the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Aug 2026
Biological imprint dominates global mesoscale air-sea flux anomalies of carbon dioxide and oxygen
Abstract. Ocean mesoscale phenomena ("mesoscale eddies") are fundamental components in mediating air-sea heat and momentum fluxes through anomalies of sea surface temperatures and currents. Yet, their impacts on air-sea gas fluxes, such as carbon dioxide, CO2, and oxygen, O2, through changing solubility and biological cycling, remain underexplored. This study aims to diagnose global patterns of ocean mesoscale CO2 and O2 flux anomalies and their drivers. To this end, we use results from an ocean eddy-rich 0.1° GFDL climate model (CM2.6), namely a preindustrial control simulation and an idealized climate change simulation with a linear increase in atmospheric CO2 until CO2 doubling is reached. Mesoscale air-sea CO2 and O2 flux anomalies are isolated from large-scale signals by applying spatial filtering to monthly averaged model results.
We find that globally mesoscale variability explains approximately 6–7 % of the variance in CO2 and O2 fluxes, with regional contributions exceeding 30 %. We present an analytical framework to attribute air-sea CO2 and O2 flux anomalies to thermally-driven solubility effects versus biological imprints, based on the sign of the correlation between CO2 and O2 flux anomalies.
We find a clear regional imprint in the mechanisms by which mesoscale eddies influence CO2 and O2 fluxes. In subtropical and mid-latitude regions, CO2 and O2 flux anomalies are predominantly of the same sign, indicating that mesoscale eddies impact gas fluxes mainly through solubility changes. In tropical and high-latitude regions the effect of mesoscale eddies on CO2 and O2 flux is mostly of biological origin, as indicated by an opposite sign of CO2 and O2 flux anomalies, either caused by changes in biological productivity or upwelling of a respiration signal from the ocean interior. Although the regions with a biological imprint globally cover an area comparable to solubility-driven regions globally, the associated flux anomalies are larger in magnitude, causing the biological drivers to dominate the globally integrated absolute mesoscale anomalies, representing approximately two-thirds of the total signal, and account for 7–9 % to global variance (compared to 5–7 % for solubility-driven regions). Under CO2 doubling, solubility-driven regions expand and fluxes intensify, increasing their relative contribution, though the large-scale patterns of drivers remain the same. The biological contribution remains dominant for O2, while the relative importance of solubility-driven processes increases for both gases under warming. Our results highlight the spatial organization of mesoscale-driven air-sea gas fluxes into distinct regimes and demonstrate that, on a global scale, a biological imprint dominates the magnitude of mesoscale CO2 and O2 flux anomalies.
- Preprint
(27670 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4347', Ford Daniel. J., 11 Sep 2026
- RC2: 'Comment on egusphere-2026-4347', Anonymous Referee #2, 14 Sep 2026
-
RC3: 'Comment on egusphere-2026-4347', Anonymous Referee #3, 17 Sep 2026
The authors use the GFDL climate model at 0.1° resolution, characterised as “mesoscale eddy-rich” setup. The biogeochemical component is miniBLING, a simplified three-tracer model (dissolved inorganic carbon, oxygen, and a generic macronutrient) chosen to keep eddy-rich biogeochemical simulations computationally tractable. Their goal is to identify the dominant drivers of mesoscale air-sea CO2 and O2 flux anomalies and characterise their global spatial patterns, both under preindustrial conditions and under an idealised CO2-doubling scenario. Methodologically, mesoscale anomalies are isolated by applying a 3°x3° spatial filter to obtain the large-scale background field, and them computing the anomaly as the difference between the total field an this background. To infer the underlying drivers of these anomalies, they build on the framework of Resplandy et al., (2015). The sign of the correlation between mesoscale CO2 and O2 flux anomalies is used to separate thermally-driven effects from non-thermal biological effects. The paper’s key methodological extension is adding a second correlation between CO2 flux anomalies and productivity anomalies, to further split the biological signal into productivity driven vs respiration-signal regimens. The main result is a clear spatial regime structure in subtropical and mid-latitude regions, thermal effects dominate the mesoscale flux anomalies, whereas in tropical and high-latitude regions, biological effects dominate. The solubility-driven regime is slightly larger (~51% of the ocean, growing to 60% under CO2 doubling), but the biologically-driven regime produces larger magnitude anomalies, such that biological effects ends up dominating the globally integrated flux anomaly for both gases.
