the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Trends and seasonal signals in Atlantic feature-based jet stream characteristics and in weather types
Abstract. Recent studies have highlighted the link between upper-level jet dynamics, especially the persistence of certain jet configurations, and extreme summer weather in Europe. The weaker and more variable nature of the jets in summer makes it difficult to apply the tools developed to study them in winter, at least not without modifications. Here, to further investigate the link between jets and persistent summer weather, we present two complementary approaches to characterize the jet dynamics in the North Atlantic sector and use them primarily on the summer circulation.
First, we apply the self-organizing map (SOM) clustering algorithm to create a 2D distance-preserving discrete feature space to the tropopause-level wind field over the North Atlantic. The dynamics of the tropopause-level wind can then be described by the time series of visited SOM clusters, in which a long stay in a given cluster relates to a persistent state and a rapid transition between clusters that are far apart relates to a sudden considerable shift in the configuration of upper-level flow.
Second, we adapt and apply a jet axis detection and tracking algorithm to extract individual jets and classify them in the canonical categories of eddy-driven and subtropical jets (EDJ and STJ, respectively). Then, we compute a wide range of jet indices on each jet to provide easily interpretable scalar time series representing upper-tropospheric dynamics.
This work will exclusively focus on the characterization of historical trends, seasonal cycles, and other statistical properties of the jet stream dynamics, while ongoing and future work will use the tools presented here and apply them to the study of connections between jet dynamics and extreme weather. The SOM allows the identification of specific jet configurations, each one representative of a large number of days in historical time series, whose frequency or persistence had increased or decreased in the last decades. Detecting and categorizing jets adds a layer of interpretability and precision to previously and newly defined jet properties, allowing for a finer characterization of their trends and seasonal signals.
Detecting jets on flattened pressure levels instead the 2PVU surface is more robust in summer, and finding wind-direction-aligned subsets of 0-contours in a wind shear field is a fast and robust way to extract jet features. Using the SOM, we highlight a trend towards more negative NAO, and isolate predictable and/or persistent circulation patterns. Using properties of the jet features, we confirm that jets get faster and narrower in winter, but not so clearly in summer, and find no significant trend in jet latitude. Finally, both methods agree on a sudden flow transition in June.
- Preprint
(8895 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 19 Nov 2024)
-
RC1: 'Comment on egusphere-2024-3029', Anonymous Referee #1, 07 Nov 2024
reply
General Comments
           This study presents two diagnostic tools applicable to jets over the North Atlantic: (1) a SOM trained on upper-level flow patterns from ERA5 output across several decades, and (2), a jet identification scheme that locates tracks individual jet axes. Each jet axis is categorized as either an instance of the polar jet/eddy-driven jet (EDJ) or the subtropical jet (STJ). The authors have developed an impressive set of methods to generate these tools, and provide many well-motivated avenues for the use of these tools in future research. The authors are also very attentive to recognizing some of the limitations of their methodology throughout the manuscript. Overall, I find this paper to be scientifically significant and I think many of the applied methods and conclusions are very valid. However, I am concerned about the validity of the methods used to categorize each jet feature as either an EDJ or STJ. There are also several moments where the writing in the paper is not as clear as I think it could be (I have pointed out these moments in my line-by-line comments).
Â
I suspect that the approach used to distinguish the EDJ and STJ using jet frequencies in mean latitude–mean longitude–pressure level space may not be as physical and therefore may not be as robust as an approach that examines how the jets are distributed in wind speed–potential temperatures space. For example, the authors show a histogram of jets in mean latitude and mean longitude in panel (a) of Figure 4. Although there are two peaks in the distribution, I wonder if the interpretation that one peak belongs to the EDJ and one peak belongs to the STJ is really appropriate. How do we know that both types of jets do not contribute to each of these peaks (i.e., these peaks are just peaks in jet occurrence in general, not peaks mostly in one type of jet or the other)? Additionally, there appears to be no bimodal distribution in the histogram shown in panel (b) of Figure 4. How is it possible to draw a line separating the EDJ and STJ from each other?
Â
Furthermore, there is often only one jet present during the summer (which is clearly reflected in the double jet index time series from Figure 10). The unimodal maximum in frequency observed along the pressure level dimension (panel (b) in Figure 4) suggests that maybe this single jet is best described as a combination of both the EDJ and STJ. Alternatively, I think one could argue that such a jet could be better represented by the physical processes supporting the STJ as opposed to the EDJ, since the Hadley Cell remains during the summer, but baroclinicity over the mid-latitudes is drastically reduced.
