| 21 Oct 2025
Solubility of high-magnesium calcite in seawater and its implementation in (Py)CO2SYS
Abstract. The calcium carbonate pump is an important part of the ocean's carbon cycle but our knowledge of it is incomplete. For example, alkalinity data suggest that carbonate mineral dissolution happens at shallow depths where bulk seawater is oversaturated with respect to calcite and aragonite. It has been hypothesised that high-Mg calcites could explain this discrepancy due to their high solubility. However, our knowledge of what depth Mg calcites start dissolving and how they might respond to continuing ocean acidification is limited because their solubility in marine environments is poorly known. Here, we develop an approach to calculate Mg calcite solubility by using published solubility data under standard laboratory conditions and adding dependencies for temperature, salinity and pressure for ranges relevant to the marine environment. We then implement this into the CO2SYS software family (Python and GNU Octave/Matlab versions) and calculate saturation states globally for Mg calcites with different Mg%. Our results reveal that, contrary to previous assumptions, the saturation horizon for many high-Mg calcites often lies deeper than that of aragonite, suggesting that high-Mg calcites are unlikely to account for shallow-water carbonate dissolution. Our model aligns with the few existing in situ particle dissolution measurements of Mg calcites, but many unknowns remain regarding the solubility of Mg calcites in marine environments and future experiments focusing on temperature and pressure dependence are needed to better constrain their role in the marine carbon cycle.
Competing interests: Some authors are members of the editorial board of Ocean Science.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
The authors use the extremely limited published equilibrium solubility data of magnesium calcites (Mg-calcite) and incorporate it into the widely used CO2sys programs allowing for calculation of saturation states (Ω) for Mg-calcite alongside values for aragonite and calcite. From this they calculate global Ω states and find, contrary to essentially every published study for the last 60 years, that high Mg-calcites are more stable in seawater than aragonite. They then conclude that Mg-calcites can therefor not explain observed total alkalinity and sediment trap data indicating CaCO3 dissolution above the Ω for aragonite. I’m skeptical of their conclusions due to lack of discussion on how to reconcile the result with conflicting studies (particularly Morse et al. 2006 which used the same equation to calculate Ω), and due to significant disagreement within the literature in how to interpret Mg-calcite solubility data leading to, in my opinion, a more likely conclusion that Ω is not appropriate for Mg-calcites and should thus not be incorporated into CO2sys.The two case studies provided are either poorly described or misleading. In general, the manuscript is not well organized and lacks significantly in background details and discussion.
I was unable to test out the software because I do not use python it was unclear where to find the matlab version even though it is indicated that it exists (L142); but as discussed throughout this review, implementation of Mg-calcites into CO2sys is probably premature.
Interpreting Mg-calcite solubility data has always been fraught with problems and challenges, most of which have never been satisfactorily resolved and are still under debate, including if thermodynamic solubility is a meaningful way in which to discuss Mg-calcites (Morse and Mackenzie 1990; Morse et al. 2006; Mucci 1983; Lafon 1978). While it would be too much for the authors to give a full detailed discussion, the introduction, particularly section 1.1, should discuss the most relevant points, and the discussion and conclusions should consider these challenges when interpreting the results and drawing conclusions. Specifically, the equations used (e.g. eqn. 2) is not the only formulation that has been proposed (Morse et al. 2006) and that a major problem with this equation is that it assumes a one component system, when it is in fact at least two (most likely more), and for a given set of conditions the aqueous solution can only be in equilibrium with a single MgCO3 content (x), which is not supported by the range of Mg content found in natural waters.
