Oyster farming acts as a marine carbon dioxide removal (mCDR) hotspot for climate change mitigation

Significance

Oyster farming not only supports global food security but also acts as a significant marine carbon dioxide removal (mCDR) mechanism by promoting primary production and organic carbon sedimentation. Our study reveals that oyster-driven organic carbon production sequesters 2.39 times more carbon than is harvested during shell formation, enhancing atmospheric CO₂ uptake and mitigating ocean acidification. These findings highlight oyster farming as a scalable, nature-based solution to climate change, offering a dual benefit of carbon sequestration and sustainable food production. This research provides critical insights for integrating bivalve farming into global carbon trading frameworks and blue carbon strategies.

Abstract

Bivalve farming, a vital component of global aquaculture, has been proposed as a potential marine carbon dioxide removal (mCDR) strategy, yet its role remains contentious. Using field mesocosms, we demonstrate that oyster filter-feeding enhances mCDR by accelerating the formation of particulate and dissolved organic carbon in the water column and promoting organic carbon deposition in sediments. This process shifts the water column toward a more autotrophic and alkaline state, effectively sequestering CO₂ from the atmosphere. Over the full culture period, the net carbon sequestered by oyster-driven organic carbon production is 2.39 times greater than the CO₂ sequestered in oyster shells. These findings position oyster farming as a scalable, nature-based solution for climate change mitigation, offering dual benefits of carbon sequestration and enhanced food security. Our results underscore the potential of oyster farming to address global challenges such as rising food demand and ocean acidification, making it a critical component of sustainable marine resource management.

As climate change and ocean acidification accelerate, the search for effective carbon sequestration strategies has become increasingly urgent (1, 2). Marine ecosystems, particularly bivalve aquaculture, present promising opportunities for carbon removal while addressing global food security (3–5). In 2022, global aquaculture produced 18.7 million tons of marine mollusks, predominantly bivalves (6), underscoring its significant role in global food security and its contributions to conserving arable land and freshwater resources (7, 8). Beyond these benefits, bivalve farming has been proposed as a potential marine carbon dioxide removal (mCDR) due to the substantial carbon sequestered in their shells, which may accelerate CO₂ invasion from the atmosphere into the ocean (3, 4, 9, 10).

However, this view remains contentious. Many studies suggest that bivalve farming may act as a net source of CO₂ to the atmosphere and coastal acidification due to the release of CO₂ during shell calcification, fueling debate over its role in carbon sequestration (11–15). These contrasting perspectives have led to the exclusion of bivalve farming from mCDR, blue carbon frameworks, and global carbon trading schemes (4, 16, 17). This ambiguity underscores the need for a clearer understanding of the carbon dynamics associated with bivalve farming to accurately assess its potential contributions to mCDR.

The carbon dynamics associated with bivalve farming are intricate. Shell formation (calcification) and respiration release CO₂, potentially elevating local partial pressure of carbon dioxide (pCO₂) and reducing atmospheric CO₂ uptake or even promoting CO₂ release to the atmosphere (14, 18). While these processes are well documented, the broader ecological role of bivalves—particularly their influences on primary production and carbon cycling at the ecosystem scale—remains less understood (16, 19, 20). Moreover, previous studies lack the holistic carbon flux measurements necessary to comprehensively evaluate the net carbon impact of bivalve farming systems (21–24).

To address these gaps, we conducted a field mesocosm study with Pacific oyster (Crassostrea gigas), monitoring carbon budget changes throughout the oyster growth period. Our approach included all measurable carbon species, air–sea CO₂ fluxes, concentrations of chlorophyll-a (Chl-a). Coupling these results with remote sensing data, we provide comprehensive insights into how oyster farming enhances the marine biological pump—a biologically driven process that sequesters carbon from the atmosphere to the ocean (25)—and contributes to relatively long-term carbon sequestration. This research highlights the dual benefits of oyster farming: bolstering global food security (7, 8) while serving as an effective method for mCDR and mitigating ocean acidification (OA)—a decline in ocean pH caused by the absorption of anthropogenic CO₂ (26). These benefits align with global sustainability goals to combat climate change and enhance food security (27).

Results and Discussion

Oyster Farming Enhances Autotrophic State.