Â
To address these concerns, I suggest examining seasonal histograms of the jet frequencies for bins of potential temperature in the vertical direction to confirm the validity of the current delineation between the EDJs and STJs. I suspect that the distribution may still, however, be unimodal. In that case, I think a sensitivity test is warranted to determine how the choice of a cutoff in potential temperature or pressure level between the categories along the vertical dimension impacts the frequencies of each category in total and geographically. If the categories are very sensitive to this choice, then maybe a third category of jet is required to describe the summer jet, especially when the double jet index is relatively low.
Â
Â
Â
Â
Specific Comments
Â
Line 49: I am not quite sure what the authors mean by “double jet state.” Maybe this is referring to the simultaneous presence of both the EDJ and STJ? If so, then maybe the phrase “a persistent double jet state” could be replaced with something a little clearer, along the lines of “the persistent simultaneous presence of the EDJ and STJ”.
Â
Line 72: I think this review of past research in the introduction, especially on projected changes in the EDJ and STJ in warmer climates, is extremely thorough and thoughtful.
Â
Line 110: I think it would be useful to mention what climatology these geopotential height anomalies are computed relative to. Maybe the authors used a daily climatology that includes all year from 1959 to 2022, for example?
Â
Lines 115-138: speaking as someone who is less familiar with SOMs than other researchers, I found the explanation of how SOM clustering works in this section to be very clear and appropriately concise.
Â
Lines 119 - 120: I think a more intuitive phrase could be used instead of “arraying the clusters on the nodes of a regular 2D grid.” For example, the authors could write “The SOM adds another layer to this algorithm, by organizing the clusters into nodes within a regular 2D grid…”
Â
Lines 165-166: I am confused by what this sentence is meant to convey. Maybe the authors are just trying to say that the transition matrices allow a metric to be computed that communicates how predictable a SOM node is following the occurrence of a prior, different SOM node? Not sure, but I think this sentence needs to be reworded for clarity. The subsequent sentences in this paragraph, however, make perfect sense to me.
Â
Lines 188-189: I think it is worth rewording this sentence that starts with “As a first difference to S17…” to be a little more clear and to make it obvious that vertical maxima in the wind speed U, specifically, is used to identify the presence of a jet. Maybe instead, the authors could write something like, “Whereas S17 identifies the presence of a jet using the vertical maximum in U on the 2PVU surface, we use vertical maxima over several high-altitude pressure levels.” I also recommend adding a brief clarification that “PVU” refers to potential vorticity units, where 1 PVU = 10-6 K kg-1 m2 s-1. I have observed that making this clarification is customary in peer-reviewed literature that references potential vorticity, even minimally.
Â
Line 195: I am not sure what a “contour library” is, and I think this term might not be commonly used in scientific literature focused on synoptic meteorology. Does this refer to a publicly-available coding package for Python users, for example? I think some text to briefly clarify what this term refers to should be added, or maybe there is a more widely used term to replace it with.
Â
Lines 204-207: I find the text here to be confusing, especially because the criteria described here sound very similar to the criteria introduced in the previous paragraph (lines 197-203). Are the criteria described in these lines applied in addition to the criteria in the previous paragraph? If so, the statement that “jets are defined” with the application of this second set of criteria seems to contradict the statement that “jets are… extracted” using the criteria outlined in the previous paragraph. It sounds like the jets that are extracted in lines 197-203 may not officially be “jets” yet. In light of all this confusion, I think it is worth modifying the text on these lines (and possibly on lines 197-203) to make the step-by-step process here clearer.
Â
Lines 212 – 214: I feel as if saying “the two differences” here makes it seem as if these are the only two differences between this algorithm and S17, whereas there is at least one other difference I see mentioned in Section 2.3.1 (the use of vertical wind speed maxima over several high-altitude levels instead of just on the 2PVU surface). I think it would be easiest for readers to follow these differences if all of them were listed together in the same paragraph after the step-by-step description of the jet identification algorithm has been completed. For example, the authors could write something along the lines of “The difference in results from our algorithm and the S17 algorithm are due to three important differences in algorithms themselves.”, and then list each of the differences (including the departure from the use of the 2PVU surface) following this sentence.
Â
Line 222: “low altitudes” may be a typing error (i.e., the authors may have meant to write “high altitudes” instead). If this is not a typing error, then why is the averaging performed at low altitudes instead of high altitudes, given that we know jet cores reside at high altitudes? I think adding some text briefly explain this might be helpful.
Â
Lines 270-271: I am struggling to understand what the sentence starting with “One instead has to rely…” despite being very familiar with the Winters and Martin 2017 study. Part of my confusion comes from the fact that I do not know what variable is used to compute the “depth” that this sentence refers to – is this the depth of the drops in the tropopause height that accompany the presence of a jet? If so, why do low-level winds and latitude impact the use of these drops in tropopause height to categorize the jet? I think this sentence in the manuscript could be reworded or some specificity could be added to provide more clarity.
