Dear Authors,
Thank you for your responses to the reviewers’ comments. Overall, your suggestions to take the Referee’s comments into account appear reasonable to me at this stage. I would thus be happy to consider a revised version for publication in Climate of the Past, which I will send back to the reviewers to have their advice for the following.
Sincerely,
Hugues Goosse

Dear Authors,

I would like first to apologize again for the abnormally long delays of this second round of review that was due to very special circumstances. Based on the evaluation of one of the reviewers of the first phase and my own reading, I consider that the revised version takes well into account most of the points raised by the reviewers or that you have argued convincily in your response on the reasons of your choices. The reviewer of the first phase has still some technical issues that must be solved. I have also listed some points below, most of them related to ones already raised by the reviewers. They are mostly minor and should not need too much time. I thus invite you to submit a new version taking into account those remarks.

Best regards,

Hugues Goosse

Additional points
Title of section 1.1. Is it reconstructions of ‘SAM variance’ or of ‘SAM variations’ ? ‘variance’ may be misleading because of the association with the statistical definition of the variance.
Lines 83-90. Maybe add some references on the interest and limitations of pseudo-proxies?
Line 85. I would use ‘mask’ instead of ‘degenerate’ and I must admit that it was not easy for me to make the link between the first half of the sentence that mentions ‘non-climatic noise’ and the second half focused on a ‘real’ change in the strength of teleconnection due to variability in the climate. I think I finally got the message but I had to read several times the sentence so maybe a rephrasing would be helpful
Line 108, it is not clear what is referred to in ‘in this instance’. Is it Karpechko et al. (2009)?
Another issue for Figure 1 is the comparison of the results of control simulations with the observations over past decades that likely includes a forced trend. It is mentioned elsewhere but I think the impact of this point should be included in the discussion of the figure 1.
Line 168. If I understand well, the proxies are selected if they meet the criteria on one of the windows, not all of them, and you are only using one time window for all the proxies at the same time for each reconstruction. Maybe it would help to state this explicitly. The wording ‘For each grid point of the model’ may suggest that the procedure is different for each point.
Line 215 (and also 253-260). The larger skill when using precipitation record is interesting. Is it possible to make a link with the correlation between those records (precipitation and temperature) with SAM, as for instance illustrated in Figure 2.
Figure 10. I may have missed it but the reference to figure 10 occurs after the one to figure 11 in the text.
Line 355. ‘less exhibiting less spread’, maybe suppress one ‘less’.
Line 368 My visual interpretation is that the difference in RMSE is small between precipitation and temperature derived reconstructions. This difference should be quantified here to confirm that it ‘again perform better’.

Dear Authors,
Thanks for the submission of the revised version of your manuscript. I consider that you have addressed well the points raised. I am thus pleased to accept your paper for publication in Climate of the Past.
Best regards
Hugues Goosse