the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Past, present and future rainfall erosivity in Central Europe based on convection-permitting climate simulations
Michael Haller
Christoph Brendel
Gudrun Hillebrand
Thomas Hoffmann
Abstract. Heavy rainfall is the main driver of soil erosion by water which is a threat to soil and water resources across the globe. As a consequence of climate change, precipitation – and especially extreme precipitation – is increasing in a warmer world, leading to an increase in rainfall erosivity. However, conventional global climate models struggle to represent extreme rain events and cannot provide precipitation data at the high spatio-temporal resolution that is needed for an accurate estimation of future rainfall erosivity. Convection-permitting simulations (CPS) on the other hand, provide high-resolution precipitation data and a better representation of extreme rain events, but they are mostly limited to relatively small spatial extents and short time periods. Here we present for the first time rainfall erosivity and soil erosion scenarios in a large modelling domain such as Central Europe based on high-resolution CPS climate data generated with COSMO-CLM. We calculate rainfall erosivity for the past (1971–2000), present (2001–2019), near future (2031–2060) and far future (2071–2100) and apply the new data set in the soil erosion model WaTEM/SEDEM for the Elbe River basin. Our results showed that future increases in rainfall erosivity in Central Europe can be up to 84 % in the river basins of Central Europe. These increases are much higher than previously estimated based on regression with mean annual precipitation. In consequence, soil erosion and sediment delivery in the Elbe River basin are also increasing strongly. Locally, changes in erosion rates can be as high as 120 %. We conclude that despite remaining limitations, convection-permitting simulations have an enormous and to date unexploited potential for climate impact studies on soil erosion. Thus, the soil erosion modelling community should follow closely the recent and future advances in climate modelling to take advantage of new CPS for climate impact studies.
- Preprint
(1401 KB) - Metadata XML
-
Supplement
(184 KB) - BibTeX
- EndNote
Magdalena Uber et al.
Status: final response (author comments only)
-
CC1: 'Comment on hess-2023-120', Panos Panagos, 19 May 2023
The manuscript is very interesting and you have addressed a topic which contributes to better research on soil erosion modelling.
However, there are two inconsistencies that somehow “speculate” differences between your erosivity estimation and the ones from other studies.
The first one has to do with the use of the equations for kinetic energy Eq.2. You have used different equations than the ones of Panagos et al (2015). Using a different equation will have as a result to have different results in the R-factor estimation. Nobody argues that your equation is worst or better than the other one. However, you cannot speculate that the others have inconsistencies or underestimate erosivity. It is simply comparing somehow different outputs.
The second problematic has to do with the temporal scaling factor of 1.9. In the European study and the sub-sequent one of Panagos et al (2016) on conversion factors between different time resolutions, the authors have estimated a conversion factor of 1.56 between 60 minutes and 30-minutes difference. Obviously, your values will be much higher since you use a much higher conversion factor.
For those 2 reasons your figure 3a comparison is not correct and you are called to correct it. In addition, you have to discuss those assumptions/differences in estimating the erosivity.
Those two issues are the most problematic ones that researchers easily get into trouble.
Citation: https://doi.org/10.5194/hess-2023-120-CC1 - AC1: 'Reply on CC1', Magdalena Uber, 30 May 2023
-
RC1: 'Comment on hess-2023-120', Anonymous Referee #1, 07 Jun 2023
In the submitted paper authors investigate past, present and future rainfall erosivity in relation to soil erosion in Central Europe. Authors also compare rainfall erosivity maps derived using 1h precipitation data and maps derived based on the annual precipitation data with the consideration of simple empirical DIN equation. Soil erosion-sediment transport modelling is also conducted using the WaTEM/SEDEM model. The topics is very interesting and within the scope of the HESS journal. The paper is very well written, easy to follow. I only have some moderate comments and suggestions.
