the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Sep 2025
Shifting water scarcities: Irrigation alleviates agricultural green water deficits while increasing blue water scarcity
Abstract. Agricultural areas often experience green water scarcity – i.e. soil moisture limitation on crop growth – due to e.g. unfavourable soil texture, high potential evapotranspiration rates, poor or inefficient crop management, and fluctuations in meteorological conditions. Driven by the growing effects of climate change and the rising water and food demands of an increasing world population, agricultural green water scarcity is becoming an increasingly important phenomenon. In this global modelling study, a plant-physiology based indicator of green water stress is applied, that quantifies the ratio between soil moisture limitation and atmospheric water demand on agricultural areas. Results show that currently (2015–2019 average) 37 % of the global agricultural area is green water stressed, where this ratio is >0.2. Hotspots are characterized by a high seasonal variability in stress conditions, and are mainly located in India and Pakistan, northern Sub-Saharan Africa, North Africa and southwestern Asia. Using an analogous blue water stress indicator – which relates human water use for households, industry and agriculture to available blue water resources in rivers, reservoirs and aquifers – current irrigation is shown to alleviate green water stress on 13 % of the total agricultural area (207 Mha) but simultaneously increases the share of areas experiencing blue water stress by 12 % (199 Mha). Moreover, on average 585 km3 yr-1 of water used for irrigation (22 % of the total water use) is found to stem from surface water resources at the expense of rivers' environmental flow requirements. This shift in water stress types highlights the importance of jointly considering the interconnected green and blue water resources and stresses in pathways towards sustainable water use in agriculture.
-
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
(1398 KB)
-
Supplement
(2234 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1398 KB) - Metadata XML
-
Supplement
(2234 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-3817', Lorenzo Rosa, 26 Sep 2025
-
AC1: 'Reply on RC1', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
-
AC1: 'Reply on RC1', Heindriken Dahlmann, 01 Dec 2025
-
RC2: 'Comment on egusphere-2025-3817', Anonymous Referee #2, 01 Oct 2025
General comments
This paper sheds light on the interlinkages between green (GWS) and blue water scarcity (BWS) as simulated by the global gridded vegetation model LPJmL. The results are presented for the 2015–2019 period with a sufficient level of detail and supporting visualisations. The main finding suggests that irrigation helps humanity to alleviate GWS impacts on crops, but this comes at the cost of increasing BWS. As such, this paper contributes to the growing body of literature highlighting the need to analyse both green and blue water use simultaneously when addressing water stress impacts on crop production worldwide.
However, as always, there are multiple things the authors can improve upon. My main concern lies in the definition of GWS and how robust the results are. Choosing 0.2 as the main GWS threshold seems rather arbitrary and leaves me wondering how different the results would be if the authors had chosen 0.1 or 0.3 instead (I explain below in detail). Adding to this, there is no clear section on the validation of the underlying global estimates by LPJmL. The full list of potential issues to address is provided below.
Specific comments:
Regarding the definition of GWS:
- Authors use the 0.2 GWS threshold as earlier applied by Rosa et al. (e.g. https://doi.org/10.1093/pnasnexus/pgad117). The assumption is that such a level of GWS would force farmers to irrigate to avoid large yield declines. However, this assumption is rather arbitrary and does not rely on actual crop yield decline simulations (as you also acknowledge in L245-246). As far as I know, LPJmL is capable of simulating crop yields, so why not using those to define reasonable thresholds per crop functional type (CFT)? I can imagine that some crops indeed can have substantial yield declines at 0.2 GWS, but others might not be so sensitive.
- Connected to the above, consider reframing the analysis from only two main GWS categories (0.2-0.4 and 0.4-1.0) to sequential steps of 0.2 or 0.25 to have graduality in GWS exposure (e.g. no GWS, minor, moderate, severe). Authors already do this for Fig. 4, for example, why not in the main text and Table 1 too?
- Authors define green water as “the plant available rainwater held in soils which sustains the growth of crops and pastures”, and then they also state that “irrigation is shown to alleviate green water stress”. I find this a bit confusing. If green water = rainwater, then how come irrigation can alleviate GWS? The total water stress (from insufficient green+blue water supply) can indeed be reduced by supplementary blue water supply via irrigation, but green water scarce areas remain green water scarce by definition of what green water is. There is still not enough rainwater in the soil even after adding irrigation, and this statement would hold unless the precipitation patterns and/or soil texture change. Therefore, please consider rephrasing the parts where such confusion can occur or perhaps use total water stress (or some other general term) to make the distinctions clearer.
Regarding the definition of BWS:
- Is there a particular reason why water use (as I understand, that means withdrawals) as opposed to water consumption (actual water volume removed from a catchment) is used? To my knowledge, most recent BWS studies use the latter since a large part of the withdrawals stays within the same catchment, and thus, should not contribute to BWS in a long term.
- It is not clear where water use estimates for domestic and industrial use are taken from, and whether they account for monthly variability. Please add more details.
- It is not clear whether the authors consider groundwater withdrawals as well as livestock, electricity generation, and mining water supplies. Please elaborate.
Regarding the model setup:
- I find the description of LPJmL inputs in the main text insufficient. This study relies entirely on results from this model, so, as a reader, I would expect to have a very detailed description of input data and main assumptions in Section 2.2, and validation, limitations, and uncertainties in Section 4. Please try to elaborate. I provide several suggestions below.
- In addressing 3a, I would recommend moving Table S1 into the main text, as a reader needs to see the main input data sources and their description. Also, consider adding a column with spatial resolution next to “Time period”.
- When describing the land use and CFTs, please explain:
- where the LandInG toolbox gets rainfed/irrigated crop maps from, and whether those maps cover spatiotemporal changes
- where crop calendars are from, and whether they are static or dynamic
- what share of primary crops from the crop list provided by FAOSTAT is covered (otherwise, it is not clear what your scope is), see L85-87. Also mention if perennial crops are covered.
