the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Mar 2025
Extreme precipitation and flooding in Berlin under climate change and effects of selected grey and blue-green measures
Abstract. This paper aims to quantify potential changes in extreme precipitation under climate change scenarios in the city of Berlin, Germany, and their resulting impacts on urban flooding in a selected flood-prone area of the city. Furthermore, it investigates the effectiveness of the existing drainage system, infiltration from unsealed surfaces, and retention roofs during extreme rainfall events under both current and future climate conditions.
The effect of climate change on the statistical distribution of extreme precipitation in Berlin is assessed by analyzing a single-model set of climate scenario simulations at convection permitting resolution (COSMO-CLM). Three 30-year periods are simulated: The historical period under observed greenhouse gas concentrations from 1971 to 2000 and two RCP8.5 scenario periods from 2031 to 2060 and from 2071 to 2100. For the historical period, the estimated 1-hour rainfall sum for a 100-year return level (referred to as ‘Historical 100a’) agrees well with the statistical values from station observations. For the period 2031–2060 under RCP8.5 conditions the respective rainfall sum of the 1-hour 100-year event (referred to as ‘Future 100a’) increases by 46 % and the strongest hourly intensity in all three simulated 30-year periods (referred to as ‘Strongest’) is increased by 123 % compared to the Historical 100a event.
The impacts of these increases in extreme precipitation on flooding characteristics in a Central-Berlin region around the Gleimtunnel, which is known for frequent pluvial flooding, are studied by conducting simulations with the 2D surface flow model hms++ coupled to a 1D drainage model. The Future 100a event result in a 51 % increase in the simulated maximum water depth, a 43 % increase in maximum surface runoff at the local flooding hotspot Gleimtunnel, and a 33 % increase in the volume of combined sewer overflow. For the Strongest event, the respective increases are 137 % (maximum water depth), 296 % (maximum surface runoff), and 74 % (combined sewer overflow).
The effects of the existing drainage system and infiltration under different rainfall scenarios are highlighted by comparing simulation results with and without their consideration. Neglecting the drainage system results in a 170 % increase for the Historical 100a event and a 110 % increase for the Strongest event, compared to the reference simulations. While the drainage system strongly reduces flooding, especially at hotspots, it cannot fully prevent severe flooding, and its effectiveness decreases with higher rainfall intensity. Studying infiltration reflects potential impacts of surface sealing or, conversely, desealing as a climate adaptation strategy. Neglecting infiltration increases the maximum water depth at the Gleimtunnel by 33 % for the Historical 100a event and 18 % for the Strongest event compared to reference simulations. Infiltration significantly reduces flooding, though its effectiveness decreases with higher rainfall intensity.
As a potential adaptation strategy, the impact of replacing all roofs with retention roofs is examined. For this best-case adaptation scenario, the maximum water depth at the local hotspot is reduced by 22–24 %, and the volume of combined sewer overflow by 15–20 % in the different scenarios. Since full retention on all roof surfaces is considered for all rainfall scenarios, the effects are almost the same. Remarkably, the retention roofs significantly reduced the maximum surface runoff in the Gleimtunnel during the Strongest event to below the stability threshold for pedestrians, which was clearly exceeded in the simulation without retention roofs.
The results of this study highlight the potential local impacts of ongoing global warming in terms of heavy rainfall and urban flooding in the city of Berlin and emphasize the need to combine grey infrastructure, retention roofs, and other blue-green measures.
Competing interests: U. Ulbrich is an editor for NHESS and K. Nissen a guest editor for the special issue.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(20500 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-445', Anonymous Referee #1, 26 Mar 2025
Review of Manuscript: “Extreme Precipitation and Flooding in Berlin under Climate Change and Effects of Selected Grey and Blue-Green Measures”
This manuscript addresses a timely and important topic, urban flooding under climate change, and evaluates the effectiveness of grey and blue-green adaptation measures for surface water flood risk management in Berlin. The study is well-structured and methodologically sound, incorporating climate projections, hydrodynamic modelling, and scenario-based analysis. The findings provide valuable insights for urban flood risk management and adaptation strategies, making the study relevant to a wide range of audience.
