the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Aug 2022
Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection
Abstract. Vb-cyclones propagating from the North-Western Mediterranean Sea (NWMS) into central Europe are often associated with extreme precipitation. This study explores the state and process chain linking the NWMS state and the Vb-cyclone precipitation in the Danube, Elbe, and Odra catchments in regional coupled atmosphere-ocean climate simulations with COSMO-CLM+NEMO. Two high-resolution simulations, an evaluation simulation (1951–2005) downscaling the centennial ERA-20C reanalysis and a continuous simulation (historical 1951–2005 + RCP8.5 future scenario 2006–2099) downscaling the EC-EARTH global climate data set are used for this purpose. The results show a good agreement in mean annual Vb-cyclone frequency between the evaluation (9.7 events/year) and the historical (10.1 events/year) simulations. But, there are significant discrepancies in the seasonal cycle. The mean cyclone intensity measured with minimum central pressure, track density, and precipitation rankings in the three catchments also show good agreement. The simulations for the future period show a basin-average SST warming of ≈ 2.5–3 K by the end of 21st century, but insignificant changes in Vb-cyclone frequency, mean intensity, and precipitation in the selected catchments. The NWMS sea surface temperature, evaporation, and wind speed anomalies corresponding to the Vb-cyclone precipitation rankings differ between the evaluation and historical simulations. In the evaluation simulation, Vb-cyclone precipitation rankings correspond with sea surface temperature, evaporation, and wind speed anomalies, while in the historical and the future simulation no such correspondence is seen. Especially the Adriatic and Ionian basins in the simulation driven by EC-EARTH show no sensitivity to the Vb-cyclone precipitation over the catchments. The change in the processes between evaluation and historical simulations might be due to the emergence of biases inherited from the driving EC-EARTH global simulation. The future simulation shows no significant process changes compared to the historical simulation.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(10030 KB)
-
Supplement
(4063 KB)
-
This preprint has been withdrawn.
- Preprint
(10030 KB) - Metadata XML
-
Supplement
(4063 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on esd-2022-24', Anonymous Referee #1, 22 Sep 2022
Review of «Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection»
by Praveen Kumar Pothapakula, Amelie Hoff, Anika Obermann-Hellhund, Timo Keber and Bodo Ahrens
submitted to: Earth System Dynamics (esd-2022-24)
The submission tries to assess Vb-cyclones in a centennial simulation downscaled using a regional climate model coupled to an ocean module to investigate precipitation related to Vb-cyclones over three different catchments. The vast amount of data provided by such a simulation provides certainly an interesting and important basis to obtain a better understanding of Vb-cyclones under global warming, and to deepen the understanding during such events. This is important because these events are often related to extreme precipitation and flooding events in central Europe. Nevertheless, I don’t see how this study helps to foster or even increase the process understanding of Vb-cyclones with respect to the Mediterranean Sea:
First, the method is not well enough described, so that I find it hard to understand how variables (e.g., evaporation, wind speed and sea surface temperature) over the Mediterranean Sea are linked to the precipitation over the different catchments. How long are the Vb-cyclone events typically? What period is taken to accumulate precipitation? Is it daily and a four day lasting Vb-cyclone is counted as four events? How is it assured that the precipitation over the catchment is actually linked to the Vb-cyclone and not to another large-scale feature over northern Europe? The fact that half of the cyclones result in a negative anomaly with respect to the climatology (Figure 1) indicates that no real cyclone situation is present over the different catchments for these negative anomalies. What is the argument for negative precipitation anomalies while a Vb-cyclone is present in the vicinity of the catchment?
Second, I find the conclusions drawn for the future too strong (particularly in the abstract), particularly in the light of obvious missing processes in the GCM, for which the RCM cannot compensate for (e.g., evaporation and SST values and patterns) in the Mediterranean Sea. The chosen climate simulation does not very well represent the SSTs over the Mediterranean Sea during the historical period and some other CORDEX simulations do a much better job. Why did the authors not consider including these simulations in the analysis as well? This would certainly add value as it allows to estimate the robustness of the results and would give an indication of the sensitivity of the results with respect to the Mediterranean SSTs. Another option could be to correct the SSTs in the boundary conditions of the RCM by a delta change approach. These missing processes at least need to be better discussed in the conclusions and in the abstract.
