the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Rossby wave packets driving concurrent and non-concurrent heatwaves in the Northern and Southern Hemisphere mid-latitudes
Abstract. Heatwaves that occur simultaneously over several regions, termed concurrent heatwaves, pose compounding threats to society and the environment. Amplified quasi-stationary circumglobal Rossby wave patterns (CGWPs) and high-amplitude transient non-circumglobal Rossby Wave Packets (RWPs) have been proposed as two possible explanations for the occurrence of heatwaves. The relation of these mechanisms for heatwaves has been investigated over different timescales, but their relevance for concurrent and non-concurrent heatwaves remains unexplored. In the present study we focus on daily time scales and investigate the relevance of the global CGWP amplitude and of the local RWP amplitude for the occurrence of concurrent and non-concurrent heatwaves over the Northern Hemisphere (NH) and Southern Hemisphere (SH) mid-latitudes. To distinguish between concurrent and non-concurrent heatwaves we apply a k-means clustering algorithm on all heatwaves detected in ERA5 reanalysis data within the 1959–2021 period. We identify 42 spatial clusters of heatwaves in the NH and 53 in the SH. In all identified clusters, mid-latitude heatwaves typically occur at the leading edge of RWPs where Rossby wave breaking takes place in the form of ridge building or block formation. No specific zonal wavenumber is more frequently related to the concurrent or to the non-concurrent heatwave category. However, for high global CGWP amplitudes concurrent heatwaves occur more often in the NH when the dominant zonal wavenumber is k = 7, and non-concurrent heatwaves occur more often in the SH for k = 5. The mid-latitude regions exhibiting increased heatwave probabilities under the influence of either global or local high wave amplitude, include western North America, central Europe, Black Sea, Tibet, the southwest coast of Australia, as well as the southern Indian and Atlantic Oceans. Over those regions, the local high amplitude RWPs increase heatwave probabilities by a factor ranging from 4 to 7, whereas the maximum factor for high global CGWP amplitude is 2. These results emphasize the importance of the daily RWP amplitude and the weak association of the global CGWP amplitude to heatwave occurrence over the NH and SH mid-latitudes. This research for the first time investigates the underlying atmospheric dynamical processes that contribute to the development of concurrent and non-concurrent heat extremes, a crucial step towards improving our understanding and ability to predict heatwave variability at weather and longer time scales.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(53701 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-3088', Anonymous Referee #1, 05 Mar 2024
General Comments
This paper investigates the role of Rossby waves on heatwaves, and contrasts the impacts of circumglobal Rossby waves with that of more localized Rossby wave packets, finding that locally amplified Rossby wave packets have a much stronger association with heatwaves than amplified circumglobal Rossby waves. I think this is an important and interesting finding. However, it seemed to me that this interesting result was somewhat hidden under details of other analysis in the paper, particularly that focussed on distinguishing between concurrent vs non-concurrent heatwaves, which I remain a little confused by. I would recommend to the authors that they consider re-structuring their manuscript to better highlight their key result. I think the comparison of concurrent vs non-concurrent heatwaves requires more thought, and perhaps a stronger separation/distinction between these two types of heatwave, and exploration of the sensitivity of results to the methodology.
Specific Comments
There is a lot of detail on the separation of the heatwaves into concurrent and non-concurrent heatwaves early on in the paper, however, it seems like mostly your results are not particularly sensitive to whether you are looking at concurrent or non-concurrent heatwaves (except perhaps the wavenumbers for the high-amplitude waves in Fig. 9), and so it seems perhaps unnecessary for readers to get through all of the information about how the concurrent vs non-current separation is performed before finding out it doesn’t really matter? I therefore recommend changing the order of your manuscript, and showing the top rows of Figs. 10 and 11 much earlier, highlighting this result, that heatwaves are associated with amplified local Rossby wave packets much more strongly than they are associated with amplified circumglobal Rossby waves. I think it is then very reasonable to hypothesise that this might be different for concurrent heatwaves, and to then explore that.
I am not convinced, based on the plots presented, that the methodology for distinguishing concurrent and non-concurrent heatwaves is an appropriate methodology. For example, in Fig. 1, there is much less difference between some of the concurrent clusters and some of the non-concurrent clusters (e.g. EU concurrent vs non-concurrent) than I was expecting given the definition. The concurrent clusters still appear to be a dominant heatwave in one location. Denoting a region that’s above +0.3 standard deviations in temperature as a heatwave seems surprising. Are the weak signals of the secondary heatwaves in this plot because it is a composite(/cluster) mean, and the cluster is really defined as a strong heatwave over one region (e.g. Europe) with some heatwaves elsewhere, but the region of the secondary heatwave varies between cluster members? I think more justification of this methodology is required here, with figures that clearly show the distinctions, or discussion around the fact that the distinctions are relatively subtle. Given that your results are mostly ‘there isn’t much difference between concurrent and non-concurrent heatwaves’ (at least that’s my understanding), I think you need to show a stronger difference between your clusters for concurrent and non-concurrent heatwave – I can’t tell whether that difference is there, and it’s just that the cluster mean is not a good way of showing it, or that there isn’t enough difference? Perhaps adjusting your method to exclude events that are close to the middle between concurrent and non-concurrent could help, I.e. defining non-concurrent events as when there is one region with >320 HW grid-points, and no other HW regions exceed 100 grid-points, rather than no other regions exceeding 320 grid-points. I realize this would reduce the number of events, but it would give you a clearer distinction, which would make your results that the distinction doesn’t generally matter seem more convincing.
I think that the Rossby wave breaking analysis is interesting, but it’s used and discussed relatively little for the amount of detail you go into in the methods, and it receives little attention in the results. Overall in section 2.5 for me it feels like not quite enough information for me to fee like I understand the method, and the information that is there is to complex to understand without additional complex. I'm unsure what's the same as previous methods, and what is new in this work. I think either explain the method more clearly, or, if the method isn't too relevant to your results, and is already explained in previous papers, just provide a clearer overview of the method with fewer details here. E.g. Line 203: I assume the elongated structures are identified in the PVU contour? Line 205: how are you defining length vs width? Line 216. It’s this depth of tropospheric PV structures that you use as your metric of RWB (i.e. in Fig 4), is that right? This needs to be stated here.
Separation into high amplitude CGRW days: the separation to only high amplitude days is not introduced clearly, and there is no explanation or justification for why this is done – was it because there were no statistically significant results for all wave amplitude days? I think it’s a reasonable approach, but it requires some explanation and introduction. I’m also interested in how much it reduces the number of heatwave days by? How much noisier does the dataset get when you restrict it to wave amplitude days > +1.5sigma? This is particularly relevant as the only statistically different results you see from climatology is when you restrict to high amplitude days. It is also unclear what the climatological probability densities are for only the high amplitude days – are the CCHW_HA statistically different from climatology_HA? I think there might be some interesting results here, but they aren’t very clear in the current format.
Other minor comments
Line 72. I think the research gap/question you’re asking could be defined more clearly here.
