Challenges and Limitations of Using Monitoring Data in Catchment-Based Models—A Case Study of Rivers Taw and Torridge, UK

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article presented the results of modeling using SWAT models  as applied to a pilot study to investigate spatial and temporal issues around monitoring of faecal indicator bacteria (Escherichia coli) in rivers of the Taw and Torridge catchments in the UK. The study makes a significant contribution to the science.  However, it could be improved based on the following comments:

1- Just below the "Materials and Methods", a detailed flowchart of the methodology is presented as follows: Figure 1 presents a detailed flowchart of the methodology used in this study. This methodology will be explained in the next subsections. Please start the methodology with "Identifying the research gap, Identifying the study area, Data collection, ......, outcomes of the study".

2- In section 2.2 (Measurement of bacteria counts in riverine samples), indicate which standards you followed in sampling, transporting, storing, and testing of the collected samples. Cite the source and list it. 

3- Discuss the modeling calibration and validation (or verification) based on the availability of other studies or field data. Discuss the accuracy of both calibration and validation (agreement with other results) or (verification (agreement between simulated and observed data).

4- Present the results of the residual analysis to prove that the developed models for each are not biased.

5- If possible, present the recommendations in a separate section.

6- Improve the quality of the figures, for example: the text used in the figures (titles of the x-axis and y-axis, etc.). 

Author Response

Thank you for taking the time to review the paper and for your useful comments. We have addressed the comments below and hope that they are to your satisfaction.

Comment. Just below the "Materials and Methods", a detailed flowchart of the methodology is presented as follows: Figure 1 presents a detailed flowchart of the methodology used in this study. This methodology will be explained in the next subsections. Please start the methodology with "Identifying the research gap, Identifying the study area, Data collection, ......, outcomes of the study".

Response. We have included a flow diagram of the study set up and analysis in the Supplementary Materials to guide the reader if needed. We did not feel that there was room to include this in the main body of the text.

Comment. In section 2.2 (Measurement of bacteria counts in riverine samples), indicate which standards you followed in sampling, transporting, storing, and testing of the collected samples. Cite the source and list it. 

Response. Thank you for drawing our attention to this.  Standard protocols were used in sampling, transporting and analysing the river water samples, and we have included text and a reference to reflect this.

Comment. Discuss the modeling calibration and validation (or verification) based on the availability of other studies or field data. Discuss the accuracy of both calibration and validation (agreement with other results) or (verification (agreement between simulated and observed data).

Response. As is standard practice for SWAT model performance we have covered the comparison of the model output for streamflow versus observed river flow data (see lines 392 to 421). We have included in this section classification of the model NSE values according to Moriasi et al 2007 to give the reader a clear interpretation of model performance.

Comment. Present the results of the residual analysis to prove that the developed models for each are not biased.

Response. The SWAT models are presented with pBIAS values to indicate how the hydrology compares with the actual values and we have included information of the efficiency and correlation of these models between simulated and sampled bacterial loads.

Comment. If possible, present the recommendations in a separate section.

Response. In the manuscript we kept to the section headings available on the template. To make it clearer where the reader can find recommendations, we have renamed the final section ‘Conclusions and recommendations’. We hope this is sufficient to cover this point.

Comment. Improve the quality of the figures, for example: the text used in the figures (titles of the x-axis and y-axis, etc.).

Response. Thanks for this comment. The figure captions for Figures 5 and 6 now include an explanation for the reference line showing exact parity between in situ and modelled values. The figures have all been improved and some supplied as vector graphics to make them clearer.

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

The manuscript describes the development of a model used for the prediction of spatial and temporal variation of faecal indicator bacteria 16 (Escherichia coli) in rivers of the Taw and Torridge catchments in the UK. The study is based on a large described methodology, the ideas are well written, well explain. As a general comment, I suggest the authors to enhances to challenges and limitations of using the developed model. 

In the attached file, you will find another comments. 

Comments for author File: Comments.pdf

Author Response

Thank you for taking the time to review the paper and for your useful comments. We have addressed the comments below and hope that they are to your satisfaction.

