Climate and Carbon Cycle Responses to a 21st century AMOC Collapse under a 2 &deg;C Stabilization Pathway

Frölicher, Thomas L.; Maier, Patrick; Burger, Friedrich A.; Silvy, Yona; Swingedouw, Didier; Hofmann Elizondo, Urs; Berger, Manon

doi:10.5194/egusphere-2026-3065

Preprints

https://doi.org/10.5194/egusphere-2026-3065

Preprints

24 Jun 2026

| 24 Jun 2026

Climate and Carbon Cycle Responses to a 21st century AMOC Collapse under a 2 °C Stabilization Pathway

Thomas L. Frölicher, Patrick Maier, Friedrich A. Burger, Yona Silvy, Didier Swingedouw, Urs Hofmann Elizondo, and Manon Berger

Abstract. The Atlantic Meridional Overturning Circulation (AMOC) is a key component of the climate system, yet the climate and carbon cycle responses to a collapse under emission pathways consistent with the Paris Agreement remain poorly understood. Using the comprehensive GFDL ESM2M Earth System Model with the Adaptive Emissions Reduction Approach, we impose a freshwater-induced strong AMOC weakening to 20 % of its preindustrial strength, initiated in year 2026 and achieved within 60 years. The counterfactual simulations without freshwater hosing otherwise follow a pathway in which global warming stabilizes at 2 °C and the AMOC weakens only modestly and partially recovers. Relative to the 2 °C scenario without AMOC collapse, a strong AMOC weakening cools global mean surface air temperature by −0.8 °C (5-member ensemble range: −0.7 to −0.9) by 2171–2200, offsetting 40 % of global warming. Pronounced cooling is simulated in the North Atlantic region, reaching up to -5.4 °C (-8.0 to -3.3) in winter over Iceland relative to 1861–1900 conditions. The global cooling is primarily driven by larger negative feedback from clouds, driven by an increase in low-level clouds in the North Atlantic region, with smaller contributions from enhanced global ocean heat storage and reduced atmospheric CO₂. The total ocean heat content increases by an additional 488 ZJ (442–531), primarily south of 20° N, associated with reduced northward heat transport and enhanced heat uptake in the North Atlantic. The additional heat increases global thermosteric sea level rise by an additional 10 % (8–12), with enhanced rise in the western and tropical North Atlantic and northern Indian Ocean, but pronounced reductions in the eastern North Atlantic. Atmospheric CO₂ declines by 13 ppm due to anomalous land carbon uptake of 44 GtC (33–53), dominated by enhanced carbon storage in the Amazon region under cooler and wetter conditions. In contrast, global ocean carbon storage decreases by 14 GtC, mainly north of 20° N. The AMOC-induced cooling temporarily breaks the near-linear relationship between cumulative CO₂ emissions and warming, increasing the remaining emission budget for limiting warming to 2 °C by 63 % (54–72). Compared to identical freshwater forcing under preindustrial conditions, the surface temperature, ocean heat content, and sea-level responses to an AMOC collapse are substantially damped in a 2 °C world, indicating reduced climate sensitivity to AMOC collapse in a warmer world. These results demonstrate that a strong AMOC weakening would profoundly alter the climate–carbon cycle system and underscore the importance of explicitly accounting for AMOC risks in long-term climate assessments.

Received: 28 May 2026 – Discussion started: 24 Jun 2026

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.

Download & links

Thomas L. Frölicher, Patrick Maier, Friedrich A. Burger, Yona Silvy, Didier Swingedouw, Urs Hofmann Elizondo, and Manon Berger

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-3065', Amber Boot, 01 Jul 2026
In this paper the authors investigate the climate and carbon cycle impacts of an AMOC collapse under climate change in an Earth System Model. Where most studies use pre-industrial conditions and don’t use an interactive carbon cycle, this study does both which adds to the existing literature. I find the paper well written and the analysis of the results robust. It is already a very nice paper that was nice to read. I have some minor comments that can help clarify parts of the paper, and one larger concern that I’d like to see addressed in the authors’ response.
Major concern:
The paper does not address whether carbon and alkalinity are conserved in the model, and to what extent the total carbon content of each model simulation differs. I would like the authors to comment on the issue. I think it’s very unlikely that it will change the results and the conclusion in a major way, but I think it is important to mention this, and discuss the differences between the model setups. Specifically, between the 2C-ref and the 2C-hos-Eref simulations.
Minor comments:
Line 59/60: I think it would be good to cite Boot et al. (2024) also in line 59 and/or line 60.