The mesocosms were designed to mimics natural systems, incorporating a continuous nutrient supply while maintaining closed horizontal boundaries, allowing for CO₂ exchange at the water surface and exposure at the bottom (Experimental Mesocosms). This setup enabled comprehensive monitoring of carbon parameters throughout the oyster growth period under various stocking densities, referring to the stocking densities of bivalves in practice (Experimental Design). It facilitated understanding carbonate dynamics over time and allowed for the calculation of the carbon budget, establishing correlations between oyster density, phytoplankton productivity, and fluxes of multiple carbon species at the ecosystem scale.

Over the 120-day experimental period, the field mesocosms (SI Appendix, Fig. S1) experienced water temperature variation from 20.7 to 29.0 °C and salinity fluctuation between 29.0 and 29.8 (Fig. 1A). The initial dissolved inorganic carbon (DIC) and total alkalinity (TA) of experimental waters were 2455 ± 19 and 2750 ± 27 μmol kg⁻¹, respectively. The observed patterns of Chl-a offered insights (Fig. 1B). During the initial 20 d (early stage), Chl-a increased most significantly in the control group (greatest), followed by progressively lower increases with higher oyster density in the order of 0.5, 1, 2, and 4 ind. m⁻² (Oy0.5, Oy1.0, Oy2.0, and Oy4.0). This pattern highlights a strong top–down control of oyster filter-feeding on phytoplankton growth. The variation in primary productivity (PP, gO₂ m⁻²d⁻¹) closely mirrored the Chl-a trends (Fig. 1C), indicating a biomass-dependent relationship and confirming the reliability of the experimental data.

Fig. 1.

Temporal patterns of environmental parameters of oyster farming ecosystems. (A) Temperature and salinity in each experimental system. (B) Chlorophyll a concentration (Chl-a) in each system. (C) Primary productions (PP) in each system (see PP in Materials and Methods). (D) The ratios of primary production to respiration (P/R) in each system. (E) Dissolved inorganic carbon (DIC) in each system. (F) Total alkalinity (TA) in each system. (G) The CO₂ fluxes over air–water interface, calculated based on the carbonate systems (F_CO2-C, see Materials and Methods) of each system. (H) The scatter plot of ΔTA against ΔDIC (relative to the control group) of oyster farming systems. (I) pH in each system. (J) Net community production (NCP) of each system, calculated based on carbonate systems (see NCP in Materials and Methods). (K) Net community calcification (NCC) of each system, calculated based on carbonate systems (see NCC in Materials and Methods). (L) Net carbon yields from harvest products (C_{net yield}) in oyster farming systems. Ctrl, the control system without oyster; Oy0.5, the oyster farming system with 0.5 ind. m⁻²; Oy1.0, the oyster farming system with1.0 ind. m⁻²; Oy2.0, the oyster farming system with 2.0 ind. m⁻²; Oy4.0, the oyster farming system with 4.0 ind. m⁻². All treatments have three replicates. Error bars are ± SEM P < 0.05. Comprehensive measurements demonstrate oyster farming system is autotrophic regardless stocking density.

In the middle stage (40 to 60 d), the Chl-a increment for oyster-containing groups was significantly higher than that of the control group. The highest Chl-a increment occurred at a stocking density of 2 ind. m⁻², followed by the densities of 1, 0.5, and 4 ind. m⁻². These results suggest that with appropriate stocking density, oyster can significantly stimulate phytoplankton growth. A similar phenomenon was observed in systems with relatively low-density hard clams (Mercenaria mercenaria) farming, where primary production doubled (28).

In the late stage of culture period, phytoplankton biomass remained relatively unchanged (Fig. 1B), indicating that a balance was reached between phytoplankton production and consumption. Notably, the 4 ind. m⁻² stocking density case consistently exhibited the lowest Chl-a and PP among all treatments, significantly lower than the control group. This result suggests that high-density stocking effectively suppressed phytoplankton biomass, resulting in “clarity effect” (see discussions below).

Our experimental results demonstrate that filter-feeding oysters can simultaneously suppress the phytoplankton growth through feeding (“feeding effect”) and enhance productivity by excreting nutrients (“nutrient effect”) (19, 20, 28). The observed patterns of Chl-a and PP variations under different stocking densities clearly illustrate the interplay between bottom–up and top–down controls on primary producers. These findings suggest that reasonable stocking densities promote PP in oyster farming ecosystems, while overstocking reduces PP by depleting phytoplankton biomass.

