the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Nov 2025
Climate Impacts on Water Resources in a High Mountain Catchment: Application of the Open-Source Modeling Workflow MATILDA in the Northern Tian Shan
Abstract. Applied glacio-hydrological modeling is crucial for the integrated water management strategies needed to effectively mitigate climate change impacts on freshwater resources fed by high mountain areas. We demonstrate the application of MATILDA-Online, an open-source toolkit for modeling glacier evolution and water resources in glacierized catchments. We showcase it's capabilities in data-scarce environments on a catchment in the Tian Shan Mountains in Kyrgyzstan, and outline a four-step multi-objective calibration strategy that integrates glacier surface mass balance, snow water equivalent, and discharge observations. Projections indicate severe glacier mass loss by 2100, significant reductions in runoff, and a shift toward earlier peak flow driven by snowmelt. The main sources of uncertainty in the catchment water balance are biases in precipitation data and inconsistencies in glacier mass balance datasets, highlighting the importance of adequate monitoring. Despite limitations in the model's representation of spatial variability and dynamic processes, MATILDA provides easy access to sophisticated modeling and can be a valuable tool for bridging the gap between advanced glacio-hydrological science and practical water resource management.
- Preprint
(32120 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-3462', Yongmei Gong, 16 Dec 2025
-
RC2: 'Comment on egusphere-2025-3462', Anonymous Referee #2, 02 Jan 2026
Review of
“Climate Impacts on Water Resources in a High Mountain Catchment: Application of the Open-Source Modeling Workflow MATILDA in the Northern Tian Shan”
Summary
This manuscript presents an application of MATILDA, an open-source glacio-hydrological modeling framework, to assess climate change impacts on water resources in the Kyzylsuu catchment in Kyrgyzstan. The authors apply a four-step, multi-objective calibration strategy that combines glacier surface mass balance, snow water equivalent, and discharge observations. The model is forced with ERA5-Land reanalysis data and an ensemble of bias-adjusted climate projections under SSP2 and SSP5 scenarios.
The paper clearly demonstrates the potential of MATILDA as an accessible and reproducible modeling framework for data-scarce mountain regions, supported by extensive documentation and a transparent workflow. The open-source, cloud-based nature of the framework is a clear strength and will be valuable for both researchers and practitioners.
However, because this study is part of a dual-publication framework and the model description is presented in a companion paper (not reviewed here), the specific contribution of this manuscript is not always clear when read on its own. The paper would also benefit from a more critical examination of the calibration and validation strategy, a clearer justification of the model structure and assumptions, and a deeper discussion of the suitability of a lumped conceptual model for century-scale projections. In particular, the very large uncertainty ranges identified warrant a more explicit discussion of their implications for interpretation and for potential use in water resource management. For these reasons, I recommend major revisions before the manuscript can be considered for publication.
Major Comments
Overall contribution. The paper is clearly written, but the specific contribution of this “Part 2” manuscript is not fully clear when read independently. While Part 1 introduces the MATILDA model itself, this paper appears to focus mainly on (i) an example application and (ii) the proposed calibration and uncertainty analysis. This distinction should be stated more explicitly in both the abstract and the introduction. At present, the reader has to infer what is new in this contribution.
Methods clarity. The calibration and validation strategy is difficult to follow in its current form. The workflow involves many steps, datasets, thresholds, and performance metrics, but several of these are not clearly defined when first introduced, or their rationale is explained only later. As a result, the reader has to reconstruct the logic of the calibration sequence and understand how individual decisions affect the final parameter sets and projections. Given that the calibration strategy is presented as a key contribution, the manuscript would benefit from a clearer and more structured presentation of the model, variables, and calibration/validation procedure.
Dual publication and model description. Even though this study is part of a dual-publication framework, the model should still be briefly described before introducing the data and experimental setup. In particular, the main assumptions, structure, and intended strengths of MATILDA should be summarized and justified on their own merits, rather than relying on the companion paper for context. This would make the manuscript more self-contained and accessible to readers who encounter it independently.
Suitability for long-term projections. The manuscript acknowledges that MATILDA is a lumped conceptual model with static parameters, limited spatial representation, and simplified glacier dynamics. However, the implications of these limitations for 80–100 year projections are not discussed in sufficient depth. Given the strong non-stationarity expected in glacier cover, snow regimes, and land-surface processes, this issue deserves a more critical examination.