While the manuscript is well organised and written, and presents meaningful contributions, there are still areas that require further improvements before publication, outlined as follows:
1. Uncertainty Quantification:
- The manuscript should provide a more detailed discussion of uncertainties associated with climate projections, model assumptions, and parameter sensitivity (e.g. sensitivity of flood simulation results to infiltration parameters/settings). Addressing these uncertainties will enhance the reliability of the findings.
- Sensitivity analysis or confidence intervals could be included to better quantify the robustness of results.
2. Generalizability of Findings:
- Since the study focuses on Berlin, the authors should discuss how the findings could be applied to other cities with similar or different urban hydrology characteristics.
- A brief comparison with existing studies from other regions would strengthen the broader applicability of the results.
3. Data and Model Transparency:
- More details on the datasets, model calibration, and validation processes should be provided to ensure reproducibility.
- How green roofs are represented/simulated in the model?
Minor issues:
- Figures and Tables: Some figures and tables could be improved in clarity – consistent font size, format and clear legends should be used for all figures/tables.
- The manuscript should be carefully proofread for minor grammatical and typo errors.
- “The computation time increased by a factor of 3.3 from 2.4 hours (5 m resolution) to 7.8 hours (2 m resolution) while using 48 CPU cores (Intel Cascade Lake Platinum 9242) for both simulations.” Did the hydrodynamic model simulations adopt adaptive timesteps? If yes, should be increasing factor of computational time from 2m to 5m resolution simulations ~10? When then resolution is doubled, the number of cells for calculation should be 4 times more and the time steps would be halved, leading to ~8 times of increase in runtime. So, increasing the resolution from 5m to 2m, the runtime should increase ~10 times?
Recommendation: Based on the above comments, I recommend moderate revisions before acceptance. Addressing the issues outlined above will significantly enhance the manuscript’s clarity and quality. The study has strong potential for publication, provided the authors refine their discussion of uncertainties and generalisability.
Citation: https://doi.org/10.5194/egusphere-2025-445-RC1 -
AC1: 'Reply on RC1', Franziska Tügel, 04 Jul 2025
We would like to thank the anonymous reviewer for her/his positive evaluation and helpful comments. In the following we list the remarks of the review and add our answers in bold.
Review of Manuscript: “Extreme Precipitation and Flooding in Berlin under Climate Change and Effects of Selected Grey and Blue-Green Measures”
This manuscript addresses a timely and important topic, urban flooding under climate change, and evaluates the effectiveness of grey and blue-green adaptation measures for surface water flood risk management in Berlin. The study is well-structured and methodologically sound, incorporating climate projections, hydrodynamic modelling, and scenario-based analysis. The findings provide valuable insights for urban flood risk management and adaptation strategies, making the study relevant to a wide range of audience.
While the manuscript is well organised and written, and presents meaningful contributions, there are still areas that require further improvements before publication, outlined as follows:
- Uncertainty Quantification:
The manuscript should provide a more detailed discussion of uncertainties associated with climate projections, model assumptions, and parameter sensitivity (e.g. sensitivity of flood simulation results to infiltration parameters/settings). Addressing these uncertainties will enhance the reliability of the findings.
We suggest to extend the conclusions:
“The study is associated with a number of uncertainties. For the extreme rainfall analysis, these include the choice of the climate scenario and the fact that only one global-regional model combination was available for the study. This limitation affects the estimation of the strength of the precipitation change, that is at the upper range of changes reported by Hundhausen et al. (2024) for the area average in southern Germany. The fact that extreme precipitation intensifies under global warming conditions can be regarded as undisputed and has been recognized by a large number of studies (IPCC, 2021; Fowler et al., 2021). Regarding the model assumptions, it has been taken care to include all relevant processes and represent them based on given data and knowledge. This includes a fully dynamic 2D robust shallow water model coupled to a 1D subsurface drainage model provided by the city’s water company, and spatially distributed friction and infiltration parameter depending on the land use, soils, and degree of sealing. Regarding the sensitivity to infiltration, one simulation completely neglecting infiltration is compared to one with infiltration based on the knowledge about the actual state of the study area. Simplifications in the model setup, such as no direct drainage from roof surfaces into the subsurface drainage system, and infiltration parameters based on block-level data, come along with some uncertainties.”