Third, the results for the future are shown in figures, but are often mentioned in only one sentence in the text. In a climate study, it is valid to evaluate the historical period against a reference and point out misrepresentations or biases, but then the focus should be on the differences between the historical and the future period. These differences allow to reveal changes in processes, which should then be discussed with respect to the bias of the model. This is widely missing in this study. In addition, the results obtained in the evaluation period are often compared to the results in Krug et al. (2021 and 2022) and it is pointed out that the results of the two studies matches (e.g., L340ff). This is not surprising as the underlying data, the method and the analysis are identical, therefore also the figures are very similar to the Krug et al. (2022) study.
Fourth, a lot of the method is based on other studies. I agree that these methods do not need to be explained in detail, but it would be very helpful for the reader if the method is summarizes the main aspects, which is still lacking, particulalry if the methods have been further adapted. I do not think that a reader wants to read several other studies to understand the underlying method of the study (e.g., cyclone tracking, track density, regressions, etc.).
Fifth, for me it is unclear how the trendlines in the Figure 1 and S1 are calculated. I would suggest using a non-parametric approach, such as the Mann-Kendall test for a monotonic trend. The trend line could even be removed and just the statistical number about the trend (slope, p-value) could be given. Since the authors have a seamless simulation that connects the historical and the future period it could also help to use the full period for the trend analysis, rather than just the two 55-year periods. One could also think to show this analysis in terms of histograms, where the evaluation, historical and future periods are overlayed into one panel. In my opinion it seems as if particularly the variability of the Vb-cyclones is changing in the future, at least in some seasons.
Sixth, I think it would be more meaningful to restrict the whole analysis to the 50-100 most intense Vb-cyclones only. This might help to obtain more meaningful results. This would further allow to cover the systems that impact the different catchment areas a bit at least.
Seventh, the language of the paper is still not very fluent, and the grammar is not very good, rendering the manuscript difficult to understand.
Finally, I would like to point out that I have reviewed this manuscript before and most of the major comments have not been adressed or implemented into the new version. Reviewing publications is quite time consuming and I think it would be respectful to the reviewers to adress or include major comments from previous submissions. To submit a less sloppy manuscript, a final critical check of the spelling, citation style (usage of brackets, L35ff meaningless citation of “vol”) and the quality of the figures (aspect ratio of maps is not correct, axis labels overlay with labels of colour bar) before submission, is highly recommended.
Some further comments are (non exhaustive):
- Why did you downscale ERA-20C reanalysis rather than ERA5, as it covers the same period, but with a much better resolution and quality?
- Did you use the ensemble mean of ERA-20C? This is not clearly statet
- To measure the intensity of a cyclone the mean sea level pressure is not the best possible variable, it would make more sense to check the Laplacian of the pressure field within the closed isobar for example.
- In Krug et al. (2022) it is pointed out that the Mediterranean is not important for Vb-cyclones unless for the most precipitation intense ones and also several other studies indicate that the Mediterranean Sea is not the most important factor (as described in the introduction). I am not sure if I understand the motivation starting at L53 and hence of the whole study.
- L86ff: is the representation of Vb-cyclones in uncoupled simulations not realistic?
- Compared to other studies, this one is not convection resolving. How does this affect the results and what is the added value?
- Method section: Information about the EC-EARTH model is completely missing. Add an extra section.
- How is precipitation associated to Vb-cyclones? (e.g., L144, L173ff, L278, L399)
Vb-cyclones are rather mesoscale features (much smaller than other cyclones) and hence their impact radius is not very large. - Cyclone tracking at mean sea level pressure is often a problem for Vb-cyclones as they are in the vicinity of Alps and can often be detected earlier in a higher level using the 900 or 850 hPa level. How sensitive are the tracks with respect to the chosen pressure level?
- I am a bit confused with the selection of the probability distribution of the transfer entropy. You have tested a lot of different distributions and then decided to use a Gaussian distribution, which does not make sense in my opinion as neither daily precipitation sums, nor anomalies are Gaussian distributed.
- L255: it is still unclear to me how the generation of a Vb-cyclone is related to the Mediterranean SSTs.
- L272ff: This might also be due to an underestimation of the climatological mean sea level pressure field in the historical simulation. As mentioned before, the minimal pressure of a cyclone is not the best estimate for its intensity as it strongly depends on its surrounding pressure field.
- Figure 3: I find the concept of the offset a bit weird. One could also use a Mann-Kendall test for the estimation of the trend.
- For Figure 4 to 11 there is always very little text and almost no information for the future. Are there significant changes between the historical and future periods?