Â
Lines 276-277: could the authors offer some explanation as to why they chose to use jet frequencies in mean latitude, mean longitude, and pressure level as opposed to jet frequencies in wind speed and potential temperature? This may help clear up some of my concern mentioned in the “General Comments” section of this document, which highlights my concern toward using the mean latitude, mean longitude, and pressure level approach. I know that the authors acknowledge that there may be some mis-categorization of jets using their method, but I worry this mis-categorization could potentially be too large using this method.
Â
Line 277: I think most readers will be familiar with the fitting of a unimodal Gaussian distribution to a 3D dataset, but I think many would be unfamiliar with what a “two-component Gaussian mixture” is refers to. Does this just involve fitting two overlapping Gaussian distributions to the 3D data? Some elaboration on what this fitting technique involves could be very useful.
Â
Line 279: I feel that text is missing from this paragraph which connects the construction of the histograms to the categorization of individual jet objects. Could a sentence or two be added that explains how the delineation between the EDJ and STJ regimes in the histograms are determined, and that an individual jet object is assigned either the STJ or EDJ classification depending on which regime it falls within?
Â
Line 321: I would argue that the EDJ appears fairly wavy in the top most row (nodes 1-6) as well, and I think this is worth noting. Unless the authors intended to use the word “columns” instead?
Â
Lines 321- 323: I recommend either adding the exhaustive list of clusters indicating each weather regime, or adding “e.g.” prior to the listing of each clusters in the sets of parentheses to make it clear that these are not the only clusters featuring each weather regime.
Â
Line 330: I wonder if the authors could offer some intuition in the manuscript as to why cluster 8 is so common, especially since this is such a stark result from Figure 6. Please ignore this comment, however, if this topic is already discussed at a later point in the manuscript (I may have missed it). Additionally, I do wonder if cluster 8 could be seen as sort of a “catch all” cluster for any upper-level flow patterns that do not fit into other nodes, especially given that Figure A1 shows that composited geopotential height anomalies for this cluster are relatively weak compared to other clusters. I imagine that a Sammon map for the SOM may help indicate whether cluster 8 is indeed a “catch-all” cluster?
Â
Lines 367 - 370: I am confused on what information led the authors to the distinction made between the interpretation of the high residence times for clusters 2, 8, and 13 and the interpretation of the high residence times for cluster 17 and 18. I may be missing something critical that was mentioned earlier in the transcript, but I think some text could be added here to clarify why clusters 2, 8, and 13 are not representative of true state persistence.
Â
Lines 403-404: I think the text which reads, “All the results are split by jet category and always colored in the same way: pink for the STJ and purple for the EDJ. The double jet index is colored in black.” would be the kind of descriptive text that belongs in the caption for Figure 10 as opposed to in the main text of the manuscript, since most readers will look to the caption first to understand the use of color in the figure.
Â
Line 489-491: I wonder if the authors could briefly offer a hypothesis or speculation in the manuscript text as to why the two methods of assessing persistence often disagree.
Â
Line 515: I think “waveguidability” is not a commonly-used word in the literature, and I myself am not quite sure what that means. Could the authors add some brief text to clarify what this term refers to?
Â
Line 546: I really appreciate that the authors added a discussion on methods they tried and eventually rejected for jet feature extraction, and felt that it was a very thoughtful addition. I think the knowledge presented here will be highly useful for readers who may be interested in developing similar datasets involving feature extraction.
Â
Figure 4 and Figure 4 caption:
- I think the word “arrayed” is not commonly used and I worry that it will not be intuitive for many readers. Could a different word or phrase be used here to describe the display of jet frequencies in each hexagonal bin?
- I am confused on why the units are arbitrary in this figure. Most plots of this nature have frequency units. For example, does panel (a) not show the count of identified summer jets per hexagonal bin? My reason for asking this is that if possible, I think adding a colorbar (or colorbars) to quantify the magnitudes of the shading here would be helpful. I would certainly be very interested in seeing the magnitude of the frequencies here.
- The cutoff between EDJs and STJs is drawn through a very high-density region of the histogram in panel (b), which makes me feel concerned that classifications of jets as either EDJ or STJ could be very sensitive to where this cutoff is drawn. Could a sensitivity test be performed to illuminate how much the choice of this cutoff impacts the frequencies of each category both overall and a geographic map (i.e., in latitude/longitude space)?
Â
Figure 5: I feel that some nodes in this figure further demonstrate my concern over how physical the delineation between the EDJ and STJ is. For example, the switch from the EDJ to the STJ categorization assigned to the relatively zonal jet in nodes 4 and 5 feels a bit arbitrary. I wonder if it could be more physically reasonable to think of most of the zonal extent of this jet as being more like a STJ in nature.
