrate studies by adding alkaline materials to cement, hypothesizing that alkaline substrates will increase local pH, which will aid in the initial settlement, survivorship, and long-term growth of corals.

We designed AE tiles made with four different chemistries (Portland limestone cement with no additions, 1% sodium bicarbonate by weight, 1% sodium carbonate by weight, and 2% sodium carbonate by weight, referred to as tile “chemistries”) and two different surface topographies (a flat tile and a textured tile with concave divots) as substrates. These AE tiles may increase alkalinity, observed as the consequential modulation of local pH by diffusion and advection within a boundary layer in the laminar flows generated using a flume (Fig. 1b–d)²¹. Carbonate and bicarbonate ions leaching out of the tiles may be sufficient to stimulate early growth of coral spat through either direct contact of the coral with the substrate or while the coral is small enough to occupy the laminar boundary layer (Fig. 1a). This AE method via substrates could be less resource-intensive than bulk additions by mixing small amounts of sodium carbonate/bicarbonate into the cement used to fabricate artificial or hybrid reefs, or by the production of tiles using these chemistries for the land-based production and grow-out of coral spat.

Tile verification: diffusion and advection

Statistical analysis of tile chemistry was performed on linear ΔH⁺ measurements to allow for analysis via linear methods (Supplementary Fig. 1), but is presented throughout this paper as the logarithmic ΔpH. Delta values refer to the change in pH (H⁺) with respect to the initial conditions of the water before a tile was added.

All tiles significantly changed pH (two-way factorial ANOVA: DF_Chemistry = 3, F_Chemistry = 8.22 × 10³, p_Chemistry < 0.0001), to differing degrees based on the presence of flow (two-way factorial ANOVA: DF_Flow = 1, F_Flow = 4.83 × 10⁴, p_Flow < 0.0001; DF_{Flow × Chemistry} = 3, F_{Flow × Chemistry} = 7.03 × 10³, p_{Flow x Chemistry} < 0.0001). Adding 1–2% sodium bicarbonate or sodium carbonate by weight to cement tiles was sufficient to consistently elevate pH by 0.2–0.5 pH units under stagnant conditions after 6 h (Fig. 1a). In experimental flumes with advective flow, the pH modifications are less stable over time than in stagnant conditions. While there appears to be no consistent change in pH over time (Fig. 1b), when the ΔpH data is averaged over the 6 h window, 1% bicarbonate and 1% carbonate addition tiles increase pH by 0.1 pH units within the laminar boundary layer (Fig. 1e). Near-substrate laminar boundary layer flows (cm scale) within a coral reef framework were modeled in flumes by matching flow speeds to those observed within a coral reef canopy (1–2 cm s⁻¹)^22,23,24,25 (Fig. 2a–c). All AE tiles significantly increased pH within the boundary layer in laminar flow; however, the effect of 2% carbonate tiles under advective flow was greatly diminished compared to the diffusive (stagnant) setting (Fig. 2c and Supplementary Table 1). The smaller increase of pH in the advection setting compared to diffusion setting occurred because the pH changes arising from the tiles was continually diluted by incoming seawater.

Longevity of tiles

Linear regressions were fitted to the change in linear ΔH⁺ concentrations, rather than the logarithmic pH (graphed as ΔpH) for each of the four different chemistries with respect to tile age (1–84 days post cure) for measurements at the tile surface and 3.26 ± 0.09 mm above the tile surface in flow conditions, to capture long-term changes above the tile surface and within the laminar boundary layer (Fig. 2d). There was one outlier in the entire dataset noted for 2% carbonate additions that occurred on day 1 and was excluded from the analysis (but included in Fig. 2d for reference). Three out of the eight total regressions were significant, indicating the declining effectiveness of tiles with no additions (p_Surface = 0.01), and 2% carbonate additions (p_Surface < 0.01, p_Above = 0.03) to increase pH over the 12-week aging experiment, while 1% bicarbonate addition (p_Surface = 0.09, p_Above = 0.20) and 1% carbonate additions (p_Surface = 0.16, p_Above = 0.09) did not change with respect to time (Fig. 2d and Supplementary Table 2).

Data points were grouped by chemistry and measurement location over the duration of the experiment to identify broad trends with respect to chemistry and height above substrate surface. The resulting pH varied with respect to chemistry (two-way factorial ANOVA: DF_Chemistry = 3, F_Chemistry = 11.50, p_Chemistry < 0.0001) and position relative to the tile’s surface (two-way factorial ANOVA: DF_Position = 1, F_Position = 35.70, p_Position < 0.0001) to different degrees for each chemistry (two-way factorial ANOVA: DF_{Chemistry × Position} = 3, F_{Chemistry × Position} = 6.95, p_{Chemistry × Position} < 0.001). At the surface of the tile, 1% bicarbonate additions and 1% carbonate additions resulted in the greatest increases in pH (Supplementary Table 3). Additionally, one-tailed Students’ t-tests compared the aggregated ΔH⁺ values to an estimated mean of zero to determine if the lower-performing tiles (no additions or 2% carbonate additions) still increased pH compared to the initial seawater pH when no tile was present. A non-significant result would indicate that there was no change in the ΔH⁺ values with the tile present compared to the initial ambient conditions. All tiles, except for those with no additions, had an average pH higher than initial ambient conditions before the tiles were added at the surface of the tile and within the laminar boundary layer (Supplementary Table 3) (Fig. 2e). Overall, these aging tests demonstrated that the AE tiles could elevate pH within the fluid boundary layer at least 12 weeks, to a maximum of 0.05–0.06 units (1% carbonate additions).