Relevance for practitioners. One of the main strengths of this work is the accessibility of the MATILDA framework, which has clear potential to support practitioners who lack the resources to assemble extensive datasets or implement complex glacio-hydrological models. The manuscript would be more convincing if the contribution were framed more explicitly as a transparent assessment of what can—and what cannot—be robustly inferred with current data and modeling approaches. This would better align the conclusions with the results and provide more realistic guidance for practical applications.
Appendix figures. There are ten appendix figures related to climate projections. I suggest condensing these into two to three figures. In particular, showing ERA5-Land together with uncorrected projections in the same plots seems unnecessary, as these biases are expected and are addressed during the bias-correction process.
Calibration and validation. Validation is a weaker point of the study, and the calibration does not explicitly evaluate interannual variability. The validation period (2018–2020) is short and affected by data gaps, which the authors acknowledge, but the implications are not fully discussed. In particular:
- Winter performance is poor, yet this is barely addressed, despite its importance for low-flow and drought-related applications.
- The discrepancy between RGI6 glacier area and the random forest estimate around 2002 (about 23%) is large and concerning. It is mentioned but not explored further, even though it directly affects confidence in the glacier evolution results. More details on the random forest model and its training would be helpful.
Specific Comments
Title and Abstract
- Consider quantifying key results directly in the abstract.
- The graphical abstract is excellent. However, showing glacier area as a “local trend” may be misleading, as it is primarily a climate-driven impact. Splitting temperature and precipitation might better preserve the two-panel layout.
Introduction
- The regional context is strong but could be slightly condensed.
- Clearly distinguish what Part 1 covers versus what this paper contributes (see major comments).
- Explicitly state research questions and objectives rather than implying them (e.g., “evaluate the proposed data products in the light of scarce observations” is rather vague).
Study Site
- I suggest removing Table 1, as many elements could be incorporated into the text or are already shown in Figure 1.
- Clarify which decades are meant by “recent decades” when referring to peak runoff.
- Briefly discuss the representativeness of the Kyzylsuu catchment for the wider Tian Shan region and explain why this specific catchment was chosen.
Methods
- See general comment on model description and framework. It is difficult to review the paper while switching between manuscripts, and the same issue is likely to affect readers.
- It is not clear which variables are used as model inputs.
- Consider adding a column to Table A1 listing the corresponding variables used from each dataset and the associated time period.
- “bias adjustement” → “bias adjustment”.
- The justification for selecting Barandun et al. (2021) as the reference SMB dataset should be strengthened.
- “See 2 → See Figure 2”.
- The sentence “The simulated and observed mean annual SMB are compared…” would fit better in the calibration procedure section.
- The calibration procedure is hard to follow without fully reading the companion paper. Several acronyms are not defined.
- The validation section should be merged with the calibration section, clearly stating from the outset that a split-sample approach is used.
- Using only 10% of the data (two years) for validation seems very limited; values closer to 30–50% are more common, especially given the non-stationarity in the catchment.
- “The hydrograph ranges for (a–d) illustrate parameter uncertainty.” Parameter uncertainty of what, exactly?
Results
- Section 4.1 (Reanalysis data) does not include figures or tables to support the comparison with the local weather station. I suggest computing KGE (or its components) for precipitation and temperature to assess whether ERA5-Land captures seasonal variability but fails mainly in bias.
- Section 4.2 (Calibration) could be shortened or removed if space is needed; Table 2 could be moved elsewhere.
- Performance metrics are very good, likely due to strong seasonality. Is the model able to reproduce interannual variability?
- The consistency checks applied to climate projections are not clearly described. It is unclear whether the climatology of the historical period in the bias-adjusted projections is expected to match ERA5-Land, and how deviations are evaluated.
Discussion
- The discussion covers the main result patterns well but often reiterates results rather than critically interpreting their implications.
- The claim that SWE calibration can compensate for precipitation biases appears overstated. SWE constraints help but cannot fully resolve structural or seasonal precipitation biases.
- The discussion of observational uncertainty is strong but could be more forward-looking by identifying which specific measurements would most effectively reduce projection uncertainty.
- While model limitations are listed, their implications for century-scale projections are not fully explored. A clearer statement on when and for what purposes the results should (and should not) be used would strengthen the discussion.
Figures and Tables
- Figure 1: The elevation color bar appears truncated; there is no reference to the yellow shading.
- Figure 2: Consider using a different color for “Simulation,” as it is very similar to Miles et al.
- Figure 3: Consider using a consistent color (e.g., black) for observations.
