Detection and global climatology of two types of cyclone clustering

Weijenborg, Chris; Spengler, Thomas

doi:10.5194/egusphere-2024-3404

Preprints

https://doi.org/10.5194/egusphere-2024-3404

Preprints

11 Nov 2024

| 11 Nov 2024

Detection and global climatology of two types of cyclone clustering

Chris Weijenborg and Thomas Spengler

Abstract. Cyclone clustering, the swift succession of multiple extratropical cyclones during a short period of time, is often associated with weather extremes and characterised by a strong atmospheric jet and enhanced baroclinicity. While several diagnostics exist to detect cyclone clustering, most focus on a regional impact. We introduce a novel global detection for cyclone clustering, inspired by the original idea of cyclone families by Bjerknes and Solberg, in which individual cyclones follow a similar track. We further subdivide cyclone clusters into two types, a 'Bjerknes' type and a stagnant type. The former is associated with cyclones that follow each other over a minimum distance, whereas the stagnant type requires a proximity over time while these cyclones do not move much in space.

We find that cyclone clustering is most frequent along the storm tracks, with more cyclone clustering during winter compared to summer. The majority of cyclone clustering occurs just south of the main storm tracks in the Atlantic and Pacific basins. In the Southern Hemisphere, most cyclone clustering is found in the South-Indian Ocean. Bjerknes type cyclone clustering is associated with stronger cyclones compared to non-clustered cyclones, while for the stagnant type this intensity difference is less pronounced. This effect is strongest for the North Atlantic and North Pacific, while clustered cyclones in the South Indian Ocean are generally not much stronger. The cyclone intensity within the Bjerknes type does not decrease during a cluster, while in contrast secondary cyclones of the stagnant type are significantly weaker than primary cyclones. This suggests that these two types of cyclone clustering are dynamically different.

Received: 31 Oct 2024 – Discussion started: 11 Nov 2024

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 8820 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (8820 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

09 Mar 2026

Detection and global climatology of two types of spatio-temporal clustering of extratropical cyclones

Chris Weijenborg and Thomas Spengler

Weather Clim. Dynam., 7, 475–488, https://doi.org/10.5194/wcd-7-475-2026,https://doi.org/10.5194/wcd-7-475-2026, 2026

Short summary

Chris Weijenborg and Thomas Spengler

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-3404', Anonymous Referee #1, 17 Dec 2024

See attached pdf.

Citation: https://doi.org/10.5194/egusphere-2024-3404-RC1
- AC1: 'Reply on RC1', Thomas Spengler, 24 Feb 2025
  
  Please see pdf for response letter.
  
  Citation: https://doi.org/10.5194/egusphere-2024-3404-AC1
RC2:
'Comment on egusphere-2024-3404', Anonymous Referee #2, 27 Jan 2025

The aim of this paper is to provide a global climatology of cyclone clustering based on Reanalysis data. With this aim, the authors apply a cyclone cluster detection approach to cyclone track data derived with the Murray and Simmonds tracking method. I am a bit puzzled about the results, because the main result is simply the well-known storm tracks, which can be derived easily from both eulerian and lagrangian methods (e.g. Hoskins and Hodges, 2002). Following the definition of Mailier et al (2006, also their Fig. 6), clustering is a synonym of overdispersion, meaning that there are periods of time where the number of occurrences within a defined period of time significantly differs from the expected value. Indeed, the areas where cyclone clustering is identified over the North Atlantic Ocean correspond to the areas on the flanks and downstream of the main storm track, and not the main storm track itself (see review by Dacre and Pinto, 2020, in particular their Figure 2). Indeed, over the core of main storm track, cyclone occurrences tends to be random, and at the its entrance underdispersed (see also their Fig. 2). Thus, the results included in this manuscript are clearly in disagreement with many studies published on this topic. So, while I understand the idea of the authors, the methodology simply identifies areas where “cyclones tend to follow each other”, which are per definition the storm tracks themselves. I would also recommend to consider a minimum intensity and lifetime for the cyclones (lines 85ff), as otherwise the whole picture may be dominated by non-synoptic-relevant systems.
Given the very fundamental question marks regarding the methodology, which I do not think it is adequate to answer the research questions and the objectives posed, I have refrained to make detailed comments on the text. Based on the above, I unfortunately cannot recommend this manuscript for publication. However, I do agree that a strongly re-worked manuscript would be an important contribution to this field of work, given the lack of studies outside of the North Atlantic Basin on cyclone clustering (see research gaps in Dacre and Pinto, 2020).

