the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Dec 2023
Surface-subsurface interaction analysis and the influence of precipitation spatial variability on a lowland mesoscale catchment
Abstract. The hydrology of the catchments is primarily shaped by the intricate and dynamic interactions between surface water and groundwater. This is particularly evident in lowland catchments, where these interactions assume a complex nature. This study investigated the complex interaction between surface water and groundwater in the transboundary catchment Aa of Weerijs, shared by the Netherlands and Belgium. A hydrological model, MIKE SHE coupled with MIKE 11, was calibrated and validated over twelve years using streamflow, groundwater levels, and evapotranspiration data. The model performance was analyzed using model efficiency parameters i.e., correlation coefficient and Nash-Sutcliffe Efficiency coefficient. The model performed well, with satisfactory simulations of streamflow, groundwater levels, and evapotranspiration dynamics. Groundwater levels rose in winter and declined from April to September due to increased evapotranspiration in summer. Precipitation drove the water balance, with 60 % lost through evapotranspiration. Base flow from subsurface drainage networks significantly contributed to river water. Spatial variability in precipitation minimally impacted streamflow but caused localized fluctuations in groundwater levels. Higher spatial resolution precipitation data led to fluctuations due to local recharge points, yet overall catchment hydrology was unaffected. The findings highlight the importance of surface water-groundwater interactions in lowland catchments. The developed model provides insights for water resource planning and climate change adaptation in the catchment.
- Preprint
(2098 KB) - Metadata XML
-
Supplement
(989 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2023-276', Anonymous Referee #1, 15 Jan 2024
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2023-276/hess-2023-276-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Muhammad Haris Ali, 25 Jan 2024
We appreciate the reviewer for dedicating time to review our manuscript and for providing feedback. However, we are disappointed to learn that the decision to reject the manuscript was made after reading only few lines of the abstract.
The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the HESS discussion are presented below.
Reviewer comment 1: lines 14-16: authors said that they used the NSE and R, so what are the values they got? As opposed to listing the values, authors said ‘model performed well, with satisfactory….’, so what do ‘well’ and ‘satisfactory’ mean?
Authors’ response: The intention of these first lines is to introduce the reader to the elaborate evaluation of the model performance, which included streamflow, groundwater levels, and evapotranspiration data (please see lines 12-13). The actual detailed values we got and insights into this evaluation are presented in section 3.2., lines 147-150. We have elaborated on the evaluation specifics, indicating that streamflow, groundwater levels, and actual evapotranspiration data were assessed at 3, 13, and 13 locations, respectively. Their geographical distribution is depicted in Fig. 2, and the corresponding NSE and R values are tabulated in Table S4. Due to space limitations in the abstract, we refrained from explicitly stating the NSE and R values, and we conveyed the model's performance as "well" and "satisfactory" based on these outcomes and overall model performance in capturing the hydrological trend in comparison to observed data. The new manuscript will reformulate the abstract, reducing in some parts and including some of these values.
Reviewer comment 2: The only description of interactions between groundwater and surface water is in line 18, a very vague expression, but authors said, ‘the findings highlight the importance of surface water groundwater interactions ….’ in line 21. Why?
Authors’ response: The major portion of the streamflow is contributed by the groundwater through sub-surface drainage flow. Additional insights are mentioned in sections 4.2.3, 4.2.4, and 4.2.5, where we discussed the average yearly catchment water balance, as well as water balances during dry and wet years, elucidating exchanges between surface and groundwater. The assertion in line 21 of the abstract, "the findings highlight the importance of surface water-groundwater interactions," is reflective of the comprehensive findings discussed throughout the paper, not solely those presented in the abstract. As mentioned previously the abstract will be reformulated to state main findings.
Reviewer comment 3: Overall, all the conclusions in abstract are fundamental knowledge of hydrology in textbooks
Authors’ response: We have mentioned the conclusions that are drawn from our model results only. Their alignment with fundamental hydrological knowledge, support the robustness of our findings. This concurrence underscores the validity and reliability of our model, providing a solid foundation for the study's outcomes. We will add model output values and numbers in the abstract finding to make it more specific to our catchment.
Reviewer comment 4: It is very vague. for example, line 26, ‘a multi-scale approach that combines various techniques’, what do the authors want to express? I got nothing.
Authors’ response: In line 26, the intention is to convey that combining different techniques and concepts related to surface water and groundwater simulation at different scales of analysis (e.g., local, regional, global) can significantly improve our understanding of the exchanges between surface water and groundwater. This approach may allow us to gain a more comprehensive and accurate understanding of how water moves and interacts within hydrosystems, as highlighted by Ntona et al. (2022) in their review paper. The new manuscript will reformulate this for better clarity eliminating the vague formulation.
Reviewer comment 5: Then lines 28–31, some other fundamental knowledge. Also, I don’t think ‘groundwater tables’ is a right terminology, it should be ‘water table’.
Authors’ response: In lines 28-31, we underscored the significance of surface water-groundwater interaction, emphasizing that variations in precipitation can significantly influence these exchanges. We have emphasized on studying groundwater and surface water together instead of analysing them independently. We can rephrase the text in these lines to increase the readability in the updated manuscript.
Regarding the term "Groundwater table," it is our understanding that the use of both "groundwater table" and "water table" is interchangeable in scientific literature. Notably, the term "groundwater table" has been consistently employed in various research articles published in HESS, such as: https://doi.org/10.5194/hess-26-5859-2022, 2022, https://doi.org/10.5194/hess-25-1905-2021, 2021, etc.
Reviewer comment 6: Then lines 32-33, what does ‘unique characteristics’ mean? I got nothing again.
Authors’ response: In lines 32-33, the term "unique characteristics" refers to the specific properties of each catchment, encompassing factors such as size, slope, topography, soil type, land use/land cover, etc. These individual features contribute to the distinctiveness of each catchment. The revised manuscript will rephrase and will contain this explanation.
Reviewer comment 7: why ‘the temporal variations’ in line 33? why not spatial?
Authors’ response: The use of the term "temporal variation" in line 33 is intended to cover the seasonal fluctuations or variations associated with wet and dry conditions.
We have not used the term 'spatial variation' because we want to convey that at any particular location, the interaction between surface water and groundwater is influenced by catchment characteristics and climate variation. We will update this in the revised manuscript.
Reviewer comment 8: lines 34-35, the cited papers are 15-20 years ago. why ‘empirical field-based and numerical modeling’, why not remote sensing and machine learning?
Authors’ response: The mention of empirical field-based techniques, such as the use of tracers, and numerical modelling in lines 34-43 is intended to highlight the conventional methods commonly employed for the analysis of surface-groundwater interaction. In the updated manuscript we will extend with a short discussion on the role of RS and ML. For both conventional and for RS and ML methods, we will include additional recent references.
Reviewer comment 9: Then from line 36 to the end of the introduction, I got nothing new and the only useful information is a modeling work was conducted in the catchment, Aa of Weerijs, based on MIKE SHE and MIKE 11. Another interesting thing is, in line 79, authors said it is a mesoscale catchment while in line 86, it becomes a small scale.
Authors’ response: The lines 36-75 are intended to identify the knowledge gap and give the reader a comprehensive overview of the literature and the rationale behind the study, not necessarily in the particular catchment considered, but overall of similar catchments. Detailing the steps of introducing the knowledge gaps: in lines 36-45, several studies employing tracers for surface water-groundwater interaction are referenced. The limitations of field methods are discussed in lines 43-50, and the advantages of using numerical models to overcome these limitations are explained. The preference for physically based hydrological models over lumped models is highlighted in lines 53-57. The importance of accurate representation of rainfall data in the modelling domain is emphasized in lines 57-74. After making the context, the identified knowledge gap and the study's objective are outlined in lines 75-85. In the updated manuscript we will consistently use the term “mesoscale”.
Reviewer comment 10: lines 102–103, ‘The hydrology of the Aa of Weerijs catchment is influenced by various factors, including topography, land use, soil type, and climate.’ I am wondering which catchment is not affected by various factors as listed by authors?
Authors’ response: In 102-103, the intention was to emphasize on the complexity of catchment’s hydrology, indicating that no single catchment characteristic dominates. This aspect will be rephrased in the updated manuscript to ensure a more concise and clear expression of the message.
Reviewer comment 11: Lines 112–121, always, in this part, the governing equations and coupling approach are introduced. However, I got nothing again in this part. Though I knew MIKE SHE and MIKE 11 well. But if I was a common reader, I do have a question ‘what MIKE SHE and MIKE 11 are?’ I can only get that they are powerful and they can simulate everything, then what?
Authors’ response: Lines 123 onward contain the modelling setup. The reference to MIKE-SHE is provided in Lines 125. We believe that it is outside the scope of this paper to reiterate existing literature that comprehensively explains the modelling structure and governing equations of MIKE-SHE and MIKE-11. We will include in the new manuscript a phrase explaining what MIKE-SHE and Mike 11 are. We will refer readers seeking detailed information about these modelling tools to the available literature and technical manuals. We will include in the new manuscript a phrase explaining what MIKE-SHE and MIKE-11 are, though we hope that the HESS papers are read by experts and not “common readers”.