General assessment
The manuscript is disciplined and easy to referee. Each result subsection maps directly onto one of the three stated research questions, which make it straightforward to check whether each one is actually answered. The authors are transparent about their own method’s limitations. The discussion section reads as the author genuinely engaging with how their result complicates part of the existing literature’s framing.
The CO2-O2 pairing idea is, in my view, the strongest conceptual choice in the paper. The stoichiometric asymmetry they’re exploiting (biology moves CO2 and O2 in opposite directions, solubility moves them together) is of course well established since Resplandy et al. (2015), but extending it to the mesoscale is non-trivial step, since mesoscale and submesoscale mixing, short eddy lifetimes, and lateral stirring could plausibly have broke down that clean signal. That it holds up, and that they can further split the biological effect into productivity vs respiration is a genuine methodological contribution rather than a rescaling exercise. Worth stating that plainly in the review rather than burying it as one strength among several.
The abstract and conclusions state the headline numbers (60-70%) with more confidence than their own sensitivity analysis (filter size, temporal resolution) strictly supports. The qualitative claim, biology dominates magnitude despite occupying less area, is well-argued and I don’t doubt it. But the specific percentages are filter and resolution-dependent in ways the body of the paper acknowledges and the abstract doesn’t.
Specific comments
- Driver classification doesn’t report significance: Fig.1 and Fig.4c clarify each grid point by the sign of the CO2-O2 Pearson r, with no significance threshold or minimum |r| reported anywhere. The regionally-averaged correlations quoted and shown are weak enough that some fraction of the map is likely being classified by sign that isn’t statistically distinguishable from zero. Please add significance mas, or report how the area/ flux percentages in Fig. 4d,e change under a minimum |r| threshold (e.g., 0.2).
- Effective degrees of freedom are not addressed: Correlations are computed from 120 monthly samples per grid box (10years x 12months). Monthly mesoscale gas-flux anomalies are almost centainly autocorrelated, since eddies persist over weeks to months, so the effective sample size is well below 120. Any significance test run on these correlations (see point 1) needs to account for this, or it will overstate how significant weak correlation are.
- Eddy resolution is weakest exactly where the biological signal is claimed strongest: They state the model only partially resolves mesoscale variability at high latitudes. Separately, they report that biological imprint is most prominent at high latitudes and the Southern Ocean. These two facts are never connected in the discussion. If eddy resolution is marginal at high latitude, part of what’s being classified as respiration-driven, weak backgrounds variance, high variance contribution could plausibly be partially unresolved dynamics rather than a genuine biological signature. Since the author already cites Pacheco et al. (2026) for EKE patterns, overlaying that EKE field on the driver classification map would be a low-cost way to check whether "respiration-driven, high variance contribution” regions coincide with poorly-resolved eddy activity rather than a genuine biological signature.
- CO2-doubling area shift conflates circulation change with biogeochemical driver change: The growth of the solubility-driven area under CO2 doubling (51%—> 60%) is attributed to “expanding subtropical gyres”. But gyre expansion under warming it itself a circulation response (poleward shift in wind stress curl, changing EKE distribution), independent of the biogeochemical framework. Fig. C3 shows the spatial reclassification but not an EKE-change overlay, so it’s not possible to tell whether the area shift reflects eddies physically relocating (a dynamical effect) vs the same eddies in the same locations flipping correlation sign (a genuine mechanistic shift). These are different claims and should be distinguished. An EKE or gyre-boundary overlay on Fig. C3 would resolve this.
- The 6-7% global variance contribution: Mesoscale variability is reported to contribute 6-7% of “total” flux variance globally, and this is linked to monitoring, reporting and verification marine CDR in the conclusions. But the total variance is almost certainly dominated by the seasonal cycle, with inflates the denominator and makes the 6-7% number look smaller than it maybe for those purposes. What actually matters for detecting an inter annual or trend signal is mesoscale variance relative to inteannual/trend variance specifically, not relative to the seasonal cycle. Maybe It would be better to report mesoscale variance contribution with the seasons cycle removed from the denominator.