Clustering algorithm: how many days have no heatwaves? Is there a ‘no heatwave’ cluster or are these days excluded from the analysis?
Line 124. how did you decide on this optimal cluster number definition, and are your results sensitive to this definition?
Line 134. is there a minimum number of gridpoint separation of the regions for two heatwaves?
Line 174. “More specifically…” this is a rather confusing sentence structure, I recommend rephrasing.
Line 178: where A(t) is the amplitude…. (add is)
Section 2.4. Why give the equations for a Fourier transform, but not for the calculation of the RWP using the fourier transform? I don’t think the equations for a Fourier transform are necessary in the main paper, and I feel like the RWP equation is the more interesting one.
Lines 188-195: might be better off in the introduction, particularly when you discuss methodologies that you don’t use in the current paper.
Line 226: How sensitive are these results to your heatwave definition?
Line 230: As mentioned above, +0.3 sigma seems quite low for a heatwave definition, so the use of this threshold needs more justification here
Line 262. “The clusters that demonstrate persistent to….” this sentence is too difficult to interpret - is this showing the clusters that are more dominated by persistent heatwaves rather than non-persistent heatwaves? Is this saying that concurrent heatwaves tend to be more persistent?? Or can you not quite say that? I'm just unsure what physical interpretation to make here.
Line 275. “Finally…” is this statement related to persistence? If so, please make the connection more directly; if not, I think the sentence is out of place here.
I don’t think Fig. 7 ever gets referenced?
Line 300: k=5 and k=6 are also the most prevalent for the climatology, i.e. non-heatwave days? Also, k=4 is similarly important to k=6 in the SH).
Fig. 9. It’s hard to see some of the NCNW_HA dots when they lie behind the NCHW ones – maybe use a different shape?
Line 319: It doesn’t feel like this connection between heatwaves and high amplitude wave events was explored in the previous section particularly – I agree that it would be interesting, and suggest re-focussing section 3.2 to make this more clear (as stated in Specific Comments)
Line 321. And k = 5 in the NH? Even if it’s not statistically significant, it might be a signal?
Line 324. By ‘the two metrics’ do you mean the global fourier transform amplitude and the local RWP amplitude? Or do you include RWB here? I think making this distinction between the two metrics clearer throughout the paper would be useful.
Lines 325-335. This is a confusing paragraph, and might be better placed in the methods, with the equation to help understand this. Since this leads to what is, based on the title of the paper and my sense of your results, your most important result, it seems more important to be clear about this methodology than, for example, giving the equations for a fourier transform.
Line 329. I don't quite understand this. So, for a particular grid-point, the concurrent heatwave days probability is the fraction of all days within concurrent heatwave clusters during which this gridbox is part of a heatwave? And then the global amplitude for concurrent heatwaves is for any concurrent heatwave, regardless of where the heatwave is located? Please rephrase to make this clearer (see previous comment)
Line 335: how do you objectively define "a ridge over that location"?
Section 3.3. The CGWP days seem to have a very similar spatial pattern as for the RWP analysis (particularly noticeable in the SH) - do you think this is some feature of heatwaves, or is simply because local waves can show up in a global Fourier transform?
Line 369: the differences were surprisingly small to me though - comparing the middle columns of Figs 2 and 3? I agree that the difference seems to not be a global wave vs a local wave packet, and more is one vs multiple RWPs, but even this isn’t clear, as multiple RWPs are visible in the non-current heatwaves too, in Fig. 3. How do you explain this? My suspicion is that it’s related to my earlier comment, that the composites of EU concurrent and EU non-concurrent are not particularly well-separated.
Line 383. There isn't a lot of difference between concurrent and non-concurrent in Fig. 9, unless you look at high-amplitude events and k=7, which is very specific. Maybe this is sensitive to the methodology of separating concurrent and non-concurrent heatwaves, which should be determined and, if not, this might be one of the key messages to this paper - rather than highlighting the small differences between them, highlight how similar they are from the atmosphere’s perspective, with some small differences?
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC1 -
RC2: 'Comment on egusphere-2023-3088', Anonymous Referee #2, 07 Mar 2024
The topic of the study is of great interest. Spatially remote but concurrent extremes are gaining increasing attention in the literature, and there has been some debate as to the large-scale atmospheric drivers of such extremes. In particular, there is an active line of research supporting the notion that circumglobal planetary wave patterns are crucial for determining the co-occurrence of remote extremes, and testing this hypothesis robustly against other large-scale triggering mechanisms is a very relevant research undertaking. Challenging previous results is essential for advancing research, yet when one draws conclusions which partially counter an established body of previous work it is important to provide robust evidence to support them. My impression after reading this study is that the methodological choices steer to a great extent the results, and that the contextualisation and discussion of such results is lacking. As such, I cannot recommend this study for publication and believe that the edits necessary to make this a valuable contribution to the ongoing discussion of the large-scale triggers of single and concurrent heatwaves go beyond the scope of a major revision.
Major Comments
- The abstract provides a misleading context for the study and exaggerates its novelty.
- The authors state that: “Amplified quasi-stationary circumglobal Rossby wave patterns (CGWPs) and high-amplitude transient non-circumglobal Rossby Wave Packets (RWPs) have been proposed as two possible explanations for the occurrence of heatwaves. The relation of these mechanisms for heatwaves has been investigated over different timescales, but their relevance for concurrent and non-concurrent heatwaves remains unexplored”. I agree that the role of RWP for concurrent events is an understudied topic; on the other hand there is ample work on circumglobal Rossby waves and concurrent heatwaves (Kai Kornhuber and co-authors have for example written several papers on the topic, but there are also several others, some of which the authors cite in the introduction). I may be misunderstanding the claim the authors are making, in which case I would suggest rephrasing the second sentence. If not, then I would suggest the authors remove the claim.
- Similary, I find the last sentence of the abstract a gross overstatement of the novelty of this study. There are very many studies that have investigated the drivers of both individual and concurrent heatwaves, simply not with the methodology adopted here.
- The authors end the abstract by claiming the relevance of their work for predictability, but this is never supported by the results nor in the conclusions of the paper (beyond again making a similar statement without any supporting evidence). While I do think that the research presented here could indirectly support work on predictability, the connection is pretty far removed (is there any interest in forecasting heatwave clusters? I assume that most forecast users would want a forecast of heatwaves, not a forecast of a cluster, which is one step removed from the actual spatial temperature pattern that may occur).
- In the introduction, the authors essentially equate QRA with circumglobal Rossby waves. QRA is a (debated) mechanism proposed to explain the amplification of planetary waves with specific wavenumbers, but not all papers looking at circumglobal planetary waves necessarily assume that these result from quasi-resonance as stipulated by Petoukhov and colleagues. Indeed, researchers such as Nili Harnik (but there are certainly also others) have both worked on circumglobal Rossby waves and criticised the QRA hypothesis. I would thus recommend reframing the discussion of circumglobal Rossby waves and QRA.