- When describing the soil, please specify whether the root zone is dynamic or kept static over the growing season, mention how many soil layers are simulated, and whether the shallow groundwater is considered (since capillary rise can support some rainfed crops).
- When describing the irrigation, clearly state what irrigation systems are considered (furrow, sprinkler, drip), how and when irrigation is applied, and whether conveyance losses are simulated.
Regarding validation, limitations, and uncertainties:
- As mentioned earlier, I would expect to see a sub-section on validation, limitations, and uncertainties in Section 4. The presently provided comparisons mainly look at GWS and BWS in terms of hectares. However, the underlying gridded estimates of green and blue ET, as well as sectoral blue water demand, are not properly validated. Please add such comparisons (global total and/or gridded levels), maybe a brief summary for the main text and an elaborated version in SI. Otherwise, it is difficult to judge whether the LPJmL model outputs are reliable before even diving into GWS/BWS analysis.
- For green and blue ET (i.e. water consumption) estimates, you can have a look at https://doi.org/10.1038/s41597-020-00612-0 and https://doi.org/10.1038/s41597-024-03051-3, but there also might be more studies published recently.
- Sectoral blue water use can be obtained from global hydrological models provided by ISIMIP as well as AQUASTAT
- Lastly, I would recommend having an additional sub-section describing 1) main limitations and uncertainties (coming from input data, LPJmL-specific ones, etc.) and 2) discussing how those limitations and uncertainties affect the reliability of the results. For example, how BWS could change if the authors used a water consumption-based approach instead of water withdrawals or how different EFR methods could affect the results. No need to run additional simulations to provide an uncertainty range, but it is always a good practice to be open about the weaknesses of the selected methodology, so the follow-up studies can address those.
Technical corrections:
- L13: Authors define GWS as “soil moisture limitation on crop growth” in the abstract without mentioning that it only concerns rainwater.
- L30: Please consider using more recent publications to support your statement.
- L32: Authors mention green water scarcity without defining what it is.
- L36: Rainfed croplands can also get blue water via capillary rise, see https://doi.org/10.1088/1748-9326/ad78e9. Also, correct the citation for (X. Liu et al., 2022) to fit journal guidelines (check through the manuscript for other instances).
- L47: Maybe use EFR instead of “environmental water requirements” since you already introduced it a few lines earlier.
- L62: Define what the LPJmL abbreviation is.
- L98-99: It is not clear why running the 1901-2019 period with a 3500-year spin-up is needed. The analysis is mainly for 2015-2019. So why starting in 1901, and why so many spin-up years? Please elaborate.
- L112: Please explain what 0.01 rate is (units? physical meaning?) and how it was selected.
- L120: Based on what 8mm/d as max is defined?
- L126-127: Please provide references to the coefficient and scaling factor (or explain how you calculated them)
- Please add units to Eq. 2 and 3.
- How do you calculate annual representative GWS values based on monthly numbers? I find it a bit unclear. Perhaps add an equation to better explain L137-139.
- L148: WUdom, WUind, and WUirr are not defined.
- L187-188: consider adding a sentence with additional global GWS estimates for each year or simply provide a range (min, max) during 2015-2019.
- In Table 1, the total area for ILIM under GWS>0.2 is larger for rainfed crops (426+213 Mha) than for all crops combined (431+161 Mha). Please double-check.
- L197-198: If 37% of all croplands are under GWS>0.2, then logically the remaining lands (63%) are under GWS<0.2. However, the authors say it is only 13%. Please double-check.
- L202-204: From the text, it is not immediately clear what the difference is with L195-197. After a few minutes, I finally realised that it is only about irrigated grid cells here (rainfed excluded). Please make it clear for the reader.
- L204: If indeed only irrigated cells are analysed, please explain why there are still large GWS areas remaining (once you switch on irrigation). Is it because some irrigated cells cannot have access to a sufficient blue water supply, so the crops remain under water deficit? If so, how realistic is this assumption? I would expect that such underirrigation is not that common globally, since farmers can always dig into groundwater reserves or simply ignore EFR (perhaps mention this in the sub-section on limitations and uncertainties).
- L296: Consider providing a few examples of green water management options.
As you can see, I took a great responsibility of being the “annoying reviewer #2”. Hope my constructive criticism will be helpful in improving the quality of this paper, which I am truly looking forward to reviewing in the next round of revisions. Good luck!
Citation: https://doi.org/10.5194/egusphere-2025-3817-RC2 -
AC2: 'Reply on RC2', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
-
RC3: 'Comment on egusphere-2025-3817', Anonymous Referee #3, 01 Oct 2025
The study leverages two indicators: green water scarcity (GWS ) and blue water scarcity (BWS), to quantify and examine the interconnectedness of green and blue water resources for irrigation. They show that while irrigation alleviates GWS, it simultaneously exacerbates BWS at the expense of environmental flow supply. The paper successfully addresses the research gap with clear narration and consistent structure throughout. Given that the results hinge on several salient parameters (Emax, threshold values), I recommend that authors consider conducting a sensitivity analysis to strengthen their analysis and to enable both themselves and the readers to better gauge the robustness of the results.
Major comments:
- There is still room to further improve the transparency of the modelling process. This could include making assumptions more explicit and providing justification or citations for modelling decisions so that readers can trace and assess their epistemic quality/reasoning. Such steps would also support future scholars in replicating the work based solely on the provided documentation. Please see the following details:
- L. 90: Could you expand on the parameterisation of irrigation efficiency? Since CFTs may be suitable to more than one irrigation method (see Jagermeyr et al 2015), how did you determine which method to apply? Was the efficiency value based on an area- or grid-cell-weighted average?