Reviewer comment 12: For the model itself, nothing is attractive. 500 resolution modeling over a domain of 346 km2
Authors’ response: It is our disappointment to see that for such a complex model nothing is attractive. We should emphasize more in abstract and introduction. We find great value in the development of a physically based, fully distributed model that utilizes both global and local datasets. This is not common procedure yet. The procedural method described for model development can be applied to model any other catchment globally. Our model includes hydrodynamic river modelling in MIKE-11, incorporating weirs and control structures to account for flow regulation—a feature not commonly found in hydrological modelling. Calibration and validation were performed using discharge, groundwater levels, and actual evapotranspiration data at 3, 13, and 13 locations, respectively. We employed a weighted objective function to consider multiple variables at various locations during the calibration and validation of the distributed hydrological model. The 500 m resolution was selected to strike a balance between achieving a comparatively short simulation time and retaining sufficient spatial detail to represent parameters and modelled processes.
Reviewer comment 13: I am really sorry. I suggest authors submit it to another journal. This is good documentation about the study of MIKE model in a real catchment. I didn’t find any advanced scientific question and anything different in terms of model and modeling. I really don’t think HESS is the right place.
Authors’ response: The primary objective of our study was not to introduce novel model structures or modelling approaches. Instead, we focused on addressing key aspects of hydrological modelling for meso-scale catchments, specifically the representation of rainfall data across the model domain. Another significant aim was to enhance our understanding of the complex interactions between surface water and groundwater for lowland meso-scale catchments. These objectives are explicitly outlined in lines 81-85 of the manuscript, with the corresponding findings presented in sections 4.2 and 4.3.
Citation: https://doi.org/10.5194/hess-2023-276-AC1
-
AC1: 'Reply on RC1', Muhammad Haris Ali, 25 Jan 2024
-
RC2: 'Comment on hess-2023-276', Anonymous Referee #2, 07 Mar 2024
-
AC2: 'Reply on RC2', Muhammad Haris Ali, 14 Mar 2024
We appreciate the reviewer for dedicating time to review our manuscript and for providing feedback. The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the finalisation of the HESS discussion, are presented below.
Reviewer comment 1: The abstract fails to adequately address key factors influencing the conceptual model, such as the regulation of streamflow and the agricultural nature of the watershed. These factors are crucial for understanding the context and conditioning of the modeling processes, yet their omission diminishes the clarity and completeness of the manuscript.
Author’s response: Thank you for this valuable comment. Indeed, the paper presents the key factors of catchment characteristics in section 2. Based on the suggestion the new version of the manuscript, will reformulate the abstract such that key factors of the catchment are included.
Reviewer comment 2: Acronyms, including MIKE SHE and MIKE 11, are frequently used throughout the manuscript without proper explanation or clarification of their respective roles and differences. This lack of clarity hinders readers' understanding of the modeling approach and its components.
Author’s response: Thank you for pointing this out. We will add a description of the acronyms MIKE SHE and MIKE 11. Their respective modelling roles are briefly described in section 3.1, lines 123-134. The new version of the manuscript will include explanations to ensure better clarity in this regard.
Reviewer comment 3: The manuscript contains numerous instances of vague language lacking both literature and quantitative support. Additionally, figures and tables are inadequately labeled, with critical information missing, such as a color scale for Figure 3. Confusing equation notations (Equation 1 and 2) and setups (line 201-202) further complicate comprehension and interpretation.
Author’s response: The new version of the manuscript will take into account the suggestion and will try to eliminate any vague sections. Though, indicating areas lacking quantitative support and literature would have helped us better to eliminate this vagueness. We will conduct a thorough read to identify where data and literature are lacking and add relevant information accordingly.
We would like to clarify further the comments about the figures: In Figure 3 we did not include a colour scale, as its purpose was not to show the quantity of rainfall, but to show the extent of the spatial distribution of precipitation over the model grid under three considered scenarios (see section 3.4, lines 241-247). we will make it clearer in the caption of the figure.
The notation used in equations is described in line 194 of the manuscript. We will check again and we will ensure that any missing notation is added in the updated version of the manuscript.
The setup explained in lines 201-202 will be reformulated to ensure better clarity. Overall, we will explain that the same weights are assigned to all considered variables (streamflow, Actual evapotranspiration, and groundwater levels) to calculate the mean NSE. However, each of these variables is measured at multiple space locations. For instance, streamflow is evaluated at three locations. To calculate the mean NSE of streamflow at multiple locations, more weight (0.4) is given to the streamflow at the outlet of the catchment due to its greater importance in water management practices, whereas a weight of 0.3 is considered at the other two locations. The details of the assigned weights are provided in the supplementary material as well (see Table S1). We will make sure the link to this material is clear.
Reviewer comment 4: Much of the modeling process appears to be subjective, with weights assigned based on authors' knowledge (line 203-209) and parameter tuning conducted manually (line 211-212) for purported insights. This subjectivity raises questions about the validity and reliability of the results, particularly considering the authors' acknowledgment that "promising" results (line 252-254) align with their mental model, suggesting potential bias and the generation of artifacts.
Author’s response: We appreciate the reviewer's concern regarding the subjective elements inherent in the modelling process, particularly regarding the assignment of weights and manual parameter tuning. The rationale for this is detailed in Section 3.3.1, lines 195-200, where the assignment of equal weight to multiple variables was based on literature, not on the author’s knowledge. However, when evaluating the same variable at multiple spatial locations, we adopted a more nuanced approach based on the knowledge of the Water Board who manages the catchment. Several sessions and interactions with stakeholders in the area took place to determine these weights through co-design. See explanations for the example of streamflow, in the previous comment. All weights were adopted based on co-designed with stakeholders. In the new version of the manuscript, we will emphasize this aspect, such that it is clear that the selection of weights is not an arbitrary choice.
Furthermore, we would like to emphasize on the other comments:
MIKE-SHE is a fully distributed physically based model that relies heavily on field data and literature for parameterization, and its setup is intricate, with simulations being time-intensive. Further, the sensitive parameters such as horizontal hydraulic conductivity and drainage time constant, are provided as spatially distributed grids over the model domain, further adding to the complexity. Given these circumstances, we opted for a one-at-a-time manual calibration approach. This method allowed us to carefully adjust individual parameters while considering their spatial distribution.
Thank you for pointing out the phrase 'promising results' in lines 252-254. It is the wrong choice of words and we will make sure to avoid using such words in the updated manuscript.
We respectfully differ in our perspective on the potential bias and generation of artifacts resulting from subjectivity, as in our analysis, we implemented several measures to mitigate this issue. Firstly, we conducted sensitivity analyses to assess the impact of parameter variations on model outputs. This allowed us to evaluate the robustness of our results and identify parameters that exert the most influence on model performance. Secondly, we thoroughly evaluated the model's performance using multiple variables at multiple space locations, using goodness-of-fit statistics metrics and validated against the independent data set. Lastly, we transparently have provided detailed descriptions of the calibration process, including the rationale behind parameter selection and weighting decisions. This transparency allows readers to understand the basis for our choices and assess the validity of the results.
We understand the importance of ensuring that our paper meets the highest standards of quality and rigor, and we are fully committed to addressing any concerns you may have. We would gladly answer any further concerns.
Citation: https://doi.org/10.5194/hess-2023-276-AC2
-
AC2: 'Reply on RC2', Muhammad Haris Ali, 14 Mar 2024
-
RC3: 'Comment on hess-2023-276', Anonymous Referee #3, 09 Apr 2024
General Comments
The manuscript describes the application of an integrated hydrologic model (MIKE-SHE) to a lowland watershed in the Netherlands and Belgium.
The authors describe several aspects of the region to which the model is applied, they note the datasets used to create model inputs, and an overview of a sensitivity analysis and calibration exercise is provided. The calibration effort emphasized use of multiple data types (streamflow, groundwater head, and actual ET) and an objective function that modified weights of these components, with specific calibration and validation results provided in the supplemental material. The model is applied to provide generalized statements about the water budget of the region and a limited-scope sensitivity analysis of how spatial variability of precipitation affects streamflow and groundwater levels.While the development of a model, including sensitivity analysis and manual calibration, is undoubtedly a substantial effort, it is the use of the model to provide insight on some aspect of hydrologic systems that is generally of considerably higher value. The information presented in the manuscript, unfortunately, falls short of addressing a substantive scientific question in a demonstrable and convincing manner. In general, the manuscript would be greatly improved by providing a more thorough explanation of the motivation and specific question(s) being investigated as well as explaining the evidence used to support claims and assertions throughout - either through citations of other relevant works or by elaborating further on the analysis already completed. Some key areas potential improvements to the manuscript include the following:
- Many references in the manuscript are made to the complex interactions of groundwater and surface water in lowland catchments like the one being analyzed, yet little specific information is provided to characterize what these complexities or "intricacies" may be in a general sense (as in, what makes lowland catchments complex in the manner asserted?) and why this catchment is a good candidate for analysis (what data or evidence suggests the Aa of Weerijs demonstrates these complex phenomena?). Some examples may relate to spatiotemporal characteristics or the interaction of atmospheric-surface-and subsurface domains: Does the complexity arise from highly spatially variable groundwater-surface water exchange in channels? or maybe high temporal variability? threshold interaction effects from runoff and subsurface drainage? unique streamflow or groundwater level outcomes that depend on precipitation intensity or antecedent conditions?). Providing a clearer idea of the characteristics of interest in the catchment, and how the model represents them, will make the manuscript much more impactful.
- Related to the point (previous paragraph) about elaborating on work already performed - the main portion of the manuscript does not use figures and graphics to convey the message of the work as effectively as could be done. The MIKE-SHE model is spatially distributed, and spatial variability is in the manuscript title, yet the reader is left with little visual demonstration of how this manifests in the model results. For example - are there patterns in the variability of recharge, groundwater contributions to channel flow, or groundwater level response (and how might those inform interpretation of the calibration, sensitivity analysis, or future application of the model?)?