One aspect of the manuscript that is very unclear is varying definitions used for the equilibrium constants, how they are converted, and which are being used where. The literature itself is often not clear and different papers use different terminology. There are three main different types of constants, thermodynamic (activities), stoichiometric (mol concentrations), and apparent stoichiometric (mol concentrations with an operationally defined a(H+)). It is not easy to interconvert between these values, particularly in seawater where even with modern Pitzer equations the activities are not known precisely enough. The nomenclature they chose to use for these different values is never clearly defined particularly because it is different from many of the works they cite (e.g. they use stoichiometric when using activities while most works, including Mucci 1983 which is in CO2sys uses stoichiometric to mean concentration based). There are also times where they appear to switch nomenclature (e.g. line 71 vs lines 422-423). Throughout the manuscript I was often confused on which were being used, particularly when it comes to how the data are incorporated into CO2sys. Importantly, the authors do not give units, which make it challenging to follow their methods because Pitzer equations are usually in mol/kg of (pure) water, while CO2sys uses and reports values in mol/kg of seawater. Where the lack of clarity on which K is being used is most confusing is when comparing the Mg-calcites to aragonite and pure calcite because CO2sys uses Mucci 1983 which are in concentration units, while in this work eqn. 2 is in activities, but eqn 36 is in concentration. The main reason CO2sys does not use activities is because they cannot be directly measured, and current Pitzer models are incomplete and still not yet accurate enough for use over the range of oceanic conditions. In particular, the uncertainties in the Pytzer model used are not adequately discussed or quantified. For example, Mucci (1983) noted the need to consider the CaHCO3- ion pair, but that is not included in equation 26. Relatedly, the carbonate dissociation constants in CO2sys are in concentration not activities. Also of note, CO2sys uses the CO2* convention, while Morse and Mackenzie (1990) do not, which would impact the calculation of 𝛾CO32-, but I do not know to what degree. It is not clear within the text how these different models are unified in order to be directly comparable with each other or the uncertainties. Or even which is incorporated into CO2sys.
One of the main conclusions is that prior studies used eqn 40 to calculate Ω Mg-calcite rather than eqn 36, leading those studies to incorrectly conclude that Mg calcite would begin dissolving above the aragonite saturation horizon. While this difference may explain the discrepancy between this work and that of Woosley et al. (2012) and Hashim et al. (2025) it does not explain the discrepancy with Morse et al (2006). Like this work, Morse et al. (2006) calculated Ω using the full equation, yet their results agree with Woosley and Hashim rather than this work. What is the explanation for this discrepancy?
The authors further contend that eqn 40 is incorrect and eqn 36 is the ‘correct’ (line 550) method. However, this is up for debate and the case studies presented do not provide the evidence needed to resolve the debate. Given the challenge and uncertainty in how to represent Mg-calcite solubility and uncertainties in the activity models, eqn 40 as done by Woosley et al. (2012) and Hashim et al. (2025) provides a method of canceling out some of the uncertainties in activities and making the K’s more directly comparable. Additionally, given the near constant Mg/Ca ratio in seawater, CO32- can act as a proxy for saturation (Bertram et al. 1991, Broker and Peng 1982). Given the large body of literature showing Mg-Calcites to be less stable (above some threshold x), and evidence that Mg-content alone is insufficient to describe their solubility (Morse et al. 2006), more evidence would need to be presented to demonstrate that eqn 40 is not actually the ‘correct’ one. Perhaps a discussion of Lafon (1978) is warranted as it criticizes the assumptions of the equilibrium model on which this is based on and could provide an alternate conclusion for the results.
Section 1.1.2
The disagreement between solubility measurements of different types of Mg-calcite is a long-standing problem that has yet to be resolved. The discussion of this problem and how the different studies are categorized and what the different categories mean is not adequately described. Particularly confusing is that that the names are different than commonly used in the literature (e.g. Morse and Mackenzie 1990). The discussion text often talks about the categories with respect to sample cleaning and handling, but the description of the categories focuses more on defects. Also, given that fish carbonates are a case study, discussion of where they fit in is very important in this section. The discussion in section 4.4 about how fish carbonates do not really fit into these categories also undermines category descriptions and the main conclusions of the study.
Table 1 should be expanded to include all of the critical details one would need to make a determination of which category their study fit into. This would include experimental conditions, temp, pressure, media, Mg contents. How the samples were obtained and prepared. More details about the sample than ‘biogenic’ or synthetic is needed. ‘Cleaned’ and ‘annealed’ are put in the same category, but rinsing a sample to remove organic matter would have different impacts on the sample than annealing, and both would address different types of ‘defects.’ The authors may have been trying to be intentionally vague because of the uncertainties, but they have done so to the point of being meaningless.