- Table 2 could be integrated into the text, as it is essentially a one-column list.
- Appendix figures are not always referenced in order.
Citation: https://doi.org/10.5194/egusphere-2025-3462-RC2
Model code and software
MATILDA Phillip Schuster, Ana-Lena Tappe, Alexander Georgi https://doi.org/10.5281/zenodo.14267360
MATILDA-Online Phillip Schuster, Alexander Georgi, Mia Janzen https://doi.org/10.5281/zenodo.15712744
Interactive computing environment
MATILDA-Online Website Phillip Schuster, Alexander Georgi, Mia Janzen https://matilda-online.github.io/jbook
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 227 | 209 | 21 | 457 | 69 | 61 |
- HTML: 227
- PDF: 209
- XML: 21
- Total: 457
- BibTeX: 69
- EndNote: 61
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
The manuscript presents a test case of the open-source toolkit MATILDA-online, which simulates glacier evolution and water resources in the Kyzylsuu Valley in the Tian Shan Mountains. The main focus of the manuscript is a four-step calibration procedure and the validation of the model results. The authors hope to provide an open-access toolkit to support data-driven water resource management and learning for local practitioners and other non-academic stakeholders in in-situ data-scarce regions.
I praise the authors' efforts in providing open-access resources to improve the capacity of local stakeholders in the Global South. However, the manuscript, in its current stage, requires a substantial revision. The biggest issue amongst the others is that Part I of the double publication is still under review in another journal. The authors did not introduce the model components in this manuscript at all, sometimes not even expanding abbreviations. Thus, I can’t fully assess the quality of the model chain and its ability to provide accurate information to the stakeholders, as mentioned by the authors as the aim of the model in the Introduction.
A few more major concerns are as follows:
Thus, I don’t think this manuscript, in its current form, is suitable for publication even after a major revision because we don't know if its 'sister' paper will be accepted at all. I see the value in the study, but perhaps the author could put more effort into the storytelling and the accessibility of the language to non-academic experts, if you really want MATILDA to fulfill its purpose. I suggest that the authors revise the manuscript carefully and resubmit it after the first manuscript is accepted.
Below are some of my more specific comments:
Section 3 Methods
There is no description of the toolkit and its model components. Some abbreviations are not even expanded. I understand that there is a 'sister' manuscript that provides additional details. However, at least some key elements should be included here so that readers don’t need to read another article to grasp what MATILDA does. In addition, the sister manuscript is still under review in another journal. It means that as a reviewer, I need to review two manuscripts to ensure the accuracy of my judgments.
Section 4 Results
The structure of this section is very confusing. ERA5-Land is also presented here, even though the authors did not produce the forcing data. Sec. 4.2 basically only summarizes Table 2. What is the message here? A lot of abbreviations are used, which made it very hard to follow what the authors wanted to convey, especially for non-academic readers, whom they addressed in their introduction. Could we consider starting the sections or paragraphs with a sentence that describes their content or key messages?
Section 5 Discussions
Shouldn’t Sect 5.2.3 be in Sect. 4?
And it will also be helpful to have some discussion on how MALTILDA is a better tool for the local stakeholders than the other glacio-hydrological tools, as it seems to be the “innovation” presented in the Introduction. From what I can see on the MATILDA website, it appears to be a model chain that combines a PDD model, a glacier volume/Area rescaling routine, and the openly assessable HBV model, combined with a significant calibration component.
Perhaps it is discussed in the first paper. However, without reviewing the first paper, I would not know if the model chain is really novel. In particular, from what I have read in the current discussion, MATILDA faces the same challenges as other glacial-hydrological models, i.e., forcing data and the modeling workflow.
Section 6 Conclusions
Overall, I am not convinced that ‘MATILDA offers a solution by supporting, educating, and empowering water management stakeholders in regions affected by climate change.’ It is not supported by what was written in the manuscript.
Figures
From Figure 2, we can see that the interannual variation of the simulated SMB does not match that of the reference study and the observations. In the conclusion, the author wrote, “The temperature index model with the modified Δh routine can provide reasonable long-term estimates of glacial contributions to runoff, including stabilizing effects at higher elevations (Schuster et al., 2025c). However, this setup fails to reproduce observed inter-annual changes and neglects important glaciological factors such as glacier dynamics and debris cover”. Does this really show the robustness of the model chain? How important are glacier dynamics and debris cover compared to the ability of the model to reproduce the observations in the time period you are looking at, i.e., 20 years?