Understanding spatio-temporal patterns of the propagation characteristics across meteorological, hydrological, and agricultural droughts and their influencing factors

Liu, Yuanrui; Hu, Tingting; Yang, Jiawen; Yu, Lei

doi:10.5194/egusphere-2025-4791

Preprints

https://doi.org/10.5194/egusphere-2025-4791

Preprints

06 Nov 2025

| 06 Nov 2025

Understanding spatio-temporal patterns of the propagation characteristics across meteorological, hydrological, and agricultural droughts and their influencing factors

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Abstract. Understanding the propagation of diverse drought conditions is necessary for drought preparedness. This study conducted a comprehensive analysis of the propagation characteristics across meteorological, hydrological, and agricultural droughts from 1958 to 2024 over global land areas, based on an ensemble of ERA5, GLDAS, and TerraClimate datasets. Using standardized drought indices at different accumulation periods, three drought propagation characteristics, including response time (RT), propagation rate (PR), and lag time (LT), were examined based on time-lag correlation analysis and multi-threshold run theory. The climatic and geographical feature factors that influence drought propagation were quantitatively evaluated using a SHapley Additive exPlanations (SHAP)-based attribution method. The results demonstrate the propagation pathways of meteorological-hydrological-agricultural drought at the global-scale, with the average RT, PR, and LT from meteorological to hydrological drought at 5.0 months, 55.3 %, and 1.23 months; from meteorological to agricultural drought at 8.7 months, 30.3 %, and 2.60 months; and from hydrological to agricultural drought at 5.8 months, 35.0 %, and 2.49 months, respectively. Notable temporal and spatial heterogeneities are observed in the drought propagation characteristics, which are closely influenced by with the regional climatic feature. Globally, temperature and potential evapotranspiration are the primary factors influencing the propagation of meteorological drought to hydrological drought, whereas precipitation plays a decisive role in the propagation from meteorological or hydrological drought to agricultural drought. The findings underscore the importance of taking climatic characteristics into account in the development and implementation of regional drought risk management.

Received: 29 Sep 2025 – Discussion started: 06 Nov 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, 3627 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (3627 KB)

Supplement (856 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

08 May 2026

Understanding meteorological, runoff, and agricultural drought propagation and their influencing factors in an ensemble of multiple datasets

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Hydrol. Earth Syst. Sci., 30, 2775–2795, https://doi.org/10.5194/hess-30-2775-2026,https://doi.org/10.5194/hess-30-2775-2026, 2026

Short summary

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2025-4791', Anonymous Referee #1, 08 Dec 2025

Review of “Understanding spatio-temporal patterns of the propagation characteristics across meteorological, hydrological, and agricultural droughts and their influencing factors” by Yuanrui Liu and co-authors
General comment
Liu and co-authors investigate drought propagation characteristics, their variations, and their controls at the global scale by using an ensemble of reanalyses and land surface models, conventional statistical analyses, and an explainable Machine Learning (ML) algorithm. By doing so, they show spatial and temporal differences in drought propagation characteristics across the globe, and the importance of climatic conditions in governing these differences. Despite an already good body of literature on this topic, the use of explainable ML is rather new and the analysis of multiple products adds strength to this work. However, I have some concerns, mainly regarding the exact novelty of this contribution, some methodological choices and the presentation of the work, as reported below.
Major comments

1. Novelty

Which is the specific research gap that this work addresses? This is currently not entirely clear from the title, abstract, and introduction. As the authors also acknowledge in the introduction, global-scale drought propagation studies are already available. I might see the originality of this work being the use of multiple datasets, but in this case, I believe this could be better worked out throughout the manuscript. Not only in the title, abstract, and introduction, to set the reader’s expectations clear, but also later in the manuscript. As a reader, for instance, I would enjoy having more discussion on the most suitable dataset(s) for drought applications. I see that providing a clear recommendation on this may be difficult from the current analyses, since you do not have here observations to benchmark the datasets with, but maybe you could still say something based on expectations on drought propagation that we have from previous observation-based studies? Also, I would find useful to have clear ranges of variation for the drought propagation characteristics from the different datasets in the abstract and conclusions, as an indication of the uncertainties in such characteristics.

2. Agricultural drought definition and propagation from hydrological to agricultural droughts

The adopted agricultural drought definition and the choice of investigating hydrological-to-agricultural drought propagation is not entirely clear to me. Agricultural droughts are introduced in the paper as ‘reduced soil moisture’ (L40), coherently with extensive previous literature (Van Loon, 2015) which refers to agricultural or soil moisture droughts as deficits in the root-zone soil moisture mainly impacting the agricultural sector, following meteorological droughts and potentially leading to hydrological droughts (i.e., deficits in runoff and groundwater, Van Loon, 2015). Previous works therefore mostly investigated the propagation from meteorological to agricultural and then to hydrological droughts, by finding shorter propagation times from meteorological to agricultural droughts than from meteorological to hydrological droughts (e.g., Odongo et al., 2023; Teutschbein et al., 2025). Could you clarify why you chose to investigate the propagation from hydrological to agricultural droughts instead of the other way round? Also, are you considering soil moisture data from the upper or deeper layers? This is not specified in the methods. In the discussion, agricultural droughts are said to affect the ‘deeper soil moisture’ (L393), which may point to the use of deep soil moisture data only, but I ask you to clarify earlier on this important piece of information for the general understanding of the work. The use of deep soil moisture would (partly) explain to me both the choice of investigating this drought propagation pathway and the results, showing longer propagation times for agricultural than for hydrological droughts with respect to the meteorological ones. Yet, if this is the case, I wonder whether the use of the term agricultural droughts is the most suitable here, given the interest in the upper soil layer by the agricultural sector.