- The question of how much the spatial variability of precipitation inputs affect hydrologic models is an interesting one and should receive more thorough analysis in this manuscript. The effect of the different precipitation methods is presented as a variation in calibration metrics, but this obscures the direct model response (e.g. do groundwater levels go up or down when using radar vs IDW?) and prevents any broader hydrologic insights. Also - if using a high resolution radar product makes no difference in both calibration metrics and shows little impact to the spatial distribution of groundwater response or streamflow, then this tells us something interesting and very valuable, leading to the question "what about this catchment (or this model of the catchment) makes the response insensitive to the spatial resolution of precipitation?". An analysis that addresses these questions would make this manuscript considerably more meaningful.
Specific comments:
Page 5: 3.1 Model Setup and Input Data:
Because the analysis in the manuscript focuses on simulated water budget components and calibration metrics relating simulated and observed streamflows, heads, and ET, this section should provide sufficient detail for the reader to understand the context of those results. There is no mention of a conceptual hydrogeologic model - presumably there is some hydrostratigraphic complexity being simplified in the single 80-m thickness of the saturated zone? This should be explained in more detail. Similarly, what is the conceptual model of the surface system and it's connection to the groundwater? Is runoff in the catchment generally viewed as saturation or infiltration excess? How is that appropriately represented in the model? Actual ET is used as one of the station-based calibration comparison variables, but no information is provided on the model configuration used or the ET process represented. Similarly, no explanation of the root zone or the specifics of the subsurface drainage network is provided, even though the latter is implicated in the interpretation of results.Page 7, Figure 2: It seems most of the stations are located on or near the channels in the watershed. Does the calibration data being concentrated at stations along the main axis of the watershed limit the robustness of the calibration?
Page 8, Line 190: Some justification of the calibration metrics chosen would be helpful. The NSE and KGE are common enough, but the value of the correlation coefficient 'R' in this application needs some more explanation. It seems the use of the correlation coefficient in the calibration process could allow bias in results. How were each of the metrics used in the calibration process?
Page 9, Line 214: A sensitivity analysis is mentioned here, but no other information is provided. Did this one-at-a-time analysis provide insights into the relative role that each factor/parameter plays in changing groundwater levels, streamflow, and ET? Was there some threshold of sensitivity used to identify parameters for calibration?
Page 11, Line 255: "Hence, a reduction in drainage levels diverts more water into the stream, resulting in an improved model performance for streamflow. However, this adjustment adversely affected the model's performance in predicting groundwater levels (GWL)."
These statements imply a potential inconsistency in the model formulation - if draining more water from the subsurface to increase streamflows causes a greater mismatch in groundwater levels, then that suggests some portion of runoff that contributes to streamflow is being underrepresented somewhere. This should be addressed and explained more fully.Page 13 - Water balances: These sections would be prime candidates for figures or maps to demonstrate the temporal or spatial nature of the water balance that supports the intent of the manuscript. That precipitation is the largest inflow and ET is largest outflow is not a surprising hydrologic outcome - what further insights can a spatially explicit integrated hydrologic model provide?
Page 14 - Section 4.3: The role of precipitation spatial variablity on model outcomes is an interesting question - the analysis here should be elaborated on to provide more meaningful insights. Presenting the analysis only as the change in NSE, KGE, and/or correlation at the available groundwater level stations limits the potential to understand the local and overall sensitivity of the spatially explicit model to forcing of varying resolution. I'd suggest developing this section further with additional analysis and figures that address these questions in more depth.
Page 15 - Conclusions - It's always helpful to include some acknowledgement and explanation of the limitations of the work presented. I'd suggest adding a brief explanation to that effect prior to the conclusions.
Citation: https://doi.org/10.5194/hess-2023-276-RC3 -
AC3: 'Reply on RC3', Muhammad Haris Ali, 02 May 2024
General Comments
Reviewer comment 1: The manuscript describes the application of an integrated hydrologic model (MIKE-SHE) to a lowland watershed in the Netherlands and Belgium. The authors describe several aspects of the region to which the model is applied, they note the datasets used to create model inputs, and an overview of a sensitivity analysis and calibration exercise is provided. The calibration effort emphasized use of multiple data types (streamflow, groundwater head, and actual ET) and an objective function that modified weights of these components, with specific calibration and validation results provided in the supplemental material. The model is applied to provide generalized statements about the water budget of the region and a limited-scope sensitivity analysis of how spatial variability of precipitation affects streamflow and groundwater levels.
While the development of a model, including sensitivity analysis and manual calibration, is undoubtedly a substantial effort, it is the use of the model to provide insight on some aspect of hydrologic systems that is generally of considerably higher value. The information presented in the manuscript, unfortunately, falls short of addressing a substantive scientific question in a demonstrable and convincing manner. In general, the manuscript would be greatly improved by providing a more thorough explanation of the motivation and specific question(s) being investigated as well as explaining the evidence used to support claims and assertions throughout - either through citations of other relevant works or by elaborating further on the analysis already completed. Some key areas potential improvements to the manuscript include the following:
Authors’ response: We thank the reviewer for dedicating time to review our manuscript, providing feedback, and encouraging the efforts we put into this work. This gives us clear ways for improvement of the presented analysis. We outlined the motivation and specific questions addressed in Section 1, lines 75-85. As suggested by the reviewer, in the revised version of the manuscript we will incorporate additional citations related to our analysis and elaborate further by including more results from the model concerning various aspects of water balances, especially incorporating further evidence from data on recharge to groundwater, contribution of groundwater to surface runoff and contribution of groundwater to streamflow under different seasons.
The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the finalisation of the HESS discussion, are presented below.
Reviewer comment 2: Many references in the manuscript are made to the complex interactions of groundwater and surface water in lowland catchments like the one being analyzed, yet little specific information is provided to characterize what these complexities or "intricacies" may be in a general sense (as in, what makes lowland catchments complex in the manner asserted?) and why this catchment is a good candidate for analysis (what data or evidence suggests the Aa of Weerijs demonstrates these complex phenomena?). Some examples may relate to spatiotemporal characteristics or the interaction of atmospheric-surface-and subsurface domains: Does the complexity arise from highly spatially variable groundwater-surface water exchange in channels? or maybe high temporal variability? threshold interaction effects from runoff and subsurface drainage? unique streamflow or groundwater level outcomes that depend on precipitation intensity or antecedent conditions?). Providing a clearer idea of the characteristics of interest in the catchment, and how the model represents them, will make the manuscript much more impactful.
Authors’ response: Recent drought events in Europe have shown the potential impacts on streamflow and groundwater levels. The 2018 drought in the Aa of Weejis catchment exemplified these challenges, manifesting in decreased streamflow and groundwater levels. The predominant land use in the catchment is agriculture which in second to rainfall relies heavily on river water and most of the decision-making activities are linked to streamflow at the outlet of the catchment. Further, Aa of Weerijs is a lowland catchment in a temperate region with porous soil characteristics (sandy, sandy loam), intense channels network, almost flat topography with very mild slope, and shallow groundwater levels where the major portion of streamflow is contributed by groundwater. All these characteristics make this catchment a good candidate to study the behaviour of interaction between groundwater and surface water under different seasons in terms of groundwater recharge, contribution of groundwater to surface runoff (overland flow), and contribution of groundwater to streamflow. Therefore, a better understanding of the flow exchanges in the catchment can contribute to better management practices.
Moreover, in the temperate lowland catchments, the groundwater and surface water are closely coupled and influenced by climate drivers such as precipitation. Thus, our study also explores the impact of various spatial representations of precipitation on hydrological processes within the modelling domain. In the revised manuscript, we will present additional findings from our model results regarding the effects of different rainfall representations on groundwater recharge, the contribution of groundwater to overland flow, and the contribution of groundwater to streamflow to strengthen our findings in the revised manuscript.
Reviewer comment 3: Related to the point (previous paragraph) about elaborating on work already performed - the main portion of the manuscript does not use figures and graphics to convey the message of the work as effectively as could be done. The MIKE-SHE model is spatially distributed, and spatial variability is in the manuscript title, yet the reader is left with little visual demonstration of how this manifest in the model results. For example - are there patterns in the variability of recharge, groundwater contributions to channel flow, or groundwater level response (and how might those inform interpretation of the calibration, sensitivity analysis, or future application of the model?)?
Authors’ response: Thank you for this valuable remark. Indeed MIKE-SHE is a distributed model and provides spatially distributed results of many variables. In the updated manuscript, we will include the most relevant figures of the results that are mentioned in the authors’ response to comment 2.
Reviewer comment 4: The question of how much the spatial variability of precipitation inputs affect hydrologic models is an interesting one and should receive more thorough analysis in this manuscript. The effect of the different precipitation methods is presented as a variation in calibration metrics, but this obscures the direct model response (e.g. do groundwater levels go up or down when using radar vs IDW?) and prevents any broader hydrologic insights. Also - if using a high-resolution radar product makes no difference in both calibration metrics and shows little impact to the spatial distribution of groundwater response or streamflow, then this tells us something interesting and very valuable, leading to the question "what about this catchment (or this model of the catchment) makes the response insensitive to the spatial resolution of precipitation?". An analysis that addresses these questions would make this manuscript considerably more meaningful.
Authors’ response: The Aa of Weerijs catchment is relatively flat, and frontal precipitation predominates the region, resulting in a relatively uniform distribution of precipitation across the catchment. Consequently, modifying the resolution of precipitation data has minimal to no impact on streamflow at the outlet. However, the spatial variability of precipitation within the catchment does affect local recharge, leading to observed variations in groundwater head at specific points. We explained this in section 4.3, lines 324-328, and Table S7 in supplementary material presented the catchment average water balances under different precipitation representations. However, in the section 4.3, most of the results are explained in terms of NSE values. In the updated manuscript we will add further evidence from the data on the groundwater level variations, contribution of water from groundwater to surface runoff, contribution from groundwater to river, and recharge to groundwater under different rainfall representations so the local variations can be represented.