I have some issue with using the terms ‘with defects’ and ‘no defects.’ Mg is actually a defect. Even in very high Mg contents that approach 50% they are not called dolomites or pro-dolomites because they do not have the ordered crystalline structure of dolomites, and the Mg is effectively randomly located within the crystalline structure. As discussed in a lot of these studies it is the strain put on the crystal (which varies by exactly where in the crystal the Mg appears) that impacts the free energy and resultant solubility. As a result, Mg content itself is insufficient to characterize the solubility, and sample preparation is insufficient to define the categories. There’s also evidence of hydrated carbonates, which should be discussed somewhere.
A value of x=0.14 is often used as a comparison point because it is the ‘average amount found in shelf sediments’ (no citation is given). However, that is what makes it a poor comparison point: it indicates that all these fish carbonates with much higher x are already dissolving and not making it to the sediments. In other words, these figures showing 5, 10, 15% are probably not useful comparisons if trying to demonstrate that Mg carbonate is not dissolving in the surface ocean because we do not expect them to be more soluble than aragonite.
L356-349: This clearly demonstrates that the uncertain in pK is the largest source of uncertainty and even the value for Woosley et al. (2012) is probably an underestimate since it only considers precision. More emphasis should be given to this, and the value should probably be increase in table 3.
Case Study: Fish-produced Mg-calcite:
The recognition of icthyocarbonates recently has renewed interest in studying Mg-calcite and highlights a potential use of this work. However, many important details are missing from the figures and discussion which will make these results easy to misunderstand or misuse. Fish carbonates are extremely diverse and range from amorphous carbonates, to aragonite, to very high Mg-calcite with morphologies that vary by species (e.g Salter et al. 2019). Presenting a global map (and discussion) as representing one single measurement and labeling it as Mg Calcite (fish) as though it represents all fish carbonates of all types and values of x is extremely misleading, even the value of x is missing from the figures and captions and the discussion of this limitation in the text is inadequate.
Lines 421-425 are very confusing. It states that eqn 36 is used to calculate saturation horizons. Equation 36 uses K*sp and molar concentrations, but the paragraph goes on to describe how the K* of Woosley et al (2012) is first converted to K (thermodynamic or activities). If K was used then were the aragonite and calcite K*’s also converted to K? Fish carbonates are described as being very important biogeochemically, yet these results are then described as not being applicable to them.
A 300 m shallower (line 430) Ωfish is still very significant and would likely lead to significant dissolution above the aragonite saturation horizon.
Paragraph lines 433-441 is confusing. It is unclear what solubility estimates are inconsistent with the ichthyocarbonate dissolution rates. They all show dissolution at Ω aragonite >1 in agreement with Woosley et al. (2012) and in disagreement with this work. The Ω at which dissolution begins varies but that is not surprising given the different Mg content, likely morphology variability, and differences in sample preparation and handling. The discussion of organic coatings is lacking. Such coatings would create a barrier slowing contact of the carbonates with seawater, and thus their dilution rates, but not their solubility. However, these coating have a very low density and would slow the sinking rate, meaning the carbonates would spend more time in corrosive waters and at least partially counteract a slower dissolution rate. Oehlert et al. (2024) still predict complete dissolution of untreated fish carbonates above 1000m (even without considering the impact of Ω on dissolution rates). Care should also be used in comparing these rates to Folkert et al’s because Folkert rinsed the carbonates with pure water and may have partially removed or otherwise altered the organic coatings even if they were not chemically removed with bleach.
Lines 439-441, most likely it is a combination of all of these factors along with many other challenges in representing Mg-Calcite solubility.
Case Study: Mg Calcite ooids
This section is based on Milliman (1977) data with used 12% Mg calcite. Even with the different uncertainties in solubility, the literature is in general agreement that this Mg content ooids would not be more soluble than aragonite. It is unclear what this section is meant to show. Given the average Mg content in shallow water sediments is 14%, it suggests that only Mg content greater than that is more soluble than aragonite.
Line 450-451. This statement is incorrect; you cannot equate kinetics and thermodynamics in this way. If that were true, the oceans would be dominated by dolomite and not calcite and aragonite.
L458 is contradictory, saying Ωcrit is the same for Mg-calcite and pure calcite and aragonite then stating that the rate change observed is different for Mg-calcite.
