the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Mar 2026
European runoff drought event types: From historical classification to projected future changes
Abstract. European drought events have intensified in recent decades, raising concerns about water resources, agriculture, and ecosystem health. Most existing continental-scale drought assessments provide limited attribution of drought changes to the processes that generate runoff drought, which constrains understanding of how these processes will evolve in a warming climate. In this study, we provide the first continental-scale assessment of future changes in runoff drought generation processes across Europe, focusing on rainfall deficit, rain-to-snow and wet-to-dry season transitions, and snow-related processes. We used a continental-scale hydrological model in combination with observed meteorological data and climate simulations (GCMs) under a middle-of-the-road emission scenario (i.e., RCP4.5, ≈2 °C global warming by 2100). Unlike widely used aggregated drought index-based studies, our mechanism-specific classification approach distinguishes the physical drivers of drought events by classifying runoff droughts into seven event types based on their severity, duration, and frequency and leverages spatial clustering analysis (Getis–Ord G∗i method) to assess the climate responses. Our analysis reveals considerable regional differences in drought mechanisms and their responses. Historical observations (1971–2000) highlight that Mediterranean Europe experiences the most severe drought conditions, dominated by rainfall deficit processes and wet-to-dry season transitions, with deficits exceeding (> 4 mm day−1), prolonged durations (> 165 days), and high event frequencies. Future projections (2070–2099) indicate further drought intensification in the Mediterranean driven by increasing rainfall deficit and wet-to-dry transition events, with runoff deficit increases by 2–6 mm day−1 and duration extensions exceeding 200 days, while Northern and Western-Central Europe show predominantly decreasing drought severity due to declining cold-snow season droughts under warming conditions. Importantly, drought related to temperature-driven processes, especially those triggered by rain-to-snow transitions, exhibit the most pronounced projected changes. These findings demonstrate that different drought mechanisms respond distinctly to climate forcing. By attributing projected drought changes to specific generation processes, our results enable region-specific interpretation of drought hazard, which is crucial for effective water-resources planning across Europe.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Hydrology and Earth System Sciences.
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
(7861 KB) - Metadata XML
-
Supplement
(5977 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-973', Yves Tramblay, 08 Mar 2026
-
AC2: 'Reply on CC1', Sadaf Nasreen, 16 Mar 2026
We thank Dr. Yves Tramblay for this important and constructive comment. We agree that the rationale for the selection of validation basins in Fig. S1 was not explained clearly enough in the original manuscript. The station-screening criteria, and their implications for spatial representativeness, were not described in sufficient detail, and we will improve this part in the revised version. Somehow during the manuscript finalization, we missed to properly explain to the readers, that the purpose of Fig. S1 was to provide a drought-focused evaluation of the mHM simulations forced by EOBS. This included checking the consistency between the finer-resolution setup (0.1°) and the 0.5° setup used for the ISIMIP-based climate change analysis. It was not intended as a comprehensive validation against all currently available European gauges.In the revised manuscript, we will describe the station-selection procedure more explicitly. This will include the screening criterion used to retain only stations for which the upstream area was represented adequately in the coarser 0.5° setup. We will also report the number of retained basins more transparently and add information on their size range and regional distribution. In addition, we will update the station information using the latest GRDC availability as of early 2026.We also agree that the selected basin set is spatially uneven, and that this should be stated more clearly. In particular, the validation coverage is denser in Central Europe than in parts of southern Europe. This reduces the strength of the validation evidence in some Mediterranean regions. We will therefore revise the text to avoid implying uniform confidence across Europe.We also thank the Reviewer for pointing us to the EStreams database, which is a valuable new resource for large-sample European hydrology. We will include more validation stations using the latest updated GRDC records. Where feasible, we will possibly explore complementary resources such as EStreams.To address this comment more fully, we will revise the manuscript and supplement by:(1) adding an explicit description of the basin-selection criteria;(2) reporting the spatial distribution of the retained basins more clearly; and(3) softening the interpretation of Mediterranean-scale conclusions in regions where observational validation coverage remains limited.Citation: https://doi.org/
