Review of “Can low-resolution CMIP6 ScenarioMIP models provide insight into future European Post-Tropical Cyclone risk?” by Sainsbury et al.

In this article the authors investigate the change in PTCs affecting Europe due to global warming in coarse resolution CMIP6 models. Because TC are intense small-scale systems the analyses of TCs and PTCs is usually done with the highest resolution models available. The authors of this article take the courage step to analyse TCs and the subsequent PTCs in present coarse CMIP6 ScenarioMIP runs. 

Although courageous this choice is defendable because apart from being small scale and intense, TCs are characterized by their different structure such as warm core and can be identified in low resolution models although with smaller intensity and larger structure as already shown by Haarsma et al 1993.

The authors provide a thorough analysis of the change in PTCs affecting Europe and the different factors that play a role in this change, like change changes in TC genesis, recurvature and reintensification. For the first time the authors provide an analysis of the complete PTC life cycle and the changes due to global warming in climate models. This analyses is compared with HURDAT2 observations and ERA5 reanalyses highlighting that the CMIP6 ScenarioMIP models are able to capture the main processes that govern the evolution of TCs to PTCs.  However, from these analyses it is also clear that the climate models still show large biases, which are not only due to the low resolution.

The strong point of this article is that it separates the different phases in the evolution of PTCs affecting Europe starting with the genesis of TCs in the Atlantic and investigates the different processes that play a dominant role in those phases and how these are affected by global warming, which has not been done before. I consider therefore this as an important paper, also for future studies with larger ensembles and higher resolution, for which this analysis can be repeated or serve as a guideline.  

The scientific question how global warming will affect future European PTC risk is far from being answered. However, this study disentangles this question into other underlying relevant questions and provides useful analyses and preliminary answers.

The article is well written, and the analyses are done thoroughly. I recommend publication. I have only a few minor comments that I will discuss below.   

1)    The authors select only five models from the large CMIP6 set, based on a criterium outlined in the supplemental material. Five models is a rather small number for any statistical robust conclusions. The selection criterion is based on a TC frequency being equal or larger than observed hurricane frequency. No motivation for this selection criterium is given. I am wondering if a weaker criterium providing a larger ensemble would give similar results and give more statistical robustness. Please provide arguments for this criterium choice.

2)    Two of the models that are analysed (CNRM and HadGEM) have resolutions comparable to the HighResMIP resolution (~50 km atmosphere, 0.25° ocean). These two models are also analysed in Baker et al. 2022. So, the sentence at line 74-76 is not correct. HighResMIP models had a somewhat different protocol than ScenarioMIP (i.e. aersol forcing, land surface scheme). Also period and spin-up are differently. I assume that CNRM and HadGEM are based on the HighResMIP models and modified and used for the ScenarioMIP. Please explain.

3)    Line 152-153 something went wrong in explaining the region of WEST. Later the regions are shown for instance in Fig. 7. May be mention that the figures are shown in the forthcoming plots?

4)    A notable difference in Fig.1 is the genesis over west Africa simulated by ERA5 and the models, but not seen in HURDAT2. Is this spurious or are no observations available over Africa for HURDAT2? Please explain or discuss this.

5)    Line 232-233. To me this is no surprise as coarse resolution models tends to increase the size of TCs and reduce the pressure gradient. May be include this argument here?

6)    Line 285. Maybe you can mention the numbers of Sainsbury et al 2020 here?

7)    Line 463-464. Here and in the supplemental material section 7 the effect of ensemble size is investigated. However, the ensemble spread is not discussed. That the ensemble size does not affect the basic results is reassuring, but I assume that there is considerable spread in TCs and also how much of those TCs recurve and translate into PTCs between individual ensemble members.

An interesting result of this study is the important role of vertical wind shear for the development of PTCs. The important role of wind shear associated with El-Nino’s for genesis and development of TCs is well known. This study also shows the large biases in wind shear in CMIP6 models. To this I want to add that recent studies also show that there is a mismatch during recent decades between the observed evolution of El-Nino and simulated by CMIP6 models (Seager et al. 2019). This will have an impact on vertical windshear over the North Atlantic and the genesis TCs and as a consequence also the PCT risk for Europe.

Haarsma, R. J., Mitchell, J. F., & Senior, C. A. (1993). Tropical disturbances in a GCM. Climate Dynamics, 8(5), 247-257.

Seager, R., M. Cane, N. Henderson, D. E. Lee, R. Abernathey and H. Zhang (2019). ‘Strengthening tropical Pacific zonal sea surface temperature gradient consistent with rising greenhouse gases’. In: Nature Climate Change 9.7, pp. 517–522. doi: 10.1038/s41558-019-0505-x.

Comments on Can low-resolution CMIP6 ScenarioMIP models provide insight into future European Post-Tropical Cyclone risk? by Sainsbury et al

The authors present a study that looks at both the performance of CMIP6 models to represent tropical cyclones in the North Atlantic, and a future projection of recurving PTCs impacting Europe.

