The North Atlantic Oscillation (NAO) is the dominant mode of atmospheric variability in the North Atlantic domain. It has a strong impact on daily weather in western Europe and eastern North America, especially in winter. The present work investigates a less studied aspect of the NAO: the evolution of the percentage of explained variability. The authors show that the NAO explains more of the SLP variance at the end of the 20^th century than at the beginning, meaning that the role played by the NAO has increased over time. The science presented here is sound. The text is well written and structured and the figures are clear. I have only a few comments listed below.

General comments:

• I understand that the authors focus on the winter NAO only, but it is not clearly stated in the text. Could the authors clearly specify that they focus on the winter season only in the introduction and specify which months are downloaded in section 2. If I am wrong and all the seasons are considered, is this change in explained variance due to one season in particular or is it observed in every season?
• Why the authors did not use more recent reanalyses? Like ERA5 instead of ERA-20C, that spans 83 entire years (1940-2022). And can we fully trust ERA-20C/NOAA-20CR in 1930?
• The authors show that the observed increase in explained variance is not visible in the CMIP6 ensemble mean, which could mean that it is not a consequence of an external forcing like global warming. Have the authors thought of any other process that could explained this change in explained variance on such a long timescale? Could the authors comment more on this in the conclusion?

Specific comments:

• Figure 1: I understand that the shading represents the standard error, but it does not appear in the caption. Please add an explanation for the shading.
• Figure 3: I do not understand why the spread in explained variability reaches negative values. Are the boxplots “shifted” so that the median is zero for each model and ERA-20C? Could the authors better describe in the caption what is plotted in Figure 3?
• Plots in the Supplementary Materials: why do the authors highlight with green a third model?

Review of “Changes in the North Atlantic Oscillation over the 20th Century” by Outten and Davy

Overall recommendation: Major revision

General comments

This paper examines CMIP6 models’ reproducibility on interannual variability mode, NAO. The study has found increasing trend of the percentage of explained variability by the NAO in the reanalysis datasets, while that is not the case in many of CMIP6 models.

I believe the scope of the research meets the scope of the journal, and the paper was thoughtfully prepared. One concern I have though is about the percentage of the explained variability (PEV) The authors mostly focus on the PEV by the NAO. While I acknowledge that it is an important measure, it could be sensitive to the magnitude of the total variance of the model because the PEV is relative value to the total variance. I think having this addressed in addition would make the paper more robust.

In the discussion section (line 180), it was discussed that the trend from the reanalysis datasets could be resulted by the combination of the natural variability and the forced response. I wonder if this point could be explored more. One possible path could be working with anomaly monthly PSL field by removing any linear trend in it. This might be helpful to provide some insights on isolating influence of forced response.

Specific comments

Line 74 “times series”: Does this be replaced by “time series”?

Fig 1: What are the shadings indicating?

Table 1 and analysis: Is one ensemble member of each model used for the paper? Please consider clarify this in the paper.

Line 130 “In general, the climate models over-estimated the importance of the NAO in the first half of the century and under-estimated its importance in the second half, when compared to the reanalyses.”:

I am not sure what “importance” indicates in this description. Does it indicate the “portion of variability explained by the NAO”?

Also, this is the percentage, not the magnitude of variance itself. I think the magnitude of variance for each 30-year epoch might also need to be analyzed to complement the analysis for the “percentage of variability explained by NAO”