the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Oct 2023
A storm-relative climatology of compound hazards in Mediterranean cyclones
Raphaël Rousseau-Rizzi
Shira Raveh-Rubin
Jennifer Catto
Alice Portal
Yonatan Givon
Olivia Martius
Abstract. Cyclones are responsible for much of the weather damage in the Mediterranean region, and while their association with individual weather hazards is well understood, their association with impactful multivariate compound hazards remains to be quantified. This study aims to establish a storm-relative climatology of three different multivariate hazards in Mediterranean cyclones. Namely, the co-occurrences of rain and wind, rain and wave, and particulate matter and heat are composited relative to storm centers. Composites are computed for various large-scale environments using a recent cyclone classification, which shows that few different large scale configurations lead to each compound event type. Compound rain and wind events are mostly associated with frontal cyclones and cyclones induced by anticyclonic Rossby wave breaking from late fall to early spring in the northern Mediterranean. Compound rain and wave events occur at similar times and locations, but are also associated with cyclonic Rossby wave breaking. Particulate matter and heat compound events are associated with heat lows, daughter cyclones and anticyclonic Rossby wave breaking in the warm season and over North-Africa. Next, we find that the probability of compounding associated with a cyclone class does not depend monotonically on the probabilities of the individual contributing hazards, but also on the goodness of their temporal and spatial correspondence. Finally, we find warm conveyor belts and cold fronts to frequently co-occur with rain and wind, and rain and wave events, while particulate matter and heat events are not strongly associated with dynamical features. These results, which systematically associate various large-scale environments and dynamical features to different compound event types, have implications for forecasting and climate risk predictions.
- Preprint
(3366 KB) - Metadata XML
- BibTeX
- EndNote
Raphaël Rousseau-Rizzi et al.
Status: open (until 30 Dec 2023)
-
RC1: 'Comment on egusphere-2023-2322', Leonardo Aragão, 21 Nov 2023
reply
The study by Rousseau-Rizzi and collaborators sought to identify and quantify the coexistence of two hazards related to cyclonic activities over the Mediterranean Sea. The analysis was based on three pairs of hazards: rain and wind, rain and waves, and concentrations of PM10 and heat. The outstanding contribution of the study lies in its separate analyses for different types of cyclones, with clusters classified by upper-air potential vorticity. Another interesting investigation concerns the probability of occurrence of each pair of hazards in situations of warm conveyor belts, cold fronts and dry intrusions, always associated with cyclone's passages. Perhaps the most delicate point concerns selecting hazard events based only on one hour of the life cycle of a Mediterranean cyclone, significantly limiting the identification of potential extreme risks if we consider that each risk has its maximum at different moments of a storm. In any case, the study is presented coherently and in detail, and its publication makes essential contributions to quantifying hazards associated with Mediterranean cyclones.
Please find below a list of some questions, suggestions and corrections regarding the manuscript:
[ Lines 23-24 ] “a broad variety of hazards ranging from sea waves”
[ Suggestion ] Perhaps this list could be extended to storm surges, coastal erosion, floods and landslides, as the passage of cyclones is also directly associated with these events.[ Line 24 ] “pm10”
[ Correction ] In the related literature, pm10 is spelled PM10. Please replace this in the following appearances.[ Lines 44-45] “citation of currently embargoed thesis by Laura Owen”
[ Correction ] Unpublished studies cannot be cited. Everything written in this manuscript will be further cited referring to this study. Consider removing this information or inviting Laura Owen as a co-author if it is relevant to a future Owen publication. This should also be applied to the following citations related to Owen's work.[ Line 49 ] “PV”
[ Correction ] Please move the acronym's introduction from line 96 to here on its first occurrence.[ Line 64 ] “... on three types of multivariate compound events,”
[ Question ] Why not wind and wave?[ Line 65 ] “heat waves”
[ Correction ] This hazard was not assessed in this study. Perhaps warm-spell is more appropriate here.[ Line 70 ] “cold fronts (CF)”
[ Correction ] To make the acronyms' identification more accessible, I suggest moving the introduction of CF to line 39, where WCB and DI are also introduced.[ Line 77 ] “Data and methods”
[ Question ] Could you clarify which dataset was used to define the study period, as several datasets covering different periods were utilized?[ Line 90 ] “ERA5 reanalysis data”
[ Correction ] You must cite Hersbach et al (2020) instead of Flaounas et al (2023).[ Line 91 ] “leads to accurate tracking of storm centers”
[ Suggestion ] In this case, I would be more accurate by adding " at the Mediterranean scale" to this sentence.[ Line 92-93 ] “... activity around Italy and southern France as well as over the Sahara and south of Turkey.”