Specific comments:
Reviewer comment 5: Page 5: 3.1 Model Setup and Input Data: Because the analysis in the manuscript focuses on simulated water budget components and calibration metrics relating simulated and observed streamflows, heads, and ET, this section should provide sufficient detail for the reader to understand the context of those results. There is no mention of a conceptual hydrogeologic model - presumably there is some hydrostratigraphic complexity being simplified in the single 80-m thickness of the saturated zone? This should be explained in more detail. Similarly, what is the conceptual model of the surface system and it's connection to the groundwater? Is runoff in the catchment generally viewed as saturation or infiltration excess? How is that appropriately represented in the model? Actual ET is used as one of the station-based calibration comparison variables, but no information is provided on the model configuration used or the ET process represented. Similarly, no explanation of the root zone or the specifics of the subsurface drainage network is provided, even though the latter is implicated in the interpretation of results.
Authors’ response: Section 3.1, Lines 123 onward contains the modelling setup. The reference to MIKE-SHE is provided in Lines 125. We did not provide details about the modelling structure and governing equations of MIKE-SHE and MIKE-11 as we thought that it was outside the scope of this paper to reiterate existing literature that comprehensively explains these processes. Readers interested in detailed information about these modelling tools can consult available literature and technical manuals, for which we give references. We will include more explicit references to governing equations and modelling structure where they are available.
For the conceptualization of the saturated zone, we referred to the REGIS II: National Hydrogeological model for the Netherlands. Given the shallow groundwater table in the area and to streamline computational efforts, we consider the saturated zone as a single layer. However, we drew upon the range of horizontal hydraulic conductivities from REGIS II to represent the hydrogeological conditions of the catchment, considering these values as maximum and minimum during calibration (see section 3.3.2, lines 224-229). This approach yielded a satisfactory simulation of groundwater levels in the catchment, obviating the need for a more complex multi-layer conceptual hydrogeological model.
On the query regarding the conceptual model of the surface system and its connection to the groundwater, it is submitted that the surface runoff in the model happens when the net inflow to a cell either from rainfall or an adjacent cell exceeds the infiltration capacity of the soil. This water is ponded on the surface and will be routed downhill toward the river streams using the 2 D diffusive wave equation. The elevation of each cell is defined using the DEM. Further, if the groundwater table reaches the surface level then exchange also takes place between the saturated zone and the overland component leading to ponding of water that is also routed downhill. In this way, both saturation and infiltration excess runoff are considered. Otherwise, in a vertical direction, the surface system is connected to the groundwater in the saturated zone via the unsaturated zone (UZ).
The UZ of the model is represented by spatially distributed soil types wherein the movement of infiltrated water is simulated using the one-dimensional Richard's equation vertically. Soil evaporation and transpiration are sink terms in the UZ near the soil surface (in the root zone), which are modelled based on spatially distributed land use featuring various vegetation types. These components contribute to the overall evapotranspiration. Furthermore, storage change in the UZ and the recharge to the aquifer is controlled based on the water content in the UZ.
Upon reaching the saturated zone (SZ), recharge initiates a rise in groundwater levels. The flow within the SZ is modelled in three dimensions using the groundwater flow equation. Its interaction with river flow (modelled in MIKE-11) is represented as: 1) flow from subsurface drainage, using conceptual drains that can only receive water from the SZ, which is then routed to the river, contingent upon the positions of groundwater levels and drain levels, and 2) direct river-aquifer exchange (in both directions, determined by differences in levels between the SZ and the river). If the saturated zone extends to the root zone of plants (resulting in high groundwater levels), there can also be direct evaporation from this zone. Once the water table exceeds or equals the ground surface, the unsaturated zone within that specific cell becomes inactive and direct exchange with the overland component of the model takes place as described above.
Subsurface drainage is a special boundary condition in MIKE-SHE used to define natural and artificial drainage systems that cannot be defined in the River Network. We used this option to represent the small river channels that cannot be modelled in MIKE-11. When groundwater levels reach the defined levels of these conceptual drains, MIKE-SHE transfers the water to the nearest river link by conceptual linear reservoir routing controlled by the reservoir time constant parameter. For the accurate representation of this feature in the model, we considered the actual bed levels of these small channels as the levels of the conceptual drains in the model, and the time constant parameter was considered during calibration.
The actual evapotranspiration in a catchment is obtained as the sum of evaporation of intercepted rainfall by the canopy, transpiration from plants, and evaporation from ponds and the soil. Interception and evapotranspiration are modelled in MIKE-SHE based on the Kristensen and Jensen model (Kristensen and Jensen, 1975). The source of input data (LAI and root depth) is mentioned in Table 1 of the manuscript. Firstly, the model calculated the maximum canopy interception based on LAI and canopy evaporation. Secondly, transpiration from the root zone is considered as a function of moisture content and plant properties (LAI and root depth). Evaporation from the ponded water is considered at the rate of potential ET.
The river is modelled hydrodynamically in MIKE-11 with a defined river network and cross-section data, where the water is routed using 1D fully dynamic wave approximations of the Saint-Venant equations. At each computational time step, water levels in river reaches are compared with conditions in corresponding MIKE-SHE grids cell, for calculating the exchanges between the river and the aquifer and also from the overland surface to the river. Then these computed exchanges are returned to MIKE-11 rivers link as lateral flows to be incorporated in the next computational time step. The routing along the overland surface in MIKE-SHE also accounts for the losses due to evaporation and infiltration along the flow path.
As suggested by the reviewer we will update section 3.1 by including the above text in the revised manuscript.
Reviewer comment 6: Page 7, Figure 2: It seems most of the stations are located on or near the channels in the watershed. Does the calibration data being concentrated at stations along the main axis of the watershed limit the robustness of the calibration?
Authors’ response: We acknowledge your observation that most of the observation points are located along the channels in the watershed. However, it was beyond our control or influence as the observed data was available at these specific locations.
Reviewer comment 7: Page 8, Line 190: Some justification of the calibration metrics chosen would be helpful. The NSE and KGE are common enough, but the value of the correlation coefficient 'R' in this application needs some more explanation. It seems the use of the correlation coefficient in the calibration process could allow bias in results. How were each of the metrics used in the calibration process?
Authors’ response: Each evaluation metric has its strengths and limitations. For instance, the NSE is a widely used metric that measures the model's ability to reproduce the observed variance relative to a perfect fit. However, it tends to be more sensitive to peak flows, potentially skewing the assessment if the focus is primarily on peak flow events. Additionally, the R measures the linear relationship between observed and simulated values, offering insights into the overall trend. Therefore, we considered R in the calibration process, especially because we were interested in how well the trends in different seasons were captured by the model for groundwater levels. While Section 3.2, lines 151-156, briefly touch upon this aspect, we will provide further explanation in Section 3.3.1 of the updated manuscript.
Reviewer comment 8: Page 9, Line 214: A sensitivity analysis is mentioned here, but no other information is provided. Did this one-at-a-time analysis provide insights into the relative role that each factor/parameter plays in changing groundwater levels, streamflow, and ET? Was there some threshold of sensitivity used to identify parameters for calibration?
Authors’ response: The same parameters that were included in the sensitivity analysis were also included in the calibration process. The range of values in which the parameters were varied is presented in Table 2. We acknowledge that the insights into the relative role of each parameter in changing groundwater levels, streamflow, and ET are not presented in the manuscript. We think that this is beyond the scope of this article.
Reviewer comment 9: Page 11, Line 255: "Hence, a reduction in drainage levels diverts more water into the stream, resulting in an improved model performance for streamflow. However, this adjustment adversely affected the model's performance in predicting groundwater levels (GWL)." These statements imply a potential inconsistency in the model formulation - if draining more water from the subsurface to increase streamflows causes a greater mismatch in groundwater levels, then that suggests some portion of runoff that contributes to streamflow is being underrepresented somewhere. This should be addressed and explained more fully.
Authors’ response: We have written this text (line 255) in a more general manner to emphasize the significance of representing the conceptual drainage network in the model and how variations in it can impact streamflow and groundwater dynamics. Notably, it is essential to mention that the drain levels in the model corresponded to the true bed levels of these minor channels in the field and were not subjected to calibration during the process. In the revised manuscript, we aim to refine this text for clarity, ensuring it is easily understood by the reader and eliminating any potential confusion.
Reviewer comment 10: Page 13 - Water balances: These sections would be prime candidates for figures or maps to demonstrate the temporal or spatial nature of the water balance that supports the intent of the manuscript. That precipitation is the largest inflow and ET is largest outflow is not a surprising hydrologic outcome - what further insights can a spatially explicit integrated hydrologic model provide?
Authors’ response: We will enhance this section by incorporating additional text as mentioned in authors’ response to comment 2.
Reviewer comment 11: Page 14 - Section 4.3: The role of precipitation spatial variablity on model outcomes is an interesting question - the analysis here should be elaborated on to provide more meaningful insights. Presenting the analysis only as the change in NSE, KGE, and/or correlation at the available groundwater level stations limits the potential to understand the local and overall sensitivity of the spatially explicit model to forcing of varying resolution. I'd suggest developing this section further with additional analysis and figures that address these questions in more depth.
Authors’ response: We will enhance this section by incorporating additional text and tests as already mentioned in authors’ response to comment 2.