10.5194/egusphere-2026-973-AC2
-
AC2: 'Reply on CC1', Sadaf Nasreen, 16 Mar 2026
-
AC1: 'Comment on egusphere-2026-973', Sadaf Nasreen, 16 Mar 2026
We thank Dr. Yves Tramblay for this important and constructive comment. We agree that the rationale for the selection of validation basins in Fig. S1 was not explained clearly enough in the original manuscript. The station-screening criteria, and their implications for spatial representativeness, were not described in sufficient detail, and we will improve this part in the revised version. Somehow during the manuscript finalization, we missed to properly explain to the readers, that the purpose of Fig. S1 was to provide a drought-focused evaluation of the mHM simulations forced by EOBS. This included checking the consistency between the finer-resolution setup (0.1°) and the 0.5° setup used for the ISIMIP-based climate change analysis. It was not intended as a comprehensive validation against all currently available European gauges.In the revised manuscript, we will describe the station-selection procedure more explicitly. This will include the screening criterion used to retain only stations for which the upstream area was represented adequately in the coarser 0.5° setup. We will also report the number of retained basins more transparently and add information on their size range and regional distribution. In addition, we will update the station information using the latest GRDC availability as of early 2026.We also agree that the selected basin set is spatially uneven, and that this should be stated more clearly. In particular, the validation coverage is denser in Central Europe than in parts of southern Europe. This reduces the strength of the validation evidence in some Mediterranean regions. We will therefore revise the text to avoid implying uniform confidence across Europe.We also thank the Reviewer for pointing us to the EStreams database, which is a valuable new resource for large-sample European hydrology. We will include more validation stations using the latest updated GRDC records. Where feasible, we will possibly explore complementary resources such as EStreams.To address this comment more fully, we will revise the manuscript and supplement by:(1) adding an explicit description of the basin-selection criteria;(2) reporting the spatial distribution of the retained basins more clearly; and(3) softening the interpretation of Mediterranean-scale conclusions in regions where observational validation coverage remains limited.Citation: https://doi.org/
10.5194/egusphere-2026-973-AC1 -
RC1: 'Comment on egusphere-2026-973', Emanuele Mombrini, 03 Apr 2026
I have found the presented paper interesting, as I feel it brings noteworthy results on the evolution of drought under climate change. Furthermore I feel that the clustering of areas in terms of projected change adds much to the legibility of a complicated subject, making dense results more easily understandable. Still, I think that the current manuscript shows some problems, particularly regarding the presentation of the methodology used, which need to be addressed.
The work seems to be based mainly on Brunner et al., 2022, sharing with it much in terms of both aims and methodology. The significant differences between this work and the above cited one are the use of a different dataset, the inclusion of future projections and the clustering through hotspot analysis. These are by no mean trivial differences, but I feel that the implications of these changes or the reason for them are not expressed clearly enough in the current manuscript. In particular, line 149 mentions that the proposed classification is modified based on Brunner’s proposed one, but the paper remains vague on what the modifications might be. If Brunner’s classification is indeed modified, it would be both interesting and necessary to explain how, especially given the ample discussion of the possible problems of this classification discussed in the 2022 article.
I find the discussion of drought identification severely lacking, in a way that affects the legibility of the subsequent results. First, it is not clear to me how the sensitivity analysis for the drought identification parameters is conducted. The supplementary figures by themselves don’t offer much explanation, and it is not clear to me in what sense event characteristics should be “stable” (line 142). The minimum length chosen is 1, which is not a value present in the figures themselves, and it’s not clear why a ml value outside the sensitivity analysis range would be chosen. I advise to expand the supplemental portion of the manuscript by including a more complete explanation, and in any case make clearer what the goal of this sensitivity analysis was.
