Hypothesis-Driven Conceptual Model for Groundwater–Surface Water Interaction at Aguieira Dam Reservoir (Central Portugal) Based on Principal Component Analysis and Hierarchical Clustering

Round 1

Reviewer 1 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

Accept in present form

Author Response

Comment 1: Accept in present form

Response 1: We very appreciate the previous reviews that were crucial for the improvement of this manuscript.

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

The manuscript titled “A conceptual model for groundwater-surface water interaction at Aguieira dam reservoir (Central Portugal) based on Principal Component Analysis and Hierarchical Clustering” presents a novel application of Rn-222 as a tracer combined with Principal Component Analysis (PCA) and Hierarchical Clustering (HC) to conceptualize groundwater-surface water interactions in a dam-controlled reservoir. The proposed piston-like conceptual model integrating structural geology and hydrochemistry in the Aguieira reservoir is innovative and addresses a significant knowledge gap in hydrogeological systems influenced by anthropogenic factors like reversible pump-turbines. However, this paper needs modifications and clarifications at some places to make it more informative to the readers.

Comments:

1. The introduction section needs more references to provide support to the statements being made.
2. The introduction section should automatically lead to the research gaps, and objectives of the study.
3. Keywords should be arranged alphabetically.

1. The absence of hydrodynamic modeling limits quantitative predictions of flow paths or fluxes. The authors are advised to address this in the manuscript.

5. Potential biases in RC interpretation (e.g., atmospheric degassing, mixing effects) are noted but not fully addressed.

1. Quantify and explicitly discuss uncertainties, especially regarding sampling depth variability and interpretation of PCA axes.
2. All references should be consistently formatted and accurately cited. The authors are advised to check the following references and may use them in the paper.

• https://doi.org/10.1007/978-3-030-60869-9_5

Author Response

Comment 1: The introduction section needs more references to provide support to the statements being made.

Response 1: Introduction have been updated with more literature, including newer references.

 

Comment 2: The introduction section should automatically lead to the research gaps, and objectives of the study.

Response 2: In the revised manuscript this has been made clearer to the reader (L70-82).

 

Comment 3: Keywords should be arranged alphabetically.

Response 3: This has been corrected in the revised manuscript.

 

Comment 4: The absence of hydrodynamic modeling limits quantitative predictions of flow paths or fluxes. The authors are advised to address this in the manuscript.

Response 4: We acknowledge that the absence of hydrodynamic modelling constrains the ability to provide quantitative predictions of flow paths or fluxes. Instead, our approach emphasizes the use of natural tracers and multivariate statistical analyses to qualitatively infer groundwater–surface water interactions. While these methods do not replace hydrodynamic modelling, they provide complementary insights, particularly in data-limited contexts such as in the studied setting, of fractured granitic rocks in an artificial and man-controlled reservoir. The proposed approach proved its worth since this work highlights the complexity of the system and endorses probable differences in groundwater flow path, related to the geological and structural setting, around the year. This is especially important since groundwater endmember is a major source of uncertainty in the quantitative models, accounting for 50-90% of the total uncertainty (L78-80).

 

Comment 5: Potential biases in RC interpretation (e.g., atmospheric degassing, mixing effects) are noted but not fully addressed.

Response 5: The referred sources of biases are very important to consider when model building. In this work they are considered and addressed throughout the text. As stated in [42] and in the newly added reference [37], vertical mixing between water layers during the stratified period is negligible due to the intense stratification. Therefore, each stratum preserves its own Rn-222 signal. The maximum of RC is also used as a tracer of the groundwater input. During the stratified period it occurs at the thermocline and/or hypolimnion, that are not subjected to atmospheric degas (L432-437]. During the winter, with higher water turbulence, radon could degas but two different scenarios are noted: one with a complete homogenous water column; and other, after heavy rains, that show an increment of Rn-222 towards the surface (L437-439]. Here we followed an approach like the one proposed by [37]: an increment in RC is a signal of groundwater input but a decrease in RC is not necessarily indicative of a diminishing of groundwater contribution. From all this previously mentioned relationships, it was proposed this conceptual model, that considers the possible mentioned biases or uncertainties and make use of their inherent properties to infer the reaction of the system.

 

Comment 6: Quantify and explicitly discuss uncertainties, especially regarding sampling depth variability and interpretation of PCA axes.

Response 6: The Rn-222 uncertainty related to water sampling was added to the methods section, as this protocol was validated through a replicate design test in another published paper [65]. The total estimated uncertainty (sampling + analysis) is 15% (95% CI). This uncertainty is expressed in the newly added figure 3. The in-depth sampling uncertainty is estimated to be 5% (95% CI).

