the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Identification of regions with a robust increase of heavy precipitation events
Abstract. Global climate change is increasingly associated with the increase and/or the intensification of extreme weather such as heat waves, droughts, or heavy precipitation events. However, the characteristics and severity of these changes can vary considerably by region and season. This study focuses on heavy and extreme precipitation events over Europe for the time period 1951–2099. The main objective is to identify regions which show a robust and therefore reliable change in such events with ongoing climate change. The study is based on daily precipitation values from 40 regional climate simulations of the EURO-CORDEX ensemble with a spatial resolution of 12 km (EUR-11).
Future changes were investigated using four different metrics, which are sensitive to alterations in the number and intensity of the detected events and consider an accumulated precipitation amount over a selected threshold and two return values for a 10- and a 100-year return period. Differences were detected between the climate scenarios RCP4.5 and RCP8.5, between summer and winter half-year, and between three different methods used to identify and quantify temporal changes from a reference period (1951–1980) to a future climate period (2070–2099). Furthermore, two criteria characterizing the robustness of the changes were used, i.e. a dominant agreement on the sign and a majority of significant changes within the full simulation ensemble. The analysis provided relative and absolute changes of the four metrics being used and the area fractions that exhibited a robust change.
With all methods applied, our study clearly confirms a significant increase in heavy and extreme precipitation in northern, central, and eastern Europe and a robust decrease in heavy, but not very extreme events in the southwest of the Euro-CORDEX domain associated with projected future climate change. For large parts of Southern Europe and the Mediterranean, a tendency towards decreasing intensities (down to -11 mm/year for the accumulated threshold exceedance) become visible but without the evidence of robustness.
Both, the intensity and the area with robust changes in heavy and extreme events prove to be significantly stronger in the RCP8.5 than in the RCP4.5 scenario. For Central Europe, for example, the accumulated threshold exceedance increases from 15 mm/year to 24 mm/year and the 100-year return value from 21 mm/day to 30 mm/day. The related robust area fractions extend from 31 % to 99 % and from 61 % to 95 %, respectively. The relative changes are substantial, even averaged over larger areas, with values greater than 100 % (50 %) for the severe events and up to 40 % (30 %) for the very extreme events in the RCP8.5 (RCP4.5) scenario. Heavy events increase relatively more in winter (for Central Europe +117 % for the accumulated threshold exceedance in RCP 8.5) than in summer (+81 %). The opposite is true for the extreme events with a weaker increase in winter than in summer (e.g. for Central Europe +26 % for the 100-year return value in winter and +43 % in summer).
- Preprint
(2137 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2023-552', Anonymous Referee #1, 26 Apr 2023
This study investigates the impact of climate change on heavy and extreme precipitation events over Europe for the period 1951-2099, using daily precipitation values from the EURO-CORDEX RCM ensemble with a spatial resolution of 12 km (EUR-11). While the topic of the study is timely and important, given the increasing occurrence of extreme events under climate change, previous studies have already examined the changes in extreme precipitation over Europe using EURO-CORDEX RCMs. As such, a new study in this field should address a critical knowledge gap to justify publication.
The topic investigated in this study is pertinent and crucial given the rise in extreme events caused by climate change. However, as the impacts of extreme precipitation over Europe using EURO-CORDEX RCMs have been examined in multiple previous studies, a new study in this field must address a critical knowledge gap to justify publication. While some of the previous studies were cited, others, such as a recent study by Ritzhaupt and Maraun (2022) identifying robust and conflicting projections of mean and extreme precipitation across Europe using different ensembles of climate models (ENSEMBLES and EURO-CORDEX RCMs, HighresMIP, and CMIP3, CMIP5, and CMIP6 GCMs), were not mentioned. Another study by Dosio and Fischer (2017) investigated the robustness of changes in extreme precipitation in Europe using EURO-CORDEX RCMs under different global warming levels. Given the existing knowledge in the literature, the objective of this study may be too narrow to justify publication in NHESS.