3. Trend analysis

Important methodological details on this are missing. The analysis is briefly described in the Results section (L251–252), but how this moving-window trend analysis exactly works is not totally clear. From my understanding, you calculate the various metrics (e.g., propagation time) for each year based on a moving window consisting of (the next?) 30 years and then apply a trend analysis on the annual values that you obtained. If this is the case, how do you deal with potential autocorrelation from partially overlapping raw data? I would suggest expanding on this point and moving the current description of this analysis to the Methods section (e.g., to a new subsection between the current 2.4 and 2.5). Please also provide full name and appropriate references for the statistical tests used here (i.e., the ‘M-K test’ currently mentioned in the text). Finally, what do you mean by ‘monotonic trend’ in the pie charts in e.g. Fig. 3? From my understanding, it refers to the greyish areas in the maps, with trend slopes close to zero. Did you set any lower and upper limits on the trend slopes to discriminate these ‘monotonic trends’? If so, please specify. And why are these monotonic trends not appearing in the pie charts in Fig. 5 and 6, even though greyish areas are reported in the corresponding maps?

4. Language and readability

I find the paper generally well structured, but rather lengthy and sometimes convoluted. I think the reading flow could be improved by reducing redundant expressions (e.g., couldn’t ‘feature factors’ be simply ‘features’ or ‘factors’?), repetitions between sections (e.g., L105–107 already said in the previous section), and rather obvious statements (e.g., ‘with positive correlation with r_P > 0, and negative correlation with r_P < 0’, L156–157). I also noticed many abbreviations, especially in Sect. 2.1 Datasets (e.g., ECMWF, CLSM, etc), which seem to me not used anymore in the paper. I would suggest removing them and making sure that abbreviations are always introduced the first time they are used (currently not the case, see e.g., ML at L73). Consistent notation throughout the text and across the text and the figures (currently not the case, see e.g., Eq.1 and Fig. 1b, d, and f) would also ease the readability of the paper. In summary, I see room for improvement, with another careful round of proofreading focused on language.

5. References

I appreciate the referencing to very recent literature on the topic, yet I believe that also additional references to (older) seminal papers on droughts and drought propagation characteristics would be appropriate (e.g., López-Moreno et al., 2013 and Barker et al., 2016 for the correlation analysis, other papers that I referred to above).
Minor comments

6. Abstract, I would appreciate introducing the datasets and methods you used in general terms (e.g., reanalyses for ERA5 and so on), to facilitate readers potentially not familiar with these specific datasets and methods.

7. L26–28, I suggest rephrasing since, from my understanding, Gebrechorkos et al. (2025) showed that increases in atmospheric evaporative demand significantly contributed to recent increases in drought severity, but not as primary factor.

8. L42–43, I believe defining here drought propagation characteristics would be beneficial for readers who may not be familiar with them and to ease the readability of the rest of the manuscript.

9. L44–48, many different methods are currently mixed together in this sentence and specifically: methods used to generate the datasets needed for drought propagation studies (e.g., hydrological models), methods used to quantify drought propagation characteristics (e.g., correlation analysis and run theory), and methods used to attribute these characteristics to their controls (e.g., ML). I suggest clarifying this point, for instance by splitting this long sentence into several ones. In addition, maybe add event-coincidence analysis as another method to quantify drought propagation characteristics as proposed by Baez-Villanueva et al. (2024)? Finally, I would suggest removing the mention to the complex network theory since this is used for spatial drought propagation, which is not the topic of this paper, or alternatively, specifying this point and what spatial drought propagation is.

10. L52, I suggest introducing the concept of groundwater droughts earlier.

11. L64–65, I would say that all the global-scale analyses cited before are ‘consistent’ and ‘comparable’ within themselves since they use common methods and datasets for the whole globe. I would suggest rephrasing this sentence to the exact research gap you are aiming at addressing with your work (see also comment #1).

12. L69–70, I do not fully agree with this sentence, which seems to me also contradicting the previous one. Literature on the factors controlling drought propagation across different climatic and geographical regions is rather vast now (see e.g., Xiong et al. 2025 and other reviews on the topic, also cited in the text).

13. L93, could you provide some references of these comparisons?

14. L105, I would argue that data quality is crucial for any study, not only drought studies. It may also be a matter of personal taste, but I do not see as really needed these very general sentences, which also contribute to making the paper quite lengthy in my view (see also comment #4).

15. L118 ‘high-spatial-resolution’, explicitly mentioning the resolution of this and all other datasets would help the readers in my opinion.

16. L121–123, I appreciate the details provided here on how potential evapotranspiration, runoff and soil moisture are computed in this dataset. Could you provide such details also for the other datasets? I think they would give to the reader more context, also on the differences you detect among them.

17. L140–143, some references to support these statements would be appreciated since the differences between a standardized approach and other methods for drought identification are well discussed in many papers now.

18. L154, which is the maximum accumulation period n that you tested? I assume 24 months from Fig. 1, but please specify.

19. L163–166, clarification on why you applied these rules for drought selection would be appreciated.

20. L195–196, I assume the factors that you use as model predictors regarding precipitation, temperature, potential evapotranspiration, runoff, soil moisture, and vegetation conditions are long-term averages, but I encourage you to specify this point in the text. If so, which period did you consider for averaging? Also, why did you choose these specific factors? Other factors regarding e.g. soil or geology would also be important in my view. In general, additional details on the models would be beneficial (e.g., training and validation periods, achieved model performances, etc).

21. Fig.1, caption, please expand the abbreviations (in other captions as well, in the supplementary figures too). Also, some details are missing in this specific caption (e.g., inner plots - where axes are not labelled – and p value for statistical significance). I recommend you having another check that all information needed to fully understand the figures are reported in the captions, including those in the supplementary.

22. L212–213, it is not entirely clear to me what you mean by ‘maximum correlation coefficients’. I suggest rephrasing. I would say that the robustness of your assessments comes from the relatively high correlation coefficients in Fig. 1 and their statistical significance.

23. Fig. 2, wouldn’t considering relative months from the start of the local water year rather than calendar months easier here? This would allow not to mix different processes occurring in the same months in the northern and southern hemispheres. Also, units for correlation coefficients are missing in the axes labels.

