the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Apr 2026
A dynamic socio-hydrology model for the assessment of time-variant feedbacks between irrigators’ adaptive responses and basin hydrology
Abstract. This study introduces a Dynamic Feedback (DF) socio-hydrology model that couples the widely used SWAT hydrological model with a microeconomic model of irrigators’ behavior (Positive Multi-Attribute Utility Programming Model). Unlike conventional static or exogenous scenario-based socio-hydrology couplings, the DF setup allows irrigators’ adaptive responses and basin hydrology to interact dynamically through time-variant two-way endogenous exchanges. The model is illustrated with an application to the Tormes catchment in Spain, where we assess the impacts of a Drought Management Plan (DMP) that introduces water caps to ensure environmental flows under alternative climate change scenarios (SSP126 and SSP585), using both a DF and no-feedback setup. Aggregate results indicate relatively small differences between the DF and no-feedback setup in annual hydrological indicators across the Tormes catchment (<0.5 %). Critically, differences become significant at sub-basin and seasonal scales, where adaptive irrigators’ responses to DMP caps during dry years in the DF setup increase summer inflows by up to 9.3 % under SSP585 as compared to the no-feedback setup, signaling higher effectiveness of DMP interventions.
- Preprint
(1763 KB) - Metadata XML
-
Supplement
(210 KB) - BibTeX
- EndNote
Status: open (until 18 Jul 2026)
- CC1: 'Comment on egusphere-2026-1502', Miao Yu, 20 Apr 2026 reply
-
RC1: 'Comment on egusphere-2026-1502', Anonymous Referee #1, 24 Jun 2026
reply
Overall Assessment
This manuscript presents a valuable contribution to socio-hydrological modeling by developing a dynamic feedback (DF) coupling protocol between the SWAT hydrological model and a microeconomic Positive Multi-Attribute Utility Programming (PMAUP) model. The authors demonstrate their approach using the Tormes catchment in Spain under climate change scenarios (SSP126 and SSP585). The work addresses a critical gap in modular socio-hydrological modeling—namely, the temporal mismatch that typically prevents true two-way feedback between human and water systems.
The manuscript is well-structured, technically sound, and the methodological innovation is significant. However, several issues require attention before the manuscript can be accepted for publication in HESS.
Decision: Major Revisions Required
General Comments:Strengths-
Methodological innovation: The DF protocol represents a genuine advance over conventional static or scenario-based couplings. The modification of SWAT Fortran code to enable interannual feedback is technically impressive and provides a template for others to follow.
-
Open science practices: The authors have made the SWAT-DF code openly available on GitHub and the PMAUP database on Zenodo, which is commendable and aligns with HESS's open science principles.
-
Scale-dependent findings: The observation that catchment-scale differences (<0.5%) mask significant sub-basin and seasonal responses (up to 9.3% increase in summer inflows) is an important insight for water resources management.
-
Relevant case study: The Tormes catchment is an appropriate testbed given its semi-arid conditions, agricultural water demands, and existing drought management framework.
-
Limited exploration of feedback mechanisms: While the DF protocol enables two-way feedback, the actual range of adaptive responses appears limited to four discrete WR scenarios (0%, 25%, 50%, 100%). This restricts the emergent behavior the model can capture.
-
Uncertainty handling: The manuscript lacks a comprehensive uncertainty analysis. While the authors use an ensemble of GCMs, the propagation of uncertainty through the coupled model is not thoroughly addressed.
-
Socioeconomic scenario limitations: The use of 2015 prices and costs throughout the 2020-2050 simulation period is a significant simplification that the authors acknowledge but do not fully address.
-
Missing sensitivity analysis: There is no systematic sensitivity analysis of the coupled model behavior to key parameters in the PMAUP model.
Major Issues1. Dynamic feedback implementation (Section 3.1)The DF protocol is described conceptually, but several technical details need clarification:
Lines 120-125: The authors state that "PMAUP provides updated crop portfolio changes, which are translated into HRU-level fractional areas (hru_fr) for each water restriction scenario (WR)." How exactly is the spatial translation performed when an AWDU spans multiple HRUs? This connects to the limitation mentioned in lines 399-405 but should be explained more clearly in the methods.