Second, it is not clear to me on which data the threshold values are calculated (i.e. if only on E-OBS or also the GMCs) and if the thresholds calculated through the chosen percentile are calculated for each time series. This is problematic since, given the use of percentile thresholds and the change of time period, this opens up questions about which reference period to use for such quantiles. Given the large time passing between 1970 and 2050, climate shifts which would change the underlying climate, and thus what constitutes a drought in statistical terms, are likely to happen. If the historical period is used as a reference for the future projection, this needs to be made explicit since it is fundamental in understanding the results themselves.
Another aspect lacking from the methodological discussion is how the timing of droughts is evaluated. Since this aspect is discussed at length in Section 3.4, a description of how this is evaluated is needed. I also find Figure 8 difficult to understand, as it often seems to show drought ending dates earlier than their starting ones (especially the GCM_hist rainfall deficit droughts). This doesn’t seem to be reflected in the discussion of the figure itself, and based on Line 197 I think the start and end panels may have been switched.
Bias in the climate models is it discussed, with the assumption that although strong biases are found, comparison between historical and future projection climate data are unaffected by such biases due to both datasets sharing them. While I don’t disagree with this argument, I do not feel confident in this assumption, and the manuscript doesn’t provide either enough references or reasoning for it. A reference is given to Gurdmundsson and Seneviratne, 2016, at line 326, but it is not clear to me how their work supports the idea of “internally consistent relative changes”. They in fact seem to confirm the results from climate models through comparisons with measured data, and write that when “the observational and the model based assessment is conflicting [...] results are inconclusive [.]”. I would either find a more focused discussion in support of the author’s assumption or make clearer the speculative nature of this argument.
Technical details:
On line 44 Brunner at al., 2021 is cited, but given that this paper cites only briefly the categorization of hydrological drought, Brunner et al., 2022 is a better reference. The same applies at line 81.
The choice of threshold for drought identification is not described clearly: I think stating that the tau value indicates the non-exceedance probability would make this part easier to understand. I would also rephrase line 131 making clear that percentiles are in fact used in this work and not just “typically” used.
Figure 2 seems to contain a mistake, as according to it wet to dry season droughts result from deficits outside summer not transitioning to dry periods. It also lacks a connection between rainfall deficits droughts and the “transition to dry period” option. If these are not errors, then an explanation for why the 2022 categorization has been changed is needed.
In Table 3 I would add also the values for the ensemble climate model which biases are showed in Figure 3. The same applies to Table 4 and Figure 4.
The use of citations on lines 240 and 243 is confusing to me. The section presents results from the current study, and the citations seem to be put there to indicate consistency with those results. I would first present the results and then indicate that they are consistent with other studies such as those cited. This also applies to line 271-2.
Line 261 feels a bit misleading, as values above the third quartile are used for the Mediterranean region drought duration increase (“durations may lengthen by more than 200 days”) compared to roughly the first and third quartile for the West-Central Europe region (“approx 25–75 days”). Given the highest values found in the Mediterranean region are remarkable, I would compare similar values while also mentioning this fact.
Citation: https://doi.org/10.5194/egusphere-2026-973-RC1 -
AC3: 'Reply on RC1', Sadaf Nasreen, 08 Jun 2026
We sincerely thank you Emanuele Mombrini for your thoughtful and constructive review of our manuscript. We greatly appreciate the time and care you took in reading the paper and providing detailed suggestions. Please find attached our point-by-point responses to your comments.
Your feedback has been very helpful in improving both the clarity and scientific presentation of the manuscript. In particular, your comments helped us strengthen the description of the drought-identification sensitivity analysis, clarify the threshold-selection procedure and the fixed 1971--2000 reference period, improve the explanation of drought onset and termination, and present the interpretation of GCM bias and projected drought changes more cautiously. We also revised the relevant figure and table explanations, improved citation placement, added ensemble values where appropriate, and clarified the discussion of Mediterranean drought-duration results.
Thank you once again for your valuable comments and for helping us improve the manuscript.
-
AC3: 'Reply on RC1', Sadaf Nasreen, 08 Jun 2026
-
RC2: 'Comment on egusphere-2026-973', Jitao Zhang, 22 May 2026
The manuscript addresses an important and timely topic: process-based classification of European runoff droughts and their projected changes under climate warming. The study has clear potential, especially because it moves beyond aggregated drought indices and attempts to attribute future changes to specific drought-generating mechanisms. However, several methodological, technical, and interpretive issues need to be resolved before the conclusions can be considered robust.
