Dear Authors, 

Please refer to the attached PDF for my review of your manuscript.

Sincerely,

Andrew Winters

This manuscript, “A climatological characterisation of North Atlantic winter jet streaks and their extremes” by Bukenberger, et al., makes an important diagnostic contribution to the field of mid-latitude atmospheric dynamics. I would certainly like to see this paper published after the authors address the following minor points: 

Line 23 – I would suggest: “The jet stream is a band of enhanced westerly winds in the mid- and upper troposphere found in both hemispheres.”

Line 24 - I would suggest adding here one of encompassing textbooks on atmospheric circulation for the benefit of wider audience, e.g., “influences daily-to-weekly weather patterns with its meanderings (Randall, 2015).”

Randal, D, 2015, An introduction to the global circulation of the atmosphere, Princeton University Press, pp 442, ISBN, 9780691148960

Line 40 – I would suggest adding here textbook reference: “Frame et al., 2011, Wilks, 2020). The response …”

Wilks, D.S., 2020, Statistical methods in atmospheric science, 4^th ed., Elsevier, https://doi.org/10.1016/C2017-0-03921-6

Line 59 – I would suggest adding here a figure/plot depicting essential 4Q-model schematic for the benefit of wider audience.

Line 74 – I would suggest adding here textbook reference: “The PV Perspective (Hoskins and James, 2014) has been employed to study …”

Hoskins B.J, and I.N.James, 2014, Fluid Dynamics of the Midlatitude Atmosphere, Wiley Blackwell, pp 408. DOI:10.1002/9781118526002

Line 101 – Please justify why are you not considering data from available pre-1979 period? Additional point, could 3-hourly data provide more temporarily resolved insight?

Line 165 – Determining K in K-means clustering approach is one of the most sensitive aspects of such clustering analysis. Hence, please justify how did you set effective/optimal K=3 in your K-means clustering analysis (why not K=2 or K=4 or ...).

Line 172 - When the neighbourhood radius in SOM is set to 0, the SOM reduces to the K-means clustering (i.e., SOM can be perceived as a constrained version of K-means clustering, Hastie, et al., 2009). Please elaborate your choice of using SOM here instead of also K-means clustering (that is also unsupervised clustering technique).  

Hastie, T., R. Tibshirani, and J. Friedman, 2009: Unsupervised learning. The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Springer, 485–585.

Line 550 – I would suggest pointing out that the optimal number of clusters can be substantially dependent on the selected validity index or set of applied validity indices. This should deserve follow-up study (or studies) that could also involve some other state-of-the-art atmospheric reanalysis products such as JMA JRA-3Q and NASA MERRA-2.