The ratio of PP to respiration (P/R) serves as an indicator of an ecosystem’s metabolic state (29). In our mesocosm experiments, the P/R ratio was consistently >1, indicating that all systems, including the control, were autotrophic. However, the Oy1.0 and Oy2.0 treatments significantly enhanced autotrophic states compared to the control group (Fig. 1D), suggesting that moderate oyster stocking increases productivity and autotrophy. By contrast, overstocking (e.g., Oy4.0) reduced P/R values, although the system remained autotrophic.

As stocking density increased, both DIC and TA decreased in our study (Fig. 1 E and F). The CO₂ flux (negative values indicate CO₂ invasion from the air to the waters) exhibited a similar pattern to DIC (Fig. 1G). A positive correlation (SI Appendix, Fig. S2) was observed between CO₂ influxes (F_CO2-C) calculated using the carbonate system approach and those measured directly by using a chamber (F_CO2-D, see Materials and Methods). The quantities of F_CO2-C were consistently lower than those of F_CO2-D in the present study, aligning with observations from previous studies elsewhere (30, 31). Nevertheless, these two independent, positive correlated measurements provide robust evidence that oyster farming promotes an autotrophic state across all stocking densities.

It is worthwhile to note that the clarity effect and overstocking phenomenon in high-density systems (Oy4.0) may not occur in culture systems of other species. Unlike filter-feeding oysters, gastropod shellfish (e.g., abalone and mud snail) exhibit fundamentally strategies that do not produce the phytoplankton-clearing effects characteristic of bivalve systems. While these gastropods lack the direct top–down control on phytoplankton populations seen in oyster farms, their sediment bioturbation activities may significantly influence ecosystem autotrophy through alternative pathways (32). The carbon dynamics of gastropod shellfish farming ecosystems need to be studied in depth in the future.

Carbonate Systems Modulated by Oyster Growth and Other Processes.

The process of biocalcification removes both TA and DIC from seawater in a 2:1 mol ratio (18) (see the dashed line in Fig. 1H). In contrast, the production of 1 mol of organic matter through photosynthesis decreases DIC by 1 mol and increases TA by 0.16 mol(18). Beyond biocalcification, oyster respiration, primary production, microbes’ respiration, and atmospheric CO₂ invasion collectively modulate the carbonate system in oyster farming environments.

Generally, in an aquatic system, when TA decreases more rapidly than DIC (ΔTA/ΔDIC < or ≈1, see the 1:1 line in Fig. 1H), water pCO₂ rises, reducing the capacity of surface water to absorb atmospheric CO₂ (14). However, in all stocking densities across the entire culture period in this study, DIC consistently decreased more rapidly than TA (see Fig. 1H, ΔTA/ΔDIC < or ≈1 except the late stage of Oy4.0). This led to an increase in pH (Fig. 1I) or alkaline state, favoring enhanced atmospheric CO₂ absorption. Conversely, the degradation of 1 mol of organic matter through respiration increases DIC by 1 mol and decreases TA by 0.16 mol (18). When coupled with calcification, the balance between production and respiration largely determines the TA/DIC slope, which in turn regulates the water’s CO₂ absorption capacity.

As previously noted, measured P/R ratios were consistently >1.2 across all our scenarios (Fig. 1D), indicating that carbonate dynamics in the mesocosms were primarily governed by biocalcification and productivity. Net community production (NCP), calculated from observed DIC and TA, was higher in oyster-containing systems compared to the control (Fig. 1J). Positive NCP reduces DIC while increasing TA in a 0.16 ratio (33). Furthermore, NCP values significantly exceeded the net community calcification (NCC, Fig. 1K) in oyster farming systems, further suggesting that oyster farming promotes the water column toward a more autotrophic state, enhancing CO₂ uptake from the atmosphere.

In the late stage of the culture period, alterations in carbonate system parameters (pCO₂, DIC, and pH change coherently) became evident, while changes in TA were less pronounced (Fig. 1F). This indicates that NCC was less influenced. A dramatic slope change in the ΔTA-ΔDIC scatter plot (Fig. 1H) was observed for late-stage samples from Oy4.0. Given that Oy4.0 exhibited the lowest P/R ratios (Fig. 1D), calcification effects become more prominent during the late stage in the ΔTA-ΔDIC scatter plot. This shift may have been driven by decreased temperature, which potentially promoted the CO₂ influx and lowered the P/R ratio by altering microbial metabolic activity without significantly impacting biocalcification.