Regarding the PCA interpretation, that relies on knowledge-based qualitative interpretation of the results it is not possible to quantify its uncertainty – the uncertainty of the interpretation is more related to the limitations of the work that are already stated in the limitations’ chapter.

 

Comment 7: All references should be consistently formatted and accurately cited. The authors are advised to check the following references and may use them in the paper.

• https://doi.org/10.1007/978-3-030-60869-9_5

Response 7: We have revised the references’ list for formatting issues. We also kindly thank the reviewer for the recent reference, that we found relevant to cite in the paper.

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

The work is promising and methodologically solid in parts, but the sampling limitations, lack of robust cluster validation, and speculative nature of the conceptual model mean that additional clarification, analysis, and discussion are needed before it reaches publishable quality. The conclusions are interesting but currently overextended relative to the data. Therefore, the manuscript requires a major revision.
Introduction includes well background study, but unfortunately major cites references are old which needs to be substitute with newer ones.
Graphical works (figures) need authors attention as some figures sometimes lack scale bars or clear axis units (especially Figure 5).
Regarding the model, there is a big if and real-life reliability questions. how real life limitation and influencing factor has been assessment? and based on what this model could be confirmed?
however, the piston effect model is plausible, but the link between precipitation and RC spikes needs stronger evidence; likewise, no hydrological flow or tracer transport modeling is used to confirm the mechanism — the model remains speculative. so it is recommended to change from a confirm statement to the preliminary and hypothesis-driven.
So add extra discussion on limitations and how this can be a developing pathway , instead of a confirmed statement.
Regarding statistical analysis and works,  no quantitative cluster validity indices (e.g., silhouette width) are reported, so the choice of 5 clusters remains subjective. Additionally, while PCA is valid here, with only 5 variables the dimensionality reduction benefit is limited; a correlation matrix or redundancy analysis could have sufficed.
so for this section it is recommended that authors Include objective validation metrics for the clustering and discuss alternative interpretations.
lack of discution about the time gap. 8-month gap between S1 and S2 means seasonal continuity is partially lost. how you would explain if there is one site, then without spatial replication, it's difficult to confirm that patterns are representative of the reservoir.

Author Response

Comment 1: Introduction includes well background study, but unfortunately major cites references are old which needs to be substitute with newer ones.

Response 1: Introduction have been updated with more literature, including newer references.

Comment 2: Graphical works (figures) need authors attention as some figures sometimes lack scale bars or clear axis units (especially Figure 5).

Response 2: We kindly say sorry, but we do not understand the concerns raised here regarding the graphical works. All the axis are identified and legible. On figure 1, geological mapping has a scale bar and rock photos have a pen to work as a visual scale. All the graphs have the corresponding axis identified and units. Figure 5 has the vertical axis identified with the corresponding unit and is showing the scaled parameters in the horizontal axis (adimensional). This is explained in the legend of the figure “values were scaled to show the relations between patterns. For original values please refer to [22]”.

Comment 3: Regarding the model, there is a big if and real-life reliability questions. how real life limitation and influencing factor has been assessment? and based on what this model could be confirmed?

however, the piston effect model is plausible, but the link between precipitation and RC spikes needs stronger evidence; likewise, no hydrological flow or tracer transport modeling is used to confirm the mechanism — the model remains speculative. so it is recommended to change from a confirm statement to the preliminary and hypothesis-driven.

So add extra discussion on limitations and how this can be a developing pathway , instead of a confirmed statement.

Response 3: We’ve decided to answer simultaneously these three important comments since they are unconditionally related.

There is no doubt that there are high uncertainties on the model or even oversimplification. All this is stated on limitations chapter (5.4). There it is also discussed the importance of this work and the future work directions, including other tracers that could help to differentiate the water flow paths. We think that these considerations are really important since at it is stated now on the revised manuscript in the introduction section (L70-82), the awareness raised from this data and this methodology is often not considered or addressed in model building (L80-82; [41]). As a good example in the literature, from the mathematical model of [37], the authors concluded that the major uncertainty is associated with the groundwater endmember, that could account for 50%-90% of the total uncertainty. Our work endorses this concerns from a multivariate perspective with knowledge-driven hypothesis, in low-permeability fractured rocks.