- L. 120: Could you provide a rationale for using Emax = 8 mm/day? Earlier studies suggest lower values (e.g., Gerten et al 2004 report 5-7 mm/day, while Rost et al, 2008 used 5 mm/day). If more recent studies support your chosen value, it would be helpful to cite them.
- L. 129: The statement “If S > D, we set S = D to ensure the result remains within [0,1]” could benefit from clearer justification. For example: “When S>D, we set S = D, because plants cannot take up more water than their transpiration demand allows.” This framing preserves the rationale while adding a physiological explanation.
- L. 143: Could you elaborate on the rationale for selecting 0.4 as the threshold to delineate highly water-scarce conditions? Since this assumption conditions the characterisation of scarcity across regions, you may consider conducting a sensitivity analysis to assess the soundness of your results to alternative threshold values.
- L. 256: It may be helpful to mention the non-representation of multiple cropping already in the methodology, rather than only in the discussion, as this limitation (as you argued) could underestimate your results.
- Given that the results are highly contingent on several parameterisations, it may be beneficial for the authors to conduct a sensitivity analysis to strengthen their conclusions and to allow stakeholders to better gauge the robustness of the findings. For example, sensitivity runs on influential elements such as Emax values (5-8 mm/day) and different thresholds (0.2, 0.3, 0.4, 0.6) may be informative and also support your choice of values.
Minor comments:
- For consistency with previous LPJmL studies, please consider standardising the equation symbols:
- L. 124: am should be written as αm
- L. 124: adjust the fraction form of gm/gc for clarity
- The authors may consider aligning the numbering of supplementary materials with the order in which they are first referenced in the manuscript. Additionally, there are several information and figures included in the supplementary material that are not explicitly mentioned in the main text. Without such references, readers may overlook these potentially important resources.
Technical comments:
- When citing, if applicable, it may be helpful to specify that exact page, figure or table for the reader’s reference. For example, L. 90: (Table 5, Jagermeyr et al. 2015)
- L. 156: Please consider removing the comma following “overuse”
- L. 156: It may improve readability to elaborate on the meaning of transgression: clarifying what it implies when a threshold has been transgressed.
- L. 264-266: The sentence “While irrigation reduces GWS (below the threshold of 0.2) on 13% of the global agricultural area, it thereby leads to an increase in areas experiencing moderate BWS as well as high BWS by 6%, respectively.” may be confusing, as “respectively” suggests a missing distinction between moderate and high BWS. Consider rephrasing for clarity, e.g., “…by 6% and 6%, respectively” or “…both by 6%”
Citation: https://doi.org/10.5194/egusphere-2025-3817-RC3 -
AC3: 'Reply on RC3', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
- There is still room to further improve the transparency of the modelling process. This could include making assumptions more explicit and providing justification or citations for modelling decisions so that readers can trace and assess their epistemic quality/reasoning. Such steps would also support future scholars in replicating the work based solely on the provided documentation. Please see the following details:
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-3817', Lorenzo Rosa, 26 Sep 2025
Revision:
“Shifting water scarcities: Irrigation alleviates agricultural green water deficits while increasing blue water scarcity”
The authors investigate how irrigation reshapes agricultural water stress worldwide by shifting pressure from green water scarcity (GWS) to blue water scarcity (BWS). They simulate crop water use with the global vegetation–crop model LPJmL at 0.5° daily resolution, contrasting a no-irrigation counterfactual (INO) with a limited-irrigation case (ILIM) where withdrawals are constrained to locally available blue water (including upstream inflow and reservoirs, but excluding long-distance transfers and fossil groundwater). GWS and BWS are then quantified with monthly indices.
The paper’s key innovation is to treat GWS and BWS jointly, which yields clear global insights: irrigation alleviates GWS on 13% of cropland are but increases BWS by ~12%, concentrating the latter in some irrigation hotspot (e.g., in India and the Mediterranean basin). The estimates of blue water overuse are broadly consistent with published studies; however, the manuscript should explicitly state what is meant by “water use” (i.e., whether it refers to water consumption or to withdrawals). Moreover, it remains unclear whether renewable groundwater is explicitly included in the accounting of available blue water. Clarifying these points would strengthen the interpretation of overshoot volumes.
The manuscript is well written, the narrative is coherent, and figures, tables, and supplementary materials effectively support the analysis. Overall, this is a solid and valuable contribution; I recommend minor revisions focused on clarifying the treatment of renewable groundwater in the blue-water budget and related sensitivity.
Specific comments:
** lines 39-40**: This sentence would benefit from bibliographic references; for example, Gleeson et al. (2020).
Gleeson, T., Wang‐Erlandsson, L., Porkka, M., Zipper, S. C., Jaramillo, F., Gerten, D., ... & Famiglietti, J. S. (2020). Illuminating water cycle modifications and Earth system resilience in the Anthropocene. Water Resources Research, 56(4), e2019WR024957.
**lines 115-116**: Adding (S) and (D) to water supply and atmospheric demand, respectively, would make Eq. (1) clearer and more accessible to readers, since these symbols are not defined immediately afterward.
**Eq.3**: In the ratio between the scaling factor and the potential canopy conductance, the division sign (—) is missing; please check.
**lines 145-147**: Here too, it would be better to introduce the symbols in Eq.4 beforehand. The symbols for the water uses, even if intuitive, should be restated in the text for the sake of completeness.
Moreover, could you please specify what is meant by “water use” in this study? Are you working with consumption or with withdrawals? For water-balance estimates of overuse, the relevant quantity is typically identified by the water consumption, as it represents the fraction removed from the system and not locally available for reuse (hence generally smaller than withdrawals). It would help to state this explicitly in the Methods.
**lines 145-162**: In these paragraphs, it is unclear why additional sectors such as livestock, electricity generation, and mining are not considered. In several regions these are as water-intensive as domestic and industrial uses. Please justify their exclusion (e.g., data gaps, scope) or discuss as a limitation.