During the final stage, CO₂ flux from the atmosphere decreased, as calcification and microbial respiration both released CO₂. Despite this, pH continued to increase, aligning with changes in other carbonate system parameters. These observations suggest that the interplay between oysters feeding and metabolism, autotrophic phytoplankton growth, and heterotrophic respiration plays a critical role in modulating the carbonate system and the CO₂ adsorption capacity of oyster farming ecosystems.

Closing the Carbon Budget of the Oyster Farming System.

The mesocosm design allowed a comprehensive closure of the carbon budget postharvest. The net carbon yield (C_{net yield}) of the oyster farming systems ranged from 1.55 ± 0.05 to 8.02 ± 0.83 gC m⁻² (Fig. 1K), with the higher the stocking density leading to significantly increased C_{net yield} (P < 0.05). On average, shell carbon accounted for 91% the total carbon yield (including both tissue and shell), a proportion significantly higher than that of other bivalves, such as clams, mussels, and scallops (4, 34, 35).

During the experiment, the changes in particulate organic carbon (ΔPOC) in the water column were closely associated with oyster stocking density (Fig. 2B and SI Appendix, Fig. S3). The ΔPOC in the Oy1.0 and Oy2.0 systems were significantly higher than that in the control system (P < 0.05), whereas Oy4.0 exhibited a “clarity effect” indicated by suppressed phytoplankton biomass. Similarly, the increments of dissolved organic carbon (ΔDOC; Fig. 2C and SI Appendix, Fig. S4) in water column followed trends comparable to those of ΔPOC. As DOC in water column is partially derived from excretion of phytoplankton (36), this lower DOC observed in Oy4.0 case can be attributed to phytoplankton biomass suppression.

Fig. 2.

Carbon sequestration potentials of experimental ecosystems. (A) Sedimentation of organic carbon (SOC, see Materials and Methods) in each system. (B) Changes of dissolved inorganic carbon (ΔDIC) in each system. (C) Increment of particulate organic carbon (ΔPOC) in each system. (D) Increment of dissolved organic carbon (ΔDOC) in each system. (E) Carbon influx metric (CIM) in each system. Please refer to Fig. 1 for the acronym of each treatment. All treatments have three replicates. The values with different letters are significantly different from each other (P < 0.05). Oyster farming significantly promotes carbon sequestration.

The net carbon yield (C_{net yield} = C_harv − C_seed) (Fig. 1L) reduced the total carbon content (TC) in the aquatic system. At the same time, sedimentation of organic carbon (SOC), approximately representing the increment in sediment organic carbon in the present studies, increased significantly with higher stocking densities (P < 0.05) (Fig. 2A), suggesting that oyster filter-feeding enhances organic carbon deposition. In the Oy4.0 treatment, the increase in SOC was particularly pronounced, as both POC and DOC (Fig. 2 B and C) were reduced. This indicates intensified grazing and respiration processes under high-density conditions. Clearly, SOC contributes to reducing TC in the water column, thereby enhancing the potential for atmospheric CO₂ influx.

Based on the law of mass conservation, the Carbon influx metric (CIM) for a holistic bivalve farming system can be calculated as

$$\begin{array}{c}CIM=(C_{\mathit{h}\mathit{a}\mathit{r}\mathit{v}}-C_{\mathit{seed}})+(C_{\mathit{biodep}}-C_{\mathit{decomp}})+\mathrm{\Delta }TC\hfill \\ =C_{\mathit{netyield}}+SOC+(\mathrm{\Delta }DIC+\mathrm{\Delta }POC+\mathrm{\Delta }DOC).\hfill \end{array}$$

When all carbon species were considered, the CIMs of oyster farming systems ranged from 18.8 to 27.5 gC m⁻², significantly higher than that of the control (Fig. 2E). The CIM variations among experimental ecosystems mirrored the patterns observed in CO₂ influx (F_CO2-C; Fig. 1G), further supporting the role of oyster farming ecosystems as biological carbon pumps and carbon sinks.

Overall, the carbon budget of our mesocosm systems can be reasonably closed. These findings highlight oyster farming as an effective accelerator of biological pump, enhancing the conversion of CO₂ into organic carbon, ultimately reducing water CO₂ levels and promoting atmospheric CO₂ influx. Over the full culture period, the net carbon sequestered by oyster-driven organic production (CIM_F) was about 2.39 and 1.39 times greater than the net carbon yield from farmed oyster shells and the carbon released from NCC, respectively (Table 1).