[ Suggestion ] In Mediterranean cyclone studies, this region is often known as "the lee of the Alps", referring Buzzi & Tibaldi (1978). The same works for the Sahara and Turkey. I suggest referring to the Atlas and the Anatolian mountain chains instead of associating these positions with country or region names.
Buzzi A, & Tibaldi S (1978). Cyclogenesis in the lee of the Alps: a case study. QJR Meteorol Soc 104:271–287.
https://doi.org/10.1002/qj.49710444004[ Line 93-94 ] “... there is more storm activity over bodies”
[ Suggestion ] Depending on the tracking method, more storms can be identified over land during the summer months. As Flaounas et al (2023) presented composite tracks, only those simultaneously identified by N methods were considered (in this study, N is also called Confidence Level), keeping only parts of tracks in which the different methods overlapped. This explains the high number of tracks over sea once there are divergences between methods in identifying tracking points over land. Perhaps you should add a comment pointing to this.[ Line 97-98 ] “... the upperlevel potential vorticity (PV)”
[ Suggestion ] Considering the importance of this clustering for constructing your analyses, it would be interesting to describe them briefly and explain why you chose so many clusters (9).[ Line 98] “... introduced by Givon et al. (2023).”
[ Suggestion ] This method dates back to the 90s and early 2000s with several works from K. Emanuel, R. Romero, and others. It is OK if you used Givon et al (2023), but I think it is essential to cite works such as:
Davis K A & Emanuel K A (1991). Potential Vorticity Diagnostics of Cyclogenesis. Mon. Wea. Rev., 119, 1929–1953.
https://doi.org/10.1175/1520-0493(1991)119<1929:PVDOC>2.0.CO;2
Romero R (2001). Sensitivity of a heavy-rain-producing western Mediterranean cyclone to embedded potential-vorticity
anomalies. Q.J.R. Meteorol. Soc., 127: 2559-2597. https://doi.org/10.1002/qj.49712757805[ Figure 1 ] ...
[ Correction ] Name the colour scale units and, if possible, enlarge the whole figure.[ Line 100 ] “... using self-organizing maps”
[ Correction ] Self-Organizing Maps (capitalize each word regarding acronyms).[ Line 101 ] “... as the associated mean PV fields can be viewed in Figs. X and Y”
[ Correction ] I think something is missing here.[ Lines 102-103 ] “... only the large scale environment of storm clusters shown to be responsible for compound events will be discussed.”
[ Question ] Is this "the large scale environment of storm clusters" defined by Givon et al (2023), or is a subsequent filter applied to keep only these storms?[ Line 106 ] “... heat wave data”
[ Correction ] ERA5 ReAnalysis does not include heatwave data. More than that, this study does not address the heatwave phenomenon. Please cite air temperature.[ Line 107 ] “... but only saved at 6h intervals”
[ Correction ] What about the spatial resolution?[ Line 118 ] “... (CAMS) reanalysis”
[ Correction ] Please, add the dataset spatial-temporal resolutions.[ Lines 127-128 ] “Note that the methods used to identify DIs and WCBs are similar, while the front identification method is very distinct.”
[ Suggestion ] Better understanding of this information is essential. Consider adding short sentences summarizing these methods.[ Lines 135-136 ] “a given hazard magnitude may be benign in one region and catastrophic in another, and local percentile thresholds are most appropriate to capture the local risk. Otherwise, a fixed threshold may be used to define the occurrence of events everywhere.”