Sensitivity analysis or confidence intervals could be included to better quantify the robustness of results.
An explanation on how the sampling uncertainties in the extreme precipitation a analysis were derived will be added: “The sampling uncertainty is determined using a bootstrap approach that randomly draws years with replacement.”. Also, we will carry out one or two more infiltration simulations to further study the sensitivity of simulation results to different infiltration settings in more detail.
- Generalizability of Findings:
Since the study focuses on Berlin, the authors should discuss how the findings could be applied to other cities with similar or different urban hydrology characteristics.
The findings generally apply also to other cities with similar hydrological characteristics. The increase in extreme rainfall intensities has been already observed in many regions of the world, and according to climate projections significant increases in the maximum daily rainfall intensity are expected for almost all land surfaces, while the percentages of increase depend on the emission scenario (IPCC, 2021). The effects of infiltration and the drainage system will also be similar depending on the design of the subsurface drainage system, given soils, and surface sealing conditions. The approach can be applied to any other city, while often the availability of high-quality and high-resolution data will be a limiting factor. Particularly, data or a model of the subsurface drainage system might often not be available. Reasonable drainage models could be generated by using AI and available data, e.g. on topography, roads, and sealed surfaces (Döring & Neuweiler, 2019). High-resolution DEMs based on LiDAR data are getting more often available.
A brief comparison with existing studies from other regions would strengthen the broader applicability of the results.
A brief comparison with more existing studies from other regions (e.g. China, Japan) will be carried out.
- Data and Model Transparency:
More details on the datasets, model calibration, and validation processes should be provided to ensure reproducibility.
A table with all used datasets will be prepared for a better overview. Due to limitations in observed data, a model calibration could not be carried out. The model is strongly based on physical relationships and available data for estimating model parameters. Model validation of the surface flow model was carried out by comparing results for other parts of the city with official flood maps from the municipality (Geoportal, 2025) as well as with simulations from other projects on urban flooding in Berlin (AMAREX, InnoMAUS, see references below). For the coupled model, plausibility checks have been carried out for selected heavy rain events comparing simulated time series of the water depth in a retention basin with observed data.
How green roofs are represented/simulated in the model?
Retention roofs are simulated by setting the rainfall in the input raster for all roof surfaces to zero. A full retention of all rainwater falling on roof surfaces is assumed. This will be clarified in the methodology.
Minor issues:
Figures and Tables: Some figures and tables could be improved in clarity – consistent font size, format and clear legends should be used for all figures/tables.
The authors will improve the figures and tables with regard to consistent font size, format, and legends.
The manuscript should be carefully proofread for minor grammatical and typo errors. The authors will carefully proofread the final revised manuscript.
“The computation time increased by a factor of 3.3 from 2.4 hours (5 m resolution) to 7.8 hours (2 m resolution) while using 48 CPU cores (Intel Cascade Lake Platinum 9242) for both simulations.” Did the hydrodynamic model simulations adopt adaptive timesteps? If yes, should be increasing factor of computational time from 2m to 5m resolution simulations ~10? When then resolution is doubled, the number of cells for calculation should be 4 times more and the time steps would be halved, leading to ~8 times of increase in runtime. So, increasing the resolution from 5m to 2m, the runtime should increase ~10 times?
Yes, in theory and with a linear solver behaviour, the runtime should increase by about 10 times. For larger simulations, we would also generally expect a stronger increase in the runtime, but not necessarily for smaller ones: If the system is not yet fully utilized anyway, then it scales even more favourably for the time being. In addition, the implementation of the solver as well as possible effects of the higher resolution on local flow characteristics affects the runtime. It is possible that the higher resolution has reduced/eliminated a numerical speed hotspot and therefore the time step has not decreased as much. Also large parts of dry areas during long periods of the simulation, which are not contributing to additional calculations, can contribute to a smaller increase in runtime. In our solver, the query for the vectorization (acceleration) of the calculation is grouped over a certain number of cells, i.e. it is not performed for individual cells. The higher the resolution, the more cells fit between two wet areas, so this may well lead to more skipped calculations.