Table 1.

The carbon budgets of oyster farming systems (mgC m⁻² d⁻¹)

Treatments	Oy0.5	Oy1.0	Oy2.0	Oy4.0
Cnet shell yield	11.83 ± 0.51d	22.32 ± 3.49c	41.98 ± 1.48b	60.48 ± 6.59a
C released from NCC	16.91 ± 4.82d	46.75 ± 3.81c	76.49 ± 5.89b	101.72 ± 4.37a
SOCF	14.57 ± 2.79c	24.30 ± 8.38c	48.54 ± 5.02b	97.77 ± 13.99a
CIMF	32.75 ± 7.25a	58.75 ± 9.67b	105.75 ± 3.92c	98.58 ± 10.67c
SOCF/Cnet shell yield	1.23	1.13	1.16	1.64
CIMF/Cnet shell yield	2.77	2.63	2.52	1.63
CIMF/C released from NCC	1.94	1.27	1.38	0.97
POC+DOC+SOC/C released from NCC	1.56	1.31	1.34	0.77

Notes: C_{net shell yield}, net carbon yield from farmed oyster shells. SOC_F, relative sedimentation of organic carbon compared to the control. CIM_F, relative carbon influx metric compared to the control. NCC, net community calcification. The values with different lowercase letters in the same line are significantly different from each other (P < 0.05). Ctrl, the control system without oyster; Oy0.5, the oyster farming system with 0.5 ind. m⁻²; Oy1.0, the oyster farming system with 1.0 ind. m⁻²; Oy2.0, the oyster farming system with 2.0 ind. m⁻²; Oy4.0, the oyster farming system with 4.0 ind. m⁻². All treatments have three replicates.

Our results also underscore the critical influence of stocking density. Overstocking, as observed in Oy4.0 system, as consumed more POC and produced less DOC compared to other stocking densities, ultimately constraining the pumping efficiency, as evidenced by the CIM values (Fig. 2E).

Note that, our calculation represents the instant budget state. A portion of SOC might be respired to overrate the sequestration potential. It is difficult to conduct the carbon budget of an entire bivalve farming ecosystem in coastal sea. The simple way to get the CIM of a bivalve farm is to quantify the correlation between the productions of a type of bivalve species and their CIM with experimental mesocosms, and then use the relationship to estimate the CIM.

Top–Down Phenomenon in Field Oyster Farms.

Remote sensing data analysis was performed to examine the horizontal distribution of Chl-a in two of the largest nearshore oyster farms in China (Fig. 3A). The farming areas, located in Rongcheng and Rushan, Shandong Province, spanned 40,000 and 20,000 hm², respectively, with annual oyster productions of 0.5 and 0.38 million tons.

Fig. 3.

The Chl-a concentration distribution detected by satellite remote sensing and in situ sampling over two typical nearshore oyster farm areas in China. (A) Location of the two oyster farms. (B) Chl-a distribution in August average 2021 over Rongcheng nearshore oyster farm area. (C) Chl-a distribution in August average 2021 over Rushan nearshore oyster farm area. (D) Chl-a and dissolved oxygen content (DO) at various in-situ sampling sites in Rushan on 23 August 2024.

Remote sensing images from August of 2021 (Fig. 3 B and C) revealed significantly reduced Chl-a levels within the farming areas, indicating a strong top–down control exerted by oyster filter-feeding on phytoplankton biomass. This finding is consistent with the clarity effect observed in experimental mesocosms and underscores the large-scale ecological impact of oyster farming on phytoplankton dynamics.

On 23 August 2024, Chl-a concentration and dissolved oxygen content (DO) were measured across a transect of the Rushan oyster farm (Fig. 3C). Results revealed a significant reduction in both Chl-a and DO content within the farm site, accompanied by a notable increase in Chl-a concentrations surrounding the farm site (Fig. 3D). This finding aligns with the Chl a-density correlation patterns observed in our mesocosms and the distribution patterns derived from satellite remote sensing data. However, the observed DO reduction (below 100% saturation) within the Rushan oyster farm suggests also a significant respiration, consistent with the Oy4.0 cases.

