The study uses the recently developed CANARI-LE large ensemble to analyse four storms that affected the UK and to compare their characteristics under the SSP3-7.0 (Regional Rivalry). They found that the total number of cyclones similar to such storms decreases in the future and the most extreme events intensify in terms of precipitation and low-level wind speed. 

The manuscript builds upon the cyclone analogue methodology introduced by Ginesta et al. (2024), and this prior study is acknowledged in the main text. However, beyond applying the framework to a different model and a new set of UK-impacting storms, it is not yet clear what additional conceptual or mechanistic understanding is gained. If the primary contribution is to assess robustness across models and storm types using this methodology, this should be stated explicitly — already in the abstract and reiterated in the conclusions — for example, by framing the study as a systematic extension aimed at testing generalisability and moving toward broader or more automated application of this analogue framework.

A clearer comparison of the results with Ginesta et al. (2024) and, most importantly, its implications, would strengthen the manuscript. Although the storms are different and not directly comparable, the authors should discuss more explicitly how their results relate to the CESM-based findings: do you find similarities with specific storms, are magnitudes or scaling different, are there model-dependent sensitivities, are the results so storm-dependent that no generalisable conclusions can yet be drawn? This would help clarify the added value of the study and if it extends or generalises the previous work meaningfully.

The ensemble size (1200 model years per period) should adequately sample internal variability. However, since the analysis relies on selected analogue events, it would be useful to check whether the analogue selection favours particular modes of variability (e.g. ENSO, NAO, AMOC). Even a check for one storm would increase confidence in the robustness of the results.

The number of analogues appears small, and for some storms very small. Why do the authors use a constant 500 km radius along the entire track, rather than a varying radius (e.g. narrower near peak intensity and wider when the storm is farther from its peak)? The authors state that sensitivity tests were performed without finding significant differences; however, if using a 700 km radius increases the number of analogues and thus the statistical sample, why not adopt that choice? In other words, how do the authors define the point at which a storm ceases to qualify as an analogue?

Precipitation: why do the authors not consider the total accumulated precipitation over the entire storm track? Alternatively, analysing 6-hourly precipitation along the track (e.g. as a histogram, cumulative distribution, or time series relative to peak intensity) could provide additional insight and account for potential shifts in the timing of max precipitation. Do precipitation changes follow Clausius–Clapeyron scaling? Is the intensification mainly thermodynamic, dynamic, or a combination of both?

Please clarify whether analogues are restricted to winter (or extended winter, or Septemebr to April as stated in a few figures). If so, this should be clearly stated; if not, possible seasonality effects should be discussed.

Definition of extreme analogues: why do you use the 90th percentile based on ERA5 and not a specific percentile according to the model, as Priestley and Catto 2021 did? This threshold might not be the same in CANARI LE.

Other comments:

• Please use colour-blind friendly colorbars (avoid rainbow)
• Use blue for precipitation increases and red for decreases.
• Adjust colorbars to correspond only to the plotted fields (e.g., Fig. 2)
• Please clarify what is meant by “dynamical intensity” in the third research objective.

The authors build on an existing approach to study changes in archetype cyclones through studying changes in their analogues. A considerable part of the paper is dedicated to evaluating the CANARI LE dataset. The paper is generally well-written. The analysis is logically structured and provides detail on the analysed cyclones.

What I am more concerned with is the interest that the paper may hold for the broader WCD readership. Given that the paper slightly modifies an existing approach and applies it to a different dataset and different cyclones, I would encourage the authors to clarify the value of the paper for those not involved in CANARI or not interested in the specific geographical subregion being studied. This is partially done in the conclusions, but more could be done in this respect in the introduction and discussion sections. For example, the authors could reflect on the feasibility of systematically applying this approach to a very large number of cyclones (would this be computationally feasible? Would the results be interpretable?) or on whether the approach could work for other atmospheric dynamical features beyond cyclones.

I provide some additional specific comments below.

1) Title: Given that, according to Fig. 1, at least two of the four cyclones being analysed appear to cause stronger winds over France than over the British Isles, and a third mainly affected the Republic of Ireland, the authors may consider revising their title to include both the British Isles and France, rather than only mentioning the UK.

2) l. 9: This sentence makes it sound as though the authors are drawing a general conclusion on all cyclones. Perhaps they could state explicitly that this and the following points apply to cyclones similar to the four events that they are analysing.

3) ll. 148 and following: It is worth mentioning here that reanalyses often heavily underestimate the local magnitudes of extreme events, including heavy precipitation and strong winds.

4) Sect. 3.2.2: You state that you identify analogues across 1200 simulated years of present and future climates, from 1980 to 2010 in the present-day climate period, and from 2070 to 2100 in the future climate period. However, in Sect. 5 you also identify analogues directly in ERA5. I would ask the authors to verify whether this apparent discrepancy issues from my own misunderstanding of the information or is indeed a case of incomplete information being provided in Sect. 3.2.2.

5) Consider whether Figs. 4 and 6 could be combined into a two-panel figure.

6) Consider adding vertical labels to each row in Fig. 7, to indicate which cyclone is being shown without having to refer to the caption.

7) Sect. 6: In general, I see a very weak link to hazards throughout this section. In the typical view of risk, hazards should be quantities that may be directly related to impacts. Here, the authors seem more intent on analysing quantities of dynamical interest than quantities directly associated with impacts. The only part of the section that truly deals with a hazard is the subsection looking at precipitation. I provide some concrete examples of this below.