- What is the causal relation between a high wind speed over the Mediterranean and the precipitation at the same time over the different catchments? Or is there a time lag in the analysis that I am missing? (L360ff)
- Figure 8: are there any significant changes between the historical and future period?
- Figures with patterns: The figures would look much nicer and cleaner if you would use only one colour bar for the nine panels as it is the same for all. In addition, the coordinates do not have units.
- Fig. S2: Units are missing
- Figs. S13 and S14: Aspect ratio is completely wrong.
Citation: https://doi.org/10.5194/esd-2022-24-RC1 -
RC2: 'Comment on esd-2022-24', Anonymous Referee #2, 13 Oct 2022
Review of “Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection” by Pothapakula et al.
This study aims to understand future projections of Vb cyclones that are often associated with extreme precipitation and floods in three catchments in central Europe. By analyzing coupled downscaled simulation from ERA-20C reanalysis (for evaluation) and EC-EARTH (both historical and future runs), links between potential drivers of Vb cyclones and their associated precipitation are explored in the northwest Mediterranean Sea, namely, SST, evaporation and wind speeds. While the aim of the study is obviously both scientifically important and relevant, it is unclear what the coupling adds to the already large body of literature on the topic. It is also not clear if the current modeling setup and methodological approach are able to address the research aims adequately, given the large cold biases that are introduced by the driving GCM, and the lack of more careful attention to the hypothesized mechanisms linking the different variables upstream and the resulting precipitation (that may require, e.g., time lagged analysis). I detailed below on specific issues that exemplify these concerns, which, unfortunately mean that I recommend to reject the paper from being published in ESD.
In addition, the manuscript is poorly written, with numerous grammatical errors and lack of clear storyline connecting the sentences and paragraphs. Also beyond the language itself, the text often mentions the results in a too minimal sense, without interpretation in light of the research goals. A major rewriting is required before specific suggestions can be pointed out regarding the language.
Major issues
1. The introduction is not coherent and does not provide the reader with the motivation for the work. In fact, I was left wondering what are the goals of this study? At the end of the introduction I assumed that the study aims to understand the connection between NWMS SST/evaporation and Vb cyclone occurrence and its associated precipitation in future projections. However, this goal is not specifically mentioned as one of the three aims in the end of the introduction, and the “process chain” also remains too vague at this stage. Moreover, there are many knowledge gaps that emerge from the cited literature, but the authors do not put a focus on them, so it remains unclear which gaps will be eventually addressed. For example, (i) Krug et al. (2022) found ~10 Vb cyclones per year (and stated that this is consistent with reanalysis), while Messmer et al. (2020) found less than 3 cyclones per year in both present and future simulations. (ii) the process chain relates sea state with Vb cyclones and precipitation. However, Krug et al. (2022) found that Mediterranean evaporation plays a marginal role in Vb cyclones-associated precipitation. Are any of these gaps going to be addressed by the current study? (iii) the seasonality of Vb cyclones and its changes is not clear (summer noted as the peak in line 27 and spring in line 33).
2. Methodology section and the “process chain”:
- The rationale for the focus on the NWMS region SST, evaporation and wind is not clear, given the insignificant role of this region as a moisture source found by Krug et al. (2022).
- The “process chain” is not immediately clear from the introduction and needs to be clearly explained before the method from information theory is described. Do the authors refer to effects of SST or its gradients on cyclone intensification or on thermodynamical processes feeding precipitation? Or both? Such a core issue in the manuscript must be clearly outlined and motivated, which is currently not the case.
- lines 170-174: precipitation is averaged in the different catchments during the Vb cyclone track lifetime. However, precipitation falling in the catchment during this whole time cannot be directly attributed to the cyclone, especially if the cyclone is still far. Thus, this approach needs to be justified.
- lines 198-203: it is not clear how the process chain can be studied by considering the mixing of all Vb cyclone days, i.e., without referring to the time evolution, and hence the processes in question which describe a system evolving in time. What is the meaning of omega=1 (units?)
- line 203: what is the Granger causality?
- line 204-205: what did Krug et al. (2021) find and what is the current study doing differently?
- lines 234-246: this important statement needs justification, as we know that e.g., upper tropospheric troughs and their lower tropospheric cyclonic organization typically govern evaporation on their rear side and precipitation ahead, at the same time, therefore undermining the applicability of this method.
