Review of "Enhancing Physically Based and Distributed Hydrological Model Calibration through Internal State Variable Constraints"

General Comments

The authors have presented a comprehensive study that investigates the impact of different calibration approaches on hydrological models using the WaSiM model. The paper explores three distinct configurations: Baseline (BL), Physical Groundwater Model (GW), and Physical Groundwater with Recharge Calibration (GW-RC) to evaluate their effectiveness in representing various hydrological variables. This research addresses an important topic in hydrological modeling by highlighting the significance of integrating internal state variables into the calibration process.

The paper is well-structured, and the authors have made a significant effort to present detailed analyses across multiple catchments. The inclusion of groundwater recharge as a calibration variable is an important approach that aligns with the growing need for multi-variable calibration frameworks in hydrological modeling. The findings makes effort to underscore the importance of considering both streamflow and internal hydrological processes for robust model performance. 

However, I have a major concern regarding the primary objective of the study, which requires clarification. The current presentation leaves the reader uncertain about whether the study aims to compare calibration strategies or assess the impact of model complexity on hydrological process representation. Addressing this ambiguity will make clear the paper's overall contribution and impact.

Major Comments

1. Unclear Research Focus: The primary research question of the paper is not clearly defined. It remains ambiguous whether the authors aim to compare calibration strategies or demonstrate the added value of increasing model complexity.

• If the goal is to compare calibration strategies, the authors should focus on showing how the constrained recharge parameter improves the realism of the results when compared to both streamflow and PACES data.
• If the goal is to assess model complexity, the paper should clearly outline what unique complexities are introduced in each configuration and how they enhance the model’s capability to represent hydrological processes.

2. Simplifying the Experimental Setup: The current experimental design includes three configurations (BL, GW, GW-RC), but most of the observed differences in results seem to be attributed to the choice of model complexity rather than calibration strategies.

• To demonstrate the impact of the constrained recharge parameter, the authors could simplify their experimental setup by comparing two GW-RC experiments: one calibrated solely to streamflow using the KGE metric and another using a modified objective function that accounts for both the mean and variability of recharge. This would directly illustrate the benefit of including internal state variables in the calibration process.

3. Clarifying the Role of GW-RC: The GW-RC configuration is described as integrating recharge calibration into the model. However, the results suggest that most of the observed improvements are due to the activation of more physically based processes rather than the calibration strategy itself. If the authors wish to emphasize the importance of incorporating internal state variables, they should isolate and highlight the specific impact of the recharge constraint.

Specific Suggestions

1. Objective Statement: In the introduction, clearly state whether the study aims to evaluate calibration strategies or model complexity.

2. Experimental Design: Consider restructuring the experimental setup to compare GW-RC configurations with and without recharge constraints. This would make the study’s focus more precise.

3. Discussion Section: Emphasize the role of model complexity in the observed differences in results. If the paper aims to introduce new complexities in the model, demonstrate their unique contribution to the model's performance.

4. Results Interpretation: Clearly distinguish between the impact of calibration strategies and model complexity in the results section. For example, highlight how much of the improvement in GW-RC is due to the activation of groundwater dynamics versus the recharge calibration constraint.

Conclusion

The paper presents valuable insights into the role of multi-variable calibration in hydrological modeling. However, the main research target needs to be more clearly defined. The authors should clarify whether their focus is on calibration strategy comparison or model complexity assessment. Additionally, simplifying the experimental setup to directly compare the impact of recharge constraints would make the study more impactful.

This manuscript titled “Enhancing physically based and distributed hydrological model calibration through internal state variable constraints” investigates the effectiveness of various calibration approaches within the Water Balance Simulation Model (WaSiM) to enhance the representation of hydrological variables. The study assesses three configurations: Baseline (BL), Physical Groundwater Model (GW), and Physical Groundwater with Recharge Calibration (GW-RC), which has an add-in of recharge calibration across 34 catchments in Southern Quebec, Canada. The research provides valuable insights into the importance of multi-variable calibration frameworks in developing robust models capable of adapting to anticipated hydrological shifts due to climate change. However, it is too long!

Major Comments:

• Abstract:

While the abstract effectively conveys the general research objective and findings, it lacks specific quantitative data. It relies heavily on vague terms and subjective assessments, making it difficult for readers to grasp the magnitude and significance of the improvements achieved fully.

Generally, I suggest you revise it.

• Methodology:

The authors should provide more details on the selection criteria for the 34 catchments used in the study. While some information is given in section 2.1, a more comprehensive explanation of why these specific catchments were chosen would strengthen the methodology. The selected basins are relatively medium-sized (between 100 and 10,000 square kilometers), which may limit the generalizability of the findings to larger or smaller basins.

The rationale behind using ERA5 reanalysis data instead of ground-based observations for meteorological inputs should be further elaborated. Since ERA5 recorded underestimating winter precipitation and bias in convective precipitation, it would be better to employ another dataset.

• Model Configurations:

While the three configurations (BL, GW, GW-RC) are described, readers would benefit from a more detailed explanation of how they differ in their treatment of groundwater processes.

The authors should consider discussing the potential limitations of each configuration and how these might impact the results.