Firstly, authors only used the RCP8.5 scenario but this is only mentioned a few times in the manuscript. Authors should definitely state this more clearly in abstract and conclusions since probably RCP2.6 and RCP4.5 would yield smaller increase in the rainfall erosivity and also in the soil erosion rates. It would be definitely very interesting to include these scenarios if input data would be available. Hence, the presented results are significantly influenced by this selection (data availability actually since COSMO-CLM (CPS-SCEN) is only available for RCP8.5). Related to this I suggest that authors add some discussion in relation to using only RCP8.5 and try to elaborate a bit more about the possible results (i.e., deviations from the presented results) using also the RCP2.6 and RCP4.5.
Secondly, part of the results is influenced by the selection of the CMIP5 model ensemble. Since CMIP6 is also available authors should at least add some discussion about the impact of using CMIP5 instead of CMIP6. This is another selection done that has probably quite significant impact on the derived results.
Thirdly, authors used median of the model ensemble, can you add some additional results (e.g., 25% or 75% or 10-90% quantiles) to the Supplement in order to show what is the variability among the included models.
Finally, the results are also significantly influenced by the data time step (1h) since conversion factor needs to be applied. I suggest that authors add more discussion about the selected temporal scaling conversion factor (i.e., 1.9) and try to elaborate about the possible impact on the derived results (i.e., rainfall erosivity and modelled soil erosion and sediment transport rates).
Some specific comments:
-L161-162: Please add more details.
-Figure 3: Maybe add R2 to the figure as well.
-Discussion in section 3.4 is very useful.
-Figure 4: Please add more details about the Erosion Index in the Material and methods section.
-L461-465: From my perspective hourly resolution is actually quite problematic especially because the applied conversion factor (only one number (fixed for the whole period)), is used for different type of rainfall events (e.g., intense storms, longer duration events). In relation to this some progress should be made in future.
-L466-471: This is only valid for the RCP8.5. It should be clearly mentioned and discussed.
Citation: https://doi.org/10.5194/hess-2023-120-RC1 - AC2: 'Reply on RC1', Magdalena Uber, 01 Sep 2023
-
RC2: 'Comment on hess-2023-120', Anonymous Referee #2, 19 Jun 2023
The study of Uber et al. is very interesting and of high importance for proper estimation of future rainfall erosivity and future soil loss in a changing climate. The topic of the submitted manuscript addresses well the scope of the HESS journal. Uber et al. use precipitation data from convection-permitting simulations (CPS) for estimation of future rainfall erosivity in Central Europe. The use of CPS-based precipitation data for estimations of future erosivity is the novelty of their study. CPS include explicit simulation of large deep convection cells what is not the case in conventional convection-parametrized climate models but important for erosivity estimations. CPS-based precipitation data are available in a spatial resolution of around 3 km and a temporal resolution of 1 h. This spatial and temporal resolution of precipitation data is considerably higher than it is the case for the output of convection-parameterized models. A high spatial and temporal resolution of precipitation is also of importance for erosivity estimations. Uber et al. describe well in their introduction the potential of CPS-based precipitation data for improved erosivity estimations for the near and far future and appropriately explain the limits of current CPS outputs for it. The major limit of CPS-based precipitation data for erosivity projections is the lack of ensembles of CPS. This lack results in an unknown uncertainty of erosivity projections. So far comprehensible, Uber et al. try to account for this by comparing single-CPS-based erosivity estimates with, first, erosivity estimations with precipitation data from CPS evaluation simulations driven by observation data, and, second, with erosivity estimations by a regression equation based on mean annual precipitation sums using precipitation data from conventional convection-parametrized climate model ensembles. Although their attempt is much appreciated, it must be pointed out that this equation (Equation 3 in the manuscript) is explicitly not applicable to current and future erosivity estimations. This is described in the DIN 19708:2022-08. In consequence, the validation approach of comparing single-CPS-based erosivity projections with erosivity projections based on Equation 3 of the manuscript seems incorrect. This approach might be possible for the past period but should be carefully applied due to the trend of increasing erosivity that has already existed in recent decades.
