Review of Earth System Dynamics manuscript egusphere-2024-1711 “Future Changes of Compound Explosive Cyclones and Atmospheric Rivers in the North Atlantic” by Lopez-Marti et al.

This manuscript investigates the concurrences of explosive mid-latitude cyclones and atmospheric rivers and their future changes in the North Atlantic. The authors use ERA5 and CMIP6 climate models, tracking software, and study the link between ECs and ARs in the present and future climate. The main finding is that the concurrences of ECs and ARs is going to increase in the future, regardless that the number of ECs themselves show a decreasing trend. 

I enjoyed reading the manuscript. I think it’s very well written and easy to follow. The language was good, and the used datasets and methods are appropriate for the design of the study. Thanks to the authors for that. I also think that the topic of the study falls within the scope of the journal. 

Despite the overall good presentation of the manuscript, I have some minor concerns related to the methods and the main result. I have listed them below, in addition to some line-to-line comments. I hope these are of help to the authors and they can address them before the paper is published. 

1. Methods. At L134 you say that the ARs are detected within a 1500 km radius from the cyclone centre. At quick thinking, it sounds quite a large distance, given that e.g. Rudeva and Gulev (2007) found that the effective radius of oceanic ETCs is about 900 km. In the Introduction, you write (probably correctly) that the release of latent heat by the moisture of the ARs is an important mechanism in deepening the ECs. For me, it feels that if the AR is located very far (such as > 1000 km) from the cyclone centre, it cannot be involved in the deepening of the system. So, how did you arrive at the 1500 km value, and have you investigated how sensitive your results are to the used distance?
2. Results. Perhaps the headline result of your study is that the concurrences of ECs and ARs show an increasing trend in a warmer climate. In Sect 5.1 you discuss that the number of individual ECs show a downward trend with climate change, and also that the change of detected ARs is very modest, almost flat, in a warming climate. At L201-204 you say that the increase of concurrence in a warmer climate points to the changes in characteristics or ARs or cyclones, i.e. as I understood, changes in the dynamics.
   
   
   
   In any case, I was still missing a more detailed explanation or mechanism of how the concurrence of ECs and ARs can increase with climate change when neither individually shows a clear upward trend (I think this is a rather important finding which I haven’t heard before!). I understand it may be challenging to find any clear explanation, but in the absence of one, it would be good to at least state out loud this dilemma clearer, for example in the conclusions or in the abstract. Now it feels like it is being swept under the rug as it is only briefly mentioned at L202-204 and not again in the conclusions.

Line-to-line comments:

L42. “Extratropical Cyclones”. Do you mean explosive cyclones? If not, please decapitalize. 

L45. “undergoes further amplification compared to surface water vapor”. I’m not really sure what you mean by further amplification and why this is the case. This sentence could be rephrased.

L73. Why did you select 2009 as the ending year? Was that because you wanted to have 30 years in the historical period, consistently with 2070-2099 for the future? Please add an explanation to the text. And also, does your time range include OND 1979, i.e. should it be ONDJFM 1979/1980 - 2008/2009?

L95. What’s the unit of NDR? Isn’t it hPa / h. Now there’s no unit after 1.

L99. Here you start to speak about the number of cyclones detected. However, I think you should more clearly repeat the domain of tracking. The domain of ERA5 was presented at L74. Is this the same domain where you apply the tracking software?

L99. How do you treat those cyclones which form or decay outside the domain and only travel across it? How can you be sure what the MDP of a given cyclone is if only part of the cyclone’s life cycle occurs inside the domain?

L117-118. These threshold values seem a bit subjective. How did you arrive at them? I think the justification for these values should be mentioned.

L118. “The detected candidates are concatenated if at least one grid point is detected as AR in sequential timesteps”. Does this mean that the AR area at the next time step must overlap spatially with the AR area at the previous time step?

L134. The presence of AR. I understand that the location of the EC is clearly defined, i.e. it means the location of the minimum SLP. But it is unclear to me how the location of the AR is defined in relation to the EC. If the AR is wide, does that mean that it is sufficient if the closest grid point of the AR is at most 1500 km from the centre of the cyclone?

Fig. 3. Does the map show Xynthia's full life cycle, the whole track? Not a big deal, but I missed the MDP of Xynthia. Can you show its location on the map? I think it would better tie Fig. 3 to the following figures where the time axis is shown in relation to MDP.

L146-147. “Initial stages of the cyclone formation / dissipation stages of the cyclones”. In Fig. 4, you show only 72 hours (3 days) of the cyclone composites. Arguably many of the tracked cyclones last longer, meaning that their formation or decay can be days before/after MDP. Figure 7 shows that, on average, the minimum SLP of the cyclones has not increased much even 36 hours after the MDP. So is it correct to speak of dissipation in this context? Could it be better to just say 36 hours after / before MDP?

