the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Sep 2025
Compound droughts have become more widespread globally
Abstract. Compound droughts, in which meteorological, hydrological, and agricultural drought occur simultaneously, have a more substantial impact on ecological and socio-economic systems than any drought type alone. Yet, the global patterns – and particularly the multi-decadal trends – of these co-occurring droughts remain poorly understood. We analyse the global co-occurrence of meteorological, agricultural, and hydrological droughts from 1961 to 2020 using run theory and empirical drought indices from three global hydrological models (GHMs), each forced with three different meteorological forcing datasets. Our results indicate that compound droughts and their characteristics show distinct spatial patterns, varying across different hydrological regions. The findings suggest that compound droughts have become more widespread globally, particularly in the hydrological regions (i.e., hydrobelts) near the equator and in the southernmost regions, where the number of days under compound drought has increased rapidly over the past 60 years. Our results also show that compound drought behaviour in the boreal hydrobelt significantly differs from all other hydrobelts, showing a general wetting trend and no increase in compound droughts. We also, however, find a high uncertainty in the ensemble, highlighting a need to global hydrological modelling aimed toward droughts specifically. The results provide valuable global insights into complex phenomena of compound droughts, helping in drought preparedness actions and planning.
- Preprint
(2547 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-3909', Anonymous Referee #1, 04 Nov 2025
-
AC1: 'Reply on RC1', Marko Kallio, 01 Jan 2026
Response to reviewers
We wish to thank all three reviewers for their contributions and for their insights. Going over and answering these comments have significantly streamlined and strengthened our manuscript. It is also notable that there is a significant overlap between the reviewer comments, highlighting the biggest weaknesses of our original manuscript. Therefore we list here the main changes we have made to the manuscript, before responding to your review comments in detail.
Main changes to the manuscript:
- We have revised the language used in the manuscript. Most notably, we change the language from compound droughts and talk about concurrent droughts instead. We also make the aim of our study more clear.
- We provide information of validation of the ensemble we use, both towards replicating droughts as well as the primary outputs of the models (streamflow, soil moisture).
- We have moved the ENSO analysis to the appendix and refer to it as an example of additional analyses our results could support.
- We provide a more detailed description of our handling of uncertainty through using an ensemble approach.
- We have shifted from Hydrobelts to using Holdridge Life Zones as the areas we use to summarise our results.
- We have revised our discussion to better reflect our results.
- We have made numerous small edits to the text based on the minor comments from each reviewer.
REVIEWER # 1
Dear reviewer, thank you for taking your time to review our work and for your insights. Please find below our responses to the points you have raised.
Comment 1.1: First, I would like to question the authors about their motivation for conducting a study on droughts at the global scale. Studies of this nature commonly require methodological simplifications to make their execution feasible, and the present manuscript is no exception. Thus, what is the real advantage of sacrificing a higher level of detail and contextualization of a specific region in exchange for a global study, whose analyses are, by nature, more limited? Who, in fact, benefits from simplified and limited information or analyses about drought events? This is especially relevant considering that droughts are managed locally or regionally, not globally.
Reply 1.1: This is a relevant question, as indeed some simplifications are needed when working on a global scale. At the same time, however, we believe that it is very important and insightful to conduct studies across scales. While on a local scale one can use very accurate and specific methodological and data choices, they give a geographical snapshot of an issue. Global scale studies, however, allow one to put the local scale trends and phenomena into a broader context, help identify hotspot areas where deeper knowledge might be necessary, and also map areas exhibiting similar patterns or trends.
Studies at a global scale are crucial for stakeholders and decision-makers in international organizations and governing bodies who require comprehensive insights for strategic planning resources allocation. Such analyses assist in comparative studies between regions, offering perspectives on vulnerabilities and strategies that can be shared globally. Global studies integrate extensive data from various regions, capturing cross-border influences and the effects of global climatic phenomena like ENSO, which local studies might not encompass fully.
Additionally, considering emerging global challenges such as climate change, studying droughts at a global scale enables us to recognize trends and patterns that transcend local boundaries and provides the insights necessary to devise coherent and coordinated global responses. We thus argue that it is imperative to work on multiple scales to truly comprehend the phenomena—in this case, the occurrence and trends of concurrent droughts—and potential changes over time. For example, if changes in climate were only assessed locally with varying degrees of methods and datasets, we would miss a much-needed overview of the changes.
In the revised manuscript we now briefly justify the need for a global study in the introduction.
Comment 1.2: Additionally, the representativeness of grouping the planet into only nine regions is questionable. Beyond the presumed meteorological similarities, do these regions truly share comparable drought impact patterns? For example, the “SST” region encompasses both the Brazilian semiarid — with an average annual rainfall of approximately 600 mm, potential evapotranspiration close to 2000 mm, and intermittent rivers, making it highly susceptible to droughts — and the country’s South and Southeast regions, characterized by a markedly humid climate that includes some of the continent’s largest rivers, where droughts are not a recurring problem. In other words, these two regions do not share similarities in terms of drought occurrence patterns and drought impacts, so they should not be analyzed in the same group. Therefore, I suggest that the authors consider grouping regions not only based on hydrometeorological similarities but also on drought impact and occurrence patterns.
Reply 1.2: This is a valid observation and we agree that the hydrobelts might not offer the ideal aggregation units for our analysis. We chose the hydrobelts because of their manageable number of zones as well as the underlying idea of hydrobelts combining the river basins with similar hydrometeorological characteristics. However, we agree with the reviewers that a drought analysis with area definitions incorporating such heterogeneous conditions is not feasible. We have therefore revised our analyses using the Holdridge Life Zones that divide a world into zones based on a combination of precipitation, temperature and aridity. This increases the number of zones from 8 (hydrobelts) to 13, using the simplifications from the original Holdridge zones made in Kummu et al. (2021). This increases the nuance we can extract from the analysis. The revised Figure 3 of the manuscript (Figure R1.1) is provided as a supplement to this reply.
Thank you also for suggesting making the zones using drought impacts and occurrence patterns. While it would be highly interesting to assess our findings against drought impacts, it is beyond our scope. The drought impacts can be assessed from multiple perspectives, such as human health, food security, ecosystems, and biodiversity - all of them being relevant. However, there exist no - in our understanding - a comprehensive global dataset for these that would allow us to create a rigorous zoning based on them. Therefore, this would warrant an independent study of its own. Regarding the occurrence patterns: this is what we are actually analysing here and thus in our opinion, it would not be justified to make zoning based on a variable that we would use the zoning to aggregate the results for.
Comment 1.3: Given the profound environmental changes caused by human activities in all parts of the world, does it still make sense to consider droughts solely as a purely natural hazard, completely disregarding the anthropogenic influence on their onset and propagation? I suggest that the authors address this discussion or clearly state the rationale for considering it more useful to consider droughts exclusively as natural hazards.
Reply 1.3: Thank you for the insightful comment! We fully agree with the reviewer that anthropogenic influences are critical to consider when assessing any climate impacts. We apologize for not clearly addressing this aspect in our initial methodology and description of underlying datasets.
Our analysis does consider anthropogenic influences: We utilize outputs from global hydrological models incorporated within the ISIMIP project—a large international model intercomparison initiative. These models use historical anthropogenic forcing, which includes, but is not limited to, climate change, land use change, water use, and the impact of major reservoirs. While models differ in how they incorporate these influences, they collectively provide a comprehensive view of both natural and human-induced factors affecting droughts.
To clarify our manuscript, we've revised the text under Section 2.1 Data to specify how anthropogenic elements are integrated. We've also included an appendix (refer to Table R1 below) detailing how various models account for human activities.
We appreciate the reviewer’s feedback on this important issue, as it has helped us enhance the clarity and accuracy of our manuscript, ultimately offering readers a more complete understanding of drought phenomena.
Table R1.1 Human forcing in ISIMIP 3a global water sector (Tiwari et al., 2025; www.isimip.org).
Model
Groundwater Withdrawal
Irrigation
Land use change effects
Dams and reservoirs
H08
Yes
Yes
Irrigation area expansion
Yes
MIROC-INTEG-LAND
No
Yes
ISIMIP3a time varying land use
Yes
WaterGAP 2-2E
Yes
Yes
Only static land use map is used but irrigation area is changing over time
Yes
Comment 1.4: Another relevant limitation concerns the temporal scale adopted for the calculation of drought indices. The study employs an empirical distribution methodology on a daily scale, later aggregated into three-month periods. What is the justification and advantage of performing a daily analysis for a type of disaster whose natural component is traditionally assessed using standardized indices at monthly scales (e.g. 3, 6, or 12 months)? What is the potential impact of this choice on identifying drought events in regions with highly seasonal rainfall regimes? I recommend converting all variables to a monthly basis and computing standardized indices using a 12-month scale, as this approach may be more appropriate for a global-scale study. Each time step would always consider one full year, which would always capture the rainy season and avoid the potential effect of the seasonality of the precipitation regime.
Reply 1.4: We fully agree that the temporal aggregation is an important factor in our methodology and that we have not adequately addressed it in our manuscript. We recognize that global studies of drought often use monthly timesteps. We chose to go to a higher temporal resolution for two reasons: 1) in tropical environments with rapidly responding catchments, even very short periods of drought can cause an impact (Nauditt et al., 2022), and 2) aggregating time series to monthly temporal resolution may not be able to detect shorter drought periods, particularly if their duration is split between two months.