Dacre and Pinto, 2020, https://doi.org/10.1038/s41612-020-00152-9
Hoskins and Hodges, 2002, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2
Mailier et al. 2006, https://doi.org/10.1175/MWR3160.1

Citation: https://doi.org/10.5194/egusphere-2024-3404-RC2
- AC2: 'Reply on RC2', Thomas Spengler, 24 Feb 2025
  
  Please see pdf for response letter.
  
  Citation: https://doi.org/10.5194/egusphere-2024-3404-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-3404', Anonymous Referee #1, 17 Dec 2024

See attached pdf.

Citation: https://doi.org/10.5194/egusphere-2024-3404-RC1
- AC1: 'Reply on RC1', Thomas Spengler, 24 Feb 2025
  
  Please see pdf for response letter.
  
  Citation: https://doi.org/10.5194/egusphere-2024-3404-AC1
RC2:
'Comment on egusphere-2024-3404', Anonymous Referee #2, 27 Jan 2025

The aim of this paper is to provide a global climatology of cyclone clustering based on Reanalysis data. With this aim, the authors apply a cyclone cluster detection approach to cyclone track data derived with the Murray and Simmonds tracking method. I am a bit puzzled about the results, because the main result is simply the well-known storm tracks, which can be derived easily from both eulerian and lagrangian methods (e.g. Hoskins and Hodges, 2002). Following the definition of Mailier et al (2006, also their Fig. 6), clustering is a synonym of overdispersion, meaning that there are periods of time where the number of occurrences within a defined period of time significantly differs from the expected value. Indeed, the areas where cyclone clustering is identified over the North Atlantic Ocean correspond to the areas on the flanks and downstream of the main storm track, and not the main storm track itself (see review by Dacre and Pinto, 2020, in particular their Figure 2). Indeed, over the core of main storm track, cyclone occurrences tends to be random, and at the its entrance underdispersed (see also their Fig. 2). Thus, the results included in this manuscript are clearly in disagreement with many studies published on this topic. So, while I understand the idea of the authors, the methodology simply identifies areas where “cyclones tend to follow each other”, which are per definition the storm tracks themselves. I would also recommend to consider a minimum intensity and lifetime for the cyclones (lines 85ff), as otherwise the whole picture may be dominated by non-synoptic-relevant systems.
Given the very fundamental question marks regarding the methodology, which I do not think it is adequate to answer the research questions and the objectives posed, I have refrained to make detailed comments on the text. Based on the above, I unfortunately cannot recommend this manuscript for publication. However, I do agree that a strongly re-worked manuscript would be an important contribution to this field of work, given the lack of studies outside of the North Atlantic Basin on cyclone clustering (see research gaps in Dacre and Pinto, 2020).

Dacre and Pinto, 2020, https://doi.org/10.1038/s41612-020-00152-9
Hoskins and Hodges, 2002, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2
Mailier et al. 2006, https://doi.org/10.1175/MWR3160.1

Citation: https://doi.org/10.5194/egusphere-2024-3404-RC2
- AC2: 'Reply on RC2', Thomas Spengler, 24 Feb 2025
  
  Please see pdf for response letter.
  
  Citation: https://doi.org/10.5194/egusphere-2024-3404-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Thomas Spengler on behalf of the Authors (11 Apr 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (29 Apr 2025) by Michael Riemer

RR by Anonymous Referee #3 (12 Jun 2025)

For final publication, the manuscript should be

accepted as is

accepted subject to technical corrections

accepted subject to minor revisions

reconsidered after major revisions

rejected

Were a revised manuscript to be sent for another round of reviews:

I would be willing to review the revised manuscript.

I would not be willing to review the revised manuscript.

Suggestions for revision or reasons for rejection

In this study the authors have developed a novel algorithm to detect cyclones that fall into a category that they refer to as cyclone families. In other words, cyclones that occur close to each other in space-time. The paper is mostly well-written and there are some interesting results. However, I think it needs to be made clearer that this paper is not representing a new and better way to identify cyclone clustering, as the authors seem to suggest. Rather it is a method to identify a specific subset of cyclones – those that follow each other or have similar tracks. The paper mentions other studies that have considered cyclone clustering from a statistical point of view, or a regional point of view. Both of these previous types of studies consider that clustering is the grouping together of cyclones in time and space that is unusual given the mean frequency of cyclones. The reason for this is the potential impacts of these clustering events. In the present study, it is not clear to me whether the large fraction of clustered cyclones in the main stormtracks has a similar impact.

