| 11 Mar 2026
Novel in situ CO2 enrichment system reveals seagrass meadows are a refugium against coastal acidification for North Atlantic bivalves
Abstract. While the accumulation of anthropogenic CO2 in the atmosphere is causing a decline in global ocean pH, many eutrophic estuaries are already experiencing acidification due to accelerated respiration driving the consumption of dissolved oxygen (DO) and production of CO2, decreasing available carbonate ions (CO32-) and threatening marine calcifiers. Here, a novel in situ CO2 enrichment system was constructed to examine the effects of coastal acidification on the growth and survival of two species of North Atlantic bivalves (Argopecten irradians and Crassostrea virginica) in two distinct estuarine habitats: a seagrass meadow and an unvegetated sandy bottom in an open water estuary. The in-situ system captured natural diel dynamics as ambient chambers displayed chemistry nearly identical to the surrounding water, while CO2-enriched, acidified chambers maintained a consistent ~Δ 0.3–0.5 pH offset. At the unvegetated sandy bottom site, A. irradians and C. virginica displayed significant reductions in growth and survival in the acidified chambers (pHT = 7.3–7.5; saturation state of aragonite, ΩAr = 0.6–0.9) relative to ambient conditions (pHT = 7.6–7.9; ΩAr = 1.6–2). At the seagrass site, while growth of A. irradians and C. virginica in the acidified treatments (pHT = 7.3–7.7; ΩAr = 0.7) receiving the same delivery of CO2 was, again, significantly slowed compared to the control (pHT = 7.5–8.1; ΩAr = 2 – 2.8), the growth reduction, mortality rates, and levels of acidification were attenuated compared to the sandy bottom experiment, evidencing the ability of seagrass to buffer seawater and serve as a potential acidification refuge for bivalves. Collectively, the novel experimental CO2 enrichment system constructed for this project demonstrates that coastal acidification can have deleterious effects on marine bivalve populations, and that future conditions as well as the habitat refuge offered by seagrasses must be considered when developing management and restoration plans for temperate estuaries.
General comments
In this manuscript, Wallace et al. started describing spatial distribution of the carbonate chemistry in seawater of the two surveyed coastal areas in August-September (open water and seagrass beds). Then, they presented a novel in situ CO2 enrichment system they used for incubating juvenile bivalves (scallops, Argopecten irradians, and oysters, Crassostrea virginica) in acidified vs ambient conditions (~∆ 0.3 - 0.5 pH) at both sites. They monitored seawater parameters (temperature, salinity, dissolved oxygen and pH) continuously, and measured dissolved inorganic carbon twice a week to reconstruct carbonate system, for about one month. Finally, they estimated the growth rate and mortality of specimens at the end of the incubations, and statistically compared both conditions and both sites (One- and Two-way ANOVA, and pairwise comparison). Through this interesting approach, Wallace et al. showed a seawater buffering in seagrass bed that mitigates significantly the growth reduction and mortality of bivalves under acidified conditions. The methodological approaches are comprehensive and the in-situ incubation system is impressive. The study aims to characterize seawater conditions and relevant biological parameters. However, there is a lack of information on the hydro-sedimentary and bathymetric contexts at both sites. This information is necessary for understanding how the two systems function overall. The authors have extensively studied this area. A brief summary, referencing their previous, more comprehensive studies, would help the reader to better understand these two environments. Some elements lack clarity and the structure of the manuscript is occasionally difficult to follow and should be revised. Occasional redundancies in the description of the results could be avoided, and certain figures should be revised to improve readability, particularly in the results section, where the density of information is high given the number of variables studied. I would also recommend that the authors upload the data used for this publication (statistical tests and figures) to an online repository (i.e., Zenodo), making it accessible to all. Overall, I think this is a robust and very interesting research paper, with a wide variety of high-quality data, which corresponds to the scope of Biogeosciences.