On the matter of seasonality in the indices, we apologize for the confusion that has arisen from our writing. We do explicitly take this into account by computing a different empirical distribution for every day of the year (i.e. we have 365 distributions), and the drought index is computed using the distribution for that specific day. Therefore, our threshold for a drought is lower when precipitation, soil moisture or streamflow are low (on average), and higher when their values are higher. We attempt to visualise this in Figure R1.2 in the supplement to this reply. In the revised manuscript we have clarified this matter and now highlight in text that our drought indices describe seasonal drought more than annual droughts, although this methodology can recognise longer term droughts as well.
Based on the literature, we chose to use the 90-day accumulation period as a trade-off between 1) our ability to capture the short-term droughts that are important in tropical catchments, and 2) reduce the strong variability in a standardized index with short accumulation periods to produce a clean signal. We also find that, based on literature, the 90-day accumulation period is a common one.
Comment 1.5: It is also important to discuss the effectiveness of the method adopted for identifying drought events. I suggest that the authors include a validation step based on historical drought records, in order to demonstrate the capability of the proposed methodology to adequately detect different types of drought events. The discussions and conclusions about the occurrence of compound droughts are only valid if it is shown, based on observed data, that the proposed approach can effectively identify the different drought types. Based on these observational data, more relevant regions of interest could then be defined for detailed analysis, replacing the nine generic regions currently considered.
Reply 1.5: We agree with the reviewer that one needs to be aware how well the data used corresponds with the actual observations. As we detailed in our response to Comment 1.3, we use outputs from an ensemble of global hydrological models. All these models have been rigorously validated against multiple observations. A summary of the performance of ISIMIP 3a Water sector is available for instance in Heinicke et al. (2024), Veldkamp et al. (2018), and Zaherpour et al. (2018). We discuss these in the answer to Comment 1.6.
A recent article by (Tiwari et al., 2025) validates ISIMIP3a model outputs and their ability to detect Terrestrial Water Storage droughts, including each of the three models we use in our work (although, they only use one climate forcing dataset). They find that, compared to GRACE satellite gravimetry, the ISIMIP3a models underestimate the magnitude of terrestrial water storage drought. But as they show (Figure R1.3 in the supplement to this reply) in their work that at the global scale, the water models correlate strongly with the GRACE signature. Because we do not analyse the intensity or severity of the drought events, and our methodology of fitting the empirical distributions to each individual grid cell independently, a high correlation is the most important metric for our application. Tiwari et al. (2025) show further validation figures in the supplementary information of their paper, showing for instance a high correlation between the terrestrial water storage signal and soil moisture.
Kumar et al. (2022) evaluate an older generation of ISIMIP2a models (including MATSIRO, the land surface component of MIROC-INTEG-LAND, and older versions of WaterGAP2 and H08) for their ability to estimate hydrological drought using 1-month Standardised Runoff Index (SRI). They found that an ensemble mean of Global Hydrological Models (GHM) performed better than individual models and detected 133 SRI events out of 136 events identified from observed data across the 8 major catchments they used. Further, the ensemble identified 233 out of 244 runoff deficit events identified from observed data. However, they do also report that the simulation of event durations were not great.
We thus argue that, globally the ISIMIP model outputs we use do sufficiently well reproduce terrestrial water storages and thus their use is justified in our article.
Comment 1.6: The authors employed a combination of global models with a spatial resolution of 0.5°, using reanalysis data as input. In this context, one may ask: what is the accuracy and performance of this model? How was it calibrated, considering the distinct regional characteristics across the globe? The authors discuss model and index-related uncertainties, but it remains unclear how these could be improved and how the results should be evaluated in light of such uncertainties.
Reply 1.6: Thank you for the very relevant point that we agree was not well covered in our original submission. The validation of the model outputs, for instance in terms of streamflow, are available in a large number of existing studies. For the revised manuscript we have added a clear summary of the validation work of the used models. This is briefly summarised below.
Heinicke et al. (2024) validate ISIMIP 3a models with 644 gauging stations across the globe. They report median correlations of 0.62, 0.61 and 0.43 for WaterGAP 2.2e, H08 and MIROC-INTEG-LAND, respectively. Veldkamp et al. (2018) is a validation study which includes earlier versions of all of the three models we use; H08, MATSIRO (the land surface component of MIROC-INTEG-LAND), and WaterGAP2. They find that H08 and MATSIRO have similar correlation coefficients at approximately 0.5 in human managed catchments, globally. WaterGAP2 performs significantly better with an average correlation coefficient of ~0.7. In near-natural conditions, correlation coefficients are significantly higher; approximately 0.65 for H08 and MATSIRO, and approx. 0.77 for WaterGAP2. Zaherpour et al. (2018) show that, overall, Watergap2 performs the best, in 27 out of 40 catchments across the world, but that there is a large variability in the performance of the models in different environments. All three validation studies show that low-flow periods have lower performance. Furthermore, Gädeke et al. (2020) perform validation of 8 arctic watersheds and find that WaterGAP2 is, overall, the most performing model. In the arctic watersheds, however, MATSIRO (MIROC-INTEG-LAND) performs the best in extreme low-flow simulations (Gädeke et al. (2020) supplementary table 7).
WaterGAP 2.2e has recently been validated by the development team (Müller Schmied et al., 2024), where they use the same ISIMIP 3a climate input datasets as we do. They report relatively high correlation coefficients with a median of 0.78 across the globe.
Both the validations we cite here and our analysis show that there is significant uncertainty, particularly in the low-flow simulations. One of our key conclusions is that there is a need for global modelling targeting drought and low-flow periods specifically.
References
Gädeke, A., Krysanova, V., Aryal, A., Chang, J., Grillakis, M., Hanasaki, N., Koutroulis, A., Pokhrel, Y., Satoh, Y., Schaphoff, S., Müller Schmied, H., Stacke, T., Tang, Q., Wada, Y., and Thonicke, K.: Performance evaluation of global hydrological models in six large Pan-Arctic watersheds, Climatic Change, 163, 1329–1351, https://doi.org/10.1007/s10584-020-02892-2, 2020.
Heinicke, S., Volkholz, J., Schewe, J., Gosling, S. N., Müller Schmied, H., Zimmermann, S., Mengel, M., Sauer, I. J., Burek, P., Chang, J., Kou-Giesbrecht, S., Grillakis, M., Guillaumot, L., Hanasaki, N., Koutroulis, A., Otta, K., Qi, W., Satoh, Y., Stacke, T., Yokohata, T., and Frieler, K.: Global hydrological models continue to overestimate river discharge, Environ. Res. Lett., 19, 074005, https://doi.org/10.1088/1748-9326/ad52b0, 2024.
Kumar, A., Gosling, S. N., Johnson, M. F., Jones, M. D., Zaherpour, J., Kumar, R., Leng, G., Schmied, H. M., Kupzig, J., Breuer, L., Hanasaki, N., Tang, Q., Ostberg, S., Stacke, T., Pokhrel, Y., Wada, Y., and Masaki, Y.: Multi-model evaluation of catchment- and global-scale hydrological model simulations of drought characteristics across eight large river catchments, Advances in Water Resources, 165, 104212, https://doi.org/10.1016/j.advwatres.2022.104212, 2022.
Kummu, M., Heino, M., Taka, M., Varis, O., and Viviroli, D.: Climate change risks pushing one-third of global food production outside the safe climatic space, One Earth, 4, 720–729, https://doi.org/10.1016/j.oneear.2021.04.017, 2021.
Müller Schmied, H., Trautmann, T., Ackermann, S., Cáceres, D., Flörke, M., Gerdener, H., Kynast, E., Peiris, T. A., Schiebener, L., Schumacher, M., and Döll, P.: The global water resources and use model WaterGAP v2.2e: description and evaluation of modifications and new features, Geoscientific Model Development, 17, 8817–8852, https://doi.org/10.5194/gmd-17-8817-2024, 2024.
Nauditt, A., Stahl, K., Rodríguez, E., Birkel, C., Formiga-Johnsson, R. M., Kallio, M., Ribbe, L., Baez-Villanueva, O. M., Thurner, J., and Hann, H.: Evaluating tropical drought risk by combining open access gridded vulnerability and hazard data products, Science of The Total Environment, 822, 153493, https://doi.org/10.1016/j.scitotenv.2022.153493, 2022.
Tiwari, A. D., Pokhrel, Y., Felfelani, F., Elkouk, A., Boulange, J., Gosling, S. N., Hanasaki, N., Koutroulis, A., Mishra, V., Schmied, H. M., Satoh, Y., Ostberg, S., Stacke, T., and Yin, J.: Underestimation of Historical Terrestrial Water Storage Droughts in Global Water Models, Geophysical Research Letters, 52, e2025GL115164, https://doi.org/10.1029/2025GL115164, 2025.