Citation: https://doi.org/10.5194/egusphere-2026-4347-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 182 | 79 | 78 | 339 | 75 | 75 |
- HTML: 182
- PDF: 79
- XML: 78
- Total: 339
- BibTeX: 75
- EndNote: 75
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Hocke et al. assess the impact of mesoscale eddies on air-sea fluxes of O2 and CO2 a eddy rich modelling approach. Their analysis uses air-sea CO2 fluxes and O2 fluxes to diagnose the dominant drivers to the air-sea flux modifications by the eddies, into a solubility or biologically driven dominant driver. The biological driver is further split into a “productivity” or “respiration” signal. They find that mesoscale variability can explain 6-7% of global flux variance in O2 and CO2, with mesoscale eddy dominated regions have much higher variance. Biologically dominated regions covered a similar area then solubility driven areas, but these biologically dominated regions had a greater influence due to the greater gas flux. Within their framework, they also assess the potential changes to these mesoscale regimes under a climate change scenario, up to a doubling of CO2 concentrations in the atmosphere. Under this scenario the solubility driven regions expand, which increases their relative contribution to the total variance. Overall, I find the manuscript to be well written and results of great interest to understanding the role of mesoscale eddies in modifying gas fluxes. I have some comments/suggestions that the authors should consider.
L94: I would suggest adding what a positive anomaly in the air-sea flux means within the methods explicitly. i.e positive anomaly indicates ingassing
L96: What is the native time resolution of the model? I understand the outputs analysed are monthly, and it’s clear there are also daily outputs based on the appendix figures, but the native resolution would be good to see here.
L112: I would suggest the authors add information on the specific gas transfer parameterisation that was used for the CO2 and O2 fluxes. The magnitude of the results would change if the parameterisation of the gas transfer coefficient were changed; for example if Nightingale et al. (2000) were used compared to Wanninkhof (2014).
L114-123/L407-420: I can see the need for MiniBLING in this computationally expensive model and appreciate the authors highlighting of the limitations in these sections. I would however suggest that some mention of the simplified nutrient setup (i.e 1 varying nutrient) with a climatology of iron could lead to a simplified response of the biology within the model. For example Song et al. (2016) highlight the contrasting impacts of mesoscale eddies on the iron within the Southern Ocean, where iron is the limiting nutrient for phytoplankton. With MiniBLING this dynamic is likely missing from the productivity estimates.
L268: I would suggest rephrasing this sentence, as the similarly suggests that they follow a similar latitude pattern. But I think you mean that they have a similar maximum variance accounted for, but these occur in different regions for the 2 gases.
L325: I would suggest, the information in the last sentence could be inserted in the sentences within the paragraph. i.e L322 could include the 7 and 10% values after biological imprint.
L386: I would highlight that the work of Ford et al. (2023) has been expanded to the global scale in Ford et al. (2026), and they find a global net effect of long-lived mesoscale eddies of 2.7 ± 1.1 Tg C yr-1. They also find a general asymmetry in the CO2 flux for anticyclonic and cyclonic eddies but of differing magnitudes regionally. I would suggest including this updated global analysis within the manuscript.
References
Ford, D. J., Tilstone, G. H., Shutler, J. D., Kitidis, V., Sheen, K. L., Dall’Olmo, G., & Orselli, I. B. M. (2023). Mesoscale Eddies Enhance the Air‐Sea CO 2 Sink in the South Atlantic Ocean. Geophysical Research Letters, 50(9), e2022GL102137. https://doi.org/10.1029/2022GL102137
Ford, D. J., Shutler, J. D., Sheen, K. L., Tilstone, G. H., & Kitidis, V. (2026). UEx-L-Eddies: decadal and global long-lived mesoscale eddy trajectories with coincident air–sea CO2 fluxes and environmental conditions. Earth System Science Data, 18(2), 969–988. https://doi.org/10.5194/essd-18-969-2026
Nightingale, P. D., Malin, G., Law, C. S., Watson, A. J., Liss, P. S., Liddicoat, M. I., et al. (2000). In situ evaluation of air-sea gas exchange parameterizations using novel conservative and volatile tracers. Global Biogeochemical Cycles, 14(1), 373–387. https://doi.org/10.1029/1999GB900091
Song, H., Marshall, J., Munro, D. R., Dutkiewicz, S., Sweeney, C., McGillicuddy, D. J., & Hausmann, U. (2016). Mesoscale modulation of air-sea CO2 flux in Drake Passage. Journal of Geophysical Research: Oceans, 121(9), 6635–6649. https://doi.org/10.1002/2016JC011714
Wanninkhof, R. (2014). Relationship between wind speed and gas exchange over the ocean revisited. Limnology and Oceanography: Methods, 12(JUN), 351–362. https://doi.org/10.4319/lom.2014.12.351