the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Feb 2026
Seasonal shifts in drought characteristics and their drivers in Italian alpine catchments under climate change
Abstract. Drought is an increasingly important hazard in Alpine regions, where snow dynamics strongly influence river flow and support reservoir filling, irrigation, tourism, and ecosystem sustainability. Declining snow and warmer winters, which increase rainfall at the expense of snowfall, are shifting Alpine catchments toward lower-elevation hydrological regimes. This study examines shifts in drought seasonality and drivers in the Adige River basin under future climate conditions. Hydrological simulations for a reference period (1989–2018) and three future horizons (near 2020–2049, mid 2045–2074, and far 2070–2099) under climate scenarios are used to analyse drought drivers, timing, duration, severity, and intensity across catchments of different elevations. Results show that high-elevation catchments progressively shift from snowmelt- and glacier-driven droughts toward rainfall-deficit dominance. Drought peaks exhibit a bimodal pattern, occurring primarily in spring and summer, with summer peaks projected to shift earlier under future warming. Drought severity rises by more than 60% in winter and spring at high elevations, while duration remains stable. These findings highlight the need for adaptation strategies that account for both seasonal and driver-specific responses to sustain Alpine water systems.
- Preprint
(2436 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 20 Mar 2026)
- RC1: 'Comment on egusphere-2026-464', Anonymous Referee #1, 15 Feb 2026 reply
-
RC2: 'Comment on egusphere-2026-464', Andrea Galletti, 25 Feb 2026
reply
The authors present an assessment of seasonal shifts in drought characteristics and drivers with reference to the Adige catchments and 10 sub-basins therein.
While the evolution of drought metrics in relation to climate change has been explored from several work, the very concept of driver-based categorization of hydrological drought is, to this day, severely under-explored while it offers great potential for drought prediction and management. For these reasons, I believe the work is timely and a valuable contribution to the scope of NHESS.
What I liked about the paper is the attempt at a spatial characterization of drought metrics and driver in a relatively narrow and morphologically complex domain. The amount of climate models employed in the analysis is sufficient, and not overwhelming compared to the size of the study. The results presented in the paper are consistent with existing literature and reflect the specificities of the study area. I think this is a (rightfully!) very ambitious work, and believe that its potential has not been fully tapped, for reasons i will try to enumerate in the following. I will follow the order of the paper and try to convey the relevance of each comment directly inside it. I think the work requires major revision to clarify some methodological choices, their implications in the results and to ultimately amplify its impact and relevance to the community.
1. The abstract ends talking about the need for adaptation strategies. While obviously true, and while it is good to mention the practical implications of a study, I think these are way beyond the scope of this paper. Indeed, I would not request the Authors to detail on what kind of strategies should be developed based on their findings. More simply, I suggest focusing on the implications of the results actually presented (shifts in drivers and metrics).
2. Though understandable, I don't really like the term "lower-elevation hydrological regime". It implies a contrast with a high-elevation regime that not all readers might be familiar with. I suggest using a more general terminology (such as rainfall-dominated or pluvio-nival, depending on the Authors' opinion and the message they want to convey)
3. The introduction presents some of the "ingredients" to the study: climate change effects on mountain river regimes, their implications for drought in mountain regions. It then cites (55-65) different drought types (per Brunner 2022 and other studies), sharply concluding that the spatial and elevational dependence of such drivers is underexplored. Finally, some emphasis is put on cryospheric processes related to drought and their uncertain evolution under continued warming. While agreeing with all of the above, I think that a clear need to understand drivers and metrics of drought in relation to climate change does not yet emerge from the introduction. Put in other words, what is the value of *jointly* analysing drought drivers, metrics, and the effects that climatic change has on them? this would offer a stronger lead into the research questions.
4. line 90: i think the area and minimum elevation reported are related to the Adige at its outlet, not at the Villa Lagarina section.
5. lines 99 and 105 state that the work is focused on 11 natural catchments. While being in stark contrast with line 94, and with my knowledge of the presented sub-basins, perhaps here the Authors are meaning that they do not model anthropogenic alterations as a design choice?
6. I would specify that Crespi 2021 does not contain observations, rather an observation-based gridded dataset, though it is somewhat implied by providing the 250m resolution.
7. line 116: A-HDT seems to reflect Hydrological Digital Twin (?) instead of Digital Hydrological Model (DHM?). perhaps check.
8. lines 132-135: From my understanding, streamflows are moving averaged (30 days, i assumed centered to 15 before and 15 after?). Then, a climatological "reference year" of Q20s is computed. I do not understand if each Q20 is really not referring to its own calendar day or also considers the +-15 window, as stated in line 134. Would this mean tha tthe same window is adopted before for smoothing and later for picking the Q20? in any case, the explanation could be clearer. Also, the authors do not mention (or I missed it) whether the thresholds are computed with respect to observation or to modelled data. I assume the latter, since it is done for historical and future time windows. Do the authors imply that the variable threshold changes for each future time window? if so, did they explore the implications on their result, as opposed to a fixed window computed on the historical 30 years?