1.The drought threshold reference period must be clarified.It is unclear whether drought thresholds are calculated separately for each dataset and period, or whether historical thresholds are fixed and then applied to future simulations. This is a critical methodological point. If future thresholds are recalculated using the future climate period, the analysis identifies droughts relative to future climatology rather than changes relative to present-day conditions. The authors should explicitly state the threshold strategy and justify it.
2.The drought classification scheme is not sufficiently reproducible.The seven drought types are central to the paper, but the classification rules are not described in enough detail. The manuscript should provide a clear algorithm, ideally as pseudocode or a decision table. Important details include the definition of “proximal rainfall deficit,” the six-pentad window, the treatment of overlapping conditions, the priority order among drought types, the calculation of SWE anomalies, and the handling of grid cells with no snow or very limited SWE.
3.The parameter selection for drought identification appears subjective.The selected values of mit, ml, and tau are said to be chosen by inspecting sensitivity analyses. This is not sufficiently objective. The authors should explain what criterion defines “stable event characteristics” and show whether the main conclusions are robust to alternative parameter choices. The runoff minimum length of eight pentads, approximately 40 days, is especially important because it may exclude shorter but hydrologically meaningful drought events.
4.Historical model biases are large and require stronger treatment.The GCM-driven simulations substantially underestimate event counts for several drought types, especially snowmelt and composite droughts. The argument that future-minus-historical changes are robust because both periods share the same bias structure is not fully convincing. The authors should show model spread, model agreement, and possibly robustness categories, rather than relying mainly on the ensemble mean.
5.The meaning and units of drought severity need clarification.The methods define severity as cumulative deficit, but the results frequently discuss mean runoff deficit in mm day⁻¹. The manuscript should clearly distinguish between cumulative severity, mean intensity, runoff deficit anomaly, and absolute change. The sign convention should also be standardized. Presenting deficits as positive magnitudes may reduce confusion.
6.Some conclusions appear stronger than the evidence supports.For example, the manuscript states that temperature-driven processes, especially rain-to-snow droughts, show the most pronounced projected changes. However, Table 4 shows large inter-model disagreement for rain-to-snow events, including both decreases and increases. The authors should moderate this claim or support it with model-agreement statistics.
7.The Getis-Ord hotspot analysis requires more methodological detail.The manuscript should specify the spatial weights matrix, distance threshold, neighborhood definition, significance levels, treatment of edge effects, and whether multiple-testing correction was applied. Without this information, claims about statistically robust hotspots are difficult to evaluate.
8.Spatial resolution and regridding procedures need to be described.EOBS is at 0.25°, ISIMIP2b is at 0.5°, and mHM may operate at another resolution. The final analysis grid, interpolation/regridding method, land mask, area weighting, and regional aggregation procedure should be clearly stated.
9.The use of RCP4.5/CMIP5 needs more careful framing.The framework is useful, but the manuscript should avoid overgeneralizing from one scenario and five CMIP5 models. The authors should frame the results as a mid-range CMIP5/ISIMIP2b case study and discuss how CMIP6, ISIMIP3b, or EURO-CORDEX could affect regional signals.
10.The article should be written scientifically. It is recommended to change 'we' and 'our' to 'this paper' or 'this study'.
Suggestions to result section. 1.Present model uncertainty more explicitly. For each drought type and region, show the ensemble median, interquartile range, and number of models agreeing on the sign of change. 2.Add a robustness criterion, for example: “robust increase” where at least four out of five GCMs agree on the sign of change. 3.Report both absolute and relative changes. Absolute changes are useful for impacts, while relative changes help compare rare and common drought types. 4.Include regional summary tables for MED, WCE, and NEU. These tables could report changes in mean deficit, duration, frequency, and model agreement by drought type. 5.Avoid interpreting ensemble-mean maps alone where individual models differ strongly. For rain-to-snow droughts in particular, model spread should be shown. 6.For drought timing, directly quantify changes in onset month, termination month, and duration rather than inferring duration from heatmap patterns alone.
