I read with interest the discussion paper by Roberto Bilbao and Colleagues on the impact of volcanic eruptions on CMIP6 decadal predictions. The study is interesting, the analysis overall sound and the manuscript is well written and organized. The manuscript is worthy of publication in ESD, also as a reference paper for the DCPP-A and -C experiments. I have just a main comment regarding abstract/conclusions and a few minor specific suggestions for the authors about some possible improvements to their manuscript.

Regarding my main comment on abstract and conclusions, it seems to me that, beyond the focus on prediction, the study revisits general aspects of volcanically forced climate variability. The latter analyses confirm previous results about, for instance, dependency of the response on magnitude and spatial structure of the forcing, the signal-to-noise ratio issue in detecting the response to moderate size eruptions, the lack of robustness in NAO response, and the tendency to produce a post-eruption El Nino response. My point is, several of the main messages in abstract and conclusions concern these aspects of volcanically forced variability, rather than volcanoes and predictions. This could still be viable, of course, but statements like “some differences across models and eruptions arise due to the varying magnitude and spatiotemporal structure of the volcanic forcing”, or [decadal prediction systems show] “a strong agreement in predicting the radiative response to the volcanic eruptions, simulating a reduction in global mean top-of-atmosphere radiation fluxes, surface temperature” appear too general and vague, and, in my opinion, do not describe new advances in our understanding of the volcano-climate relationship. I feel that the paper would benefit from a narrower focus, in abstract and conclusions, on novel information about volcanic impacts on decadal predictions that emerges from this approach/analysis: In the abstract, this key part appears mainly in the statement that “including the volcanic forcing results in overall better predictions” and in the last sentence. I recommend the authors to consider revisiting these important parts of their paper.

Some specific comments:

Line 182-185: It is worth comparing these results with Zanchettin et al. (2022): results from volc-pinatubo-full indicate “maximum expected cooling ranging across models between about -0.27 and -0.38 and a multi-model mean of about -0.33”. The good overlap suggests minor influence from initial states and background conditions?

Line 240-241: aren’t points (1) and (2) different flavors of the same explanation?

Section 3.5: a deeper discussion about the implications of the constructive/destructive superposition mechanism by Swingedouw et al. (2015) might be worthy here.

Line 469: typo (is does)

Figure 5: could you extend the plots to encompass the whole hindcast period? At least for El Chichon observations appears to be in a colder anomalous state than hindcasts: does this discrepancy hold since the beginning? What if you use the pre-eruption period as baseline to calculate the anomalies (e.g., the five years before each eruption)? This comment stands also for the SSTs in Figure 14.

Regarding the NAO and ENSO responses: analysis of volc-pinatubo-full experiment yields overall similar conclusions to this study regarding the lack of NAO and, to a lesser extent, ENSO response to the Pinatubo eruption, using the analog approach of paired anomalies, but with some noticeable differences (for instance regarding ENSO response in IPSL-CM6A-LR. Maybe it is worth commenting on this?

This article presents results from a new set of decadal forecast models which allows the authors to isolate the effect of the three most recent large volcanic eruptions on climate. The analysis is interesting and very clearly presented, the paper is well written and methodology seems sounds – so I think that this article would be worthy of publication is in this journal.

I have two major concerns.

The first is due to the small sample size (only three eruptions) and the comparison to the observations. While the large ensemble sizes allow for proper verification of model results, it is not clear to me how best to interpret the comparison to observations. It the signal to noise ratio is low (as it presumably is for regional results). Including a volcano could improve skill in general while degrading the fit to certain aspects of climate in a specific case. Equally the models could have no skill while by chance a better fit to observations could occur. I think more discussion regarding this is needed in the context of predictability, so that the conclusions are clear and the results are not over interpreted.

The second is regarding novelty. Certainly, all findings do have novelty as they represent results from a new set of model simulations. However, much of this article details changes in aspects of the climate by volcanic eruptions which are already well documented in many other studies - ENSO, NAO, winter warming, AMOC etc as outlined in the introduction. I therefore think that more should be added to put them into the context of previous studies. What is new and what confirms or contradicts existing work?

Minor points:

Line 127-128. “on” the 30^th October in both cases

Paragraph L130. Would it be possible to include some discussion about the uncertainty in the volcanic forcing. Agung in particular is pre-satellite, so are the uncertainties larger? I note that the reference given only seems to covers the satellite era.

Line 143 – could you give more information regarding the ozone depletion as this isn’t that clear to me.

L155-156 – were these the infilled version of HadCRUT5?

Figure 8 – could you specify what period the anomalies are respect to?