Drought hazard assessment across Sweden's diverse hydro-climatic regimes

Canedo Rosso, Claudia; Nyberg, Lars; Pechlivanidis, Ilias

doi:https://doi.org/10.5194/egusphere-2025-1843

Preprints

https://doi.org/10.5194/egusphere-2025-1843

Preprints

06 May 2025

| 06 May 2025

Drought hazard assessment across Sweden's diverse hydro-climatic regimes

Claudia Canedo Rosso, Lars Nyberg, and Ilias Pechlivanidis

Abstract. In recent years severe droughts have significantly impacted the water-dependent sectors including water supply, agriculture, energy, and forestry. This study aims to assess the meteorological, agricultural and hydrological drought risk in Sweden, with a focus on hazard assessment using a set of indicators, including the Standardized Precipitation Index (SPI), Standardized Precipitation and Evapotranspiration Index (SPEI), Standardized Soil Moisture Index (SSMI), and Standardized Streamflow Index (SSI). The indicators were computed at time scales of 1, 3, 6, 12, and 24 months using daily precipitation, evapotranspiration, soil moisture, and streamflow simulations (1975–2021) from the national S-HYPE hydrological model at about 13 km² spatial resolution for almost 40,000 sub-catchments. The drought events were next identified and characterized based on their intensity, duration, and frequency, following this a trend analysis was performed for these indicators and events. Assessing the spatial similarities in soil moisture anomaly led to the categorization of the Swedish river systems into five clusters further improving the understanding of the identified spatial variability of drought indicators and trends. Our findings showed drier conditions and an increasing frequency of droughts in central- and south-eastern Sweden. Significant negative trends in these regions, along with increasingly wet conditions in northern and western Sweden, were observed when analysed using the SPEI, SSMI, and SSI. Based solely on precipitation (SPI), similar significant wetter conditions were observed in northern and western Sweden; however, no significant decreasing precipitation trends were found in parts of central-eastern Sweden and Gotland Island. The findings of this study can improve climate services and early warning systems of droughts, better understanding the link to sectoral impacts and guiding mitigation practices and adaptation policies.

Received: 18 Apr 2025 – Discussion started: 06 May 2025

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, 2499 KB)

Supplement (1371 KB)

Download & links

Claudia Canedo Rosso, Lars Nyberg, and Ilias Pechlivanidis

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-1843', Antonia Longobardi, 13 Jun 2025

The paper presents a modelling application which, although not new in terms of methodology, is very well organised and written, follows a proper and logical flow of practical steps, is exhaustively commented in the framework of current specific literature and provides a comprehensive analysis of the state of drought hazard in Sweden. In my opinion the manuscript should be accepted as is.

Citation: https://doi.org/10.5194/egusphere-2025-1843-RC1
- AC1: 'Reply on RC1 Assoc. Prof. Antonia Longobardi', Claudia Canedo Rosso, 29 Jul 2025
  
  We sincerely thank Assoc. Prof. Antonia Longobardi for the positive and encouraging feedback. We appreciate the recognition of the manuscript's structure, clarity, and thorough contextualization within the current literature, as well as the relevance of the analysis for understanding drought hazard in Sweden. We are pleased that the referee finds the manuscript suitable for publication without further revision.
  
  Citation: https://doi.org/10.5194/egusphere-2025-1843-AC1
RC2:
'Comment on egusphere-2025-1843', Anonymous Referee #2, 25 Jun 2025
Using hydrologic data from simulations (S-HYPE model; L96), the manuscript provides detailed insights into trends in Sweden's water budget components—precipitation, evapotranspiration, and surface and soil storages—at a spatial resolution of approximately 40,000 subcatchments. A suite of standardized drought indicators is employed, including the Standardized Precipitation Index (SPI), Standardized Precipitation–Evapotranspiration Index (SPEI), Standardized Soil Moisture Index (SSMI), and Standardized Streamflow Index (SSI).
Drought events of varying severities (e.g., extreme, severe, moderate) are identified based on when these indicators fall below predefined critical thresholds (Table 1). Due to the distinct temporal variability of each indicator, the identified drought periods differ accordingly (L217, L429). In addition to SPI, SPEI, SSMI, and SSI, droughts are also detected based on Soil Moisture Anomaly (SMA) values, using the same critical thresholds (Table 1; Figure 2).
Monthly SMA values are used to perform a cluster analysis, resulting in the identification of five regions exhibiting distinct monthly SMA patterns (Figure 1). Overall, the detected trends in drought indices (Figures 4, 5), drought periods, and regional patterns align with prior literature on droughts in Sweden (L207; L288; L330).
I find the study insightful, particularly the identification of regions with homogeneous monthly SMA variations and the trend analysis of standardized drought indicators. However, I have some concerns regarding the methodology used to identify droughts based on each standardized drought indicator independently. While fluctuations in these indicators reflect seasonal transitions from wetter to drier conditions, they do not, in principle, constitute sufficient evidence for the occurrence of drought (see comments #1, #2, and #3).
For a more rigorous assessment, I recommend exploring ways to integrate information from multiple drought indicators to identify drought events, using documented historical droughts as a reference. Alternatively, the authors could consider shifting the study’s focus slightly—from identifying droughts to analyzing trends in established drought indicators and the occurrence of periods with low values, without labeling them as “droughts.” This could be complemented by qualitatively noting the observed agreement between such low-indicator periods and historic droughts, as already discussed in the current manuscript. The latter approach would primarily require a revision in terminology.
A more detailed list of comments is provided below.