Line 77: ‘However, their analysis …’ How I think I should understand the sentence here, I think that’s correct. The sentence is a bit confusing since Boot et al. (2024) did use emission-driven simulations. I think the sentence can also be read as if they didn’t. I suggest to clarify the sentence.

Section 2.1: I suggest extending the model description with respect to TOPAZv2.0 and LM3.0 to include a short description of processes relevant for section 3.5.

Line 129: I suggest adding what the (approximate) absolute threshold values for an AMOC collapse are in this model.

Line 130: ‘With the 2C-ref experiment …’ I suggest making clearer that the AMOC collapse signal becomes apparent from comparing the two simulations (2C-ref and 2C-hos-Eref).

Line 145: I’m not sure what Is meant with ‘… or global salinity drifts.’

Line 159: I suggest adding references here for the model biases (they are already somewhere else in the manuscript).

Line 253: What is the reason for the shift into regime 3? From the feedbacks it appears changes in cloud cover. Do you know why this happens? I think it would make for a nice addition to the paper if this could be explained in more detail.

Line 302: Is the cloud feedback different in other ESMs? And if so, would you expect different qualitative results for those models?

Line 319: I suggest referencing to Fig. 5 here.

Figure 5: I think it would be insightful to add the absolute numbers of the two simulations as well in the figure. And for clarity, I suggest adding a ‘+’ sign in front of the positive differences to highlight it’s a number representing a difference.

Figure 8: the land response in southern Finland is quite different from the response in the surrounding area. It is also qualitatively different compared to the results in Boot et al. (2024) in the same region. From the air temperature and precipitation I do not see a clear reason why this region would behave differently. Why does this region show this response?

Figure 9: Similar as for Figure 5.

Figure 9: For clarity it might be nice to add the word atmosphere above the ocean boxes so that we know the fluxes up there are air-sea exchanges without having to read the caption.

I suggest discussing Lerner and Romanou (2026), even though it’s still a preprint.

Figure A1: I suggest adding to the caption that the y-axis differs per subpanel.

Equations A5-A7: I suggest keeping the order of the feedbacks consistent.

Figure B2: I suggest adding a contour line at 0 Sv.

References:
Paul Lerner, Anastasia Romanou. Cascading Impacts of AMOC weakening and collapse on Marine Biogeochemistry, 08 May 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-9044876/v1]
Citation: https://doi.org/10.5194/egusphere-2026-3065-RC1
RC2: 'Comment on egusphere-2026-3065', Anonymous Referee #2, 06 Aug 2026

Summary:
The manuscript imposes a freshwater-forced AMOC collapse against the backdrop of an emission-driven 2°C stabilization pathway in GFDL ESM2M, using adaptive emissions to target a 2°C global warming level under different AMOC scenarios. Relative to the counterfactual, AMOC collapse strongly cools global mean temperature by 0.8°C, increases thermosteric sea level rise, lowers atmospheric CO2 through Amazonian land uptake, and raises the remaining carbon budget for staying below 2°C. Similar to previous literature, the authors find that all climatic responses to AMOC weakening are state-dependent and are, except for atmospheric carbon concentrations, moderated by the warmer background climate.
Overall assessment:
The simulation setup is carefully designed and executed and the adaptive emissions method is a real advance over concentration-driven hosing experiments; the decomposition of the global temperature signal and the heat and carbon budget decompositions are thorough and very insightful. My reservations concern framing and interpretation rather than the modelling itself. Two results that carry weight in the abstract and discussion, the seasonal cooling pattern in selected cities, and the TCRE slope change under AMOC collapse, are stated more confidently than corroborated by current figures and statistics. Additionally, the gap the paper sets out to fill is somewhat ambiguous. I recommend minor revisions to this already very nice paper.
I really appreciate the structured and clear writing of the paper, which made it easy to follow and pleasant to read.
Major points:
1. Winter amplification of AMOC-induced cooling
- "reveals a pronounced winter amplification of the AMOC-induced cooling" (L273)