- I have serious concerns on the methodology, and to some extent get the impression that the results are as much determined by the methodological choices as by the underlying data. Moreover, the current description of the methods is not sufficient to ensure reproducibility of the results.
- To define concurrent versus “isolated” heatwaves, the authors first implement a k-means clustering and then impose extent thresholds on the heatwave regions. A first point I struggle with is the term “region” (e.g. as used on ll. 129 and following). How is a “region” defined? Is it simply a set of connected heatwave gridpoints for a given cluster? Is any spatial clustering implemented, e.g. if there are two contiguous heatwave regions separated by a one gridpoint gap?
- Related to this, from the description on ll. 128-134 it is difficult to understand whether the areal extent is applied to the heatwave binary data (as has been commonly done in the literature), or to the clusters. If the latter, then this is one step removed from the areal extent of the actual heatwaves included in the cluster.
- The areal extent is imposed in terms of gridpoints on a regular lat-lon grid. However, the authors analyse a relatively broad latitudinal extent, and the size of a gridpoint varies significantly between 30N and 70N. I see no physical justification to impose extent in terms of gridpoints rather than area – as done in much of the heatwave literature.
- Always in Sect. 2.2, the authors state that: “We show the results for the most frequent clusters per region, except for the case of the NH concurrent, for which we show clusters with a prominent global signal of RWPs.” Is this not directly determining one of the key results of the study, namely that RWPs project strongly on concurrent heatwave occurrences? Presumably if clusters were instead selected so that they do not show a prominent global signal of RWPs (which would be an equally arbitrary and thus equally justifiable choice), then the conditional importance of RWP for triggering heatwaves would be much reduced.
- The authors perform clustering on all mid-latitude gridpoints and then impose the relatively weak constraint of selecting clusters with “the most frequent heat-wave occurrence regions over land”. As outlined by the authors themselves in the introduction, the interest in studying heatwaves is typically linked to impacts over land. The interest in studying concurrent heatwaves is often even more specifically associated with food production and agricultural yields, which by definition are restricted to land areas. Marine heatwaves are an active topic of study, but their dynamics and timescales are very different from the types of events studied here. Selecting clusters that have a non-negligible part of the heatwaves over ocean areas somewhat weakens the link to the paper’s motivation as outlined in the introduction.
- Related to this, the drivers of heatwaves (and their links and interplays with atmospheric drivers) can differ significantly between land and ocean regions, and clustering the two together may conflate these different processes and confound the link to the wave patterns that the authors are studying. The magnitudes of heatwaves are also likely very different over land and ocean regions, although this is masked by the choice of showing results exclusively in terms of standardised anomalies.
- Always concerning the methodology, the authors highlight that the literature has frequently linked amplified waves with k = 4-9 to heatwave occurrences. They then define the high global wave amplitude day by summing over k = 4-15. A first question is what is the rationale behind summing over a different range from the one that the authors themselves highlight as having been found in the literature to have the strongest link to heatwaves.
- A second question related to the above is whether there is the need to sum at all, as much of the literature singles out specific wavenumbers (as the authors also find in their own results in Sect. 3.2). See also my comment below on Sect. 3.3
- A further minor point concerning the methodology is that in reading Sect. 2.4 I got the impression that the authors were only considering planetary wave amplitude summed over a range of wavenumbers, since they write: “We calculate the amplitude of the daily global CGWP amplitude, by summing the amplitudes of wavenumbers k=4–15”. When they then mention how they compute the amplitude for specific wavenumbers, I interpreted these as being used to then yield the summed metric for all wavenumbers. However, the analysis in Sect. 3 partly focuses on single wavenumbers. Perhaps this could be clarified by updating the text to: “We calculate the amplitude of the daily global CGWP amplitude, by summing the amplitudes of wavenumbers k=4–15 for each hemisphere, but also consider amplitudes for each wavenumber within this range in isolation” or similar.
- A more general comment regards the choice of using a clustering algorithm. There are good examples from the literature of rule-based detection algorithms for concurrent heatwaves which could be used here. Clustering adds an intermediate step between the physical phenomenon being studied and the data actually being analysed, and is typically used when there is a need for dimensionality reduction. In this case however, one could identify concurrent and isolated heatwaves by using some existing rule-based algorithm from the literature (or by designing a new one) and then conduct essentially the same analysis as done in Sect. 3.2 and 3.3 but using actual heatwave occurrence at every gridbox rather than cluster occurrence.
- Sect. 3.1
- 1 I myself often use standardised anomalies, but for this specific case I would find it helpful to also see some numbers in K, since heatwaves are a phenomenon of which we have a good intuitive quantitative understanding. Taking for e.g. a typical location in N. America or W. Europe, how many K does a standardised anomaly of 0.3 correspond to on average?
- 227 and following. Are these % referred to the total number of heatwave gridpoints detected in the NH midlatitudes and all clusters (even the ones not shown in this section) or to the total number of heatwave gridpoints detected in the NH midlatitudes and only the clusters shows in this section? If the former, I would suggest also indicating the % for the clusters included in this section, so that the readers can get an idea of how large a fraction of all heatwaves you draw your conclusions from. It could also be interesting to indicate what % of all land heatwave gridpoints is included in the clusters you study, as those are presumably the heatwaves that we are most interested in.
- 269 and following. Same question as above.
- The authors barely comment on the central column in Figs. 2 and 3, which is one of the more interesting results here.
- In Fig. 3, some of the non-concurrent clusters display high-amplitude RWPs at multiple locations. This is an important point to comment on and discuss further, also in the perspective of the extent to which RWPs connect to the heatwave clusters as defined by the authors.
- 249-250 Is there a way to quantify this extent? Just by eyeballing Figs. 2 and 3, I would for example argue that the top-left panel in Fig. 3 (non-concurrent) has a wave pattern of similar or even larger extension than the top-left panel in Fig. 2 (concurrent).
- 4 and 8 are also barely commented on. If there is nothing more to say about these figures than the fact that areas of high RWB frequency display heatwaves (which is not such a surprising result and one that has been repeatedly noted in the literature), then do they really have a place in the main part of the paper?
- Sect. 3.2
- Is this analysis for all clusters or only for the same clusters as selected for analysis in Sect. 3.1? If the former, then this makes it difficult to compare the results for circumglobal wavepatterns versus RWP, as they are conducted on different sets of heatwaves.
- The comparison here is done on the distribution of dominant wavenumber on days with heatwaves versus climatology. Would it not be more logical to do a comparison of amplitude for the different wavenumbers during heatwave days versus climatology? To explain my reasoning: one could construct an artificial dataset where, on a given set of days, specific wavenumbers have an extremely high amplitude, but also ensure that this set of days has exactly the same distribution of dominant wavenumbers as climatology. In other words, the analysis here does not say whether the waves are actually amplified during heatwaves. It just says that the dominant wavenumbers do not change much relative to climatology.
- Why show a different wavenumber range in Fig. 9 from that highlighted on l. 167 as having been found relevant for heatwaves in the literature and also from the one that the authors themselves selected to defined the CGWP amplitude?