It is not specified how the Flörke et al. (1950–2010) database covers the full simulation period, particularly after 2010. Were values held constant at 2010 levels thereafter, as in Rosa & Sangiorgio (2025) and Citrini et al. (2025)? This should be stated explicitly in the Methods and acknowledged as a limitation.
Please clarify why renewable groundwater inflows to grid cells are not included in the blue-water budget (they appear to be excluded in Eq. 4 and Eq. 5). If intentionally omitted, explain the rationale and, in the Discussion section, the implications for overuse estimates.
It would help to add a brief note on the routing module: how upstream–downstream relationships among cells are represented when computing overuse, including whether deficits are calculated locally or propagated, and any reservoir/return-flow assumptions.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F., & Alcamo, J. (2013). Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study. Global Environmental Change, 23(1), 144-156.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**Lines 155-156**: Stenzel et al.’s approach to estimating blue-water overuse appears closely aligned with earlier assessments of unsustainable water use (e.g., Mekonnen & Hoekstra, 2016; Mekonnen & Hoekstra, 2020; Citrini et al., 2025; Rosa & Sangiorgio, 2025 and many others). It would strengthen the manuscript to situate the method explicitly within this literature (briefly clarifying similarities and differences) and, where feasible, to prioritize citations to peer-reviewed, published studies.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Mekonnen, M. M., & Hoekstra, A. Y. (2016). Four billion people facing severe water scarcity. Science advances, 2(2), e1500323.
Mekonnen, M. M., & Hoekstra, A. Y. (2020). Blue water footprint linked to national consumption and international trade is unsustainable. Nature Food, 1(12), 792-800.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**Figure2**: The figure is excellent. One suggestion would be to revisit the colorbar and its tick labels. Because the text consistently refers to thresholds at 0.2–0.4–0.6–0.8–1.0, harmonizing the colorbar bins/ticks with those intervals (similar to the style used in Fig. S11) would improve readability and make cross-references more immediate. If a full reclassification is not desired, introducing color breakpoints at those thresholds would still create a clear visual link to the classes cited in the text. Please take this as an optional refinement to consider.
It would also help to state explicitly in the caption that the colorbar applies to all panels in the figure (as you did in the caption of Figure 4).
**lines 181-184**: Another enhancement to consider is splitting Figure 2 into five panels (2a–e) instead of two, so the manuscript can reference each component explicitly. For example: “The green water stress patterns show a high seasonal variability over the year due to changing weather conditions but also season-specific growing seasons (Fig. 2b-e). Europe and North America do experience less (or even no) GWS during the winter months since the water demand of the crops grown during that time is very low (Fig. 2b). During the summer months (Fig. 2d), however, especially southern regions in Europe and the western US are highly green water stressed (GWS >0.4). Large regions in Brazil change into GWS hotspots from June to November where pulses, rapeseed and sugarcane are especially green water stressed (Fig. 2d-e). India, by contrast, does not experience high GWS from June to November (Fig. 2d-e), when most crops are grown.”
**Figure 5**: The figure appears very similar to the patterns reported by Citrini et al. (2020) for the 2001–2010 baseline (see their Supplementary Fig. 2a). It would strengthen the paper to briefly position your map against those results - highlighting key consistencies or divergences and the likely reasons in the Discussion section. This seems especially pertinent if your overuse metric is restricted to surface-water resources, whereas Citrini et al. used a source-agnostic approach (i.e., not distinguishing whether scarcity arises from groundwater or surface water).
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
**Discussion section**: The Discussion would benefit from engaging with the most recent literature published this year, e.g., Rosa & He (2025) for GWS, and Rosa & Sangiorgio (2025) together with Citrini et al. (2025) for blue-water overuse. Situating your findings alongside these studies (noting agreements, differences, and methodological nuances) would further strengthen the contribution. Could you clarify what drives the difference between the estimates (585 km3yr-1 vs ~460 km3yr-1)? Is the difference mainly due to (i) restricting blue-water availability to surface sources, (ii) using withdrawals rather than consumptive use, or (iii) other methodological choices (e.g., treatment of return flows, reservoirs, EFRs, routing, baseline years)?
In addition, the sentence at lines 302–304 should be supported with appropriate references.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Rosa, L., & He, L. (2025). Global multi-model projections of green water scarcity risks in rainfed agriculture under 1.5° C and 3° C warming. Agricultural Water Management, 314, 109519.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**lines 314-315**: It would be helpful to briefly outline how such these new case studies would be implemented. For example, would higher spatial and/or temporal resolution be required? If so, please indicate the additional data needs (e.g., finer-resolution water use, irrigation, hydrologic, and management datasets) and whether the availability of such data currently constrains feasibility.
**Reference list**: Just a minor formatting note, likely governed by the journal’s template rather than the authors: applying a hanging indent to the paragraph/list would make the items much easier to scan and review.
Thank you for the opportunity to revise the study. Again, this is great work (congratulations!) and minor clarificatory revisions are requested.
Lorenzo Rosa
Citation: https://doi.org/10.5194/egusphere-2025-3817-RC1 -
AC1: 'Reply on RC1', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
-
AC1: 'Reply on RC1', Heindriken Dahlmann, 01 Dec 2025
-
RC2: 'Comment on egusphere-2025-3817', Anonymous Referee #2, 01 Oct 2025
General comments
This paper sheds light on the interlinkages between green (GWS) and blue water scarcity (BWS) as simulated by the global gridded vegetation model LPJmL. The results are presented for the 2015–2019 period with a sufficient level of detail and supporting visualisations. The main finding suggests that irrigation helps humanity to alleviate GWS impacts on crops, but this comes at the cost of increasing BWS. As such, this paper contributes to the growing body of literature highlighting the need to analyse both green and blue water use simultaneously when addressing water stress impacts on crop production worldwide.
