Review of ‘Characteristics and dynamics of extreme winters in the Barents Sea in a changing climate’ - WCD manuscript egusphere-2024-878

Overview:

This manuscript examines the changing nature of extreme winter seasons in the Barents sea region between the last 10 years of the 20th century and the last 10 years of the 21st century under RCP8.5. An extreme season is determined by the combination of 6 variables via a PCA-based method, from which 3 types of extreme seasons are identified for each time period by the clustering of seasons in a biplot. The seasonal-mean large scale anomaly patterns that contribute to these clusters are analyzed, and two example seasons are studied to understand the time evolution of daily weather systems and how they contribute. They conclude that surface boundary conditions become a less important contributor to the formation of extreme winters in the future.

General:

I think the method used here is very interesting and not one I was previously aware of, and in particular I like the idea of defining extreme seasons via a combination of factors, which to me is more intuitive for the way in which people actually experience the weather. The figures are mostly very good and contain a lot of information. It is a very thorough breakdown of the formation of extreme seasons in the Barents Sea. However, the paper is very long in its current form and often reads like a list of results, which I think is quite common for a paper based on a PhD thesis (I know I had the same comment on a paper of mine based on my thesis!) To be publishable, I believe the authors need to really pick out the salient results and think about what the story of the paper is. I therefore recommend major revisions due to the need to clarify and rework the paper’s messaging rather than much need for additional analysis.

Somewhat more specifically, the intro is quite long and I think contains some irrelevant info, such as the 3rd paragraph on linkages, while other paragraphs are repeating info already stated.   

The methods are a bit sparse and without reading Rothlisberger et al 2021 and Hartmuth et al 2022 the reader can’t really understand the simulations or what the physical interpretation of the PCA method is.  

The term ‘seasonal-mean anomaly’ is used often in the text and I’m not certain if it means the daily mean anomaly from a seasonal average, averaged across the season, or if it means how the seasonal mean value differs from the seasonal average, so this should be made clear. Similarly, ‘key variables’ or ‘key surface parameters’ is used quite a few times but why are the chosen variables ‘key’? The term is also used before the variables are ever defined (e.g. line 32). I think that ‘surface parameters’ on its own is probably sufficient most of the time.  

Section 3 is a good overview of the statistical differences between ERA5/S2000 and S2000/S2100 but I have trouble linking this to a physical meaning. For example, what is the implication of less variance being explained by PC1 and PC2 in S2000 compared to ERA5? Why is dM less for S2100 than ERA5 if there are more seasons to choose from, also what does dM then mean physically, and does it changing from 2.47 to 2.40 in the future have any physical relevance?

Are the two different example seasons for each cluster necessary? It might be a way to cutback on the ‘list of results’ feel to the paper. I assume you  probably can’t present a mean time series because of how different each season’s evolution of individual weather systems is, but what is added by providing two seasons?

In the conclusions, paragraphs beginning at 566 and 595 both seem to just state things that are very much expected, like wet winters having a surplus of cyclones and surface boundary conditions both drive and are driven by temperature anomalies. I’m not sure these things are new results in any way, and it makes it seem like there aren’t any interesting conclusions to be drawn from quite a lot of analysis. There’s little here in the way of considering past work and placing results within the context of existing literature.

Some Specific comments:

Abstract:

1: remove comma after temperatures, move to after time on line 2.

2: predestined is a very strong word to use here, maybe something like ‘anticipated to be a 

7-10 slightly confusing and long sentence

Introduction: 

26: strongly affects -> has strongly affected

31-33 do the simulations show differing trends between models or do you mean this is in a single model?

37: heat transported by the ocean?

59: remove thereof

64: wettening -> wetting

67: susceptible to ice loss?

69: partially -> particularly

82-83: final sentence is superfluous

85: have been focusing -> have focused 

101: whereby -> and

122-126: I’m not sure I understand what the difference between Q’s 1 & 3 are, and for point 2 should probably specify that this is in a model, since it was already answered it for ERA5 in your previous paper. 

130 simulations in S2000 - > S2000 simulations

Data and Methods:

143: might want to be explicit about how the surface energy budget is defined

Paragraph at line 146: I’m a bit confused why the Barents Sea is defined by its sea-ice cover, it seems this could be influencing your results. Does its definition then change between ensemble members of S2000, or is it defined by an ensemble average sea ice cover? Why not just use lat/lon bounds?

Figure 1: In terms of physical interpretation, is it important that C3 in S2100 covers two quadrants? Might want to use different colours for each cluster to make it easier to read.

174: why is the interplay between the variables largely affected by the surface type?

Paragraph at line 196: Why define the clusters differently in S2000 and S2100, wouldn’t finding the nearest 10 seasons to the S2000 cluster in the S2100 phase space be a more interesting question to examine? They’re all extreme seasons, so it’s not like a similar type doesn’t happen in the future, we just don’t see such a tight cluster (as long as I’ve understood correctly.) Also, since you are choosing 30/50 seasons for the clusters, it seems a bit arbitrary to claim that extremes are of a different type in the future,

205: less -> fewer

Interannual variability:  -- Many of these sections could use more descriptive titles! Specifying this section is about a comparison, for example. Substructure was confusing to me, it’s a time series analysis, etc.

215-216: Why use different region than past work, seems just to complicate the comparison.

229 (& elsewhere): ‘such’ is often used and it’s not often necessary to make the sentence clear.

243-246: not necessary to include this info

248 existent -> present

252: why does the increase in variability of other precursor variables give a similar decrease?