- Sect. 3.3
- I very much struggled to follow the logic of this section. In the first paragraph the authors highlight that both past work and their analysis in Sect. 3.2 highlights that the link between circumglobal Rossby waves and heatwaves only holds for specific wavenumbers. In the next paragraph, they comment on results obtained by summing wave amplitude over a very wide range of wavenumbers. There is clearly a disconnect between the results obtained in Sect. 3.2 and the analysis conducted in Sect. 3.3. Once again, I feel that the methodological and analysis choices dictate the results of the study.
- Conclusions
- 375-376 I don’t think that the analysis supports this claim, as in Fig. 9 you never show the amplitude for specific wavenumbers but only which one is dominant.
- 379-380 These results warrant a much more detailed contextualisation relative to the literature, as they are one of the key results of the study and counter the notion that has been put forth in previous studies that circumglobal wave patterns are important for triggering concurrent extremes,
Minor Comments
- 65 and following You may also want to cite the nice summary work of Teng and Branstator (2017) on the topic. Teng, H. and Branstator, G. (2017). Connections Between Heat Waves and Circumglobal Teleconnection Patterns in the Northern Hemisphere Summer. In Climate Extremes (eds S.-Y.S. Wang, J.-H. Yoon, C.C. Funk and R.R. Gillies). https://doi.org/10.1002/9781119068020.ch11
- 190-191 There are a number of studies that have linked RWB to surface windstorms, e.g. papers by Rodrigo Caballero and co-authors and by Iñigo Gómara and co-authors. Many of these studies have used different RWB detection methods from those listed by the authors, for example Elisabeth Barnes’ absolute vorticity-based method
- 310 and following. This text gives the impression that we can see the values for individual clusters in Fig. 9, but this is not the case. I would suggest rephrasing.
- 325 and following. Simply writing out the equation would help here. The textual explanation is quite convoluted and hard to follow.
- 331-332 Should “(non-concurrent)” be removed here?
- 10, 11 Since the cut-off here between enhanced/decreased frequency is 1, I would find it useful to have a clear colour distinction between values >1 and values <1. Making this distinction between two shades of grey makes interpretation of the figures difficult.
- I would recommend giving all panels in all figures a letter, to avoid having to reference panels using combinations of left-central-right and top-middle-bottom.
- 377-378 “we find a statistically significant increase of the zonal wavenumber k=7…” Most readers would interpret this as if you find an increased amplitude in this wavenumber. However, what you show is that there is an increased conditional frequency of days when k=7 is the dominant wavenumber. These two things are not the same. I would suggest rephrasing to clarify this.
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC2 - The abstract provides a misleading context for the study and exaggerates its novelty.
-
RC3: 'Comment on egusphere-2023-3088', Anonymous Referee #3, 08 Mar 2024
This manuscript investigates the occurrence of both concurrent and non-concurrent heatwaves and how they are associated with the occurrence of either circum-global Rossby wavetrains or more regional Rossby wave packets. The topic is relevant, the paper is timely, and the analysis has the potential to further the discussion. Also, I think that WCD would be an appropriate journal.
At the same time, I feel that there is a substantial lack of logic and clarity such that it was very hard for me to find out what was really done, what is new, and what the greater implications are.
Let me give a few examples.
(1) One of the major results (as far as I can see) is the fact that the association between heatwaves and synchronous Rossby waves is much stronger for regional Rossby wave packets than for circum-global Rossby waves. This insight has been obtained earlier in a paper by one of the coauthors, namely Fragkoulidis et al. (2018). To be sure, the authors go beyond the work of Fragkoulidis et al. (2018) in that they distinguish between concurrent and non-concurrent heatwaves. However, in this respect the results are much less robust: the differences between concurrent and non-concurrent heatwaves are quite small, and it did not become clear to me to what extent the results depend on the exact definition of “concurrent” and “non-concurrent”. In other words, it remained unclear to me (i) what is new in the obtained insight, and (ii) how robust the results are regarding the distinction between “concurrent” and “non-concurrent”.
(2) The description of Rossby wave resonance around line 60 is fairly obscure and does certainly not capture the essence of this phenomenon. This is too bad, as resonance is a well-known topic in dynamical meteorology and has been described very lucidly early on, e.g., by Haurwitz (1940, J. Mar. Res., 3, 254–267). However, what’s more, the topic of resonance (for what I could find) is irrelevant to anything that comes later in this paper, so it did not become clear to me why it is described here at all.
(3) I think that the results are interpreted in an inflated manner. For instance, the sentence on line 65/66 suggests that the circum-global wave causes concurrent heatwaves. It could, however, be just as well the other way around: Concurrent heatwaves, that have been caused by whatever mechanism or just by chance, may be responsible for (may “cause”) a certain wavenumber in the Fourier decomposition to dominate. Remember: correlation does not imply causation. Along similar lines: The term “driving” (in the title or on line 392) implies a causal relationship. However, as far as I can see, in your work you only analyzed correlations and conditional probabilities (the latter of which can be considered as generalized correlations). Again, correlation does not imply causation. Therefore, I fear that your claim in the second to last sentence of the summary is not supported by the results from your study. To be sure, it may well be that Rossby wave packets “drive” heatwaves in specific cases, but that is not what you have shown in this study.
(4) There is a lot of detail in the paper which makes it hard for the reader to see what is important, what is not, and what eventually should be learned from the study. For example, there is the connection with Rossby wave breaking, which is mentioned rather loosely in the results section and where it has not become clear to me how it contributes to the main point of the paper. We have known for decades that a heatwave may be associated with blocking and blocking may be associated with wave breaking, but how does this manuscript contribute new insight in this respect? Along similar lines: a conclusion section is most useful if it summarizes the results on a higher level of abstraction and tells the reader what can be learned from the study. In my eyes, the current conclusion section does not satisfy this expectation. Rather, the authors present a condensed and quite detailed version of outcomes from the specific analysis, and the reader is left to wonder what one should take away from it.
All in all, I think that this manuscript is not suitable for publication in WCD at this point. I encourage the authors to reconsider this work and eventually have it published. But I fear that this would require quite some time and effort, and the final result would have to be very different from the current version. For this reason, I cannot recommend publication of this manuscript in WCD.