L157 and hereafter. For me, inter-seasonal means variations or comparisons between different seasons within a single year. So basically changes and differences from one season to another within the same year. Whereas inter-annual refers to variations between the same periods in different years, i.e. year-to-year differences. Do you mean inter-annual here and later in the manuscript?

L174. What does the internal variability of the datasets mean in this context? I think the sentence needs rephrasing.

L177. quantitative

Fig. 4. Could it be written to the panels the a-b represent averages over 30 years, and c-d standard deviations? It took me a while to understand that the panels c-d are standard deviations, especially as I got confused about the meaning of “inter-seasonal” in their title (see the comment a few comments back).

L184 and so on: 12 % or 12 percentage points?

L192. “for almost all the models”. Well that’s one way of saying it if 4 out of 6 models are showing an increase of AR tracks between SSP5-8.5 and historical. In general, the values in Table A3 seem very unchanged, so I think it could be honest to say that there is not really a systematic change at all between the scenarios and historical. It could be just random variation (internal climate variability).  

L199. Does this sentence refer to Fig. 5a? The reference to the figure could be added. 

L200. By the number of compound events do you mean a situation where at least one time step of the track of the cyclone centre is closer than 1500 km to the AR?

L207. the inter-model spread of the SSP5-8.5 scenario

L228. CMIP6

L249. “... is very limited”. I think you could continue this sentence by for example “as the coloured lines in Fig. 7 are close to each other” or similar. It took me some time to understand where you got this conclusion. 

Fig. 7e. I think you do not discuss at all why CMIP6 models seem to be more sensitive to the ARs than ERA5? Or did I understand it correctly? Why the coloured lines in Fig. 7e go much lower before MDP and much higher after MDP when compared to ERA5? 

Conclusions. Currently, I think the conclusions (and in fact the whole paper) paper puts quite a lot of emphasis on the high emission SSP5-8.5 scenario. However, it has been shown to be unrealistic (https://www.nature.com/articles/d41586-020-00177-3), and the world is currently roughly on the path of the SSP2-4.5 scenario. It might be appropriate to add a few sentences of discussion on this, stating that the results should be interpreted always with the scenario in mind, and that the results of the SSP2-4.5 scenario are more likely in the future than those of SSP5-8.5.

L283. SSP5-8.5 scenario

Table A2-A3. It could be helpful to add the periods (the year ranges) used for historial and SSP scenarios also here. 

References

Rudeva and Gulev (2007): https://journals.ametsoc.org/view/journals/mwre/135/7/mwr3420.1.xml

Review of the manuscript intitled “Future Changes of Compound Explosive Cyclones and Atmospheric Rivers in the North Atlantic” by Lopez-Marti et al.

This manuscript aims to evaluate the current climatology and assess changes under future climate scenarios of the concurrence between atmospheric rivers and extratropical cyclones undergoing explosive development – frequently referenced as explosive cyclones – in the North Atlantic. Being the explosive development of extratropical cyclones and atmospheric rivers crucial in driving extreme weather in the mid-latitudes, this topic is relevant, deserves to be investigated and it fits the scope of the Earth System Dynamics journal.

The manuscript is well-structured and well written and it is pleasant to read. It applies well-known datasets and detection and tracking methods previously published and discussed in the literature. However, in my opinion, some points are too succinct and need further details and explanations before the manuscript is accepted for publication.

Some points that need further clarification are:

• The authors use TempestExtremes Code for Detecting and Tracking Extratropical Cyclones and Atmospheric Rivers (ARs) for the North Atlantic region [25-65ºN; 80◦W-10◦E]. In Appendix “A4 Number of Cyclones and ARs detected in ERA5 and CMIP6” the words Cyclone and track are used as synonyms; in Figure 1 “EC track density climatology” is presented and “Units are the number of cyclones per 1.5◦ spherical cap per month”; and in Figure 2 “AR frequency climatology” is presented with “Units are the percentage of time-steps detected as AR”. Before a climatology can be presented and discussed, a clarification must be presented, and methods should detail the definitions and how this has been computed. In both cases, the authors must clarify if the systems are being tracked and considered or if the systems’ timesteps are considered independently. All the processes to produce the cyclones’ tracks and ARs datasets must be better explained. From my understanding, the systems are considered, but this is not clear from the discussion and captions of Figures 1, 2 and 8.
• The North Atlantic region [25-65ºN; 80◦W-10◦E] is considered. The authors should discuss the artefact over the western boundaries of the domain. A buffer area should be considered for the identification and tracking of the systems.
• The method to identify the concurrences of Extratropical Cyclones and ARs also needs further explanation: it is presented through Figure 3 and the Xynthia case study, but this example elucidates the doubts and need for clarification on the methods. From this example, five timesteps are consistent with the concurrence of the cyclone under explosive development and the occurrence of the AR. The shaded areas in Figure 3 that depict the regions identified as ARs should have some correspondence with the cyclone track and should be described in the text as well. It is not clear to me how many times cyclone Xynthia and the concurrent AR are considered for the climatological assessment. I would say Figures 1 and 2 correspond to timesteps – and not Cyclones/AR.
• As mentioned previously, the method to identify the concurrences (Ln 130-134) must be further detailed. Please discuss the choice of the Maximum Deepening Point (MDP). An explanation should be given for the choice of the 1500 km threshold. It is not clear if a sensitivity analysis was performed, nor if this metric is constant for all cyclone’s sequential timesteps. To the best of my understanding, each detected AR candidate may have more than one grid point being detected as AR in the same timestep. Certainly, the authors have considered this and all these aspects should be presented and discussed in the methods section. Additionally, how this method differs from Eiras-Barca et al. (2018) should be highlighted.
• The clarification of these methodological aspects is vital for the discussion of the results: authors should clearly state if this study evaluates “the concurrence of ECs and ARs in the ERA5 reanalyses and we compare it with those obtained in climate models” considering only once each EC and AR.
• All the analysis is performed for the extended winter period (October to March). This should be indicated in the figure captions. Please clarify, in the methodology, how the inter-seasonal variability is defined if only the extended winter season is considered. I suppose the authors mean interannual variability.
• CMIP6 models’ information and discussion of results: additional detail should be included for the choice of one single member for each model and not the ensemble – how this particular member has been selected and how this choice may affect the final results. This should be included in the methodology and the discussion. Please refer to whether one may state that model X overestimates/underestimates the results or if model Y is more adequate for the analysis if only one member has been used. Please also discuss if the biases quantification is reliable. To the best of my knowledge, this assessment is not enough to make a comparison between models. A multimodel ensemble framework with varied combinations of GCMs is extremely useful and allows for reducing the uncertainty in climate projections for future scenarios and for a tendency assessment, but it can hardly be used to intercompare models when only one member is used. Please, define “the internal variability of the datasets” (ln 161) in the methodology section and how it is assessed in this manuscript.
• It would be useful if the results presented in the Appendix should be accompanied by a short description and discussion. Please avoid using expressions like “little change” (ln 190-191) or “lower-than-expected increase” (ln 194)– please quantify. A percentage could be added to tables.
• Ln 200 – this sentence deserves additional information or a reference. It is out of context in this paragraph. These would be relevant results but evidence must be shown.
• Conclusions: please discuss what is the novelty, for the present period, from the literature. The sentence “The fact that AR concurrences are larger after the MDP suggests that mature ECs (when they are deeper) can facilitate the formation of ARs in their surroundings” (ln 272-273) deserves to be further discussed and justified. Firstly, it is well known that the detecting and tracking methods still have large uncertainty in detecting the absolute minimum central pressure of an extratropical cyclone; secondly, the difference should be quantified; finally, and most importantly, the only conclusion that these results allow us to obtain, in this state, is that additional AR are detected – we cannot state that they only formed at that particular timestep. 

Summary

In this study, the authors investigated the future changes of compound explosive cyclones (ECs) and atmospheric rivers (ARs) in the North Atlantic using the simulation data from six CMIP6 models. Different from previous papers, this study focused on the future changes of compound ECs and ARs, which usually develop rapidly due to the strong diabatic feedback and are closely related to extreme precipitation and wind. The authors found that there is a significant and systematic future increase in the EC-AR concurrences, especially over Western Europe in the high-emission scenario. Overall, this study investigated the future changes of ECs and ARs from a novel perspective and the paper is well written and organized. However, I have some major concerns, especially for the methods, and general comments listed below for the authors’ consideration.

Major Concerns:

(1) My biggest concern is the sensitivity of the conclusions in this study to the AR and cyclone tracking methods. For example, there are many different AR detection methods with large differences as summarized in some papers from the Atmospheric River Tracking Method Intercomparison Project (ARTMIP; e.g., O’Brien et al. 2020, 2022; Shields et al. 2019). If a different AR detection method or a different cyclone tracking method is used, will that have a significant impact on the conclusions about the EC-AR concurrences?