Veldkamp, T. I. E., Zhao, F., Ward, P. J., Moel, H. de, Aerts, J. C. J. H., Schmied, H. M., Portmann, F. T., Masaki, Y., Pokhrel, Y., Liu, X., Satoh, Y., Gerten, D., Gosling, S. N., Zaherpour, J., and Wada, Y.: Human impact parameterizations in global hydrological models improve estimates of monthly discharges and hydrological extremes: a multi-model validation study, Environ. Res. Lett., 13, 055008, https://doi.org/10.1088/1748-9326/aab96f, 2018.
Zaherpour, J., Gosling, S. N., Mount, N., Schmied, H. M., Veldkamp, T. I. E., Dankers, R., Eisner, S., Gerten, D., Gudmundsson, L., Haddeland, I., Hanasaki, N., Kim, H., Leng, G., Liu, J., Masaki, Y., Oki, T., Pokhrel, Y., Satoh, Y., Schewe, J., and Wada, Y.: Worldwide evaluation of mean and extreme runoff from six global-scale hydrological models that account for human impacts, Environ. Res. Lett., 13, 065015, https://doi.org/10.1088/1748-9326/aac547, 2018.
-
AC1: 'Reply on RC1', Marko Kallio, 01 Jan 2026
-
RC2: 'Comment on egusphere-2025-3909', Anonymous Referee #2, 04 Nov 2025
Review of the manuscript „Compound droughts have become more widespread globally“
This manuscript presents the analysis of simultaneous co-occurrence of droughts in different compartments (i.e., precipitation, soil moisture and streamflow) based on simulations of several global models. The study examines the likelihood of such co-occurrences around the globe and quantifies their trends.
Although investigating simultaneous co-occurrences of metrological, hydrological and agricultural droughts can indeed provide valuable insights for water resources management, the novelty of this study given a large body of drought propagation work is not clear. Moreover, lack of validation of model simulations combined with several methodological ambiguities in implementing standardized drought indexes likely result in high spatially-variable uncertainty of detected trends that is neither quantified nor reported in this study. Finally, very general conclusion that lacks any novel insights further indicates that more effort is needed to clarify the novelty of this manuscript before publishing. Please find below my detailed comments.
Major comments
Terminology and novelty: In the manuscript the terms of drought propagation and compounding are used in a very confusing and inconsistent manner. The authors define compounding drought as overlap of droughts in different terrestrial compartments: precipitation, soil moisture and streamflow. However, it is not clear how their study is different from a large body of studies on drought propagation and why it is justifiable and useful to use the compound event framework here. Although I find that studying the question whether or not drought occurs simultaneously in all three compartments and highlighting the regions where this is more likely – very interesting, I believe that the inconsistent terminology used in the manuscript makes its novelty rather unclear. While the introduction is more focused on compounding framework, the discussion refers to propagation. Given ambiguous terminology, the novelty of the study with regard to existing drought propagation studies is not clear, and on the other side it is not clear how the results of the study are comparable with the findings of studies that consider such events from compounding perspective.
Lack of model validation and incomplete model description: Several global models are used to demonstrate probabilities and trends in simultaneous cross-compartment droughts; however, no information is provided on the ability of these models in combinations with the different input datasets to actually simulate droughts. Moreover, very little information is provided about model structure and processes that they are able (or not able to simulate). Considerable uncertainty across models reported in this study, together with the reports on the poor functional performance of global models (Gnann et al., 2024 https://doi.org/10.1038/s44221-023-00160-y) indicates that reliability of these models has to be critically examined. Moreover, it is not clear to what extent human activities are considered in each of the models. All these information is necessary to judge the reliability of presented results.
Drought event definition: I think standardized drought indices can be a very useful tool for detecting drought anomalies. However, I am skeptical that drought events across different compartments can be effectively identified using the same aggregation period for precipitation, soil moisture and streamflow. The latter have clearly different temporal structure from precipitation and associated with much higher persistency. This point has to be clarified and the optimal aggregation period for each compartment has to be critically evaluated. Moreover, the implemented method seems to be suboptimal for dry regions with long consequent periods of no flow as the authors report themselves (Page 16 Line 8-9).
Inconsistent ENSO analysis: The analysis of ENSO suddenly appears in the Methods, only very roughly presented in the results and is neither discussed nor introduced properly. It is not clear why it was selected as a single climatic driver to be examined and it is not clear what was the rationale for this analysis. Consider either extending the scope to possible large-scale climatic drivers of drought co-occurrence across compartments or remove this part from the manuscript for clarity.
Conclusions: The conclusion section is very vague and very general. It barely touches on the actual results obtained in this study. The implications mentioned there are also very general and could have been resulted from almost any drought study again indicating that the novelty of the presented study is not quite clear. Please revise.
Detailed comments
Page 1 Line 29-31: From this sentence it sounds as precipitation deficit propagates into precipitation drought. Please revise.
Page 2 Line 5-10: At this point the authors should clearly clarify why is it worth to consider simultaneously co-occurring droughts across different compartments as compounding events and what is an added value of using such frameworks over drought propagation concept. I believe this can clarify the novelty of this study. Moreover, towards the Discussion section the authors converge back to discussing propagation rather than compoundness (e.g., Page 12 Line 9) making the added value of implementing compounding framework even more unclear.
Page 2 Line 18-20: It would be more instructive to summarize in this part what is known about co-occurrence of droughts across different compartments globally, instead of merely stating what was done without reflecting on their findings. Please revise.
Page 2 Line 25-40: I definitely agree that it is very related to the analysis of drought types. This also makes the novelty of this study rather unclear (see also my major comment). Introduction should identify existing gap more clearly and show how the analysis performed in this study aims to close it.
Page 3 Line 5-7: From this statement it is not clear what additional insights the global analysis will bring compared to the array of the regional studies listed above. Please clarify. See also my comment above on the need of clarifying the existing gap more clearly to highlight the novelty of this study.
Page 3 Line 7: It would be useful for the reader to already highlight in the Introduction that this is a global modeling study and reflect on the current accuracy of these tools and possible challenges associated with using them for global drought analysis.
Page 3 Line 21: This comes rather surprising. From the Introduction it was not clear that the connection to large-scale atmospheric drivers is in the scope of this manuscript. Without it the choice of this particular driver is not justified, i.e., it is not clear why only this one was used and no other alternative was considered.
Figure 1: Portions of panel B and C conserving events are identical. Please correct or clarify how these two analyses differ. Please also explain all acronyms in the caption.
Page 5 Line 1-12: The description of the models and the input is rather poor (see major comment). Moreover, it is not clear whether and how the models were calibrated/validated.
Page 5 Line 11-12: What about all other glaciers abundantly present in the European Alps, High Asia Mountains and in the mountainous ranges in North and South Americas? How is the glacier problem treated there? Please clarify.
Page 5 Line 22-27: Please clarify why two different periods had to be considered here.
Page 5 Line 23: Is the same accumulation period justifiable given the known difference in temporal persistence of precipitation, soil moisture and streamflow? Please provide rationale for your choice of particularly 3 moths as well.
Page 6 Line 1-2: Please provide reference for this statement.
Page 6 Line 5-10: Moreover, given this description it does not seem to me that these are events, these seem to be anomalies (see also my major comment).
Page 6 Line 13-15: References are needed to motivate the choice.
Page 6 Line 26: Run theory is mentioned on several occasions in the text, but is actually never properly introduced. Please add.
Page 7 Line 5-10: This is a very confusing description. Please clarify how the start and the end was selected and which compartment was chosen to define the overlapping periods.
Page 7 Line 20-25: This description is unclear. How droughts and phases was matched? Please clarify.
Page 8 Line 2: Why the word events is in brackets? Because it is actually not an event, but rather an anomaly?
Page 8 Line 5-10: It is not clear which count is meant here, seems that the definition differs, please revise.
Section 3: The explanation of the observed geographical patterns is missing here. Please provide it here or in the discussion section.
Page 9 Line 14: Are these patterns agree with observations? See Seo et al., 2025 doi: 10.1126/science.adq6529
Figure 3: Please explain all abbreviations.
Page 11 Line 11-15: These findings seem to be a direct result of the definitions used in this study. Please clarify.
Page 11 Line 25: Isn’t it very much expected that there are more precipitation droughts? I think it is well established that large portion of droughts never propagates (e.g., Brunner and Chartier-Rescan, 2025 https://doi.org/10.1029/2023GL107918).
Page 12 Line 11-12: I miss a discussion here on why this is observed in these regions.
Figure 4: Model uncertainty and regional variability has to be displayed in the aggregated bars of the panel G. Please revise
Page 14 Line 19: Please revise the sub-title, it is not the most eloquent choice.
Page 15 Line 24-29: I find this discussion not very helpful. Merely stating which threshold was chose in a different study is not very helpful for putting the results of this study into perspective. Please clarify here how the choice of the threshold in this study might have affected the outcomes.
Page 16 Line 8-9: If this is the known problem, why this method was not applied in this study? I would assume that intermittent flows is quite a common challenge across all arid regions. Are any of the observed trends possible artifacts of this limitation?
Page 16 Line 20-25: Which calibration is meant here? No calibration was mentioned in the Methods section. Please clarify and revise the Method section if needed.
Page 16 Line 25-30: Does it mean that the observed trends might be the artefacts of the dataset used? Please clarify.