Paragraph line 254: much is repetitive from previous sections

264: why ‘therefore?’

270: I don’t think they’re disparate at all, this seems too strong of a word, see my point about paragraph at line 196.

Extreme winters in S2000

289: ‘which lets assume’, not sure what this means

Figure 2: I like the different colour maps for the different variables, but the yellow and orange lines are hard to see.

Figure 3: include a colour bar for the frequency anomalies as well as bigger labels

331: are occurring -> occur

369-371: confusing sentence

397: North -> north

Last paragraph belong in the next section

Extreme winters in S2100

438: I think the small SST anomaly looks like a dipole or a shift in the gradient

466: sigma range-> within a standard deviation

489: does it become apparent from their anomalies?

Discussion and Conclusion

537: I think the citation is doubled

Figure 8: Really like this figure!

557: as key region - > as a key region

Review of “Characteristics and dynamics of extreme winters in the Barents Sea in a changing climate” by Hartmuth et al. 

This study uses a multivariate and cluster approach (considering surface air temperature, precipitation and surface energy fluxes) to identify extreme winters over the Barents Sea in the current (1990-2000) and future climate (2090-2100) from CESM large ensemble simulation. During the current and future climate, the role of atmospheric circulation (in terms of frequency of cyclones and anticyclones) and boundary surface conditions (sea ice cover and sea surface temperature) in affecting extreme winters are explored. The main conclusion is that when the sea ice edge retreats northwards in a warming climate, the boundary surface conditions play a less important role in controlling the surface variables and the extreme winters.

General comments.

The findings and methodology are valid; the analysis is thorough. Understanding the physical processes that contribute to the Arctic's extreme winters in the future climate is useful to the community. However, the paper has a lot of information and is quite long. Overall the authors should rewrite some parts (especially the Abstract and Discussion) to make the paper more readable, and better organize the main messages. I suggest a major revision. 

Specific comments

Abstract

Line 8: surface temperature → surface air temperature

Line 9: in a warmer climate (S2100)

Line 11: Substructure: This wording is new to me. Does it simply mean the synoptic (or daily) variability over a winter season?

Line 14-17: I wonder if this part is important to put in the Abstract.

Line 20: “decreasing magnitude in seasonal-mean anomalies” → I don’t think this statement is correct. The magnitude of anomalies seems to be comparable in Figures 4 and 7. I do agree that the variability of the anomalies decreases, as shown in Figure 8.

Introduction

Lines 41-43: The interannual and decadal variability is also driven by anomalous atmospheric circulation. See Siew et al. 2023 and Liu et al. 2022. 

• Siew, P., Wu, Y., Ting, M., Zheng, C., Clancy, R., Kurtz, N.T. and Seager, R., 2023. Physical Links from Atmospheric Circulation Patterns to Barents–Kara Sea Ice Variability from Synoptic to Seasonal Timescales in the Cold Season. Journal of Climate, 36(22), pp.8027-8040.
• Liu, Z., Risi, C., Codron, F., Jian, Z., Wei, Z., He, X., Poulsen, C.J., Wang, Y., Chen, D., Ma, W. and Cheng, Y., 2022. Atmospheric forcing dominates winter Barents-Kara sea ice variability on interannual to decadal time scales. Proceedings of the National Academy of Sciences, 119(36), p.e2120770119.

Line 43: Should also include Siew et al. 2023 in the reference.

Line 47-59: This paragraph regarding Arctic-midlatitude teleconnection is not relevant to this study. The whole paragraph can be condensed into 1-2 sentences. 

Lines 75-83: This paragraph has some repetitive information and it can be combined with the previous paragraph. 

Line 88-90: “we” refers to HA2022? It might be fine if the author list is the same. However, using “we” is confusing as readers might think the ERA5 analysis is done in this study. So I suggest using HA2022 to avoid confusion. 

Line 101: Add Screen 2014 in the reference. 

• Screen, J.A., 2014. Arctic amplification decreases temperature variance in northern mid-to high-latitudes. Nature Climate Change, 4(7), pp.577-582.

Data and method:

Line 147-151: This part is unclear. Is the area of the Barents Sea fixed in the definition, or changing in different winters? Also, a bigger Barents Sea region is used for S2100 simulation, as mentioned on line 366 and Figure 5. 

Line 140: 2-metre temperature

Line 182: How many winters are there in total? 

Line 201-203: The two-step procedure for identifying the 3 clusters on the PCA biplot is unclear. Could you explain that more clearly in the text?

Section 3:

Section 3.1 is not very useful (what do we learn by comparing those details between the method applied on CESM and ERA5 in HA2022?). Section 3.2 can also be largely shortened. 

Line 215: What does KBS refer to?

Section 4.1 

Figure 2: The dashed yellow line (climatological sea ice edge) is invisible over the Barents Sea region due to the overlapping with the Barents Sea box.

Line 284: Authors should define the direction of positive energy fluxes (Es). 

Line 295-297: The statement related to CAOs comes from nowhere without going to Figures 3 or 4. 

Section 4.2 and Sections 5.2

Line 317, 328: How these specific cases are picked (out of 10 samples ) is not mentioned.

Figures 3 and 6: Overall I think the fc, fa and Fcao analyses  (the heatmaps at the bottom of the subplots) are not very helpful. Their roles are largely accomplished by Figures 4 and 7. What do we learn here by examining their daily evolution? Also, how do these frequencies be defined on daily timescales? 

Line 493: sigma → standard deviation 

Section 6

Line 559: Reduced spread of ES anomalies.

Line 565-590: I think these two paragraphs can be shortened and combined.