Apart from the lack of novelty, there are other issues in the paper that require explanation and/or justification:
- The authors mentioned that block maxima (BMM) is associated with high statistical uncertainty in extreme value analysis, and therefore used the peak-over-threshold (POT) method in their analysis. This was due to the short time period of their analysis (30 years). However, looking at Fig. 5 of Tabari (2021), it is clear that the difference between BMM and POT methods for projected changes in extreme precipitation in terms of magnitude and spatial distribution is very small for Europe.
- The end of the 21st century (2070-2099) was used for future climate, and the period 1951-1980 was used as the reference period. However, this reference period is not representative of either the current past or the pre-industrial era. The periods 1971-2000 or 1976-2005 are commonly used for climate change studies using EURO-CORDEX RCMs, and using these periods facilitatesa comparison of the results with previous studies.
- The authors used 26 and 14 simulations, respectively, for the RCP8.5 and RCP4.5 scenarios. Therefore, the difference in projected changes between the two scenarios may be due in part to their significantly different number of simulations.
- The authors conducted trend analysis using the Sen slope estimator and tested its significance using the Mann-Kendall test. However, the existence of auto-correlation in time series can influence the results of these methods. For example, positive auto-correlation increases the chance of rejecting the null hypothesis of no trend and vice versa. It is unclear whether the authors checked for auto-correlation in the time series and, in the case of significant auto-correlation, whether its influence was taken into account in trend analysis methods.
References
Dosio, A., & Fischer, E. M. (2018). Will half a degree make a difference? Robust projections of indices of mean and extreme climate in Europe under 1.5 C, 2 C, and 3 C global warming. Geophysical Research Letters, 45(2), 935-944.
Ritzhaupt, N., & Maraun, D. (2023). Consistency of seasonal mean and extreme precipitation projections over Europe across a range of climate model ensembles. Journal of Geophysical Research: Atmospheres, 128(1), e2022JD037845.
Citation: https://doi.org/10.5194/egusphere-2023-552-RC1 - AC1: 'Reply on RC1', Veronika Ettrichrätz, 10 Jul 2023
-
RC2: 'Comment on egusphere-2023-552', Anonymous Referee #2, 30 May 2023
Review of the manuscript:
Identification of regions with a robust increase of heavy precipitation events
The manuscript investigates future changes of different indices related to heavy and extreme daily precipitation based on an ensemble of regional climate simulations for Europe. The focus is on the robustness of changes within the ensemble. They compare their results with references from the E-OBS observation data as well as with a short-term forecast precipitation product from the ERA5 reanalysis. Afterwards they compare the climate change signal for different methods.
The topic of future changes of heavy and extreme precipitation is relevant. However, this topic has already been covered in different studies. The methods and data applied need a better justification. The results section is mostly descriptive and partly misses important or valuable information (see specifics below). I my opinion this manuscript needs major revision, before it can be deemed as a valuable contribution to the field.
Major points
There is a clear shift in the quality of argumentation between the chapters 2-4 und chapter 5. In the chapters 2-4 the information given often left me puzzled why the authors choose to do it that way (unusual choice of the reference period, reference data with clear deficiencies for the purpose, partly somewhat arbitrary metrics,…). Finally, in chapter 5 some of it is put into a context. I would suggest to give some of the context earlier at the appropriate places and consider it in the result section. This might change some of the sparsely given explanation in the result section and put the descriptions there in some context earlier on.
The four indices given lack a systematic approach and proper discussion. The authors distinguish between “heavy” and “extreme” precipitation. However, there seems to be no systematic approach to that, for instance covering the range of return periods. Furthermore, for “heavy” precipitation, they use a) linear trends and b) the difference between two time slices. The manuscript does not explore thoroughly, how the differences in methodology affect the results. Is the trend always linear, for all ensemble members? Does the trend depend on the climate sensitivity of the GCM/RCM chain? Is there an effect of long-term climate variability? A regional shift in the sign of the change over time? Going into some of these topics would make the paper more valuable.
