the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Synoptic weather patterns conducive to compound extreme rainfall-wave events in the NW Mediterranean
Juan Carlos Peña
Sotiris Assimenidis
José Antonio Jiménez
Abstract. The NW Mediterranean coast is an area at high risk of impacts generated by extreme rainstorms and coastal (wave) storms, which generally result in flash floods, coastal erosion, and flooding along a highly urbanised territory. In many cases, these storms occur simultaneously, and compound events can increase the local impacts (when they occur in the same place) or accumulate the impact throughout the territory (when they occur in different locations). In both cases, multivariate and spatially compound events represent a challenge for risk management because they can overwhelm the capacity of emergency services. In this study, we analysed the prevailing atmospheric conditions during the occurrence of different types of extreme episodes to produce the first classification of synoptic weather patterns (SWPs) conducive to compound events (heavy rainfall and storm waves) in the Spanish NW Mediterranean. For this purpose, we developed a methodological framework by combining an objective synoptic classification method based on principal component analysis and k-means clustering with a Bayesian Network to characterise the nonlinear relationships between SWPs and different variables characterising storms. This method was applied to a dataset of 562 storm events recorded over 30 years, of which 112 were compound. The obtained SWPs were grouped into three main types, of which the Cut-Off was dominant in terms of multivariate event occurrence and was also the situation under which the most severe compound events occurred. The position and depth of the upper-level cold air pools and surface lows affect the relative weight and spatial distribution of the terrestrial (rain) and maritime (waves) components. Finally, the Bayesian Network allowed for a quantitative assessment of the obtained SWP classification by measuring the prediction skill of the target storm variables (i.e. daily precipitation or maximum wave height). Reasonably good skill results were obtained using the SWP as a predictor when accompanied by an additional variable capturing seasonality and event duration. These findings contribute to the overall understanding of compound terrestrial-maritime phenomena in the study area and may assist in the development of effective risk management strategies.
- Preprint
(1514 KB) - Metadata XML
-
Supplement
(3109 KB) - BibTeX
- EndNote
Marc Sanuy et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2023-104', Anonymous Referee #1, 14 Jun 2023
Review of the manuscript: Synoptic weather patterns conducive to compound extreme rainfall-wave events in the NW Mediterranean by Marc Sanuy, Juan C. Peña, Sotiris Assimenidis, and José A. Jiménez
The paper analyses weather events inflicting hazardous impacts over the Spanish northwest (NW) Mediterranean, such as floods and coastal storms characterized by high waves. The article analyses synoptic weather patterns (SWPs) conducive to compound events by combining an objective synoptic classification method based on principal component analysis and k-means clustering with Bayesian Networks (BNs). As the first method is a rather traditional method used in classifying synoptic patterns, the main innovation is adding BNs analysis. By adding BNs skills analysis to their classification method, the authors claim its advantage is characterizing the nonlinear relationship between SWPs and different variables for predicting compound extremes. The subdivision and research were done to contribute to understanding compound terrestrial-maritime phenomena in the study area and to assist in developing predictive and effective risk management strategies.
Dear Authors,
Your research is innovative by adding BNs to traditional methods. Your combined methodological framework shows promising results. However, my main issues are the work presentation, which is sometimes difficult to interpret, and your classification procedure. For example, the number of clusters you choose to describe the atmosphere seems too large. I.e., 18 weather types to describe 112 compound events? I have included my comments and suggestions below for you.
Abstract
1. Line 26 – What do you mean by ‘reasonably’? Please give some quantitative estimates.
Introduction
2. Line 56 – A few recent review articles on extreme weather in the Mediterranean region are missing from your reference list.
Flaounas, E., Davolio, S., Raveh-Rubin, S., Pantillon, F., Miglietta, M. M., Gaertner, M. A., Hatzaki, M., Homar, V., Khodayar, S., Korres, G., Kotroni, V., Kushta, J., Reale, M., and Ricard, D.: Mediterranean cyclones: current knowledge and open questions on dynamics, prediction, climatology and impacts, Weather Clim. Dynam., 3, 173–208, https://doi.org/10.5194/wcd-3-173-2022 , 2022.Hochman A, Marra F, Messori G, Pinto JG, Raveh-Rubin S, Yosef I, Zittis G,. 2022. Extreme weather and societal impacts in the Eastern Mediterranean. Earth System Dynamics 13(2): 749-777. https://doi.org/10.5194/esd-2021-55
Zittis, G., Almazroui, M., Alpert, P., Ciais, P., Cramer, W., Dahdal, Y., et al. (2022). Climate change and weather extremes in the Eastern Mediterranean and Middle East. Reviews of Geophysics, 60, e2021RG000762. https://doi.org/10.1029/2021RG000762