24. Fig. 3, caption, which correlation is not statistically significant in blank grid cells? If I understand this analysis correctly (see comment #3), you computed multiple correlations here. Please specify.

25. Fig. 4, which dataset does this figure refer to? The same comment applies also to other figures. I assume all the figures in the main text refer to the ensemble mean, but I would recommend specifying this somewhere. With respect to this specific figure, I would also suggest correcting the label in the colour bars in panels a, c, and e to ‘Propagation rate’ rather than ‘Response rate’ for consistency with the rest of the manuscript and specifying in the caption the different y-axis in panel b as compared to panels d and f. An additional comment on the analysis behind this figure: could the very low propagation rates in panels c and e be due to the time scale that you use? From my understanding, you are not considering any time lag between drought types, even though you show that some drought types can occur well after others in e.g., Fig. 1.

26. Fig. 5, why are the blank areas here the same across the three maps? This is not the case in Fig. 3, which makes sense to me. Also, please add more information to this caption, in a similar way to what done for Fig. 3.

27. Fig. 6 shows decreasing trends in the lag time between meteorological and hydrological droughts in large portions of Europe and northern Asia that are not reflected in the propagation time though (Fig. 3). Do you have an explanation for such discrepancies? Also, I see that attributing these trends to their causes might be outside of the scope of this current paper, but I think that some discussion on potential causes of these trends would still be valuable.

28. L327–328, ‘the SHAP value indicates that high temperatures have shortened the response time of meteorological drought to hydrological drought’, I stumbled a bit here. I suggest rephrasing, for instance by using present tenses, not to evoke changes over time, which is not what you are looking at with your SHAP-analysis. This comment applies also to L409–411.

29. Fig. 9 and corresponding text, I would suggest turning ‘key feature factors’ to ‘dominant factors’ to enhance the clarity of which factors you are looking at here. Also, for Fig. 9, do you have an explanation for the non-monotonic behaviour of meteorological-to-hydrological drought propagation characteristics across different quantiles of the considered features?

30. L429–435, this part sounds to me more like a discussion of the implications rather than of uncertainties. I suggest moving to a new subsection or rename the current one.

31. L438–440, I suggest some rephrasing here to improve the clarity of this sentence.
Technical corrections

32. L45, I cannot find the reference (Zhu et al., 2021) in the reference list, please add it. In addition, include also Xiong et al. (2025), already cited in the text, among the global-scale studies?

33. L47, there are two entries for both (Yang et al., 2024) and (Shi et al., 2022) in the reference list. Which one are you referring to here? Please specify here and elsewhere in the manuscript.

34. L127, I assume you mean here ‘elevation’ rather than ‘evaluation’. Please correct it, here and throughout the manuscript.

35. L165, ‘with on month’ -> ‘with one month’? Else, the sentence sounds strange to me. Please check.

36. L190, ‘formula’ -> ‘formulated’?

37. L193, ‘influencing on the model predictions’ -> ‘influencing the model predictions’

38. L282, ‘the highest PR_MH and LT_MH values’ -> the highest PR_MH and lowest LT_MH values?

39. Fig. 8 and 9, please correct the labels in panels c, f, and i with the subscript ‘HA’ instead of ‘MA’.

40. L303–304, I suggest rephrasing or removing, since this wording does not sound fitting to this subsection to me.

41. L344–345, the first ‘hydrological’ should probably be ‘agricultural’.

42. L413, ‘reasons’ -> ‘seasons’?

43. L424, Figures 11-13, these figures are reported in the supplement. Please correct.

44. L442, ‘finding’ -> ‘findings’?
References

Baez-Villanueva, O. M., Zambrano-Bigiarini, M., Miralles, D. G., Beck, H. E., Siegmund, J. F., Alvarez-Garreton, C., Verbist, K., Garreaud, R., Boisier, J. P., & Galleguillos, M. (2024). On the timescale of drought indices for monitoring streamflow drought considering catchment hydrological regimes. Hydrology and Earth System Sciences, 28(6), 1415–1439. https://doi.org/10.5194/hess-28-1415-2024

Barker, L. J., Hannaford, J., Chiverton, A., & Svensson, C. (2016). From meteorological to hydrological drought using standardised indicators. Hydrology and Earth System Sciences, 20(6), 2483–2505. https://doi.org/10.5194/hess-20-2483-2016

Gebrechorkos, S. H., Sheffield, J., Vicente-Serrano, S. M., Funk, C., Miralles, D. G., Peng, J., Dyer, E., Talib, J., Beck, H. E., Singer, M. B., & Dadson, S. J. (2025). Warming accelerates global drought severity. Nature, 642(8068), 628–635. https://doi.org/10.1038/s41586-025-09047-2

López-Moreno, J. I., Vicente-Serrano, S. M., Zabalza, J., Beguería, S., Lorenzo-Lacruz, J., Azorin-Molina, C., & Morán-Tejeda, E. (2013). Hydrological response to climate variability at different time scales: A study in the Ebro basin. Journal of Hydrology, 477, 175–188. https://doi.org/10.1016/j.jhydrol.2012.11.028

Odongo, R. A., De Moel, H., & Van Loon, A. F. (2023). Propagation from meteorological to hydrological drought in the Horn of Africa using both standardized and threshold-based indices. Natural Hazards and Earth System Sciences, 23(6), 2365–2386. https://doi.org/10.5194/nhess-23-2365-2023

Teutschbein, C., Grabs, T., Giese, M., Todorović, A., & Barthel, R. (2025). Drought propagation in high-latitude catchments: Insights from a 60-year analysis using standardized indices. Natural Hazards and Earth System Sciences, 25(7), 2541–2564. https://doi.org/10.5194/nhess-25-2541-2025

Van Loon, A. F. (2015). Hydrological drought explained. WIREs Water, 2(4), 359–392. https://doi.org/10.1002/wat2.1085