[ Suggestion ] Check if a citation is needed. There is something in this sentence that looks like a déjà vu.[ Line 139 ] “... 99th percentile”
[ Question ] Have you computed the 99th percentile considering all hourly data, or did you filter hours without precipitation (< 1 mm/h)?
Hours without precipitation often accumulate 70-90% of the distribution. Including this in percentile computation can mask the actual threshold.[ Line 148 ] “... and interestingly, wherever Mediterranean cyclone track density is high”
[ Suggestion ] From my point-of-view, it is expected since the cyclogenesis and cyclolysis hotspots in the Mediterranean regions along the mountain chains. Try to rewrite and add this information.[ Line 162 ] “30 C”
[ Correction ] 30°C?[ Lines 161-162 ] “Finally, for heat events, we chose a fixed 30 C threshold to be consistent with previous studies of increased mortality in the
event of co-occurring dust and heat extremes (Katsouyanni et al., 1993).”
[ Correction ] Divergence. Until now, the authors introduced a heat component described by heatwave events. But here, the threshold definition seems more similar to a warm-spell description and changes the final number of events a lot. Please note that the previous information regarding this component must be adjusted.[ Line 176 ] “... only at the time of minimum central pressure.”
[ Question ] How much of the total storm damage do the authors estimate to cover using a single time to represent a storm track as a whole? Consider that until here, it is not clear what the timestep used (1h or 6h). Also, consider that during a storm life-cycle, there are several moments of severe weather (precipitation, wind, waves, etc), and often it doesn't happen at the same time as the minimum central pressure.[ Lines 176-178] “Since the time resolution of the storm tracks is sometimes higher than that of the hazard data, we select the minimum pressure over the times when hazard data is available.”
[ Correction ] It is impossible, considering the information the authors have given up to this point (hourly hazard data X 6-hourly cyclone tracks data).[ Line 183 ] “... Storm ID”
[ Question ] It is not clear what Storm ID is.[ Line 183 ] “... the Cyclone ID dimension.”
[ Question ] What is Cyclone ID dimension?[ Line 196 ] “... over the Mediterranean”
[ Correction ] Do you mean, Mediterranean Sea?[ Line 233 ] “... the horizontal (latitude-longitude) dimensions of a 20×20 degree box around the cluster center.”
[ Question ] As the search box is always a 20x20 degree area, can we consider these "horizontal dimensions" a unique and constant value for all storms?[ Line 234 ] “... which may be interpreted as the fraction of time”
[ Correction ] In this case, it is not a matter of time but of cases, so it must be a "fraction of all cases".[ Lines 246-247 ] “This leads us to the observation that rain-wind compound probabilities are systematically multiple times smaller than either rain probabilities or wind probabilities alone.”
[ Question ] Wasn't this expected?[ Lines 249-250 ] “Notably, cluster 2 which exhibits the highest compounding probabilities, has lower individual probabilities of both precipitation and wind events than cluster 4.”
[ Question ] At this point, we return to not considering the entire cyclone's life cycle. Each cyclone type has moments of lower pressure in the centre, more significant precipitation, intense winds and higher waves. Likewise, each of these extremes is known to occur at particular times during the cyclone's life. So, I wonder if the compounding probability's magnitude is relevant or just its existence is enough. In other words, does defining the compounding probability from 0 to 1 give us information different from true or false?[ Lines 255-256 ] “As a result, the rain probabilities shown here are slightly different than the ones shown in Fig.4.”
[ Correction ] This information would be relevant if we knew the number of events identified and the number of years that the study covers. As readers, we have no idea whether these numbers are relevant on a climatological scale or represent ten years when the Mediterranean Sea SST was extraordinarily high, for example.[ Line 266 ] “Note the case of cluster 2,4? ...”
[ Correction ] I think something is missing here.[ Lines 268-269 ] “In fact, none of the clusters most important for wind and rain compositing (2, 4 and 1) are associated with heat event field significance.”
[ Question ] What correlation with cyclones would the authors like to highlight in analysing the pm10-heat? From the manuscript's beginning, it is unclear how this hazard composition is associated with cyclone passages. In theory, the most significant risks concerning this composition are expected to occur specifically in the absence of cyclones.[ Figure 4 ] ...