So, I recommend some clarification that this algorithm is specifically looking for a subset of cyclones that fall into the “cyclone families” group (although since it does not consider secondary cyclogenesis specifically, it is a somewhat expanded group).

I give some specific comments below.

1. Perhaps the title could be changed to say “cyclone families” instead of “cyclone clustering”.
2. Introduction: the review of the literature as it is presented seems to suggest that the current study is less restrictive than the dispersion diagnostic or the regional studies. However, I would argue that this new method is quite restrictive in a different way. By requiring the tracks to follow similar paths, it excludes some of the serial clustering that would occur from cyclones coming from different directions (which seems plausible if a cyclone developed on a trailing cold front). Perhaps the introduction could be slightly reframed to discuss what has gone before and to highlight that this is a new and different way of looking at cyclone families, rather than a better way to identify clustering.
3. Lines 52-57: This paper does not seem particularly relevant to the current study. It is only considering that tracks might have similar characteristics, but does not consider any time constraint.
4. Line 75: “cyclone mechanism” should perhaps be “cyclone development mechanism”. However, I don’t think the paper does look at this, so mentioning the latent heating here is misleading.
5. Figure 1: The panels need labels as given in the caption, and the caption contains unnecessary “in” on the second line.
6. Line 98: “Conceptionally” should be “conceptually”.
7. Line 145: Wherever the “Laplacian” is mentioned, it should really be the Laplacian of mean sea level pressure. This is important for the figures and captions also.
8. Line 168: The cyclone clustering is not abnormal. The frequency of clustering is highest in the end of the stormtracks. Please change the wording here, and check the bibliography/referencing in this line.
9. Line 194: I think the figures of 35-40% quoted here are incorrect, looking at Figure 2f, where the values look much higher than this.
10. Line 226: I find the term “length of cluster” to be confusing since it actually refers to the number of tracks within a cluster. Surely size of cluster would be more appropriate?
11. Line 247: “stronger” should be “larger”.
12. Line 258: “between” should be “of”.
13. Figure 7: The figures need units of the Laplacian of MSLP.
14. Figure 9: The labels on these panels are too small to read on an A4 page.
15. Line 267: Check the parentheses for the referencing.
16. Conclusions: Again, I think it is important here to make it clear that this new study is not offering an algorithm that is better than others, rather it is detecting a specific subset of cyclones that are within the extratropical stormtracks. I think it should not be compared so closely to the previous studies as it is specifically choosing something different to look at. I also think that since this method does not use a front identification method, it cannot distinguish secondary cyclogenesis as it is defined in the literature. Therefore, the conclusions need some work to demonstrate the novelty of this method and the implications for future work.

Hide

RR by Anonymous Referee #2 (22 Jun 2025)

For final publication, the manuscript should be

accepted as is

accepted subject to technical corrections

accepted subject to minor revisions

reconsidered after major revisions

rejected

Were a revised manuscript to be sent for another round of reviews:

I would be willing to review the revised manuscript.

I would not be willing to review the revised manuscript.

Suggestions for revision or reasons for rejection

Formal review for egusphere-2024-3404 (2nd round)
Detection and global climatology of two types of cyclone clustering
Chris Weijenborg and Thomas Spengler

The authors have revised their manuscript but have largely disregarded my comments regarding the approach and the choice of nomenclature, which is unfortunate. Thus, my main concerns remain (see my initial comments below), and in its current form I cannot recommend this article to be accepted for publication, but rather a rejection. Primarily, the nomenclature used and the discussion with the available literature are not adequate.

Nevertheless, I would like to suggest to the editor and the authors a way forward, so that the paper can be revised and hopefully published. I do think the manuscript contains valuable material, but it should be presented without contradicting the current nomenclature and available literature. Two major points are indicated, and would imply a major revision of the manuscript (also my decision on the online system):

1. There is a 2020’ review paper on the topic of cyclone clustering (Dacre and Pinto, 2020) that I have suggested in my previous review and the authors have chosen not to consider it. I would like to insist that the authors consider the wide material included in this review, regarding the different types and metrics of clustering, followed by climatologies, considerations on different time scales, associated large-scale dynamics and a gap analysis (see also my comment on this in the initial review). I am sure the review article will be helpful to rework the introduction and rephrase aims of this study in terms of the available literature. Also the differences between “cyclone families” and “cyclone clustering” are discussed in this review paper.