However, as always, there are multiple things the authors can improve upon. My main concern lies in the definition of GWS and how robust the results are. Choosing 0.2 as the main GWS threshold seems rather arbitrary and leaves me wondering how different the results would be if the authors had chosen 0.1 or 0.3 instead (I explain below in detail). Adding to this, there is no clear section on the validation of the underlying global estimates by LPJmL. The full list of potential issues to address is provided below.
Specific comments:
Regarding the definition of GWS:
- Authors use the 0.2 GWS threshold as earlier applied by Rosa et al. (e.g. https://doi.org/10.1093/pnasnexus/pgad117). The assumption is that such a level of GWS would force farmers to irrigate to avoid large yield declines. However, this assumption is rather arbitrary and does not rely on actual crop yield decline simulations (as you also acknowledge in L245-246). As far as I know, LPJmL is capable of simulating crop yields, so why not using those to define reasonable thresholds per crop functional type (CFT)? I can imagine that some crops indeed can have substantial yield declines at 0.2 GWS, but others might not be so sensitive.
- Connected to the above, consider reframing the analysis from only two main GWS categories (0.2-0.4 and 0.4-1.0) to sequential steps of 0.2 or 0.25 to have graduality in GWS exposure (e.g. no GWS, minor, moderate, severe). Authors already do this for Fig. 4, for example, why not in the main text and Table 1 too?
- Authors define green water as “the plant available rainwater held in soils which sustains the growth of crops and pastures”, and then they also state that “irrigation is shown to alleviate green water stress”. I find this a bit confusing. If green water = rainwater, then how come irrigation can alleviate GWS? The total water stress (from insufficient green+blue water supply) can indeed be reduced by supplementary blue water supply via irrigation, but green water scarce areas remain green water scarce by definition of what green water is. There is still not enough rainwater in the soil even after adding irrigation, and this statement would hold unless the precipitation patterns and/or soil texture change. Therefore, please consider rephrasing the parts where such confusion can occur or perhaps use total water stress (or some other general term) to make the distinctions clearer.
Regarding the definition of BWS:
- Is there a particular reason why water use (as I understand, that means withdrawals) as opposed to water consumption (actual water volume removed from a catchment) is used? To my knowledge, most recent BWS studies use the latter since a large part of the withdrawals stays within the same catchment, and thus, should not contribute to BWS in a long term.
- It is not clear where water use estimates for domestic and industrial use are taken from, and whether they account for monthly variability. Please add more details.
- It is not clear whether the authors consider groundwater withdrawals as well as livestock, electricity generation, and mining water supplies. Please elaborate.
Regarding the model setup:
- I find the description of LPJmL inputs in the main text insufficient. This study relies entirely on results from this model, so, as a reader, I would expect to have a very detailed description of input data and main assumptions in Section 2.2, and validation, limitations, and uncertainties in Section 4. Please try to elaborate. I provide several suggestions below.
- In addressing 3a, I would recommend moving Table S1 into the main text, as a reader needs to see the main input data sources and their description. Also, consider adding a column with spatial resolution next to “Time period”.
- When describing the land use and CFTs, please explain:
- where the LandInG toolbox gets rainfed/irrigated crop maps from, and whether those maps cover spatiotemporal changes
- where crop calendars are from, and whether they are static or dynamic
- what share of primary crops from the crop list provided by FAOSTAT is covered (otherwise, it is not clear what your scope is), see L85-87. Also mention if perennial crops are covered.
- When describing the soil, please specify whether the root zone is dynamic or kept static over the growing season, mention how many soil layers are simulated, and whether the shallow groundwater is considered (since capillary rise can support some rainfed crops).
- When describing the irrigation, clearly state what irrigation systems are considered (furrow, sprinkler, drip), how and when irrigation is applied, and whether conveyance losses are simulated.
Regarding validation, limitations, and uncertainties:
- As mentioned earlier, I would expect to see a sub-section on validation, limitations, and uncertainties in Section 4. The presently provided comparisons mainly look at GWS and BWS in terms of hectares. However, the underlying gridded estimates of green and blue ET, as well as sectoral blue water demand, are not properly validated. Please add such comparisons (global total and/or gridded levels), maybe a brief summary for the main text and an elaborated version in SI. Otherwise, it is difficult to judge whether the LPJmL model outputs are reliable before even diving into GWS/BWS analysis.
- For green and blue ET (i.e. water consumption) estimates, you can have a look at https://doi.org/10.1038/s41597-020-00612-0 and https://doi.org/10.1038/s41597-024-03051-3, but there also might be more studies published recently.
- Sectoral blue water use can be obtained from global hydrological models provided by ISIMIP as well as AQUASTAT
- Lastly, I would recommend having an additional sub-section describing 1) main limitations and uncertainties (coming from input data, LPJmL-specific ones, etc.) and 2) discussing how those limitations and uncertainties affect the reliability of the results. For example, how BWS could change if the authors used a water consumption-based approach instead of water withdrawals or how different EFR methods could affect the results. No need to run additional simulations to provide an uncertainty range, but it is always a good practice to be open about the weaknesses of the selected methodology, so the follow-up studies can address those.
Technical corrections:
- L13: Authors define GWS as “soil moisture limitation on crop growth” in the abstract without mentioning that it only concerns rainwater.
- L30: Please consider using more recent publications to support your statement.
- L32: Authors mention green water scarcity without defining what it is.
- L36: Rainfed croplands can also get blue water via capillary rise, see https://doi.org/10.1088/1748-9326/ad78e9. Also, correct the citation for (X. Liu et al., 2022) to fit journal guidelines (check through the manuscript for other instances).
- L47: Maybe use EFR instead of “environmental water requirements” since you already introduced it a few lines earlier.
- L62: Define what the LPJmL abbreviation is.
- L98-99: It is not clear why running the 1901-2019 period with a 3500-year spin-up is needed. The analysis is mainly for 2015-2019. So why starting in 1901, and why so many spin-up years? Please elaborate.
