| 15 Oct 2025
Marine particles and their remineralization buffer future ocean biogeochemistry response to climate warming
Abstract. Transport and fate of particulate organic carbon (POC) and nutrients through marine particles co-determine the future response of ocean biogeochemistry and oceanic carbon uptake under climate warming. This makes the parametrization of the biological carbon pump in Earth system models (ESMs) an important model component and motivates us to compare a recently developed new sinking scheme (M4AGO; Maerz et al. 2020) to the current CMIP6 default Martin curve-like sinking scheme in MPI-ESM1.2-LR (see Mauritsen et al. 2019) under the future shared socio-economic pathway high-emission scenario SSP5-8.5. In their global response, the two model versions are similar, showing a decrease of integrated net primary production between the historical (1985–2014) and future (2070–2099) period of about 8.1 % and 9.7 % for the CMIP6 and M4AGO version, respectively. However, the models response differs latitudinally. In M4AGO, the temperature-dependent remineralization offsets the future increase in sinking velocity caused by a higher CaCO3 to POC ratio in the low latitudes. There, M4AGO thus buffers the export loss of nutrients to the mesopelagic, visible in little future changes of the export to net primary production ratio (the p ratio), while the CMIP6 version shows more pronounced changes with regionally declining or increasing p ratio. In the Arctic Ocean, the projected future increase of net primary production in the CMIP6 version is diminished with M4AGO through its higher POC transfer efficiency in high latitude regions. Hence, the more mechanistic and to environmental changes-responding M4AGO scheme shows a stronger buffering regional response to climate warming than the CMIP6 model version. The higher transfer efficiency also impinges on higher CO2 uptake in high latitude regions while the tropical regions turn later into a net sink with M4AGO compared to the standard CMIP6 version. Next to ballasting, we identified the particle microstructure as vigorous determinant for future changes of POC sinking velocity. Microstructure co-determines particle porosity and particle density. Processes governing the microstructure thus can be regarded as decisive to understand for reducing uncertainty of future POC fluxes.
The study by Maerz et al. provide an extensive analysis of the M4AGO parametrisation in a context of climate change. This parametrization includes temperature-dependant remineralization, oxygen limitation of remineralization, sea water viscosity, ballasting (composition) and a microstructure (fractal dimension / porosity) representation with aggregation/desagregation processes including particle density, size and stickiness. Sinking velocity is ultimately considering sea water viscositiy, particle composition (of ballast material), particle density, porosity and size. The study is very well written (although very verbose and some times convoluted), and provide a dense, comprehensive, well referenced, honest (especially about the limited impacts on global air-sea CO2 fluxes and limitations in general) and transparent analysis of this ambitious parametrization. The authors demonstrate a very high level of mastery in their disciplines.
They found little influence and global scale but highlighted regional important differences such as in the Arctic Ocean.
The review of this article was challenging. About 23 pages that must also include the lengthy study of Maerz et al. 2020 (another very lengthy and technical paper introducing the M4AGO parametrization). The writing is sometimes lengthy and technical. The paper in general would deserve a more synthetic and accessible bite. The problem with that is I really wonder who is able to read and actually digest this article beyond the small BGC modeling community.
The other problem is related to the microstructure parametrization which the authors claim is an important factor elucidating regional patterns of the BCP. If all other parametrizations are relatively simple and are developed similarly in other models, the microstructure parametrization increase complexity substantially with a lot of under- (or non-) -constrained parameters (see Maerz et al. 2020). I am aware that the authors already acknowledge this, guaranty computational efficiency and provide quantitative effects. Still, this is very hard to proof what is done here especially noting that the code is not open. How can this parametrization could be evaluated with observations? (The comparison with CO2 fluxes does not necessarily show an improvement to be honest). How could we constrain more the numerous parameters? (although if not achievable now, what could be used in the future?) I know they acknowledge there is little information available so far, but how could we proceed then? Are we sure the regional patterns are more realistic?