Xiong, H., Han, J., & Yang, Y. (2025). Propagation From Meteorological to Hydrological Drought: Characteristics and Influencing Factors. Water Resources Research, 61(4), e2024WR037765. https://doi.org/10.1029/2024WR037765

Citation: https://doi.org/10.5194/egusphere-2025-4791-RC1
- AC1: 'Reply on RC1', Yuanrui Liu, 03 Feb 2026
  
  We sincerely thank the reviewer for taking the time to review our manuscript and for providing these valuable comments and suggestions. They have been very helpful in improving the quality of our work. Our point-by-point responses are provided in supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2025-4791-AC1
RC2:
'Comment on egusphere-2025-4791', Yao Wang & Haiyun Shi (co-review team), 20 Dec 2025
General Comments: This study systematically analyzes the propagation characteristics across meteorological, hydrological, and agricultural droughts on a global scale using multi-source datasets and explores their driving factors. The research topic holds clear scientific value, and the ensemble analysis of multiple datasets effectively reduces the uncertainties associated with single-source data. However, I believe the manuscript has not yet reached a consensus on several critical issues, which require further revision and clarification by the authors. Therefore, I recommend a Major Revision. My specific comments are as follows:

Main Critique on Methodology and Physical Interpretation: The study employs both the Response Time (RT) based on time-lag correlation and the Lag Time (LT) based on run-theory event identification to analyze drought propagation from the dual dimensions of statistical association and event evolution. However, I contend that there is a fundamental difference in their underlying physical mechanisms. RT reflects the overall synchronicity or "statistical memory" between long-series precipitation, runoff, and soil moisture. Its values are typically larger (e.g., 5–8 months in this study), primarily capturing the integrated system response driven by seasonal cycles, multi-year climate oscillations (e.g., ENSO), and the long-term water storage capacity of basins. In contrast, LT, based on discrete event tracking, focuses on the physical evolution of specific drought pulses within the hydrological cycle, reflecting the instantaneous triggering mechanism of drought signals penetrating from the atmosphere to the land surface; thus, its values are usually much smaller (e.g., 1.2–2.6 months).

The authors must go beyond simply listing these inconsistent indices in tables and provide a rigorous physical explanation for this "numerical gap" in the Discussion section. Specifically, does the long-period RT represent the smoothing effect of basin storage or seasonal cycles on drought signals, while the short-lived LT captures the non-linear rapid response mechanism when the system exceeds a threshold under extreme stress? Without clarifying why statistical correlation and event evolution differ so significantly in magnitude from a physical perspective, readers will find it difficult to judge which indices are more valuable for early warning, and may even question the robustness of the results. Therefore, I expect the authors to add a dedicated section in the revised manuscript to deeply discuss the physical coupling behind these methodological differences and clearly indicate how to weigh the use of these distinct propagation indices under different management needs.

Specific Comments:

The paper analyzes three pathways: M→H, M→A, and H→A. To what extent is the propagation of H→A independent of M? That is, if meteorological drought (M) has already directly driven agricultural drought (A), is the contribution of hydrological drought (H) to A merely a "shadow" of M?

In desert regions with extremely low precipitation, the correlation between SPI and SRI is often meaningless. How were these extreme climatic zones handled in your global assessment, and are the conclusions applicable there?

The authors selected eight factors for attribution. What was the rationale for selecting these specific factors? Why were underlying surface characteristics, such as land use types, not included? These physical surface features often have a more direct impact on drought propagation (especially PR and LT) than NDVI.

Global grid data exhibit strong spatial autocorrelation. If all grid points are fed directly into the XGBoost model, the model may suffer from overfitting or yield erroneous significance levels. Have the authors attempted to prove the robustness of the model?

In the Introduction, please emphasize that meteorological, hydrological, and agricultural systems are not isolated but are coupled through the hydrological cycle.

Add a mention of the "Propagation Threshold" in the Introduction.

In the Data section, the spatial resolution of different datasets should be clarified, and any resampling operations must be mentioned.

More details need to be added to Section 2.4.

It is suggested to add a brief explanation of "Non-significant areas" in Section 3.1.

Figure 2 uses a unified global timeline. Since seasons are opposite in the Northern and Southern Hemispheres, the high response values in February–April might be entirely driven by the Northern Hemisphere. Should these be discussed separately?

Human activities can significantly alter the PR and LT of drought propagation. Have the authors considered quantifying human activities? Although this is challenging, I suggest a rough discussion on this topic.

I strongly recommend placing the propagation maps generated by each individual dataset in the Supplementary Materials.
Citation: https://doi.org/10.5194/egusphere-2025-4791-RC2
- AC2: 'Reply on RC2', Yuanrui Liu, 03 Feb 2026
  
  We sincerely thank the reviewer for taking the time to review our manuscript and for providing these valuable comments and suggestions. They have been very helpful in improving the quality of our work. Our point-by-point responses are provided in supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2025-4791-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2025-4791', Anonymous Referee #1, 08 Dec 2025

Review of “Understanding spatio-temporal patterns of the propagation characteristics across meteorological, hydrological, and agricultural droughts and their influencing factors” by Yuanrui Liu and co-authors
General comment
Liu and co-authors investigate drought propagation characteristics, their variations, and their controls at the global scale by using an ensemble of reanalyses and land surface models, conventional statistical analyses, and an explainable Machine Learning (ML) algorithm. By doing so, they show spatial and temporal differences in drought propagation characteristics across the globe, and the importance of climatic conditions in governing these differences. Despite an already good body of literature on this topic, the use of explainable ML is rather new and the analysis of multiple products adds strength to this work. However, I have some concerns, mainly regarding the exact novelty of this contribution, some methodological choices and the presentation of the work, as reported below.
Major comments

1. Novelty

Which is the specific research gap that this work addresses? This is currently not entirely clear from the title, abstract, and introduction. As the authors also acknowledge in the introduction, global-scale drought propagation studies are already available. I might see the originality of this work being the use of multiple datasets, but in this case, I believe this could be better worked out throughout the manuscript. Not only in the title, abstract, and introduction, to set the reader’s expectations clear, but also later in the manuscript. As a reader, for instance, I would enjoy having more discussion on the most suitable dataset(s) for drought applications. I see that providing a clear recommendation on this may be difficult from the current analyses, since you do not have here observations to benchmark the datasets with, but maybe you could still say something based on expectations on drought propagation that we have from previous observation-based studies? Also, I would find useful to have clear ranges of variation for the drought propagation characteristics from the different datasets in the abstract and conclusions, as an indication of the uncertainties in such characteristics.