Chemistry and topography

The tile topographies in this experiment included a flat surface (a “flat” tile), and a surface textured with a 4 × 4 grid of cylindrical (diameter = 3.0 mm, depth = 1.86 mm; Supplementary Fig. 2) divots (a “textured” tile). There were minimal differences in the flow above the tiles between these two surface topographies with regard to velocity and vorticity, indicating minimal breakdown of the laminar boundary layer resulting from the tile divots (Supplementary Fig. 3). This was because the smooth, laminar flow within the flumes did not generate strong recirculation regions.

One important limitation of this flow quantification is the inability to measure flow within the divots. While any flow within the divots cannot be observed or measured, it is reasonable to conclude that flow in the divots is reduced compared to that over the surface of the tile (Supplementary Fig. 3). Streamlines of particle image velocimetry (PIV) measurements indicate fairly laminar flows over the tiles. However, there is likely weak recirculation within the divots arising from boundary layer separation of fluid due to pressure differences within and outside of the divot²⁶.

Even though there were negligible differences in the hydrodynamic environment between flat and textured tiles, the textured tiles featured a more complex chemical landscape about them. Within the textured tiles, ΔpH varied with respect to chemistry and location (in a divot or on the surface of the tile) (two-way factorial ANOVA: DF_Chemistry = 3, F_Chemistry = 9.97 × 10³, p_Chemistry < 0.0001; DF_Location = 1, F_Location = 1.53 × 10⁶, p_Location < 0.0001; DF_{Chemistry × Location} = 3, DF_{Chemistry × Location} = 8.35 × 10³, p_{Chemistry × Location} < 0.0001 (Fig. 2f, g, Supplementary Table 4). Because there was minimal flow into the divots, resulting in minimal advection, pH in the divots was significantly higher than at the surface. Hence, the divots provide a region which enhanced the chemical effects from the tile.

Settlement

To verify the biological effectiveness of AE tiles, O. faveolata was settled on AE tiles in stagnant conditions with CCA and later grown in flumes where the chemical environment about the tiles is defined. To identify any settlement preference, the number of O. faveolata settlers per tile was analyzed on day 0 with respect to chemistry and topography. There was no significant difference in settlement based on chemistry (Kruskal–Wallis: DF = 2, χ² = 1.46, p = 0.48), however, textured tiles had higher settlement numbers than flat tiles (one-tailed t-test: DF = 13, t = 4.78, p < 0.001, Fig. 3a, Supplementary Table 5). On textured tiles, where corals had the option of settling on the surface or in a divot, corals preferred settling in divots (Wilcoxon: DF = 11, Z = −2.32, p = 0.02, Fig. 3b).

Survivorship

Total survivorship at the end of the growth experiment (139 d) was analyzed with chemistry and topography as factors. The effect of chemistry on total survivorship across all tiles is almost significant (one-way ANOVA: DF = [2,12], F = 3.59, p = 0.06), and the effect of topography on total survivorship is greater on flat tiles than textured tiles (one-tailed t-test: DF = 13, t = −2.73, p < 0.01) (Fig. 3c). There is an interactive effect between chemistry and topography (one-way ANOVA: DF = [4,10], F = 4.27, p = 0.03), as 2% carbonate addition tiles performed significantly better than textured 1% carbonate tiles (Tukey’s HSD: t = −3.40, p = 0.04; Supplementary Table 5).

Within the textured tiles, there was no significant difference in survivorship between coral that settled on the flat surface with those that settled within a divot (one-tailed t-test: DF = 10, t = 1.44, p = 0.09), although it appears that survivorship may be worse in divots. Survivorship on flat, exposed surfaces was the same across tile topographies for tiles with no additions (one-tailed t-test: DF = 4, t = 0.02, p = 0.55) and 1% carbonate additions (one-tailed t-test: DF = 4, t = −1.28, p = 0.13), and so when location survivorship is compared across all tiles, larvae that settled in the divots have a significantly lower survivorship than those that settled on the exposed surface (one-tailed t-test: DF = 19, t = 2.58, p < 0.01). Similar to the total survivorship analysis, an interactive effect between chemistry and settlement location (one-way ANOVA: DF = [4,16], F = 3.15, p = 0.04) reveals that coral settled on the flat surface of 2% carbonate tiles has a greater survivorship than those that settled in the divots of 1% carbonate tiles (Tukey’s HSD: t = −3.12, p = 0.04).

Growth

Absolute change in growth measured in three dimensions – vertical height, contact surface area with the tile, and cylindrical volume—of O. faveolata was analyzed with respect to chemistry, topography, and an interactive effect of chemistry and topography. Ultimately, the AE tiles had no effect on absolute change in growth (Table 1). The strongest trend observed was that O. faveolata contact surface area appears to vary between tile topographies (two-tailed t-test: DF = 10, t = 1.40, p = 0.19), although this observation is not statistically significant. This trend may result from negative interactions with algae growing in the divots. Indeed, lower growth on textured tiles in general may be because coral growing in divots experiences higher competition from other corals and/or algae in these confined spaces or were spatially limited in the area to which they could grow. However, we could not test this directly because our profile imaging for measuring vertical growth could not measure the height of corals growing in divots; individuals settled in divots were excluded from growth analyses.