3. Results: Vb cyclone occurrence:
- the two historical simulations demonstrated huge biases in the seasonal distribution of simulated Vb cyclones, with 41% underestimation of winter events, and 49% overestimation of summer events in the historical simulation, compared to the evaluation simulation. This remark is left without further discussion, but it strongly undermines the validity of the future simulation results. Summer Vs. winter cyclones are expected to be dynamically different in terms of governing mechanisms and associated precipitation (see, e.g., recent review Flaounas et al. 2022). The different seasonality likely affects some of the differences observed in Fig. S3, as summer cyclones may be shallower, as noted in line 272, however this can be easily checked, rather than noting that it ‘might be attributed…’
Flaounas, E., Davolio, S., Raveh-Rubin, S., Pantillon, F., Miglietta, M. M., Gaertner, M. A., ... & Ricard, D. (2022). Mediterranean cyclones: Current knowledge and open questions on dynamics, prediction, climatology and impacts. Weather and Climate Dynamics, 3(1), 173-208.
4. Results: precipitation, SST, evaporation:
- Fig. S7 and accompanying text: the systematic bias of the historical simulation is not mentioned in the text. Is it related to the bias in Vb cyclone density (Fig. S2)?
- The text related to Fig. 3 does not discuss the variation of SST with ranked precipitation, beyond the slight negative anomalies for the high ranks. The figure is left simply there, but no description or interpretation are provided.
- are there trends or large variations in the SST climatological values that is subtracted to produce the anomaly graph for each simulation period? If so, are they reflected in the anomaly results of Fig. 3? How is the different seasonality affecting the result (STD of SST likely varies with season).
- Fig. 4 and accompanying text: does the historical-evaluation difference relate to the seasonal bias? Or to cyclone location differences? These aspects can and should be checked, rather than simply stated.
- line 318: unclear which process linking negative SST anomalies and precipitation the authors refer to. What do we learn from the results shown in Fig. 5 about the process chain?
- lines 324-328: the notion contradicts the findings in Krug et al. (2022).
- Generally in this section when discussing Figs. 4, 7, 10, S8, S9, S10 it is unclear if/how the authors relate to significant biases, comparing the two historical simulations, when interpreting the future simulation? If major biases exist, also with regard to the information exchange (Figs. 5,8), what credence do we have for the future projection?
- lines 355-362: what is the underlying process chain? It is unlikely that the newly-evaporated moisture feeds the precipitation directly, as there is no time lag between the two, and these are separate airmasses. It seems more likely that both winds speeds/evaporation and precipitation are in turn affected by the intensity of the cyclone itself.
- lines 376-377: what process links wind speeds in the NWMS and precipitation in the catchments? How do the authors interpret the similar information transfer among simulations despite the differences in SST and evaporation signals, if the process assumed connects wind speeds to evaporation, which in turn reduces SST and increases precipitation?
5. conclusions:
- There is missing reference to past literature mentioned in the introduction which found different results regarding the future projections of Vb cyclones and their precipitation, and a deeper explanation as to why this may be.
- relation between NWMS variables and precipitation: again it is not discussed what can be concluded from a simulation which already does not pass the evaluation step in terms of relationship of the variables (representing some mechanistic understanding, see also major point #2). Do we trust the mostly insignificant future changes given the large biases in the historical simulation, and the large SST biases of EC-EARTH driving model?
- the relatively good agreement between the historical simulations in terms of cyclone numbers in contrast to the very large biases in terms of the variable correlations, suggests that overall they may not be critical to estimate the “right” frequency? This is not further discussed.
Minor comments
Line 31: missing reference to the 4-10 numbers
line 167-169: which box plot is it referred to?
line 175: it is confusing that the uptake region is defined only later in line 180
line 250: the trends for each period are not significant, but the authors do not mention here if there is a statistically significant difference between the historical and future simulations?
Line 253: Fig. S1 should contain also the overall mean numbers and STD, rather than showing only the interannual variability which does not contain relevant information if the trends are not significant. Are any of the seasonal-dependent trends significant? (e.g., summer in the evaluation simulation? Is a similar trend observed in reanalysis?)
Line 254-256: this statement needs supporting evidence. High SST is not known to often support the occurrence of Mediterranean cyclones.
Line 260: this is inconsistent with Messmer et al. (2020) mentioned in the introduction (line 60).
Line 402: change ‘2006’ with ‘2045’.
Figures:
Fig. S2: missing units
Fig. S4: too small titles, typo in “autumn”
Figs. 2 and S5: since the aim is to contrast simulations, I suggest to separate the panels by catchments, and plot the different simulations for each.