2. Agricultural drought definition and propagation from hydrological to agricultural droughts

The adopted agricultural drought definition and the choice of investigating hydrological-to-agricultural drought propagation is not entirely clear to me. Agricultural droughts are introduced in the paper as ‘reduced soil moisture’ (L40), coherently with extensive previous literature (Van Loon, 2015) which refers to agricultural or soil moisture droughts as deficits in the root-zone soil moisture mainly impacting the agricultural sector, following meteorological droughts and potentially leading to hydrological droughts (i.e., deficits in runoff and groundwater, Van Loon, 2015). Previous works therefore mostly investigated the propagation from meteorological to agricultural and then to hydrological droughts, by finding shorter propagation times from meteorological to agricultural droughts than from meteorological to hydrological droughts (e.g., Odongo et al., 2023; Teutschbein et al., 2025). Could you clarify why you chose to investigate the propagation from hydrological to agricultural droughts instead of the other way round? Also, are you considering soil moisture data from the upper or deeper layers? This is not specified in the methods. In the discussion, agricultural droughts are said to affect the ‘deeper soil moisture’ (L393), which may point to the use of deep soil moisture data only, but I ask you to clarify earlier on this important piece of information for the general understanding of the work. The use of deep soil moisture would (partly) explain to me both the choice of investigating this drought propagation pathway and the results, showing longer propagation times for agricultural than for hydrological droughts with respect to the meteorological ones. Yet, if this is the case, I wonder whether the use of the term agricultural droughts is the most suitable here, given the interest in the upper soil layer by the agricultural sector.

3. Trend analysis

Important methodological details on this are missing. The analysis is briefly described in the Results section (L251–252), but how this moving-window trend analysis exactly works is not totally clear. From my understanding, you calculate the various metrics (e.g., propagation time) for each year based on a moving window consisting of (the next?) 30 years and then apply a trend analysis on the annual values that you obtained. If this is the case, how do you deal with potential autocorrelation from partially overlapping raw data? I would suggest expanding on this point and moving the current description of this analysis to the Methods section (e.g., to a new subsection between the current 2.4 and 2.5). Please also provide full name and appropriate references for the statistical tests used here (i.e., the ‘M-K test’ currently mentioned in the text). Finally, what do you mean by ‘monotonic trend’ in the pie charts in e.g. Fig. 3? From my understanding, it refers to the greyish areas in the maps, with trend slopes close to zero. Did you set any lower and upper limits on the trend slopes to discriminate these ‘monotonic trends’? If so, please specify. And why are these monotonic trends not appearing in the pie charts in Fig. 5 and 6, even though greyish areas are reported in the corresponding maps?

4. Language and readability

I find the paper generally well structured, but rather lengthy and sometimes convoluted. I think the reading flow could be improved by reducing redundant expressions (e.g., couldn’t ‘feature factors’ be simply ‘features’ or ‘factors’?), repetitions between sections (e.g., L105–107 already said in the previous section), and rather obvious statements (e.g., ‘with positive correlation with r_P > 0, and negative correlation with r_P < 0’, L156–157). I also noticed many abbreviations, especially in Sect. 2.1 Datasets (e.g., ECMWF, CLSM, etc), which seem to me not used anymore in the paper. I would suggest removing them and making sure that abbreviations are always introduced the first time they are used (currently not the case, see e.g., ML at L73). Consistent notation throughout the text and across the text and the figures (currently not the case, see e.g., Eq.1 and Fig. 1b, d, and f) would also ease the readability of the paper. In summary, I see room for improvement, with another careful round of proofreading focused on language.

5. References

I appreciate the referencing to very recent literature on the topic, yet I believe that also additional references to (older) seminal papers on droughts and drought propagation characteristics would be appropriate (e.g., López-Moreno et al., 2013 and Barker et al., 2016 for the correlation analysis, other papers that I referred to above).
Minor comments

6. Abstract, I would appreciate introducing the datasets and methods you used in general terms (e.g., reanalyses for ERA5 and so on), to facilitate readers potentially not familiar with these specific datasets and methods.

7. L26–28, I suggest rephrasing since, from my understanding, Gebrechorkos et al. (2025) showed that increases in atmospheric evaporative demand significantly contributed to recent increases in drought severity, but not as primary factor.

8. L42–43, I believe defining here drought propagation characteristics would be beneficial for readers who may not be familiar with them and to ease the readability of the rest of the manuscript.

9. L44–48, many different methods are currently mixed together in this sentence and specifically: methods used to generate the datasets needed for drought propagation studies (e.g., hydrological models), methods used to quantify drought propagation characteristics (e.g., correlation analysis and run theory), and methods used to attribute these characteristics to their controls (e.g., ML). I suggest clarifying this point, for instance by splitting this long sentence into several ones. In addition, maybe add event-coincidence analysis as another method to quantify drought propagation characteristics as proposed by Baez-Villanueva et al. (2024)? Finally, I would suggest removing the mention to the complex network theory since this is used for spatial drought propagation, which is not the topic of this paper, or alternatively, specifying this point and what spatial drought propagation is.

10. L52, I suggest introducing the concept of groundwater droughts earlier.

11. L64–65, I would say that all the global-scale analyses cited before are ‘consistent’ and ‘comparable’ within themselves since they use common methods and datasets for the whole globe. I would suggest rephrasing this sentence to the exact research gap you are aiming at addressing with your work (see also comment #1).