Lines 116ff, Table 1: The EURO-CORDEX ensemble used here is more an ensemble of opportunity than a balanced one (cf. Sobolowski et al., 2021, DOI: 10.5281/zenodo.7673400). There are for instance about twice as much rcp8.5 than rcp4.5 simulation included. In addition, some GCMs and RCMs are more often represented than others. Who does this affect the robustness of the results? Does the lack of consistency in some regions arise from more the GCM or the RCM spread?
Lines 127-128: For the Mediterranean extreme precipitation season the summer to winter separation might be disadvantageous, since it often starts in September (e.g. Grazzini et al., 2019). Could shifts in the heavy precipitation period between the ensemble members affect the lack of “robustness” for that area?
Chapter 2/4.1/5.: The problems with the quality of the reference data are discussed in chapter 5. You might consider if you use some of the statements from chapter 5 for your analysis, which might exclude some areas from a comparison at least with E-Obs. Furthermore, how adequate is ERA5 as a reference? Since - as stated in chapter 5 – ERA5 precipitation is a forecast product and does not include assimilated observed precipitation directly. It would be good to explain your choice of references already in chapter 2. Furthermore, in chapter 4.2 the manuscript shows, that the reference data are outside the range of the observations over large areas. These areas partly differ for both reference data sets. E.g. or E-OBS it seems, that for Germany, where the dataset is based on many stations, the result is quite different compared to Eastern Europe, where it is based fewer station data. In addition, the comparison with KOSTRA-DWD for Germany is not that bad. It seems necessary to put these findings into a perspective. Is the ensemble not suitable or are the references not suitable for the specific requirements? Or do you deem it as not crucial?
Chapter 3.3, Line 199ff: The text is a bit confusing. Consider reformulation. Since the 100-year return values are far outside the range of the 30-year input data, confidence intervals should be given.
The authors give criteria for the non-applicability of the methods or the exclusion of certain grid points. How does this affect the “robustness” of the results in such areas?
Chapter 4 structure: Chapter 4.1 and 4.2 are mostly about evaluation, whereas chapter 4.3 is about the climate change signals. You could consider separating them.
Minor points
Line 40: the people died due the flooding caused by extreme precipitation.
Line 95: Sørland et al. (2021) is not about CPM simulations
Line 151ff: With a threshold of 3 events per 6 month, you consider a 2-monthly return period in method 1 and 2. Consider stating that to get an easier distinction between your terms “heavy” and “extreme” precipitation
Line 240: “..extreme events are lowest in SC…” Consider reformulation like e.g. “..least intense..”
Line 248: “…median s of extreme events are greater..” should be changed to “higher”
Line 316-318: Long somewhat confusing sentence. Consider reformulation.
Reference
Grazzini et al. (2019) Extreme precipitation events over northern Italy. Part I: Asystematic classification with machine-learning techniques. QJRMeteorolSoc. 2020;146:69–85, DOI: 10.1002/qj.3635
Citation: https://doi.org/10.5194/egusphere-2023-552-RC2 - AC2: 'Reply on RC2', Veronika Ettrichrätz, 10 Jul 2023
Status: closed
-
RC1: 'Comment on egusphere-2023-552', Anonymous Referee #1, 26 Apr 2023
This study investigates the impact of climate change on heavy and extreme precipitation events over Europe for the period 1951-2099, using daily precipitation values from the EURO-CORDEX RCM ensemble with a spatial resolution of 12 km (EUR-11). While the topic of the study is timely and important, given the increasing occurrence of extreme events under climate change, previous studies have already examined the changes in extreme precipitation over Europe using EURO-CORDEX RCMs. As such, a new study in this field should address a critical knowledge gap to justify publication.