Uber et al. also use their CPS-based erosivity estimates for estimation of future changes of soil loss rates by water erosion and sediment delivery. For this, they select the Elbe River basin. Their USLE-based soil loss estimates for the near and far future consider changes in the R-factor exclusively. Although Uber et al. calculated the erosion index from their erosivity projections (chapter 3.1.3), they don’t consider changes in the C-factor for their future soil loss estimations. Also, potential changes in crop growth due to prolonged vegetation periods are not considered or changes of the other factors. Only future changes in the R-factor are considered in their erosion modelling. In consequence, estimates of future relative changes in soil loss are equal to relative changes in erosivity. This is because the USLE is a multiplicative approach. So, the purpose of the efforts of Uber et al. for the chapter on simulated erosion rates is unclear. I recommend to either consider, at least, changes in the erosion index or to strongly shorten the chapter on future erosion rates. The manuscript would benefit from a concentration on single-CPS-derived erosivity projections and an in-depth discussion of e. g. the applicability of scaling factors accounting for spatial and temporal resolutions of precipitation data lower than those from e.g. 1-min precipitation data from rain gauges. This would also address the comment by Panos Panagos (https://doi.org/10.5194/hess-2023-120-CC1) who criticises the applied temporal scaling factor 1.9 for being too high. Panagos et al. (2015) found a lower temporal scaling factor, but this was used to correct only to 30-min resolution. Considering in addition their temporal scaling factor to correct from 30-min resolution to 5-min resolution (they calculated no factor for 1-min resolution), it results in a factor of even slightly higher than 1.9. In consequence, the temporal scaling factor from Fischer et al. (2018) applied by Uber et al. seems to be in accordance with the temporal scaling factors found by Panagos et al. (2015). Nevertheless, the question could be discussed whether the temporal scaling factor is still valid when rainfall intensities further increase in the near and far future. Moreover, it could be of interest to explain/discuss why no spatial scaling factor was necessary to account for the underestimation of erosivity despite the spatial resolution of around 3 km. A possible reason might be the overestimation of hourly extreme precipitation intensities mentioned in the discussion (page 19, line 444).
Although the manuscript is very well written, it could benefit from a separation of the validation of precipitation and erosivity from CPS projection runs by precipitation and erosivity from CPS evaluation runs, and the results of erosivity from CPS projection runs and their discussion in context to results from other studies.
All in all, the manuscript could benefit from revisions as suggested above and in more detail in the follwoing list below.
Detailed comments:
page 2, line 32: The citation of Nearing et al., 2017 for the definition of rainfall erosivity gives the impression that rainfall erosivity was no topic before. I suggest referring to an appropriate earlier publication from the 1950s to 1970s, at least in addition.
page 2, line 38: Same as comment above; an earlier publication is recommended at least in addition to Wilken et al., 2018
page 2, line 40/41: The “Thus,” is irritating as it gives the impression that there is a conclusion, but it is not obvious from what this conclusion is taken. It doesn’t seem to be taken from the preceding sentence. Please, make it clearer.
page 2, line 42: It might be rather the suitability of the R-factor equations to express rainfall erosivity which depends on the temporal resolution of the precipitation data.
page 2, line 47: “This approach” is not clear as you refer in the preceding sentence to “low-resolution approaches” which suggests that there are several approaches.
page 2, line 49/50: Please add a reference to the expectation of changes in the frequency distribution of rainfall events.
page 2, line 56: Every error needs to be assessed critically. This sentence has not really information. Please clarify.
page 2, line 57: Please clarify that an increase in precipitation (intensity and annual total) is likely rather for wet than for dry global regions according to your reference Sun et al., 2007. Moreover, to my understanding, changes in precipitation are the result of the changes in temperature and both, changes in temperature and precipitation are the changed climate. So, changes in temperature and precipitation are not the result of climate change but they are the changed climate itself which, simplified, results from increased concentrations of greenhouse gases in the atmosphere.
page 2, line 59/60: Please clarify what specifically is increasing. Do you mean the number of extreme events?