- "Cooling is substantially weaker in summer" (L280)
In Figure 3f only Reykjavik shows a visually clear seasonal cycle in the anomaly. For London, Bern, Bergen and Beijing the seasonal contrast is modest relative to their mean cooling and comparable to the ensemble spread. The stronger statements above do not seem supported by Fig. 3f and I would like to see statistical tests for difference between winter and summer months if this general statement is to be maintained. Given the pronounced winter signal in Reykjavik and, in Fig. B4, the larger variability for Reykjavik, I am also curious about the influence of sea ice. Despite sea ice albedo effects being small in comparison to cloud effects, I still wonder whether some of the regional and seasonal temperature effects can be (partly) explained by sea ice. A map of March sea ice mean state in the 2171-2200 period might also help to visually see which land areas might be directly affected, especially in light of the results by van Westen and Baatsen (2025).
2. TCRE slope difference with a collapsed AMOC
"the slope in 2°C-ref is 1.00 °C (0.65 to 1.40) per 1000 GtC, which is slightly smaller than the corresponding slope in 2°C-hos-Eref of 1.25 (1.03 to 1.52) °C per 1000 GtC" (L389-390)
Looking at Fig. 10b does not visually suggest a different slope. The two confidence intervals also overlap substantially, yet the discussion builds on the difference: "AMOC-induced changes in ocean circulation modify to some extent the partitioning of heat and carbon uptake by the global ocean, thereby altering the transient warming response to cumulative CO2 emissions" (L466-468). Can the authors test whether the slopes are statistically distinguishable? Absent such a test, L466-468 should be marked as a conjecture.
An idea might also be to pool both experiments in a single regression that controls for AMOC strength, so that it would yield an overall TCRE conditional on the AMOC state, but this is just a suggestion.
Generally, the discussion statement "this near-linear relationship breaks down when the AMOC collapses" (L462) raises the question of whether this is the first known mechanism to break TCRE linearity or whether there are analogous processes that have been previously studied. The permafrost literature comes to mind, for instance Steinert and Sanderson (2025, ESD).
3. What exactly is the gap in the literature?
"A systematic assessment of the climatic and carbon cycle consequences of a collapsed AMOC under emission-driven warming is therefore still lacking." (L79-80)
I understand that the contribution of this paper is manifold, in that it has an interactive carbon cycle in an emissions-driven setup, looks at global AMOC impacts under ongoing non-idealised warming scenarios, and makes the genuinely novel link to TCRE and carbon budgets. Yet I found it a bit hard to parse which aspects of the contribution are considered the key innovations and would encourage a slight reformulation of this paragraph. After reading it several times, I appreciate that all the aspects of the contribution and the comparisons to other relevant papers are there, but I found it hard to understand during the first read.
Additionally, comparisons with Romanou et al. (2023) and Orbe et al. (2023) as well as with Nielsen et al. (2019) and Nian et al. (2026) could help situate this paper in the literature and clarify the gap that it fills.
Minor points:
L16, L19-20: The abstract and results disagree on two headline numbers. Land carbon uptake is 44 GtC (33-53) here but +49.6 GtC (44.5-59.9) at L353. The remaining budget increase is 63% (54-72) here but 60% (49-73) at L386.
L21-22: "reduced climate sensitivity to AMOC collapse in a warmer world" brings up connotations with ECS or TCR in an abstract. L403-404 already has the clearer wording, "a reduced sensitivity of the climate system to AMOC collapse under warmer background conditions".
L27: Several reference entries seem to have issues. "University of Potsdam and Rahmstorf, S." and "Griffies, S. M., NOAA, and Laboratory, G. F. D." parse institutions as authors; the AERA protocol appears as "Thomas, F., Jens, T., Fortunat, J., and Yona, S.", with given and family names swapped; I suggest checking all entries again.
L93, L104, L105: "Dunne et al. (2013)" is cited as the ESM2M description and carbon-cycle evaluation, but the reference list gives "Reductions in Labour Capacity from Heat Stress under Climate Warming". The intended paper is presumably Dunne et al. (2013), J. Climate 26, 2247-2267.
L157: "string model biases" should probably be "strong" or "existing".
L214: The choice of reported variables is not motivated. Sea level is analysed in detail while precipitation enters only through NPP. Relatedly, the gap statement promises "climatic and carbon cycle consequences" (L79-80), and it is not obvious how that maps onto the quantities actually analysed, which are the drivers of global temperature and thermosteric sea level. A sentence in the introduction stating which quantities are the focus and why would resolve this.
L256: "although with a delay because emission adjustments are based on the warming and cumulative emissions since preindustrial". Could this delay be avoided, and would the budget results change if AERA tracked the cooling more directly and returned to the reference pathway earlier? Since the 60% increase is a headline number, it matters how much of it is a property of the AMOC response and how much of the AERA implementation.