3. Line 61 – Do you mean ‘objective’ rather than ‘subjective’?
4. Lines 56 – 62 - A few articles on synoptic weather classification and their physical grounding are missing from your reference list. Please consider adding them.For example:
The special issue entitled: Circulation-type classifications in Europe: results of the COST 733 Action
I would mention COST733 and describe its contributions in the introduction.Hochman A, Messori G, Quinting J, Pinto JG, Grams C. 2021. Do Atlantic-European weather regimes physically exist? Geophysical Research Letters 48: e2021GL095574. https://doi.org/10.1029/2021GL095574
Data
5. Lines 120 – 125 – Please add more information on how wave height reconstruction was done.
6. Line 130 – Please add latitude and longitude to Figure 1.
7. Figure 1 – Please increase the fonts of country labels. Please add the topography to the map.
Methods
8. Line 146 – Please add detailed information in the caption of Figure 2 for the reader to be able to interpret your framework without looking it up in the main text. This comment can be applied to most of your figures.
9. Line 161 – Typo remove ‘as.’
10. In Line 185 and throughout the text, I think you mean ‘trough’ rather than ‘through.’
11. Line 205 – Please add more information in the Figure 3 caption for the reader to be able to interpret without looking it up in the main text.
12. Line 236 – Add information in the Figure 4 caption for the reader to be able to interpret without looking it up in the main text.
Results
13. Line 246 – remove ‘affected.’
14. Line 264 – Are 18 weather types for 112 events too large? How many clusters do you have, and what is the explained variance? The issue of selecting a priori number of groups is essential, and you should discuss it.
For example:
Falkena, S. K., de Wiljes, J., Weisheimer, A., Shepherd, T. G. (2020), Revisiting the identification of wintertime atmospheric circulation regimes in the Euro‐Atlantic sector. Quarterly Journal of the Royal Meteorological Society, 146, 2801–2814. https://doi.org/10.1002/qj.3818
15. Section 4.2 – I suggest significantly reducing the text amount in this section. It isn’t easy to read.
16. Table 3 – Why is this table included in the text? Is it necessary, or can it be moved to the supplementary information?
17. Figures 5, 6, and 7 – Please consider using anomalies in the figures rather than absolute values.
18. In all figures, please use letters (a, b, c..) for each panel so the reader knows what panel you are referring to in the text. Also, you mention these letters in the captions of the figures, but they need to be shown on the figure.
Discussion
19. The discussion section can be significantly shorter.
Supporting information
20. There are too many panels in the supporting information figures, which could be clearer to interpret.Citation: https://doi.org/10.5194/hess-2023-104-RC1 - AC1: 'Reply on RC1', Marc Sanuy, 02 Aug 2023
-
RC2: 'Comment on hess-2023-104', Anonymous Referee #2, 04 Jul 2023
General comments
The paper by Sanuy et al. “Synoptic weather patterns conducive to compound extreme rainfall-wave events in the NW Mediterranean” presents (1) a synoptic-climatological assessment of compound extreme events over the NE part of Spain coupled with (2) the use of Bayesian networks that quantify the nonlinear links between the synoptic types and various variables describing the extreme events.
Overall, I find the study interesting and potentially worthy of publication, but only once the authors have successfully addressed the major issues. These issues comprise the methodology, the balance between the weather typing and its subsequent application, and the clarity of presentation (both text and figures).
Specific comments
In synoptic climatology, it is well known that links between synoptic-scale circulation and any conditioned surface variable are sensitive to how the circulation domain is defined, both in terms of its size and localization relative to one's area of interest. Based on the presented results, I do not understand why for studying extremes mainly dependent on close-by lows or troughs the authors decided to analyse atmospheric circulation over such a broad area. This is particularly striking for precipitation variables, which require smaller domains for good skill (Beck et al. 2016; IJC 36:7). This issue demonstrates itself when the authors train their networks, the skill of which is very low for precipitation. The authors suggest that including other variables including large-scale teleconnections may help, but I suggest that their primary focus be on smaller rather than larger scales. I strongly recommend that the authors experiment with the size and location (i.e. centre versus off-centre) of their circulation domain and assess the sensitivity of the networks’ skill to these changes. If classifications are trained independently on each type of event, there is even no reason to use an identical region for each of them.
On a similar note, I am not convinced that including MSLP, Z500 and 10m wind components adds in this particular case any synoptic skill. The authors claim they were motivated by Miró et al. (2018), who however did not analyse events related to this analysis, as the authors claim in 344, but rather cold-air pools in which decoupling between local (site) and regional circulation systems was crucial. As part of discussion, the authors should include a sensitivity study showing how their BN outputs respond to inclusion of multiple mutually strongly correlated circulation variables.
Based on the description, it is not clear whether the circulation variables were standardized prior to PCA decomposition – the authors only mention they used anomalies, which would not be enough to account for the differences in variables’ variance. In such a case, only one variable would have a dominant impact on the classification output anyway.
