Historical and future transitions between opposing UK hydrological extremes

Armitage, Rachael; Magee, Eugene; Chevuturi, Amulya; Chan, Wilson; Hannaford, Jamie

doi:10.5194/egusphere-2026-3154

Preprints

https://doi.org/10.5194/egusphere-2026-3154

Preprints

11 Jun 2026

| 11 Jun 2026

Historical and future transitions between opposing UK hydrological extremes

Rachael Armitage, Eugene Magee, Amulya Chevuturi, Wilson Chan, and Jamie Hannaford

Abstract. Transitions between droughts and floods can exacerbate the impacts of the individual events and present a complex challenge for water resource management: sudden or frequent transitions between dry and wet conditions can negatively impact water availability, water quality, agricultural productivity, and cause damage to water infrastructure. Despite these potentially severe impacts, such transitions have, until recently, received less attention in the international literature than their component extremes. In the UK, there has been no systematic assessment of the occurrence of transitions, despite growing interest given a series of recent swings between floods and droughts.

Given this gap, we assess present-day and future transitions using national river flow and precipitation projections from the enhanced future Flows and Groundwater (eFLaG) dataset for 1989-2079 over 200 UK catchments. We identify transition events as the period between consecutive yet opposite extremes at seasonal timescales, using a threshold method to demarcate extreme wet and dry events for both river flow and precipitation to understand the magnitude, duration and frequency of both hydrological and meteorological transitions.

Our results reveal the spatial distribution of transitions in the UK, with higher intensity transitions in the north-west and longest durations in the south-east. We compare hydrological and meteorological transitions and find similar spatial patterns between the two but a stronger seasonality and generally shorter durations for meteorological transitions. Most regions of the UK are projected to see an increase in transition magnitude, a decrease in duration and therefore more intense transitions in the future. The south-east sees the largest decreases in transition duration under future projections. The frequency of hydrological transitions is projected to increase in the north-west and in all regions for meteorological transitions. Our findings demonstrate the risk of increasing hydrological volatility across the UK, with implications for water resources management and climate adaptation.

Received: 01 Jun 2026 – Discussion started: 11 Jun 2026

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.

Download & links

Preprint (PDF, 1937 KB)

Supplement (351 KB)

Download & links

Rachael Armitage, Eugene Magee, Amulya Chevuturi, Wilson Chan, and Jamie Hannaford

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-3154', Anonymous Referee #1, 25 Jun 2026
This manuscript presents the first systematic, nationally-consistent assessment of transitions between hydrological (and meteorological) wet and dry extremes across the UK, using the eFLaG dataset spanning 1989–2079. The authors define transitions as the period between consecutive opposite-sign extremes identified via SSI/SPI threshold crossings at a monthly timescale, characterizing them by magnitude, duration, intensity, and frequency for both dry-wet and wet-dry directions. Historically, the authors find a clear northwest-southeast gradient, with the NW showing higher-intensity, shorter-duration, more frequent transitions and the SE showing longer, lower-intensity transitions (linked to groundwater-dominated catchments). Under future RCP8.5 projections, most regions show increasing transition magnitude and intensity and decreasing duration, with the largest duration decreases in the SE and the largest frequency increases in the NW. With a few relatively minor revisions, I think the manuscript will present the main results of the study in a way that is easier for the reader to digest.
Major comments:
The paper repeatedly contrasts itself with “drought-to-flood” transition studies (Götte and Brunner, 2024; the companion Yang, Villarini & Yang CONUS paper; Anderson et al., 2025) but the distinction between a “state-switch” (SPI/SSI threshold crossing at seasonal timescales) and a “true” DtF/FtD transition (event-based, sub-seasonal, 90-day window) could be made more explicit and earlier in the Introduction, not just in section 4.1. This distinction is now considerably sharper given the newly available companion results: Yang et al. (2025, in review for Earth's Future) report that across 2,000+ CONUS stations, drought-to-flood (DtF) events occur nearly three times more frequently than flood-to-drought (FtD) events, with DtF transitions occurring within ~41 days versus ~57 days for FtD, and with DtF intensity significantly higher than FtD (Mann-Whitney U, P<0.05). This is a striking asymmetry. By contrast, the present UK paper finds wet-dry and dry-wet transitions to be broadly symmetric in magnitude/duration/intensity changes (though differing in seasonality). Is this symmetry a property of the seasonal SSI/SPI approach (which by construction looks at smoothed, longer-duration anomalies, merging events separated by gaps) versus the asymmetry that emerges when looking at sub-seasonal/rapid threshold-crossing events with explicit drought/flood definitions tied to the 2-year-recurrence bankfull discharge and 7-day low-flow statistics? The authors should discuss this methodological dependency as a substantive finding, not just a caveat. It may be the single most important point of contrast between the UK seasonal-switch framework and the rapid-whiplash framework, and is directly relevant to readers trying to reconcile the two literatures.