Adding 1–2% sodium carbonate or bicarbonate by weight to cement tiles was sufficient to elevate the pH in the boundary layer by 0.2–0.5 pH units under stagnant conditions. When flow was introduced, 1% carbonate and 1% bicarbonate tiles yielded the largest increase in pH (0.1 units) across all chemistries (Fig. 2). This smaller increase of pH in the advection setting compared to diffusive setting for all chemistries is expected because the carbonate and bicarbonate ions leaching out of the cement will be diluted by the volume of total seawater. The 2% carbonate tiles have a significantly smaller impact on local pH in flow conditions compared to diffusive conditions in both the advection verification and in the aging experiments (Fig. 2).

The reduced effect of 2% carbonate addition tiles on pH in the flow compared to diffusion is unexpected. However, similar ΔpH values for 2% carbonate on the surface and above the surface indicate a more uniform chemical environment above the tile. This becomes apparent when estimating the Péclet number (Pe) for carbonate ions in this environment. Pe represents the ratio of the advective transport rate to the diffusive transport rate of a substance and is calculated as

where u represents the average bulk flow speed (0.012 m s⁻¹), L is the characteristic length scale with respect to ion transport (0.10 m, the water depth of the flume), and D is the diffusion coefficient for a given substance, which can be estimated as

where k_B is the Boltzmann constant (1.38 × 10⁻²³J K⁻¹), T is the absolute temperature of the system (293 K), μ is the dynamic viscosity (8.9 × 10⁻⁴Pa s), and r_i is the radius of ion in question. When Pe < 1, diffusion dominates transport, when Pe = 1, advection and diffusion are of equal importance, and when Pe > 1, advection dominates transport²⁶. Here, Pe values for carbonate (r_i = 178 pm) and bicarbonate (r_i = 156 pm) are both much greater than 1 (Pe_Carbonate = 6.7 × 10⁴, Pe_Bicarbonate = 7.5 × 10⁴), indicating that advection is dominant over diffusion within the boundary layer. This is observed in the weaker effect that tiles have on pH when in advective conditions compared to diffusive conditions.

Under flow conditions, the 2% carbonate tiles did not increase pH to the same extent as the 1% carbonate tiles, despite containing a higher proportion of added carbonate. This pattern was consistently observed throughout the aging experiment, suggesting that the response was reproducible rather than anomalous. Although the carbonate content differed among tile types, Portland limestone cement contains additional reactive constituents (e.g., CaO, SiO₂, Al₂O₃, and Fe₂O₃) that may influence local carbonate chemistry at the cement–water interface. In addition, all tiles developed visible surface precipitates, likely calcium carbonate, which may have altered near-surface chemical conditions under flow. Notably, despite the limited pH modification observed for the 2% carbonate tiles, measurements of Ca, DIC, and TA showed the largest increases for this treatment (Supplementary Fig. 4 and Supplementary Table 6), indicating substantial chemical modification of the local environment. Further work is needed to resolve the coupled chemical and mineralogical processes governing this response.

In this study, we focused on unidirectional flow conditions to establish a baseline understanding of how AE tiles modify local chemistry and create a stable microenvironment. However, unidirectional flow represents a simplified model of reef hydrodynamics and does not fully capture the complexity of natural reef flows. Future studies to better understand the relationship between the chemistry, flow, and biological response to AE tiles should expand upon this work to study the effectiveness of AE tiles in oscillatory flow conditions, which are more representative of small-scale reef flows.

Initial settlement of O. faveolata was not strongly affected by the carbonate chemistry of the tiles tested; however, settlement was affected by topography. Orbicella faveolata preferred settling on textured tiles rather than flat tiles, and within textured tiles, a strong preference to settle within the confines of a divot rather than on the surface (Fig. 3). Coral recruits are more likely to settle in topographical features closer to their size to maximize their area for attachment, a phenomenon coined “attachment point theory (APT)”^27,28. In this experiment, the 3 mm diameter of the divot was approximately six times the diameter of the recruits, which provides more surface area as the coral settler can attach to the divot bottom or the divot wall. While the size of the divots tested here is larger than that of the recruits, thereby not strongly supporting attachment point theory, O. faveolata did prefer settling in divots rather than on the surface, an observation consistent with those of Whalan, et al.²⁸.

Although O. faveolata preferred settling on textured tiles, and specifically, within the divots on textured tiles, O. faveolata that settled in divots suffered higher mortality rates than coral settled on flat, exposed surfaces, regardless of AE additions. It is possible the environment within the divots was too sheltered from the outer flow. Minimal flow was observed in the divots, creating relatively static flow conditions. With no flow to drive advective processes, Pe « 1, and diffusion dominates the chemical landscape within the divots. A purely diffusive environment limits access to ambient nutrients introduced by the flow and restricts the environment’s ability to flush waste products away from the coral. While it is possible that there is minimal flow within the divots resulting from boundary layer separation²⁶ and/or coral-generated ciliary flows for nutrient turnover²⁹, these local flows cannot generate the magnitude of turnover required to flush carbon dioxide and other waste products away while bringing in nutrients and dissolved oxygen for growth, resulting in an anoxic environment. Minimal flow in the divots also allowed pH to be significantly higher in the divots compared to the surface. While the experiment intended to expose coral to increased pH conditions, it is possible that the chemical environment within the divots was too extreme. Another consequence of using the flumes to grow coral was that tiles had to be cleaned by hand rather than co-reared with herbivores to control algal growth. Scraping algae out of the divots during cleaning, while not harming coral, was a delicate process. Moreover, the algae in the divots could have served as additional competition for resources in an already resource-limited environment³⁰.