- L112: Please explain what 0.01 rate is (units? physical meaning?) and how it was selected.
- L120: Based on what 8mm/d as max is defined?
- L126-127: Please provide references to the coefficient and scaling factor (or explain how you calculated them)
- Please add units to Eq. 2 and 3.
- How do you calculate annual representative GWS values based on monthly numbers? I find it a bit unclear. Perhaps add an equation to better explain L137-139.
- L148: WUdom, WUind, and WUirr are not defined.
- L187-188: consider adding a sentence with additional global GWS estimates for each year or simply provide a range (min, max) during 2015-2019.
- In Table 1, the total area for ILIM under GWS>0.2 is larger for rainfed crops (426+213 Mha) than for all crops combined (431+161 Mha). Please double-check.
- L197-198: If 37% of all croplands are under GWS>0.2, then logically the remaining lands (63%) are under GWS<0.2. However, the authors say it is only 13%. Please double-check.
- L202-204: From the text, it is not immediately clear what the difference is with L195-197. After a few minutes, I finally realised that it is only about irrigated grid cells here (rainfed excluded). Please make it clear for the reader.
- L204: If indeed only irrigated cells are analysed, please explain why there are still large GWS areas remaining (once you switch on irrigation). Is it because some irrigated cells cannot have access to a sufficient blue water supply, so the crops remain under water deficit? If so, how realistic is this assumption? I would expect that such underirrigation is not that common globally, since farmers can always dig into groundwater reserves or simply ignore EFR (perhaps mention this in the sub-section on limitations and uncertainties).
- L296: Consider providing a few examples of green water management options.
As you can see, I took a great responsibility of being the “annoying reviewer #2”. Hope my constructive criticism will be helpful in improving the quality of this paper, which I am truly looking forward to reviewing in the next round of revisions. Good luck!
Citation: https://doi.org/10.5194/egusphere-2025-3817-RC2 -
AC2: 'Reply on RC2', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
-
RC3: 'Comment on egusphere-2025-3817', Anonymous Referee #3, 01 Oct 2025
The study leverages two indicators: green water scarcity (GWS ) and blue water scarcity (BWS), to quantify and examine the interconnectedness of green and blue water resources for irrigation. They show that while irrigation alleviates GWS, it simultaneously exacerbates BWS at the expense of environmental flow supply. The paper successfully addresses the research gap with clear narration and consistent structure throughout. Given that the results hinge on several salient parameters (Emax, threshold values), I recommend that authors consider conducting a sensitivity analysis to strengthen their analysis and to enable both themselves and the readers to better gauge the robustness of the results.
Major comments:
- There is still room to further improve the transparency of the modelling process. This could include making assumptions more explicit and providing justification or citations for modelling decisions so that readers can trace and assess their epistemic quality/reasoning. Such steps would also support future scholars in replicating the work based solely on the provided documentation. Please see the following details:
- L. 90: Could you expand on the parameterisation of irrigation efficiency? Since CFTs may be suitable to more than one irrigation method (see Jagermeyr et al 2015), how did you determine which method to apply? Was the efficiency value based on an area- or grid-cell-weighted average?
- L. 120: Could you provide a rationale for using Emax = 8 mm/day? Earlier studies suggest lower values (e.g., Gerten et al 2004 report 5-7 mm/day, while Rost et al, 2008 used 5 mm/day). If more recent studies support your chosen value, it would be helpful to cite them.
- L. 129: The statement “If S > D, we set S = D to ensure the result remains within [0,1]” could benefit from clearer justification. For example: “When S>D, we set S = D, because plants cannot take up more water than their transpiration demand allows.” This framing preserves the rationale while adding a physiological explanation.
- L. 143: Could you elaborate on the rationale for selecting 0.4 as the threshold to delineate highly water-scarce conditions? Since this assumption conditions the characterisation of scarcity across regions, you may consider conducting a sensitivity analysis to assess the soundness of your results to alternative threshold values.
- L. 256: It may be helpful to mention the non-representation of multiple cropping already in the methodology, rather than only in the discussion, as this limitation (as you argued) could underestimate your results.
- Given that the results are highly contingent on several parameterisations, it may be beneficial for the authors to conduct a sensitivity analysis to strengthen their conclusions and to allow stakeholders to better gauge the robustness of the findings. For example, sensitivity runs on influential elements such as Emax values (5-8 mm/day) and different thresholds (0.2, 0.3, 0.4, 0.6) may be informative and also support your choice of values.
Minor comments:
- For consistency with previous LPJmL studies, please consider standardising the equation symbols:
- L. 124: am should be written as αm
- L. 124: adjust the fraction form of gm/gc for clarity
- The authors may consider aligning the numbering of supplementary materials with the order in which they are first referenced in the manuscript. Additionally, there are several information and figures included in the supplementary material that are not explicitly mentioned in the main text. Without such references, readers may overlook these potentially important resources.
Technical comments:
- When citing, if applicable, it may be helpful to specify that exact page, figure or table for the reader’s reference. For example, L. 90: (Table 5, Jagermeyr et al. 2015)
- L. 156: Please consider removing the comma following “overuse”
- L. 156: It may improve readability to elaborate on the meaning of transgression: clarifying what it implies when a threshold has been transgressed.
- L. 264-266: The sentence “While irrigation reduces GWS (below the threshold of 0.2) on 13% of the global agricultural area, it thereby leads to an increase in areas experiencing moderate BWS as well as high BWS by 6%, respectively.” may be confusing, as “respectively” suggests a missing distinction between moderate and high BWS. Consider rephrasing for clarity, e.g., “…by 6% and 6%, respectively” or “…both by 6%”
Citation: https://doi.org/10.5194/egusphere-2025-3817-RC3 -
AC3: 'Reply on RC3', Heindriken Dahlmann, 01 Dec 2025
We would like to thank all three reviewers for taking the time to read our manuscript and for the comments and suggestions.