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC3 -
EC1: 'Editor recommendation on egusphere-2023-3088', David Battisti, 24 Mar 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2023-3088/egusphere-2023-3088-EC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-3088', Maria Pyrina, 13 Apr 2024
We would like to thank very much the editor and reviewers for the time invested to review our manuscript. We understand the concerns raised and, therefore, we have decided to withdraw this manuscript. However, we would like to answer to some of these major comments:
Cluster analysis has been previously used to cluster atmospheric weather regimes and is used in this manuscript in a similar manner to cluster “Heatwave regimes”. We provide sensitivity tests on the number of clusters concidered, latitudes selected, and number of consecutive heatwave days that go into the composite mean, but we understand that more tests are needed to create clusters that separate the concurrent and no concurrent heatwaves even more. Regarding the choice of wavenumbers and index used to define CGRW we have followed exactly the literature supporting the major assosiation of concurent heatwaves and CGRW. Finally, we have not manually selected the threshold value (> 0.3 sigma) to define a heatwave region. Recall that we have clustered heatwaves according to occurrence and not according to intensity. This +0.3 region is just the standartised temperature value of the temperature composite mean which we saw that corresponds to regions where the cluster cetroid shows maximum heatwave occurrence. We had not plotted the cluster centroids for not including more plots into the manuscript, as the maximum occurrence in the centroids corresponds to the places with maximum temperature shown by the composite mean (being above +0.3), as stated in the caption of Figure 1 caption. Finally, we do not conclude that CGRWs are irrelevant for concurrent heatwaves as we also show that there is an association, which is however lower compared to daily RWPs, as we here investigate daily time scales and not monthly means.
Thank you again very much for the time and helpful comments.
Citation: https://doi.org/10.5194/egusphere-2023-3088-AC1
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-3088', Anonymous Referee #1, 05 Mar 2024
General Comments
This paper investigates the role of Rossby waves on heatwaves, and contrasts the impacts of circumglobal Rossby waves with that of more localized Rossby wave packets, finding that locally amplified Rossby wave packets have a much stronger association with heatwaves than amplified circumglobal Rossby waves. I think this is an important and interesting finding. However, it seemed to me that this interesting result was somewhat hidden under details of other analysis in the paper, particularly that focussed on distinguishing between concurrent vs non-concurrent heatwaves, which I remain a little confused by. I would recommend to the authors that they consider re-structuring their manuscript to better highlight their key result. I think the comparison of concurrent vs non-concurrent heatwaves requires more thought, and perhaps a stronger separation/distinction between these two types of heatwave, and exploration of the sensitivity of results to the methodology.
Specific Comments
There is a lot of detail on the separation of the heatwaves into concurrent and non-concurrent heatwaves early on in the paper, however, it seems like mostly your results are not particularly sensitive to whether you are looking at concurrent or non-concurrent heatwaves (except perhaps the wavenumbers for the high-amplitude waves in Fig. 9), and so it seems perhaps unnecessary for readers to get through all of the information about how the concurrent vs non-current separation is performed before finding out it doesn’t really matter? I therefore recommend changing the order of your manuscript, and showing the top rows of Figs. 10 and 11 much earlier, highlighting this result, that heatwaves are associated with amplified local Rossby wave packets much more strongly than they are associated with amplified circumglobal Rossby waves. I think it is then very reasonable to hypothesise that this might be different for concurrent heatwaves, and to then explore that.
I am not convinced, based on the plots presented, that the methodology for distinguishing concurrent and non-concurrent heatwaves is an appropriate methodology. For example, in Fig. 1, there is much less difference between some of the concurrent clusters and some of the non-concurrent clusters (e.g. EU concurrent vs non-concurrent) than I was expecting given the definition. The concurrent clusters still appear to be a dominant heatwave in one location. Denoting a region that’s above +0.3 standard deviations in temperature as a heatwave seems surprising. Are the weak signals of the secondary heatwaves in this plot because it is a composite(/cluster) mean, and the cluster is really defined as a strong heatwave over one region (e.g. Europe) with some heatwaves elsewhere, but the region of the secondary heatwave varies between cluster members? I think more justification of this methodology is required here, with figures that clearly show the distinctions, or discussion around the fact that the distinctions are relatively subtle. Given that your results are mostly ‘there isn’t much difference between concurrent and non-concurrent heatwaves’ (at least that’s my understanding), I think you need to show a stronger difference between your clusters for concurrent and non-concurrent heatwave – I can’t tell whether that difference is there, and it’s just that the cluster mean is not a good way of showing it, or that there isn’t enough difference? Perhaps adjusting your method to exclude events that are close to the middle between concurrent and non-concurrent could help, I.e. defining non-concurrent events as when there is one region with >320 HW grid-points, and no other HW regions exceed 100 grid-points, rather than no other regions exceeding 320 grid-points. I realize this would reduce the number of events, but it would give you a clearer distinction, which would make your results that the distinction doesn’t generally matter seem more convincing.
I think that the Rossby wave breaking analysis is interesting, but it’s used and discussed relatively little for the amount of detail you go into in the methods, and it receives little attention in the results. Overall in section 2.5 for me it feels like not quite enough information for me to fee like I understand the method, and the information that is there is to complex to understand without additional complex. I'm unsure what's the same as previous methods, and what is new in this work. I think either explain the method more clearly, or, if the method isn't too relevant to your results, and is already explained in previous papers, just provide a clearer overview of the method with fewer details here. E.g. Line 203: I assume the elongated structures are identified in the PVU contour? Line 205: how are you defining length vs width? Line 216. It’s this depth of tropospheric PV structures that you use as your metric of RWB (i.e. in Fig 4), is that right? This needs to be stated here.
Separation into high amplitude CGRW days: the separation to only high amplitude days is not introduced clearly, and there is no explanation or justification for why this is done – was it because there were no statistically significant results for all wave amplitude days? I think it’s a reasonable approach, but it requires some explanation and introduction. I’m also interested in how much it reduces the number of heatwave days by? How much noisier does the dataset get when you restrict it to wave amplitude days > +1.5sigma? This is particularly relevant as the only statistically different results you see from climatology is when you restrict to high amplitude days. It is also unclear what the climatological probability densities are for only the high amplitude days – are the CCHW_HA statistically different from climatology_HA? I think there might be some interesting results here, but they aren’t very clear in the current format.
Other minor comments
Line 72. I think the research gap/question you’re asking could be defined more clearly here.
Clustering algorithm: how many days have no heatwaves? Is there a ‘no heatwave’ cluster or are these days excluded from the analysis?
Line 124. how did you decide on this optimal cluster number definition, and are your results sensitive to this definition?
Line 134. is there a minimum number of gridpoint separation of the regions for two heatwaves?
Line 174. “More specifically…” this is a rather confusing sentence structure, I recommend rephrasing.
Line 178: where A(t) is the amplitude…. (add is)
Section 2.4. Why give the equations for a Fourier transform, but not for the calculation of the RWP using the fourier transform? I don’t think the equations for a Fourier transform are necessary in the main paper, and I feel like the RWP equation is the more interesting one.
Lines 188-195: might be better off in the introduction, particularly when you discuss methodologies that you don’t use in the current paper.
Line 226: How sensitive are these results to your heatwave definition?
Line 230: As mentioned above, +0.3 sigma seems quite low for a heatwave definition, so the use of this threshold needs more justification here
Line 262. “The clusters that demonstrate persistent to….” this sentence is too difficult to interpret - is this showing the clusters that are more dominated by persistent heatwaves rather than non-persistent heatwaves? Is this saying that concurrent heatwaves tend to be more persistent?? Or can you not quite say that? I'm just unsure what physical interpretation to make here.