The large overlapping ensemble spreads (Figures 8–9) are acknowledged, but the paper would benefit from a more quantitative treatment — e.g., what fraction of ensemble members agree on the sign of change for each metric/region? A simple “fraction of RCMs showing increase” map or table would help readers gauge robustness more than visual inspection of overlapping boxplots. As written, statements like “transition magnitude is projected to increase for most regions” are somewhat qualitative given the stated caveat that “internal variability… dominate the projected signal.”

I think the main paper would benefit from figures showing a comparison of hydrological transition metrics of SimRCM minus SimOBS so the reader can clearly see where the RCM best matches simulated observations, ad where the two diverge.

Another set of panels on existing figures could show the difference in median NF and FF transition metrics as compared to BL metrics. It takes a lot of work looking back and forth between the boxplots to currently see this. The boxplots are very helpful for looking at the range and variability of metrics by scenario, but simple panels showing difference in medians would draw out some of the key results of the paper.

The Discussion (line 386-387) flags that temperature/evaporative demand differences could explain hydrological vs. meteorological transition differences, and suggests SPEI as future work. Given that PET-driven changes are central to the future projections (warming climate, RCP8.5), this omission is now harder to justify as a minor limitation. The Swain et al. (2025, Nature Reviews Earth & Environment) review, which the authors already cite, explicitly builds its entire “hydroclimate whiplash” framework around SPEI rather than SPI precisely because “definitions incorporating evapotranspiration-related ('demand side') variables are increasingly favored,” and demonstrates that potential evapotranspiration substantially amplifies whiplash increases over continents relative to precipitation alone, particularly at high latitudes. Given that Swain et al. report interannual whiplash increases of 52% over land at 3°C warming using SPEI, and that the present paper's RCP8.5 scenario corresponds to a comparable warming trajectory, the omission of an SPEI-based comparison is a substantive gap rather than a future-work afterthought. At minimum, the Discussion should explicitly frame the SPI/SSI-only approach as likely underestimating future transition magnitude/frequency changes relative to what an SPEI-based analysis would show, citing Swain et al.'s finding that evapotranspiration-inclusive metrics are “not detected in precipitation-only metrics.”

Minor comments:
Line 32-37: The list of “pseudonyms” (hydrological whiplash, hydroclimate volatility, drought-to-flood events) is useful, but could also mention “compound drought-flood events” (Matanó et al.) and “drought termination” (Parry et al.) explicitly here since both are discussed later. Could help orient readers to the terminological landscape upfront.

Line 66-69: The assumption that “following a flood, sufficient water would remain stored to buffer the impacts of a subsequent drought” is asserted but not cited. Is there a reference for this being a stated assumption in the literature, or is this the authors' inference?

Line 150-156: The description of SimObs vs SimRCM and their respective roles is clear, but it might help to add a sentence on the magnitude of “inherent biases” mentioned (line 155-156) — is there a citable RMSE/bias statistic from Hannaford et al. (2023) that could be quoted to give readers a sense of scale?