Reviewer comment 12: Page 15 - Conclusions - It's always helpful to include some acknowledgement and explanation of the limitations of the work presented. I'd suggest adding a brief explanation to that effect prior to the conclusions.
Authors’ response: Thank you for mentioning it. we will add limitations of the study in Section 5 of the updated manuscript.
References
Kristensen, K. J., & Jensen, S. E. (1975). A model for estimating actual evapotranspiration from potential evapotranspiration. Hydrology Research, 6(3), 170-188, doi.org/10.2166/nh.1975.0012.
Citation: https://doi.org/10.5194/hess-2023-276-AC3
Status: closed
-
RC1: 'Comment on hess-2023-276', Anonymous Referee #1, 15 Jan 2024
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2023-276/hess-2023-276-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Muhammad Haris Ali, 25 Jan 2024
We appreciate the reviewer for dedicating time to review our manuscript and for providing feedback. However, we are disappointed to learn that the decision to reject the manuscript was made after reading only few lines of the abstract.
The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the HESS discussion are presented below.
Reviewer comment 1: lines 14-16: authors said that they used the NSE and R, so what are the values they got? As opposed to listing the values, authors said ‘model performed well, with satisfactory….’, so what do ‘well’ and ‘satisfactory’ mean?
Authors’ response: The intention of these first lines is to introduce the reader to the elaborate evaluation of the model performance, which included streamflow, groundwater levels, and evapotranspiration data (please see lines 12-13). The actual detailed values we got and insights into this evaluation are presented in section 3.2., lines 147-150. We have elaborated on the evaluation specifics, indicating that streamflow, groundwater levels, and actual evapotranspiration data were assessed at 3, 13, and 13 locations, respectively. Their geographical distribution is depicted in Fig. 2, and the corresponding NSE and R values are tabulated in Table S4. Due to space limitations in the abstract, we refrained from explicitly stating the NSE and R values, and we conveyed the model's performance as "well" and "satisfactory" based on these outcomes and overall model performance in capturing the hydrological trend in comparison to observed data. The new manuscript will reformulate the abstract, reducing in some parts and including some of these values.
Reviewer comment 2: The only description of interactions between groundwater and surface water is in line 18, a very vague expression, but authors said, ‘the findings highlight the importance of surface water groundwater interactions ….’ in line 21. Why?
Authors’ response: The major portion of the streamflow is contributed by the groundwater through sub-surface drainage flow. Additional insights are mentioned in sections 4.2.3, 4.2.4, and 4.2.5, where we discussed the average yearly catchment water balance, as well as water balances during dry and wet years, elucidating exchanges between surface and groundwater. The assertion in line 21 of the abstract, "the findings highlight the importance of surface water-groundwater interactions," is reflective of the comprehensive findings discussed throughout the paper, not solely those presented in the abstract. As mentioned previously the abstract will be reformulated to state main findings.
Reviewer comment 3: Overall, all the conclusions in abstract are fundamental knowledge of hydrology in textbooks
Authors’ response: We have mentioned the conclusions that are drawn from our model results only. Their alignment with fundamental hydrological knowledge, support the robustness of our findings. This concurrence underscores the validity and reliability of our model, providing a solid foundation for the study's outcomes. We will add model output values and numbers in the abstract finding to make it more specific to our catchment.
Reviewer comment 4: It is very vague. for example, line 26, ‘a multi-scale approach that combines various techniques’, what do the authors want to express? I got nothing.
Authors’ response: In line 26, the intention is to convey that combining different techniques and concepts related to surface water and groundwater simulation at different scales of analysis (e.g., local, regional, global) can significantly improve our understanding of the exchanges between surface water and groundwater. This approach may allow us to gain a more comprehensive and accurate understanding of how water moves and interacts within hydrosystems, as highlighted by Ntona et al. (2022) in their review paper. The new manuscript will reformulate this for better clarity eliminating the vague formulation.
Reviewer comment 5: Then lines 28–31, some other fundamental knowledge. Also, I don’t think ‘groundwater tables’ is a right terminology, it should be ‘water table’.
Authors’ response: In lines 28-31, we underscored the significance of surface water-groundwater interaction, emphasizing that variations in precipitation can significantly influence these exchanges. We have emphasized on studying groundwater and surface water together instead of analysing them independently. We can rephrase the text in these lines to increase the readability in the updated manuscript.
Regarding the term "Groundwater table," it is our understanding that the use of both "groundwater table" and "water table" is interchangeable in scientific literature. Notably, the term "groundwater table" has been consistently employed in various research articles published in HESS, such as: https://doi.org/10.5194/hess-26-5859-2022, 2022, https://doi.org/10.5194/hess-25-1905-2021, 2021, etc.
Reviewer comment 6: Then lines 32-33, what does ‘unique characteristics’ mean? I got nothing again.
Authors’ response: In lines 32-33, the term "unique characteristics" refers to the specific properties of each catchment, encompassing factors such as size, slope, topography, soil type, land use/land cover, etc. These individual features contribute to the distinctiveness of each catchment. The revised manuscript will rephrase and will contain this explanation.
Reviewer comment 7: why ‘the temporal variations’ in line 33? why not spatial?
Authors’ response: The use of the term "temporal variation" in line 33 is intended to cover the seasonal fluctuations or variations associated with wet and dry conditions.
We have not used the term 'spatial variation' because we want to convey that at any particular location, the interaction between surface water and groundwater is influenced by catchment characteristics and climate variation. We will update this in the revised manuscript.
Reviewer comment 8: lines 34-35, the cited papers are 15-20 years ago. why ‘empirical field-based and numerical modeling’, why not remote sensing and machine learning?
Authors’ response: The mention of empirical field-based techniques, such as the use of tracers, and numerical modelling in lines 34-43 is intended to highlight the conventional methods commonly employed for the analysis of surface-groundwater interaction. In the updated manuscript we will extend with a short discussion on the role of RS and ML. For both conventional and for RS and ML methods, we will include additional recent references.
Reviewer comment 9: Then from line 36 to the end of the introduction, I got nothing new and the only useful information is a modeling work was conducted in the catchment, Aa of Weerijs, based on MIKE SHE and MIKE 11. Another interesting thing is, in line 79, authors said it is a mesoscale catchment while in line 86, it becomes a small scale.
Authors’ response: The lines 36-75 are intended to identify the knowledge gap and give the reader a comprehensive overview of the literature and the rationale behind the study, not necessarily in the particular catchment considered, but overall of similar catchments. Detailing the steps of introducing the knowledge gaps: in lines 36-45, several studies employing tracers for surface water-groundwater interaction are referenced. The limitations of field methods are discussed in lines 43-50, and the advantages of using numerical models to overcome these limitations are explained. The preference for physically based hydrological models over lumped models is highlighted in lines 53-57. The importance of accurate representation of rainfall data in the modelling domain is emphasized in lines 57-74. After making the context, the identified knowledge gap and the study's objective are outlined in lines 75-85. In the updated manuscript we will consistently use the term “mesoscale”.
Reviewer comment 10: lines 102–103, ‘The hydrology of the Aa of Weerijs catchment is influenced by various factors, including topography, land use, soil type, and climate.’ I am wondering which catchment is not affected by various factors as listed by authors?
Authors’ response: In 102-103, the intention was to emphasize on the complexity of catchment’s hydrology, indicating that no single catchment characteristic dominates. This aspect will be rephrased in the updated manuscript to ensure a more concise and clear expression of the message.
Reviewer comment 11: Lines 112–121, always, in this part, the governing equations and coupling approach are introduced. However, I got nothing again in this part. Though I knew MIKE SHE and MIKE 11 well. But if I was a common reader, I do have a question ‘what MIKE SHE and MIKE 11 are?’ I can only get that they are powerful and they can simulate everything, then what?
Authors’ response: Lines 123 onward contain the modelling setup. The reference to MIKE-SHE is provided in Lines 125. We believe that it is outside the scope of this paper to reiterate existing literature that comprehensively explains the modelling structure and governing equations of MIKE-SHE and MIKE-11. We will include in the new manuscript a phrase explaining what MIKE-SHE and Mike 11 are. We will refer readers seeking detailed information about these modelling tools to the available literature and technical manuals. We will include in the new manuscript a phrase explaining what MIKE-SHE and MIKE-11 are, though we hope that the HESS papers are read by experts and not “common readers”.
Reviewer comment 12: For the model itself, nothing is attractive. 500 resolution modeling over a domain of 346 km2
Authors’ response: It is our disappointment to see that for such a complex model nothing is attractive. We should emphasize more in abstract and introduction. We find great value in the development of a physically based, fully distributed model that utilizes both global and local datasets. This is not common procedure yet. The procedural method described for model development can be applied to model any other catchment globally. Our model includes hydrodynamic river modelling in MIKE-11, incorporating weirs and control structures to account for flow regulation—a feature not commonly found in hydrological modelling. Calibration and validation were performed using discharge, groundwater levels, and actual evapotranspiration data at 3, 13, and 13 locations, respectively. We employed a weighted objective function to consider multiple variables at various locations during the calibration and validation of the distributed hydrological model. The 500 m resolution was selected to strike a balance between achieving a comparatively short simulation time and retaining sufficient spatial detail to represent parameters and modelled processes.
Reviewer comment 13: I am really sorry. I suggest authors submit it to another journal. This is good documentation about the study of MIKE model in a real catchment. I didn’t find any advanced scientific question and anything different in terms of model and modeling. I really don’t think HESS is the right place.