Line 275. “Finally…” is this statement related to persistence? If so, please make the connection more directly; if not, I think the sentence is out of place here.
I don’t think Fig. 7 ever gets referenced?
Line 300: k=5 and k=6 are also the most prevalent for the climatology, i.e. non-heatwave days? Also, k=4 is similarly important to k=6 in the SH).
Fig. 9. It’s hard to see some of the NCNW_HA dots when they lie behind the NCHW ones – maybe use a different shape?
Line 319: It doesn’t feel like this connection between heatwaves and high amplitude wave events was explored in the previous section particularly – I agree that it would be interesting, and suggest re-focussing section 3.2 to make this more clear (as stated in Specific Comments)
Line 321. And k = 5 in the NH? Even if it’s not statistically significant, it might be a signal?
Line 324. By ‘the two metrics’ do you mean the global fourier transform amplitude and the local RWP amplitude? Or do you include RWB here? I think making this distinction between the two metrics clearer throughout the paper would be useful.
Lines 325-335. This is a confusing paragraph, and might be better placed in the methods, with the equation to help understand this. Since this leads to what is, based on the title of the paper and my sense of your results, your most important result, it seems more important to be clear about this methodology than, for example, giving the equations for a fourier transform.
Line 329. I don't quite understand this. So, for a particular grid-point, the concurrent heatwave days probability is the fraction of all days within concurrent heatwave clusters during which this gridbox is part of a heatwave? And then the global amplitude for concurrent heatwaves is for any concurrent heatwave, regardless of where the heatwave is located? Please rephrase to make this clearer (see previous comment)
Line 335: how do you objectively define "a ridge over that location"?
Section 3.3. The CGWP days seem to have a very similar spatial pattern as for the RWP analysis (particularly noticeable in the SH) - do you think this is some feature of heatwaves, or is simply because local waves can show up in a global Fourier transform?
Line 369: the differences were surprisingly small to me though - comparing the middle columns of Figs 2 and 3? I agree that the difference seems to not be a global wave vs a local wave packet, and more is one vs multiple RWPs, but even this isn’t clear, as multiple RWPs are visible in the non-current heatwaves too, in Fig. 3. How do you explain this? My suspicion is that it’s related to my earlier comment, that the composites of EU concurrent and EU non-concurrent are not particularly well-separated.
Line 383. There isn't a lot of difference between concurrent and non-concurrent in Fig. 9, unless you look at high-amplitude events and k=7, which is very specific. Maybe this is sensitive to the methodology of separating concurrent and non-concurrent heatwaves, which should be determined and, if not, this might be one of the key messages to this paper - rather than highlighting the small differences between them, highlight how similar they are from the atmosphere’s perspective, with some small differences?
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC1 -
RC2: 'Comment on egusphere-2023-3088', Anonymous Referee #2, 07 Mar 2024
The topic of the study is of great interest. Spatially remote but concurrent extremes are gaining increasing attention in the literature, and there has been some debate as to the large-scale atmospheric drivers of such extremes. In particular, there is an active line of research supporting the notion that circumglobal planetary wave patterns are crucial for determining the co-occurrence of remote extremes, and testing this hypothesis robustly against other large-scale triggering mechanisms is a very relevant research undertaking. Challenging previous results is essential for advancing research, yet when one draws conclusions which partially counter an established body of previous work it is important to provide robust evidence to support them. My impression after reading this study is that the methodological choices steer to a great extent the results, and that the contextualisation and discussion of such results is lacking. As such, I cannot recommend this study for publication and believe that the edits necessary to make this a valuable contribution to the ongoing discussion of the large-scale triggers of single and concurrent heatwaves go beyond the scope of a major revision.
Major Comments
- The abstract provides a misleading context for the study and exaggerates its novelty.
- The authors state that: “Amplified quasi-stationary circumglobal Rossby wave patterns (CGWPs) and high-amplitude transient non-circumglobal Rossby Wave Packets (RWPs) have been proposed as two possible explanations for the occurrence of heatwaves. The relation of these mechanisms for heatwaves has been investigated over different timescales, but their relevance for concurrent and non-concurrent heatwaves remains unexplored”. I agree that the role of RWP for concurrent events is an understudied topic; on the other hand there is ample work on circumglobal Rossby waves and concurrent heatwaves (Kai Kornhuber and co-authors have for example written several papers on the topic, but there are also several others, some of which the authors cite in the introduction). I may be misunderstanding the claim the authors are making, in which case I would suggest rephrasing the second sentence. If not, then I would suggest the authors remove the claim.
- Similary, I find the last sentence of the abstract a gross overstatement of the novelty of this study. There are very many studies that have investigated the drivers of both individual and concurrent heatwaves, simply not with the methodology adopted here.
- The authors end the abstract by claiming the relevance of their work for predictability, but this is never supported by the results nor in the conclusions of the paper (beyond again making a similar statement without any supporting evidence). While I do think that the research presented here could indirectly support work on predictability, the connection is pretty far removed (is there any interest in forecasting heatwave clusters? I assume that most forecast users would want a forecast of heatwaves, not a forecast of a cluster, which is one step removed from the actual spatial temperature pattern that may occur).
- In the introduction, the authors essentially equate QRA with circumglobal Rossby waves. QRA is a (debated) mechanism proposed to explain the amplification of planetary waves with specific wavenumbers, but not all papers looking at circumglobal planetary waves necessarily assume that these result from quasi-resonance as stipulated by Petoukhov and colleagues. Indeed, researchers such as Nili Harnik (but there are certainly also others) have both worked on circumglobal Rossby waves and criticised the QRA hypothesis. I would thus recommend reframing the discussion of circumglobal Rossby waves and QRA.
- I have serious concerns on the methodology, and to some extent get the impression that the results are as much determined by the methodological choices as by the underlying data. Moreover, the current description of the methods is not sufficient to ensure reproducibility of the results.
- To define concurrent versus “isolated” heatwaves, the authors first implement a k-means clustering and then impose extent thresholds on the heatwave regions. A first point I struggle with is the term “region” (e.g. as used on ll. 129 and following). How is a “region” defined? Is it simply a set of connected heatwave gridpoints for a given cluster? Is any spatial clustering implemented, e.g. if there are two contiguous heatwave regions separated by a one gridpoint gap?
- Related to this, from the description on ll. 128-134 it is difficult to understand whether the areal extent is applied to the heatwave binary data (as has been commonly done in the literature), or to the clusters. If the latter, then this is one step removed from the areal extent of the actual heatwaves included in the cluster.
- The areal extent is imposed in terms of gridpoints on a regular lat-lon grid. However, the authors analyse a relatively broad latitudinal extent, and the size of a gridpoint varies significantly between 30N and 70N. I see no physical justification to impose extent in terms of gridpoints rather than area – as done in much of the heatwave literature.