12. L69–70, I do not fully agree with this sentence, which seems to me also contradicting the previous one. Literature on the factors controlling drought propagation across different climatic and geographical regions is rather vast now (see e.g., Xiong et al. 2025 and other reviews on the topic, also cited in the text).

13. L93, could you provide some references of these comparisons?

14. L105, I would argue that data quality is crucial for any study, not only drought studies. It may also be a matter of personal taste, but I do not see as really needed these very general sentences, which also contribute to making the paper quite lengthy in my view (see also comment #4).

15. L118 ‘high-spatial-resolution’, explicitly mentioning the resolution of this and all other datasets would help the readers in my opinion.

16. L121–123, I appreciate the details provided here on how potential evapotranspiration, runoff and soil moisture are computed in this dataset. Could you provide such details also for the other datasets? I think they would give to the reader more context, also on the differences you detect among them.

17. L140–143, some references to support these statements would be appreciated since the differences between a standardized approach and other methods for drought identification are well discussed in many papers now.

18. L154, which is the maximum accumulation period n that you tested? I assume 24 months from Fig. 1, but please specify.

19. L163–166, clarification on why you applied these rules for drought selection would be appreciated.

20. L195–196, I assume the factors that you use as model predictors regarding precipitation, temperature, potential evapotranspiration, runoff, soil moisture, and vegetation conditions are long-term averages, but I encourage you to specify this point in the text. If so, which period did you consider for averaging? Also, why did you choose these specific factors? Other factors regarding e.g. soil or geology would also be important in my view. In general, additional details on the models would be beneficial (e.g., training and validation periods, achieved model performances, etc).

21. Fig.1, caption, please expand the abbreviations (in other captions as well, in the supplementary figures too). Also, some details are missing in this specific caption (e.g., inner plots - where axes are not labelled – and p value for statistical significance). I recommend you having another check that all information needed to fully understand the figures are reported in the captions, including those in the supplementary.

22. L212–213, it is not entirely clear to me what you mean by ‘maximum correlation coefficients’. I suggest rephrasing. I would say that the robustness of your assessments comes from the relatively high correlation coefficients in Fig. 1 and their statistical significance.

23. Fig. 2, wouldn’t considering relative months from the start of the local water year rather than calendar months easier here? This would allow not to mix different processes occurring in the same months in the northern and southern hemispheres. Also, units for correlation coefficients are missing in the axes labels.

24. Fig. 3, caption, which correlation is not statistically significant in blank grid cells? If I understand this analysis correctly (see comment #3), you computed multiple correlations here. Please specify.

25. Fig. 4, which dataset does this figure refer to? The same comment applies also to other figures. I assume all the figures in the main text refer to the ensemble mean, but I would recommend specifying this somewhere. With respect to this specific figure, I would also suggest correcting the label in the colour bars in panels a, c, and e to ‘Propagation rate’ rather than ‘Response rate’ for consistency with the rest of the manuscript and specifying in the caption the different y-axis in panel b as compared to panels d and f. An additional comment on the analysis behind this figure: could the very low propagation rates in panels c and e be due to the time scale that you use? From my understanding, you are not considering any time lag between drought types, even though you show that some drought types can occur well after others in e.g., Fig. 1.

26. Fig. 5, why are the blank areas here the same across the three maps? This is not the case in Fig. 3, which makes sense to me. Also, please add more information to this caption, in a similar way to what done for Fig. 3.

27. Fig. 6 shows decreasing trends in the lag time between meteorological and hydrological droughts in large portions of Europe and northern Asia that are not reflected in the propagation time though (Fig. 3). Do you have an explanation for such discrepancies? Also, I see that attributing these trends to their causes might be outside of the scope of this current paper, but I think that some discussion on potential causes of these trends would still be valuable.

28. L327–328, ‘the SHAP value indicates that high temperatures have shortened the response time of meteorological drought to hydrological drought’, I stumbled a bit here. I suggest rephrasing, for instance by using present tenses, not to evoke changes over time, which is not what you are looking at with your SHAP-analysis. This comment applies also to L409–411.

29. Fig. 9 and corresponding text, I would suggest turning ‘key feature factors’ to ‘dominant factors’ to enhance the clarity of which factors you are looking at here. Also, for Fig. 9, do you have an explanation for the non-monotonic behaviour of meteorological-to-hydrological drought propagation characteristics across different quantiles of the considered features?

30. L429–435, this part sounds to me more like a discussion of the implications rather than of uncertainties. I suggest moving to a new subsection or rename the current one.

31. L438–440, I suggest some rephrasing here to improve the clarity of this sentence.
Technical corrections

32. L45, I cannot find the reference (Zhu et al., 2021) in the reference list, please add it. In addition, include also Xiong et al. (2025), already cited in the text, among the global-scale studies?

33. L47, there are two entries for both (Yang et al., 2024) and (Shi et al., 2022) in the reference list. Which one are you referring to here? Please specify here and elsewhere in the manuscript.

34. L127, I assume you mean here ‘elevation’ rather than ‘evaluation’. Please correct it, here and throughout the manuscript.