This relates to my other comment. I do not understand why were the weather types defined independently for each of the extremes' types, why the authors decided to define that many types (I do not claim that it is wrong but – again – no evaluation of the effect on the networks’ skill was carried out), especially considering the fact that each classification was subsequently (manually) clustered into three “supertypes” that have long been known to link to the studied extremes in the region.
Compared with the synoptic types, the description of which seems unnecessarily too lengthy and overly detailed to me, the text describing the networks seems way too short. Note that the classification and its descriptive analysis represent an interesting exercise. However, the added value consists (or, should consist) in the subsequent analysis.
I strongly suggest decreasing the number of abbreviations used in the text. For instance, why using C/SR/SW instead of simply compound, rain and waves? In some parts, where these are combined with abbreviations of variables and types, the text is extremely hard to read. Last, please try to simplify your terminology, better explain it, and use it consistently.
The quality of the figures is not great. It is practically impossible to see the background maps on screen, let alone when printed.
Technical corrections/queries
61 in what sense are PCA- or CA-based classifications "more subjective than those" mentioned above?
71 One may argue that the study by Sanuy et al. (2021, HESS) already did this, albeit to a limited extent
73 It is not clear what "different mechanisms" mean; different to what?
74-76 Please reword this sentence, it is hard to understand
90 did you mean "forcing on"?
2.1 Consider moving 89-93 (local scale) after the description of the larger-scale factors of extreme events
128-129 why was MSLP abbreviation defined in 82 but z500 was not?
136 Are waves also a meteorological driver or is it a typo?
137 I do not understand what "weather classifications associated SWP" means. Aren't SWP a synonym for weather (circulation) classification/types?
138 Isn't the classification method used to identify dominant SWPs? Also, what is "dominant", "critical", "target variables", and "SWP system"? I like the inclusion of the general framework and Fig. 2, but at this point they are not very clear, i.a. due to inclusion of abbreviations that have not been defined
144 Was P24h explained in Sect. 2?
158 P24 and P24h: what is the difference?
161 "as as"
174 Does "separately" mean that each variable's anomaly fields were decomposed separately?
175 What do you mean by "fundamental" modes and how does they relate to the anomalies?
179 So how many PCs/clusters were finally selected?
181 How were they grouped? You need to show the patterns and refer to them from here. Also this text suggests that you join the types in three final types for which you present results, which is not the case. Please reword. Furthermore, I recommend using a different term for your overarching three patterns, such as "supertypes", to clearly distinguish them from SWPs.
197 all?
200 what is parent?
201 is "SWP system's proficiency" the same as "classification's ability"? If the term system is important, please explain it sooner.
204 correct "(SWPs)"
209 Why are not all of these parameters/variables mentioned in Sect.2?
223 Standard deviation of the skill?
230 Why do you repeatedly define abbreviations that you then don't even consistently use?
235 Please add panel captions into the figures, and refer directly to individual panels from the text as much as possible.
252 How is there a decrease from North to South is the values are identical for the central and southern regions?
263 What does "their" refer to?
266 What was verified?
266 The sentence "highlighting...severity" makes no sense
273 Are local multivariate events the same as compound events? I am becoming lost in the terminology.
282 "The dominant..." > "The two dominant..." ... "all of them" means those in Figs 5+6 or something else? Please explain/reword "minimum relative cumulative spatial temporal all basins probability"
285 What do you mean by "unstable conditions"?
287 In all seven? I can hardly check because of the quality of the map, but it seems unlikely that all seven types (than even do not belong to the same “supertype”) have the same feature.
297 Please explain the abbreviations of types in the text, or use generic names (e.g., type1, etc.). It is impossible to remember which type is which.
346 There are more objective alternatives to visual checking for similarity. One can calculate a pattern-correlation matrix, or a distance matrix, project all SWPs to the first principal component plane, or alternatively one can use e.g. Sammon mapping to test whether SPWs that you identify with different supertypes truly occupy distinct parts of the data space without the linear/orthogonal PCA constraints.
370 is it the same skill as in 221? I suggest adding a reference to the section where it is explained
393 remove comma before "leading". Better explain "dominant" and "general"
424, 737 etc. What is "weather configuration"? Please select a clear terminology and use it consistently.
429 This describes the results of the reference, or yours?
448 consider changing "this work" to "their work" or similar
Citation: https://doi.org/10.5194/hess-2023-104-RC2 - AC2: 'Reply on RC2', Marc Sanuy, 02 Aug 2023
Marc Sanuy et al.
Marc Sanuy et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
336 | 81 | 16 | 433 | 30 | 5 | 3 |
- HTML: 336
- PDF: 81
- XML: 16
- Total: 433
- Supplement: 30
- BibTeX: 5
- EndNote: 3
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1