Coral survivorship over the 139-day growth experiment, regardless of settlement location, increased by a factor of 2.5–2.9-fold on the 1% and 2% carbonate tiles compared to the tiles not receiving any AE additions (Fig. 3) however, growth had minimal variation with respect to chemistry (Table 1). We initially hypothesized that any increase in survivorship would be a result of increased growth. Ocean acidification makes it difficult for calcifying organisms to build their carbonate skeletons, particularly during early stages of development, by raising the energetic demands for growing skeletons^31,32,33,34. For corals, the first deposition of CaCO₃ for skeletal growth occurs during the 1–2 weeks after metamorphosis. The quicker this initial deposition occurs, the greater the success of the organism surviving to reproductive maturity³⁴.

However, we did not observe an increase in growth in the corals tested. When planning for these experiments, we envisioned that AE tiles would elevate pH in the laminar boundary layer and corals would benefit from that environment while they were small enough to occupy that zone by growing faster. The observation that survivorship is improved on AE tiles without significantly faster growth suggests that corals benefit from the AE tile chemistry without augmenting their skeletal growth. It is possible that corals survived better because their energy was reallocated to other important processes rather than skeletal growth that were unaccounted for in the experiment, such as tissue growth, or it could be that skeletal density rather than areal extension was higher under AE enhancement, enhancing attachment strength and eventual survivorship.

Corals may also benefit from tile chemistry through direct contact with the substrate, in addition to benefiting via flow. This is evidenced by the enhanced survivorship on tiles that had the least effect on pH under flow – 2% carbonate addition tiles (Fig. 3). Post-hoc measurements of total alkalinity (TA), dissolved inorganic carbon (DIC) and calcium (Ca²⁺) from all AE tiles under stagnant conditions revealed decreased values compared to tiles with no carbonate additions (Supplementary Fig. 4). All tiles during multi-day experiments, including the tiles with no carbonate additions, developed a visible precipitate on the surface. The reduction in TA, DIC, and Ca²⁺ likely reflects the leaching of bicarbonate and carbonate ions from the tiles and their subsequent reaction with Ca²⁺ to form solid calcium carbonate. In contrast, under advective flow conditions in the flume experiments, continuous flows would have minimized precipitation on the tile surfaces. It is plausible that some of the calcium carbonate formed in the flumes was incorporated into the coral skeletons, contributing to the enhanced lateral and vertical skeletal growth observed.

It is possible that microbiome effects may also have changed the local chemistry of the tiles, which were unaccounted for in the current experiment. We only studied the interaction between substrate modifications and coral response, but this may not be a direct relationship, and there are additional factors to consider. Bacterial communities are sensitive to pH and alkalinity changes³⁵, and community structure varies with local physicochemical properties³⁶. Ocean acidification, while directly affecting corals through previously described means, also indirectly affects corals, because biofilms that aid in marine larval recruitment and settlement also decline in response to reducing pH³⁷. However, there is evidence that suggests the inverse is also true. On substrates with calcium carbonate additions, bacteria that induce settlement are more abundant³⁸, and healthy microbial films on artificial substrates may persist throughout seasons³⁹. Future research on AE substrates should include quantification of the microbiome to further elucidate the effect of substrates with the local ecosystem.

In summary, we consider this work to be a foundational proof-of-concept study establishing AE tile-based local pH modulation as a promising approach for coral restoration. Our results show that AE tile substrates can be an effective method of improving early coral recruit survivorship in laboratory settings. These AE tile substrates could be potentially attached to artificial reef base structures in the ocean to accelerate the development of coral reef biota on the structure. In addition, settlers grown in land-based spawning and propagation facilities could benefit from higher survivorship on AE tile substrates prior to being outplanted to natural reef surfaces. While field-based demonstrations are needed to demonstrate the scalability of AE tile substrates for reef restoration and hybrid reef deployments, our results provide promising indications that new technical advancements, such as those presented here, can aid in meeting the coral restoration challenge while also helping build coastal resilience. In ongoing and future work, we will examine the effects of oscillatory flow regimes, refine boundary-layer modeling, and further explore the mechanistic links between substrate chemistry and coral skeletal growth.

Tile fabrication

Three different chemical modifications to Portland limestone cement (Titan America, Medley, FL) were tested in addition to unmodified cement with 0.4 water-to-cement ratio. The chemical modifications utilized were 1% sodium bicarbonate (Ward’s Science, Product number: 470302-440, Rochester, NY), 1% sodium carbonate (Sigma-Aldrich, Product Number: 223530), or 2% sodium carbonate by weight to cement. Chemical modifications and cement were dry-mixed using a spatula for 4 min to achieve homogeneity, followed by mixing with water for an additional 4 min. All mixtures with added chemicals showed minor reductions in workability from visual observations. Tiles (3 × 3 × 1 cm) were cast from these four types of AE cement—no additions, 1% bicarbonate, 1% carbonate, and 2% carbonate—referred to as “chemistries”. Two different surface topographies were fabricated using 3D printed female molds—one where the surface of the tile contained no topographical modifications (a “flat” tile), and one where the tile featured a 4 × 4 grid of cylindrical, 3 mm-diameter, 1.86 mm deep divots spaced 8 mm from each other (divot center to divot center), centered on the tile (a “textured” tile; Supplementary Fig. 2b). Molds were filled with paste and placed on a vibration table to remove entrapped air, compacting the mix. The tiles were demolded 24 h after casting by drilling a hole in the bottom of the mold and applying compressed air. Subsequently, tiles were cured in a high relative humidity room (>95% relative humidity) at 23 °C for 28 d to ensure adequate hydration of cementitious materials.