Authors’ response: The primary objective of our study was not to introduce novel model structures or modelling approaches. Instead, we focused on addressing key aspects of hydrological modelling for meso-scale catchments, specifically the representation of rainfall data across the model domain. Another significant aim was to enhance our understanding of the complex interactions between surface water and groundwater for lowland meso-scale catchments. These objectives are explicitly outlined in lines 81-85 of the manuscript, with the corresponding findings presented in sections 4.2 and 4.3.
Citation: https://doi.org/10.5194/hess-2023-276-AC1
-
AC1: 'Reply on RC1', Muhammad Haris Ali, 25 Jan 2024
-
RC2: 'Comment on hess-2023-276', Anonymous Referee #2, 07 Mar 2024
-
AC2: 'Reply on RC2', Muhammad Haris Ali, 14 Mar 2024
We appreciate the reviewer for dedicating time to review our manuscript and for providing feedback. The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the finalisation of the HESS discussion, are presented below.
Reviewer comment 1: The abstract fails to adequately address key factors influencing the conceptual model, such as the regulation of streamflow and the agricultural nature of the watershed. These factors are crucial for understanding the context and conditioning of the modeling processes, yet their omission diminishes the clarity and completeness of the manuscript.
Author’s response: Thank you for this valuable comment. Indeed, the paper presents the key factors of catchment characteristics in section 2. Based on the suggestion the new version of the manuscript, will reformulate the abstract such that key factors of the catchment are included.
Reviewer comment 2: Acronyms, including MIKE SHE and MIKE 11, are frequently used throughout the manuscript without proper explanation or clarification of their respective roles and differences. This lack of clarity hinders readers' understanding of the modeling approach and its components.
Author’s response: Thank you for pointing this out. We will add a description of the acronyms MIKE SHE and MIKE 11. Their respective modelling roles are briefly described in section 3.1, lines 123-134. The new version of the manuscript will include explanations to ensure better clarity in this regard.
Reviewer comment 3: The manuscript contains numerous instances of vague language lacking both literature and quantitative support. Additionally, figures and tables are inadequately labeled, with critical information missing, such as a color scale for Figure 3. Confusing equation notations (Equation 1 and 2) and setups (line 201-202) further complicate comprehension and interpretation.
Author’s response: The new version of the manuscript will take into account the suggestion and will try to eliminate any vague sections. Though, indicating areas lacking quantitative support and literature would have helped us better to eliminate this vagueness. We will conduct a thorough read to identify where data and literature are lacking and add relevant information accordingly.
We would like to clarify further the comments about the figures: In Figure 3 we did not include a colour scale, as its purpose was not to show the quantity of rainfall, but to show the extent of the spatial distribution of precipitation over the model grid under three considered scenarios (see section 3.4, lines 241-247). we will make it clearer in the caption of the figure.
The notation used in equations is described in line 194 of the manuscript. We will check again and we will ensure that any missing notation is added in the updated version of the manuscript.
The setup explained in lines 201-202 will be reformulated to ensure better clarity. Overall, we will explain that the same weights are assigned to all considered variables (streamflow, Actual evapotranspiration, and groundwater levels) to calculate the mean NSE. However, each of these variables is measured at multiple space locations. For instance, streamflow is evaluated at three locations. To calculate the mean NSE of streamflow at multiple locations, more weight (0.4) is given to the streamflow at the outlet of the catchment due to its greater importance in water management practices, whereas a weight of 0.3 is considered at the other two locations. The details of the assigned weights are provided in the supplementary material as well (see Table S1). We will make sure the link to this material is clear.
Reviewer comment 4: Much of the modeling process appears to be subjective, with weights assigned based on authors' knowledge (line 203-209) and parameter tuning conducted manually (line 211-212) for purported insights. This subjectivity raises questions about the validity and reliability of the results, particularly considering the authors' acknowledgment that "promising" results (line 252-254) align with their mental model, suggesting potential bias and the generation of artifacts.
Author’s response: We appreciate the reviewer's concern regarding the subjective elements inherent in the modelling process, particularly regarding the assignment of weights and manual parameter tuning. The rationale for this is detailed in Section 3.3.1, lines 195-200, where the assignment of equal weight to multiple variables was based on literature, not on the author’s knowledge. However, when evaluating the same variable at multiple spatial locations, we adopted a more nuanced approach based on the knowledge of the Water Board who manages the catchment. Several sessions and interactions with stakeholders in the area took place to determine these weights through co-design. See explanations for the example of streamflow, in the previous comment. All weights were adopted based on co-designed with stakeholders. In the new version of the manuscript, we will emphasize this aspect, such that it is clear that the selection of weights is not an arbitrary choice.
Furthermore, we would like to emphasize on the other comments:
MIKE-SHE is a fully distributed physically based model that relies heavily on field data and literature for parameterization, and its setup is intricate, with simulations being time-intensive. Further, the sensitive parameters such as horizontal hydraulic conductivity and drainage time constant, are provided as spatially distributed grids over the model domain, further adding to the complexity. Given these circumstances, we opted for a one-at-a-time manual calibration approach. This method allowed us to carefully adjust individual parameters while considering their spatial distribution.
Thank you for pointing out the phrase 'promising results' in lines 252-254. It is the wrong choice of words and we will make sure to avoid using such words in the updated manuscript.
We respectfully differ in our perspective on the potential bias and generation of artifacts resulting from subjectivity, as in our analysis, we implemented several measures to mitigate this issue. Firstly, we conducted sensitivity analyses to assess the impact of parameter variations on model outputs. This allowed us to evaluate the robustness of our results and identify parameters that exert the most influence on model performance. Secondly, we thoroughly evaluated the model's performance using multiple variables at multiple space locations, using goodness-of-fit statistics metrics and validated against the independent data set. Lastly, we transparently have provided detailed descriptions of the calibration process, including the rationale behind parameter selection and weighting decisions. This transparency allows readers to understand the basis for our choices and assess the validity of the results.
We understand the importance of ensuring that our paper meets the highest standards of quality and rigor, and we are fully committed to addressing any concerns you may have. We would gladly answer any further concerns.
Citation: https://doi.org/10.5194/hess-2023-276-AC2
-
AC2: 'Reply on RC2', Muhammad Haris Ali, 14 Mar 2024
-
RC3: 'Comment on hess-2023-276', Anonymous Referee #3, 09 Apr 2024
General Comments
The manuscript describes the application of an integrated hydrologic model (MIKE-SHE) to a lowland watershed in the Netherlands and Belgium.
The authors describe several aspects of the region to which the model is applied, they note the datasets used to create model inputs, and an overview of a sensitivity analysis and calibration exercise is provided. The calibration effort emphasized use of multiple data types (streamflow, groundwater head, and actual ET) and an objective function that modified weights of these components, with specific calibration and validation results provided in the supplemental material. The model is applied to provide generalized statements about the water budget of the region and a limited-scope sensitivity analysis of how spatial variability of precipitation affects streamflow and groundwater levels.While the development of a model, including sensitivity analysis and manual calibration, is undoubtedly a substantial effort, it is the use of the model to provide insight on some aspect of hydrologic systems that is generally of considerably higher value. The information presented in the manuscript, unfortunately, falls short of addressing a substantive scientific question in a demonstrable and convincing manner. In general, the manuscript would be greatly improved by providing a more thorough explanation of the motivation and specific question(s) being investigated as well as explaining the evidence used to support claims and assertions throughout - either through citations of other relevant works or by elaborating further on the analysis already completed. Some key areas potential improvements to the manuscript include the following:
- Many references in the manuscript are made to the complex interactions of groundwater and surface water in lowland catchments like the one being analyzed, yet little specific information is provided to characterize what these complexities or "intricacies" may be in a general sense (as in, what makes lowland catchments complex in the manner asserted?) and why this catchment is a good candidate for analysis (what data or evidence suggests the Aa of Weerijs demonstrates these complex phenomena?). Some examples may relate to spatiotemporal characteristics or the interaction of atmospheric-surface-and subsurface domains: Does the complexity arise from highly spatially variable groundwater-surface water exchange in channels? or maybe high temporal variability? threshold interaction effects from runoff and subsurface drainage? unique streamflow or groundwater level outcomes that depend on precipitation intensity or antecedent conditions?). Providing a clearer idea of the characteristics of interest in the catchment, and how the model represents them, will make the manuscript much more impactful.
- Related to the point (previous paragraph) about elaborating on work already performed - the main portion of the manuscript does not use figures and graphics to convey the message of the work as effectively as could be done. The MIKE-SHE model is spatially distributed, and spatial variability is in the manuscript title, yet the reader is left with little visual demonstration of how this manifests in the model results. For example - are there patterns in the variability of recharge, groundwater contributions to channel flow, or groundwater level response (and how might those inform interpretation of the calibration, sensitivity analysis, or future application of the model?)?
- The question of how much the spatial variability of precipitation inputs affect hydrologic models is an interesting one and should receive more thorough analysis in this manuscript. The effect of the different precipitation methods is presented as a variation in calibration metrics, but this obscures the direct model response (e.g. do groundwater levels go up or down when using radar vs IDW?) and prevents any broader hydrologic insights. Also - if using a high resolution radar product makes no difference in both calibration metrics and shows little impact to the spatial distribution of groundwater response or streamflow, then this tells us something interesting and very valuable, leading to the question "what about this catchment (or this model of the catchment) makes the response insensitive to the spatial resolution of precipitation?". An analysis that addresses these questions would make this manuscript considerably more meaningful.