Major comments:

The authors identify drought events based on the occurrence of negative Soil Moisture Anomaly (SMA) values (L200-209; Fig. 2), but this approach may not be rigorous in principle. According to their own definition (L25), a drought is a period in which water deficits have significant negative impacts on society and ecosystems. However, given the way SMA is defined (Eq. 1; L115-124), it primarily captures monthly fluctuations of average soil moisture around the annual mean value. Seasonal variability naturally leads to SMA fluctuations, but minima in the time series do not necessarily indicate drought conditions. For instance, consider a hypothetical region unaffected by droughts but characterized by strong seasonal climate variations: SMA would still show fluctuations, yet its minima would reflect merely dryer-than-average months—not actual droughts. Within the framework of the current analysis, the regional classification based on distinct patterns of monthly SMA (Fig.1) is compelling. However, the authors should clarify that using SMA alone to identify droughts is not yet a standardized approach and requires further validation. Acknowledging this is particularly important given the ongoing discourse surrounding drought definitions in the scientific community. Practical suggestions for addressing this are outlined in comment #2.

As discussed above, a period characterized by markedly negative SMA values may not, in itself, constitute a drought event. Consequently, using SMA as a standalone criterion for drought identification should be supported by a validation procedure that compares low-SMA periods with documented historical droughts. The authors briefly address this comparison (L207; Fig. 2) by applying SMA thresholds of -1.5 and -2 to define “severe” and “extreme” droughts, respectively (Fig. 2 caption). Most of the identified “severe” and “extreme” events—highlighted with black rectangles—correspond to historic droughts (L207–208), with exceptions such as the 2002 and 2006 events. I recommend either expanding this analysis—for example, by providing more information on historical droughts and assessing correlations between drought impacts and the severity of soil moisture anomalies—or revising the terminology in the manuscript. In the latter case, the authors could note that “low-SMA periods below -2 show good agreement with historic drought occurrences, suggesting that monthly SMA may have potential for drought detection; however, rigorous validation of this approach lies beyond the scope of the present study.”

L126, L216: Similar concerns raised in comments #1 and #2 apply to the use of some of the standardized drought indicators considered. Among all indicators, SPI and SPEI are widely recognized in the literature, while SSI and SSMI offer a direct representation of drought impacts through its reflection of surface and soil water availability, respectively. My primary criticism lies in the methodological approach: the authors identify drought events independently from each indicator’s time series, assigning drought periods based on when each individual index drops below a predefined threshold (Tab. 1; L217, L224, L429). However, aside from SPI and SPEI, these indicators reflect symptoms—not causes—of drought. Their fluctuations alone do not necessarily imply the presence of a drought, as illustrated in the hypothetical case-study region described in comment #1. Rather than treating indicators independently, I recommend conducting an a-posteriori analysis of documented historical droughts to investigate how patterns across indicators align during known events. Can the interplay between SPI, SPEI, SSMI, and SSI be used synergistically to more robustly identify droughts? The observed discrepancies in drought periods across indicators (L217, L429) suggest that, when considered alone, none provides sufficient information for unambiguous drought classification. While the manuscript appears to touch on this integrative analysis (L228–233; L431), its treatment is preliminary and partly relegated to the supplementary materials. I strongly encourage the authors to elevate this discussion to the main text, as it could form the basis of a substantial contribution to the field of drought monitoring. Alternatively, if such integration is beyond the intended scope, the authors should avoid referring to low-indicator-value periods as "droughts" and revise the manuscript accordingly to reflect this distinction.