Specific comments
Mat & Meth section
Line 115: At what month and what time of day did you take the samples to build the map? Were all the samples collected on the same day, or did you time your campaign over a specific period? I ask because we can expect significant variability in the measurements depending on the month, the time of day and the photosynthetic activity of the seagrass bed. You also specify that the Peconic Estuary map was done in August (l.128), and that you sampled at 3:00 PM during the incubations (l. 238).
Line 135: You specify that sampling is done one meter below the surface, but how far from the substrate in this case? Why not use the same approach as at the seagrass bed site to obtain truly comparable maps?
Lines 165-166: You state that you are targeting a pH of 0.3–0.5 points below ambient conditions to simulate acidified conditions, and that this range reflects the projected pH levels by 2100. However, you also note that these values are “commonly observed in eutrophic NY estuaries during the summer”. In this case, is it possible that future pH projections for this location could be lower than expected globally? Did the available data not enable to simulate a local decrease in pH by 2100?
Line 179: the value of adding a description of these parameters for both sites.
Line 210: Why are the experimental periods at the two sites different? This raises the question of whether this difference has an impact on bivalve mortality.
Lines 210-212: This information should be included in a paragraph providing context for the two sites.
Line 216: It's a shame there weren't more size categories available for the oysters! It would have been very interesting to include them.
Line 228: Have you included dead specimens in the growth rate calculation? If not, please specify (you write l.495: “the growth rates of the survivors”). If so, these specimens should be excluded, as they distort the growth rate, which is calculated over the entire incubation period. It is impossible to determine when their growth ceased due to death rather than the chemical conditions.
Line 254: Just out of curiosity, why did you choose the dissociation constants from Millero et al. (2010) over those from Cai and Wang (1998)? The latter are also relevant in estuarine environments.
Results section
The results section could be clearer if it were restructured. Section 3.1 presents the spatial distribution of chemical parameters at each site. It would then be logical to present a comparison of these parameters across substrates and treatments (change of scale; Section 3.4), followed by a comparison of treatments for each substrate (first paragraphs of Sections 3.2 and 3.3). Finally, after providing an overview of the carbonate parameters, the responses of the bivalves to the different incubation conditions could be presented (Section 3.5) and detailed for each substrate (second paragraphs of Sections 3.2 and 3.3). This would make the results easier to read by limiting repetition and helping to maintain focus on the objectives of each section of this study.
Line 296: How many square kilometers does the seagrass bed cover? Is it patchy? This kind of information should be included in a paragraph that describes the context of the site.
Lines 300-306: It is unclear which areas are covered by the seagrass bed and which are not. This information should be shown in Figure 2.
Line 311: What about ΩAr in the open water site?
Lines 333-334: I’m less certain about ΩAr mean value in the seagrass bed under acidic conditions: it appears to have been just saturated during the incubation period. The associated uncertainties are high (± 27 %), suggesting an alternation of under and oversaturated periods. Do you have any insight into this alternation? Could it be a day-night effect? This point is not sufficiently developed in the discussion (lines 469–475).
Line 414: You wrote “With respect to mortality, all bivalves demonstrated mortal sensitivity to acidification”, even though, as you pointed out l. 420–421, this is only truly significative for the 5-mm bay scallops.
Discussion section
I would also recommend restructuring the discussion to mirror the format of the results section. This would involve beginning by highlighting the 'halo' effect of seagrass beds as potential refuge habitats compared to the open environment, by discussing their buffering capacity and dynamics. The next step would be to move on to the bivalves responses to these different conditions, as well as the ecological and economic vulnerabilities under the tested conditions (acidified treatments). This would make the discussion easier to read, as it would limit repetition between sections and help maintain focus on the objectives of each one.