On the other hand, the goal of the article was to explore an unconventional approach to gather preliminary information regarding the hydrogeological system (understudied). From the novel database that explored the relationships in the reservoir (parameter wise and atmospherically wise) it was possibly to identify some major underlying conditions (PCs). They are indeed interpretations, but that are consistent with the patterns found on surface water along the year, the geological setting and the structural network. Furthermore, the observed RC in the water column (that is consistent with other parameters’ patterns) and its variation related to external pressures is not explained by other simple conceptual models.

Nevertheless, we totally agree with the suggestion of the reviewer on emphasizing that this is a combination of some knowledge-driven with mainly hypothesis-driven conceptual model and not a model-driven study, starting from the title of the article to include this concept, but also along the revised manuscript. L512-530 also address the limitations related to this preliminary conceptual model, based on the available knowledge and data, and its importance.

 

Comment 4: Regarding statistical analysis and works,  no quantitative cluster validity indices (e.g., silhouette width) are reported, so the choice of 5 clusters remains subjective. Additionally, while PCA is valid here, with only 5 variables the dimensionality reduction benefit is limited; a correlation matrix or redundancy analysis could have sufficed.
so for this section it is recommended that authors Include objective validation metrics for the clustering and discuss alternative interpretations.

Response 4: This is partially addressed in chapter 5.4 (Limitations and future research directions). PCA is not just a valuable technique for dimensionality reduction but also for exploratory analysis. This is addressed in L503-511. Here, we do not used this technique to reduce the dimensionality, but to relate the variables with the major dimensions for physical interpretation of the major stressors over the surface water. As also to visualize the dispersion of the clusters according to those stressors. This kind of work would be difficult to interpret from a correlation matrix.

Regarding the cluster validity, we did not include the silhouette width, since the result is not very conclusive. According to that index a major separation occurs at k=2 and stabilizes around k=5. This is indicative that a larger number of clusters is still holding structure. Other indices are just an extension of this and reflect their priority according to the calculus: indices that prioritize compactness would tend to lower k (e.g., Davies–Bouldin, Dunn) and indices that prioritize structure would tend to higher k (e.g., Calinski–Harabasz). We’ve now included silhouette width into the revised manuscript with a short explanation for relying on the elbows method (L234-242 and L265-277).

Comment 5: lack of discution about the time gap. 8-month gap between S1 and S2 means seasonal continuity is partially lost. how you would explain if there is one site, then without spatial replication, it's difficult to confirm that patterns are representative of the reservoir.

Response 5: It is stated in L170-172 that S1 (first survey) was not included in the analysis. Actually, it would be interesting to include it on the analysis to confirm that S1 stratums would be grouped accordingly to the clusters defined for S7, namely how would it be distributed between the third and fourth cluster and how it would relate to fifth cluster. Unfortunately, it is not possible due to the lack of ORP variable.

Author Response File: Author Response.pdf

Reviewer 4 Report (New Reviewer)

Comments and Suggestions for Authors

It seems that a substantial part of the paper was published as [22], perhaps because of this the authors do not show the Rn-222 data in a simple scheme (the number of samples and sampling periods was not so huge…). Readers would welcome such an overview, perhaps without other (EC, ORP, T…) data as given in Fig. 5.

to let the discussion to be complete, relevant properties of groundwater and surface water and their seasonal changes should be also presented in the paper, at least in a simple table form, as apart from statistics, there are not too many quantitative considerations physically describing groundwater flow and its variability… The reader would not be then surprised by statements like “Groundwater discharge increases RC and EC in the surface water and decreases the water pH, temperature should also increase…” (lines 275 – 276) that require close knowledge of the [22] paper.

lines 329-330: some characterisation of the groundwater sampling objects would be desirable: were these wells / boreholes (of what depth), or springs?

lines 365-366: comparison with karst conduits is partly misleading interpretation: faults can be widened, interconnected, but only exceptionally serving as quick turbulent flow media as known in karst aquifers... the same for line 400: it would be better to be satisfied with "faults and joints" characteristics of groundwater routes, as "channels" are too far-fetched in this context.

Figure 7: groundwater table should be incorporated also into the main picture, and enhanced in the appropriate detail...

Author Response

Comment 1: It seems that a substantial part of the paper was published as [22], perhaps because of this the authors do not show the Rn-222 data in a simple scheme (the number of samples and sampling periods was not so huge…). Readers would welcome such an overview, perhaps without other (EC, ORP, T…) data as given in Fig. 5.

Response 1: We tried to make the manuscript as concise as possible, directing the reader to the other published article. But the authors agree with this comment, that making this data directly available to the reader is desirable. As, so this has been included in the manuscript, as also a short description of how Rn-222 is in the water column along the year.