Recommendation: Based on the above comments, I recommend moderate revisions before acceptance. Addressing the issues outlined above will significantly enhance the manuscript’s clarity and quality. The study has strong potential for publication, provided the authors refine their discussion of uncertainties and generalisability.
References
Döring, A., Neuweiler, I. (2019). Generation of Stormwater Drainage Networks Using Spatial Data. In: Mannina, G. (eds) New Trends in Urban Drainage Modelling. UDM 2018. Green Energy and Technology. Springer, Cham. https://doi.org/10.1007/978-3-319-99867-1_99Fowler, H.J., Lenderink, G., Prein, A.F. et al. (2021). Anthropogenic intensification of short-duration rainfall extremes. Nat Rev Earth Environ 2, 107–122. https://doi.org/10.1038/s43017-020-00128-6
Hundhausen, M., Feldmann, H., Kohlhepp, R., and Pinto, J. G. (2024). Climate change signals of extreme precipitation return levels for Germany in a transient convection-permitting simulation ensemble, International Journal of Climatology, 44, 1454 – 1471, https://api.semanticscholar.org/CorpusID:270475096
IPCC (2021). Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change[Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, In press. https://doi.org/10.1017/9781009157896
Websites:
Geoportal (2025). Hinweiskarte Starkregengefahren, URL: https://www.geoportal.de/map.html?map=tk_04-hinweiskarte-starkregengefahren-be-bb (last access: 04 July 2025)
Inno_MAUS project, University of Potsdam, URL: https://www.uni-potsdam.de/de/inno-maus/ (last access: 4 July 2025)
AMAREX project, RPTU Kaiserslautern-Landau, URL: https://www.amarex-projekt.de/de (last access: 4 July 2025)
Citation: https://doi.org/10.5194/egusphere-2025-445-AC1
-
RC2: 'Comment on egusphere-2025-445', Anonymous Referee #2, 23 May 2025
Review of “Extreme Precipitation and Flooding in Berlin under Climate Change and Effects of Selected Grey and Blue-Green Measures” by Franziska Tügel et al.
The manuscript presents a study on urban flooding and the influence of climate change on heavy precipitation and the consequent urban flooding. Moreover, it evaluates factors and measures to reduce flooding, like the effect of infiltration on existing unsealed areas on flooding, the actual effect of the sewer system, which is often neglected in the derivation of urban flood hazard maps, and the effect of storage roofs.
The manuscript is well written and structured, and provides useful insights into potential adaptation measures. However, there are a few issues to be dealt with to improve the manuscript even further:
- The scenario analyzing the effect of infiltration is only "retrospective/negative", means that only the effect of infiltration on flooding from existing unsealed areas is considered. While this is a useful insight into the dimension of flood reduction by the existing unsealed areas, it does not show the potential of further de-sealing for flood reduction. This would be an important information to the city authorities within the frame of the flood management and climate change adaptation, because de-sealing is an essential element of the sponge city concept. Thus, I suggest to include also a scenario with additional infiltration by de-sealing where it is likely possible in reality (e.g. parking lots, pedestrian areas etc.). This would be much more helpful for the city and the wider audience, because it shows the potential of what can be achieved with additional de-sealing, in addition to showing that the current infiltration is effective in reducing urban flooding. Particularly because the other option of increaed drainage, i.e. increasing the sub-surface drainage capacity (sewer system), is not that easily achievable in practice, and very costly.
- The conclusions are very short. Please elaborate more on recommendations for flood proofing the city based on your results, and also on the transferability to other cities. This can be done on general reasoning and literature.
- I see the way the flood volume is calculated problematic. What you actually calculate is the sum of maximum water depths, which is definitively not an estimation of the actual flood volume. Sure, you can show how this reduced by the individual measures, but this is not showing the reduction in flood volume and thus the percentages you provide have limited meaning. I recommend to use the sum of the modelled surface water at the end of the rain event. This is a much more realistic estimation of the actual surface flood volume, that considers the drainage by infiltration and the sewer system.