Line 184-185: “we use the 3-month accumulation period (SPI-3 and SSI-3)” — please justify this choice more explicitly. A sensitivity check (or at least a citation) showing that SPI-1/SSI-1 or SPI-6/SSI-6 would yield qualitatively different transition counts/durations would strengthen the methodological robustness, especially given how central the “moderate/severe/extreme” thresholds and accumulation period are to all subsequent results.

Line 257-262: “A few catchments in the southeast show notably high duration and low intensity transitions… situated in the chalk aquifer north of London” — this is a nice mechanistic explanation. Could the authors quantify this more precisely, e.g., by overlaying baseflow index (BFI) or an aquifer productivity classification on Figure 5, or computing a correlation between BFI and transition duration across all 200 catchments? This would convert an anecdotal/visual observation into a more quantitative structural finding.

Line 422-423: “Infrastructure needs to be designed not only for more extreme or intense wet and dry events than currently specified, but also have flexibility built in to adapt for transitions” — this is a good high-level recommendation, but feels slightly generic; given the SE-specific finding about a shift from multi-annual to seasonal transitions (line 314-318), could the Discussion include a more targeted recommendation specific to chalk/groundwater-dominated catchment management (e.g., implications for conjunctive use schemes, aquifer recharge management)?

Line 488-492 (Data Availability): Good that eFLaG and SPI/SSI source data are referenced, but given the extensive transition-identification methodology developed here, would the authors consider releasing the derived transition event catalogue (onset/end dates, magnitude, duration, intensity per catchment/scenario) as a data product? This could substantially increase the paper's value for follow-on UK water-management research .
Citation: https://doi.org/10.5194/egusphere-2026-3154-RC1
RC2: 'Comment on egusphere-2026-3154', Anonymous Referee #2, 14 Jul 2026

