the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Mar 2026
A diverse community constitutes global coccolithophore calcium carbonate stocks
Abstract. Coccolithophores are diverse calcifying plankton, yet most research has focused on a single species, Gephyrocapsa huxleyi, with the global contributions of other species hitherto unexplored. Since coccolithophores account for the majority of marine calcium carbonate (CaCO3) production, this narrow focus risks biasing our understanding of CaCO3 cycling, as other species differ in their distributions, CaCO3 production and response to climate change. Using a global, species-resolved machine-learning approach, we show that G. huxleyi contributes only about 7 % of estimated coccolithophore CaCO3 stock, while a morphologically and functionally diverse assemblage dominates. Since stock contributions are a good proxy for contribution to production, our findings challenge the view that G. huxleyi underpins CaCO3 cycling and show that lab and in situ datasets centred on this species capture only a small fraction of coccolithophore calcification. Our work identifies key species and regions to guide future laboratory, in situ, and modelling efforts, laying the groundwork for more realistic representations of CaCO3 cycling under climate change.
Status: open (until 08 Jul 2026)
- RC1: 'Comment on egusphere-2026-1632', Anonymous Referee #1, 27 Apr 2026 reply
Data sets
Abundance model output Joost de Vries, Fanny Monteiro, Alex Poulton, Nicola Wiseman, and Levi Wolf https://doi.org/10.5281/zenodo.16886603
CIC model ouput Joost de Vries, Fanny Monteiro, Alex Poulton, Nicola Wiseman, and Levi Wolf https://doi.org/10.5281/zenodo.16887386
COC model output Joost de Vries, Fanny Monteiro, Alex Poulton, Nicola Wiseman, and Levi Wolf https://doi.org/10.5281/zenodo.16986899
Model code and software
Abil software Joost de Vries, Nicola Wiseman, and Levi Wolf https://doi.org/10.5281/zenodo.16886568
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 500 | 0 | 1 | 501 | 0 | 0 |
- HTML: 500
- PDF: 0
- XML: 1
- Total: 501
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of de Vries et al “A diverse community constitutes global coccolithophore calcium carbonate stocks”
Summary:
de Vries et al use a machine learning based approach to estimate abundances of 58 coccolithophore species throughout the global ocean. Then they estimate organic and inorganic carbon stocks for each species. They find that E. hux does not account for much of the standing inorganic carbon stock (~7%), despite being the most widely studied coccolithophore species. They find that there is a large cocco CaCO3 stock in the subtropics, including in the sub-euphotic zone. They find that heavily calcified coccolithophore species contribute a lot to the PIC stock. They also examine species diversity finding the highest diversity in the subtropics.
This paper is generally well-written and is an interesting exercise using machine learning. I enjoyed reading it. Undoubtedly a lot of detailed work went into parameterizing these species-specific models and developing the machine learning approach. I find the results of broad interest by the coccolithophore community. However, there are some important details that I think need to be addressed before publication.
Major comments:
I think that it needs to be emphasized that this paper is only capturing coccolithophore stocks. One cannot simply extrapolate stocks to production/cycling, as so many factors contribute to these rates (bottom-up environmental factors that influence growth rate and species-specific growth rate bounds). What if E. hux stocks are small but they have fast turnover times (and shed coccoliths rather than hanging on to them) so still contribute more to CaCO3 cycling than other species? This question cannot be answered in the present study. The authors reference one study (Daniels et al., 2014) that found that E hux doesn’t have that much faster of a growth rate than two other species, but that’s under lab conditions. I think rather than just saying it doesn’t matter, just acknowledge that it’s a limitation and that’s ok. This is still a super important piece of research on the marine carbonate cycle despite not having CaCO3 production rates.
The large confidence intervals should at least be mentioned when describing the results. (like when describing parts of Figs 3, 4, 5). I think it’s good that the large Cis are explained a bit in section 3.8, but just being humbler when writing about the results would be nice.
I found that the Results & Discussion part of manuscript was missing species-specific discussion, despite the use of species-specific models. As a coccolithophore enthusiast, I am interested in species adaptation to certain environmental niches. This paper has so much fun coccolithophore information but it’s kind of brushed over with so much focus on carbon stocks. I would really like to see distribution maps for each species in the appendix. For example, what species are dominating surface subtropics? Umbellosphaera ones? Do you get certain species appearing in regions where they have never been observed? What conditions appear to select for certain species?
Also, is each species model independent from each other? I.e., does the presence of one coccolithophores species influence how much of another species can be present? I think they are independent, based on what I’ve read, and it would be interesting to evaluate community structure: are species that are regularly observed together have similar distributions in your models? Do you see any surprising results of species assemblages repeatedly appearing together?
Lastly, I’m not sure if the machine learning approach allows this, but it would be good to know which of the environmental variables you used to predict species distributions ended up being the most important for each species… this would very useful and interesting because it would help constrain what environmental conditions lead to more or less of a certain coccolithophore species (and thus CaCO3 stocks) and actually help develop process-based biogeochemical ocean models. Can you draw any broad relationships between coccolithophore assemblages/CaCO3 stocks and environmental conditions? Having just a bit of information about the oceanic conditions leading to high/low CaCO3 stocks would make the paper much more impactful.
Detailed comments:
Line 2: please put “(also known as Emiliania huxleyi)” after “Gephyrocapsa huxleyi”
Line 5: I think you need to clearly state what science question the machine-learning approach is testing. For example: We tested the hypothesis that G. huxleyi is a minor contribution to marine CaCO3 production/cycling by employing a machine learning approach…
Lines 8-10: To make this work useful to lab and modelling efforts, it would be good to know which environmental variables are leading to resulting distributions (see last major comment above)
Line 25: Could you please add a reference for when the name of E. hux changed officially? Who decided this?