35. L165, ‘with on month’ -> ‘with one month’? Else, the sentence sounds strange to me. Please check.

36. L190, ‘formula’ -> ‘formulated’?

37. L193, ‘influencing on the model predictions’ -> ‘influencing the model predictions’

38. L282, ‘the highest PR_MH and LT_MH values’ -> the highest PR_MH and lowest LT_MH values?

39. Fig. 8 and 9, please correct the labels in panels c, f, and i with the subscript ‘HA’ instead of ‘MA’.

40. L303–304, I suggest rephrasing or removing, since this wording does not sound fitting to this subsection to me.

41. L344–345, the first ‘hydrological’ should probably be ‘agricultural’.

42. L413, ‘reasons’ -> ‘seasons’?

43. L424, Figures 11-13, these figures are reported in the supplement. Please correct.

44. L442, ‘finding’ -> ‘findings’?
References

Baez-Villanueva, O. M., Zambrano-Bigiarini, M., Miralles, D. G., Beck, H. E., Siegmund, J. F., Alvarez-Garreton, C., Verbist, K., Garreaud, R., Boisier, J. P., & Galleguillos, M. (2024). On the timescale of drought indices for monitoring streamflow drought considering catchment hydrological regimes. Hydrology and Earth System Sciences, 28(6), 1415–1439. https://doi.org/10.5194/hess-28-1415-2024

Barker, L. J., Hannaford, J., Chiverton, A., & Svensson, C. (2016). From meteorological to hydrological drought using standardised indicators. Hydrology and Earth System Sciences, 20(6), 2483–2505. https://doi.org/10.5194/hess-20-2483-2016

Gebrechorkos, S. H., Sheffield, J., Vicente-Serrano, S. M., Funk, C., Miralles, D. G., Peng, J., Dyer, E., Talib, J., Beck, H. E., Singer, M. B., & Dadson, S. J. (2025). Warming accelerates global drought severity. Nature, 642(8068), 628–635. https://doi.org/10.1038/s41586-025-09047-2

López-Moreno, J. I., Vicente-Serrano, S. M., Zabalza, J., Beguería, S., Lorenzo-Lacruz, J., Azorin-Molina, C., & Morán-Tejeda, E. (2013). Hydrological response to climate variability at different time scales: A study in the Ebro basin. Journal of Hydrology, 477, 175–188. https://doi.org/10.1016/j.jhydrol.2012.11.028

Odongo, R. A., De Moel, H., & Van Loon, A. F. (2023). Propagation from meteorological to hydrological drought in the Horn of Africa using both standardized and threshold-based indices. Natural Hazards and Earth System Sciences, 23(6), 2365–2386. https://doi.org/10.5194/nhess-23-2365-2023

Teutschbein, C., Grabs, T., Giese, M., Todorović, A., & Barthel, R. (2025). Drought propagation in high-latitude catchments: Insights from a 60-year analysis using standardized indices. Natural Hazards and Earth System Sciences, 25(7), 2541–2564. https://doi.org/10.5194/nhess-25-2541-2025

Van Loon, A. F. (2015). Hydrological drought explained. WIREs Water, 2(4), 359–392. https://doi.org/10.1002/wat2.1085

Xiong, H., Han, J., & Yang, Y. (2025). Propagation From Meteorological to Hydrological Drought: Characteristics and Influencing Factors. Water Resources Research, 61(4), e2024WR037765. https://doi.org/10.1029/2024WR037765

Citation: https://doi.org/10.5194/egusphere-2025-4791-RC1
- AC1: 'Reply on RC1', Yuanrui Liu, 03 Feb 2026
  
  We sincerely thank the reviewer for taking the time to review our manuscript and for providing these valuable comments and suggestions. They have been very helpful in improving the quality of our work. Our point-by-point responses are provided in supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2025-4791-AC1
RC2:
'Comment on egusphere-2025-4791', Yao Wang & Haiyun Shi (co-review team), 20 Dec 2025
General Comments: This study systematically analyzes the propagation characteristics across meteorological, hydrological, and agricultural droughts on a global scale using multi-source datasets and explores their driving factors. The research topic holds clear scientific value, and the ensemble analysis of multiple datasets effectively reduces the uncertainties associated with single-source data. However, I believe the manuscript has not yet reached a consensus on several critical issues, which require further revision and clarification by the authors. Therefore, I recommend a Major Revision. My specific comments are as follows:

Main Critique on Methodology and Physical Interpretation: The study employs both the Response Time (RT) based on time-lag correlation and the Lag Time (LT) based on run-theory event identification to analyze drought propagation from the dual dimensions of statistical association and event evolution. However, I contend that there is a fundamental difference in their underlying physical mechanisms. RT reflects the overall synchronicity or "statistical memory" between long-series precipitation, runoff, and soil moisture. Its values are typically larger (e.g., 5–8 months in this study), primarily capturing the integrated system response driven by seasonal cycles, multi-year climate oscillations (e.g., ENSO), and the long-term water storage capacity of basins. In contrast, LT, based on discrete event tracking, focuses on the physical evolution of specific drought pulses within the hydrological cycle, reflecting the instantaneous triggering mechanism of drought signals penetrating from the atmosphere to the land surface; thus, its values are usually much smaller (e.g., 1.2–2.6 months).

The authors must go beyond simply listing these inconsistent indices in tables and provide a rigorous physical explanation for this "numerical gap" in the Discussion section. Specifically, does the long-period RT represent the smoothing effect of basin storage or seasonal cycles on drought signals, while the short-lived LT captures the non-linear rapid response mechanism when the system exceeds a threshold under extreme stress? Without clarifying why statistical correlation and event evolution differ so significantly in magnitude from a physical perspective, readers will find it difficult to judge which indices are more valuable for early warning, and may even question the robustness of the results. Therefore, I expect the authors to add a dedicated section in the revised manuscript to deeply discuss the physical coupling behind these methodological differences and clearly indicate how to weigh the use of these distinct propagation indices under different management needs.

Specific Comments:

The paper analyzes three pathways: M→H, M→A, and H→A. To what extent is the propagation of H→A independent of M? That is, if meteorological drought (M) has already directly driven agricultural drought (A), is the contribution of hydrological drought (H) to A merely a "shadow" of M?

In desert regions with extremely low precipitation, the correlation between SPI and SRI is often meaningless. How were these extreme climatic zones handled in your global assessment, and are the conclusions applicable there?

The authors selected eight factors for attribution. What was the rationale for selecting these specific factors? Why were underlying surface characteristics, such as land use types, not included? These physical surface features often have a more direct impact on drought propagation (especially PR and LT) than NDVI.

Global grid data exhibit strong spatial autocorrelation. If all grid points are fed directly into the XGBoost model, the model may suffer from overfitting or yield erroneous significance levels. Have the authors attempted to prove the robustness of the model?

