| 21 Oct 2026
A Storm-driven Reversal of the Springtime North Atlantic CO2 Sink
Abstract. The North Atlantic Ocean is a major sink for atmospheric carbon dioxide (CO2), particularly in spring when phytoplankton blooms intensify CO2 uptake. Its magnitude and variability affect how much anthropogenic CO2 is sequestered. Yet short-term variability in this sink remains poorly constrained. Energetic events such as storms are rarely sampled, leaving their contribution to air-sea CO2 exchange unresolved. Here we report a reversal of this sink during spring 2023, captured by validated measurements aboard a racing sailboat crossing a storm. Surface ocean CO2 increased abruptly, reaching on average 34 µatm above atmospheric levels (magnitude comparable to the ~40 µatm undersaturation that typically sustains spring uptake) as strong winds drove vertical mixing and Ekman suction that ventilated an unusually shallow carbon-rich subsurface reservoir. The anomaly was further amplified by a concurrent marine heatwave. Our observations provide evidence that storms can ventilate subsurface carbon reservoirs and temporarily reverse CO2 fluxes even during peak biological drawdown. They reveal a mode of short-lived but intense carbon outgassing that is unresolved in coarse resolution models. As storms and marine heatwaves intensify under climate change, such events will become an increasingly important component of the ocean carbon cycle and may weaken the North Atlantic CO2 sink.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Biogeosciences.
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.
Olivier et al. present an interesting study using high-frequency surface ocean fCO2 measurements collected aboard a racing sailboat in the North Atlantic during a storm. The authors report positive air-sea CO₂ fluxes (outgassing) during the event, contrasting with the strong CO2 sink typically observed in the subpolar North Atlantic during spring.
By combining these in situ observations with ocean reanalysis and climatological data, the authors propose that storm-driven vertical mixing and Ekman suction ventilated a shallow subsurface carbon-rich reservoir, with the 2023 marine heatwave further contributing to the outgassing.
I agree with the authors that the ocean response to short-lived extreme events, such as storms, remains poorly characterized. Understanding these processes is particularly relevant in the context of climate change and potential increase in the frequency and intensity of extreme weather events. In this context, the dataset presented here is valuable, and the use of sailing vessels and other moving observing platforms have the potential to improve our understanding of air-sea CO₂ exchange during such events, for which data remain scarce.
Overall, I also agree with the authors' general hypothesis: the observed changes in air-sea CO2 exchange likely resulted from storm-driven processes, and similar events could potentially make an important contribution to regional carbon flux variability. However, I have some qualms regarding the strength of the evidence supporting the proposed mechanisms. In particular, several of the conclusions rely primarily on indirect estimates and climatological datasets rather than contemporaneous observations of subsurface biogeochemical and physical conditions. I also believe that certain methodological aspects require further clarification.
Major comments
1. Additional observational evidence supporting a subsurface source of the observed CO2 outgassing
While I acknowledge the limited availability of in situ observations, it is important to take into account that the proposed mechanism relies heavily on climatological data.
For instance, a key component of the authors hypothesis is based on the GLODAPv2 climatology, which represents the average biogeochemical conditions from a substantially earlier period (most data likely from a few decades ago) than the May 2023 event. Although these climatological fields provide valuable information on the background biogeochemical conditions, their representativeness of the upper ocean during May 2023 remains uncertain.
In particular, the upper-ocean carbonate system may exhibit substantial seasonal and interannual variability. Therefore, using more recent observations, preferably representative of springtime conditions, would strengthen the proposed interpretation.
Furthermore, assuming that XLD > MLD necessarily implies the entrainment of carbon-rich subsurface waters is not entirely correct. Nicholson et al. (2022) derived both MLD and XLD from glider observations, providing more direct constraints and potentially lower uncertainty than the methods used in the present manuscript. Even with these observations, they demonstrated that XLD exceeding MLD is not sufficient to ensure the entrainment of carbon-rich waters.
In particular, Nicholson et al. (2022) documented a strong wind event during which XLD exceeded MLD but did not reach the subsurface carbon reservoir, resulting in a limited biogeochemical response. They therefore emphasized that the relevant criterion was whether XLD reached the maximum winter mixed-layer depth (MLDmax), which defined the upper boundary of the subsurface DIC-rich reservoir in their study, rather than simply exceeding the instantaneous MLD.
Consequently, the comparison between XLD and the instantaneous MLD in the present manuscript may not be sufficient to demonstrate the proposed entrainment mechanism. The authors should consider whether XLD reached the depth of the subsurface carbon-rich reservoir, rather than relying solely on XLD exceeding MLD.