Page 18 Line 24-26: I do agree that this is out of the scope, but reporting the results on drought trends without providing any statement on the reliability of the models for drought simulation is not very helpful (see my major comment).
Page 19 Line 2-3: I am not quite convinced that this is true. At least in Europe and Northen America soil moisture-based drought monitors are used for governance. Please reflect and revise.
Editorial comments
Page 1 Line 30 and elsewhere: please double-check the punctuations. On several occasions the commas are missing in the manuscript.
Page 5 Line 3: Why three? This is not yet clear at this point. Please revise the sentence.
Page 5 Line 11: Please use consistent acronyms in the text (ICE vs Ice)
Page 5 Line 13: This sentence makes no sense, please revise
Page 11 Line 20: Something is missing in this sentence.
Citation: https://doi.org/10.5194/egusphere-2025-3909-RC2 -
AC2: 'Reply on RC2', Marko Kallio, 01 Jan 2026
Response to reviewers
We wish to thank all three reviewers for their contributions and for their insights. Going over and answering these comments have significantly streamlined and strengthened our manuscript. It is also notable that there is a significant overlap between the reviewer comments, highlighting the biggest weaknesses of our original manuscript. We therefore list here the main changes we have made to the manuscript, before responding to your review comments in detail.
Main changes to the manuscript:
- We have revised the language used in the manuscript. Most notably, we change the language from compound droughts and talk about concurrent droughts instead. We also make the aim of our study more clear.
- We provide information of validation of the ensemble we use, both towards replicating droughts as well as the primary outputs of the models (streamflow, soil moisture).
- We have moved the ENSO analysis to the appendix and refer to it as an example of additional analyses our results could support.
- We provide a more detailed description of our handling of uncertainty through using an ensemble approach.
- We have shifted from Hydrobelts to using Holdridge Life Zones as the areas we use to summarise our results.
- We have revised our discussion to better reflect our results.
- We have made numerous small edits to the text based on the minor comments from each reviewer.
REVIEWER #2
Comment 2.1: This manuscript presents the analysis of simultaneous co-occurrence of droughts in different compartments (i.e., precipitation, soil moisture and streamflow) based on simulations of several global models. The study examines the likelihood of such co-occurrences around the globe and quantifies their trends.
Reply 2.1: Thank you for reviewing our manuscript and lending your expertise with us. Please find our responses to the points you have raised below.
Comment 2.2: Although investigating simultaneous co-occurrences of metrological, hydrological and agricultural droughts can indeed provide valuable insights for water resources management, the novelty of this study given a large body of drought propagation work is not clear. Moreover, lack of validation of model simulations combined with several methodological ambiguities in implementing standardized drought indexes likely result in high spatially-variable uncertainty of detected trends that is neither quantified nor reported in this study. Finally, a very general conclusion that lacks any novel insights further indicates that more effort is needed to clarify the novelty of this manuscript before publishing. Please find below my detailed comments.
Reply 2.2: Thank you for summarising the main points. Please find a summary of our responses to these here, while more detailed replies to each comment can be found below.
Novelty: Novelty of our work comes from a combination of 1) A global study with daily timestep, 2) long timeseries cover 60 years, 3) analysis covering the co-occurrence of drought with three standardised indices, and 4) an ensemble approach with model runs from state-of-the-art global water models.
Validation: Validation of the models was indeed missing from the manuscript. We add validation information with regards to primary outputs from the three models. We also refer to studies validating the performance against Terrestrial Water Storage drought as well as hydrological drought (Standardised Runoff Index).
Uncertainty: We indeed do not discuss the uncertainty at lengths in our results section. Instead, we had opted to discuss the uncertainty in the discussion section 4.2 and in Figure 5. Fig5A shows a bivariate plot of median probability of compounding and the coefficient of variation of all models in the ensemble across the globe. The Figure also attempts to communicate uncertainty in the compounding across different hydrobelts in panel B, and panels C-F show differences in the SSI in different locations across the globe. Furthermore, Figure 3A shows the time series for each of the ensemble members in addition to the ensemble median in an attempt to show uncertainty in our results. Nonetheless, we agree that the uncertainty can be elaborated on more in the results section, and therefore we discuss it further in the results. We also provide additional figures in the supplementary information, showing the spread and uncertainty for all maps in Figure 2.
Major comments
Comment 2.3: Terminology and novelty: In the manuscript the terms of drought propagation and compounding are used in a very confusing and inconsistent manner. The authors define compounding drought as overlap of droughts in different terrestrial compartments: precipitation, soil moisture and streamflow. However, it is not clear how their study is different from a large body of studies on drought propagation and why it is justifiable and useful to use the compound event framework here. Although I find that studying the question whether or not drought occurs simultaneously in all three compartments and highlighting the regions where this is more likely – very interesting, I believe that the inconsistent terminology used in the manuscript makes its novelty rather unclear. While the introduction is more focused on compounding framework, the discussion refers to propagation. Given ambiguous terminology, the novelty of the study with regard to existing drought propagation studies is not clear, and on the other side it is not clear how the results of the study are comparable with the findings of studies that consider such events from compounding perspective.
Reply 2.3: Thank you for your observations with regards to a confusion between the compound event framework and drought propagation. The terms are indeed related, and we also acknowledge that in the introduction where we discuss concepts that are adjacent to our application of the compound event framework. In the revised manuscript, we have modified the introduction and now state clearly our focus on concurrent droughts, not propagation or the mechanisms in which propagation occurs. We also take care to edit the discussion so that it does not appear that we are discussing propagation instead of compounding.
Novelty: To our knowledge, there is no other study which combines 1) daily temporal resolution, 2) three indices, 3) consistent ensemble framework where each model is internally consistent with closed water balance, and 4) a long 60-year temporal period. We have made sure that the novelty of our study is clearly communicated in the introduction section. We further agree that, while drought propagation is an important adjacent concept to the compound event framework, it has a different focus.
Comment 2.4: Lack of model validation and incomplete model description: Several global models are used to demonstrate probabilities and trends in simultaneous cross-compartment droughts; however, no information is provided on the ability of these models in combinations with the different input datasets to actually simulate droughts. Moreover, very little information is provided about model structure and processes that they are able (or not able to simulate). Considerable uncertainty across models reported in this study, together with the reports on the poor functional performance of global models (Gnann et al., 2024 https://doi.org/10.1038/s44221-023-00160-y) indicates that reliability of these models has to be critically examined. Moreover, it is not clear to what extent human activities are considered in each of the models. All these information is necessary to judge the reliability of presented results.
Reply 2.4: We agree that general information on the models we use and their performance was an important omission from our manuscript, and we have included this information in the manuscript. For your convenience, we add the information here in our response.
Model structures for the three selected models are shown in Müller Schmied et al. (2025), for their model versions participating in the previous ISIMIP phase 2. Schmied et al. (2025) report that there are only minor differences in the models participating in phase 3 experiments which we use as the source of our data. These changes are documented in Heinicke et al. (2024). Table R2.1 below also shows which anthropogenic influences are included in each of the models (Tiwari et al., 2025; www.isimip.org). We provide a short description of our chosen models in section 2.1 where we introduce the data we use.
There are deficiencies in the GHMs, as the study of Gnann et al. (2023) clearly shows, stemming from the simple model structures and for the fact that a single model structure likely cannot model all hydrological environments across the globe. This is the reason we use an ensemble to base our estimates on. The ensemble mean (or median, as we use) is shown by many authors to give a reasonable estimate in global studies in hydrology (see e.g. Kumar et al. (2022) or Kallio et al. (2021), among others). We agree that one needs to be aware how well the global model outputs correspond with the actual observations. All these models have been rigorously validated against multiple observations. A summary of the performance of ISIMIP 3a Water sector is available for instance in Heinicke et al. (2024), Veldkamp et al. (2018), and Zaherpour et al. (2018). We discuss these in our answer to reviewer #1, but we copy the text from there for your convenience.
Table R2.1 Human forcing in ISIMIP 3a global water sector (Tiwari et al., 2025; www.isimip.org).
Model
Groundwater Withdrawal
Irrigation
Land use change effects
Dams and reservoirs
H08
Yes
Yes
Irrigation area expansion
Yes
MIROC-INTEG-LAND
No
Yes
ISIMIP3a time varying land use
Yes
WaterGAP 2-2E
Yes
Yes
Only static land use map is used but irrigation area is changing over time
Yes
A recent article by (Tiwari et al., 2025) validates ISIMIP3a model outputs and their ability to detect Terrestrial Water Storage droughts, including each of the three models we use in our work (although, they only use one climate forcing dataset). They find that, compared to GRACE satellite gravimetry, the ISIMIP3a models underestimate the magnitude of terrestrial water storage drought. But as they show (Figure R2.1 in the supplement to this reply) in their work that at the global scale, the water models correlate strongly with the GRACE signature. Because we do not analyse the intensity or severity of the drought events, and our methodology of fitting the empirical distributions to each individual grid cell independently, a high correlation is the most important metric for our application. Tiwari et al. (2025) show further validation figures in the supplementary information of their paper, showing for instance a high correlation between the terrestrial water storage signal and soil moisture.