After curing, tiles were transported from the University of Miami Coral Gables campus (Coral Gables, FL) to the Rosenstiel School of Marine, Atmospheric, and Earth Science campus (Virginia Key, FL). Tiles for diffusion experiments were stored in plastic bags until use. Tiles for advection and aging experiments were stored in a 38 L saltwater tank with a bubbler to ensure constant mixing. Tiles for biological experiments were further conditioned for 30 d in running seawater with live rock for colonization by CCA and biofilms that would provide attractive cues for O. faveolata larvae to encourage settlement^40,41.

Total alkalinity (TA), dissolved inorganic carbon (DIC) and calcium (Ca) measurements

Eight glass jars with wide-mouth screw-top lids were filled with 300 mL of seawater from Bear Cut inlet (25.733 °N, 80.158 °W). Each jar received one cement tile and a Teflon magnetic stir bar. All jars were stirred at the same speed setting. An initial water sample was collected on day 0 of the experiment. The salinity (32.6) and temperature (22.5 °C) of the seawater remained constant throughout the experiment. After 11 days, the tiles were removed from the jars and the seawater was filtered through 0.45 μm GF/F Whatman 47 mm glass fiber filters. Twenty mL of the filtrate was transferred to 50 mL Falcon screw-top centrifuge tubes. These samples were analyzed by ICP-OES for Ca²⁺, Mg²⁺, and Sr²⁺ using a slightly modified method as used by ref. ⁴². The remainder of the sample was analyzed for total alkalinity (TA) via potentiometric open-cell titration⁴³, and total dissolved inorganic carbon (DIC) was analyzed using a Dissolved Inorganic Carbon Analyzer (AS-C3, ApolloSciTech, USA).

Seawater samples were diluted 1:100 (v/v) with 4% HNO₃ (OMNITrace). The concentrations of Ca²⁺, Mg²⁺, and Sr²⁺ were analyzed using inductively coupled plasma optical emission spectroscopy (Varian VISTA-PRO CCD Simultaneous ICP-OES). Elemental concentrations were determined from calibration curves generated with external standards. Data were collected at the following wavelengths: 279.800 nm for Mg²⁺, 216.596 nm for Sr²⁺, and 317.933 nm for Ca²⁺. Seawater ion concentrations were quantified using four standards composed of IAPSO seawater diluted from 0.22 to 1.12 mmol kg⁻¹ for Mg²⁺, 0.37 to 1.85 μmol kg⁻¹ for Sr²⁺ and 0.04 to 0.21 mmol kg⁻¹ for Ca²⁺. Precision was determined with an IAPSO standard diluted 1:100, run with the samples. Precision reported as one standard deviation was ±0.8 mmol kg⁻¹ for Mg²⁺, ± 0.01 mmol kg⁻¹ Sr²⁺ and 0.2 mmol kg⁻¹ for Ca²⁺ (Supplementary Fig. 4 and Supplementary Table 6).

pH measurements in stagnant conditions: diffusive quantification

To characterize the chemical effect of the four different AE cements, pH was measured using UNISENSE microelectrodes (pH-N; 1.1 × 40 mm² needle electrode with external reference electrodes, Supplementary Fig. 5b). To estimate the diffusive potential of the AE tiles and confirm they can change pH, one randomly selected flat tile of each of the four chemistries was measured continuously over a 24 h period to quantify the diffusive capacity of the tiles to change pH in the water column. For measurements, the tile was transferred to 1000 mL of water. Microelectrodes were positioned at the surface and 2 mm above the surface of the tile. Sensors sampled at a frequency of 1 Hz for 24 h.

Simulating the hydrodynamic landscape

The hydrodynamic and chemical landscape quantifications in this experiment required the design of a flume that could: (1) achieve flows representative of natural coral reef environments^{22,23,24,25,44}, (2) be used in PIV experiments to quantify flow fields, and (3) be used in chemistry experiments to easily observe the effects of AE tiles without needing to make changes to the bulk water chemistry. For reasons (2) and (3), we utilize two configurations of a modular flume – a flow-through configuration and a recirculating configuration²¹. Both configurations feature the same experimental chamber where measurements are carried out. The flow-through configuration is open-ended, allowing for a constant flow of water through the experimental chamber and the recirculating configuration is fully enclosed, allowing for easy PIV measurements. A brief description of the modular flume, configurations, and flow conditions is provided here, with a more detailed description of the flume’s design, testing, and validation in Ruszczyk, et al.²¹

The experimental chamber in both the flow-through and recirculating configurations is comprised of a 30 × 10 × 14.5 cm transparent acrylic tank with inlet and outlet points made from 3 in, schedule-40 PVC pipe (Fig. 1b). An 8 cm (diameter) flow straightener with a hexagonal grid (hexagon circumscribed diameter = 1 cm) is mounted in the upstream direction of the experimental chamber. A 29 × 9.8 × 1.9 cm³ plate made of ¾ in PVC was made to hold a 1 × 3 configuration of three tiles along the length of the experimental chamber. Three indentations of dimensions 3.2 × 3.2 × 1.3 cm³ were made in the center of the plate, separated by 4.5 cm centered in the middle, to recess tiles such that the surface of the tiles was exposed to flow. The top of the tile plate rested flush with the bottom of the PVC inlet and outlet pipes. In the flow-through and recirculating flume configurations, the experimental chamber was filled such that the surface of the water was flush with the top of the inlet and outlet PVC, resulting in a volume of 2.28 L in the acrylic experimental chamber.