Line 444: I completely agree with the authors. The problem is that these factors (diurnal, tidal and seasonal) lack of context in this study. Without this context, it is difficult to fully grasp the results of this otherwise very interesting study, especially given that these factors are critical to the 'halo' effect associated with seagrass beds (Cyronak et al., 2018). Thus, we can assume that the refuge effect no longer exists in winter because the eelgrass loses its leaves. Therefore, the discussion should place greater emphasis on this limited seasonal timeframe.
Line 466: As mentioned earlier, given the magnitude of the associated deviation and the lack of information regarding the sources of this variability, I would be very cautious about saying that ΩAr is saturated, even on average.
Lines 489-490: Based on the results you present, there is no significant effect of the acidified treatment on oyster mortality.
Line 513-519: I'm not sure why you're changing the buffer indicator by introducing the DIC/AT ratio here. Why not apply this to the seagrass substrate as well? Alternatively, why not use this ratio instead of Rf throughout the manuscript? This ratio should then be added to Tables 1 and 2, or at least included in the supplementary material. Also, the sentence about ΩAr (l. 515-516) appears out of the blue when you are actually discussing the buffer capacity of the site; it should come after that as a transition.
Lines 564-566: Do you have any idea what proportion of alkalinization is contributed by sediment compared to seagrasses? Are the carbonate mineral compositions similar between the two sites? It could be assumed that the dissolution of carbonate minerals in the sediment also contributes to the alkalinization of seawater. In that case, the observed substrate effect might not just be a seagrass effect, but rather a combination of the nature of the sediment and the seagrass. It would be interesting to repeat the experiment in winter, when almost all the eelgrass leaves have fallen and photosynthesis has virtually ceased, to see how the bivalves respond to river discharge, which reduce the seawater buffering capacity. This would allow us to determine whether the sediment can then take over from the eelgrass bed as a buffer under these conditions.
Lines 577-580: Out of interest, are you aware of any seagrass restoration projects in the Long Island area? The same question applies to future shellfish harvesting activities.
Figures and Tables
Regarding the figures, I suggest that the authors reduce the number of them by merging some of them (e.g., Fig. 2/Fig. 3, and Fig. 6/Fig. 7/Fig. 8) to make it easier to compare the diversity of conditions, sites and organisms, and to improve clarity when reading the results. Similarly, Tables 3 and 4 should be reorganized by substrate, with one table summarizing growth rates and mortality in seagrass beds and another for open waters. This would follow the structure of the results. I suppose this is due to the document being generated automatically upon submission, but the figures and Tables are not always placed in the corresponding results sections, which makes the document harder to read. This should be corrected.
Figure 2: despite the high-resolution imagery, we are unable to identify the seagrass beds in the figure. It would be helpful to add their boundaries to the map. The same applies when adding the black box to maps B and C.
Figure 3: although the resolution is indicated in the caption, a scale bar should be added to images A and B/C, along with the north orientation (or latitude/longitude coordinates, as in Figure 2).
Figure 4: A) I’m not sure if the system diagram is to scale, but it would be helpful to include the dimensions. B) The axis labels are too small to be easily read by everyone.
Figures 6-7-8: The axis labels are also too small to be easily read. To ensure consistency with the tables and the terminology used in the text to refer to the treatment, I suggest replacing “elevated” with “acidified”.
Tables 3 and 4: specify again the number of specimens used under each condition. Due to the uncertainty values associated with the mean, it is not very appropriate to keep the decimal point.
Table S1: what is the difference between “true ambient” and “ambient”?
Technical corrections
As a Latin expression, “in situ” should be written in italic overall the text.
Line 120: Section 2.5 instead of 2.4.
Line 224: please cite the source of the software.
Line 262: please precise that all the statistical test results can be found in the supplementary material.
Line 416: “mortality” instead of “morality”.
Lines 418-420: the two sentences seem the same.
Line 430: “two-month” instead of “two-year”.
Lines 570/581: Sections 4.4 and 4.5 have the same name. I suppose section 4.5 is the conclusion, but I would recommend merging these two sections.
Line 586: I suggest also heat waves?