Sect. 6: Most studies of cyclone-related hazards focus on the hazards over land, where the overwhelmingly largest exposure is located. Here in some cases the largest changes in the quantities that the authors look at are over the ocean, making these quantities of limited relevance from a risk perspective.

l. 323 What is the logic of using maximum RV to study hazards? Would it not be more informative to select the day of largest cumulative hazard over land regions (quantified through whichever metric the authors may prefer)?

Sect. 6.2: I would not call wind at 850 hPa a hazard. If the authors want to investigate hazards, they should focus on 10m wind gusts, 10m winds if gust data is not available from the model, or similar.

8) Statistical significance testing; In some of the figures, the authors conduct statistical tests at individual gridpoints and assign a significance level to these. As this is a repetition of the same test a large number of times, I would ask the authors to confirm that they have applied a multiple testing correction to prevent spurious statistical significance results. If this has not been applied, then it is necessary in an eventual revision of the paper.

9) l. 388 I find “exceptionally unique” an odd turn of phrase. Unique is not a comparable adjective.

10) Dataset availability: Could the authors specify whether any researcher can obtain access to JASMIN or whether this is restricted to e.g. UK-based researchers or researchers participating in specific projects?

Review of “Climate change effects on analogues of contrasting extratropical cyclones over the UK”

The manuscript by Farrell et al., investigates the influence of climate change on four historical cyclones impacting the UK, using cyclone track analogue approach. The study uzelizes the CANARI Large Ensemble, assessing both the present climate (1980–2010) and a high-emission future scenario (SSP3–7.0, 2070–2100). By focusing on these four cyclone case-studies, the study provides insights on cyclone-specific responses, which may be important for assessing their regional impact. Results show that track analogues exhibit contrasting responses to anthropogenic warming and explores the drivers of changes in intensity and meteorological hazards.

Overall, the manuscript is well written and presents some interesting results on projected changes in extratropical cyclones under anthropogenic warming. There are however some points that need to be addressed. My suggestion is for minor revisions, as detailed in the following.

Major comments:

This study is situated within the CANARI (Climate Change in the Arctic-North Atlantic Region and Impacts on the UK) research programme (Schiemann et al., 2026). As such, a significant part of the manuscript (e.g., analysis shown in figs. 3,4,5,6) is dedicated to a general evaluation of model biases and model response to climate change. The rest is dedicated to a process-based analysis of four selected cyclone analogues.

It is however not clear to me how the analysis of four cyclone analogues should it be interpreted in the context of model biases in the CANARI LE simulations. This dual purpose of the paper affects structure of the paper and perhaps masks the key findings. Particularly, in several places throughout the manuscript, it states that the goal is to verify the performance of the CANARI LE in simulating North Atlantic cyclones, while in other parts, including the abstract, the research goal is to examine the influence of climate change on four contrasting historical cyclones impacting the UK.

I would suggest better justifying why this study is needed and how adopting a UK-impact perspective allows us to gain new insights, which were not in the original study. Re-organizing the sections concerning the model biases can be better integrated (e.g., plots 3-6 (or some of them) can move to the supplementary, to help the paper become more concise and focused on its process-based analysis of these cyclones in present and future climates.

Novel aspects of this work: In what aspect does this study go beyond the findings of Ginesta et al. (2024), given that the same methodology and approach are used (replacing CESM1 with CANARI LE)? The manuscript can strongly benefit from a comparison of the results with exciting CESM (which was used in the Ginesta et al., 2024 study) and from clarify how using the LE improved the analysis. This can be added to the discussion, providing a clearer context to this work and strengthen the manuscript. Specifically, it would be interesting to know if model biases in cyclone frequency, intensity, extremes storms (if possible, cyclone analogues) show similar results. A deeper discussion or analysis can help to generalise the conclusions beyond the 4 cyclone analogues and enhance the contribution of this study.

Sample size of analogues:  Across the 74-year period of ERA5, relatively few analogues are found for each of the selected cyclones (14, 27, 17, and 5). This makes evaluating statistical significance a challenge for the cyclone track analogue approach. Is there a way to increase the number of analogues? Would it affect the results? Would a selection of analogues based on percentile (e.g., top 10% of matches) can increase the number of selected cyclones in these cases? A sensitivity test or discussion about the Methodology can help to clarify this issue.

In addition, the definition of extreme analogues is based on the maximum RV higher than the 90th percentile of cyclones in the Northern Hemisphere between 1980 and 2010 in ERA5. This threshold can be adapted perhaps for the North Atlantic or Euro-Atlantic sector, and can be implemented using model climatology rather than ERA5.

Minor/technical comments:

Abstract: “such projections average together cyclones with a range of contrasting dynamical characteristics potentially obscuring climate change effects on particular types of cyclones and the airstream structures within them.” This sentence remains bit unclear to me. What makes the cyclone characteristic contrasting?

Abstract: Are the airstream structures discussed in the paper? Perhaps rephrase.

Fig.2 shows candidate-track density for the Great Storm of 1987. Not sure if this figure is needed, and perhaps can be moved to the appendix?

Figures 4 and 6 can be combined into one (multi-panel) figure, to allow a better comparison on historical vs. ERA5, and future vs. historical.

Scientific phrasing: I suggest rephrasing the text in a more scientific writing style. Some examples: “Ophelia is included because..”, “likely exacerbated damages”, etc.

Line 110: “This trajectory likely exacerbated damages to infrastructure..” Is it possible to examine the actual range of the loss? (for example, in Met Office reports or other sources such as Perils/EM-DAT should have some loss estimations and discussion on sources of damage).

Reference: https://www.perils.org/files/News/2017/Loss-Annoucements/Ophelia/PERILS-Press-Release-Ex-Hurricane-Ophelia-27-Nov-2017.pdf

Line 82: south of Englance -> England?