Line 35: aren’t loss and removal the same thing?
Line 38: which paper assumes fast and slow growing coccolithophores species?
Line 45: I’m not sure what defines a coccolithophores as “heavily calcified” (e.g. PIC/POC > 1?) or lightly calcified? G hux has morphotypes that are more heavily calcified so this statement needs more nuance.
Line 51: they have large 95% CI, so is “statistically robust” the correct word here?
Figure 2: could you color code the dots on the maps with how many species were actually observed at each point? I appreciate the number of samples by depth in panel B but this doesn’t give information about the species number.
Line 80: why was oxygen included? I don’t see how this would influence coccolithophores
Line 87: 25-fold is very big! Could you provide a few examples on the extreme ends of the spectrum? which species are these?
Line 93: is voxel a machine-learning term for gridcell? I’ve never heard this before
Sections 2.2.2 to 2.2.6: I am not a machine learning method expert so I will not comment on these Methods sections. However, I’m impressed by how many steps/statistical methods/code packages are needed to produce this dataset. Were all these steps planned, or did you discover the need for additional steps while doing this research? Just curious how this type of study develops. It might be good to give an overview sentence at the end of section 2.2.1 describing how you got to such complexity or if you followed other similar studies. Using ML seems quite opaque for those of us who have not used it so being as frank and open as possible about the process would be interesting to many readers (I think).
Line 220-222: I see the figure A4 shows where the observational data need to be extrapolated. So, are the yellow areas the most uncertain? In other words, the North Pacific is a location where the results are most uncertain?
Line 223: It would be great to see maps of all 58 species somewhere, beyond just what’s shared in Figure 5. These maps are not in the 2024 Scientific data paper either unfortunately. Seeing these maps would be great for readers curious about specific species. Perhaps add a big (multi-page) figure of 58 little maps (one for each species) in the appendix?
Line 258: I feel like not including detached coccoliths could also influence the result that G. hux is a minor contributor to CaCO3 stocks. Do other species shed coccolithos like G. hux or is it primarily G. hux? Perhaps include a sentence on which species shed coccoliths so the reader has an idea of how this could influence results.
Figure 3: is panel A also showing inorganic carbon stock? does the color scale to the right of panel B go with the panel A too? Also, panel E is described as integrated stocks but it’s a percentage. could you please clarify?
Section 3.2: I think that you can just refer readers to Table 1 for the confidence intervals. Maybe mention as well that they are large CIs and mention that it’s just part of the method of using sparse observations. Having all these extra numbers is distracting.
Line 267-268: is this what’s shown in Figure 3C? perhaps reference this figure panel
Section 3.2 overall: the title of this section seems to be only a minor point in the paragraphs within the section. I think more attention needs to be placed on the subtropical component if you are keeping this section title. For example, why would the inorganic carbon stocks be so high in subtropics? These are highly nutrient limited, and it’s been shown that coccolithophores may continue to calcify even if photosynthesis is limited. could it be that these subtropical species are heavily calcified for that reason? which species are dominating? or is the big subtropical contribution to the standing stock just because the subtropical regions have a large spatial footprint? or maybe it’s because these regions have the highest CaCO3 saturation state/pH and thus are conducive to calcification?
Line 280: could you put dotted lines for breaking up the upper and lower euphotic zone in Figure 3, panel B? or just combine them here in the text for clarity. It’s surprising that there is more PIC stock in the sub-euphotic than the lower euphotic, but I don’t know if that’s just because the sub-euphotic is a thin layer.
Linear 283-288: Again, please include species specific information. Poulton et al 2017 observed certain species at depth. is your model capturing these species? It seems like F. profunda might be the big contributor here based on Figure B, but this figure isn’t called out here.
Line 297: Please report the picpoc ratio used for G. hux in this sentence. the color scale on fig 1 is hard to read for a specific number. E. hux picpoc ratios can vary a great deal so this is an important detail. Also, since you only use one picpoc ratio per species, then how did you decide this? is it the mean or median of all available observations?
Line 300: I don’t see why this sentence needs to be in a separate paragraph
Line 308: I feel like there is some confusion between carbon cycling and standing stocks. When I read carbon cycling, it means more rates and dynamics, whereas the model you’re describing is more about stocks. e.g., what if these heavily calcified species are slower growing or they just naturally show up in regions where growth is limited by temperature or nutrients? Could this be discussed? I think the language needs to be tailored to what you are reporting. (see first major comment)
Line 358-360: It sounds kind of strange to be discussing how important variable PICPOC ratios are when the models you use here have fixed PICPOC ratios for each species. Did you derive any relationships between environmental conditions and the resulting coccolithophore community picpoc ratio? This would be useful information to develop better models, like the studies you listed. Also worth noting the coccolithophore inorganic to organic carbon production ratio in Krumhardt et al 2019 is variable based on environmental conditions (carbonate chemistry, nutrient status, temperature): were these important predictor variables?
Line 380: I would argue that environmental conditions (nutrient & light availability and temperature) are also important drivers of growth rates.
Line 383: I would think that bloom-forming coccolithophores could have the potential to grow faster in nature. Rather than just saying growth rates are similar based on one laboratory study, please just acknowledge that it’s a limitation of your study. Since you don’t report CaCO3 production rates, this comment is just to make sure that your results on CaCO3 stocks are interpreted correctly by readers.