• Calibration and Validation:

The split-sample approach for calibration and validation is appropriate, but the authors should discuss any potential impacts of climate non-stationarity on this approach, given the study’s focus on climate change adaptation.

Given that you modified the lower and upper boundaries of the model parameter by 10% (L310), a direct comparison with the calibration results of the BL configuration using default parameters might not be entirely fair. Simulating WaSim for all configurations using the adjusted parameter range would be beneficial to ensure a more consistent evaluation.

Given that the manuscript focuses on WaSim performance, including the computational cost and time associated with each configuration is crucial. This information will be highly valuable for other researchers, allowing them to estimate the resource investment required to achieve comparable improvements in water balance closure.

• Results Presentation:

Figure 4 highlights the significant shift in the proportion of surface runoff and interflow. Please elaborate on the specific factors that influenced this shift during calibration, particularly considering the inclusion of groundwater recharge in the model.

The results presented in Figures 3-10 are generally clear, but some figures (e.g., Figures 5, 6, 7) could benefit from additional explanation in the text to help readers interpret the complex information presented.

A more in-depth discussion of the spatial variability in model performance across the 34 catchments would enhance the study’s insights, especially when compared with PACES.

My understanding differs from your conclusion in Figure 10. None of the configurations are aligned with PACES, except for a case in Noire. I would say that the lowest difference is between GW-RC and PACES. In general, I found PACES recharge different than the applied three configurations in this research, according to Figure C1.

• Climate Change Implications:

While the study mentions the importance of the findings for climate change adaptation, a more specific discussion on how the improved model configurations might be applied in climate change impact assessments would strengthen the paper’s relevance.

Minor Comments:

• Abstract:

              o   It exhibits some redundancy, such as the repetition of “on the representation of hydrological variables” in lines 9 and 11.

• The abstract could benefit from a more precise statement of the study’s objectives and a more concise summary of the key findings.
• Vague Language:
  • “significantly refines the model’s ability to depict subsurface processes”
  • “minimal emphasis on recharge”
  • “small and targeted calibration adjustments”
  • “marked improvement”
  • “enhancing the precision” 
• Lack of Quantifiable Results:
  • No specific metrics are mentioned (to quantify the improvement in model performance.
  • No specific values are given for the improvement in groundwater recharge representation.
  • No indication of how the “minimal emphasis” on recharge was defined or quantified.

• Methodology:

• Figure1: Visualising the selected case studies within a coarser-level basin delineation would be beneficial. This would provide context, as the presence of a river traversing the study area can significantly influence catchment behaviour.
• Table5: Table’s style is totally different from the other presented tables.
• Could you elaborate on the rationale behind conducting 1000 simulations at 1000 m resolution and only 50 at 250 m resolution? What factors influenced the selection of these specific numbers?
• Given the widespread familiarity of the KGE metric within the research community, a detailed definition in section 2.5.1 may be redundant. Furthermore, as per comment CC1 received on November 23rd, the assigned weights in section 2.5.2  (L349) require more comprehensive scientific justification and supporting literature.
• The manuscript would benefit from considering alternative objective functions besides KGE for streamflow. As suggested in this paper (https://gmd.copernicus.org/articles/11/1873/2018/), using SPAtial EFficiency (SPAEF) could enable the evaluation of multiple hydrological components when you utilise distributed hydrological models. This would provide a more comprehensive assessment of model performance.

• Results

• I would like to know if the same results would be obtained by switching the calibration and validation periods, as indicated in Figure 2. Given that the KGE values for all three setups are relatively close, I am uncertain about the potential benefits of using GW.
• In Figure3, consider adding each variable’s total mean or sum of observations to enhance the visual comparison. This will allow readers to contextualise the calibration and validation boxplots by providing a reference point for the overall data distribution.
• The manuscript should provide an explanation for the lack of differentiation in baseflow between configurations GW and GW-RC, as noted in L430.
• Figure 5 reports a difference of around 200 mm/y across all variables among the three configurations for all 34 catchments. To facilitate water balance closure assessment, consider adding a subplot for precipitation data for each basin.
• The current explanation of Figure 6 was neither informative nor relevant to my perspective. It needs to emphasise the significance of the figure.
• Figure7, is the x-axis long-term mean of Q, or are they for a given year? The problem is between October to December in validation period. In the rest, I see no significant differences. Maybe you could drop this figure. Also, since you have gaps in some of the frozen months (L130), how did you consider them in the likely monthly discharge time series?

• Language and Style:

             o   The manuscript is generally well-written, but there are occasional instances of complex sentence structures that could be simplified for clarity. Thank you for acknowledging the use of ChatGPT-4. The presence of long sentences with numerous commas can be indicative of revised text by an LLM-AI.

• Conclusion:

              o   I remain uncertain about the meaning of lines 659-660.

This study contributes to hydrological modelling by demonstrating the importance of incorporating internal state variables, particularly groundwater recharge, into model calibration. The authors designed their model well and developed it to have three configurations and further calibrations.

The findings highlight the potential for improved representation of hydrological processes, which is crucial for water resource management and climate adaptation strategies. However, addressing the major and minor comments outlined above would further strengthen the manuscript and enhance its impact on the scientific community. Overall, with appropriate revisions, this paper has the potential to be an important addition to the literature on hydrological model calibration and process representation.