Markus Adloff and colleagues assess the sensitivity of the AMOC to changes in radiative forcing in the intermediate complexity model Bern3D. The range of the radiative forcing is representative of the last 800 kyr. The radiative forcing comprises orbital forcing, greenhouse gases, ice-sheet induced albedo changes and dust forcing. The strength of the radiative forcing is scaled by the maximum dust forcing at the LGM. The authors identify four stable AMOC states, a strong interglacial state, a weak glacial state and two less stable intermediate states. The magnitude of the radiative forcing determines the time the AMOC spends in each of the respective states. The authors analyse the characteristics of the AMOC states and assess the underlying mechanisms through further more idealised simulations. Comparison with available proxy data for sea surface temperature, AMOC strength and climate variability indicate that the simulations contain realistic AMOC behaviour (depending on the forcing strength) and that valuable insights on thermal AMOC thresholds throughout the glacial cycles can be obtained from them.

Overall the paper is of high quality, well written and definitely of interest for a wide audience in the CP community and beyond. Testing for thermal AMOC thresholds in itself is not new, but the length of the simulations and the large covered range of forcing scenarios that can only be achieved through the intermediate complexity of Bern3D provide enough novel insights.

My comments are mostly minor, asking for more clarification or context. I recommend publication after minor revisions.

---------------------------------------------------------------------------------------------------------------------

Minor Comments:

Page 2:

l.3-4: This is the first time future AMOC stability is mentioned. It may be worthwhile to add a few sentences linking past and future AMOC stability.

l.45: It would also be helpful to provide a bit more context on thermal thresholds. The previous two paragraphs mostly talk about the haline part (i.e. surface freshwater input and salinity redistribution). Which models have been used to analyse thermal AMOC thresholds and for which climate states? And could you comment on whether the AMOC in intermediate complexity models tends to be more or less or similarly stable as in fully coupled earth system models (e.g. the AMOC in ocean-only models is known to be more prone to instabilities than in coupled GCMs).

Page 4:

l.16-17: Can you briefly explain why it is a useful approximation to use the LR04 stack as a scaling for the dust radiative forcing?

Page6:

Fig1. Could you also show the combined radiative forcing of all three forcings? That would make it easier to identify periods of changing radiative forcing.

Page7:

l.24-26: How do you assess stability here?

Page8:

Fig.3 could you rotate the maps in the upper panels by 45°, so that the perspective on the North Atlantic becomes more easily comparable to the lower panels?

l.22 and 26: Do you show stratification? The lower panles of Fig.3 only show surface density changes. Would it make sense to show stratification? Or do you infer increased stratification simply because of the lighter surface waters?

Page 9:

l.5-10: I found this paragraph difficult to read and follow. If none of the differences is statistically significant, would it not be sufficient to report that MBT has no statistically significant effect on the AMOC response?

Page 11:

l.16-19: Does this refer to Fig.5 d? And in general: more specific references to Fig.5 could be made though out this page, to make it easier to follow. It is not always clear  whether the text on this page refers to Fig.5, some other Figure or to results not shown.

l.20-21: Can you name the two processes and timescales explicitly? I guess they are N.Atl. freshwater changes (fast) and AABW propagation (slow), but it would be good to have them spelled out.

Page 12:

l.20-24: I do not really see the further reduction in NADW export. To me, the distributions of all three water masses look almost identical at 23 and 24.5 kyr. Also, I do not really see NADW replacing AAIW, the upper NADW boundary does not seem to change and if anything, the southward extent of NADW also decreases.

Page 13:

l.25: What about the strong variance at 6kyr?

Page 14:

l.32-40: Is this part meant in contrast to other models? The last sentence is also almost impossible to follow. Please consider a clearer formulation.

Page 16/17:

Meta stable AMOC modes: How are the excitable/metastable states defined? By increased AMOC variance as in Fig.5? How do the metastable states relate to the four AMOC states I-IV from the beginning? Also, please consider adding the corresponding kyrs behind MIS3/4/5e etc, so that it is easier to identify the right parts of Fig.9 for those readers who do not have those numbers at the top of their heads.

Page 18:

l.24-27: This would be very interesting indeed. I look forward to the follow-up :)

----------------------------------------------------------------------------------------------------------------------------

Technical/Editorial Comments:

Page 1

l.34: delete “boundary” after “Atlantic”

l.36-42: Very long and hard to read sentence. Consider reformulating for better readability. Also: which climate is being referred to at the end of the sentence? Probably North Pacific climate but it is not immediately clear.

l.43-47: same as comment above. Also: Does the last half sentence (“and by modulating atmospheric greenhouse gas concentrations”) still correctly belong to the beginning of the sentence (“It influences deep ocean nutrient and oxygen concentrations”)?

Page 2:

l.2: “which had regional [...]” instead of “and had regional [...]”

Page 4:

l.13: What is meant with “rest of the past 800kyr”? Rest with respect to what? The spin up state?

Page 10:

Fig.5: Please increase the font size for better readability.

Page 12:

Fig.6: Please increase the font size for better readability.