To create a flow-through configuration (Fig. 1c) where water is constantly turned over, the flumes were connected to a saltwater tap that supplied seawater filtered from the neighboring Biscayne Bay. Four flow-through flumes were installed on one tap line, branching off from a ¾ in PVC mainline. A ball valve at the top of each individual flume’s line allowed for volumetric flow control over each flume placed on the line. After the ball valve, the water flowed through a 1.9–19 L min⁻¹ (0.5–5 gal min⁻¹) clear in-line flow meter (Omega Engineering) to observe the incoming volumetric flow rate. The water then traveled through a 0.60 m section of 3 in PVC pipe before reaching the inlet of the experimental chamber. The outlet flow was controlled by adjusting a 3 in PVC elbow joint fixed to the outlet of the experimental chamber, and was set such that surface of the water was flush with the top of the inlet and outlet pipes of the experimental chamber.

The recirculating configuration was made by connecting the inlet and outlet pipes of the experimental chamber via a 3 in PVC pipe. An inverted tee joint at the back of the track allowed for a DC motor (Almencla RC Jet Boar Motor Engine Propellor) mounted to PVC pipe to be inserted. The motor was powered by an external power supply to achieve various flow rates. A laminar flow in the experimental chamber with a bulk flow speed of 1.20 cm s⁻¹ was generated (Reynolds number ≈ 800) in all flow-based experiments, matching reef-like conditions within a coral canopy^{22,23,24,25,45} (Fig. 1d).

Particle image velocimetry of tiles

The hydrodynamic landscape over flat and textured tiles was quantified in the recirculating flume using 2-dimensional planar PIV. To record the flow over different tile topographies, a tile was placed in the upstream position of a tile holder. The recirculating flume was seeded with 60 μm diameter polyamide particles (LaVision) and illuminated with a 1350 lumen LED illumination unit (LaVision) mounted above the tank creating a light sheet (Supplementary Fig. 6). The thickness of the light sheet was reduced to 1 cm by manually blocking the light entering the tank. The flow was set to a bulk flow velocity of 1.20 cm s⁻¹ using pre-established settings. Once the flow was steady, a 10 s video of the flow above the tile was recorded. Recording took place using an Imager CS2 5 Camera mounted with a L 60 mm focal length lens (F/2.8, 2:1 macro, LaVision) at 40 Hz. Flow quantification over tile topographies took place over an 8 × 6 cm² window parallel to flow bisecting the tile, resulting in a resolution of approximately 320 px cm⁻¹.

pH measurements in laminar conditions: advective quantification

To measure the chemical environment about the AE tiles in flow conditions, the pH at the tile surface for the four different chemistries was measured in a flow-through flume. One tile of each chemistry was transferred to a flow-through flume and placed in the upstream tile position. Sampling took place via UNISENSE pH microelectrodes positioned at the surface of the tile (Supplementary Fig. 5). Microelectrodes measured the pH at 1 Hz for 6 h to generate a time series of the influence of chemistries on the local pH at and above the tile surface.

pH measurements in laminar conditions: tile longevity

To estimate the longevity of the AE tiles’ enhancement effect, 12 flat cement tiles—three of each chemistry—was measured in an 84-day aging experiment. Tiles were stored in a 3 × 4 grid in a surplus flow-through flume filled with seawater and a bubbler to ensure mixing while experiments were not occurring—a storage flume. The storage flume was connected to the experimental flume such that when experiments occurred, all tiles were exposed to experimental flow conditions. The pH was measured at the surface of the tile and 3.26 ± 0.09 mm above the surface of the tile to capture changes at various heights within the laminar boundary layer. AE tiles were measured in the upstream slot of the tile holder for 90 min in a flow-through flume with a volumetric flow rate of 0.05 L s⁻¹ (1.20 ± 0.3 cm s⁻¹) for one of each of the four different chemistries daily for 30 d. After the first 30 d, experiments were performed every other day. The experimental tiles on a given day were randomly selected with no replacements until all 12 tiles had been recorded. After all tiles were measured, the tile position in the storage tank was shuffled, and all tiles were allowed to be selected for experiments again. This ensured a sample size of n = 3 for each of the four different chemistries.

Microelectrodes and reference electrodes were held in the experimental flume via dual micromanipulators (UNISENSE). Microelectrodes were positioned at the center of the downstream edge of the tile (Supplementary Fig. 5b–d) and the reference electrodes and temperature sensor were placed downstream in the experimental flume to not disturb the flow above the experimental tile. At the beginning of each day, flow was turned on to the flow-through system for the storage flume and the experimental flumes. Microelectrodes were positioned in the tank and allowed to acclimate for up to 30 min or until they provided a steady pH reading of ambient fluid for 10 min. After this acclimation period, a tile was added to the upstream tile position in the tile plate, and microelectrodes were deployed to record data. Microelectrodes ran for 90 min. After this time, the tile was removed, and the microelectrodes were allowed to reacclimate to ambient water for 10 min before the next tile was added. This was repeated such that one tile of each chemistry type was measured per day. Microelectrodes were recalibrated with buffer solutions of standard pH values once per week.

pH measurements in laminar conditions: topography quantification

The effect of topography on chemistry was tested by concurrently sampling the pH of textured tiles on the surface of the tile and within a divot on the same tile in a flow-through flume (Supplementary Fig. 5e). The first microelectrode was positioned to measure at the bottom of the divot, and the second was positioned within 10 mm of the divot such that it was touching the surface of the tile. Microelectrode positions were measured by photographing the profile and the top-down angle with known reference measurements. One tile of each of the four different chemistries was measured three times over the course of three days for 90 min each day to account for any daily variation in the tile.