Lines 130-135: Figure 4 shows the feedback loop but does not specify the exact timing of information exchange within each annual timestep. Does the PMAUP model receive SWAT outputs at the beginning or end of each water year? This timing could significantly affect results.
Computational cost: The authors mention this as a limitation of modular models but do not provide any metrics on the additional computational cost of the DF implementation compared to noDF. This information would be valuable for potential users.
2. Climate change scenario analysis (Section 4.2)Lines 260-270: The authors use an MMEA approach that aggregates outputs from ten GCMs. However, this masks the inter-model spread, which is particularly important for extreme events. While Figure A1 shows the inter-model range, the main analysis uses only ensemble means. I recommend including at least one additional analysis using individual GCM outputs to demonstrate the robustness of the DF effect.
The period 2020-2050 is relatively short. Given that the authors calibrate SWAT using data through 2014, the overlap period (2020) is essentially an extension. Why was 2050 chosen as the endpoint? A longer simulation period would strengthen the conclusions about climate change impacts.
3. Socioeconomic modeling (Section 3.3 and 4.3.3)Line 370-375: The authors state that "there are only four possible adaptive responses/crop portfolios" corresponding to the four WR levels. This is a significant limitation of the current implementation. In reality, farmers respond continuously to changing water availability, not just to discrete policy thresholds. The authors should either:
a) Justify why discrete responses are appropriate for this system, or
b) Discuss how the model could be extended to allow continuous adaptation.The PMAUP calibration (Table 5) shows that α1=1.0 for many AWDUs, effectively making them single-attribute profit maximizers. This suggests that the multi-attribute utility framework is not being fully utilized. Does this reflect the actual behavior of farmers in the Tormes catchment, or does it indicate that the calibration data do not provide enough information to estimate risk and complexity parameters?
4. Results interpretationLines 335-340: The authors attribute the small catchment-scale differences to the limited geographic extent of irrigated agriculture (2.5% of total area). This is a valid point, but it raises a question: given this small extent, why does the DF effect on summer inflows reach 9.3%? This seems disproportionate and warrants further explanation.
Figure 9 shows percentage changes in summer inflow but does not include absolute values. What is the actual magnitude of these increases (in MCM)? This would help readers assess the practical significance of the DF effect.
5. Discussion of limitationsLines 390-405: The authors identify several limitations (static prices, constant non-agricultural demands, etc.) but the discussion remains somewhat superficial. For each limitation, the authors should:
6. Missing or incomplete elements
a) Provide a more detailed justification for the simplification,
b) Discuss how the limitation might bias results in a particular direction, and
c) Suggest specific approaches for addressing it in future work.Drought Management Plan implementation: The manuscript describes the DMP thresholds (Table 1) but does not explain how the WR percentages are translated into actual water allocations to individual AWDUs. Is the reduction applied uniformly, or is it proportional to existing water rights? This affects the socioeconomic impacts.
Environmental flows: The abstract mentions that the DMP is designed "to ensure environmental flows," but the analysis focuses on reservoir storage and inflows. There is no direct assessment of whether environmental flow requirements are actually met under the different scenarios. This seems like a missed opportunity, especially given the title's emphasis on environmental flows.
The "no-feedback" setup (noDF) is described as "SWAT run dynamically with no yearly interannual interaction with the PMAUP model" (Line 285). However, it is unclear what land use inputs are used in the noDF simulation. Are crop portfolios fixed at baseline levels, or do they change based on pre-run scenarios? This needs clarification.
Minor IssuesTechnical writing-
Line 55: "clmate" → "climate"
-
Line 85: "i.e. socioeconomic agents" → should be "i.e., socioeconomic agents" (comma)
-
Lines 142-149: These lines appear to be a formatting artifact (repeated numbers) and should be removed.