In response to the reviews we propose three main changes in our analysis that address concerns raised by several reviewers:
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
Based on these yield declines, we will reframe the GWS categories to sequential steps of 20% yield decline (0-20%, 20-40%, …), in order to show a graduality in the GWS exposure. - We will recalculate irrigation water consumption since the former analysis was not based on consumption but on total applied irrigation (i.e. water use).
- We propose to change the input dataset for sectoral water consumption: Instead of taking the input data from Flörke et al. (2013), we will use the more up-to-date dataset of Huang et al. (2018) that includes more sectors than Flörke et al. (2013).
Please find the more detailed point-by-point responses in the attached supplementary document.
References
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F. and Alcamo, J.: Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study, Global Env. Change, 23, 1, 144-156, https://doi.org/10.1016/j.gloenvcha.2012.10.018, 2013.
Huang, Z., Hejazi, M., Li, X., Tangl, Q., Vernon, C., Leng, G., Liu, Y., Döll, P., Eisner, S., Gerten, D., Hanasaki, N. and Wada, Y.: Reconstruction of Global Gridded Monthly Sectoral Water Withdrawals for 1971–2010 and Analysis of Their Spatiotemporal Patterns. HESS, 22, 4, 2117–33, https://doi.org/10.5194/hess-22-2117-2018, 2018.
- Instead of using the fixed green water scarcity thresholds 0.2 (green water stressed) and 0.4 (highly green water stressed), we will recalculate the thresholds based on yield responses modeled with LPJmL. We conduct two new LPJmL simulations, both implemented with unrestricted nitrogen availability and nitrogen-insensitive maintenance respiration. In the first scenario (INOthreshold), all cropland is treated as rainfed. In the second scenario (IALLthreshold), all agricultural areas are assumed to be fully irrigated and therefore do not experience GWS. Under conditions of full water and nitrogen availability, the IALLthreshold simulation represents an optimal scenario in which maximum attainable yields can be achieved. By comparing the IALLthreshold yields with those of the INOthreshold run, we calculate the percentage yield decline from IALLthreshold to INOthreshold attributable to GWS. For each CFT, we generate scatterplots illustrating how increasing GWS values correspond to increasing yield declines. This relationship is modeled with a logistic regression for each CFT. From these fitted models, we derive CFT-specific green water stress levels at which yields decline by 10%, 20%, 30%, etc. (in gC m⁻²) due to GWS.
- There is still room to further improve the transparency of the modelling process. This could include making assumptions more explicit and providing justification or citations for modelling decisions so that readers can trace and assess their epistemic quality/reasoning. Such steps would also support future scholars in replicating the work based solely on the provided documentation. Please see the following details:
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 15,984 | 1,923 | 213 | 18,120 | 467 | 222 | 202 |
- HTML: 15,984
- PDF: 1,923
- XML: 213
- Total: 18,120
- Supplement: 467
- BibTeX: 222
- EndNote: 202
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Heindriken Dahlmann
Lauren S. Andersen
Sibyll Schaphoff
Fabian Stenzel
Johanna Braun
Christoph Müller
Dieter Gerten
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1398 KB) - Metadata XML
-
Supplement
(2234 KB) - BibTeX
- EndNote
- Final revised paper
Revision:
“Shifting water scarcities: Irrigation alleviates agricultural green water deficits while increasing blue water scarcity”
The authors investigate how irrigation reshapes agricultural water stress worldwide by shifting pressure from green water scarcity (GWS) to blue water scarcity (BWS). They simulate crop water use with the global vegetation–crop model LPJmL at 0.5° daily resolution, contrasting a no-irrigation counterfactual (INO) with a limited-irrigation case (ILIM) where withdrawals are constrained to locally available blue water (including upstream inflow and reservoirs, but excluding long-distance transfers and fossil groundwater). GWS and BWS are then quantified with monthly indices.
The paper’s key innovation is to treat GWS and BWS jointly, which yields clear global insights: irrigation alleviates GWS on 13% of cropland are but increases BWS by ~12%, concentrating the latter in some irrigation hotspot (e.g., in India and the Mediterranean basin). The estimates of blue water overuse are broadly consistent with published studies; however, the manuscript should explicitly state what is meant by “water use” (i.e., whether it refers to water consumption or to withdrawals). Moreover, it remains unclear whether renewable groundwater is explicitly included in the accounting of available blue water. Clarifying these points would strengthen the interpretation of overshoot volumes.
The manuscript is well written, the narrative is coherent, and figures, tables, and supplementary materials effectively support the analysis. Overall, this is a solid and valuable contribution; I recommend minor revisions focused on clarifying the treatment of renewable groundwater in the blue-water budget and related sensitivity.
Specific comments:
** lines 39-40**: This sentence would benefit from bibliographic references; for example, Gleeson et al. (2020).
Gleeson, T., Wang‐Erlandsson, L., Porkka, M., Zipper, S. C., Jaramillo, F., Gerten, D., ... & Famiglietti, J. S. (2020). Illuminating water cycle modifications and Earth system resilience in the Anthropocene. Water Resources Research, 56(4), e2019WR024957.
**lines 115-116**: Adding (S) and (D) to water supply and atmospheric demand, respectively, would make Eq. (1) clearer and more accessible to readers, since these symbols are not defined immediately afterward.
**Eq.3**: In the ratio between the scaling factor and the potential canopy conductance, the division sign (—) is missing; please check.
**lines 145-147**: Here too, it would be better to introduce the symbols in Eq.4 beforehand. The symbols for the water uses, even if intuitive, should be restated in the text for the sake of completeness.