Â
Figure 7: I think it would be useful to know whether the population of each cluster is in units of days or timesteps. Could a phrase be added to the Figure 7 caption that mentions the units, and could the units be added to the colorbar in the figure as well?
Â
Note for all figures: I really appreciate how clean and aesthetically pleasing all of the figures are. I found them very easy to read.
Â
Â
Technical Corrections
Â
Line 19: missing “of” after “instead”
Â
Line 21: missing “a” after “towards”
Â
Line 36: missing connecting word after “clear”; maybe this should be modified to read “… always clear, since both sources of momentum…”
Â
Line 47: I suggest using “favor” instead of “favorise”
Â
Line 57: I suggest replacing “The signal is however weak…” with “However, the signal is weak…”
Â
Line 89: I think the authors meant to write “stationarity” instead of “stationary”
Â
Line 95: I suggest replacing “works” with “work”
Â
Line 140: I suggest inserting “(June, July August)” after “JJA” to clearly define the acronym for readers
Â
Line 151: misspelled “matrix”
Â
Line 174: I recommend replacing “With a large SOM with sometimes…” with “With a large SOM that sometimes has…”
Â
Line 182: I believe a reference to some other section of the paper and a closing parentheses bracket might be missing here
Â
Lines 182-183: I recommend rewording “This is why we apply it to the full dataset…” to instead read “Therefore, we apply our detection method to all seasons within our dataset…”
Â
Line 188: I think the authors may have meant to write something along the lines of “As done prior to training the SOM,” instead of “As for the SOM”
Â
Line 212: I suggest replacing “difference” with “differences”
Â
Line 273: I suggest replacing “categorization bins and counts” with “categorization involving binning and counting”
Â
Line 283: misspelled “characterized”
Â
Line 286: I think the word “algorithm” may be missing after “A straight forward feature tracking”
Â
Line 303: I recommend spelling out the word “hour” in “6H-timesteps” to follow the convention used in previous sentences
Â
Line 308: I suggest replacing “result” with “results”
Â
Line 317: I recommend replacing “high-zonal-overlap double jet states” with “the double jets with high zonal overlap” to improve sentence flow here
Â
Line 319: I recommend replacing “centre right” with “center-right” to follow the convention used in the previous sentence, which uses “center-left”
Â
Line 328: I recommend replacing “honeycomb” with “hexagonal” to follow conventions used in previous paragraphs
Â
Line 332: I recommend replacing “panel b” with “panel (b),” and following this convention at other similar places in the manuscript text
Â
Line 342: missing “in” after “resulting”
Â
Line 353: misspelled “occurrence”; I recommend spelling out “JA” as “June and August,” or clarifying that “JA” refers to “July and August” in order to establish the meaning of the acronym
Â
Line 354: I recommend replacing “…in early June and cluster 2…” with “…in early June, cluster 2…”; I also recommend replacing “…early June, cluster 8…” with “…early June, and cluster 8…”
Â
Line 363: I recommend replacing “to flow” with “the large-scale flow pattern” or “flow pattern”
Â
Lines 363-364: I recommend replacing “…into an next state and the 95th percentiles…” with “…into the next state, while the 95th percentiles…”
Â
Lines 397: I recommend replacing “way” with “ways”
Â
Line 400: I recommend replacing “width’s” with “width”
Â
Line 412: I recommend replacing “…between STJ latitude and Hadley cell edge…” with “…between STJ latitude and the Hadley cell edge…”
Â
Line 420: I recommend replacing “questions” with “question”
Â
Line 433 – 434: I recommend rewriting “Some trends. like for the double jet index, even change signs between summer and all-year.” as “Some trends exhibit different signs when comparing the summer period to the full year, such as for the double jet index.” to improve the sentence flow and fix punctuation errors
Â
Line 435: I suggest replacing “accord” with “agreement”
Â
Line 449: misspelled “across”
Â
Line 450: “June” was not capitalized
Â
Line 481: I think the word “known” is not supposed to be here
Â
Line 478: I recommend adding “in upper-level flow patterns” after “seasonal shift” to be more a little more precise
Â
Line 485: I recommend defining STJ and EDJ as referring to the subtropical jet and eddy-driven jet prior to their first use in Section 4: Discussion and summary
Â
Lines 493-494: I recommend rewriting “Computing the properties the jet features of every time step” as “Computing properties of the jet features at every time step”
Â
Line 539: misspelled “expectedly”
Citation: https://doi.org/10.5194/egusphere-2024-3029-RC1
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
120 | 25 | 9 | 154 | 4 | 6 |
- HTML: 120
- PDF: 25
- XML: 9
- Total: 154
- BibTeX: 4
- EndNote: 6
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1