I suggest considering the following:
- Consider using the recently released GLODAPv3 observational dataset to assess whether more recent measurements support the vertical biogeochemical structure presented in the manuscript. Where sufficient observations are available, particular attention should be given to springtime conditions, especially April-May.
- BGC-Argo profiles could provide complementary observational evidence of the subsurface biogeochemical structure. For example, float WMO 4903365 collected a profile on 25 May 2023, coinciding with the storm event and also profiles before and after it. Other floats in the region and profiles from comparable periods in previous years could also be considered.
- Try to identify the depth of the carbon-rich reservoir from the most recent observations available in GLODAP and BCG-Argo and compare XLD with the maximum winter MLD.
2. Distinguishing spatial from temporal variability along the sailboat transect
Given that the sailboat was continuously moving across space and time, I believe it is important to distinguish between spatial and temporal variability in the observed fCO2 signal.
The sailboat crossed a dynamically complex region in which horizontal gradients associated with fronts, eddies, and different water masses may also contribute to changes in surface fCO2. This it must be clearer that the observed increase in fCO2 resulted primarily from storm-induced vertical mixing and not from spatial variability encountered along the transect.
Satellite SST, SSH and geostrophic currents observations could help characterize the oceanographic conditions during the event. These data could also help investigate whether the cyclonic circulation inferred from climatological fields in Figure 3 was present during May 2023. Also, SST before and after the storm would give an idea if it the SST field behaviour are consistent with storm-induced cooling or with the crossing of different water masses.
3. Correlation between wind forcing and surface fCO2
The authors report a strong correlation between wind forcing and surface fCO2 (r=0.8) However, several methodological aspects require clarification:
First, why was a 26h moving window selected? Considering that the main storm event lasted approximately two days (or the sailboat crossed the storm during that period), this window represents a substantial fraction of the event duration. It would be useful to justify this choice and assess whether the results are sensitive to the window length.
Second, temporal autocorrelation is likely to be important in both wind forcing and surface fCO2 measurements. This should be considered when assessing statistical significance.
Third, as mentioned above, the observations were collected along a moving transect, meaning that spatial gradients and temporal changes are not independent. Therefore, a strong correlation does not necessarily establish a direct causal relationship between wind intensification and the observed fCO2 increase.
Finally, I would expect that the response of surface fCO2 to wind-driven vertical mixing could involve a temporal lag. Was any lagged correlation analysis performed?
4. Regional extrapolation of the potential CO₂ outgassing
I believe the authors should be more cautious when extrapolating the observed CO2 outgassing to the regional scale.
An important consideration is that the area affected by a storm does not necessarily correspond to the area experiencing positive CO2 fluxes. As suggested by the authors themselves, the observed outgassing may depend on specific subsurface carbonate gradients, stratification, and circulation features that are unlikely to be uniform across the entire storm footprint.
There also appears to be an inconsistency in the units used for the flux calculations. Figure 4c presents air-sea CO2 fluxes in mol m-2 yr-1, whereas the extrapolation in lines 196-199 uses a mean flux anomaly of approximately 1.5 mol m-2 day-1. Is the unit in Figure 4c incorrect? If the annual units shown in the figure are correct, the estimated carbon release of 5-50 TgC per event would be substantially overestimated.
Minor comments
Lines 48, 164: The authors repeatedly refer to the event occurring during the peak of the spring phytoplankton bloom, although line 101 instead mentions the tail end of the spring bloom. However, no observational evidence, such as chlorophyll-a data, is presented to support these statements.
Spring bloom phenology in the North Atlantic can be highly variable in terms of timing, duration, and magnitude, particularly during anomalous years such as 2023, which was characterized by an intense marine heatwave. I would suggest examining satellite chlorophyll-a observations before and during the event to better characterize the bloom conditions and support these statements.
Lines 105-106: The presence of a climatological cyclonic recirculation does not necessarily demonstrate that this feature had the same position or intensity during May 2023. As suggested above, contemporaneous satellite altimetry could provide useful additional evidence.
Figure 4: The estimated Ekman pumping velocities appear to correspond to vertical displacements of only a few metres over the duration of the event (based on Figure 4). While these displacements may contribute to the proposed mechanism, particularly in combination with wind-driven mixing, it is unclear how important Ekman pumping was relative to direct turbulent entrainment.
Appendix A1, line 244: The climatological reference period is given as 1986-2023, whereas the main text (line 71) mentions 1982-2023. Please clarify which period was used.
Appendix A3: What is the behaviour of XLD
Appendix A5: The authors state that the moving correlation was calculated using wind magnitude, whereas the caption of Figure 1 refers to wind stress. Please clarify which variable was used in the analysis.