- I cannot really follow how the sub-hourly rainfall intensities are derived (section 4.3 and Fig. 5). Please revise this section to make this clearer. There are more comments/questions in the annotated pdf.
- As there was only one combination of global-regional climate model available, it would be very useful to provide an assessment of how representative the result on the increase of rainfall intensities in future is/can be expected in relation to a full ensemble of climate projection by different models. Particularly the finding that the far-future projection gives reduced rainfall intensities compared to the near future is unexpected, as you also state. This can be done by literature references. If it can be shown that this result is in line with other studies/projections, this would substantiate your findings and conclusions considerably.
Next to this I made some additional minor comments or raised questions in the annotated pdf of the manuscript.
-
AC2: 'Reply on RC2', Franziska Tügel, 04 Jul 2025
We would like to thank the anonymous reviewer for her/his positive evaluation and helpful comments. In the following we list the remarks of the review and add our answers in bold.
The manuscript presents a study on urban flooding and the influence of climate change on heavy precipitation and the consequent urban flooding. Moreover, it evaluates factors and measures to reduce flooding, like the effect of infiltration on existing unsealed areas on flooding, the actual effect of the sewer system, which is often neglected in the derivation of urban flood hazard maps, and the effect of storage roofs.
The manuscript is well written and structured, and provides useful insights into potential adaptation measures. However, there are a few issues to be dealt with to improve the manuscript even further:
The scenario analyzing the effect of infiltration is only "retrospective/negative", means that only the effect of infiltration on flooding from existing unsealed areas is considered. While this is a useful insight into the dimension of flood reduction by the existing unsealed areas, it does not show the potential of further de-sealing for flood reduction. This would be an important information to the city authorities within the frame of the flood management and climate change adaptation, because de-sealing is an essential element of the sponge city concept. Thus, I suggest to include also a scenario with additional infiltration by de-sealing where it is likely possible in reality (e.g. parking lots, pedestrian areas etc.). This would be much more helpful for the city and the wider audience, because it shows the potential of what can be achieved with additional de-sealing, in addition to showing that the current infiltration is effective in reducing urban flooding. Particularly because the other option of increased drainage, i.e. increasing the sub-surface drainage capacity (sewer system), is not that easily achievable in practice, and very costly.
We decided to compare a simulation with infiltration settings representing the actual state to the one of complete sealing, which already provides useful information on the effect of infiltration. Constructing a realistic de-sealing scenario is beyond the scope of this study, but we will add one or two more simulations with different infiltration settings. For future work, effects of realistic de-sealing scenarios would be very interesting, especially when conducted by the local authorities that have better data on potential areas for de-sealing measures.
The conclusions are very short. Please elaborate more on recommendations for flood proofing the city based on your results, and also on the transferability to other cities. This can be done on general reasoning and literature.
The conclusions will be extended with regard to recommendations for flood proofing and the transferability to other cities.
I see the way the flood volume is calculated problematic. What you actually calculate is the sum of maximum water depths, which is definitively not an estimation of the actual flood volume. Sure, you can show how this reduced by the individual measures, but this is not showing the reduction in flood volume and thus the percentages you provide have limited meaning. I recommend to use the sum of the modelled surface water at the end of the rain event. This is a much more realistic estimation of the actual surface flood volume, that considers the drainage by infiltration and the sewer system.
We thank you for raising this point. We used the sum of maximum water depths to compare the effects of different settings on the worst-case situation (max. water depth) in the entire domain by one single value. But to call this “flood volume” is indeed misleading. We included now a comparison of flood volumes after 30 minutes (time of the peak water depths in the Gleimtunnel), and 60 minutes (end of rainfall event, as suggested by the reviewer in another comment). Possibly, we will also add the values at the end of the simulation time after 120 minutes.
I cannot really follow how the sub-hourly rainfall intensities are derived (section 4.3 and Fig. 5). Please revise this section to make this clearer.