- Always in Sect. 2.2, the authors state that: “We show the results for the most frequent clusters per region, except for the case of the NH concurrent, for which we show clusters with a prominent global signal of RWPs.” Is this not directly determining one of the key results of the study, namely that RWPs project strongly on concurrent heatwave occurrences? Presumably if clusters were instead selected so that they do not show a prominent global signal of RWPs (which would be an equally arbitrary and thus equally justifiable choice), then the conditional importance of RWP for triggering heatwaves would be much reduced.
- The authors perform clustering on all mid-latitude gridpoints and then impose the relatively weak constraint of selecting clusters with “the most frequent heat-wave occurrence regions over land”. As outlined by the authors themselves in the introduction, the interest in studying heatwaves is typically linked to impacts over land. The interest in studying concurrent heatwaves is often even more specifically associated with food production and agricultural yields, which by definition are restricted to land areas. Marine heatwaves are an active topic of study, but their dynamics and timescales are very different from the types of events studied here. Selecting clusters that have a non-negligible part of the heatwaves over ocean areas somewhat weakens the link to the paper’s motivation as outlined in the introduction.
- Related to this, the drivers of heatwaves (and their links and interplays with atmospheric drivers) can differ significantly between land and ocean regions, and clustering the two together may conflate these different processes and confound the link to the wave patterns that the authors are studying. The magnitudes of heatwaves are also likely very different over land and ocean regions, although this is masked by the choice of showing results exclusively in terms of standardised anomalies.
- Always concerning the methodology, the authors highlight that the literature has frequently linked amplified waves with k = 4-9 to heatwave occurrences. They then define the high global wave amplitude day by summing over k = 4-15. A first question is what is the rationale behind summing over a different range from the one that the authors themselves highlight as having been found in the literature to have the strongest link to heatwaves.
- A second question related to the above is whether there is the need to sum at all, as much of the literature singles out specific wavenumbers (as the authors also find in their own results in Sect. 3.2). See also my comment below on Sect. 3.3
- A further minor point concerning the methodology is that in reading Sect. 2.4 I got the impression that the authors were only considering planetary wave amplitude summed over a range of wavenumbers, since they write: “We calculate the amplitude of the daily global CGWP amplitude, by summing the amplitudes of wavenumbers k=4–15”. When they then mention how they compute the amplitude for specific wavenumbers, I interpreted these as being used to then yield the summed metric for all wavenumbers. However, the analysis in Sect. 3 partly focuses on single wavenumbers. Perhaps this could be clarified by updating the text to: “We calculate the amplitude of the daily global CGWP amplitude, by summing the amplitudes of wavenumbers k=4–15 for each hemisphere, but also consider amplitudes for each wavenumber within this range in isolation” or similar.
- A more general comment regards the choice of using a clustering algorithm. There are good examples from the literature of rule-based detection algorithms for concurrent heatwaves which could be used here. Clustering adds an intermediate step between the physical phenomenon being studied and the data actually being analysed, and is typically used when there is a need for dimensionality reduction. In this case however, one could identify concurrent and isolated heatwaves by using some existing rule-based algorithm from the literature (or by designing a new one) and then conduct essentially the same analysis as done in Sect. 3.2 and 3.3 but using actual heatwave occurrence at every gridbox rather than cluster occurrence.
- Sect. 3.1
- 1 I myself often use standardised anomalies, but for this specific case I would find it helpful to also see some numbers in K, since heatwaves are a phenomenon of which we have a good intuitive quantitative understanding. Taking for e.g. a typical location in N. America or W. Europe, how many K does a standardised anomaly of 0.3 correspond to on average?
- 227 and following. Are these % referred to the total number of heatwave gridpoints detected in the NH midlatitudes and all clusters (even the ones not shown in this section) or to the total number of heatwave gridpoints detected in the NH midlatitudes and only the clusters shows in this section? If the former, I would suggest also indicating the % for the clusters included in this section, so that the readers can get an idea of how large a fraction of all heatwaves you draw your conclusions from. It could also be interesting to indicate what % of all land heatwave gridpoints is included in the clusters you study, as those are presumably the heatwaves that we are most interested in.
- 269 and following. Same question as above.
- The authors barely comment on the central column in Figs. 2 and 3, which is one of the more interesting results here.
- In Fig. 3, some of the non-concurrent clusters display high-amplitude RWPs at multiple locations. This is an important point to comment on and discuss further, also in the perspective of the extent to which RWPs connect to the heatwave clusters as defined by the authors.
- 249-250 Is there a way to quantify this extent? Just by eyeballing Figs. 2 and 3, I would for example argue that the top-left panel in Fig. 3 (non-concurrent) has a wave pattern of similar or even larger extension than the top-left panel in Fig. 2 (concurrent).
- 4 and 8 are also barely commented on. If there is nothing more to say about these figures than the fact that areas of high RWB frequency display heatwaves (which is not such a surprising result and one that has been repeatedly noted in the literature), then do they really have a place in the main part of the paper?
- Sect. 3.2
- Is this analysis for all clusters or only for the same clusters as selected for analysis in Sect. 3.1? If the former, then this makes it difficult to compare the results for circumglobal wavepatterns versus RWP, as they are conducted on different sets of heatwaves.
- The comparison here is done on the distribution of dominant wavenumber on days with heatwaves versus climatology. Would it not be more logical to do a comparison of amplitude for the different wavenumbers during heatwave days versus climatology? To explain my reasoning: one could construct an artificial dataset where, on a given set of days, specific wavenumbers have an extremely high amplitude, but also ensure that this set of days has exactly the same distribution of dominant wavenumbers as climatology. In other words, the analysis here does not say whether the waves are actually amplified during heatwaves. It just says that the dominant wavenumbers do not change much relative to climatology.
- Why show a different wavenumber range in Fig. 9 from that highlighted on l. 167 as having been found relevant for heatwaves in the literature and also from the one that the authors themselves selected to defined the CGWP amplitude?
- Sect. 3.3
- I very much struggled to follow the logic of this section. In the first paragraph the authors highlight that both past work and their analysis in Sect. 3.2 highlights that the link between circumglobal Rossby waves and heatwaves only holds for specific wavenumbers. In the next paragraph, they comment on results obtained by summing wave amplitude over a very wide range of wavenumbers. There is clearly a disconnect between the results obtained in Sect. 3.2 and the analysis conducted in Sect. 3.3. Once again, I feel that the methodological and analysis choices dictate the results of the study.
- Conclusions
- 375-376 I don’t think that the analysis supports this claim, as in Fig. 9 you never show the amplitude for specific wavenumbers but only which one is dominant.
- 379-380 These results warrant a much more detailed contextualisation relative to the literature, as they are one of the key results of the study and counter the notion that has been put forth in previous studies that circumglobal wave patterns are important for triggering concurrent extremes,
Minor Comments
- 65 and following You may also want to cite the nice summary work of Teng and Branstator (2017) on the topic. Teng, H. and Branstator, G. (2017). Connections Between Heat Waves and Circumglobal Teleconnection Patterns in the Northern Hemisphere Summer. In Climate Extremes (eds S.-Y.S. Wang, J.-H. Yoon, C.C. Funk and R.R. Gillies). https://doi.org/10.1002/9781119068020.ch11
- 190-191 There are a number of studies that have linked RWB to surface windstorms, e.g. papers by Rodrigo Caballero and co-authors and by Iñigo Gómara and co-authors. Many of these studies have used different RWB detection methods from those listed by the authors, for example Elisabeth Barnes’ absolute vorticity-based method
- 310 and following. This text gives the impression that we can see the values for individual clusters in Fig. 9, but this is not the case. I would suggest rephrasing.