The tidal current moves the water masses along the coastline where the oyster farms are located, bringing natural food for farmed bivalves from areas much larger than the farm site (37, 38). Similarly, the ecological influence of these bivalves also extends far beyond the farm site (19, 23, 24). A notable increase in Chl-a concentrations in waters surrounding the farm site compared to adjacent normal sea areas reflects enhanced primary production from regenerated nutrients (39). These observations further corroborate our findings that oyster farming ecosystem acts as effective accelerators of the biological pump.

The net carbon sequestrated in bivalve shells (C_{net shell yield}) has been suggested as a carbon trading metric (3–5, 9). However, this metric might overlook the substantial biodeposition effect of farmed bivalves (21, 22). Our results show that the introduction of oysters into an aquatic system effectively enhances its biological pump function, that is, transfer DIC into SOC, POC, DOC, and the carbon in shells. The total carbon converted was about 2.39 times the C_{net shell yield} (Table 1), indicating that using C_{net shell yield} as a carbon trading metric indeed overlooks the overall effect of oyster farming.

Summary.

Nutrient fertilization, artificial upwelling, seaweed cultivation, ocean alkalinity enhancement, electrochemical engineering, and recovery of marine ecosystems have been recognized as mCDR approaches (17). Our mesocosm experiments reveal that oyster-driven organic carbon production (POC + DOC) and sedimentation (SOC) exceed CO₂ release by a factor of 1.39 resulting in a negative carbon emission (Table 1). The entire culture system was moved toward a more autotrophic and alkaline state, effectively sequestering CO₂ from the atmosphere.

However, there is a spatial heterogeneity in Chl-a distribution and CO₂ fluxes across field oyster farming ecosystem (Fig. 4), driven by the trade-offs among three ecophysiological processes: photosynthesis, calcification, and respiration in the oyster farming ecosystem. At the individual scale, oyster biocalcification and respiration dominate the carbon dynamics. However, at the ecosystem scale, photosynthesis is significantly enhanced and emerges as the dominant process.

Fig. 4.

CO₂ flux pattern over oyster farming ecosystem. There is a spatial heterogeneity of CO₂ fluxes over an entire oyster farming ecosystem: Oyster farming exhibits the function of CO₂ efflux compared to adjacent normal sea areas at individual scale. However, at the whole ecosystem scale oyster farming demonstrates a CO₂ influx just as showing by our mesocosms, simulating both the farm site area plus the broader area affected by the farmed oysters in coastal sea.

Reasonably, optimal stocking densities allow oyster farming systems to function as an efficient biological pump, driving CO₂ from the atmosphere into the ocean by promoting primary production and organic carbon sedimentation. In some farming areas with high turbidity and significant inflow of allochthonous organic matter, primary productivity might not be enhanced due to light limitation; while respiration and calcification turn out to be the major processes to drive the carbonate system. In this case, oysters’ filter feeding can still act as an efficient carbon pump to transfer organic carbon into SOC. We also prove that through moving the culture system toward an autotrophic state, oyster farming promotes water pH for OA mitigation. Therefore, bivalve farming should be included in mCDR or blue carbon frameworks and global carbon trading schemes.

From a life cycle assessment (LCA) perspective, bivalve farming can also be considered a CDR approach. For instance, carbon emissions from farming operations of Manila clam (Ruditapes philippinarum) and Mediterranean mussel (Mytilus galloprovincialis) have been shown to account for 11.8% and 37.7% of the carbon sequestered within their shells, respectively (9, 40). It follows that bivalve farming can achieve net carbon sequestration or CDR when viewed holistically and in terms of LCA.

China, the world’s leading producer of aquaculture bivalves, produced 15.1 million metric tons (Mt) of mariculture bivalves in 2022 (41). This production is equivalent to approximately 4.04 Mt of shell weight and about 1.89 Mt of carbon in shells (42). Notably, our findings suggest that bivalve farming sequestered approximately 4.09 Mt of carbon in China, with more than half of this carbon stored as SOC (Table 1). These results emphasize the substantial contribution of oyster farming to carbon sequestration, far exceeding the shell-based metric alone.

Globally, 18.7 Mt of mariculture mollusks, predominantly bivalves, were produced in 2022 (6). Looking ahead, over 1.5 million km² of sea areas could potentially be developed for bivalve farming worldwide (43). By 2050, economically viable mariculture bivalve production could reach an estimated 80.5 Mt (44). Considering its effectiveness as a biological pump accelerator and its role in carbon cycling, the global significance of bivalve farming as a mCDR and OA mitigation strategy should not be overlooked.