We suggest to rephrase the paragraph (lines 221-236) to:
“Since the 100-year events for the near and far future simulations were very similar and the near future event even a bit stronger, only the 100-year event of the near future period was chosen for the hydrodynamic simulations. For the construction of the Euler-2 design rainfall, intensities for duration classes of less than one hour are needed. As already explained in Sect. 4.1, the precipitation output from the climate simulations is one hour and extrapolating sub-hourly intensities from the modelled probability distribution leads to strong overestimations of sub-hourly intensities compared to values from KOSTRA. To overcome this problem, a Euler-2 time series is constructed using the statistical hourly and sub-hourly 100-year return values from KOSTRA-DWD-2020. For constructing the Euler-2 design, KOSTRA values for a 100-year return period and durations of 5, 10. 15, 20, 30, 45, and 60 minutes are used, while the overall rainfall sum equals the sum of the 60 min event. In the Euler-2 design storm, the peak intensity equals the intensity of a 5-minute duration and occurs between 15 and 20 minutes. In this study, each 5-minute time step generated by rainfall intensities from KOSTRA is then scaled by a factor, calculated as the ratio between the 1-hour rainfall sum of the respective event (as listed above) and the 1-hour rainfall sum of the 100-year event from KOSTRA-DWD-2020 (48.5 mm). The resulting scaling factors are 0.98 (47.7/48.5) for the Historical 100a event, 1.44 (69.8/48.5) for the Future 100a event, and 2.20 (106.7/48.5) for the Strongest event. For the Strongest event, the maximum intensity calculated with this approach is 45 mm in 5 min (almost 9 mm/min). This is considered to be unrealistically high as the maximum 5-min rainfall measured at the station in Berlin-Dahlem during the period 1979 to 2023 was 12.3 mm. For an additional comparison with less extreme maximum intensities, all selected events were also simulated with a constant rainfall intensity. This also allows to analyse the influence of the temporal rainfall distribution. Fig. 5 shows the Euler-2 and constant rain events for the three aforementioned rainfall sums.”
There are more comments/questions in the annotated pdf.
As there was only one combination of global-regional climate model available, it would be very useful to provide an assessment of how representative the result on the increase of rainfall intensities in future is/can be expected in relation to a full ensemble of climate projection by different models. Particularly the finding that the far-future projection gives reduced rainfall intensities compared to the near future is unexpected, as you also state. This can be done by literature references. If it can be shown that this result is in line with other studies/projections, this would substantiate your findings and conclusions considerably.
Please find our answers and suggestions for a revision below in points L 205 and L 488.
Next to this I made some additional minor comments or raised questions in the annotated pdf of the manuscript.
Further comments from the annotated pdf:
L 50: Some details on this would be helpful here. A figure showing the Euler-2 sub-hourly rainfall distribution would also be helpful. We suggest to alter the description as follows: “The Euler-2 evolution is characterized by gradually increasing intensities, reaching the 5-minute maximum after 20 minutes, and a subsequent strong decrease down to a small residual amount (as illustrated in Sect. 4.3). The time series is constructed using, statistical hourly and sub-hourly precipitation intensities from the coordinated heavy precipitation regionalization and evaluation (KOSTRA, 2020) of the German Weather Service (DWD; see Sect. 2)”
L. 154: surface or sewer drainage? Sewer drainage catchment; will be clarified in the description.
L. 166: I am missing a description of how the buildings are treated in both models. Are they excluded from hms++ model, or considered as building blocks in the DEM? They are considered with increased elevation in the DEM. We will describe this clearly in the model setup.
What happens to rainfall falling on roofs? Is this added to the sub-surface drainage and simulated as such in SWMM? No, it is added directly to the surface as the buildings are represented just by increasing the elevation. This has been only mentioned in the discussion, but will now be explained in the model setup.
Information on this is essential for interpreting the no-subsurface-drainage and the retention roof scenarios.
L.173: for the study area or whole Berlin? For the study area; clarification will be added.
L. 191: where is this shown? We will add the reference to Fig. 3a
Fig. 3 Please include also the KOSTRA area-averaged estimations of the 100-year event for other durations. This would give a more encompssing view of the realiability of the historic simulations vs. the observation-based KOSTRA data; it should also show then the described deviation of the d-GEV fitting from KOSTRA for the durations < 1 h
We will add further KOSTRA values and change the y axis label for (b) as suggested to “Intensity Change”.