(2) Section 3.3: “an extratropical cyclone (EC or non-EC) is linked to an AR by detecting the presence of an AR within a 1500 km of the cyclone center” (lines 133-134). It may be oversimplified to use 1500 km distance to determine the concurrence between cyclone and AR. If a cyclone and an AR are dynamically associated with each other, the AR is usually located over the south to southeast side (the position of the low-level jet stream ahead of the cold front) of the cyclone center. In that case, it makes sense to define the concurrence if an AR exists within 1500 km. However, if an AR is located over the north or northwest side of a cyclone center, I don’t think it is reasonable to say that the AR is dynamically associated with the cyclone even if the distance is within 1500 km.

(3) Fig.4, the description in Section 4, and many other places throughout the manuscript: It is a smart way to use the maximum deepening point (MDP) as a reference point. However, the lifetime of extratropical cyclones has a large variability, varying from a couple of days to over one week. So it is very arbitrary to say that 36 hours before MDP (-36 h) is “the initial stages of cyclone formation”, +36 h is “the dissipation stages of the cyclones”, and from -36 h to +36 h is “the lifetime of the ECs”.

(4) The domain for analysis is 25N-65N and 80W-10E in this study (line 74). Did the authors use only the data within this domain for AR/cyclone tracking and EC-AR concurrence determination? If yes, there will be a boundary issue, especially for the cyclones and ARs near the boundaries, since both AR and cyclone detections have thresholds for moving distance and existing time. For example, in Fig.1 the EC track density is unreasonably high along the western boundary of the domain; in Fig.8 the rate of coincidence tends to be very small close to the boundaries. My concern is that the boundary issue may have impacts on the conclusions, especially for Western Europe, the area around the south tip of Greenland, and other places close to the boundary.

(5) For the horizontal resolution, ERA5 is 0.25 degrees while the six GCMs are quite different, varying from ~0.7 degrees to ~2.0 degrees. Did the authors interpolate the data to a common grid before analysis? Will the different horizontal resolutions have any impact on the conclusion? For example, the AR intensity is defined as the maximum IVT, but the different horizontal resolutions may have an impact on the maximum IVT across different models and ERA5. As a result, some differences across different models and ERA5 might be a data resolution issue, not the real model bias.

(6) A few concerns/questions about the AR and cyclone tracking methods.

Line 118: In addition to detecting ridges in the IVT field, there is a IVT minimum threshold of 250 kg/m/s. However, the IVT values have large variability from low to high latitudes. Will the 250 kg/m/s minimum threshold be too high for the ARs at high latitudes, like the area around the south tip of Greenland, near or higher than 60N?

Line 118: The AR candidates should have an area larger than 4 × 10^5 km^−2. Are there any requirements for the AR shape? ARs are usually defined as a long and narrow corridor of strong water vapor transport.

Line 119: “... concatenated if at least one grid point …” Similar to major comment (5), if the data were not interpolated to a common grid, it is a concern since models have quite different horizontal resolution, which means the “one grid” threshold is different across different models (~0.7 degrees to ~2.0 degrees) and ERA5 (0.25 degrees).

Lines 91 and 92: “not exceed 6 GCD degrees” and “at least 12 GCD degrees”. Are there any specific reasons for using 6 GCD and 12 GCD?

Minor Comments

(1) Line 11: “worst-case scenario”, I think it would be better to use high-emission scenario.

(2) Why did the authors select those six CMIP6 models while there are many other GCMs in CMIP6?

(3) Line 101: “These results agree with Priestley and Catto (2022) and Zappa et al. (2013) …” It’s worth noting that Zappa et al. 2013 used CMIP5, not CMIP6.

(4) Line 103: “Figure 1c,d”, do you mean Figure 1 c and f?

(5) Line 134: delete “a” before “1500 km”.

(6) Fig.1 and Fig.2: It would be very helpful to show the percentage difference in addition to the absolute difference of cyclone track density and AR frequency. Same for the other difference figures.

(7) Fig.2: The unit of AR frequency is % in the figure but the values are fraction (0.00~0.12).

(8) Fig.4: The difference across different lines (models) is not very clear. Maybe use different colors for different models?

(9) Line 144 and some other places: “MDP point”. “Point” is redundant since MDP is maximum deepening point.

(10) Line 149: “… favours the detection of an AR in its surroundings”. “detection” is not suitable here, maybe change it to “existence”.

(11) Line 158: “the standard deviation of the rate of coincidence”. Is that calculated using the coincidence rate at MDP or from -36 h to +36 h?