[ Suggestion ] Consider presenting Figures 4, 5, and 6 in a single figure to facilitate comparing these results and following the analysis described in the manuscript.[ Lines 279-280 ] “... the density plots capture similar high-density 280 areas as Fig.1, which represented the density of tracks at all times.”
[ Correction ] This is mandatory as your data has been filtered for cyclone tracks.[ Lines 283 ] “... and craters to zero in mid-to-late summer.”
[ Correction ] To provide a more accurate understanding, let's highlight that the dataset used to track cyclone tracks has certain limitations during the summer months. These limitations arise due to the need to converge different methods to identify and track cyclones. By acknowledging these limitations, we can work towards finding more effective solutions to track cyclonic activities accurately in all seasons.[ Lines 296 ] “(Givon et al. 2023)”
[ Correction ] If the results in Figure 8 were not produced by Givon et al (2023), there is no reason to cite this study.[ Figure 7 ] ...
[ Correction ] Name the colour scale units and, if possible, enlarge the whole figure.[ Lines 303 ] “... cluster 8 ORM TO ME ALSO ACWB, CL7 LOOKS CYCLONIC may be associated ...”
[ Correction ] Check this sentence, please.[ Figure 8 ] Caption "... potential? sea-level pressure"
[ Correction ] Check this sentence, please.[ Figure 8 ] ...
[ Correction ] Name the colour scale units, add numbers to the black contours, and, if possible, enlarge the whole figure.[ Lines 323 ] "... occurs at a given location and in a given cluster."
[ Correction ] This given location does not refer to a random one but to a specific position around the cyclone's centre at the moment of lowest pressure. It is essential to highlight this, as this result does not apply to all situations.[ Lines 338-340 ] “In general, compound risk peaks to the west-north-west of the storm center, with a crown of (lower) risk encircling the storm center. Despite having somewhat smaller and weaker wind and rain footprints than cluster 4, cluster 2 is associated with the highest rain-wind compound risk.”
[ Question ] Considering that Mediterranean cyclones present a wide range of trajectories which often change direction during their life cycle, how generic can we consider Fig9 for cyclones at their moment of lowest centre pressure? Remember that the most remarkable precipitation areas are usually downwind of the cyclone's centre and trajectory. For example, can we expect that Fig9 adequately describes cyclones rising at the lee of the Alps and those with cyclogenesis along the Libyan coast, even if both head to the Ionian Sea? In this case, we expect the maximum precipitation surrounding the southern area of the Alps cyclone's centre and the northern area of the Libyan cyclone.[ Lines 350-351 ] “In clusters 2 and 4, the wave event footprints tend to extend to the northeast of the storm center."
[ Correction ] This information is incorrect. Cluster 2 shows a wave's probability peak westward, while Cluster 4 peaks southwest.[ Figure 9 ] Caption "The black contour identifies the statistically significant area for each hazard."
[ Correction ] Add the threshold used.[ Figure 10 ] Caption "The black contour identifies the statistically significant area for each hazard."
[ Correction ] Add the threshold used.[ Figure 11 ] Caption "The black contour identifies the statistically significant area for each hazard."
[ Correction ] Add the threshold used.[ Line 406 ] “... a heat event is co-occurring.”
[ Suggestion ] Given that heat events were defined by a fixed threshold (30°C) instead of a percentile based on local temperatures and the clusters related to the pm10-heat composition occur over North Africa, every analysed event is a heat event, making the probability of occurrence of this hazard composition almost exclusively dependent only on PM10. Maybe you should mention it.[ Line 407 ] “... Simultaneity and Overlap, with two interesting exceptions.”
[ Question ] Is there any explanation for this?Citation: https://doi.org/10.5194/egusphere-2023-2322-RC1
Raphaël Rousseau-Rizzi et al.
Raphaël Rousseau-Rizzi et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 128 | 65 | 4 | 197 | 4 | 3 |
- HTML: 128
- PDF: 65
- XML: 4
- Total: 197
- BibTeX: 4
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1