- 325 and following. Simply writing out the equation would help here. The textual explanation is quite convoluted and hard to follow.
- 331-332 Should “(non-concurrent)” be removed here?
- 10, 11 Since the cut-off here between enhanced/decreased frequency is 1, I would find it useful to have a clear colour distinction between values >1 and values <1. Making this distinction between two shades of grey makes interpretation of the figures difficult.
- I would recommend giving all panels in all figures a letter, to avoid having to reference panels using combinations of left-central-right and top-middle-bottom.
- 377-378 “we find a statistically significant increase of the zonal wavenumber k=7…” Most readers would interpret this as if you find an increased amplitude in this wavenumber. However, what you show is that there is an increased conditional frequency of days when k=7 is the dominant wavenumber. These two things are not the same. I would suggest rephrasing to clarify this.
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC2 - The abstract provides a misleading context for the study and exaggerates its novelty.
-
RC3: 'Comment on egusphere-2023-3088', Anonymous Referee #3, 08 Mar 2024
This manuscript investigates the occurrence of both concurrent and non-concurrent heatwaves and how they are associated with the occurrence of either circum-global Rossby wavetrains or more regional Rossby wave packets. The topic is relevant, the paper is timely, and the analysis has the potential to further the discussion. Also, I think that WCD would be an appropriate journal.
At the same time, I feel that there is a substantial lack of logic and clarity such that it was very hard for me to find out what was really done, what is new, and what the greater implications are.
Let me give a few examples.
(1) One of the major results (as far as I can see) is the fact that the association between heatwaves and synchronous Rossby waves is much stronger for regional Rossby wave packets than for circum-global Rossby waves. This insight has been obtained earlier in a paper by one of the coauthors, namely Fragkoulidis et al. (2018). To be sure, the authors go beyond the work of Fragkoulidis et al. (2018) in that they distinguish between concurrent and non-concurrent heatwaves. However, in this respect the results are much less robust: the differences between concurrent and non-concurrent heatwaves are quite small, and it did not become clear to me to what extent the results depend on the exact definition of “concurrent” and “non-concurrent”. In other words, it remained unclear to me (i) what is new in the obtained insight, and (ii) how robust the results are regarding the distinction between “concurrent” and “non-concurrent”.
(2) The description of Rossby wave resonance around line 60 is fairly obscure and does certainly not capture the essence of this phenomenon. This is too bad, as resonance is a well-known topic in dynamical meteorology and has been described very lucidly early on, e.g., by Haurwitz (1940, J. Mar. Res., 3, 254–267). However, what’s more, the topic of resonance (for what I could find) is irrelevant to anything that comes later in this paper, so it did not become clear to me why it is described here at all.
(3) I think that the results are interpreted in an inflated manner. For instance, the sentence on line 65/66 suggests that the circum-global wave causes concurrent heatwaves. It could, however, be just as well the other way around: Concurrent heatwaves, that have been caused by whatever mechanism or just by chance, may be responsible for (may “cause”) a certain wavenumber in the Fourier decomposition to dominate. Remember: correlation does not imply causation. Along similar lines: The term “driving” (in the title or on line 392) implies a causal relationship. However, as far as I can see, in your work you only analyzed correlations and conditional probabilities (the latter of which can be considered as generalized correlations). Again, correlation does not imply causation. Therefore, I fear that your claim in the second to last sentence of the summary is not supported by the results from your study. To be sure, it may well be that Rossby wave packets “drive” heatwaves in specific cases, but that is not what you have shown in this study.
(4) There is a lot of detail in the paper which makes it hard for the reader to see what is important, what is not, and what eventually should be learned from the study. For example, there is the connection with Rossby wave breaking, which is mentioned rather loosely in the results section and where it has not become clear to me how it contributes to the main point of the paper. We have known for decades that a heatwave may be associated with blocking and blocking may be associated with wave breaking, but how does this manuscript contribute new insight in this respect? Along similar lines: a conclusion section is most useful if it summarizes the results on a higher level of abstraction and tells the reader what can be learned from the study. In my eyes, the current conclusion section does not satisfy this expectation. Rather, the authors present a condensed and quite detailed version of outcomes from the specific analysis, and the reader is left to wonder what one should take away from it.
All in all, I think that this manuscript is not suitable for publication in WCD at this point. I encourage the authors to reconsider this work and eventually have it published. But I fear that this would require quite some time and effort, and the final result would have to be very different from the current version. For this reason, I cannot recommend publication of this manuscript in WCD.
Citation: https://doi.org/10.5194/egusphere-2023-3088-RC3 -
EC1: 'Editor recommendation on egusphere-2023-3088', David Battisti, 24 Mar 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2023-3088/egusphere-2023-3088-EC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-3088', Maria Pyrina, 13 Apr 2024
We would like to thank very much the editor and reviewers for the time invested to review our manuscript. We understand the concerns raised and, therefore, we have decided to withdraw this manuscript. However, we would like to answer to some of these major comments:
Cluster analysis has been previously used to cluster atmospheric weather regimes and is used in this manuscript in a similar manner to cluster “Heatwave regimes”. We provide sensitivity tests on the number of clusters concidered, latitudes selected, and number of consecutive heatwave days that go into the composite mean, but we understand that more tests are needed to create clusters that separate the concurrent and no concurrent heatwaves even more. Regarding the choice of wavenumbers and index used to define CGRW we have followed exactly the literature supporting the major assosiation of concurent heatwaves and CGRW. Finally, we have not manually selected the threshold value (> 0.3 sigma) to define a heatwave region. Recall that we have clustered heatwaves according to occurrence and not according to intensity. This +0.3 region is just the standartised temperature value of the temperature composite mean which we saw that corresponds to regions where the cluster cetroid shows maximum heatwave occurrence. We had not plotted the cluster centroids for not including more plots into the manuscript, as the maximum occurrence in the centroids corresponds to the places with maximum temperature shown by the composite mean (being above +0.3), as stated in the caption of Figure 1 caption. Finally, we do not conclude that CGRWs are irrelevant for concurrent heatwaves as we also show that there is an association, which is however lower compared to daily RWPs, as we here investigate daily time scales and not monthly means.
Thank you again very much for the time and helpful comments.
Citation: https://doi.org/10.5194/egusphere-2023-3088-AC1
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
530 | 177 | 50 | 757 | 23 | 18 |
- HTML: 530
- PDF: 177
- XML: 50
- Total: 757
- BibTeX: 23
- EndNote: 18
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1