Materials and Methods

Experimental Mesocosms.

Manipulations of ecosystems through the construction of mesocosms are the most informative experiments to disentangle how ecosystems work (45, 46). Fifteen mesocosms (water volume of 10 × 10 × 1.5 m³ each) were established in a large seawater pond (100 × 150 m²) at a mariculture farm near the southern coast of Bohai Bay, China (37.0736°−37.0747°N, 19.4819°−19.4836°E, SI Appendix, Fig. S1). The mesocosms were lined with waterproof polyethylene-coated (PE) fabric material.

Prior to the experiment, seawater was introduced into the pond and then submersed U-shaped siphons were used to introduce seawater from the pond into mesocosms. One end of the siphon was covered with an 0.18 mm-mesh screen to prevent seston from entering the mesocosms. Throughout the experiment, water exchange between the mesocosms and the external pond environment was minimal, with only minor additions of seawater to compensate for evaporation and leakage. To prevent epiphytic organisms, the inner walls of the mesocosms were cleaned every other day.

Experimental Organisms.

The Pacific oysters (Crassostrea gigas) used in this study were procured from a shellfish farm in Shandong Province, China. The oysters were selected for uniform size (24.48 ± 0.08), normal morphology, physique integrity, and robust health to minimize variability (SI Appendix, Table S1) since carbon sequestration capacity depends on both oyster size and stocking density, as demonstrated by our density-gradient experiments (0.5 to 4.0 ind. m⁻²). Prior to the experiment, the oysters were acclimated in the experimental pond for 2 wk.

Experimental Design.

There were five treatments in the experiment: a control group without oysters (Ctrl) and four oyster density treatments (0.5 ind. m⁻², denoted by Oy0.5; 1.0 ind. m⁻², Oy1.0; 2.0 ind. m⁻², Oy2.0; 4.0 ind. m⁻², Oy4.0). There were three replicates for each treatment. The oysters were randomly assigned to the different treatment groups according to the specified densities (SI Appendix, Table S1). Artificial fertilizers were applied to all groups to meet the nutritional needs of the aquatic ecosystem (47, 48). Specifically, after each sampling event, 1.0 mg L⁻¹ of nitrogen (ammonium chloride) and 0.25 mg L⁻¹ of phosphorus (potassium dihydrogen phosphate) were apportioned to perforated PE slow-release containers in the mesocosms.

In aquaculture practice, the open sea area of Zhangzidao, Liaoning province, China, produced 40000t of Japanese scallop Patinopecten yessoensis in 2005 (49). The bottom sowing density of the scallop was 9.6 ind. m⁻² or 0.63 ind. m⁻³. It takes 8.45 d for the farming waters to be flushed, so the water volume maintaining the scallops in the area is around that of Oy0.5 in the present experiment. As another case, Sanggou Bay, Shandong Province, China, is a semienclosed bay with dense raft-culture of oysters. The density of the oysters was 59 ind. m⁻² or 3.9 ind. m⁻³ in 2014 (50). The time of water half-exchange in the bay is 33 d(51). The water volume maintaining the oysters in the area is between those of Oy2.0 and Oy4.0 in the present experiment.

Sample Collection and Analysis.

The experiment was conducted from June 8 to October 6, 2023, spanning a total of 120 d. Initial sampling was performed on the first day of the experiment, followed by subsequent samplings at 20-day intervals. All samples were collected between 7:00 and 10:00 am.

Water Temperature, Salinity, and pH.

The temperature (accuracy ± 0.2 °C) and salinity (accuracy ± 0.1 ppt) of the surface water (0.2 m depth) within each mesocosm were measured using a calibrated multiparameter water quality meter (YSI ProPlus, USA). The pH of surface water was measured using a PHC101 pH electrode (accuracy ± 0.02). The electrode was calibrated using pH 4.00, 6.68, and 9.18 buffer solutions before measurement.

Chlorophyll-a (Chl-a).

The concentrations of phytoplankton Chl-a in the mesocosms were determined following the method described by Lee et al (52).

Primary productivity (PP).

A method was conducted using the light and dark bottle method (53). The DO was measured using a portable meter (Multi 3510IDS, WTW Company, Germany).

CO₂ Fluxes Over Air–Water Interface.

The CO₂ fluxes over the air–water interface of the mesocosms were measured using an automated multichannel portable CO₂ concentration analyzer (PS-9000, Beijing LICA United Technology Limited, China) (30, 31).