2. The seminal work by Bjerkness and colleagues never mentions the word “cyclone clustering”, but rather the concept of “cyclone families”. I agree that the methodology the authors have developed is good in identifying exactly that, namely “cyclone families”. So I would like to strongly suggest that the authors change the title of the manuscript and generally the wording used everywhere to “cyclone families”. This would be a much more precise description of what the authors are actually analyzing with their methodology and would avoid confusion with many papers published about cyclone clustering over the last twenty years.

Additionally, I would like to add a third by minor point.
3. Please note that Pinto et al. (2014) have not only considered secondary cyclogenesis in the trailing cold fronts (upstream cyclone development) as a possible reason for cyclone clustering, but rather also of the role of the eddy driven jet stream, Rossby wave-breaking, and in particular the possibility of downstream cyclone developments (their figure 8). Moreover, clustering can also happen by chance. This is also discussed in the review paper from Dacre and Pinto, 2020). Thus, the paragraph starting in line 45 should be updated.

I would be willing to revise the manuscript upon re-submission

Dacre and Pinto 2020: https://doi.org/10.1038/s41612-020-00152-9
Pinto et al., 2014: https://doi.org/10.1002/2014JD022305

-------------------------------------------------------------------
Comments from the first formal review:
The aim of this paper is to provide a global climatology of cyclone clustering based on Reanalysis data. With this aim, the authors apply a cyclone cluster detection approach to cyclone track data derived with the Murray and Simmonds tracking method. I am a bit puzzled about the results, because the main result is simply the well-known storm tracks, which can be derived easily from both eulerian and lagrangian methods (e.g. Hoskins and Hodges, 2002). Following the definition of Mailier et al (2006, also their Fig. 6), clustering is a synonym of overdispersion, meaning that there are periods of time where the number of occurrences within a defined period of time significantly differs from the expected value. Indeed, the areas where cyclone clustering is identified over the North Atlantic Ocean correspond to the areas on the flanks and downstream of the main storm track, and not the main storm track itself (see review by Dacre and Pinto, 2020, in particular their Figure 2). Indeed, over the core of main storm track, cyclone occurrences tends to be random, and at the its entrance underdispersed (see also their Fig. 2). Thus, the results included in this manuscript are clearly in disagreement with many studies published on this topic. So, while I understand the idea of the authors, the methodology simply identifies areas where “cyclones tend to follow each other”, which are per definition the storm tracks themselves. I would also recommend to consider a minimum intensity and lifetime for the cyclones (lines 85ff), as otherwise the whole picture may be dominated by non-synoptic-relevant systems.
Given the very fundamental question marks regarding the methodology, which I do not think it is adequate to answer the research questions and the objectives posed, I have refrained to make detailed comments on the text. Based on the above, I unfortunately cannot recommend this manuscript for publication. However, I do agree that a strongly re-worked manuscript would be an important contribution to this field of work, given the lack of studies outside of the North Atlantic Basin on cyclone clustering (see research gaps in Dacre and Pinto, 2020).

Dacre and Pinto, 2020, https://doi.org/10.1038/s41612-020-00152-9
Hoskins and Hodges, 2002, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2
Mailier et al. 2006, https://doi.org/10.1175/MWR3160.1

Citation: https://doi.org/10.5194/egusphere-2024-3404-RC2

Hide

ED: Publish subject to revisions (further review by editor and referees) (07 Jul 2025) by Michael Riemer