-
Line 395: "with those year's prices" → "with that year's prices"
-
Line 420: "full-ledge" → "full-fledged"
-
Figure 1: The resolution is low; text is difficult to read. Please provide a higher-quality figure.
-
Figure 7: The legend is confusing. The filled rectangles and circles indicate WR activation, but the colors are not described in the caption. What do the different colors represent?
-
Figure 10: The hashed areas for Alert and Emergency are described in the caption but are not visible in the figure. This suggests a rendering issue.
-
Line 495-496: The reference formatting for Fuka et al. (2014) appears incomplete. Please check all references for consistency and completeness.
-
Several references appear to be missing DOIs (e.g., IGN, 2020; CEDEX, 2023). Please add these where available.
Model calibration and validationThe SWAT model calibration (Section 4.1.1) appears sound, with NSE values of 0.70 (calibration) and 0.65 (validation), which are acceptable for monthly streamflow simulation. However:
-
The validation period (2010-2014) is only 5 years, which is relatively short. How representative is this period of the hydroclimatic conditions expected under climate change?
-
The PBIAS values (-18.8% and 19.54%) are at the upper end of "satisfactory" according to Moriasi et al. (2007). This suggests some systematic bias in the model that could affect the WR threshold calculations.
The PMAUP calibration errors (Table 5) are mostly below 10%, which is acceptable. However:
-
AWDU 10 shows an error of 10.98%, which exceeds the threshold mentioned by Gómez-Limón et al. (2016). Why is this AWDU performing worse, and how does this affect the aggregated results?
-
The calibration uses only 2015 data. What is the justification for this single-year calibration given the interannual variability in agricultural conditions?
The availability of SWAT-DF on GitHub and the PMAUP database on Zenodo is excellent. However:
1. The GitHub repository should include:
A README file with installation instructions
A quick-start guide for running the model
Documentation of the code modifications made to SWAT
Example input files and test cases
2. The Zenodo database should include:
Metadata describing each file
Units for all variables
The source code for the PMAUP model, not just the input data
3. Reproducibility:
Given the complexity of the modeling framework, the authors should provide a Docker container or similar environment to ensure complete reproducibility of the results.
Recommendations for RevisionMandatory revisions-
Clarify the feedback mechanism timing (Section 3.1) and spatial translation from AWDUs to HRUs.
-
Justify the use of discrete WR scenarios rather than continuous adaptation, or discuss how the model could be extended.
-
Provide sensitivity analysis of the PMAUP parameters and the effect of model uncertainty on the DF results.
-
Include absolute values for the summer inflow increases (Figure 9) to allow readers to assess practical significance.
-
Expand the discussion of limitations with more explicit consideration of how simplifications affect the conclusions.
-
Fix the formatting issues (lines 142-149, Figure 10 hashed areas).
-
Provide a clearer description of the noDF simulation setup, including the land use inputs used.
-
Ensure all references are complete and consistent.
-
Add an analysis using individual GCM outputs rather than just the ensemble mean to demonstrate robustness.
-
Include a longer simulation period or justify the 2020-2050 timeframe more thoroughly.
-
Provide computational cost metrics for the DF vs. noDF simulations.
-
Assess environmental flow compliance explicitly, given this is mentioned as a key objective of the DMP.
-
Prepare a Docker container or other reproducibility environment.
Citation: https://doi.org/10.5194/egusphere-2026-1502-RC1 -
Data sets
Tormes_Socioeconomic_Datasets Osama Hassan https://doi.org/10.5281/zenodo.18543335
Model code and software
SWAT-DF model Osama Hassan https://github.com/oshs/SWAT-DF
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 214 | 62 | 22 | 298 | 32 | 16 | 15 |
- HTML: 214
- PDF: 62
- XML: 22
- Total: 298
- Supplement: 32
- BibTeX: 16
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This appears to be a valuable contribution to dynamic socio-hydrology. However, the manuscript would benefit from citing recent related work on adaptive crop area and farmer response dynamics in hydrological models (e.g., Yoon et al., 2024; Umer et al., 2025), which are closely aligned with the topic. Including these would help better position the study within the current literature.
References