Orbicella faveolata settlement on tiles

To test how coral recruits respond to AE cement tiles, O. faveolata larvae were settled and grown on AE tiles. Not all chemistries, or combinations of chemistry and topography type could be tested due to spatial limitations in the flumes. We excluded bicarbonate additions for the biology experiment because bicarbonate was not as effective at altering pH as carbonate, therefore maximizing the chance of observing any differences in growth related to chemistry. The three chemistry treatments were no additions, 1% carbonate additions, and 2% carbonate additions. Each chemistry included a set of flat tiles, and the no additions and 1% carbonate additions chemistries included a set of textured tiles, resulting in 5 different treatments of chemistry and topography used in the growth experiment.

Orbicella faveolata gametes were collected at Horseshoe Reef (25.1399 °N, 80.2946 °W) in Key Largo, FL, during a natural spawning event on August 7–8, 2023, and fertilized in vitro. Buoyant gamete bundles containing both egg and sperm were collected via tent collectors with a removable inverted 50 mL collection tube at the apex, as they were released from parent colonies. Sample jars from 8 to 10 colonies were brought to the boat and combined in a larger vessel where the bundles broke up and gametes from different parent colonies mixed for cross-fertilization. This fertilization solution was diluted by eye to be just slightly milky, approximating a sperm concentration between 10⁵ and 10⁷mL⁻¹ for optimal fertilization⁴⁶. After 1–1.5 h and return to the laboratory, excess sperm was rinsed away and resulting embryos were placed into three polystyrene settlement bins at a density of approximately 300 larvae per bin. Each bin contained a random assortment of 25 treatment tiles and 5 mL of crushed Hydrolithon boergesenii (CCA)⁴¹ in UV-sterilized, 1 μm filtered seawater. Settlement bins were kept in a temperature-controlled incubation room at 27 °C and daily 75% water changes were conducted while the larvae settled on the cement tiles over 5 weeks. Larvae were characterized as settled upon achieving full metamorphosis on the cement tiles, distinguished by a more circular, squat morphology in contrast to the elongated shape of freeform larvae. Settlers were counted and spatially mapped across all experimental tiles using a dissection microscope (Supplementary Fig. 7a). Tiles were labeled using waterproof paper tags affixed on the side of each tile using CorAffix gel coral glue. The three tiles of each of the five treatments with the highest settlement densities on the upper surface were selected for grow out in the flow-through flumes, totaling 180 coral recruits across 15 tiles.

Orbicella faveolata were inoculated with symbionts three times per week for 10 weeks. Inoculation occurred 5 weeks prior to relocation in flumes and the first 5 weeks residing in the flumes. Orbicella faveolata were inoculated with Durusdinium algal culture isolated from Key Largo, FL in the early 2000s. Durusdinium cultures were maintained in an incubator at 27 °C in f/2 media with a 14:10 light:dark photocycle under 40 μE light at a density of 10⁶ cells mL⁻¹. To inoculate O. faveolata, tiles were transferred from their respective tanks to polystyrene bins containing 3000 cells mL⁻¹ of Durusdinium algal culture for 6 h. Inoculation continued until each polyp on the tiles exhibited color indicative of symbiont saturation. During the inoculation period, tiles received 30 μmol s¹ m⁻² of photosynthetically available radiation (PAR) on a 12:12, light:dark photocycle.

Orbicella faveolata grow out in flumes

Tile position within the flume was randomized 3 times per week to reduce possible downstream effects from the AE tiles. Tile groups were assigned to different flumes twice per week to ensure randomness across flumes. Twice per week, tiles were relocated to 19 L tanks in a 28 °C, temperature-controlled bath for 45 min, and fed with Golden Pearls coral food. In traditional coral tanks, herbivores are co-reared with corals to control algae growth. Herbivores were not used in this experiment due to unsteadiness in flows around the additional organisms, and the threat that they would maneuver outside of the experimental chamber in the flow-through flumes. Instead of relying on herbivores to control algae growth, tiles were manually cleaned twice per week using a 25-gauge needle under a dissection microscope to ensure all algae were removed from the surface of the tiles.

To consistently identify corals throughout the experiment, settlement maps were made by taking an areal image of the full tile and updated weekly. Each recruit was assigned a unique ID, comprised of its tile number and a letter. Settlement maps provided confidence in individual coral measurements for the entirety of the experiment across multiple researchers collecting data.

Coral measurements occurred weekly. Survivorship was visually assessed by observing polyp activity during cleaning and measurement periods. For areal measurements, tiles were transferred to a 100 mL dish and imaged under a dissecting microscope. Images included a ruler in the background for scaling. For vertical growth measurements, corals were transferred to the recirculating flume with no flow present and imaged in a vertical profile using the PIV camera. Corals were located on the tile by mounting a calibration wand to a micromanipulator positioned above the tank. Referencing the settlement maps, the calibration wand was moved to the approximate position of the coral. The image was refined by adjusting the lens on the camera until the coral was in focus. The calibration wand was then physically adjusted to be in focus with the coral, ensuring that a known reference measurement was present in each image. As the image library was generated, old photos were cross-referenced to ensure consistent camera positioning throughout the experiment (Supplementary Fig. 7c, d).

Data processing

Chemistry data for tiles was processed using a custom-written MATLAB code. Chemical data collected from microelectrodes was first transformed from the logarithmic pH measurements to linear [H⁺] measurements by

allowing for processing and analysis using linear techniques. To account for slight differences in calibration between microelectrodes and daily differences in ambient pH in experiments, the change in [H⁺] (ΔH⁺) is used, rather than absolute [H⁺] measurements. To calculate ΔH⁺, the time series data was transformed such that the average [H⁺] in the 10 min calibration period of each time series is equal to zero. To account for sensor drift in the microelectrodes, data was detrended using a natural logarithmic function fitted to the calibration period of the data.