The topic investigated in this study is pertinent and crucial given the rise in extreme events caused by climate change. However, as the impacts of extreme precipitation over Europe using EURO-CORDEX RCMs have been examined in multiple previous studies, a new study in this field must address a critical knowledge gap to justify publication. While some of the previous studies were cited, others, such as a recent study by Ritzhaupt and Maraun (2022) identifying robust and conflicting projections of mean and extreme precipitation across Europe using different ensembles of climate models (ENSEMBLES and EURO-CORDEX RCMs, HighresMIP, and CMIP3, CMIP5, and CMIP6 GCMs), were not mentioned. Another study by Dosio and Fischer (2017) investigated the robustness of changes in extreme precipitation in Europe using EURO-CORDEX RCMs under different global warming levels. Given the existing knowledge in the literature, the objective of this study may be too narrow to justify publication in NHESS.
Apart from the lack of novelty, there are other issues in the paper that require explanation and/or justification:
- The authors mentioned that block maxima (BMM) is associated with high statistical uncertainty in extreme value analysis, and therefore used the peak-over-threshold (POT) method in their analysis. This was due to the short time period of their analysis (30 years). However, looking at Fig. 5 of Tabari (2021), it is clear that the difference between BMM and POT methods for projected changes in extreme precipitation in terms of magnitude and spatial distribution is very small for Europe.
- The end of the 21st century (2070-2099) was used for future climate, and the period 1951-1980 was used as the reference period. However, this reference period is not representative of either the current past or the pre-industrial era. The periods 1971-2000 or 1976-2005 are commonly used for climate change studies using EURO-CORDEX RCMs, and using these periods facilitatesa comparison of the results with previous studies.
- The authors used 26 and 14 simulations, respectively, for the RCP8.5 and RCP4.5 scenarios. Therefore, the difference in projected changes between the two scenarios may be due in part to their significantly different number of simulations.
- The authors conducted trend analysis using the Sen slope estimator and tested its significance using the Mann-Kendall test. However, the existence of auto-correlation in time series can influence the results of these methods. For example, positive auto-correlation increases the chance of rejecting the null hypothesis of no trend and vice versa. It is unclear whether the authors checked for auto-correlation in the time series and, in the case of significant auto-correlation, whether its influence was taken into account in trend analysis methods.
References
Dosio, A., & Fischer, E. M. (2018). Will half a degree make a difference? Robust projections of indices of mean and extreme climate in Europe under 1.5 C, 2 C, and 3 C global warming. Geophysical Research Letters, 45(2), 935-944.
Ritzhaupt, N., & Maraun, D. (2023). Consistency of seasonal mean and extreme precipitation projections over Europe across a range of climate model ensembles. Journal of Geophysical Research: Atmospheres, 128(1), e2022JD037845.
Citation: https://doi.org/10.5194/egusphere-2023-552-RC1 - AC1: 'Reply on RC1', Veronika Ettrichrätz, 10 Jul 2023
-
RC2: 'Comment on egusphere-2023-552', Anonymous Referee #2, 30 May 2023
Review of the manuscript:
Identification of regions with a robust increase of heavy precipitation events
The manuscript investigates future changes of different indices related to heavy and extreme daily precipitation based on an ensemble of regional climate simulations for Europe. The focus is on the robustness of changes within the ensemble. They compare their results with references from the E-OBS observation data as well as with a short-term forecast precipitation product from the ERA5 reanalysis. Afterwards they compare the climate change signal for different methods.
The topic of future changes of heavy and extreme precipitation is relevant. However, this topic has already been covered in different studies. The methods and data applied need a better justification. The results section is mostly descriptive and partly misses important or valuable information (see specifics below). I my opinion this manuscript needs major revision, before it can be deemed as a valuable contribution to the field.
Major points
There is a clear shift in the quality of argumentation between the chapters 2-4 und chapter 5. In the chapters 2-4 the information given often left me puzzled why the authors choose to do it that way (unusual choice of the reference period, reference data with clear deficiencies for the purpose, partly somewhat arbitrary metrics,…). Finally, in chapter 5 some of it is put into a context. I would suggest to give some of the context earlier at the appropriate places and consider it in the result section. This might change some of the sparsely given explanation in the result section and put the descriptions there in some context earlier on.