L158, L159, L160: The physical interpretation of the “accumulated drought intensity” and “accumulated weighted drought severity” metrics remains unclear. Could the authors clarify what these metrics represent, conceptually and physically? Moreover, as “accumulated drought intensity,” “accumulated drought severity,” and “accumulated weighted drought severity” are inherently dependent on the length of the record period, they are expected to increase with longer record periods. This raises the question: why not normalize these metrics, for example, by expressing them as annual means? Doing so may facilitate more meaningful interannual or interregional comparisons. Additionally, it appears that these metrics are introduced but not subsequently used or discussed in later sections of the manuscript. If they do not contribute to the core analysis, their inclusion should be reconsidered or more clearly justified.

Sub-section 4.2 of the Discussion offers a broad overview of current limitations in drought characterization and emphasizes the need for more advanced methodologies capable of capturing the diversity of drought impacts across spatial and temporal scales. While this is an important topic, such content may not be entirely appropriate for the Discussion section, which is typically reserved for interpreting and contextualizing the study’s own findings. In this case, the inclusion of this subsection appears somewhat out of place, as the manuscript does not specifically explore or propose novel approaches for drought characterization. If the authors wish to retain this perspective, it might be better positioned in the Introduction—provided they clearly frame it as background context for the current study. Otherwise, this section could be considerably shortened or omitted for clarity and focus.

L96-112: It is unclear whether the authors conducted the S-HYPE simulations themselves or relied on results from previous studies. If the simulations were performed by the authors, they should provide all necessary details to ensure reproducibility, ideally supported by access to the relevant datasets and code. Alternatively, if the simulations were sourced from prior studies, the authors should clearly cite these sources and ensure they are appropriately referenced in the Data and Code Availability Statement. Clarifying this point is essential for transparency and reproducibility.

Fig. 3: The authors should clarify how drought severity was calculated. Is it defined as a cumulative value over the considered period for each standardized drought indicator and timescale?

Minor comments:

Fig. 3: for the first “severity” category, express the range as "<-30" instead of "from -infinite to -30".

L126-128: Include the equations for SPI, SPEI, SSMI, and SSI in the main text.

L146, L152: I think that “drought characteristics” would be better than “drought parameters”, when referring to drought duration, severity, intensity, and frequency.

L443: As the study does not deal with drought forecasting, I disagree that its findings can enhance early warning systems.