L263: Almost all results are reported for 2171-2200. The AMOC is already close to its collapsed state around 2100 (Fig. 2a), so the more decision-relevant outcomes for this century are available and should at least be briefly mentioned. Impacts in this century might be more useful for risk assessments, which are hinted at in the closing statement.
L265-266: the strongest anomalies "extend zonally into Western and Central Europe", but the next sentence introduces a separate over-land statement, which reads as if the first were not over land.
L276: "Reykiavik"
L279: Vienna is not among the cities shown in Figure 3 (Reykjavik, London, Bern, Beijing, Bergen), so the winter cooling attributed to it cannot be checked against the figure.
Fig. 10b: I really like this figure. I think it is highly important and very suggestive.
L409: The larger CO2 decrease under 2°C warming is attributed to different sensitivities of ocean and terrestrial carbon uptake, but I am wondering whether this could also reflect different equilibration timescales, since 200 years is still very transient for the ocean carbon cycle, in contrast to atmospheric responses.
L418, L447: Parsons et al. (2014, Geophys. Res. Lett. 41, 146-151) ran hosing experiments in GFDL ESM2M with LM3V and found Amazonian carbon storage rising by about 47% through the same mechanism described here. This is the same model and the same result and should be cited and compared. They also find a large offsetting loss in Central America, so it would be useful to report the tropical American decomposition regionally and to say whether that signal appears here. Agreement within one model does not establish robustness, and the manuscript should be explicit that the +44 GtC may partly reflect the ESM2M/LM3V response.
L435: The contrast is drawn with Boot et al. (2024), but Nian et al. (2026) find CO2 increasing by 47-83 ppm with about +0.2°C of additional warming, the opposite sign and an order of magnitude larger. That paper, along with Nielsen et al. (2019) and Parsons et al. (2014), is worth engaging with directly, especially on the South American land carbon response.
L558: The author contributions credit the temperature decomposition to "DW", which matches none of the author initials. Presumably it should be DS?
References:
Boot, A. A., von der Heydt, A. S., and Dijkstra, H. A. (2024). Response of atmospheric pCO2 to a strong AMOC weakening under low and high emission scenarios. Climate Dynamics, 62, 7559-7574. https://doi.org/10.1007/s00382-024-07295-y
Dunne, J. P., John, J. G., Shevliakova, E., Stouffer, R. J., Krasting, J. P., Malyshev, S. L., Milly, P. C. D., Sentman, L. T., Adcroft, A. J., Cooke, W., Dunne, K. A., Griffies, S. M., Hallberg, R. W., Harrison, M. J., Levy, H., Wittenberg, A. T., Phillips, P. J., and Zadeh, N. (2013). GFDL's ESM2 global coupled climate-carbon Earth system models. Part II: Carbon system formulation and baseline simulation characteristics. Journal of Climate, 26, 2247-2267. https://doi.org/10.1175/JCLI-D-12-00150.1
Nian, D., Willeit, M., Wunderling, N., Ganopolski, A., and Rockström, J. (2026). Collapse of the Atlantic meridional overturning circulation would lead to substantial oceanic carbon release and additional global warming. Communications Earth & Environment, 7, 295. https://doi.org/10.1038/s43247-026-03427-w
Nielsen, S. B., Jochum, M., Pedro, J. B., Eden, C., and Nuterman, R. (2019). Two-timescale carbon cycle response to an AMOC collapse. Paleoceanography and Paleoclimatology, 34, 511-523. https://doi.org/10.1029/2018PA003481
Orbe, C., Rind, D., Miller, R. L., Nazarenko, L. S., Romanou, A., Jonas, J., Russell, G. L., Kelley, M., and Schmidt, G. A. (2023). Atmospheric response to a collapse of the North Atlantic circulation under a mid-range future climate scenario: A regime shift in Northern Hemisphere dynamics. Journal of Climate, 36, 6669-6693. https://doi.org/10.1175/JCLI-D-22-0841.1
Parsons, L. A., Yin, J., Overpeck, J. T., Stouffer, R. J., and Malyshev, S. (2014). Influence of the Atlantic Meridional Overturning Circulation on the monsoon rainfall and carbon balance of the American tropics. Geophysical Research Letters, 41, 146-151. https://doi.org/10.1002/2013GL058454
Romanou, A., Rind, D., Jonas, J., Miller, R., Kelley, M., Russell, G., Orbe, C., Nazarenko, L., Latto, R., and Schmidt, G. A. (2023). Stochastic bifurcation of the North Atlantic circulation under a midrange future climate scenario with the NASA-GISS ModelE. Journal of Climate, 36, 6141-6161. https://doi.org/10.1175/JCLI-D-22-0536.1
Steinert, N. J., and Sanderson, B. M. (2025). Normalizing the permafrost carbon feedback contribution to the Transient Climate Response to Cumulative Carbon Emissions and the Zero Emissions Commitment. Earth System Dynamics, 16, 1711-1721. https://doi.org/10.5194/esd-16-1711-2025
van Westen, R. M., and Baatsen, M. L. J. (2025). European temperature extremes under different AMOC scenarios in the Community Earth System Model. Geophysical Research Letters, 52, e2025GL114611. https://doi.org/10.1029/2025GL114611

