I would like to thank the authors for their thorough revision and both their intermediate and final comments in response to the reviews. The paper is improved and clearer, and I'm satisfied that the major comments have been addressed. It is nice that the authors have performed several new calculations to test the robustness of their methods, and have revised text and figures for clarity. The links between different timescales have been thoroughly discussed and this is a key strength of this paper.

It is also interesting that the regional EOFs are found to describe the model responses better because they allow for more degrees of freedom. I wonder if another comparison might be interesting - do the two pairs of regional EOFs work better than the set of four leading hemispheric EOFs?

The reviewers also provided detailed minor comments which has allowed the paper to reach a near-final state in this revision. I just have a few brief suggestions for the authors to consider in a minor revision:

1. I think it is satisfactory to use v winds rather than the wind perpendicular to the waveguide in this case, but perhaps you could mention the approach of Wolf et al as an alternative.

2. It would be good to clarify in the main text in which cases EOFs 1 and 2 are / are not clearly separated from EOF 3, as this seems quite useful (just in the NCEP 1 data?).

3. I didn't find figure S4 in the supplementary material. Please check this and link to it from the main text.

4. I would reconsider the use of the word 'bias' throughout the paper, as it seems to be used more generally to refer to a difference between models than is common in the literature. I would restrict bias to the mean discrepancy between model and observations over a defined period.


Typos:

Line 104: brought about by changes
Line 285: non-zero
Line 329: missing date in reference