 

Comment 2: to let the discussion to be complete, relevant properties of groundwater and surface water and their seasonal changes should be also presented in the paper, at least in a simple table form, as apart from statistics, there are not too many quantitative considerations physically describing groundwater flow and its variability… The reader would not be then surprised by statements like “Groundwater discharge increases RC and EC in the surface water and decreases the water pH, temperature should also increase…” (lines 275 – 276) that require close knowledge of the [22] paper.

Rsponse 2: We opted to change the groundwater subchapter position to the beginning of the discussion and included the table 3 with this information, in comparison to the surface water, to better characterize the groundwater.

 

Comment 3: lines 329-330: some characterisation of the groundwater sampling objects would be desirable: were these wells / boreholes (of what depth), or springs?

Response 3: They are natural springs that freely discharge along the year. This was included in the revised manuscript, including statistical parameters from those waters.

 

Comment 4: lines 365-366: comparison with karst conduits is partly misleading interpretation: faults can be widened, interconnected, but only exceptionally serving as quick turbulent flow media as known in karst aquifers... the same for line 400: it would be better to be satisfied with "faults and joints" characteristics of groundwater routes, as "channels" are too far-fetched in this context.

Response 4: We agree that the English used could be misleading. Of course, it wasn’t our intention to compare the scale of the phenomenon between both geological contexts, as they are clearly different. Our intention was to make an exaggerated abstraction to textually give an idea to the reader. As so we also stated that the faults and joints do not act as wide-open conduits.  But we totally agree that this is not the best approach. We’ve revised those statements to make the approach more scientific in detriment of abstraction.

 

Comment 5: Figure 7: groundwater table should be incorporated also into the main picture, and enhanced in the appropriate detail...

Response 5: We’ve revised the figure to include: i) the water table in the weathered granite; ii) the short and longest water path through the faults’ system.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

I read the revised manuscript. I accept the revised manuscript in its present form. 

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

It can be accepted to be published at current form.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. A sentence should be added at the end of the abstract about the significance of this study.
2. At the end of the introduction should be added about the purpose, significance and innovation of this research.
3. Why is heading 2 the same as heading 3? Also, the headings are numbered in the wrong order.
4. What is the significance of the author's interpretation of Dimension 3 when only 2 dimensions were obtained in the principal component analysis based on the principle of eigenvalues greater than 1?
5. Lines 278-280: In fact, in principal component analysis, the eigenvalues of Dimension 3 are less than 1, which means that RC is not effective in the analysis. Is it appropriate for the author to establish a conceptual model based on such results?
6. This conceptual model feels too qualitative. Can the author provide some quantitative content?

Reviewer 2 Report

Comments and Suggestions for Authors

This study proposes a conceptual model to describe the interaction between groundwater and surface water. This model is important for studying the transformation relationships between groundwater and surface water in areas with similar hydrogeological conditions.It is recommended that the authors make minor revisions. The specific details are as follows:

 

1) The significance of this conceptual model needs to be supplemented in the abstract section.

 

2) The introduction section needs to supplement the advantages and disadvantages of the current models for the transformation relationships between groundwater and surface water, and highlight the significance of this study.

 

3) Sections 2 and 3 of the paper both cover materials and methods, and it is recommended to combine them into one section.

 

4) The discussion section needs to be reorganized and detailed arguments should be presented for several issues in the study. Currently, there is little difference between the writing and the description of the results.

 

4)  It is recommended to add a section in the discussion to address the limitations of the model and future research directions.

 

5) The conclusion section generally does not cite references.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

at the moment this paper is nothing more than a draft of a publication.

 

Abstract to be revised - i.e. first sentence is pretty arguable.

Introduction

Please introduce and detail  sw-gw exchanges meaning in the context of the paper

Are there similar works?

M&M

create a section 2.1 site description.

NOt clear about monitoring points. Is it just one point? the one n the map?

Can you discuss about the limitation of this appproach? Is the point representative?

Have you performed gw sampling? it does not look like... is it just sampling of surface water?

Please provide a table for sampling dates

How did you then define the GW end-member?

Results

bad quality pictures (i.e. fig 4)

Where are the data for Gw you used to discuss your data - even if they were pubished in another journal please sum up the results here ... otherwise is not possible to understand your work.

The discussion bringing to the conceptual model is very week. You should also look at other works pointing in the same directions.

Comments on the Quality of English Language

Hello,

This is the draft of a manuscript.

The authors have to duly revised it, prior it goes to revision.