Fig. S6: a logarithmic scale for the y axis should make the lines more distinguishable.
Figs. 3, 6, 9: mark the +0.75C and +1.5C lines for SST in Fig. 3 and the other constants in Fig. 6, 9 for serving as a clearer reference.
Figs. 4, 7, 10, S7, S8, S9: I don’t understand these plots. If these are maps of precipitation/evaporation/SST or wind speed anomalies in the domain during Vb cyclone days (unclear if any lag considered?), then they don’t refer to the specific catchment. Is this why the 3 columns look ~identical?
Citation: https://doi.org/10.5194/esd-2022-24-RC2
Interactive discussion
Status: closed
-
RC1: 'Comment on esd-2022-24', Anonymous Referee #1, 22 Sep 2022
Review of «Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection»
by Praveen Kumar Pothapakula, Amelie Hoff, Anika Obermann-Hellhund, Timo Keber and Bodo Ahrens
submitted to: Earth System Dynamics (esd-2022-24)
The submission tries to assess Vb-cyclones in a centennial simulation downscaled using a regional climate model coupled to an ocean module to investigate precipitation related to Vb-cyclones over three different catchments. The vast amount of data provided by such a simulation provides certainly an interesting and important basis to obtain a better understanding of Vb-cyclones under global warming, and to deepen the understanding during such events. This is important because these events are often related to extreme precipitation and flooding events in central Europe. Nevertheless, I don’t see how this study helps to foster or even increase the process understanding of Vb-cyclones with respect to the Mediterranean Sea:
First, the method is not well enough described, so that I find it hard to understand how variables (e.g., evaporation, wind speed and sea surface temperature) over the Mediterranean Sea are linked to the precipitation over the different catchments. How long are the Vb-cyclone events typically? What period is taken to accumulate precipitation? Is it daily and a four day lasting Vb-cyclone is counted as four events? How is it assured that the precipitation over the catchment is actually linked to the Vb-cyclone and not to another large-scale feature over northern Europe? The fact that half of the cyclones result in a negative anomaly with respect to the climatology (Figure 1) indicates that no real cyclone situation is present over the different catchments for these negative anomalies. What is the argument for negative precipitation anomalies while a Vb-cyclone is present in the vicinity of the catchment?
Second, I find the conclusions drawn for the future too strong (particularly in the abstract), particularly in the light of obvious missing processes in the GCM, for which the RCM cannot compensate for (e.g., evaporation and SST values and patterns) in the Mediterranean Sea. The chosen climate simulation does not very well represent the SSTs over the Mediterranean Sea during the historical period and some other CORDEX simulations do a much better job. Why did the authors not consider including these simulations in the analysis as well? This would certainly add value as it allows to estimate the robustness of the results and would give an indication of the sensitivity of the results with respect to the Mediterranean SSTs. Another option could be to correct the SSTs in the boundary conditions of the RCM by a delta change approach. These missing processes at least need to be better discussed in the conclusions and in the abstract.
Third, the results for the future are shown in figures, but are often mentioned in only one sentence in the text. In a climate study, it is valid to evaluate the historical period against a reference and point out misrepresentations or biases, but then the focus should be on the differences between the historical and the future period. These differences allow to reveal changes in processes, which should then be discussed with respect to the bias of the model. This is widely missing in this study. In addition, the results obtained in the evaluation period are often compared to the results in Krug et al. (2021 and 2022) and it is pointed out that the results of the two studies matches (e.g., L340ff). This is not surprising as the underlying data, the method and the analysis are identical, therefore also the figures are very similar to the Krug et al. (2022) study.
Fourth, a lot of the method is based on other studies. I agree that these methods do not need to be explained in detail, but it would be very helpful for the reader if the method is summarizes the main aspects, which is still lacking, particulalry if the methods have been further adapted. I do not think that a reader wants to read several other studies to understand the underlying method of the study (e.g., cyclone tracking, track density, regressions, etc.).
Fifth, for me it is unclear how the trendlines in the Figure 1 and S1 are calculated. I would suggest using a non-parametric approach, such as the Mann-Kendall test for a monotonic trend. The trend line could even be removed and just the statistical number about the trend (slope, p-value) could be given. Since the authors have a seamless simulation that connects the historical and the future period it could also help to use the full period for the trend analysis, rather than just the two 55-year periods. One could also think to show this analysis in terms of histograms, where the evaluation, historical and future periods are overlayed into one panel. In my opinion it seems as if particularly the variability of the Vb-cyclones is changing in the future, at least in some seasons.