(12) Line 173 and 175: 0.08 and 0.05 are the model biases of what?  Coincidence rate? For the average of all models or the model with the maximum bias?

(13) Line 227: “The AR intensity for ERA5 is larger than any model for the historical period because ERA5’s resolution is almost 4 times higher than the CIMP6 models, and attains larger values of IVT-max.” So the difference of AR intensity between ERA5 and models might be a data resolution issue? This is the same question as my major concern (5).

(14) Fig.6: Does the Non-EC AR means the ARs without an explosive cyclone or the ARs without any cyclone (no matter weak or explosive)?

(15) Fig.8e, there are many areas show a large increase in coincidence rate with model agreement. Why did the authors only emphasize the western Europe in the Abstract (line 12)?

(16) Fig.8 a and b, the brown-green color map gives the readers the impression that the brown and green areas are opposite rather than low to high. I would suggest using a different color map.

The title of this manuscript does well to summarize what to expect. The overall quality of the manuscript is good. The writing is clear. The authors have carried out a substantial amount of application of existing Lagrangian tracking algorithms to reanalysis and climate model data, and then they have done some interesting sorting of the data. Ultimately the results suggest only a small signal amidst the noise of midlatitude storms. However, for the important issue of explosive cyclones, perhaps a null result is still useful. As ever, I do think we need to be cautious because there is always the lingering doubt about these model’s ability to capture the physics of explosive cyclones.

I appreciate the author’s choice on method of tracking ARs using the Laplacian of the IVT, so that they are not just picking up the thermodynamic signal. However, I have a fundamental issue with the way in which some concepts are explained in the introduction, and some questions about the interpretation of the results. These issues and questions are described below. 

Major Comment:

Lines 37 – 41: This is a section in the introduction in which the authors seek to make a physical explanation for why the presence of atmospheric rivers (ARs) impact explosive cyclones (ECs). However, I do not think these studies prove cause versus effect. I posit that in many, or perhaps half of the cases, it might be the case that rapidly intensifying cyclones have substantial upper-level forcing that drives more poleward transport of water vapor. This would lead to more ARs found in the surroundings of ECs, but the cause is not the upper-level circulation, not the latent heat release. (Isn’t this substantiated by your result that more ARs are found to be associated with the cyclones after their maximum deepening point? – line 147.)

I want to make clear about my point: If the upper-level circulation is held fixed (e.g., in a modeling study for a single event or a baroclinic wave), then the storm intensity and intensification rate will increase with more water vapor (i.e., the presence of a stronger AR). However, that is different from saying that the presence of ARs leads to explosive cyclones. For me, the explanation provided by the authors in this section needs more nuance and explanation.

Relatedly, the papers being referenced in this section all state that their results “suggest” a relationship, but none of them claim it to be conclusive. So, I request that the authors add more caveats and details to this explanation. This would impact the introduction, the interpretation of results and the conclusions.    

Minor Comments:

Line 180: Figure 4 (and all similar plots): I suggest you replace h with the word hours to reduce any chances for confusion from a viewer. 

Line 215: I am a bit puzzled by the AR intensity analysis in Section 5.2. In the methods section, you do a good job of explaining why the use of the Laplacian is important. Now you are back to working with IVT itself. Why? Given that storm forcing from latent heating (e.g., the change in diabatic potential vorticity) is related to the gradient of the heating, not the absolute value, this choice of defining AR intensity based on the absolute value should be explained in more detail.

Line 247-8: Here you state: 

“The results from ERA5 show the same behaviour for both types of cyclones but with lower intensity”. Could you clarify this sentence to explain what intensity is referring to? Is it the intensity of the relationship or the intensity of the cyclones? If it is the intensity of the relationship, then perhaps you should also include a sentence or two here reminding the readers of the multiple reasons for potential biases in the models. 

Additional papers on water vapor and storm intensity that must be cited and discussed when discussing the results, given the nature of this manuscript: 

Pfahl, S. and Sprenger, M.: On the relationship between extratropical cyclone precipitation and intensity, Geophys. Res. Lett., 43, 1752–1758, 2016 https://doi.org/10.1002/2016GL068018

Booth, J. F., Naud, C. M., and Jeyaratnam, J.: Extratropical Cyclone Precipitation Life Cycles: A Satellite-Based Analysis, Geophys. Res. Lett., 45, 8647–8654, 2018

https://doi.org/10.1029/2018GL078977

Sinclair, V. A. and Catto, J. L.: The relationship between extratropical cyclone intensity and precipitation in idealised current and future climate, Weather and Climate Dynamics, vol. 4, no. 3, pp. 567–589. doi:10.5194/wcd-4-567-2023, 2023