Dear authors,
Your manuscript has been scrutinized by three expert reviewers in the subject of cyclone clustering. (Unfortunately, one of the reviewers from the first round could not follow up in the second round, and thus I brought in a new reviewer in the second round.) All reviewers agree that your manuscript contains interesting results and valuable information. All three reviewers further agree, however, that the definition of cyclone clustering used in your manuscript is not consistent with a definition that the community has established over the previous couple of decades. Dacre and Pinot (2020) provide a review that seems to be well received by the community (based on citation metrics). All reviewers agree that your definition is consistent with the concept of cyclone families. I trust the expert reviewers that the distinction between cyclone clustering and cyclone families is important, and I agree with the reviewers that blurring an important distinction leads to confusion. Referee #3 provides a clear example of a scenario, in which cyclone families (and your perspective) are not congruent with clustering: “By requiring the tracks to follow similar paths, it excludes some of the serial clustering that would occur from cyclones coming from different directions (which seems plausible if a cyclone developed on a trailing cold front).”
Both reviewers of the second round suggest that the way forward for this manuscript is to refrain from the claim to identify cyclone clustering but rather to use the term cyclone families. I follow the clear (and strong) statement of reviewer #2 in that “the methodology the authors have developed is good in identifying exactly that, namely “cyclone families”. So I would like to strongly suggest that the authors change the title of the manuscript and generally the wording used everywhere to “cyclone families”. This would be a much more precise description of what the authors are actually analyzing with their methodology and would avoid confusion with many papers published about cyclone clustering over the last twenty years.”
Taking a “fresh look” at “established” science problems is an integral part of scientific rigor and advancement. Alternative approaches should not be dismissed lightly. The reviews, however, are convincing in arguing that the manuscript does not take a fresh look at cyclone clustering, clearly put by reviewer #3: “… it is important here to make it clear that this new study is not offering an algorithm that is better than others, rather it is detecting a specific subset of cyclones that are within the extratropical stormtracks. I think it should not be compared so closely to the previous studies as it is specifically choosing something different to look at.” (my emphasis) To be accepted for publication, I would like the authors to follow the consensus recommendations of the reviewers and change terminology. The introduction and the conclusion sections are suitable places to discuss the relation between cyclone families and clustering, where such a discussion would need to acknowledge the definition of clustering as accepted by the community, e.g., as reviewed by Dacre and Pinot (2020).

Best wishes, Michael

Hide

AR by Thomas Spengler on behalf of the Authors (29 Oct 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (05 Nov 2025) by Michael Riemer

RR by Anonymous Referee #3 (26 Nov 2025)

ED: Publish subject to revisions (further review by editor and referees) (19 Dec 2025) by Michael Riemer

AR by Thomas Spengler on behalf of the Authors (16 Jan 2026) Author's response Author's tracked changes Manuscript

ED: Publish as is (29 Jan 2026) by Michael Riemer

AR by Thomas Spengler on behalf of the Authors (19 Feb 2026)

Journal article(s) based on this preprint

09 Mar 2026

Detection and global climatology of two types of spatio-temporal clustering of extratropical cyclones

Chris Weijenborg and Thomas Spengler

Weather Clim. Dynam., 7, 475–488, https://doi.org/10.5194/wcd-7-475-2026,https://doi.org/10.5194/wcd-7-475-2026, 2026

Short summary

Chris Weijenborg and Thomas Spengler

Viewed

Total article views: 768 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
559	168	41	768	34	56

HTML: 559
PDF: 168
XML: 41
Total: 768
BibTeX: 34
EndNote: 56

Views and downloads (calculated since 11 Nov 2024)

Month	HTML	PDF	XML	Total
Nov 2024	116	10	8	134
Dec 2024	33	11	1	45
Jan 2025	31	6	4	41
Feb 2025	39	20	3	62
Mar 2025	20	7	0	27
Apr 2025	12	13	0	25
May 2025	18	4	3	25
Jun 2025	27	7	5	39
Jul 2025	26	7	1	34
Aug 2025	19	6	1	26
Sep 2025	13	3	1	17
Oct 2025	24	9	1	34
Nov 2025	39	19	4	62
Dec 2025	49	16	4	69
Jan 2026	69	13	5	87
Feb 2026	18	9	0	27
Mar 2026	6	8	0	14

Cumulative views and downloads (calculated since 11 Nov 2024)

Month	HTML	PDF	XML	Total
Nov 2024	116	10	8	134
Dec 2024	33	11	1	45
Jan 2025	31	6	4	41
Feb 2025	39	20	3	62
Mar 2025	20	7	0	27
Apr 2025	12	13	0	25
May 2025	18	4	3	25
Jun 2025	27	7	5	39
Jul 2025	26	7	1	34
Aug 2025	19	6	1	26
Sep 2025	13	3	1	17
Oct 2025	24	9	1	34
Nov 2025	39	19	4	62
Dec 2025	49	16	4	69
Jan 2026	69	13	5	87
Feb 2026	18	9	0	27
Mar 2026	6	8	0	14

Viewed (geographical distribution)

Total article views: 753 (including HTML, PDF, and XML) Thereof 753 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 10 Mar 2026

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (8820 KB)
Metadata XML

Short summary

The swift succession of storms, referred to as cyclone clustering, is often associated with weather extremes. We introduce a detection scheme for these events and subdivide these into two types. One type is associated with storms that follow each other in space, whereas the other type requires a proximity over time. Cyclone clustering is more frequent during winter and the first type is associated with stronger storms, suggesting that the two types emerge due to different mechanisms.


Total:	0
HTML:	0
PDF:	0
XML:	0