Sixth, I think it would be more meaningful to restrict the whole analysis to the 50-100 most intense Vb-cyclones only. This might help to obtain more meaningful results. This would further allow to cover the systems that impact the different catchment areas a bit at least.
Seventh, the language of the paper is still not very fluent, and the grammar is not very good, rendering the manuscript difficult to understand.
Finally, I would like to point out that I have reviewed this manuscript before and most of the major comments have not been adressed or implemented into the new version. Reviewing publications is quite time consuming and I think it would be respectful to the reviewers to adress or include major comments from previous submissions. To submit a less sloppy manuscript, a final critical check of the spelling, citation style (usage of brackets, L35ff meaningless citation of “vol”) and the quality of the figures (aspect ratio of maps is not correct, axis labels overlay with labels of colour bar) before submission, is highly recommended.
Some further comments are (non exhaustive):
- Why did you downscale ERA-20C reanalysis rather than ERA5, as it covers the same period, but with a much better resolution and quality?
- Did you use the ensemble mean of ERA-20C? This is not clearly statet
- To measure the intensity of a cyclone the mean sea level pressure is not the best possible variable, it would make more sense to check the Laplacian of the pressure field within the closed isobar for example.
- In Krug et al. (2022) it is pointed out that the Mediterranean is not important for Vb-cyclones unless for the most precipitation intense ones and also several other studies indicate that the Mediterranean Sea is not the most important factor (as described in the introduction). I am not sure if I understand the motivation starting at L53 and hence of the whole study.
- L86ff: is the representation of Vb-cyclones in uncoupled simulations not realistic?
- Compared to other studies, this one is not convection resolving. How does this affect the results and what is the added value?
- Method section: Information about the EC-EARTH model is completely missing. Add an extra section.
- How is precipitation associated to Vb-cyclones? (e.g., L144, L173ff, L278, L399)
Vb-cyclones are rather mesoscale features (much smaller than other cyclones) and hence their impact radius is not very large. - Cyclone tracking at mean sea level pressure is often a problem for Vb-cyclones as they are in the vicinity of Alps and can often be detected earlier in a higher level using the 900 or 850 hPa level. How sensitive are the tracks with respect to the chosen pressure level?
- I am a bit confused with the selection of the probability distribution of the transfer entropy. You have tested a lot of different distributions and then decided to use a Gaussian distribution, which does not make sense in my opinion as neither daily precipitation sums, nor anomalies are Gaussian distributed.
- L255: it is still unclear to me how the generation of a Vb-cyclone is related to the Mediterranean SSTs.
- L272ff: This might also be due to an underestimation of the climatological mean sea level pressure field in the historical simulation. As mentioned before, the minimal pressure of a cyclone is not the best estimate for its intensity as it strongly depends on its surrounding pressure field.
- Figure 3: I find the concept of the offset a bit weird. One could also use a Mann-Kendall test for the estimation of the trend.
- For Figure 4 to 11 there is always very little text and almost no information for the future. Are there significant changes between the historical and future periods?
- What is the causal relation between a high wind speed over the Mediterranean and the precipitation at the same time over the different catchments? Or is there a time lag in the analysis that I am missing? (L360ff)
- Figure 8: are there any significant changes between the historical and future period?
- Figures with patterns: The figures would look much nicer and cleaner if you would use only one colour bar for the nine panels as it is the same for all. In addition, the coordinates do not have units.
- Fig. S2: Units are missing
- Figs. S13 and S14: Aspect ratio is completely wrong.
Citation: https://doi.org/10.5194/esd-2022-24-RC1 -
RC2: 'Comment on esd-2022-24', Anonymous Referee #2, 13 Oct 2022
Review of “Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection” by Pothapakula et al.
This study aims to understand future projections of Vb cyclones that are often associated with extreme precipitation and floods in three catchments in central Europe. By analyzing coupled downscaled simulation from ERA-20C reanalysis (for evaluation) and EC-EARTH (both historical and future runs), links between potential drivers of Vb cyclones and their associated precipitation are explored in the northwest Mediterranean Sea, namely, SST, evaporation and wind speeds. While the aim of the study is obviously both scientifically important and relevant, it is unclear what the coupling adds to the already large body of literature on the topic. It is also not clear if the current modeling setup and methodological approach are able to address the research aims adequately, given the large cold biases that are introduced by the driving GCM, and the lack of more careful attention to the hypothesized mechanisms linking the different variables upstream and the resulting precipitation (that may require, e.g., time lagged analysis). I detailed below on specific issues that exemplify these concerns, which, unfortunately mean that I recommend to reject the paper from being published in ESD.