9. line 136+: I think the choice of the threshold (quantile level) and moving average window have a clear effect on the statistics of the droughts that they identify. They affect the severity, here defined as the deficit from the reference, as well as the population of events (duration-frequency). Moreover, droughts driven by different processes do have different hydrological signatures, therefore even the dominant drivers might be affected by this choice. While the values chosen (Q20 and 30 day moving average) are absolutely reasonable, I suggest to not downplay their implications, and rather present them as a reasonable design choice.
10. line 143: this definition of intensity is not the most common. Here, it is presented as the absolute minimum flow during each event, regardless of when it happens. More often, intensity is presented as the (average or largest) gap between reference curve and event. Perhaps the authors could elaborate on this choice and its implications. Then, I really do not understand the concept of "highest intensity" provided soon after in line 144, and whether it is used in the paper or not. Finally, if runoff is given in mm it should already be considered catchment-specific, so i do not see the point of normalising it again vs the catchment area.
11. line 148: this is a key point. I hope I am getting the definition correctly and I apologise for my next comments if i missed parts or the entirety of it. Here, every drought is basically concentrated in its "peak low flow day". However, this does not take into account the natural hydrologic regime of the river. For the Adige, lowest flows are in late winter-spring; this means that any event that spans these seasons, even if it started months earlier or continued months later (as testified by durations of 60-120 days in fig 4), it would inevitably (or very likely) be attributed to a low-flow month. The main issue is that the majority of the deficit does NOT happen in these months, but rather in months of high (reference) flows, where the absolute deviation from reference can be much larger. Honestly, I do not really agree with this way of treating droughts as "point events", as too many subsequent results on seasonality depend on this and are treated as if each event was related to a specific season, and to that only. I encourage the Authors to explore the implications of this design choice, and if they choose to retain, more careful wording of the results is necessary, as they are mostly related to peak low flow and not to the full extent of the drought. Likewise, even the link with drivers resents of the definition, and this should be clearly acknowledged when describing them. Alternatively, if the authors wish to preserve the single-day attribution of drought seasonality, perhaps as an alternative is to place the peak when the maximum relative deviation from reference flow occurs (which still doesn't talk about deficit magnitude, but I have nothing else on my mind now).
12. paragraph 3.1.2 basically re-explains Brunner 2022. To make it more relevant, i would contrast-relate all decisions with the specificities of the Adige. Was any decision parameter changed with respect to Brunner's method? Are all quantities involved in the decisions basin-wide? what is the normal climatology of the Adige so that seasons such as wet-dry-snow can be put into more context?
13. the Authors say (179) that results are contrasted with precipitation and ET. is ET potential? how is it computed? And importantly, why not also contrasting them with temperature or temperature anomaly as T directly drives 2 decisions?
14. no letters in Fig 2 sub-panels. also y axis is "propability". Also I'd like having some (not all) months on the x axis, instead of calendar days, since the discussion is entirely seasonal. seasons would do as well.
15. line 187: what supports the statement that single annual peak is characteristic of low elevation catchments? Does it imply that the peak related to spring melt is missing? Also, the behavior is not retained in RCP8.5 and even in 4.5 the variability envelop seems to still hint at a dual peak. Indeed, the Fersina also sees decent snow cover during winter, albeit it has no glaciers.
16. line 202: How can the adopted definition for drought timing capture the likely occurrence of droughts caused by earlier snowmelt?
17. line 214: The term "persistency" is not appropriate since what is being presented is a PDF generated by 1-day events. even if some high probability persists, it is due to two events whose peak is nearby, not to a single, persistent, one.
18. San Bernardo Rabbi has 2 b's.
19. line 218: again, if we are talking about 1-day events, how does this link to the "month with highest drought probability"? For this, I'd rather analyze the number of "in drought" days that each month has -regardless of which events they belong to-, to provide a more objective measurement.
20. for all section 4.2, I have found it pretty hard to follow all the changes and metrics with reference to one or more catchments. Between that, the RCP's, the seasons and the metric-change dicotomy, it can get overwhelming. have the authors thought about presenting their results by elevation classes (presented in the case study but rarely used after), instead of by-catchment? of course the images can stay unchanged.
21. line 235: see my comment about annual hydrologic regime. being close to the annual minimum does not imply that the drought is severe.
22. line 248: the wording "maximum duration" is misleading, since median is being presented
23. the "agreement" definition is not clear: is it about the magnitude of change (line 248) or just about the sign of change (line 258)?