The Acknowledgments and Data and Code Availability Statement are missing.
Citation: https://doi.org/10.5194/egusphere-2025-1843-RC2
- AC2:
  'Reply on RC2', Claudia Canedo Rosso, 29 Jul 2025
  Publisher’s note: this comment was edited on 1 August 2025. The following text is not identical to the original comment, but the adjustments were minor without effect on the scientific meaning.
  We thank Referee #2 for the thoughtful and constructive review. We appreciate the positive assessment of the manuscript’s insights, particularly regarding the regional patterns in soil moisture anomalies and the trend analysis of drought indicators. We also acknowledge the reviewer’s concerns regarding the methodology for identifying droughts based on individual indicators, which we plan to fully address in the revised manuscript. We also plan to provide a more cautious interpretation and clearer terminology to improve the clarity and robustness of the analysis. Below, we address each of the specific comments in detail.
  Responses to major comments:
  1.
  We thank the referee for raising this important point regarding the interpretation and use of Soil Moisture Anomaly (SMA) as a drought indicator. As defined in the manuscript (L25–26), “drought is a natural hazard characterized by periods of drier-than-normal conditions with wide-ranging and cascading impacts across societies, ecosystems and economies.” In line with this, we identified drier-than-normal periods using a suite of drought indicators, including SPI, SPEI, SSMI, SSI, and also including SMA, which provides valuable insights into long-term soil moisture variability.
  However, we acknowledge the reviewer’s concern that SMA minima may, in some cases, reflect seasonal variability rather than drought conditions with societal or ecological im pacts. We agree that using SMA alone to identify droughts is not yet a standardized approach and requires further validation, as the reviewer rightly points out. Please refer to response to the 2^nd major comment for a detailed explanation.
  2.
  In response to the 2^nd major comment, we will follow the reviewer’s suggestion and revise the manuscript to clarify that the identification of droughts based solely on negative SMA values should be interpreted with caution. Specifically, we will adjust the terminology throughout the text to refer to “low-SMA periods” rather than labelling them as “droughts” unless supported by historical validation.
  We will also highlight that many of these low-SMA periods show good agreement with known historical droughts (as noted in L206-209 and Figure 2). As suggested by the referee, we have carried out a systematic comparison between the identified low-SMA periods and documented historical drought events in Sweden. This is presented in the Table AC1 in supplementary material which will also be introduced in Section 3.1. This comparison includes an expanded discussion of the temporal correspondence and agreement between severe/extreme SMA events and known droughts, providing a more rigorous assessment of SMA’s potential for drought detection in the Swedish context.
  The systematic comparison supporting the potential of low SMA values for drought detection is primarily based on documented drought events in Sweden from 1975 to 2021 (please see Table AC1 in the supplementary material).
  3.
  We fully acknowledge the concerns raised regarding the independent treatment of standardized drought indicators and the limitations of using each indicator in silo to define drought periods. In response, we will revise the manuscript to clarify this methodological point and adjust the terminology accordingly. Specifically, we will avoid referring to low-indicator-value periods as "droughts" unless supported by historical validation.
  4.
  We agree that the metrics “accumulated drought intensity”, “accumulated drought severity”, and “accumulated weighted drought severity” lack clear physical interpretation and are not sufficiently integrated into the core analysis. Given their dependence on record length and the absence of follow-up discussion in the manuscript, we will remove these metrics from the study to maintain focus and clarity. We appreciate the suggestion and believe this adjustment will strengthen the overall coherence of the manuscript.
  5.
  We agree that Section 4.2, while addressing an important and timely topic, extends beyond the direct scope of our study. To maintain clarity and focus in the Discussion, we will shorten this section and limit it to aspects that are directly relevant to our findings. This revision will help ensure that the discussion remains aligned with the objectives and contributions of the present work. Section 4.2 will be restructured as outlined in the supplementary material.
  6.
  The S-HYPE simulation data used in the study are available from SMHI, part of the national hydrological service. We will cite the original source of the data in the revised manuscript and provide the following Data and Code Availability statement to support transparency and reproducibility. Please see the Response to the 5^th minor comment for a detailed response.
  7.
  Specifically, for each standardized drought indicator and selected timescale, we identify drought events as periods during which the indicator value remains less than or equal to -1. The duration of a drought event corresponds to the number of consecutive time steps (months) during which this condition is met.
  The severity of a drought event is then calculated as the sum of the indicator values over this consecutive period—i.e., the cumulative sum of all values ≤ -1 during the event. This means that severity reflects both the intensity and length of the drought: a longer or more intense event will result in a higher cumulative severity value. This calculation is performed separately for each standardized indicator and timescale used in the analysis.
  We will clarify this explanation in the revised manuscript to ensure it is clearly understood by the reader.
  Minor comments:
  In Figure 3, we will express the range as "<-30" instead of "from -infinite to -30".
  
  Considering that SPI, SPEI, SSMI, and SSI equations are standard, we will include them in the Supplementary material SM1.
  
  “Drought parameters” will be used instead of “drought characteristics”, when referring to drought duration, severity, intensity, and frequency.
  
  We agree that the study does not directly contribute to operational early warning systems, as it focuses exclusively on the characterization of historical drought conditions. However, we believe that the results can still provide indirect support for the development or refinement of early warning systems by improving the understanding of how different drought indicators behave across regions and timescales. In particular, the identification of spatial patterns, trends, and indicator thresholds may help inform which variables are most useful for early detection or risk mapping in future system design. We will revise the manuscript to clarify this distinction and avoid overstating the study's relevance.
  
  We will add Acknowledgments and Data and Code Availability Statement in the revised manuscript, as outlined below:
  