In addition, the manuscript is poorly written, with numerous grammatical errors and lack of clear storyline connecting the sentences and paragraphs. Also beyond the language itself, the text often mentions the results in a too minimal sense, without interpretation in light of the research goals. A major rewriting is required before specific suggestions can be pointed out regarding the language.
Major issues
1. The introduction is not coherent and does not provide the reader with the motivation for the work. In fact, I was left wondering what are the goals of this study? At the end of the introduction I assumed that the study aims to understand the connection between NWMS SST/evaporation and Vb cyclone occurrence and its associated precipitation in future projections. However, this goal is not specifically mentioned as one of the three aims in the end of the introduction, and the “process chain” also remains too vague at this stage. Moreover, there are many knowledge gaps that emerge from the cited literature, but the authors do not put a focus on them, so it remains unclear which gaps will be eventually addressed. For example, (i) Krug et al. (2022) found ~10 Vb cyclones per year (and stated that this is consistent with reanalysis), while Messmer et al. (2020) found less than 3 cyclones per year in both present and future simulations. (ii) the process chain relates sea state with Vb cyclones and precipitation. However, Krug et al. (2022) found that Mediterranean evaporation plays a marginal role in Vb cyclones-associated precipitation. Are any of these gaps going to be addressed by the current study? (iii) the seasonality of Vb cyclones and its changes is not clear (summer noted as the peak in line 27 and spring in line 33).
2. Methodology section and the “process chain”:
- The rationale for the focus on the NWMS region SST, evaporation and wind is not clear, given the insignificant role of this region as a moisture source found by Krug et al. (2022).
- The “process chain” is not immediately clear from the introduction and needs to be clearly explained before the method from information theory is described. Do the authors refer to effects of SST or its gradients on cyclone intensification or on thermodynamical processes feeding precipitation? Or both? Such a core issue in the manuscript must be clearly outlined and motivated, which is currently not the case.
- lines 170-174: precipitation is averaged in the different catchments during the Vb cyclone track lifetime. However, precipitation falling in the catchment during this whole time cannot be directly attributed to the cyclone, especially if the cyclone is still far. Thus, this approach needs to be justified.
- lines 198-203: it is not clear how the process chain can be studied by considering the mixing of all Vb cyclone days, i.e., without referring to the time evolution, and hence the processes in question which describe a system evolving in time. What is the meaning of omega=1 (units?)
- line 203: what is the Granger causality?
- line 204-205: what did Krug et al. (2021) find and what is the current study doing differently?
- lines 234-246: this important statement needs justification, as we know that e.g., upper tropospheric troughs and their lower tropospheric cyclonic organization typically govern evaporation on their rear side and precipitation ahead, at the same time, therefore undermining the applicability of this method.
3. Results: Vb cyclone occurrence:
- the two historical simulations demonstrated huge biases in the seasonal distribution of simulated Vb cyclones, with 41% underestimation of winter events, and 49% overestimation of summer events in the historical simulation, compared to the evaluation simulation. This remark is left without further discussion, but it strongly undermines the validity of the future simulation results. Summer Vs. winter cyclones are expected to be dynamically different in terms of governing mechanisms and associated precipitation (see, e.g., recent review Flaounas et al. 2022). The different seasonality likely affects some of the differences observed in Fig. S3, as summer cyclones may be shallower, as noted in line 272, however this can be easily checked, rather than noting that it ‘might be attributed…’
Flaounas, E., Davolio, S., Raveh-Rubin, S., Pantillon, F., Miglietta, M. M., Gaertner, M. A., ... & Ricard, D. (2022). Mediterranean cyclones: Current knowledge and open questions on dynamics, prediction, climatology and impacts. Weather and Climate Dynamics, 3(1), 173-208.
4. Results: precipitation, SST, evaporation:
- Fig. S7 and accompanying text: the systematic bias of the historical simulation is not mentioned in the text. Is it related to the bias in Vb cyclone density (Fig. S2)?
- The text related to Fig. 3 does not discuss the variation of SST with ranked precipitation, beyond the slight negative anomalies for the high ranks. The figure is left simply there, but no description or interpretation are provided.