Specific comments:
Page 5: 3.1 Model Setup and Input Data:
Because the analysis in the manuscript focuses on simulated water budget components and calibration metrics relating simulated and observed streamflows, heads, and ET, this section should provide sufficient detail for the reader to understand the context of those results. There is no mention of a conceptual hydrogeologic model - presumably there is some hydrostratigraphic complexity being simplified in the single 80-m thickness of the saturated zone? This should be explained in more detail. Similarly, what is the conceptual model of the surface system and it's connection to the groundwater? Is runoff in the catchment generally viewed as saturation or infiltration excess? How is that appropriately represented in the model? Actual ET is used as one of the station-based calibration comparison variables, but no information is provided on the model configuration used or the ET process represented. Similarly, no explanation of the root zone or the specifics of the subsurface drainage network is provided, even though the latter is implicated in the interpretation of results.Page 7, Figure 2: It seems most of the stations are located on or near the channels in the watershed. Does the calibration data being concentrated at stations along the main axis of the watershed limit the robustness of the calibration?
Page 8, Line 190: Some justification of the calibration metrics chosen would be helpful. The NSE and KGE are common enough, but the value of the correlation coefficient 'R' in this application needs some more explanation. It seems the use of the correlation coefficient in the calibration process could allow bias in results. How were each of the metrics used in the calibration process?
Page 9, Line 214: A sensitivity analysis is mentioned here, but no other information is provided. Did this one-at-a-time analysis provide insights into the relative role that each factor/parameter plays in changing groundwater levels, streamflow, and ET? Was there some threshold of sensitivity used to identify parameters for calibration?
Page 11, Line 255: "Hence, a reduction in drainage levels diverts more water into the stream, resulting in an improved model performance for streamflow. However, this adjustment adversely affected the model's performance in predicting groundwater levels (GWL)."
These statements imply a potential inconsistency in the model formulation - if draining more water from the subsurface to increase streamflows causes a greater mismatch in groundwater levels, then that suggests some portion of runoff that contributes to streamflow is being underrepresented somewhere. This should be addressed and explained more fully.Page 13 - Water balances: These sections would be prime candidates for figures or maps to demonstrate the temporal or spatial nature of the water balance that supports the intent of the manuscript. That precipitation is the largest inflow and ET is largest outflow is not a surprising hydrologic outcome - what further insights can a spatially explicit integrated hydrologic model provide?
Page 14 - Section 4.3: The role of precipitation spatial variablity on model outcomes is an interesting question - the analysis here should be elaborated on to provide more meaningful insights. Presenting the analysis only as the change in NSE, KGE, and/or correlation at the available groundwater level stations limits the potential to understand the local and overall sensitivity of the spatially explicit model to forcing of varying resolution. I'd suggest developing this section further with additional analysis and figures that address these questions in more depth.
Page 15 - Conclusions - It's always helpful to include some acknowledgement and explanation of the limitations of the work presented. I'd suggest adding a brief explanation to that effect prior to the conclusions.
Citation: https://doi.org/10.5194/hess-2023-276-RC3 -
AC3: 'Reply on RC3', Muhammad Haris Ali, 02 May 2024
General Comments
Reviewer comment 1: The manuscript describes the application of an integrated hydrologic model (MIKE-SHE) to a lowland watershed in the Netherlands and Belgium. The authors describe several aspects of the region to which the model is applied, they note the datasets used to create model inputs, and an overview of a sensitivity analysis and calibration exercise is provided. The calibration effort emphasized use of multiple data types (streamflow, groundwater head, and actual ET) and an objective function that modified weights of these components, with specific calibration and validation results provided in the supplemental material. The model is applied to provide generalized statements about the water budget of the region and a limited-scope sensitivity analysis of how spatial variability of precipitation affects streamflow and groundwater levels.
While the development of a model, including sensitivity analysis and manual calibration, is undoubtedly a substantial effort, it is the use of the model to provide insight on some aspect of hydrologic systems that is generally of considerably higher value. The information presented in the manuscript, unfortunately, falls short of addressing a substantive scientific question in a demonstrable and convincing manner. In general, the manuscript would be greatly improved by providing a more thorough explanation of the motivation and specific question(s) being investigated as well as explaining the evidence used to support claims and assertions throughout - either through citations of other relevant works or by elaborating further on the analysis already completed. Some key areas potential improvements to the manuscript include the following:
Authors’ response: We thank the reviewer for dedicating time to review our manuscript, providing feedback, and encouraging the efforts we put into this work. This gives us clear ways for improvement of the presented analysis. We outlined the motivation and specific questions addressed in Section 1, lines 75-85. As suggested by the reviewer, in the revised version of the manuscript we will incorporate additional citations related to our analysis and elaborate further by including more results from the model concerning various aspects of water balances, especially incorporating further evidence from data on recharge to groundwater, contribution of groundwater to surface runoff and contribution of groundwater to streamflow under different seasons.
The responses to specific issues raised, and how they will be addressed in the version of the manuscript to be submitted after the finalisation of the HESS discussion, are presented below.
Reviewer comment 2: Many references in the manuscript are made to the complex interactions of groundwater and surface water in lowland catchments like the one being analyzed, yet little specific information is provided to characterize what these complexities or "intricacies" may be in a general sense (as in, what makes lowland catchments complex in the manner asserted?) and why this catchment is a good candidate for analysis (what data or evidence suggests the Aa of Weerijs demonstrates these complex phenomena?). Some examples may relate to spatiotemporal characteristics or the interaction of atmospheric-surface-and subsurface domains: Does the complexity arise from highly spatially variable groundwater-surface water exchange in channels? or maybe high temporal variability? threshold interaction effects from runoff and subsurface drainage? unique streamflow or groundwater level outcomes that depend on precipitation intensity or antecedent conditions?). Providing a clearer idea of the characteristics of interest in the catchment, and how the model represents them, will make the manuscript much more impactful.
Authors’ response: Recent drought events in Europe have shown the potential impacts on streamflow and groundwater levels. The 2018 drought in the Aa of Weejis catchment exemplified these challenges, manifesting in decreased streamflow and groundwater levels. The predominant land use in the catchment is agriculture which in second to rainfall relies heavily on river water and most of the decision-making activities are linked to streamflow at the outlet of the catchment. Further, Aa of Weerijs is a lowland catchment in a temperate region with porous soil characteristics (sandy, sandy loam), intense channels network, almost flat topography with very mild slope, and shallow groundwater levels where the major portion of streamflow is contributed by groundwater. All these characteristics make this catchment a good candidate to study the behaviour of interaction between groundwater and surface water under different seasons in terms of groundwater recharge, contribution of groundwater to surface runoff (overland flow), and contribution of groundwater to streamflow. Therefore, a better understanding of the flow exchanges in the catchment can contribute to better management practices.
Moreover, in the temperate lowland catchments, the groundwater and surface water are closely coupled and influenced by climate drivers such as precipitation. Thus, our study also explores the impact of various spatial representations of precipitation on hydrological processes within the modelling domain. In the revised manuscript, we will present additional findings from our model results regarding the effects of different rainfall representations on groundwater recharge, the contribution of groundwater to overland flow, and the contribution of groundwater to streamflow to strengthen our findings in the revised manuscript.
Reviewer comment 3: Related to the point (previous paragraph) about elaborating on work already performed - the main portion of the manuscript does not use figures and graphics to convey the message of the work as effectively as could be done. The MIKE-SHE model is spatially distributed, and spatial variability is in the manuscript title, yet the reader is left with little visual demonstration of how this manifest in the model results. For example - are there patterns in the variability of recharge, groundwater contributions to channel flow, or groundwater level response (and how might those inform interpretation of the calibration, sensitivity analysis, or future application of the model?)?
Authors’ response: Thank you for this valuable remark. Indeed MIKE-SHE is a distributed model and provides spatially distributed results of many variables. In the updated manuscript, we will include the most relevant figures of the results that are mentioned in the authors’ response to comment 2.
Reviewer comment 4: The question of how much the spatial variability of precipitation inputs affect hydrologic models is an interesting one and should receive more thorough analysis in this manuscript. The effect of the different precipitation methods is presented as a variation in calibration metrics, but this obscures the direct model response (e.g. do groundwater levels go up or down when using radar vs IDW?) and prevents any broader hydrologic insights. Also - if using a high-resolution radar product makes no difference in both calibration metrics and shows little impact to the spatial distribution of groundwater response or streamflow, then this tells us something interesting and very valuable, leading to the question "what about this catchment (or this model of the catchment) makes the response insensitive to the spatial resolution of precipitation?". An analysis that addresses these questions would make this manuscript considerably more meaningful.
Authors’ response: The Aa of Weerijs catchment is relatively flat, and frontal precipitation predominates the region, resulting in a relatively uniform distribution of precipitation across the catchment. Consequently, modifying the resolution of precipitation data has minimal to no impact on streamflow at the outlet. However, the spatial variability of precipitation within the catchment does affect local recharge, leading to observed variations in groundwater head at specific points. We explained this in section 4.3, lines 324-328, and Table S7 in supplementary material presented the catchment average water balances under different precipitation representations. However, in the section 4.3, most of the results are explained in terms of NSE values. In the updated manuscript we will add further evidence from the data on the groundwater level variations, contribution of water from groundwater to surface runoff, contribution from groundwater to river, and recharge to groundwater under different rainfall representations so the local variations can be represented.
Specific comments:
Reviewer comment 5: Page 5: 3.1 Model Setup and Input Data: Because the analysis in the manuscript focuses on simulated water budget components and calibration metrics relating simulated and observed streamflows, heads, and ET, this section should provide sufficient detail for the reader to understand the context of those results. There is no mention of a conceptual hydrogeologic model - presumably there is some hydrostratigraphic complexity being simplified in the single 80-m thickness of the saturated zone? This should be explained in more detail. Similarly, what is the conceptual model of the surface system and it's connection to the groundwater? Is runoff in the catchment generally viewed as saturation or infiltration excess? How is that appropriately represented in the model? Actual ET is used as one of the station-based calibration comparison variables, but no information is provided on the model configuration used or the ET process represented. Similarly, no explanation of the root zone or the specifics of the subsurface drainage network is provided, even though the latter is implicated in the interpretation of results.