Kumar et al. (2022) evaluate an older generation of ISIMIP2a models (including MATSIRO, the land surface component of MIROC-INTEG-LAND, and older versions of WaterGAP2 and H08) for their ability to estimate hydrological drought using 1-month Standardised Runoff Index (SRI). They found that an ensemble mean of Global Hydrological Models (GHM) performed better than individual models and detected 133 SRI events out of 136 events identified from observed data across the 8 major catchments they used. Further, the ensemble identified 233 out of 244 runoff deficit events identified from observed data. However, they do also report that the simulation of event durations were not great.
We thus argue that, globally the ISIMIP model outputs we use do sufficiently well reproduce terrestrial water storages and thus their use is justified in our article.
Moreover, WaterGAP 2.2e has recently been validated by the development team (Müller Schmied et al., 2024), where they use the same ISIMIP 3a climate input datasets as we do. They report relatively high correlation coefficients with a median of 0.78 across the globe.
Both the validations we cite here and our analysis show that there is significant uncertainty in the low-flow simulations. This is one of the key conclusions of our study: there is a need for global modelling targeting drought and low-flow periods specifically.
Comment 2.5: Drought event definition: I think standardized drought indices can be a very useful tool for detecting drought anomalies. However, I am skeptical that drought events across different compartments can be effectively identified using the same aggregation period for precipitation, soil moisture and streamflow. The latter have clearly different temporal structure from precipitation and associated with much higher persistency. This point has to be clarified and the optimal aggregation period for each compartment has to be critically evaluated. Moreover, the implemented method seems to be suboptimal for dry regions with long consequent periods of no flow as the authors report themselves (Page 16 Line 8-9).
Reply 2.5: Accumulation period: Based on the literature, we chose to use the 90-day accumulation period as a tradeoff between 1) our ability to capture the short term droughts that are important in tropical catchments, and 2) reduce the strong variability in a standardised index with short accumulation periods to produce a clean signal. We also find that, based on literature, the 90-day accumulation period is a common one. We fully agree that the accumulation period is an important factor in drought propagation. We have opted for using the same seasonal accumulation period for each variable for a few reasons: 1) simplicity of comparisons (as recommended in Van Loon, 2015), 2) the optimal accumulation period necessarily varies across different regions (as shown e.g. by the correlation maps in Liu et al., 2023), and 3) an ambiguity in potential goals for the optimisation. Furthermore, a small sensitivity analysis seemed to produce consistent results for this accumulation period.
Dry areas: As acknowledged in our manuscript and in your review point, there is a significant deficiency in our methodology in dry areas. Using a standardized index with empirical distribution, we are not able to distinguish drought events when the 90-day accumulated data used to fit the empirical distribution has more than ~16% zero values. This value comes from our drought detection threshold of -1 (one standard deviation away from the 0 contains approximately 32% of all values, and because drought is only one of two tails in the normal distribution, the drought is detected when the value is among the 16% lowest values).
We conducted an additional analysis of the locations where this condition affects our ability of detecting compound droughts. Figure R2.2 (in the supplement to this reply) shows all the grid cells with zero values (of either precipitation, soil moisture or streamflow) for more than 30 consecutive days. We use 30 days as the cutoff, because our methodology combines individual identified drought events if there is less than 30 days in between them. In the locations highlighted, there are one or more periods in a year where a drought event cannot be identified because of our detection threshold and our use of the empirical distribution. This limits the maximum length of an event we can identify. In the maps shown in Figure 2 of the manuscript, these locations appear as having compound drought events, as this limitation does not apply year-round and thus allows the detection of some compound drought events.
Comment 2.6: Inconsistent ENSO analysis: The analysis of ENSO suddenly appears in the Methods, only very roughly presented in the results and is neither discussed nor introduced properly. It is not clear why it was selected as a single climatic driver to be examined and it is not clear what was the rationale for this analysis. Consider either extending the scope to possible large-scale climatic drivers of drought co-occurrence across compartments or remove this part from the manuscript for clarity.
Reply 2.6: We agree that the ENSO analysis seems out of scope of the analysis we provide. We intended it as a showcase of what kind of analysis our global study can support, but this was not well communicated. In order to streamline the manuscript, we decided to move the ENSO analysis to the supplementary information and away from the main text.
Comment 2.7: Conclusions: The conclusion section is very vague and very general. It barely touches on the actual results obtained in this study. The implications mentioned there are also very general and could have been resulted from almost any drought study again indicating that the novelty of the presented study is not quite clear. Please revise.
Reply 2.7: Thank you for the comment. The conclusion was revised and shortened. Most of the more general implications were deleted. The new conclusion now better reflects our results and the novelty of our analysis.
Comment 2.8: Detailed comments
Reply 2.8: We sincerely thank you for your detailed and extensive minor comments. We have addressed each of them in our revised manuscript.
References
Gnann, S., Reinecke, R., Stein, L., Wada, Y., Thiery, W., Müller Schmied, H., Satoh, Y., Pokhrel, Y., Ostberg, S., Koutroulis, A., Hanasaki, N., Grillakis, M., Gosling, S. N., Burek, P., Bierkens, M. F. P., and Wagener, T.: Functional relationships reveal differences in the water cycle representation of global water models, Nat Water, 1, 1079–1090, https://doi.org/10.1038/s44221-023-00160-y, 2023.
Heinicke, S., Volkholz, J., Schewe, J., Gosling, S. N., Müller Schmied, H., Zimmermann, S., Mengel, M., Sauer, I. J., Burek, P., Chang, J., Kou-Giesbrecht, S., Grillakis, M., Guillaumot, L., Hanasaki, N., Koutroulis, A., Otta, K., Qi, W., Satoh, Y., Stacke, T., Yokohata, T., and Frieler, K.: Global hydrological models continue to overestimate river discharge, Environ. Res. Lett., 19, 074005, https://doi.org/10.1088/1748-9326/ad52b0, 2024.
Kallio, M., Guillaume, J. H. A., Virkki, V., Kummu, M., and Virrantaus, K.: Hydrostreamer v1.0 – improved streamflow predictions for local applications from an ensemble of downscaled global runoff products, Geoscientific Model Development, 14, 5155–5181, https://doi.org/10.5194/gmd-14-5155-2021, 2021.
Kumar, A., Gosling, S. N., Johnson, M. F., Jones, M. D., Zaherpour, J., Kumar, R., Leng, G., Schmied, H. M., Kupzig, J., Breuer, L., Hanasaki, N., Tang, Q., Ostberg, S., Stacke, T., Pokhrel, Y., Wada, Y., and Masaki, Y.: Multi-model evaluation of catchment- and global-scale hydrological model simulations of drought characteristics across eight large river catchments, Advances in Water Resources, 165, 104212, https://doi.org/10.1016/j.advwatres.2022.104212, 2022.
Liu, Y., Shan, F., Yue, H., Wang, X., and Fan, Y.: Global analysis of the correlation and propagation among meteorological, agricultural, surface water, and groundwater droughts, Journal of Environmental Management, 333, 117460, https://doi.org/10.1016/j.jenvman.2023.117460, 2023.
Müller Schmied, H., Trautmann, T., Ackermann, S., Cáceres, D., Flörke, M., Gerdener, H., Kynast, E., Peiris, T. A., Schiebener, L., Schumacher, M., and Döll, P.: The global water resources and use model WaterGAP v2.2e: description and evaluation of modifications and new features, Geoscientific Model Development, 17, 8817–8852, https://doi.org/10.5194/gmd-17-8817-2024, 2024.
Müller Schmied, H., Gosling, S. N., Garnsworthy, M., Müller, L., Telteu, C.-E., Ahmed, A. K., Andersen, L. S., Boulange, J., Burek, P., Chang, J., Chen, H., Gudmundsson, L., Grillakis, M., Guillaumot, L., Hanasaki, N., Koutroulis, A., Kumar, R., Leng, G., Liu, J., Liu, X., Menke, I., Mishra, V., Pokhrel, Y., Rakovec, O., Samaniego, L., Satoh, Y., Shah, H. L., Smilovic, M., Stacke, T., Sutanudjaja, E., Thiery, W., Tsilimigkras, A., Wada, Y., Wanders, N., and Yokohata, T.: Graphical representation of global water models, Geoscientific Model Development, 18, 2409–2425, https://doi.org/10.5194/gmd-18-2409-2025, 2025.
Tiwari, A. D., Pokhrel, Y., Felfelani, F., Elkouk, A., Boulange, J., Gosling, S. N., Hanasaki, N., Koutroulis, A., Mishra, V., Schmied, H. M., Satoh, Y., Ostberg, S., Stacke, T., and Yin, J.: Underestimation of Historical Terrestrial Water Storage Droughts in Global Water Models, Geophysical Research Letters, 52, e2025GL115164, https://doi.org/10.1029/2025GL115164, 2025.
Van Loon, A. F.: Hydrological drought explained, WIREs Water, 2, 359–392, https://doi.org/10.1002/wat2.1085, 2015.
