Author Comment (AC) on acp-2021-877

The authors study the efficacy of different geoengineering on ameliorating the AMOC reduction under GHGs forcing using ESM simulations. While I suspect the author’s analyses were constrained by what’s available in the GeoMIP output, could you explain why G1 and G1oa were used to counter 4xCO2 forcing whereas G4 and G4cdnc were to counter RCP4.5 scenario? The authors are fully aware that GHG forcing in 4xCO2 and RCP4.5 is very different, and the geoengineering forcing strength is also different between G1, G1oa, G4, G4cdnc. These differences render the comparison across G1/G1oa and G4/G4cdnc somewhat arbitrary, and this is true whether you are talking about an absolute anomaly (e.g., table 2), or a ratio (as in equation 3), or ratio’s ratio (as in equation 4). But, if it has to be done this way, you should provide more justification and/or motivation. Alternatively, you can compare G1 with G1oa, and G4 with G4cdnc without the crossgroup comparisons. The presentation is otherwise generally clear, except for a few places (see specific comments below).

forcing due to the representative concentration pathway 4.5 (RCP4.5) scenario are partly offset. These experiments are explained in the following table, which we add in the text. The comparisons between G1oa and G1 separately, and G4 and G4cdnc are done throughout the paper, but this is a little elementary. Taking the ratios is the only way of compensating for the differing signal strengths, as the applied forcing is much larger in the abrupt4xCO2 scenario and the geoengineering scenarios associated with it (G1 and G1oa) than in the RCP4.5 scenario and its associated geoengineering scenarios (G4 and G4cdnc). Taking the ratio of ratios then allows us to compare across the differing geoengineering methods, which is really the point of the paper. Ratios are a standard way of normalizing results with different forcing, for example in medicine: Curran-Everett 2013 Adv Physiol Educ. 37(3):213-9. https://doi.org/10.1152/advan.00053.2013). The only arbitrary choice here is the selection of TOA radiation as the metric used to quantify the strength of the forcing, another choice might have been mean global mean temperature, but we think that TOA radiation was a little more fundamental since it is used in the definition of model climate sensitivity. The following has been added to the manuscript to clarify these choices: In the following analysis, we make comparisons between G1oa and G1, and G4 and G4cdnc separately as they do not use the same greenhouse gas forcing backgrounds (Table 2). But we are also interested in comparing the different geoengineering types and doing this can be done with the ratios of their response, e.g. . The different ESM also have different climate sensitivities, and we also account for this by considering their top of atmosphere radiative forcing (TOA).

Specific Comments
Comment No.1: Line 187-188, "Generally, mitigation of AMOC weakening under G4cdnc is more than with G4, but weaker than G1 solar dimming": But mitigation of G1 solar diming was applied to 4xCO2 not RCP4.5, so this comparison is not apples-to-apples.
Reply: Yes, as we do say in the same sentence immediately following the part the referee quotes: "but these scenarios were not designed to have identical forcing, so we shall discuss their relative efficacy later in the Discussion." This is because the level of forcing applied under the G4 scenarios is weaker than under the G1, as is clear from the new Table 2 shown above. We need to use the normalized ratios to make cross-group comparisons of the efficacy of the different types of geoengineering, and for ranking of these scenarios. The motivation for this cross-group comparison comes from Ji et al. However, my major comment is about the mechanism proposed to explain the AMOC response differences in the experiments, that is the sea ice-driven response. The main evidence used to support this inference is the mainly correlation between AMOC strength and Sea ice extent. They argue that the correlation should be negative if the sea ice extent is caused by the AMOC, but the correlation found here is positive. The expected negative AMOC-sea ice extent correlation is based on the assumption that an increase in the AMOC should transport more heat into the Arctic and thus reduce sea ice extent. However, several studies have shown that that heat transport into the Arctic increases with AMOC weakening under global warming. In fact, this heat transport increase into the Artic is also seen in Figure 3, poleward of 60N and agrees with sea ice extent differences between the experiments. Under this scenario, it could also be argued that a positive correlation AMOC -sea ice extent is caused by the AMOC.
Reply: Yes, as the Reviewer pointed out, it can be seen in Figure 3e that the Northward Heat Transport changes sign at about 60° N, and this does require some discussion of our interpretation of the sea ice extent mechanism.
Therefore, we analyzed the correlation between the North Atlantic heat transport across 60° N (0-700m) and the Arctic September sea ice extent (a new Fig. S3). The correlation between the change of North heat transport at 60° N and the Arctic September sea ice extent is not significant. This lack of correlation can be compared with that in Fig. 11 where only HadGem2-ES has a lower R 2 than 0.5.
We therefore include the following text: The slopes of the regression lines in Fig. 11 are positive, meaning that greater AMOC strength is correlated with greater ice extent. However, Fig. 3e also shows that heat transport anomalies under the geoengineering scenarios change sign at about 60°N, with reductions in heat transport in the south coinciding with increases to the north of 60°N. But correlations of heat transport across 60°N with sea ice extent for separate ESM across scenarios are all insignificant and vary in sign (Fig. S3), in stark contrast to the regression lines in Fig. 11.
It is, however, true that the G4 and G4cdnc and the RCP4.5 experiments are significantly correlated (which we now include as fig S4). The reason for the relation is presumably as the referee suggested-the increased heat flux north of 60°.
For individual scenarios, there are significantly anticorrelations only for the RCP4.5, G4 and G4cdnc scenarios (Fig. S4). In this respect, the behaviour is similar, although less robust, as for wind forcing in Fig. 5, where scenario impacts as expected, but a consistent relation between scenarios simulated by each ESM is not present. The stronger sea ice correlation with increased AMOC suggests that sea ice may be driving changes in AMOC through the change in fresh water budget.
General Comments (continued) The earlier paper also cited to support this mechanism (Li and Fedorov 2021) is also primary forced by sea ice changes rather than the radiative forcing in the experiments in this study, so the conclusions from this study do not necessarily carry over. The authors should provide more evidence support the causality they're inferring from this study. (2021), contrary to the Reviewer, is that their experiments were made with perturbations in radiative forcing, represented as an imposed radiative flux imbalance at the sea ice surface, and that is driving the changes. Here are several quotes from Li and Fedorov (2021) to support this, e.g. in the abstract: "Here, we examine global ocean salinity response to such changes of Arctic sea ice using simulations wherein we impose a radiative heat imbalance at the sea ice surface"; in Section 2 "Sea ice surface radiative balance is altered by either reducing sea ice surface albedo to increase shortwave absorption (named "SW" experiment) or reducing the sea ice surface emissivity to restrain outgoing long wave radiative fluxes (named "LW" experiment)." and "Two additional experiments are also conducted with stronger shortwave absorption ("strong-SW") and weaker longwave emission ("weak-LW"). All simulations start from a quasi equilibrium preindustrial control climate. Sea ice perturbations are initiated from the beginning of each simulation and maintained for 200 years. The magnitude of maximum sea ice reduction is roughly proportional to the strength of sea ice radiative perturbations." Additionally, the experiments shown in their Fig. 1 all seem to be radiative forcing designs. Thus, it does seem to be conceptually comparable with the radiative forcing differences we examine in our geoengineering experiments.