24. what about flipping the median change legend? it seems counter-intuitive to have the decrease upwards and vice-versa, though i understand it comes from more-water, less-water reasoning. Anyway, what does 100% decrease mean? can the Authors provide the formulas for these % changes in the Appendix and refer to them in the Methods?
25. lines 269-271: in two mid-elevation catchments (1500-2000) i expect that snow (or snowmelt) deficits play the main part, while T driven deficits contribute more marginally. the final sentence of line 271-272 (about elevation) is kind of hanging there. finally, minimum severity in winter droughts happens not only because of lower PET stress, but also because of very low absolute deviation from reference, as per my main comment.
26. lines 275-277 is true only if T>0, which should be clarified.
27. line 281: same as comment 25. Also, i see a potential issue here. If an event is attributed to spring due to peak occurrence, then even the whole severity of the event, likely happening in summer, is attributed to the spring season. The Authors should clarify this implication in their analysis.
28. line 288-289: while useful to compare across different basins, % changes are very hard to relate to drought magnitude, which varies wildly across seasons. While debating absolute changes might be beyond the scope of the work, i encourage the Authors to stress this aspect: for this same reasons, % changes should be discussed within-season and not contrasting different seasons (or do it, with care).
29. line 300: under RCP8.5 high warming is expected, again the attribution to winter and spring seems to point at the definition of peak drought, while i would expect the event to be more severe in the melt peak season (may-july)
30. line 304+: why are the results for the mid future presented here? and only here?
31. line 311: i struggle to understand the definition of minimum flow used. The values are very low and I failed at converting them to m3/s meaningfully. can the Authors clarify the timestep related to these streamflow values, and any additional elaboration occurred to them? This also helps clarifying the comparison being done across catchments of very different size
32. lines 327-328 discuss drivers, though what have been presented so far are only metrics.
33. line 335: Maybe just my understanding,: the term "relative" would make me think of a piechart with all drought drivers. while the relative (spatial) importance is shown, Fig 8 only displays the dominant type.
34. line 353: nov-mar is perhaps the snow season? the Authors called it wet.
35. line 353-360 kind of goes over what each drought type means, without relating it to the results. I think this should either be paired with some results or be merged with what is already present in Section 3.1.2.
36. line 358+ again the concern that i raised in comment 16. I do not see an immediate link between drought occurrence and peak day as used here.
37. A general comment looking at figure 8 and at the decision tree is that, whenever there is no P deficit, most streamflow deficits will be redirected (peak..) towards nov-mar: together with T<0, this makes me expect a great dominance of cold snow season droughts, though i would not be so sure all of them are.
38. line 384: the second sentence is unclear to me/incomplete.
39. line 388-391: the relationship of streamflow drought trends with P trends in a snow-dominated area is not warranted, unless the catchment goes fully rain-dominated, hopefully not the case.
40. lines 393-395 seem in contrast with 395-397, where the former state a decrease in summer-autumn deficit due to convective rainfall, while the latter state a negative correlation between temperature increase and precipitation change (i.e., P decrease?).
41. line 404-405 is true in RCP 4.5 but not in 8.5
FORMATTING (citations, typos): check lines 36, 52, 109-110, 120 (semi distributed?), 135,336 (snowmelT)
Citation: https://doi.org/10.5194/egusphere-2026-464-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 148 | 74 | 17 | 239 | 10 | 15 |
- HTML: 148
- PDF: 74
- XML: 17
- Total: 239
- BibTeX: 10
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The study analyzed the hydrological drought characteristics (duration, severity, intensity, and occurrence) under climate change scenarios (RCP 4.5/8.5).
According to my revision of the article, it lacks severe rigor, mainly in the methods section (see my comments below). My recommendation in the current status is that the article should be rejected.
Major revision
The study does not provide a clear definition of how the researchers are measuring drought. Are they using a streamflow drought index? I think not. Because the figures of the results display absolute values in millimeters (mm). Or, are they just analyzing the streamflow variable? This step is crucial for what is then analyzed.
The results indicate that the methodology is incomplete. The methodology does not fully address the findings presented in the results. For example, the core of the article is the analysis of the characteristics of drought (duration, intensity, severity, etc), but none of them are declared in the methods section.
All the above makes it very hard to follow what then is coming in the results and discussion.
Other comments.
In the title, instead of using just "drought", it should be replaced by "hydrological drought". Because drought is a broader concept, of which hydrological drought is just one type. Besides, in this study, hydrological drought is analyzed.
Methods
Drought identification.
What is the specific definition of hydrological drought in this case? Instead of just referencing it, you need to provide a clear definition here. The text discusses a time series variable, but it does not specify which variable is being referred to. Should I assume that the variable in question is a streamflow anomaly?
Hydrological drought classification
It is not clear what the purpose of using this classification is. Then in the results, these types of drought are called drivers and analyzed as such. How could the variable analyzed be the driver of the same event (drought)?