On the regional aspect, the authors put a lot of emphasis on the Arctic Ocean (and OMZ). If conclusions rather make sense for most regions to me, I was still puzzled by the conclusions drawn for the Arctic mostly because of the lack of synthesis capacities of the authors. Explanations are scattered around which makes a lot of work for the reader to re-assemble the results and conclusions. They show that M4AGO allows higher transfer efficiency compared with CMIP6 (Appendix D show maximum sinking velocity in the Arctic? Why?). But climate induce change towards:
=> If I understand well, this overall has the effect of decreasing the sinking velocity in the Arctic (Appendix D)
However, this is combined with:
I finally got the sense of the overall message: The total effect is RLS & transfer efficiency decrease despite increase in NPP (positive feedback loop). The authors should wrap this up somewhere better, it’s not an intuitive result. Same for other regions eventually.
This article is certainly worth publishing, but I would recommend a few changes and clearer explanations before doing so.
I have noted point-by-point comments below:
Line 107: What are the limitations of such hypothesis ? In general this does not stand in case of strong lateral advection.
Section 2.2: Why not adjusting calcite ?
Line 161: physical internal variability is not assessed, is there any differences in the physical fields ?
Line 165: Not true. Stratification only decrease in the Atlantic sector of the Arctic. Fix also statement line 227.
Line 174: With all due respect, this sentence is too complicated. There is sea ice now and there always will be … in winter. You are talking about summer sea ice. Seasonality, I guess you refer to the winter polar night (absence of light -> no NPP). And yes the Arctic is a small ocean but what the point if you discuss relative changes in % ?
Line 176: 100m is not the euphotic depth. It is a simplified threshold depth considered as the euphotic depth. Of course, much less accurate that an actual calculation of the euphotic depth (variable in time and space) to derive the export production. It’s fine ! But reformulate.
Line 179: While still using the SSP585 while we know this is not the way to go ?
Hausfather, Z. & Peters, G. P. Emissions – the ‘business as usual’ story is misleading. Nature 577, 618–620 (2020).
Line 204: Did I miss the obvious or the remineralization is not shown ?
Line 233: Sequestration. I have also used this word wrongly for while, I am not blaming, but could we fix that? You can refer to the nice Visser 2025 which clarifies:
“carbon sequestration is synonymous with an offset of carbon emissions”
https://doi.org/10.1002/lol2.70053
replace by storage at greater depth or similar.
Line 255: Arctic Ocean amplification, ref:
Shu, Q. et al. Arctic Ocean amplification in a warming climate in CMIP6 models. Sci. Adv. 8, eabn9755 (2022).
I agree but this is counter-intuitive for most reader and non-experts. Can you clarify here quickly what is meant ? You mean that there is more POM and therefore, relatively, less ballast material in the composition of particles if I refer to Appendix C. Why seasonal average ?
Line 260: If the inter-annual variability is represented by the STD, say it.
Line 270-272: needed ?
Line 274: Even a flux cannot ! Only change in storage.. See article by Frenger, I. et al. Misconceptions of the marine biological carbon pump in a changing climate: thinking outside the ‘export’ box. Glob. Change Biol. 30, e17124 (2024).
Line 309: you mean vertical DIC gradient right ? fix through the text.
Line 316: I can understand why (simulations from data product or your simulation) internal variability is a problem, but why the mean of the observational product is ?
Line 355: time-cumulative ?? you mean yearly integrated ?
Line 365: It is appreciated that the authors acknowledge that physico-chemical process dominate air-sea CO2 fluxes dynamics. Although this is repeated several times in the manuscript.
Line 370: Yes the BCP if responsible for the most part of the vertical DIC gradient. Rephrase.
Line 410: I don’t understand how more detritus production necessarily leads to less compact & bigger particles.
Line 420 : Explain me how temperature dependant remineralization has a direct effect on particles density and porosity ? You mean temperature in general ? I don’t understand this sentence.
Line 338: variable distribution slope ? you mean the size distribution? Not clear to me.
Line 500: Likely true. Positive feedback loop maybe see Oziel et al. 2025. Not represented in CMIP6 models… not so sure, prove it.
Line 516: between
Line 545: “more realistic” in terms of process maybe, but in terms of model performance ? Not sure.