In the Introduction, please emphasize that meteorological, hydrological, and agricultural systems are not isolated but are coupled through the hydrological cycle.

Add a mention of the "Propagation Threshold" in the Introduction.

In the Data section, the spatial resolution of different datasets should be clarified, and any resampling operations must be mentioned.

More details need to be added to Section 2.4.

It is suggested to add a brief explanation of "Non-significant areas" in Section 3.1.

Figure 2 uses a unified global timeline. Since seasons are opposite in the Northern and Southern Hemispheres, the high response values in February–April might be entirely driven by the Northern Hemisphere. Should these be discussed separately?

Human activities can significantly alter the PR and LT of drought propagation. Have the authors considered quantifying human activities? Although this is challenging, I suggest a rough discussion on this topic.

I strongly recommend placing the propagation maps generated by each individual dataset in the Supplementary Materials.
Citation: https://doi.org/10.5194/egusphere-2025-4791-RC2
- AC2: 'Reply on RC2', Yuanrui Liu, 03 Feb 2026
  
  We sincerely thank the reviewer for taking the time to review our manuscript and for providing these valuable comments and suggestions. They have been very helpful in improving the quality of our work. Our point-by-point responses are provided in supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2025-4791-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (12 Feb 2026) by Manuela Irene Brunner

AR by Yuanrui Liu on behalf of the Authors (13 Feb 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (19 Feb 2026) by Manuela Irene Brunner

RR by Yao Wang & Haiyun Shi (co-review team) (03 Mar 2026)

RR by Anonymous Referee #1 (20 Mar 2026)

Suggestions for revision or reasons for rejection

Re-review of “Understanding meteorological, hydrological, and agricultural drought propagation and their influencing factors in an ensemble of multiple datasets” by Yuanrui Liu and co-authors

I thank the authors for their responses to my conceptual and methodological concerns and for addressing my previous comments regarding the presentation of their results. The manuscript has now improved in my view. I still have however a couple of comments on the manuscript for the authors, the major being on the use of the term hydrological drought.

1. Thank you for clarifying the considered soil layers and your reasoning for investigating the propagation from hydrological to agricultural droughts, this helps. To avoid any confusion around the term hydrological drought though, as you also acknowledge at L517–520, I would recommend using the term ‘runoff drought’ instead of ‘hydrological drought’ throughout the manuscript (including in the title).

2. L13–15, I see what you mean by the terms ‘meteorological’, ‘hydrological’ and ‘agricultural’ in the brackets, but their placement in the sentence looks rather strange to me, I would suggest some rephrasing.

3. L45–48, what is drought propagation threshold? This is not clear to me from this sentence. I would suggest briefly describing these characteristics here, rather than just listing them.

4. L126, could you please introduce also the spatial resolutions of the raw datasets? I believe this is what comment #8 from the other reviewers referred to as well.

5. L194, the header of this subsection is rather cryptic to me, I would suggest some rephrasing.

6. L526, ‘these regions’ seems now referring to hyper-arid regions only, while it should refer to all the regions listed in L531–533 in my view. Consider some rephrasing. Also, in L533 there is still an ‘evaluation’ instead of ‘elevation’.

7. L545–547, I would recommend supporting this statement with some references. Disentangling the effects of human activities on drought propagation at a global scale is surely challenging, but we do have some evidence on the effects of human activities for some specific case studies and drought types at least.

8. L556, ‘finding’ should be ‘findings’.

Hide

ED: Publish subject to minor revisions (review by editor) (14 Apr 2026) by Manuela Irene Brunner

AR by Yuanrui Liu on behalf of the Authors (22 Apr 2026) Author's response Author's tracked changes Manuscript

ED: Publish subject to technical corrections (24 Apr 2026) by Manuela Irene Brunner

AR by Yuanrui Liu on behalf of the Authors (28 Apr 2026)

AR by Yuanrui Liu on behalf of the Authors (28 Apr 2026) Author's response Manuscript

Journal article(s) based on this preprint

08 May 2026

Understanding meteorological, runoff, and agricultural drought propagation and their influencing factors in an ensemble of multiple datasets

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Hydrol. Earth Syst. Sci., 30, 2775–2795, https://doi.org/10.5194/hess-30-2775-2026,https://doi.org/10.5194/hess-30-2775-2026, 2026

Short summary

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Supplement

https://doi.org/10.5194/egusphere-2025-4791-supplement

Yuanrui Liu, Tingting Hu, Jiawen Yang, and Lei Yu

Viewed

Total article views: 2,510 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
1,731	602	177	2,510	266	108	160

HTML: 1,731
PDF: 602
XML: 177
Total: 2,510
Supplement: 266
BibTeX: 108
EndNote: 160

Views and downloads (calculated since 06 Nov 2025)

Month	HTML	PDF	XML	Total
Nov 2025	690	130	49	869
Dec 2025	305	180	55	540
Jan 2026	206	104	15	325
Feb 2026	252	97	40	389
Mar 2026	202	67	15	284
Apr 2026	61	20	3	84
May 2026	15	4	0	19

Cumulative views and downloads (calculated since 06 Nov 2025)

Month	HTML	PDF	XML	Total
Nov 2025	690	130	49	869
Dec 2025	305	180	55	540
Jan 2026	206	104	15	325
Feb 2026	252	97	40	389
Mar 2026	202	67	15	284
Apr 2026	61	20	3	84
May 2026	15	4	0	19

Viewed (geographical distribution)

Total article views: 2,497 (including HTML, PDF, and XML) Thereof 2,497 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 May 2026

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (3627 KB)
Metadata XML

Short summary

Understanding drought propagation is vital for disaster preparedness and risk management. This study presents a comprehensive analysis of various drought conditions across global land areas. Interpretable machine learning technique is employed to identify the key factors influencing drought propagation. Results reveal large-scale propagation pathways of meteorological-hydrological-agricultural droughts, and highlight how climatic characteristics affect these dynamics.


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