It is not clear what you mean with an “intensified hydrologic regime”. Please edit.page 2, line 60/61: “This is due to the fact…” suggests that you are describing a cause-and-effect relationship, but, actually, it just clarifies the preceding sentence. I would suggest combining both sentences (line 59 – 61) into one but clear sentence.
page 3, line 75: Do you mean 11 out of the 196 studies?
page 3, line 95: Do you mean with “at the time scales needed…” the length of the time series? If yes, it might be clearer to write “for the length of time series” as ‘time scales’ may be used rather in context to the temporal resolution of the data. If you agree, please keep it in this way throughout the manuscript.
page 4, line 108: Please add the reference where to find the published data.
page 4, line 110: Was the temporal resolution of CPS already mentioned? If not, please add so that it is clear what you mean with high temporal resolution.
page 4, line 120: Which size of regions to do mean? In context of the globe (mentioned in the same sentence), a region could also be Europe, but you may rather think of smaller areas? Maybe it is possible to provide the typical spatial extent in km2 or to refer to a specific geographic region as an example.
page 6, line 170: It should be specified that you mean the final CPS model output as you mentioned several different models before.
page 6, line 172: What is the ‘FPS-convection contribution’?
page 7, line 182: Please add that the point that these must be 6 hours without any precipitation.
page 7, line 186: In case of 𝐼 ≥ 76.2 𝑚𝑚 ℎ−1, the kinetic energy should be 28.33 * 10-3; please correct this factor.
page 7, line 194: If ‘grid cell’ is meant to be equal to ‘grid point’ then it may be good to use the same wording throughout the manuscript for the sake of simplicity.
page 8, line 208: It is not clear why erosivity is also calculated by Equation 3. “For comparison” is not enough as a reason. Is it meant for validation of the R-factors derived from single-CPS-based precipitation data? This equation is not valid for calculation of current and future erosivity as described in the beginning.
page 8, line223: The value 0.2 mm/a for simulated erosion rate should be based on more information. Is it a multi-year average and of which time period - past, present, future? What means ‘locally’? Is it simulated for a single field? Or averaged for e.g. a county? Please specify to what this value relates.
page 8, line 220-230: This text section may be rather part of the introduction. Think about shifting it to page five, line 131.
page 9, line 245: You write that the aim is to identify the impact of climate change on soil erosion and therefore you just consider changes in the R-factor. But also, the erosion index changes and the soil loss ratios by e.g. prolonged growing season (even when the management itself is not changed). Climate change might also affect soil erodibility (e. g. by changes in soil organic matter). So, you should not write that you aim to identify the impact of climate change but rather the impact of changes in the R-factor by climate change on soil erosion estimates.
page 9, line 247f: Sediment load measurements of which years were used for calibration and validation of the model? Please add this information.
page 9, line 255f: Why don’t you distinguish between past 1971-2000 and present 2001-2019? Would be good to have average annual erosivity of the entire modelling domain for both periods separately.
Moreover, don’t you mean that average annual erosivity in the lowlands is between 50 and 90 N h-1 a-1 rather than “erosivity in the lowlands is on average about 50 – 90 N h-1 a-1”? The term ‘on average’ lets one expect a single value.page 9, line 257: Similar to the comment above; how to interpretate the range 90 – 96 N h-1 a-1 as a mean? Is this range arising from a confidence interval around the mean? “The mean of the entire modeling domain” let one expect a single value which is then also given for the near and far future. Please edit respectively.
page 9, line 258: You are inviting people to use your R-factor results for USLE-based soil loss modelling. But for this, the erosion index is missing. So, you would need to provide these data as well. Moreover, I like to recommend providing most important information about the data and development in a table and pointing out that the results are based on a single climate model and not on a model ensemble using the RCP8.5 scenario.
page 11, line 294f: Is it possible to calculate in addition the mean annual R-Factor from the evaluation runs for the respective time periods used in the other studies and to compare these? Moreover, the differences could be discussed in context to changes of the R-factor in the last decades as discovered by other studies already.
page 12, line 300f: What is the reason for comparing results of other studies for the R-factor with the results of the evaluation run but for the erosion index with the results of the projection run?
page 12, line 300f: How does the seasonal distribution of erosivities from evaluation runs and projection runs fit together for the two periods respectively? Please add information about this.
page 12, line 316f: Should it not be relatively easy to analyse the erosion index restricted to the area of the individual countries? This should be done instead of guessing that the changes in the intra-annual distribution of erosivity level out across the modelling domain.