- are there trends or large variations in the SST climatological values that is subtracted to produce the anomaly graph for each simulation period? If so, are they reflected in the anomaly results of Fig. 3? How is the different seasonality affecting the result (STD of SST likely varies with season).
- Fig. 4 and accompanying text: does the historical-evaluation difference relate to the seasonal bias? Or to cyclone location differences? These aspects can and should be checked, rather than simply stated.
- line 318: unclear which process linking negative SST anomalies and precipitation the authors refer to. What do we learn from the results shown in Fig. 5 about the process chain?
- lines 324-328: the notion contradicts the findings in Krug et al. (2022).
- Generally in this section when discussing Figs. 4, 7, 10, S8, S9, S10 it is unclear if/how the authors relate to significant biases, comparing the two historical simulations, when interpreting the future simulation? If major biases exist, also with regard to the information exchange (Figs. 5,8), what credence do we have for the future projection?
- lines 355-362: what is the underlying process chain? It is unlikely that the newly-evaporated moisture feeds the precipitation directly, as there is no time lag between the two, and these are separate airmasses. It seems more likely that both winds speeds/evaporation and precipitation are in turn affected by the intensity of the cyclone itself.
- lines 376-377: what process links wind speeds in the NWMS and precipitation in the catchments? How do the authors interpret the similar information transfer among simulations despite the differences in SST and evaporation signals, if the process assumed connects wind speeds to evaporation, which in turn reduces SST and increases precipitation?
5. conclusions:
- There is missing reference to past literature mentioned in the introduction which found different results regarding the future projections of Vb cyclones and their precipitation, and a deeper explanation as to why this may be.
- relation between NWMS variables and precipitation: again it is not discussed what can be concluded from a simulation which already does not pass the evaluation step in terms of relationship of the variables (representing some mechanistic understanding, see also major point #2). Do we trust the mostly insignificant future changes given the large biases in the historical simulation, and the large SST biases of EC-EARTH driving model?
- the relatively good agreement between the historical simulations in terms of cyclone numbers in contrast to the very large biases in terms of the variable correlations, suggests that overall they may not be critical to estimate the “right” frequency? This is not further discussed.
Minor comments
Line 31: missing reference to the 4-10 numbers
line 167-169: which box plot is it referred to?
line 175: it is confusing that the uptake region is defined only later in line 180
line 250: the trends for each period are not significant, but the authors do not mention here if there is a statistically significant difference between the historical and future simulations?
Line 253: Fig. S1 should contain also the overall mean numbers and STD, rather than showing only the interannual variability which does not contain relevant information if the trends are not significant. Are any of the seasonal-dependent trends significant? (e.g., summer in the evaluation simulation? Is a similar trend observed in reanalysis?)
Line 254-256: this statement needs supporting evidence. High SST is not known to often support the occurrence of Mediterranean cyclones.
Line 260: this is inconsistent with Messmer et al. (2020) mentioned in the introduction (line 60).
Line 402: change ‘2006’ with ‘2045’.
Figures:
Fig. S2: missing units
Fig. S4: too small titles, typo in “autumn”
Figs. 2 and S5: since the aim is to contrast simulations, I suggest to separate the panels by catchments, and plot the different simulations for each.
Fig. S6: a logarithmic scale for the y axis should make the lines more distinguishable.
Figs. 3, 6, 9: mark the +0.75C and +1.5C lines for SST in Fig. 3 and the other constants in Fig. 6, 9 for serving as a clearer reference.
Figs. 4, 7, 10, S7, S8, S9: I don’t understand these plots. If these are maps of precipitation/evaporation/SST or wind speed anomalies in the domain during Vb cyclone days (unclear if any lag considered?), then they don’t refer to the specific catchment. Is this why the 3 columns look ~identical?
Citation: https://doi.org/10.5194/esd-2022-24-RC2
Data sets
Vb-cyclones and associated North-Western Mediterranean Sea state in regional coupled climate simulations: evaluation and projection Praven Kumar Pothapakula https://doi.org/10.5281/zenodo.6585342
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 635 | 163 | 43 | 841 | 73 | 47 | 40 |
- HTML: 635
- PDF: 163
- XML: 43
- Total: 841
- Supplement: 73
- BibTeX: 47
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Praveen Kumar Pothapakula
Amelie Hoff
Anika Obermann-Hellhund
Timo Keber
Bodo Ahrens
This preprint has been withdrawn.
- Preprint
(10030 KB) - Metadata XML
-
Supplement
(4063 KB) - BibTeX
- EndNote