I think bost aspects of the study are relevant and new, as well as show useful results with opportunities for future work.

The methods are overall well explained, the manuscript is well-written and structurally sound, and the results properly assessed.

There is room for some improvement in the motivation of model choice and general assessment of their TC skill, to better understand their present and future skill regarding North Altantic TCs.

Some more detailed comparison to existing studies using high resolution models in particular would be useful, showing which of the expected trends are consistent and therefore useful between different sets of models.

This may be at least partly the subject of some previous or future work, but in that case it should be better clarified.

Regardless, the manuscript is of high quality and would only require limited adjustments for publication.

General comments:

• A selection of CMIP6 models is used, based on the present TC frequency over the North Atlantic Ocean. While this is probably a good measure, it is tough to say whether these models accurately show TC frequency for the right reasons and even more so if they are able to correctly show future trends.
  
  Especially with contrasting mechanisms determining the trend, this may be an important concern. It would therefore be useful to show some general metrics of these models, e.g. how they reproduce overal TC formation/intensification measures, or how they compare versus (limited) available high resolution simulations.
• I would like to see some more information regarding the thresholds used to define a TC, both for tracking as for impacting Europe (does this include wind speed, or just vorticity?).
  
  Although previous methods are mostly being used and well documented, this could help avoid some confusion going through the results.
• In the results section, the first paragraph of most subsections is more of a motivation/methods part. This can be a deliberate choice, but makes the results section slightly more tedious and less focussed.
• The results cover different statistics about TC recurvature, frequency and track, which is very useful. There is, however, an important distinction between considering a fraction of North Atlantic TCs (recurving, impacting Europe etc. e.g. Table 2, Figure 10), or rather cyclones impacting Europe as a start (i.e. how many of those are PTCs e.g. Figures 5,9). This distinction could be made clearer and considered more deliberately at times, to avoid confusion.

Specific comments:

• L46 and following: it is important to clarify that the Haarsma and Baatsen studies only focus on PTCs impacting Europe with storm-force winds, which is only a minor fraction of all PTCs.
• L123 & L130: Besides earlier work and data limitations, could you motivate the different choices between the datasets and the potential limitations for the results? (e.g. weaker vorticity signatures at T42)
• L166-184: It would be useful to perform some of these analyses (or just e.g. the GPI) on the full CMIP6 ensemble, to get a better understanding why the selected models are better representing TC frequency.
• L208: To test whether the models dissipate existing TCs too rapidly or rather convert too few seeds into TCs, you can also consider the average track length and duration of the considered TCs.
• L219: Can you explain this underestimation, while models seem to represent overall TC frequency well? Are conditions too hostile in the north or rather the models unable to represent physical processes correctly towards ET?
  
  Especially for cases of re-intensification after ET, it could be useful to check SST fields (and gradients) in the models.
• L237: It would surprise me if baroclinic forcing would play a major role over the Gulf of Mexico during much of the Hurricane season, as would model resolution differences between 10-20N and 25-30N. What about e.g. biases in SST, RH or wind shear?
• L256: As mentioned in the general comments, I believe a more detailed comparison would be insightful here, especially regarding the environmental conditions for TC formation and maintenance.
• L324: This seems to happen mostly towards the end of the 21st century, following an increase in intense storms that originate from the tropics but are not PTCs (Baatsen et al 2015).

Figures:

• Figure 1: The HURDAT2 contours are rather tough to see, particularly as they conflict with the coastlines. Consider e.g. making the coastlines a lighter gray for clarity.
• Figure 1&2: As the lower values are masked out, it would make more sense to use a reversed colouring i.e. darker colours for higher values?
  
  The highest values run beyond the scale, which is probably chosen to make for an easy overview. Consider adding some larger steps to the top of the scale to be able to distinguish the maxima as well. Right now, one cannot tell whether this value is 10 or more like 50.
• Figure 5: It would be useful to include some information regarding total counts as well, both on the cyclone frequencies as on wind speed percentiles.
  
  It is unclear to me how the percentage given in each panel relates to the shown bins: I understand this is the total fraction of Europe-impacting cyclones that are PTCs during Hurricane season. In that case, the number should correspond to that of the 0th-percentile, as this would include all storms. Clearly it does not, so I am missing something here.
• Figure 6: This is an insightful and nice overview figure, it could benefit from a CMIP6 ensemble mean comparison.
  
  Figure S4 is a great addition in the supplement, so it would make sense referring here as well.
• L405: Term 3 seems a bit unintuitive, being the combination of terms 1 and 2? While still referring to the appendix for more detailed information, it would be nice to have a bit more explanation on how this combination works.

Technical remarks/typos:

• L17: similar occurring twice
• L276: 'is show'
• L291: consider making the list numbers bold font for clarity
• Table 3: this is quite an extensive table, is there a possibility to include some (simple) coloured shading to the cells to facilitate visual comparison?