Page 14:

l.23: The name of the ocean model is COCO (the ocean component of MIROC)

Page 15:

Fig.7: Please increase the font size for better readability.

Page 16:

l.28: wrong Figure reference? Should be Figure 1?

Page 18:

l.14: delete “but”

see the PDF

Summary

This study investigates the thermal thresholds of AMOC over the last 800kyr in the Bern3D climate model. By controlling the amplitude of the global cooling via dust forcing, they extract the timing of occurrences of the thermal threshold of AMOC. In addition, they manage to extract four stable and meta-stable AMOC modes, which differ in the locations of NADW formation regions and sea ice extent over the North Atlantic. The paper further explores mechanisms causing the AMOC weakening due to the reduction in the radiative forcing by means of hysteresis experiments. At the end of the paper, the authors provide a discussion on the limitations of the study by not including the orographic effect of the ice sheet in their simulations.

The experiments and results presented in this paper are very interesting (especially figs 2 and 9!) and are of interest of readers of the Climate of the Past. Therefore, I think this paper should be published. However, the explanation of the mechanism of the AMOC changes appears unclear to me (also noted by other reviewers). Additionally, there are some ambiguity in the experimental setup. Addressing these points would improve the overall quality of the manuscript. The comments are summarized below.

Main comments:

1. Mechanisms by which the reduction of radiative forcing weakens the AMOC: Other reviewers have pointed out many points, so I’ll just list potential ways to improve the manuscript. 1. Separate the paragraph explaining the effect of the Southern Ocean and North Atlantic. Section 3.2 goes back and forth between the role of NA and Southern Ocean. This makes it hard to follow the discussion. Related to this, Buizert and Schmittner (2015) provides a nice summary on the role of Southern Ocean. Ando and Oka (2021, GRL) also gives useful insight on the role of sea ice and heat transport on the stability of the AMOC. These two studies further performed hysteresis experiments with freshwater forcing. While the way of hysteresis experiment is not the same as in this study, I feel these studies should be cited and included in the discussion of the mechanism. 2. Use Fig. 3 to help explain the mechanism. For example, it would be more convincing for me if the authors explain the mechanism in the following manner “reduction of radiative forcing first weakens the convection in the Labrador Sea (Fig. 3) by increasing transport of sea ice from the arctic and by reducing the northward heat transport (Fig. 4). However, intensified surface cooling initiates the deepwater formation close to UK (Fig. 3), causing a shift of the AMOC into the second phase. Further reduction in radiative forcing ….” Obviously this is not a perfect example but please consider modifying the manuscript in this way. 3. Relation of heat transport and the AMOC is alway tricky. They vary together and also the heat transport can either weaken or strengthen the AMOC depending on the background condition (e.g. Paul and Schulz 2002, https://doi.org/10.1007/978-3-662-04965-5_5, Ando and Oka 2021, GRL). Please cite these paper when discussing the effect of heat transport on AMOC and explain why it should work in that sense.
2. Experimental setup: I think the authors need to explain why they decide to vary the magnitude of the dust related radiative forcing but not others in their sensitivity experiments (I’m not saying that’s bad!). I don’t fully understand how this model works, but isn’t there another way to do similar experiments, e.g. changing the magnitude of the emissivity of the atmosphere or the magnitude of the ice sheet related radiative forcing? Effect of dust forcing is of course uncertain, but so are others (Tierney et al. 2020).
3. Related to 2, another question I have is that “Does the radiative forcing by dust affect the global and local temperatures in the same way as the GHG do in this model?” Looking at results from GCMs (e.g. Kawamura et al. 2017 Science Advances, Ohgaito et al. 2018 CP), it is shown that GHG and dust affect the local temperatures in a different way. This information is important especially when we want to use the insight from this study to better understand results of AOGCMs.
4. NADW formation in Norwegian/Greenland sea: This might be related to comments from other reviewers, but some previous studies have suggested the importance of cessation/resumption of convection over the Norwegian Sea when considering the thermal threshold of the AMOC (Oka et al. 2012). Please describe this feature in the Introduction and add some discussion wherever appropriate.

Specific comments:

P1L25-26: Given the limitations in the model, I think it would be safe to add “in this model” at the end of the sentence.

P1L31-33: Isn’t this the other way round; relatively salty water gets cooled by the atmosphere, the vertical density gradient weakens, and the water sinks and forms the NADW.

P2L34: Perhaps “sensitive” -> “dependent”?

P3L13-25: So many references are missing in this paragraph. Please add the appropriate reference for each sentence. (e.g. references for Bern3D model, references for freshwater flux corrections)

P4L11-13: How did you define the maximum ice extent? Is it from the LGM?

P5L23-24: Better to say “stable”->”monostable”, “unstable”->”bistable” here.

P11L11-15: I could not understand this sentence. Can you further elaborate on this, please?

P13L4-5: Not quite sure what this positive feedback means here. In general, a surface cooling will reduce the SST and hence increase the surface density while the cooler SST reduces evaporation and causes a reduction in surface salinity and surface density. So isn’t it a negative feedback?

P13L6: Isn’t the sea ice feedback a positive feedback?

Figs.2 and 9: Very nice figures.