L. 198: +/- 1 Kelvin for a climatological mean temperature is quite a lot. Can you comment on this? Has these implications on the results and conclusions?
For the hydrodynamic simulations conducted in our study temperature does not play a role. As the extreme precipitation statistics agree well with observations we are confident that shortcomings of other variables will have no significant implications for the results and conclusions. For other impact studies relying on realistic temperatures a bias adjustment would be needed.
L. 205: can this be physically explained? or is it just a random result? would different slightly different realizations of CCLM-CPS or MIROC5 (nudging with different starting points or varying parameter sets, etc.) give the same or different results?
Answering this question could get lengthy, I know, but the presented result is counter-intuitive, on the first glance at least. Thus a few sentences to help interpreting/judging the findings would be very helpful. Citing other studies with the same or similar results might help here.
Or is this an artefact of the d-GEV fitting?
This is a random result and caused by the strong variability in the simulations (variability of extremes is predicted to increase under climate change conditions). The result demonstrates that 30 years are a rather short period for a robust GEV analysis aimed at estimating 100-year return periods. The variability can be seen e.g. in Fig. 7 of Hundhausen et al. (2024). Hundhausen et al. (2024) analysed precipitation extremes in an ensemble of continuous simulations at convection permitting resolution covering southern Germany. We will extend the description to “ This is a result of the high variability in extreme precipitation that is superimposed on the overall increase (e.g. Hundhausen et al. 2024) and illustrates the need for long simulations for robust estimates.” We might include a comparison between different realizations in the revised version.
L 226: Can this be compared/used? Does KOSTRA provide sub-hourly rainfall distributions? To my knowledge, KOSTRA presents the statistical evaluation of rainfall frequencies of different duration, which is not necessarily a quantitative estimation of the sub-hourly rainfall distribution of an 1-hour event. Please provide explanation. This means you scale the 5-min rainfall estimates to match the 1-hour 100-year event sum of KOSTRA, correct? Overall, this whole section is not well understandable and needs revision. A graphical sketch of the workflow would certainly help.
The Euler-2 temporal evolution is constructed using the statistical intensities (not a realistic sub-hourly disaggregation of the hourly event). KOSTRA does provide the necessary statistical sub-hourly statistical intensities. It is correct that we scale the 5-min rainfall estimates from KOSTRA. As a scaling factor we use the ration between the simulations and KOSTRA determined for the 100-year hourly event. As we already have a high number of figures, we would like to avoid adding another one. Please see our suggestion to rephrase the paragraph lines 221-236 above.
L. 293: What would be realistic for retention roofs? can they be realized on the typical gable roofs in Berlin? My assumption is not, but only flat roofs can be converted to retention roofs. Is this correct? A few words on the plausibility of the assumption that all roofs can be converted to retention roofs are advisable.
It is correct, that only flat roofs or roofs with a small slope can be converted to retention roofs. As stated in l. 312 ff “While the complete retention during all considered heavy rain events could hardly be realized by modifying all roofs in the model domain to become green or retention roofs, it could - in theory - be achieved with combinations of green and/or retention roofs and cisterns for all buildings in that area.”. So, instead of retention/green roofs, the retention of the rainwater could also be achieved by draining the water from the roofs into (subsurface) cisterns. An analysis of the potential for retention roofs in the study area is beyond the scope of this study, as we focused to investigate the theoretically maximum possible effect by capturing rain water falling on roof surfaces.
L.361: That's absolutely correct and opens the question why not the actual flood volume is used. This can be easily calculated summing up the surface water depths at the end of the rainfall (i.e. t = 60 min). This would be a realistic number considering the direct drainage of rainfall. I recommend to use this value in the analysis.
As mentioned in an answer to an earlier comment, we will adapt that and compare the volumes at different time steps.
L.371: These numbers are based on that unrealistic calculation of flood volume, thus they have very limited practical meaning, unless you show that the flood volume as you calculated is the same as the flood volume calculation I suggest above. If you use the actual flood volume as I suggest, you can avoid a discussion like this and the volume evaluation in relation to rainfall sum would be much more robust and meaningful.