Flow data for tiles was processed in DaVis 11.0.0.196 software (LaVision). For PIV over tile topographies, a mask was applied to the bottom of the tank to exclude any areas where the tile may have protruded above the tile holder. PIV was performed using a bicubic interpolation to interpolate vectors, 5 × 5 denoising, a symmetrical shift correction mode. Pixels were interpolated using a spline interpolation mode with a window interrogation size of 16 × 16 px, a direct correlation algorithm was used between frames, and cell size weighted as a round cell. These parameters resulted in an average correlation value of 1.00 px.

Orbicella faveolata data was first scanned to exclude any chimeras (n = 5)—instances of two adjacent settlers fusing to form one large entity—from the analysis. Data was analyzed using three response variables: settlement preference, survivorship, and growth. Settlement preference was assessed by counting the total number of O. faveolata settlers per tile, parsed by settlement location (i.e., on the flat surface or within a divot). Settlement location preference on textured tiles was calculated as the percentage of larvae that settled in each location.

Two different measures of survivorship were calculated and analyzed. Total survivorship with respect to the tile was calculated each week as the ratio of all surviving O. faveolata on a given tile to the initial number of settlers on that tile, regardless of settlement location. Survivorship was partitioned for textured tiles into the ratio of surviving O. faveolata settlers on flat surface versus within divots, relative to the initial number of settlers in those locations – location survivorship. For flat tiles, only survivorship of flat surface settlers was calculated, as there were no divots present.

Polyp surface area was calculated by analyzing images in ImageJ software (version 1.54 g). The polyps in photos were outlined twice with the freehand tool and the area was calculated and measured twice. The radius of the coral, r, was estimated from this area, assuming the fitted shape was a circle. Vertical measurements were analyzed in DaVis 11.0.0.196 software. Vertical height was measured from the surface of the tile to the visible skeletal height, identifiable by a change in contrast in the image. Volumetric growth was calculated using the areal and profile images. Coral volume (mm³) was modeled as a cylinder, using the radius of the fitted circle calculated in areal measurements and the skeletal height of the coral as

where r is the radius (mm) of the areal measurements and h is the skeletal height (mm) of the coral (Fig. 1b).

The absolute change in coral size was calculated for all surviving coral with respect to vertical height (Δ vertical height), surface area (Δ surface area), and volume (Δ volume) as the change in size between the first and last days of the experiment.

Statistical analysis

Statistical analysis for chemical characterization of tiles was performed on ΔH⁺ to allow for analysis using linear techniques. Results reporting chemical differences are presented as ΔpH values, for ease of understanding. The influence of tiles on local pH via diffusion and advection was assessed using a fully crossed, two-way factorial ANOVA on one 6 h time series for each chemistry type for each flow condition. Aging data was modeled using linear regressions fitted to the average daily ΔH⁺ at the surface and above the surface for each chemistry. Aging data was further compared by aggregating the daily ΔH⁺ and comparing using a fully crossed, two-way factorial ANOVA using chemistry and sensor position as factors. The effect of topography on chemistry was assessed using a fully crossed, two-way factorial ANOVA on one 1.5 h time series for each chemistry.

Flows over surface topographies were compared via velocity and vorticity measurements at similar spatial coordinates. The data was visually checked for outliers resulting from PIV analysis, and outliers (greater than or less than 4 standard deviations from the mean) were removed from the dataset.

Orbicella faveolata data were analyzed with respect to tile (n = 15) using a series of one-way ANOVAs to mimic a factorial ANOVA (chemistry, topography, and an interactive effect of chemistry and topography). A full factorial ANOVA would not be appropriate for the data as not all combinations of chemistry and topography were present in the data (there were no 2% carbonate textured tiles), and several variables required non-parametric analyses. On the textured tiles, to investigate location-based effects (i.e., on the surface or within a divot) on settlement and survivorship, the percentage of settlers based on location (settlement with respect to location) for the textured tiles was calculated as the amount settled in each location per tile (i.e., on the exposed surface or in a divot) and was analyzed with settlement location as a factor.

Settlement preference was analyzed by the initial number of settlers (0 d). Total survivorship was analyzed at the conclusion of the experiment (139 d). All percentage data (settlement location preference ratio and survivorship) were transformed using the angular transformation, asin(sqrt(Y)), prior to analysis⁴⁷. Growth data was filtered to only include colonies which survived the duration of the experiment to remove any selective bias in growth data from smaller corals dying. Growth analysis only included measurements for corals grown on the flat surfaces of tiles, as vertical measurements could not be obtained for corals which settled in the divots of textured tiles. Change in growth data was calculated as the change in size in either the vertical, areal, or volumetric size of the coral between the beginning and end of the experiment.

Statistical analysis was performed in JMP 16.0 (SAS Institute 2023) and R 4.4.1 (R Core Team 2022). Assumptions for normality and heteroscedasticity were tested on all the data using Anderson-Darling tests and Levene’s tests (Supplementary Tables 7 and 8). Parametric ANOVAs, t-tests, and Tukey’s HSD tests or non-parametric Kruskal-Wallis, Wilcoxon Rank, and Steel-Dwass all pairs tests were performed using the DHARMa package⁴⁸ in R. All data is presented as mean ± standard error unless otherwise stated.

Reporting summary

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