L460: citation for this equation needed. It is also important to note that buried within the k is the surface area to volume ratio which is a major controlling parameter in the rate.
L461: statement that Mg calcite has a similar reaction order to calcite and aragonite is very misleading. Yes, the values are all around 3, but the difference in k between an n of 2.5 and 3.5 is almost 3 times. Walter and Morse (1984) meant this to be more in comparison to other minerals where the values can be 16. It does not mean that the differences between the three CaCO3 are insignificant.
Global Saturation States:
L381 notes that category 1 is undersaturated at the surface. As category 1 would be most representative of the natural environment, how does this not contradict the conclusions that Mg carbonate is not less stable than aragonite?
Minor comments:
Eqn. 3 – 5, use of the 1-3 subscripts could be confusing as the carbonate dissociation constants are often defined as K1 and K2. Should also define what ranges of x these are applicable to.
Figure 1. Cite values for aragonite and calcite. Should indicated that all of these are in pure water.
L4: add ‘at least partially’ explain
L8: how the salinity dependence is determined is not well explained in the methods
L19: ‘without significantly affecting pH’ is highly subjective
L24: carbonate counter pump is called CaCO3 pump in the abstract
L25-27: sentence awkwardly worded, difficult to follow.
L51-59: poorly worded paragraph. Also, saturation states might not be the best way to address this problem.
L229: should be noted that although this uses magnesite (as commonly done), magnesite is actually very rare in the environment, leading to uncertainties in the applicability of this assumption.
L281: needs a citation. While this assumption may be necessary, it is unlikely to be true.
L284: poorly worded sentence. Looks like 2.8 is multiplied by ∆V.
L289: Unclear what formula units and unit cell are.
L299: Need to give the valid range of x
L320: explain what ‘a certain amount’ means.
L330: why is this assumption needed? Why not just include them all?
L334: why is this only for the surface ocean?
L339: stating that 6000db is aprox 6000m is unnecessary.
L331: Need a citation for x=0.14.
L356: What is meant by “quality of these three fits?” Also, they all seem to overestimate measured values which should be noted.
Fig. 4. The colors are too similar and difficult to distinguish in the figures.
L386: Why not use the Boron value of Lee which has been shown to be more internally consistent?
L389: Which sulfate value?
L394: Why is 10-18% most relevant?
Fig. 5. Everywhere else category 1 is called fresh not minimally prepared.
L481: Clearly category is important but have not adequately described which category is most relevant biogeochemically.
L491: column ‘and’ contribute
L492: what is meant by ‘repackaged’?
L503: Discussion of how fish carbonates were implemented and how they fit in with the categories is needed.
L504-505: Needs more development. Every sample requires some amount of prep and equating all forms of cleaning and prep as being equal is misleading.
L509: What is meant by low concentrations of foreign ions? Seawater is full of them.
L518: Incorrect, do not equate solubility and kinetics, they are very different.
L521: why aren’t Bertram et al. (1991) included in any of the categories?
L533: There are 3 categories, how would experiments on only two help resolve the discrepancies?
L536: Incomplete sentence.
L543: Such experiments would need a way to account for incongruent dissolution.
L551: ‘correctly’ is debatable
L552: What is meant by ‘parts of’?
References:
Bertram et al. (1991) Influence of temperature on the stability of magnesian calcite Am. Minerology.
Broker, W. S., and T. S. Peng. (1982) "Tracers in the sea."
Hashim et al. (2025) DOI: 10.1029/2024GB008387
Lafon, G. M. "Equilibrium criteria for two-component solids reacting with fixed composition in an aqueous phase; example, the magnesian calcites; discussion." American Journal of Science 278.10 (1978): 1455-1468.
Morse, J. W., & Mackenzie, F. T. (1990). Geochemistry of Sedimentary Carboantes (Vol. 48). Elsevier.
Morse et al. (2006) DOI:10.1016/j.gca.2006.08.017
Mucci (1983) DOI: 10.2475/ajs.283.7.780
Oehlert et al. (2024) DOI: 10.1029/2024GB008176
Salter et al. (2019) DOI: 10.1002/lno.11339
Woosley et al (2012) DOI:10.1029/2011JC007599