Suggestions to discussion section. 1.Discuss more explicitly how threshold choice affects future drought interpretation. This should be a central limitation because threshold methodology can strongly influence projected frequency and duration. 2.Moderate claims about “resilience” or “improvement” in Northern and Western-Central Europe. A decrease in one drought type does not necessarily imply lower overall water-resource risk. 3.Better connect the process-based results to adaptation implications. For example, Mediterranean rainfall-deficit and wet-to-dry droughts imply different management needs than snowmelt or rain-to-snow droughts in northern regions. 4.Discuss the implications of using naturalized/modelled runoff without anthropogenic water withdrawals, reservoirs, irrigation, or land-use change. 5.Expand the discussion of PET uncertainty. Since the manuscript interprets some changes through evaporative demand, sensitivity to PET formulation should be treated as more than a minor limitation. 6.Distinguish clearly between robust findings and exploratory findings. The Mediterranean intensification appears well supported; changes in rare or snow-related drought types are more uncertain and should be discussed accordingly. 7.Reframe the contribution more precisely. Rather than claiming a fully definitive continental-scale assessment, the paper is strongest as a transferable framework for process-based drought typology applied to CMIP5/ISIMIP2b projections over Europe.
Citation: https://doi.org/10.5194/egusphere-2026-973-RC2 - RC3: 'Comment on egusphere-2026-973', Anonymous Referee #3, 28 May 2026
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,108 | 678 | 99 | 1,885 | 286 | 69 | 106 |
- HTML: 1,108
- PDF: 678
- XML: 99
- Total: 1,885
- Supplement: 286
- BibTeX: 69
- EndNote: 106
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
I read this manuscript with interest. It proposes a classification of drought types in Europe, both historically and in future ISIMIP2b simulations using the mHM hydrological model.
However, looking at supplementary figure S1, I am very surprised by the choice of validation basins for the mHM simuations, that constitute the core of the study. No justification is given on the criteria to select these stations for model validation, even though there are now large river discharge databases in Europe that would allow for much more comprehensive validation. As an example, weekly discharge in 17,130 basins in Europe:
do Nascimento, T.V.M., Rudlang, J., Höge, M. t al. EStreams: An integrated dataset and catalogue of streamflow, hydro-climatic and landscape variables for Europe. Sci Data 11, 879 (2024). https://doi.org/10.1038/s41597-024-03706-1
There is also no mention of the size of the basins considered in this work, while the results suggest a homogeneous coverage of the whole of Europe. However, with data such as EOBS at 0.125° spatial resolution, which is also not homogeneous in quality across Europe, it is obvious that this data is more tailored for large basins.
The stations shown in Figure S1 reveal a strong location bias towards Central Europe. However, this happens to be the region with the highest density of stations in EOBS data (see for example https://doi.org/10.1002/joc.7269 and https://doi.org/10.1029/2017JD028200, notably Figure 1 of the latter).
In fact, this figure S1 shows that there are only 7 basins in France and 11 in Spain (with some nested within the Ebro), and none in Italy or Greece. However, the results place a strong emphasis on Mediterranean regions. I find this problematic because we see low performances for some basins in Spain.
It seems to me that this validation should be reinforced to give more confidence in the results, particularly in the basins of southern Europe where it is well known that, in addition to climatic influences, there are various factors that play a role in the reduction of low flows:
Vicente-Serrano, S. M., Kenawy, A. E., Peña-Angulo, D., Lorenzo-Lacruz, J., Murphy, C., Hannaford, J., Dadson, S., Stahl, K., Noguera, I., Fraquesa, M., Fernández-Duque, B., & Domínguez-Castro, F. (2025). Forest expansion and irrigated agriculture reinforce low river flows in southern Europe during dry years. Journal of Hydrology, 653, 132818. https://doi.org/10.1016/j.jhydrol.2025.132818