Moreover, could you please specify what is meant by “water use” in this study? Are you working with consumption or with withdrawals? For water-balance estimates of overuse, the relevant quantity is typically identified by the water consumption, as it represents the fraction removed from the system and not locally available for reuse (hence generally smaller than withdrawals). It would help to state this explicitly in the Methods.
**lines 145-162**: In these paragraphs, it is unclear why additional sectors such as livestock, electricity generation, and mining are not considered. In several regions these are as water-intensive as domestic and industrial uses. Please justify their exclusion (e.g., data gaps, scope) or discuss as a limitation.
It is not specified how the Flörke et al. (1950–2010) database covers the full simulation period, particularly after 2010. Were values held constant at 2010 levels thereafter, as in Rosa & Sangiorgio (2025) and Citrini et al. (2025)? This should be stated explicitly in the Methods and acknowledged as a limitation.
Please clarify why renewable groundwater inflows to grid cells are not included in the blue-water budget (they appear to be excluded in Eq. 4 and Eq. 5). If intentionally omitted, explain the rationale and, in the Discussion section, the implications for overuse estimates.
It would help to add a brief note on the routing module: how upstream–downstream relationships among cells are represented when computing overuse, including whether deficits are calculated locally or propagated, and any reservoir/return-flow assumptions.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Flörke, M., Kynast, E., Bärlund, I., Eisner, S., Wimmer, F., & Alcamo, J. (2013). Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study. Global Environmental Change, 23(1), 144-156.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**Lines 155-156**: Stenzel et al.’s approach to estimating blue-water overuse appears closely aligned with earlier assessments of unsustainable water use (e.g., Mekonnen & Hoekstra, 2016; Mekonnen & Hoekstra, 2020; Citrini et al., 2025; Rosa & Sangiorgio, 2025 and many others). It would strengthen the manuscript to situate the method explicitly within this literature (briefly clarifying similarities and differences) and, where feasible, to prioritize citations to peer-reviewed, published studies.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Mekonnen, M. M., & Hoekstra, A. Y. (2016). Four billion people facing severe water scarcity. Science advances, 2(2), e1500323.
Mekonnen, M. M., & Hoekstra, A. Y. (2020). Blue water footprint linked to national consumption and international trade is unsustainable. Nature Food, 1(12), 792-800.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**Figure2**: The figure is excellent. One suggestion would be to revisit the colorbar and its tick labels. Because the text consistently refers to thresholds at 0.2–0.4–0.6–0.8–1.0, harmonizing the colorbar bins/ticks with those intervals (similar to the style used in Fig. S11) would improve readability and make cross-references more immediate. If a full reclassification is not desired, introducing color breakpoints at those thresholds would still create a clear visual link to the classes cited in the text. Please take this as an optional refinement to consider.
It would also help to state explicitly in the caption that the colorbar applies to all panels in the figure (as you did in the caption of Figure 4).
**lines 181-184**: Another enhancement to consider is splitting Figure 2 into five panels (2a–e) instead of two, so the manuscript can reference each component explicitly. For example: “The green water stress patterns show a high seasonal variability over the year due to changing weather conditions but also season-specific growing seasons (Fig. 2b-e). Europe and North America do experience less (or even no) GWS during the winter months since the water demand of the crops grown during that time is very low (Fig. 2b). During the summer months (Fig. 2d), however, especially southern regions in Europe and the western US are highly green water stressed (GWS >0.4). Large regions in Brazil change into GWS hotspots from June to November where pulses, rapeseed and sugarcane are especially green water stressed (Fig. 2d-e). India, by contrast, does not experience high GWS from June to November (Fig. 2d-e), when most crops are grown.”
**Figure 5**: The figure appears very similar to the patterns reported by Citrini et al. (2020) for the 2001–2010 baseline (see their Supplementary Fig. 2a). It would strengthen the paper to briefly position your map against those results - highlighting key consistencies or divergences and the likely reasons in the Discussion section. This seems especially pertinent if your overuse metric is restricted to surface-water resources, whereas Citrini et al. used a source-agnostic approach (i.e., not distinguishing whether scarcity arises from groundwater or surface water).
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
**Discussion section**: The Discussion would benefit from engaging with the most recent literature published this year, e.g., Rosa & He (2025) for GWS, and Rosa & Sangiorgio (2025) together with Citrini et al. (2025) for blue-water overuse. Situating your findings alongside these studies (noting agreements, differences, and methodological nuances) would further strengthen the contribution. Could you clarify what drives the difference between the estimates (585 km3yr-1 vs ~460 km3yr-1)? Is the difference mainly due to (i) restricting blue-water availability to surface sources, (ii) using withdrawals rather than consumptive use, or (iii) other methodological choices (e.g., treatment of return flows, reservoirs, EFRs, routing, baseline years)?
In addition, the sentence at lines 302–304 should be supported with appropriate references.
Citrini, A., Sangiorgio, M., & Rosa, L. (2025). Global multi-model trends of unsustainable irrigation under climate change scenarios. Environmental Research Letters, 20(10), 104011.
Rosa, L., & He, L. (2025). Global multi-model projections of green water scarcity risks in rainfed agriculture under 1.5° C and 3° C warming. Agricultural Water Management, 314, 109519.
Rosa, L., & Sangiorgio, M. (2025). Global water gaps under future warming levels. Nature Communications, 16(1), 1192.
**lines 314-315**: It would be helpful to briefly outline how such these new case studies would be implemented. For example, would higher spatial and/or temporal resolution be required? If so, please indicate the additional data needs (e.g., finer-resolution water use, irrigation, hydrologic, and management datasets) and whether the availability of such data currently constrains feasibility.
**Reference list**: Just a minor formatting note, likely governed by the journal’s template rather than the authors: applying a hanging indent to the paragraph/list would make the items much easier to scan and review.
Thank you for the opportunity to revise the study. Again, this is great work (congratulations!) and minor clarificatory revisions are requested.
Lorenzo Rosa