This comment refers to all other numbers in this volume comparison section.
As mentioned in an answer to an earlier comment, we will adapt that and compare the volumes at different time steps.
L433: But you used data from the soil survey, thus the simulation of the infiltration based on the surveyed conductivities should not be that uncertain.
But you are right, for the simulation of de-sealing measures this is an issue, as the soil under buildings is not surveyed. But this has not been done in this study.
You may want to change the statement towards this argument.
The available soil data is relatively coarse (on a block level), so local details might not be captured. But you are right, the uncertainties on the overall effects in the considered study area should not be too large.
L.435: I made a comment about this in the method section. It is important to state this earlier in the methods, not only here in the discussion.
We will now state this already in the section about the model setup.
L. 436: is this simulated like this in the hms++ model? this also needs to be stated in the method section.
We will clarify this in the section about the model setup.
Moreover, the statement that "most" of intense rainfall is overspilling the roof drainage is debatable. Is there any evidence for this, e.g. design values for roof drainage vs. rainfall? To my knowledge roof drainage is as designed to the same standard as street drainage.
Means, neglecting roof drainage is actually the same as assuming no subsurface drainage for the whole area, as it is usually done in pluvial flood mapping. But drainage is important, as you correctly state in the introduction. Thus, why is roof drainage neglected? Wouldn't it be closer to reality to consider 100% rood drainage than none?
The available drainage model data does not include the roof drainages. We could make assumptions on it, but this would mean a lot of effort to connect all roof drainages correctly to the subsurface drainage network in the model. In our case, the rain is flowing from the roofs to the adjacent areas, in case of roads, the water can be drained into the subsurface drainage system by street inlets, which can be considered as good approximation of the actual roof drainage directly into the subsurface drainage system. Some of the runoff also flows to backyards, where no connection to the subsurface drainage system is present in the model. This part of surface flooding might indeed be to some extent an overestimation, as at least some part of the water would be drained into the subsurface drainage system instead of reaching the backyard. At the same time, this means some kind of underestimation of water present in the subsurface drainage system and therefore also combined sewer overflow. We will discuss these effects due to the simplified assumptions without roof drainage. A inclusion of roof drainage into the subsurface drainage system is not feasible within this study.
L. 488: add this paragraph to the paragraph stating uncertainties above. The conclusion is based on the assumption that the single used global-regional climate model combination is representative for the actual climate development/the usually used ensemble (median). You should prove that this increase in heavy precipitation is very likely in the text above (literature), that the model represents this well, and state this here to support your conclusion.
We suggest to extend the paragraph:
“The study is associated with a number of uncertainties. For the extreme rainfall analysis, these include the choice of the climate scenario and the fact that only one global-regional model combination was available for the study. This limitation affects the estimation of the strength of the precipitation change, that is at the upper range of changes reported by Hundhausen et al. (2024) for the area average in southern Germany. The fact that extreme precipitation intensifies under global warming conditions can be regarded as undisputed and has been recognized by a large number of studies (IPCC, 2021; Fowler et al., 2021).”
References
Fowler, H.J., Lenderink, G., Prein, A.F. et al. (2021). Anthropogenic intensification of short-duration rainfall extremes. Nat Rev Earth Environ 2, 107–122. https://doi.org/10.1038/s43017-020-00128-6IPCC (2021). Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change[Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, In press. https://doi.org/10.1017/9781009157896
Hundhausen, M., Feldmann, H., Kohlhepp, R., and Pinto, J. G. (2024). Climate change signals of extreme precipitation return levels for Germany in a transient convection-permitting simulation ensemble, International Journal of Climatology, 44, 1454 – 1471, https://api.semanticscholar.org/CorpusID:270475096
Citation: https://doi.org/10.5194/egusphere-2025-445-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 392 | 136 | 18 | 546 | 13 | 29 |
- HTML: 392
- PDF: 136
- XML: 18
- Total: 546
- BibTeX: 13
- EndNote: 29
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1