The four indices given lack a systematic approach and proper discussion. The authors distinguish between “heavy” and “extreme” precipitation. However, there seems to be no systematic approach to that, for instance covering the range of return periods. Furthermore, for “heavy” precipitation, they use a) linear trends and b) the difference between two time slices. The manuscript does not explore thoroughly, how the differences in methodology affect the results. Is the trend always linear, for all ensemble members? Does the trend depend on the climate sensitivity of the GCM/RCM chain? Is there an effect of long-term climate variability? A regional shift in the sign of the change over time? Going into some of these topics would make the paper more valuable.
Lines 116ff, Table 1: The EURO-CORDEX ensemble used here is more an ensemble of opportunity than a balanced one (cf. Sobolowski et al., 2021, DOI: 10.5281/zenodo.7673400). There are for instance about twice as much rcp8.5 than rcp4.5 simulation included. In addition, some GCMs and RCMs are more often represented than others. Who does this affect the robustness of the results? Does the lack of consistency in some regions arise from more the GCM or the RCM spread?
Lines 127-128: For the Mediterranean extreme precipitation season the summer to winter separation might be disadvantageous, since it often starts in September (e.g. Grazzini et al., 2019). Could shifts in the heavy precipitation period between the ensemble members affect the lack of “robustness” for that area?
Chapter 2/4.1/5.: The problems with the quality of the reference data are discussed in chapter 5. You might consider if you use some of the statements from chapter 5 for your analysis, which might exclude some areas from a comparison at least with E-Obs. Furthermore, how adequate is ERA5 as a reference? Since - as stated in chapter 5 – ERA5 precipitation is a forecast product and does not include assimilated observed precipitation directly. It would be good to explain your choice of references already in chapter 2. Furthermore, in chapter 4.2 the manuscript shows, that the reference data are outside the range of the observations over large areas. These areas partly differ for both reference data sets. E.g. or E-OBS it seems, that for Germany, where the dataset is based on many stations, the result is quite different compared to Eastern Europe, where it is based fewer station data. In addition, the comparison with KOSTRA-DWD for Germany is not that bad. It seems necessary to put these findings into a perspective. Is the ensemble not suitable or are the references not suitable for the specific requirements? Or do you deem it as not crucial?
Chapter 3.3, Line 199ff: The text is a bit confusing. Consider reformulation. Since the 100-year return values are far outside the range of the 30-year input data, confidence intervals should be given.
The authors give criteria for the non-applicability of the methods or the exclusion of certain grid points. How does this affect the “robustness” of the results in such areas?
Chapter 4 structure: Chapter 4.1 and 4.2 are mostly about evaluation, whereas chapter 4.3 is about the climate change signals. You could consider separating them.
Minor points
Line 40: the people died due the flooding caused by extreme precipitation.
Line 95: Sørland et al. (2021) is not about CPM simulations
Line 151ff: With a threshold of 3 events per 6 month, you consider a 2-monthly return period in method 1 and 2. Consider stating that to get an easier distinction between your terms “heavy” and “extreme” precipitation
Line 240: “..extreme events are lowest in SC…” Consider reformulation like e.g. “..least intense..”
Line 248: “…median s of extreme events are greater..” should be changed to “higher”
Line 316-318: Long somewhat confusing sentence. Consider reformulation.
Reference
Grazzini et al. (2019) Extreme precipitation events over northern Italy. Part I: Asystematic classification with machine-learning techniques. QJRMeteorolSoc. 2020;146:69–85, DOI: 10.1002/qj.3635
Citation: https://doi.org/10.5194/egusphere-2023-552-RC2 - AC2: 'Reply on RC2', Veronika Ettrichrätz, 10 Jul 2023
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
404 | 203 | 47 | 654 | 24 | 42 |
- HTML: 404
- PDF: 203
- XML: 47
- Total: 654
- BibTeX: 24
- EndNote: 42
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1