Summary: The study analyses the response of volcanic eruptions in a set of climate model simulations , comparing them with proxy-based temperature and circulation reconstructions, over the past millennium. The study distinguishes between tropical and extratropical eruptions in the Northern Hemisphere. The main results are, in my interpretation, that the qualitative response is , first, strongly dependent on whether or not the eruptions have been tropical; it also strongly depends on the model, and on the forcing data set of past volcanic aerosols.

The study is motivated by a series of prior publications that also target the response of the atmospheric circulation to volcanic eruptions, which have found somewhat contradictory results. Whereas it seemed clear that the reconstructions did show a shift towards a positive NAO state and thus an Eurasian warming after eruptions. The modelled response was not that clear. Some studies, e.g., Tejedor et al. and Polvani et al., suggested that there was no response at all, and that the signal from the reconstructions, at least at the surface level, is just random internal variability.

This study offers an alternative interpretation, suggesting that the unclear signal derived from simulations may be due to differences among models and to differences in the volcanic aerosol data sets used to force the models. They find that the more recent forcing data, along with a focus on the most realistic models, allow them to identify a consistent signal, particularly for tropical eruptions.

Recommendation: I found the study interesting, and the manuscript is well-written. It can be somewhat controversial, as it does not agree with those previous studies, but I am happy to recommend it for publication. I have a few suggestions that the authors may want to consider:

1) If my interpretation of the framing of the study is correct, I would suggest stating more clearly in the abstract

the conceptual links to those prior analyses, explaining briefly why this study is important. The current abstract sounds correct, but it does not clearly convey the present backdrop and where this study fits.

2) In my opinion, one key figure of the manuscript is Figure 5 (also Figure 6). However, I had to stare at Figure 5 for a long time to fully grasp the message. First, it won't be easy for many readers to see all the details. The purportedly grey dashed lines showing the sigma and 2xsigma bounds can barely be discerned; actually, what the reader sees are other grey dashed lines marking the zero anomalies for both axes. Additionally, the zero grey lines are not visible in all panels, and the reader may wonder if there is a hidden meaning behind this omission. The explanation of the circle colour is also left to the figure inlet in the first panel, and it is not mentioned in the caption. The caption also states that the circle sizes, displaying the magnitude of the eruptions, are normalised to the Samalas eruption. Thus each panel should have one largest red circle of the same size (the Samalas eruption). However, this does not seem to be the case for all panels, e.g., not for MPI-ESM-P CEA or ACCES evolv2k.

My recommendation is that the authors give further thought to this complex yet important figure.

3) One of the previous hypotheses is that there is no NAO response or Eurasian temperature response. This study retorts that the choice of model and forcing data is, or can be, important, and that the signal may have been smeared out by different model responses and differences in the forcing data. One way to contribute to clarifying this question is to look at a one-model simulation ensemble. If the MPI-ESM-P model, according to the authors, is one of the more realistic models in this regard, would it be possible to examine the ensembles of historical simulations with this model for the Pinatubo eruption? What is the spread in that ensemble? Marking the Pinatubo eruption (perhaps with green colour) in the panels of Figure 5  would also help.

4) The resolution of the figures needs to be improved. Many of them are multiple panels of relatively small size. A finer resolution would help the reader

Publisher’s note: this comment was edited on 3 September 2025. The following text is not identical to the original comment, but the adjustments were minor without effect on the scientific meaning.

Understanding the differing responses to tropical and Northern Hemisphere volcanic forcing is crucial. This knowledge not only aids us in making informed predictions regarding potential future eruptions but also serves as a valuable analogy for stratospheric aerosol injection. This paper provides a thorough comparison between multiple reconstructions and simulations. The consistency observed in both reconstructions and simulations reveals positive NAO anomalies and Eurasian warming following tropical volcanic eruptions. However, significant discrepancies emerge in the case of Northern Hemisphere eruptions. These findings are intriguing, and I believe this paper could be published in Climate of the Past following major revisions.

My primary concern lies in the definition of the eruption year. The impact of a volcanic eruption on climate is contingent upon when and where its stratospheric aerosols evolve. Specifically, tropical to mid-latitude lower stratospheric aerosols absorb shortwave radiation, leading to warming, which in turn increases the meridional temperature gradient and enhances the polar vortex. The role of ozone deflection is not adequately represented in these PMIP models when explaining the enhanced polar vortex following tropical eruptions. For instance, consider the Samalas eruption, which is identified as occurring in 1258 in the GRA dataset, while it is recorded as 1257 in both the CEA and eVolv2k datasets. This discrepancy means that comparisons made for the winter NAO refer to the 58/59 winter in GRA, whereas they refer to the 57/58 winter in CEA and eVolv2k. In Fig. 4c, we can see that the middle-latitude lower tropospheric aerosol forcing is significantly large in GRA six months prior to the 58/59 winter, while it only appears one or two months before the 57/58 winter in CEA and eVolv2k. This suggests that the aerosol warming effect did not have sufficient time to exert its influence in the latter case. Moreover, for many historical eruptions, pinpointing the exact eruption month can be challenging. Therefore, I recommend defining the eruption year based on the maximum annual aerosol production in simulations and making subsequent comparisons accordingly.

Abstract: The results of this paper are not sufficiently summarized and emphasized in the abstract. It is crucial to clarify the nature of the NAO response observed following tropical and Northern Hemisphere eruptions, as well as to outline the consistencies and discrepancies between the models and reconstructions. We can confidently conclude that tropical eruptions tend to enhance the NAO positive phase. However, there is no consistent conclusion regarding the effects of Northern Hemisphere eruptions.

Section 2.2: To compare reconstructions and simulations, the criteria for event selection were similar to those used in previous work (Liu et al. 2022 Nature communications).

Lines 245: The composite aerosol forcing for these three datasets, as shown in Figs. 4c and 4d, is valuable for understanding the differing responses observed in the simulations.