Citation: https://doi.org/10.5194/egusphere-2026-3065-RC2
RC3: 'Comment on egusphere-2026-3065', Andreas Schmittner, 11 Aug 2026

This manuscript describes simulations with the GFDL model of impacts of an AMOC collapse on climate and carbon cycle. The paper is very well written and illustrated and contributes new information on an important topic. I think it could be published with minor revisions.
Line 100: I’d like more information on the terrestrial carbon model, specifically if vegetation is dynamic such that plant functional types can change, or if it is static. Also, can vegetation adapt to climatic changes or not?
Line 358: how is land-use affected by AMOC changes?
Line 363: what temperature dependency of NPP is used in the model? Could this result depend on model parameters or structure? Is vegetation able to change and/or adapt?
Line 367: why is northward carbon transport reduced? Is this dominated by anthropogenic carbon? Is there northward carbon transport in the preindustrial? I'd expect that the AMOC transports DIC southward since NADW is colder and thus more carbon rich than warm upper ocean waters.
Lines 409-410: why is this not quantified and reported? The results should be available.
Line 495: Intermediate complexity models confirm long (multi-millennial) timescales of ocean carbon equilibration affecting CO2 after AMOC collapse, e.g. Schmittner and Boling (2025).

Schmittner, A., & Boling, M. (2025). Impact of Atlantic Meridional Overturning Circulation Collapse on Dissolved Inorganic Carbon components in the ocean. Global Biogeochemical Cycles, 39, e2025GB008526, https://doi.org/10.1029/2025GB008526.

Citation: https://doi.org/10.5194/egusphere-2026-3065-RC3

Thomas L. Frölicher, Patrick Maier, Friedrich A. Burger, Yona Silvy, Didier Swingedouw, Urs Hofmann Elizondo, and Manon Berger

Viewed

Total article views: 355 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
276	69	10	355	4	5

HTML: 276
PDF: 69
XML: 10
Total: 355
BibTeX: 4
EndNote: 5

Views and downloads (calculated since 24 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	167	56	7	230
Aug 2026	109	13	3	125

Cumulative views and downloads (calculated since 24 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	167	56	7	230
Aug 2026	109	13	3	125

Viewed (geographical distribution)

Total article views: 249 (including HTML, PDF, and XML) Thereof 249 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Aug 2026

Short summary

We simulate climate and carbon-cycle responses to a strong AMOC weakening under a 2 °C stabilization pathway. A collapse to 20 % of preindustrial strength cools global climate by 0.8 °C and winter climate in Northern Europe by up to 5 °C, enhances ocean heat uptake and sea-level rise, and alters land and ocean carbon storage. The cooling temporarily increases the remaining carbon budget for limiting warming to 2 °C.


Total:	0
HTML:	0
PDF:	0
XML:	0