Veldkamp, T. I. E., Zhao, F., Ward, P. J., Moel, H. de, Aerts, J. C. J. H., Schmied, H. M., Portmann, F. T., Masaki, Y., Pokhrel, Y., Liu, X., Satoh, Y., Gerten, D., Gosling, S. N., Zaherpour, J., and Wada, Y.: Human impact parameterizations in global hydrological models improve estimates of monthly discharges and hydrological extremes: a multi-model validation study, Environ. Res. Lett., 13, 055008, https://doi.org/10.1088/1748-9326/aab96f, 2018.
Zaherpour, J., Gosling, S. N., Mount, N., Schmied, H. M., Veldkamp, T. I. E., Dankers, R., Eisner, S., Gerten, D., Gudmundsson, L., Haddeland, I., Hanasaki, N., Kim, H., Leng, G., Liu, J., Masaki, Y., Oki, T., Pokhrel, Y., Satoh, Y., Schewe, J., and Wada, Y.: Worldwide evaluation of mean and extreme runoff from six global-scale hydrological models that account for human impacts, Environ. Res. Lett., 13, 065015, https://doi.org/10.1088/1748-9326/aac547, 2018.
-
AC2: 'Reply on RC2', Marko Kallio, 01 Jan 2026
-
RC3: 'Comment on egusphere-2025-3909', Yannis Markonis, 07 Nov 2025
Thank you for the opportunity to read this work. The manuscript addresses an important and timely question on how different drought types interact within the hydro-climate system, and it is clear that substantial thought and technical care have gone into the analysis and experimental design. My comments below are intended to help sharpen the framing, align terminology with the broader compound-events literature, and better showcase the strengths that are already present.
1. Terminology and definition
This is a valuable and timely analysis, and I appreciate that you already acknowledge neighboring concepts (propagation, two-type co-occurrence, multi-hazard compounding in Page 2, Lines 14–16). One place where a small editorial change could greatly improve readability is the term “compound drought.” In the extremes literature, “compound events” usually means multi-hazard combinations (e.g., heat and drought), whereas this paper focuses on simultaneous occurrence of multiple drought types within the hydro-climate system.
My suggestion to help readers land where you intend would be to consider adopting “concurrent multi-type drought” throughout (as you did in your previous work in Ahopelto et al., 2020). This small shift will better signal that the contribution is about the facets of drought rather than cross-hazard interactions, without changing your analysis or findings. It also helps to clarify what is being measured (so your strengths are obvious). As implemented, your core metric reads as the fraction of drought-affected days on which all three types co-occur at the grid-day scale, and “durations” are runs of such concurrent days nested within broader droughts, not separate “tri-type events.” Making this explicit will help readers interpret your trends correctly. For example, increases in concurrent days can reflect tighter synchronization (timing) even when overall drought frequency is flat or drops, while longer but less synchronized droughts could lower the concurrent share. Calling this out will showcase a key insight of your study.
If you prefer to retain the term “compound drought,” please consider these suggestions to improve the manuscript’s clarity:
- You could use a phase descriptor, e.g., “compound-drought phase/period”. In this manner you can help distinguish it from a parent drought event.
- You could add one-sentence definition (early in Introduction/Methods), similar to:
“We define a compound (concurrent multi-type) drought as the simultaneous co-occurrence of meteorological, agricultural (soil-moisture), and hydrological (streamflow) drought at the grid-day scale, evaluated with strict concurrence.” - You could add 2–3 sentences that clearly separate concurrent drought from drought propagation and from multi-hazard compound events, and most importantly explain why the concurrent framing is scientifically/operationally useful here (e.g., it pinpoints periods when multiple sectors face simultaneous stress).
- You should cite Wu (2022) when introducing/defining “compound drought,” and acknowledge the broader compound-events canon (e.g., Zscheischler et al., 2020) to demonstrate terminological awareness and to contrast with multi-hazard usage.
- You could add a small schematic illustrating a long meteorological drought with nested soil-moisture and streamflow deficits, highlighting the tri-type overlap segment analyzed (for example Fig.2 in AghaKouchak et al., 2023[1]).
These changes will sharpen the manuscript’s scope and ensure readers interpret your statistics and trends in the intended, operationally meaningful way.
2. Uncertainty in Methods/Results; ENSO out-of-scope
At present, uncertainty appears mainly in the Discussion, while the Methods and Results do not articulate a formal uncertainty framework. In contrast, ENSO receives dedicated methodological space that feels tangential to the central objective of characterizing concurrent multi-type drought climatology and trends. I recommend giving uncertainty a more visible and structured role and, if needed, streamlining or relocating the ENSO analysis.
In the Methods, you could introduce a short subsection titled “Ensemble and uncertainty framework” that explains how you use the multi-model ensemble, multiple forcings, or other choices to assess robustness. This may include, for example, how model spread is summarized, how sensitive results are to thresholds or definitions, and how agreement or disagreement across GHMs and forcings is interpreted. In the Results, it would be beneficial if each main finding is accompanied directly by an indication of uncertainty (for instance, ranges, confidence in sign of trend, or agreement levels), rather than having uncertainty emerge for the first time in the Discussion. This will strengthen readers’ trust in your conclusions and make the Discussion more focused and efficient.
Regarding ENSO, the stratification is interesting, but it is not essential for your core message on concurrent multi-type drought behavior and may distract from it or occupy space better used to formalize uncertainty. A reasonable solution would be either to remove the ENSO-focused analysis from the main text or move it to the Supplementary Material with a brief pointer in the main manuscript. This keeps the narrative tight while preserving the analysis for interested readers.
3. Discussion of Results
Your Results already contain rich and informative patterns; the Discussion will be more impactful if it stays tightly anchored to those findings and avoids opening threads that extend beyond what the analyses can support (Section 4.3 in particular currently feels somewhat beyond the demonstrated evidence).
A clear, evidence-led structure for each subsection of the Discussion could be as follows: first, restate a specific empirical result with explicit reference to the relevant figure or table; second, interpret this result mechanistically, drawing on known processes; third, connect it to existing literature that has ideally already been introduced in the Introduction, noting where your findings confirm, nuance, or challenge previous work; fourth, recognize key limitations or assumptions that affect confidence in this interpretation; and finally, where appropriate, outline practical or conceptual implications. Adopting this pattern will naturally transform the information from the Results into knowledge claims grounded in both data and process understanding.
Several questions arise directly from your reported patterns and could help structure the Discussion without expanding its scope. For example, why do concurrent multi-type drought days and durations increase in equatorial and southern hydrobelts while the boreal belt shows little change or a tendency toward wetting? Why do some regions show pronounced increases in concurrent days without commensurate increases in overall drought frequency, and to what extent might tighter synchronization among precipitation, soil moisture, and streamflow explain this? Why do trend magnitudes and even signs differ across models and forcings, and what does that say about structural or forcing-related uncertainties? Why are changes in duration sometimes stronger than changes in counts; does hydrological memory or catchment storage act to prolong concurrent phases once initiated? Why do coastal or monsoon-dominated regions behave differently from nearby continental interiors? And why do snow- or permafrost-affected regions deviate from global tendencies; is this due to genuinely distinct processes or to limitations in representing cryospheric processes and seasonality in the models? You do not need to address every one of these in depth, but framing selected subsections around such targeted questions will keep the Discussion tightly linked to figures, mechanisms, and literature, and ensure that interpretation remains firmly evidence-based.
In the same spirit, there are a few specific features in the figures that would benefit from direct comment. The step-like jumps visible in the Figure 3A time series (for example in NDR) should be discussed: please clarify whether they reflect genuine hydro-climatic regime shifts, arise from methodological or processing decisions, or could point to data inconsistencies. Applying or citing a suitable change-point or non-stationarity assessment would help to evaluate whether these shifts may bias trend estimates. Similarly, the apparent non-stationarities in Figure 5C and the striking similarity in statistical properties between H08 and WaterGAP warrant a short explanation. It would help readers if you could distinguish what aspects of cross-model similarity are likely linked to shared structural or parameterization choices and what aspects may be inherited from harmonized forcings or experimental setup. A concise clarification here will prevent over-interpretation of agreement as independent confirmation.
Finally, consider revisiting subsection 4.3 with a critical eye on whether every element is directly supported by your analyses. Content that is more speculative or forward-looking could be condensed into a final “Outlook” paragraph that clearly signals its role as perspective, leaving the core of the Discussion focused on results-backed insights.
4. Minor comments:
Page 2, Line 18: I could not locate the reference “Kallio, M., Heino, M., Kinnunen, P., Fallon, A., and Ahopelto, L.: Identifying Global Co-occurrence of Hydrological, Meteorological and Agricultural Droughts, 2019.” Please verify the citation details or provide additional information.
Page 5, Line 26: Please clarify whether “forcing” is used in the sense of “transforming” or whether you refer to external climate forcings; the current phrasing is ambiguous.
Page 6, Line 27: The phrase “probability of occurrence” suggests an annual probability that a compound drought will occur in a given year. Since your metric appears to describe the share of time steps with concurrence, a term such as “fraction of concurrency” or similar would likely be more transparent.
Page 10, Line 1: Figure 3A is difficult to read in its current composite form. You may wish to facet by hydrobelt with independent vertical axes so that regions with low variance are no longer compressed by those with higher variance. If time series by forcing and GHM remain crowded, consider separating them into distinct figures (for example, one by forcings and one by GHMs).