Authors’ response: Section 3.1, Lines 123 onward contains the modelling setup. The reference to MIKE-SHE is provided in Lines 125. We did not provide details about the modelling structure and governing equations of MIKE-SHE and MIKE-11 as we thought that it was outside the scope of this paper to reiterate existing literature that comprehensively explains these processes. Readers interested in detailed information about these modelling tools can consult available literature and technical manuals, for which we give references. We will include more explicit references to governing equations and modelling structure where they are available.
For the conceptualization of the saturated zone, we referred to the REGIS II: National Hydrogeological model for the Netherlands. Given the shallow groundwater table in the area and to streamline computational efforts, we consider the saturated zone as a single layer. However, we drew upon the range of horizontal hydraulic conductivities from REGIS II to represent the hydrogeological conditions of the catchment, considering these values as maximum and minimum during calibration (see section 3.3.2, lines 224-229). This approach yielded a satisfactory simulation of groundwater levels in the catchment, obviating the need for a more complex multi-layer conceptual hydrogeological model.
On the query regarding the conceptual model of the surface system and its connection to the groundwater, it is submitted that the surface runoff in the model happens when the net inflow to a cell either from rainfall or an adjacent cell exceeds the infiltration capacity of the soil. This water is ponded on the surface and will be routed downhill toward the river streams using the 2 D diffusive wave equation. The elevation of each cell is defined using the DEM. Further, if the groundwater table reaches the surface level then exchange also takes place between the saturated zone and the overland component leading to ponding of water that is also routed downhill. In this way, both saturation and infiltration excess runoff are considered. Otherwise, in a vertical direction, the surface system is connected to the groundwater in the saturated zone via the unsaturated zone (UZ).
The UZ of the model is represented by spatially distributed soil types wherein the movement of infiltrated water is simulated using the one-dimensional Richard's equation vertically. Soil evaporation and transpiration are sink terms in the UZ near the soil surface (in the root zone), which are modelled based on spatially distributed land use featuring various vegetation types. These components contribute to the overall evapotranspiration. Furthermore, storage change in the UZ and the recharge to the aquifer is controlled based on the water content in the UZ.
Upon reaching the saturated zone (SZ), recharge initiates a rise in groundwater levels. The flow within the SZ is modelled in three dimensions using the groundwater flow equation. Its interaction with river flow (modelled in MIKE-11) is represented as: 1) flow from subsurface drainage, using conceptual drains that can only receive water from the SZ, which is then routed to the river, contingent upon the positions of groundwater levels and drain levels, and 2) direct river-aquifer exchange (in both directions, determined by differences in levels between the SZ and the river). If the saturated zone extends to the root zone of plants (resulting in high groundwater levels), there can also be direct evaporation from this zone. Once the water table exceeds or equals the ground surface, the unsaturated zone within that specific cell becomes inactive and direct exchange with the overland component of the model takes place as described above.
Subsurface drainage is a special boundary condition in MIKE-SHE used to define natural and artificial drainage systems that cannot be defined in the River Network. We used this option to represent the small river channels that cannot be modelled in MIKE-11. When groundwater levels reach the defined levels of these conceptual drains, MIKE-SHE transfers the water to the nearest river link by conceptual linear reservoir routing controlled by the reservoir time constant parameter. For the accurate representation of this feature in the model, we considered the actual bed levels of these small channels as the levels of the conceptual drains in the model, and the time constant parameter was considered during calibration.
The actual evapotranspiration in a catchment is obtained as the sum of evaporation of intercepted rainfall by the canopy, transpiration from plants, and evaporation from ponds and the soil. Interception and evapotranspiration are modelled in MIKE-SHE based on the Kristensen and Jensen model (Kristensen and Jensen, 1975). The source of input data (LAI and root depth) is mentioned in Table 1 of the manuscript. Firstly, the model calculated the maximum canopy interception based on LAI and canopy evaporation. Secondly, transpiration from the root zone is considered as a function of moisture content and plant properties (LAI and root depth). Evaporation from the ponded water is considered at the rate of potential ET.
The river is modelled hydrodynamically in MIKE-11 with a defined river network and cross-section data, where the water is routed using 1D fully dynamic wave approximations of the Saint-Venant equations. At each computational time step, water levels in river reaches are compared with conditions in corresponding MIKE-SHE grids cell, for calculating the exchanges between the river and the aquifer and also from the overland surface to the river. Then these computed exchanges are returned to MIKE-11 rivers link as lateral flows to be incorporated in the next computational time step. The routing along the overland surface in MIKE-SHE also accounts for the losses due to evaporation and infiltration along the flow path.
As suggested by the reviewer we will update section 3.1 by including the above text in the revised manuscript.
Reviewer comment 6: Page 7, Figure 2: It seems most of the stations are located on or near the channels in the watershed. Does the calibration data being concentrated at stations along the main axis of the watershed limit the robustness of the calibration?
Authors’ response: We acknowledge your observation that most of the observation points are located along the channels in the watershed. However, it was beyond our control or influence as the observed data was available at these specific locations.
Reviewer comment 7: Page 8, Line 190: Some justification of the calibration metrics chosen would be helpful. The NSE and KGE are common enough, but the value of the correlation coefficient 'R' in this application needs some more explanation. It seems the use of the correlation coefficient in the calibration process could allow bias in results. How were each of the metrics used in the calibration process?
Authors’ response: Each evaluation metric has its strengths and limitations. For instance, the NSE is a widely used metric that measures the model's ability to reproduce the observed variance relative to a perfect fit. However, it tends to be more sensitive to peak flows, potentially skewing the assessment if the focus is primarily on peak flow events. Additionally, the R measures the linear relationship between observed and simulated values, offering insights into the overall trend. Therefore, we considered R in the calibration process, especially because we were interested in how well the trends in different seasons were captured by the model for groundwater levels. While Section 3.2, lines 151-156, briefly touch upon this aspect, we will provide further explanation in Section 3.3.1 of the updated manuscript.
Reviewer comment 8: Page 9, Line 214: A sensitivity analysis is mentioned here, but no other information is provided. Did this one-at-a-time analysis provide insights into the relative role that each factor/parameter plays in changing groundwater levels, streamflow, and ET? Was there some threshold of sensitivity used to identify parameters for calibration?
Authors’ response: The same parameters that were included in the sensitivity analysis were also included in the calibration process. The range of values in which the parameters were varied is presented in Table 2. We acknowledge that the insights into the relative role of each parameter in changing groundwater levels, streamflow, and ET are not presented in the manuscript. We think that this is beyond the scope of this article.
Reviewer comment 9: Page 11, Line 255: "Hence, a reduction in drainage levels diverts more water into the stream, resulting in an improved model performance for streamflow. However, this adjustment adversely affected the model's performance in predicting groundwater levels (GWL)." These statements imply a potential inconsistency in the model formulation - if draining more water from the subsurface to increase streamflows causes a greater mismatch in groundwater levels, then that suggests some portion of runoff that contributes to streamflow is being underrepresented somewhere. This should be addressed and explained more fully.
Authors’ response: We have written this text (line 255) in a more general manner to emphasize the significance of representing the conceptual drainage network in the model and how variations in it can impact streamflow and groundwater dynamics. Notably, it is essential to mention that the drain levels in the model corresponded to the true bed levels of these minor channels in the field and were not subjected to calibration during the process. In the revised manuscript, we aim to refine this text for clarity, ensuring it is easily understood by the reader and eliminating any potential confusion.
Reviewer comment 10: Page 13 - Water balances: These sections would be prime candidates for figures or maps to demonstrate the temporal or spatial nature of the water balance that supports the intent of the manuscript. That precipitation is the largest inflow and ET is largest outflow is not a surprising hydrologic outcome - what further insights can a spatially explicit integrated hydrologic model provide?
Authors’ response: We will enhance this section by incorporating additional text as mentioned in authors’ response to comment 2.
Reviewer comment 11: Page 14 - Section 4.3: The role of precipitation spatial variablity on model outcomes is an interesting question - the analysis here should be elaborated on to provide more meaningful insights. Presenting the analysis only as the change in NSE, KGE, and/or correlation at the available groundwater level stations limits the potential to understand the local and overall sensitivity of the spatially explicit model to forcing of varying resolution. I'd suggest developing this section further with additional analysis and figures that address these questions in more depth.
Authors’ response: We will enhance this section by incorporating additional text and tests as already mentioned in authors’ response to comment 2.
Reviewer comment 12: Page 15 - Conclusions - It's always helpful to include some acknowledgement and explanation of the limitations of the work presented. I'd suggest adding a brief explanation to that effect prior to the conclusions.
Authors’ response: Thank you for mentioning it. we will add limitations of the study in Section 5 of the updated manuscript.
References
Kristensen, K. J., & Jensen, S. E. (1975). A model for estimating actual evapotranspiration from potential evapotranspiration. Hydrology Research, 6(3), 170-188, doi.org/10.2166/nh.1975.0012.
Citation: https://doi.org/10.5194/hess-2023-276-AC3
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 312 | 100 | 39 | 451 | 32 | 26 | 26 |
- HTML: 312
- PDF: 100
- XML: 39
- Total: 451
- Supplement: 32
- BibTeX: 26
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1