  Acknowledgments:
  The study was supported by the Centre of Natural Hazards and Disaster Science (CNDS) and the Centre for Societal Risk Research (CSR) at Karlstad University. We gratefully acknowledge Swedish Meteorological and Hydrological Institute (SMHI) for providing the climatological and hydrological simulations utilized in this research.
  Data and Code Availability:
  The HYPE model code, which was used in the national S-HYPE model setup, is available from the HYPEweb portal (https://hypeweb.smhi.se/model-water/; (SMHI, 2025b)). The meteorological data used for driving the S-HYPE model can be obtained upon contact with SMHI, and the hydrological data used are available from the Vattenwebb portal (https://vattenwebb.smhi.se/; (SMHI, 2025c)).
  The R scripts used to compute the drought indicators, along with the resulting datasets, are openly available at a FAIR-aligned public repository via Zenodo: https://doi.org/10.5281/zenodo.16539104 (Canedo Rosso, 2025).
  References
  Canedo Rosso, C.: Drought indicators across Sweden (1975-2021) (R version 4.2.0 or later), https://doi.org/10.5281/ZENODO.16539104, 2025.
  SGU: Lowest measured groundwater level (Lägst uppmätta grundvattennivåer), 2025. https://www.sgu.se/grundvatten/grundvattennivaer/om-grundvattennivaer/lagst-uppmatta-grundvattennivaer/
  SMHI: The precipitation and humidity climate of Sweden during the vegetation period (Nederbörds och humiditetsklimatet i Sverige under vegetationsperioden). Report 46 on Meteorology and Climatology by Bertil Eriksson, 1986. https://www.diva-portal.org/smash/get/diva2:948001/FULLTEXT01.pdf
  SMHI: Weather and Water 1994 (Väder och Vatten 1994). Report by Carla Eggertsson Karlström, 1994 https://www.smhi.se/download/18.18f5a56618fc9f08e832ee6f/1717826327996/Väder%20och%20Vatten%20Nr%201-Nr%2013%20Väderåret%201994.pdf
  SMHI: Water 2003 (Vattenåret 2003). Report 18 by Torbjörn Jutman, 2004 https://www.smhi.se/download/18.598b468c190544b12305785/1719987136086/faktablad_vattenaret_2003
  SMHI: July 2006 - Temperature and precipitation (Juli 2006 - Temperatur och nederbörd), 2006a. https://www.smhi.se/klimat/klimatet-da-och-nu/manadens-vader-och-vatten-i-sverige/manadens-vader-och-vatten-i-sverige/2006-08-02-juli-2006---temperatur-och-nederbord
  SMHI: July 2006 - Water flow, groundwater and groundwater (Juli 2006 - Vattenföring, markvatten och grundvatten) , 2006b. https://www.smhi.se/klimat/klimatet-da-och-nu/manadens-vader-och-vatten-i-sverige/manadens-vader-och-vatten-i-sverige/2006-08-02-juli-2006---vattenforing-markvatten-och-grundvatten
  SMHI: Historical drought periods (Historiska torrperioder), 2025a. https://www.smhi.se/kunskapsbanken/hydrologi/historiska-torrperioder
  SMHI: Hypeweb. https://hypeweb.smhi.se/, 2025b.
  SMHI: Vattenwebb. https://vattenwebb.smhi.se/, 2025c.
  SVT: Fem av de värsta somrarna de senaste 100 åren, 2018.https://www.svt.se/nyheter/inrikes/fem-av-de-varsta-torkorna-de-senaste-100-aren
  
  Citation: https://doi.org/10.5194/egusphere-2025-1843-AC2

Claudia Canedo Rosso, Lars Nyberg, and Ilias Pechlivanidis

Supplement

https://doi.org/10.5194/egusphere-2025-1843-supplement

Claudia Canedo Rosso, Lars Nyberg, and Ilias Pechlivanidis

Viewed

Total article views: 560 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
454	91	15	560	45	21	42

HTML: 454
PDF: 91
XML: 15
Total: 560
Supplement: 45
BibTeX: 21
EndNote: 42

Views and downloads (calculated since 06 May 2025)

Month	HTML	PDF	XML	Total
May 2025	96	33	4	133
Jun 2025	87	16	5	108
Jul 2025	65	16	2	83
Aug 2025	104	24	4	132
Sep 2025	102	2	0	104

Cumulative views and downloads (calculated since 06 May 2025)

Month	HTML	PDF	XML	Total
May 2025	96	33	4	133
Jun 2025	87	16	5	108
Jul 2025	65	16	2	83
Aug 2025	104	24	4	132
Sep 2025	102	2	0	104

Viewed (geographical distribution)

Total article views: 580 (including HTML, PDF, and XML) Thereof 580 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 12 Sep 2025

Short summary

Severe droughts have increasingly affected water supply, farming, and forestry in Sweden. This study explored how drought risks have changed over time and across regions using meteorological and hydrological data. Results showed that droughts are becoming more frequent in central and south-eastern Sweden, while northern areas are getting wetter. These insights can support early warnings and help guide decisions on drought preparedness and climate adaptation.


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