Smedsrud et al. (2017) point to the fact that the AO - Fram Strait export correlation is non-stationary in time. The results presented here are in line with what Smedsrud et al. report for the 21st century. Yet, I am curious about the earlier period, do you see any correlation between drift and AO in earlier periods of the FESOM simulations?

I find this section very convincing!

One additional suggestion, still. What do the sea ice area export curves look like in the FESOM simulations? I am assuming that if you were to look at area export (function of ice drift and ice concentration), you would get a similar picture to Fig. 3c - i.e.: slightly positive trends, with the variability almost entirely explained by winds. Is it possible to show that (additional panel in Fig. 3)? I believe it would bring yet an another argument supporting that the long term trend in volume export has to be thickness driven.

I would like to see just a little bit more detail as to why PIOMAS shows no trend. Selyuzhenok et al. (2020) report that the systematic error in PIOMAS (smoothed difference between thick and thin ice) might play an important role in this discrepancy. Do you have any other insight as to why the Selyuzhenok et al. PIOMAS trends stand out compared to the other?

Same comment here. At Fram Strait - average thickness across Fram Strait is implied?

Remove 'the', unless there is a missing word following Greenland?

I find the introduction very clear and well structured.

I would consider including a word on Fram Strait sea ice area export (which is reported to be increasing (Smedsrud et al. 2017)), which would help set a contrast with volume export, which is decreasing. 'sea ice export' is used to describe area or volume export, depending on the paper. Although it is perfectly clear through the manuscript that we refer to volume export, a short note on area export in Fram Strait could be of interest for readers.

I don't understand this part of the sentence. What does 'both the SLP gradients' mean? Was it meant to refer to the mode that better describe the thickness variability and the sea ice drift variability?

This is in line with the findings of Williams et al. (2016) Fig. 3, who demonstrate co-variability between sea ice area export in the Eurasian sector and sea ice area export in the Fram Strait, and further show the correlation with the AO index.

typo

I understand this is the sea ice drift averaged across Fam Strait? Furthermore - is this the component of the drift perpendicular to the Fram Strait gate? I think these details are important to precise in-text.

Why not calculate correlations with SLP over land? There might be some interesting connections when thinking about high pressure systems above Greenland?

Can you provide more information as to how the winds induce sea ice thickness variability in the Fram Strait (in the wind_vari setup)? In the following paragraph you make the case for convergence of sea ice causing an increase in thickness. Could other mechanisms, such as advection of thicker ice from regions with typically thicker ice, due to a change in the atmospheric forcing pattern - also be at play?

I do not see the correlation between ice thickness and drift reported anywhere?

This is the annual mean anomaly with respect to the 1992-2014 mean, is that correct? Maybe precise this in the legend.

Please precise the definition of the Fram Strait used for producing the results. Is it a line of constant latitude (82°N)? Or a straight line between Greenland and Svalbard?

The Fram Strait gate could also be illustrated on the map of averaged thickness and sea ice drift I suggested above.

In Fig. 2b - the observations are the Cryosat-2 thicknesses, correct?

I understand that CMST assimilates sea ice thickness from Cryosat-2. Therefore, why is there such an important discrepancy in the ice thickness in Fram Strait between CMST and the observations?

More precision on how FESOM simulates sea ice should be included. Are there any know biases?

Also, I think a useful additional figure would be a map of averaged thickness and sea ice drift over the 2001-2019 period for the control run (a figure similar to Fig. 1a in Min et al. (2019)).

My comment above applies here - a map of averaged ice thickness and sea ice drift over the 2001-2019 period would be very useful in supporting this point.

How is that being handled? Are the fields replaced by their climatology above a certain latitude (60°?), or is there a finer definition of an Arctic Ocean domain?

Also are certain criterions imposed for continuity in the climatology/dynamic-fields transition zones?

What does the sea ice model use as an Ice Thickness Distribution (ITD)? ITDs will influence the representation of ice thickness in the model, and since ice thickness is of major importance in this study, I believe this is useful information.

I would find it interesting to have little bit more information here, on the changes in thickness reported by Hansen et al. 2013 and Renner et al. 2014. - i.e.: very briefly report on the change in modal thickness, or % change over recent decades.

What is being referred to as a long time scale? Decadal? Multi-decadal?

A reference to Williams et al. (2016), Dynamic Preconditioning of the Minimum September Sea-Ice Extent, doi: 10.1175/JCLI-D-15-0515.1; should also be included. The authors make the connection between Fram Strait sea ice export and September minimum sea ice extent.