Carbon Pools and TA in Water Column.

Surface water samples were collected and filtered using pretreated (acid-washed and burned at 450 °C for 4 h) 0.7 μm GF/F filter membranes (Whatman International Ltd., Maidstone, England). One portion of the filtrate was used for on-site TA measurement, while another portion was stored in 50 mL centrifuge tubes with added saturated HgCl₂ solution for DOC and DIC analysis (54, 55). The membranes were preserved for POC analysis (56). TA was determined using the acid–base titration method (accuracy ±4 μmol kg⁻¹). DIC and DOC were measured with a TOC analyzer (multi N/C 2100, Jena, Germany). TOC analyzers measure inorganic carbon (IC) by acidifying the sample and detecting released CO₂ via NDIR, while organic carbon (OC) is determined by high-temperature combustion oxidation of the acidified residue followed by NDIR detection of CO₂. The calibration curves for DIC and DOC were established using seven standard solutions at varying concentrations, and the measurement accuracy was ±3 μmol kg⁻¹. POC was measured with an elemental analyzer (Vario EL III, Elementar, Germany). The POC was determined using an elemental analyzer (Elementar vario EL III, Elementar Company, Germany) and was estimated based on the analysis of the standard sample acetanilide, with an accuracy of 0.05%.

Net Community Production (NCP) and Net Community Calcification (NCC).

The NCP and NCC was calculated as follows (57, 58):

$$\mathrm{\Delta }DIC=-NCP-NCC+F_{CO2}/h$$

$$\mathrm{\Delta }TA=0.16NCP-2\times NCC,$$

where ΔDIC represents the change in dissolved inorganic carbon between the sampling intervals; ΔTA denotes the change in total alkalinity between the sampling intervals; F_CO2 is the mean value of CO₂ fluxes over air–water interface between the sampling intervals, and h indicates the water depth.

Net Carbon Yield From Oysters.

The carbon contents in oyster shells and soft tissues were analyzed using an elemental analyzer (59). Net carbon yields in oyster shells and soft tissue were calculated based on biomass growth, carbon content, and dry-to-wet weight ratios (60).

Sedimentation of Organic Carbon.

At the start of the experiment, cylindrical sediment traps (21.4 cm in diameter and 21 cm in height) were placed on the bottom in each mesocosm to collect sediment (61). At the end of the experiment, the sediment traps were retrieved. The sediment carbon contents were analyzed using an elemental analyzer to determine the sedimentation of organic carbon (SOC) over the experimental period.

Spatial Distribution of Chl-a Concentration At Nearshore Oyster Farming Areas.

Based on the Google Earth Engine platform, the Copernicus Sentine-2 Level-1C Top-of-Atmosphere Reflectance Dataset was utilized to calculate the Normalized Difference Chlorophyll Index (NDCI) and chlorophyll-a concentration (62). The summertime was set as August. Each image in the dataset was used to calculate the NDCI image and further laid out the resulting image—Chl-a image in Rushan and Rongcheng nearshore aquaculture areas. The Chl-a concentration grayscale image was displayed in single-band pseudocolor between min-value 0 and max-value 25 μg L⁻¹. The spatial resolution of the Chl-a image from Sentinel-2 was 20 m by 20 m per grid cell. The basic Chl-a concentration map from GOCI II was the mean result image for July (63), and it was also displayed in single-band pseudocolor between min-value 0 and max-value 25 μg L⁻¹. The spatial resolution of the basic Chl-a map was 250 m by 250 m per grid cell.

Chl-a Concentration and DO at Rushan Oyster Farm.

The in-situ sampling was conducted on 23 August 2024. Measurement was performed using CTD48M within 20 cm of the sea surface.

Statistical Analysis.

Data were analyzed using SAS 9.4 (SAS Institute Inc., Cary, NC) and visualized with Origin Pro 2022 (OriginLab Corporation, Northampton, MA). Normality and homogeneity of variance were tested using the Shapiro–Wilk test and Levene’s test, respectively. When data were normally distributed and variances were homogeneous, one-way ANOVA followed by Tukey’s multiple comparison test was used to analyze differences. If the data were normally distributed but variances were not homogeneous, the Kruskal–Wallis rank-sum test was used. Statistical significance was set at P < 0.05, and results were reported as mean ± SD (Mean ± SD).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.