General comments
This manuscript presents the first systematic assessment of present-day and future hydrological transitions (between wet and dry extremes) across the UK. It sufficiently identifies the current gap in the literature and sets out to address this gap in a scientifically coherent and robust way, sufficiently addressing a relevant scientific question within the scope of HESS. It presents a novel approach using existing datasets and adopts scientifically robust methodologies that are reproducible, and generally states valid methods and assumptions, with possible minor additions (S1). The findings and conclusions are substantial and the results generally support the conclusions made, albeit these could be further refined with minor amendments (S2). The title clearly reflects the contents of the paper, however it could be improved to mention the analysis of meteorological transitions as well as hydrological transitions, for clarity (S3). The overall presentation of the methods and results is clear and well structured; however, some minor clarifications/amendments could substantially improve the coherence of the results (S1). The number and quality of references is appropriate, and supplementary material is provided; however, this could include additional code/datasets/catalogue of results (S1).
Specific comments
S1: Methods and assumptions, reproducibility
Equations and definitions (or references to literature where they are defined) could be added for the calculation of the standardised indices, the fitted distributions, and for the four-transition metrics themselves (duration, magnitude, intensity, frequency).
L179: "By extending our methodology to include calculating SSI, we reduce model differences in comparison to transitions identified from raw modelled flows." There are both benefits and disbenefits to using raw modelled time series versus standardised indices. These should be explained here and the choice of SSI appropriately justified. Assumptions inherent to the use of standardised indicators, in particular, temporally stationary distribution parameters, should be stated explicitly as a potential limitation. The SSI/SPI parameters are calibrated on 1982–2018 and applied through to 2080 under RCP8.5. If winters become wetter and summers drier, citing Hannaford et al. (2025), then under a fixed baseline, future winter SSI-3 will drift systematically positive and summer SSI-3 systematically negative, even with no change whatsoever in variability. The fixed baseline is a defensible choice, arguably the right one for water-management, but the authors could either provide a diagnostic separating the mean-shift contribution from the variability contribution (e.g. recomputing transitions with a transient/moving-window standardisation and comparing frequency, duration and seasonality against the fixed-baseline results), or state explicitly in the Discussion what a fixed baseline can and cannot demonstrate.
L183: "SPI…was fitted with the standard gamma distribution instead". Can the authors provide justification or a reference for this methodological decision?
L185: "we use the 3-month accumulation period (SPI-3 and SSI-3) to focus on meteorological and hydrological transitions at a seasonal time scale". Can the authors provide justification for the choice of a 3-month accumulation period? Has a sensitivity analysis with other accumulation periods (e.g. SPI-1/SSI-1, SPI-6/SSI-6) been undertaken? Given how central the accumulation period is to all subsequent transition counts and durations, this would strengthen the methodological choice.
Threshold sensitivity. While the threshold level method has not been applied to the time series specifically, has a sensitivity analysis been undertaken on the different levels/deviations from "normal conditions"? Does this influence the results, and more specifically across both meteorological and hydrological variables? Note that Section 2.3 (L214-216) defines three threshold categories (moderate, severe, extreme), but the figures use ±1.6 only. Either present the other two, the supplement would be an appropriate place or justify ±1.6 directly.
Model selection: Why were these specific hydrological models selected? Is it possible to provide a quantitative analysis, or a reference to a paper that quantifies, how well SimObs performs in relation to observations? A citable RMSE or bias statistic would give readers a sense of the scale of the "inherent biases" mentioned at L155-156.
L202: "Here, we aim to focus on transitions on a broadly seasonal or longer timescale rather than sudden transitions over shorter timescales of days or weeks as these are more relevant to water resource management, where such significant shifts have the greatest impact on water availability planning." While this is valid justification, a more quantitative explanation of "seasonal or longer" timescales would be beneficial. This also bears on the chalk-aquifer interpretation (L257-262): are those long durations genuinely slow, connected transitions, or simply long gaps in catchments where extremes are rare?
Ephemeral and intermittent flows: while transition events are identified as the period between the end of one and the start of another consecutive extreme, have the authors considered aspects such as ephemeral or intermittent flows in the definition of a dry extreme? The Tweedie distribution accommodates zero flows, but the implications for SSI behaviour in intermittent catchments are not discussed.
The monthly timescale means this analysis does not capture flashy responses. This could be caveated adequately as it may have implications, particularly for the NW upland catchments where the fastest transitions are reported and where the monthly resolution is most likely to be limiting. Is this something for future research?
L126: "for both time slices and transient changes, using multiple hydrological models". Clarity is needed on what time slices and transient changes are analysed.
Abstract, L15: "enhanced future Flows and Groundwater (eFLaG) dataset for 1989-2079 over 200 UK catchments", while L135 states "river flow projections across the UK, for 1980-2080". These should be made consistent.
There is currently no Code Availability statement. The Data Availability section points to eFLaG and the UK Water Resources Portal. Given that the transition identification and characterisation framework is itself a substantive contribution of this paper, this would be beneficial to future research.
S2: Results and findings
Is "volatility" the right term? Section 1.1, L49 uses "hydrological volatility" with reference to Swain et al. (2025a) — but that framework is built on SPEI rather than SPI, because evapotranspiration-related variables are increasingly favoured, and Swain et al. show that including PET substantially amplifies projected whiplash increases relative to precipitation alone. Given that evaporative demand is central under RCP8.5 could the Discussion section explain that an SPI/SSI-only approach may understate future changes in transition magnitude and frequency.
Section 1.1, L72: "wet-dry transitions can still be particularly impactful, for example wet-dry transitions across the growing season and wildfire season have been linked to an elevated wildfire risk (Swain et al., 2025b)". This link operates through evaporative demand, humidity and other hydroclimatic variables, not through hydrological variables alone, worth making explicit, as it reinforces the SPEI point above.
Figs. 4 and 8 boxplots are slightly difficult to read. Difference-in-medians panels (NF−BL, FF−BL) alongside the boxplots would draw out the key results effectively.
Uncertainty quantification would improve the robustness of the projected transitions. The authors state at L307-310 that internal variability and inter-model uncertainty dominate the projected signal, however explicit quantification of uncertainty may be beneficial in the Results.
Area-averaging across UK regions does not show the full catchment-by-catchment spread. How does this affect the interpretability of the results, and what about inter-region heterogeneity?
The SimObs/SimRCM comparison is currently visual only (L247-256, Fig. 5). Difference maps (SimRCM − SimObs) for each metric would show immediately where the RCM-driven runs match the observation-driven simulations and where they diverge, and a compact table of regional bias would be beneficial.
The BFI link can easily be made quantitative (L257-262). The mechanistic explanation for the long-duration, low-intensity SE catchments is one of the most interesting findings. BFI is already plotted in Fig. S1, so a scatter of BFI against mean transition duration across all 200 catchments, with a perhaps a correlation coefficient, would improve this finding quantitatively.
Modal season may be unstable where transitions are rare (Fig. 7). In low-frequency SE catchments over a 30-year SimObs baseline, the modal season may rest on very few events. Reporting the underlying event counts (marker size) would let readers judge which parts of Fig. 7 are well constrained.
The state-switch vs. rapid-transition distinction could appear earlier in the paper. This is currently confined to Sect. 4.1 (L425-438), but may be useful in the Introduction to compare this work with Götte and Brunner (2024), Matanó et al. (2024) and Hammond et al. (2025). The smoothing inherent in SSI-3, limit detection of rapid, asymmetric behaviour that event-based methods capture. This is an interesting methodological finding in its own right and could be presented as such.
S3: Title suitability
The title accurately reflects the hydrological analysis, but does not mention meteorological transitions. This is clear from the abstract itself, L18: "both river flow and precipitation to understand the magnitude, duration and frequency of both hydrological and meteorological transitions". I would suggest amending the title to address both, for example "Historical and future transitions between opposing hydrological and meteorological extremes in the UK" or similar.
Technical corrections
L47/48: full stop needed after (IPCC, 2023).
L81: spelling correction to "Millennium".
L431: "methods that capture much more rapid changes transitions", suggest removing the word changes.
L448: "Chan et al., 2025" — 2025a or 2025b?
Abstract, L16: study period given as 1989–2079, while Sect. 2.1 gives 1980–2080 for the data.
Figs. 8 and S4: rotated region labels are very small and hard to read at print size, abbreviations or larger figures/text might help.