page 13, line 321f: Mention the mean annual rainfall erosivities for near and far future so that it is easier to compare the results of evaluation and projection run.
page 14, line 339: It is not the age but changes in the precipitation characteristics and the fact that it does not consider rainfall intensities but only total rainfall amount.
page 14, line 341: The DIN 19708 also explicitly states that the regression equations are not suitable for calculation of R-factors of (even) the presence, not to mention the future.
page 15, line 364f: Discard the part of the sentence which refers to the comparison with the results from Equation 3.
page 15, line 366: Please add the references which report higher future increase in erosivity in comparison with your results from CPS.
page 16, line 370f: Please indicate the period to which your results on erosion rate estimates relate here.
page 17, line 391f: It is unclear what the message of the first sentence is. Relative changes in erosion rate are equal or slightly higher than relative changes in sediment delivery to water bodies? Moreover, it is not indicated to which simulations (‘CPS’ or ‘MAP’ approach?) the mentioned changes in sediment load in the near and far future relate. Please revise.
page 17, line 397: With respect, your results are not necessary to know that USLE-based simulations of future changes in soil erosion are highly sensitive to changes in the R-factor when just the R-factor is changed while the other factors are assumed to remain equal.
page 18, line 410f: Please indicate that the resolution of precipitation data of 3 km is just high in comparison to most other projected precipitation data (as this is not the case in comparison to measured precipitation data by e. g. ground-based radar).
page 18, line 427f: The conclusion of the results from Chapman et al. (2021) cited by you would be that projected monthly (? - what is “the same temporal resolution”?) precipitation sums are higher from the convection-permitting model than from the conventional convection-parameterized model as changes in erosivity resulting from changes in the rainfall intensity are not necessarily reflected in changes of e.g. monthly precipitation sums. What is the reason for that?
page 19, line 435f: Here, you discuss limitations of the USLE although the intension of chapter 3.4 is to discuss potentials and limitations of CPS for calculation of erosivity. Delete the part on limitations of the USLE and concentrate on discussing the limitations by CPS-derived erosivities as calculated in your study. For example:
Do you expect an underestimation of erosivity in near and far future by using a constant temporal scaling factors of 1.9 for accounting of erosivity underestimation by using 60-min data?
In the introduction you mentioned that CPS can simulate large deep convection cells but not smaller shallow convection (line 84). Which consequence do you expect from this for projected near and far future erosivity estimates?
Are there any indicators which allow a guess whether your single-CPS-based results of erosivity in the near and far future are laying in the lower, middle, or upper range of erosivities derived from future ensembles of CPS?page 19, line 442f: The comparison of modelled and measured precipitation data is an important result. What is the reason that you don’t provide this information earlier, e. g. when you compare your erosivity estimations from projection runs with those from evaluation runs? What is the reason for the overestimation of hourly extreme precipitation intensities? How large is this overestimation?
page 19, line 446: What do you mean with “estimates of future precipitation”? Do you mean hourly precipitation sums, or monthly, or yearly?
page 20, line 468: Which spatial scale do you mean with “locally”? Is it a single ‘pixel’ of a certain size? How can the increase in soil erosion be higher than the increase in erosivity although all other USLE factors remain constant? Do you refer once (line 467) to an average across an area and once (line 468) to a single ‘pixel’?
According to language/grammar:
Please write either “modelling” or “modeling” throughout the manuscript.
page 2, line 34: A comma might be necessary before “and” in “…its derivates and models…”.
page 4, line 108: Please use either singular or plural of ‘data’ consequently throughout the manuscript.
page 5, line 131: “…the new rainfall erosivity maps” sounds as it will be official maps. I suggest to revise this sentence. Maybe like “Furthermore, modelled rainfall erosivities of these periods were used in the USLE-based model WaTEM/SEDEM to estimate changes in soil erosion and sediment delivery to the Elbe River.”
page 5, line 137: With ‘or results’ you may mean ‘our results’. Please correct.
page 7, line 182: Instead of “We use”, it may be “We used”.
page 9, line 246: ‘r’ is missing in ‘future’.
page 9, line 254: Should be ‘Alps’ instead of ‘Alpes’. Please adjust also Fig.1 accordingly.
page 9, line 255: Should be ‘Alps’ instead of ‘Alpes’.
page 12, line 315: Why do you write “…despite…” here? Isn’t it logical that the erosion index needs to decrease in other months when there is an increase in May to October?