Page 10, Lines 2–9: Please add the full names of the hydrobelt regions in the caption here and similarly for Figures 4 and 5.
Page 15, Lines 8–9: The statement about “climate change” influencing increases in droughts would benefit from specifying which aspects (for example, shifts in precipitation regimes, increased evaporative demand) and briefly indicating the mechanism.
Page 16, Lines 5–7: When referring to dataset uncertainties, please make explicit that uncertainties differ across regions and environments and briefly indicate how this might affect interpretation [2].
Page 16, Lines 14–19: This section invites the question “why are we seeing this?” Adding one or two sentences that link the pattern to plausible processes or existing studies would strengthen it.
Page 16, Line 21: Please specify what kind of problems ERA5 is experiencing in this context (for instance, biases in precipitation, issues in particular regions or seasons).
Page 17, Line 1: It is not entirely clear from the legend whether Forcing Data are included in Figure 5B. Please clarify the labeling so that readers can distinguish all elements.
Page 18, Line 17: Clarify what you mean by “non-stationary” (for example, changes in mean, variance, or dependence over time), so readers understand which property of the series is being discussed.
Page 19, Lines 1–15: Please reconsider whether subsection 4.3 is fully aligned with the main findings and scope. If it is only loosely supported by your results, it may be more effective to shorten or reposition it as part of a brief Outlook rather than a core Discussion element.
References[1] AghaKouchak, A., Huning, L. S., Sadegh, M., Qin, Y., Markonis, Y., Vahedifard, F., ... & Kreibich, H. (2023). Toward impact-based monitoring of drought and its cascading hazards. Nature Reviews Earth & Environment, 4(8), 582-595.
[2] Markonis, Y., Vargas Godoy, M. R., Pradhan, R. K., Pratap, S., Thomson, J. R., Hanel, M., ... & Papalexiou, S. M. (2024). Spatial partitioning of terrestrial precipitation reveals varying dataset agreement across different environments. Communications Earth & Environment, 5(1), 217.Citation: https://doi.org/10.5194/egusphere-2025-3909-RC3 -
AC3: 'Reply on RC3', Marko Kallio, 01 Jan 2026
Response to reviewers
We wish to thank all three reviewers for their contributions and for their insights. Going over and answering these comments have significantly streamlined and strengthened our manuscript. It is also notable that there is a significant overlap between the reviewer comments, highlighting the biggest weaknesses of our original manuscript. We therefore list here the main changes we have made to the manuscript, before responding to your review comments in detail.
Main changes to the manuscript:
- We have revised the language used in the manuscript. Most notably, we change the language from compound droughts and talk about concurrent droughts instead. We also make the aim of our study more clear.
- We provide information of validation of the ensemble we use, both towards replicating droughts as well as the primary outputs of the models (streamflow, soil moisture).
- We have moved the ENSO analysis to the appendix and refer to it as an example of additional analyses our results could support.
- We provide a more detailed description of our handling of uncertainty through using an ensemble approach.
- We have shifted from Hydrobelts to using Holdridge Life Zones as the areas we use to summarise our results.
- We have revised our discussion to better reflect our results.
- We have made numerous small edits to the text based on the minor comments from each reviewer.
Review by Professor Yannis Markonis
Comment 3.1: Thank you for the opportunity to read this work. The manuscript addresses an important and timely question on how different drought types interact within the hydro-climate system, and it is clear that substantial thought and technical care have gone into the analysis and experimental design. My comments below are intended to help sharpen the framing, align terminology with the broader compound-events literature, and better showcase the strengths that are already present.
Reply 3.1: Thank you very much for reviewing our work and providing insightful comments. We have addressed each of them as you’ll find in our replies below.
- Terminology and definition
Comment 3.2: This is a valuable and timely analysis, and I appreciate that you already acknowledge neighboring concepts (propagation, two-type co-occurrence, multi-hazard compounding in Page 2, Lines 14–16). One place where a small editorial change could greatly improve readability is the term “compound drought.” In the extremes literature, “compound events” usually means multi-hazard combinations (e.g., heat and drought), whereas this paper focuses on simultaneous occurrence of multiple drought types within the hydro-climate system.
My suggestion to help readers land where you intend would be to consider adopting “concurrent multi-type drought” throughout (as you did in your previous work in Ahopelto et al., 2020). This small shift will better signal that the contribution is about the facets of drought rather than cross-hazard interactions, without changing your analysis or findings. It also helps to clarify what is being measured (so your strengths are obvious). As implemented, your core metric reads as the fraction of drought-affected days on which all three types co-occur at the grid-day scale, and “durations” are runs of such concurrent days nested within broader droughts, not separate “tri-type events.” Making this explicit will help readers interpret your trends correctly. For example, increases in concurrent days can reflect tighter synchronization (timing) even when overall drought frequency is flat or drops, while longer but less synchronized droughts could lower the concurrent share. Calling this out will showcase a key insight of your study.
If you prefer to retain the term “compound drought,” please consider these suggestions to improve the manuscript’s clarity:
- You could use a phase descriptor, e.g., “compound-drought phase/period”. In this manner you can help distinguish it from a parent drought event.
- You could add one-sentence definition (early in Introduction/Methods), similar to:
“We define a compound (concurrent multi-type) drought as the simultaneous co-occurrence of meteorological, agricultural (soil-moisture), and hydrological (streamflow) drought at the grid-day scale, evaluated with strict concurrence.” - You could add 2–3 sentences that clearly separate concurrent drought from drought propagation and from multi-hazard compound events, and most importantly explain why the concurrent framing is scientifically/operationally useful here (e.g., it pinpoints periods when multiple sectors face simultaneous stress).
- You should cite Wu (2022) when introducing/defining “compound drought,” and acknowledge the broader compound-events canon (e.g., Zscheischler et al., 2020) to demonstrate terminological awareness and to contrast with multi-hazard usage.
- You could add a small schematic illustrating a long meteorological drought with nested soil-moisture and streamflow deficits, highlighting the tri-type overlap segment analyzed (for example Fig.2 in AghaKouchak et al., 2023[1]).
These changes will sharpen the manuscript’s scope and ensure readers interpret your statistics and trends in the intended, operationally meaningful way.
Reply 3.2: We sincerely thank you for this insight. We have followed your suggestion and changed the language of the manuscript to concurrent (multi-type) droughts in order to avoid confusion with the most common use of the compound drought. We find that the compound event framework is useful in assessing drought events, but since we only use the multivariate compound event type, it makes sense to simply discuss concurrent drought types. We, however, still make the connection between concurrent multi-type droughts and multivariate compound events in the introduction.
Thank you also for the suggestion of adding a schematic figure illustrating multi-type droughts. We have added such an example in a revised Figure 1 in our manuscript.
Comment 3.3: 2. Uncertainty in Methods/Results; ENSO out-of-scope
At present, uncertainty appears mainly in the Discussion, while the Methods and Results do not articulate a formal uncertainty framework. In contrast, ENSO receives dedicated methodological space that feels tangential to the central objective of characterizing concurrent multi-type drought climatology and trends. I recommend giving uncertainty a more visible and structured role and, if needed, streamlining or relocating the ENSO analysis.
In the Methods, you could introduce a short subsection titled “Ensemble and uncertainty framework” that explains how you use the multi-model ensemble, multiple forcings, or other choices to assess robustness. This may include, for example, how model spread is summarized, how sensitive results are to thresholds or definitions, and how agreement or disagreement across GHMs and forcings is interpreted. In the Results, it would be beneficial if each main finding is accompanied directly by an indication of uncertainty (for instance, ranges, confidence in sign of trend, or agreement levels), rather than having uncertainty emerge for the first time in the Discussion. This will strengthen readers’ trust in your conclusions and make the Discussion more focused and efficient.
Reply 3.3: Thank you for these observations! We agree that discussing our way to address uncertainty is important for the reader and that this was indeed missing from our manuscript. We have included a section under the methods which explains our use of the ensemble and handling of uncertainty in the ensemble. We also address the uncertainty already in the results section and do not wait until the discussion to discuss it. We also include new supplementary information which includes additional maps of the spread of the ensemble for all maps in Figure 2 of the manuscript.
Reviewer #2 also raised questions about the applicability of our methodology in dry areas. In response, we have made new maps to illustrate where our method is not able to properly address droughts all year round.
Comment 3.4: Regarding ENSO, the stratification is interesting, but it is not essential for your core message on concurrent multi-type drought behavior and may distract from it or occupy space better used to formalize uncertainty. A reasonable solution would be either to remove the ENSO-focused analysis from the main text or move it to the Supplementary Material with a brief pointer in the main manuscript. This keeps the narrative tight while preserving the analysis for interested readers.
Reply 3.4: Thank you for this suggestion. ENSO analysis does indeed seem out of scope. We have followed your suggestion and moved the ENSO analysis in the supplementary information in order to streamline the main text.
Comment 3.5: 3. Discussion of Results
Your Results already contain rich and informative patterns; the Discussion will be more impactful if it stays tightly anchored to those findings and avoids opening threads that extend beyond what the analyses can support (Section 4.3 in particular currently feels somewhat beyond the demonstrated evidence).