Citation: https://doi.org/10.5194/egusphere-2026-3154-RC2

Rachael Armitage, Eugene Magee, Amulya Chevuturi, Wilson Chan, and Jamie Hannaford

Supplement

https://doi.org/10.5194/egusphere-2026-3154-supplement

Rachael Armitage, Eugene Magee, Amulya Chevuturi, Wilson Chan, and Jamie Hannaford

Viewed

Total article views: 137 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
87	41	9	137	12	6	7

HTML: 87
PDF: 41
XML: 9
Total: 137
Supplement: 12
BibTeX: 6
EndNote: 7

Views and downloads (calculated since 11 Jun 2026)

Month	HTML	PDF	XML	Total
Jun 2026	60	22	4	86
Jul 2026	27	19	5	51

Cumulative views and downloads (calculated since 11 Jun 2026)

Month	HTML	PDF	XML	Total
Jun 2026	60	22	4	86
Jul 2026	27	19	5	51

Viewed (geographical distribution)

Total article views: 126 (including HTML, PDF, and XML) Thereof 126 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Jul 2026

Short summary

Transitions between droughts and floods can exacerbate the impacts of the individual events and provide challenges for water resource management. Here, we use river flow and climate projections to assess transitions in the UK over historical and future periods. Our results reveal the characteristics of and spatial pattern in current transitions, and how these are projected to change in the future, with wider implications for water resource management.


Total:	0
HTML:	0
PDF:	0
XML:	0