References
DIN-Normenausschuss Wasserwesen: DIN 19708:2022-08 Bodenbeschaffenheit - Ermittlung der Erosionsgefährdung von Böden durch Wasser mit Hilfe der ABAG, https://dx.doi.org/10.31030/3365455, 2022.
Fischer, F. K., Winterrath, T., and Auerswald, K.: Temporal-and spatial-scale and positional effects on rain erosivity derived from point-scale and contiguous rain data, HYDROL EARTH SYST SC, 22, 6505-6518, https://doi.org/10.5194/hess-22-6505-2018, 2018.
Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., Tadić, M. P., Michaelides, S., Hrabalíková, M., and Olsen, P.: Rainfall erosivity in Europe, SCI TOTAL ENVIRON, 511, 801-814, https://doi.org/10.1016/j.scitotenv.2015.01.008, 2015.
Citation: https://doi.org/10.5194/hess-2023-120-RC2 - AC3: 'Reply on RC2', Magdalena Uber, 01 Sep 2023
-
CC2: 'Comment on hess-2023-120', Shuiqing Yin, 29 Jun 2023
The study proposed the advantage of Convection-Permitting Simulations (CPS) in simulating extreme convective precipitation, which plays an critical role in the soil erosion process. Uber et al. compared rainfall erosivity for the past (1971-2000), present (2001-2019), near future (2031-2060), and far future (2071-2100) periods in the Central European region based on CPS data as well as soil erosion in the Elbe River basin. The study follows in the scope of HESS and the use of CPS data for the estimation and projection of future erosivity is the novelty of the study. Following are the main concerned issues:
- Uber et al. used “CMIP5 driven CPS” to compare with “ERA5 driven CPS” for the evaluation simulation, which can not demonstrate the benefit of CPS in simulating extreme precipitation. High spatial-temporal resolutions of precipitation observations should be used for the purpose.
- Regression equation based on mean annual precipitation sums can not be used to project future rainfall erosivity for it can not fully reflect the change of daily and hourly precipitation intensity along with the warming climate.
- This reviewer agrees with the RC2 that the temporal scaling factor of 1.9 is reasonable (Yue et al. (2020) as a reference, Effect of time resolution of rainfall measurements on the erosivity factor in the USLE in China) . Nearing et al. (2017, Rainfall erosivity: An historical review) reported the differences among three KE-I equations in USLE and its revisions, which may be useful for you to discuss differences between your study and Panagos et al. (2015c). Breakpoint or less than 5-min interval observation data are suggested to be obtained for the determination of the scaling factor of the study area.
- This reviewer agrees with the RC2 that the chapter on future erosion rates for the Elbe River basin is strongly shorten or even deleted. Instead, the study focuses on the projection of rainfall erosivity and fully demonstrates the advantage of CPS data to highlight the novelty.
Citation: https://doi.org/10.5194/hess-2023-120-CC2 - AC5: 'Reply on CC2', Magdalena Uber, 01 Sep 2023
-
RC3: 'Comment on hess-2023-120', Anonymous Referee #3, 19 Jul 2023
- AC4: 'Reply on RC3', Magdalena Uber, 01 Sep 2023
Magdalena Uber et al.
Data sets
Past, present and future rainfall erosivity in Central Europe Magdalena Uber, Michael Haller, Christoph Brendel, Gudrun Hillebrand, and Thomas Hoffmann https://doi.org/10.5281/zenodo.7628957
Magdalena Uber et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
478 | 137 | 29 | 644 | 40 | 5 | 7 |
- HTML: 478
- PDF: 137
- XML: 29
- Total: 644
- Supplement: 40
- BibTeX: 5
- EndNote: 7
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1