A clear, evidence-led structure for each subsection of the Discussion could be as follows: first, restate a specific empirical result with explicit reference to the relevant figure or table; second, interpret this result mechanistically, drawing on known processes; third, connect it to existing literature that has ideally already been introduced in the Introduction, noting where your findings confirm, nuance, or challenge previous work; fourth, recognize key limitations or assumptions that affect confidence in this interpretation; and finally, where appropriate, outline practical or conceptual implications. Adopting this pattern will naturally transform the information from the Results into knowledge claims grounded in both data and process understanding.
Reply 3.5: We have streamlined our discussion in the pattern you suggest. We agree that this does help in the flow of the discussion.
Comment 3.6: Several questions arise directly from your reported patterns and could help structure the Discussion without expanding its scope. For example, why do concurrent multi-type drought days and durations increase in equatorial and southern hydrobelts while the boreal belt shows little change or a tendency toward wetting? Why do some regions show pronounced increases in concurrent days without commensurate increases in overall drought frequency, and to what extent might tighter synchronization among precipitation, soil moisture, and streamflow explain this? Why do trend magnitudes and even signs differ across models and forcings, and what does that say about structural or forcing-related uncertainties? Why are changes in duration sometimes stronger than changes in counts; does hydrological memory or catchment storage act to prolong concurrent phases once initiated? Why do coastal or monsoon-dominated regions behave differently from nearby continental interiors? And why do snow- or permafrost-affected regions deviate from global tendencies; is this due to genuinely distinct processes or to limitations in representing cryospheric processes and seasonality in the models? You do not need to address every one of these in depth, but framing selected subsections around such targeted questions will keep the Discussion tightly linked to figures, mechanisms, and literature, and ensure that interpretation remains firmly evidence-based.
Reply 3.6: We sincerely thank you for the suggestions for discussion items. We have incorporated some of them in our revised discussion. One point, we feel, was already incorporated into the discussion: the wetting of the boreal hydrobelt may be due to the melting of permafrost and reduction in (frozen) water storage in boreal basins. In response to a comments made by reviewer #1, we have changed our summarising regions from Hydrobelts to Holdridge Life Zones, increasing the number of zones from 8 to 13 (we use simplifications from Kummu et al. (2021) to come from 38 to 13 zones). This has increased our ability to find nuance in the zones. Under the Holdridge Life Zones, we find that there is strong wetting in Polar Deserts, Boreal Forest, and Tundra. At the same time, there is a clear increase in drought days per year in Polar Deserts and Tundra, which suggests tightening synchronisation of the three drought types in these regions. This seems logical, considering that these regions are the most northern ones lying mostly above 60°N. This, and other points are covered in the newly revised discussion, and new supplementary diagnostic tables and figures are added to support the discussion.
Comment 3.7: In the same spirit, there are a few specific features in the figures that would benefit from direct comment. The step-like jumps visible in the Figure 3A time series (for example in NDR) should be discussed: please clarify whether they reflect genuine hydro-climatic regime shifts, arise from methodological or processing decisions, or could point to data inconsistencies. Applying or citing a suitable change-point or non-stationarity assessment would help to evaluate whether these shifts may bias trend estimates. Similarly, the apparent non-stationarities in Figure 5C and the striking similarity in statistical properties between H08 and WaterGAP warrant a short explanation. It would help readers if you could distinguish what aspects of cross-model similarity are likely linked to shared structural or parameterization choices and what aspects may be inherited from harmonized forcings or experimental setup. A concise clarification here will prevent over-interpretation of agreement as independent confirmation.
Reply 3.7: We will apply a simple change point analysis to the data presented in Figure 3 of the manuscript (timeseries and trends in days under drought, and the general trend in the drought indices). With the change in summarising regions, Holdridge Life Zones provide a more nuanced view of the trends in different regions of the world, but may also provide more evidence for an analysis of what causes the step-like jumps. We will also provide an analysis of whether the jumps originate from the models, their forcings, or whether they are a real phenomenon.
In your minor comments, you suggest adding less crowded additional versions of Figure 3. We have included such plots in the supplementary materials, revealing that there are more agreement than disagreement among the three models. Figure R3.1 (in the supplement to this reply) shows the timeseries of days under concurrent droughts for the 13 Holdridge Life Zones and for a single climate forcing (20CRv3-ERA5). The figure shows, however, that the disagreements between the three different models are more pronounced in Tropical Forests and in the cold environment, suggesting that the differences among the models are the most meaningful in these environments. We will include this figure (and others) in the supplementary materials, and discuss them in the main text.
Comment 3.8: Finally, consider revisiting subsection 4.3 with a critical eye on whether every element is directly supported by your analyses. Content that is more speculative or forward-looking could be condensed into a final “Outlook” paragraph that clearly signals its role as perspective, leaving the core of the Discussion focused on results-backed insights.
Reply 3.8: We shortened and streamlined the sub-section 4.3 and removed the subtitle and added the text to the end of section 4.2. We now start the paragraph with a sentence “Lastly, we raise the discussion to a more forward-looking general drought management level.” to clarify the meaning of the paragraph and to differentiate it from the results-oriented discussion. We also removed the last paragraph about adaptive governance to make the storyline better.
Comment 3.9: 4. Minor comments
Reply 3.9: Thank you for the numerous minor comments. We have addressed each in our work. We wish to comment on the issue of not finding our earlier report online (Kallio et al., 2019) here. This is indeed the case; it was not previously available in online repositories. We have now deposited it in Zenodo, and you can find the report at: https://zenodo.org/records/18082574.
References
Kallio, M., Heino, M., Kinnunen, P., Fallon, A., and Ahopelto, L.: Identifying Global Co-occurrence of Hydrological, Meteorological and Agricultural Droughts, 2019.
Kummu, M., Heino, M., Taka, M., Varis, O., and Viviroli, D.: Climate change risks pushing one-third of global food production outside the safe climatic space, One Earth, 4, 720–729, https://doi.org/10.1016/j.oneear.2021.04.017, 2021.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,799 | 140 | 35 | 1,974 | 34 | 36 |
- HTML: 1,799
- PDF: 140
- XML: 35
- Total: 1,974
- BibTeX: 34
- EndNote: 36
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
First, I would like to question the authors about their motivation for conducting a study on droughts at the global scale. Studies of this nature commonly require methodological simplifications to make their execution feasible, and the present manuscript is no exception. Thus, what is the real advantage of sacrificing a higher level of detail and contextualization of a specific region in exchange for a global study, whose analyses are, by nature, more limited? Who, in fact, benefits from simplified and limited information or analyses about drought events? This is especially relevant considering that droughts are managed locally or regionally, not globally.
Additionally, the representativeness of grouping the planet into only nine regions is questionable. Beyond the presumed meteorological similarities, do these regions truly share comparable drought impact patterns? For example, the “SST” region encompasses both the Brazilian semiarid — with an average annual rainfall of approximately 600 mm, potential evapotranspiration close to 2000 mm, and intermittent rivers, making it highly susceptible to droughts — and the country’s South and Southeast regions, characterized by a markedly humid climate that includes some of the continent’s largest rivers, where droughts are not a recurring problem. In other words, these two regions do not share similarities in terms of drought occurrence patterns and drought impacts, so they should not be analyzed in the same group. Therefore, I suggest that the authors consider grouping regions not only based on hydrometeorological similarities but also on drought impact and occurrence patterns.
Given the profound environmental changes caused by human activities in all parts of the world, does it still make sense to consider droughts solely as a purely natural hazard, completely disregarding the anthropogenic influence on their onset and propagation? I suggest that the authors address this discussion or clearly state the rationale for considering it more useful to consider droughts exclusively as natural hazards.
Another relevant limitation concerns the temporal scale adopted for the calculation of drought indices. The study employs an empirical distribution methodology on a daily scale, later aggregated into three-month periods. What is the justification and advantage of performing a daily analysis for a type of disaster whose natural component is traditionally assessed using standardized indices at monthly scales (e.g. 3, 6, or 12 months)? What is the potential impact of this choice on identifying drought events in regions with highly seasonal rainfall regimes? I recommend converting all variables to a monthly basis and computing standardized indices using a 12-month scale, as this approach may be more appropriate for a global-scale study. Each time step would always consider one full year, which would always capture the rainy season and avoid the potential effect of the seasonality of the precipitation regime.
It is also important to discuss the effectiveness of the method adopted for identifying drought events. I suggest that the authors include a validation step based on historical drought records, in order to demonstrate the capability of the proposed methodology to adequately detect different types of drought events. The discussions and conclusions about the occurrence of compound droughts are only valid if it is shown, based on observed data, that the proposed approach can effectively identify the different drought types. Based on these observational data, more relevant regions of interest could then be defined for detailed analysis, replacing the nine generic regions currently considered.
The authors employed a combination of global models with a spatial resolution of 0.5°, using reanalysis data as input. In this context, one may ask: what is the accuracy and performance of this model? How was it calibrated, considering the distinct regional characteristics across the globe? The authors discuss model and index-related uncertainties, but it remains unclear how these could be improved and how the results should be evaluated in light of such uncertainties.