Comment. Line 182: Why did the authors use different SWAT models? What differences are there between the two river basins that could interfere in the catchment modeling? Please state that in the manuscript.

Response. Thank you for the comment. The reason for separate SWAT models being generated is covered in section 2.3.

Comment. Line 217: The figure is not very clear. The authors should try a different approach in describing the cattle density (maybe different colors). Moreover, in the figure caption, it is hard to follow the logic in the sentence. If authors state first a statement about Taw catchment and after that a statement of Torridge catchment, please be consistent throughout the explanation.

Response. Thank you for the comment. Changes to the caption text have been made. We did not choose to use different colours to represent the different livestock density because this can be an issue for accessibility. Therefore, we feel the use of symbol size to differentiate between density categories is more appropriate here.

Comment. Line 357: It is not very clear how the study area is divided into sites A, B, C, D, etc.

Response. The river sample site locations and how/why they were chosen is provided in Figure 1, and in the text under 2.2.

Comment. Line 379: Where do they start? 

Response.  Unfortunately, we were unable to resolve this comment as it was not clear what needed to be addressed.

Comment. Line 468: Please explain in a more comprehensive way the points that deviate from the trend in Figure 6a.

Response. We have included information about the trend line in the figure caption.

Comment. Line 492: The authors should provide all abbreviation meanings at first mention.

Response. Thank you for the comment. LMM is first mentioned in section 2.7 and is contained in the abbreviations list.

Comment. Line 522: Please increase the font used in pictures. It is very hard to read axis captions.

Response. Thanks for this comment. The figure captions for Figures 5 and 6 now include an explanation for the reference line showing exact parity between in situ and modelled values. The figures have all been improved and supplied as vector graphics to make them clearer.

Comment. Line 576: Could the authors give examples of some studies that combine both physico-chemical and microbiological parameters in Europe?

Response. Based on other reviewers comments this has been removed from the discussion and therefore has not been addressed.

Comment. Line 683: Please list the advantages and disadvantages of your model.

Response. We have added text in the discussion about the limitations of the model, especially around the difficulty in modelled episodic events such as intense rainfall.

Comment. Line 726: Please state future plans for improving the methodology and investigating spatial-temporal distribution of bacterial counts.

Response. We have included potential future work in the discussion.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper “Challenges and limitations of using monitoring data in catchment-based models - a case study from Rivers Taw and Torridge, UK” presents modelling approach into water quality. The paper is well written, however there are some issues that needs to be addressed.

In the lines 57-60 the authors state that catchment models have had "less frequent application in the routine assessment of bacterial loadings to waterbodies." This generalization does not fully reflect the existing work in this area. Numerous studies and models, such as SWAT, HSPF, and others, have been applied to model microbial or fecal indicator organisms (e.g., E. coli, fecal coliforms) in catchments, particularly for water quality and public health assessments.

 

In the lines 91-95 authors provide a general description of the SWAT model, referencing its ability to simulate a wide range of hydrological and transport processes. However, it remains unclear whether all these components, including soil percolation, lateral flow, and groundwater discharge, were actually used in the current study. The focus appears to be limited to precipitation, land cover, and cattle presence, which raises concerns about the modelling for fecal bacteria transport.

In modeling fecal bacteria, the transport pathways are critical. Bacteria can move via: Direct deposition to streams (e.g., by grazing cattle), surface runoff during rainfall events, subsurface flows (though this pathway is less common and highly dependent on soil characteristics and calibration), resuspension from streambed sediments (which SWAT can only partially simulate).

In lines 121-124 the authors present the sparsity of bacterial monitoring data and its seasonal/rainfall-driven variability as a primary challenge (Challenge 1). While this is certainly true, it is also a widely recognized and well-documented issue in catchment hydrology and microbial water quality studies. For example, multiple studies (e.g., by Kay et al., 2008; Gannon et al., 2005; Whitehead et al., 2009) have demonstrated the episodic nature of bacterial contamination, especially during storm events.

In lines 163-165 the authors state that 12 water samples were collected in six months in 2023. However, there is no accompanying information about the hydro-meteorological context of that year. Was 2023 a hydrologically typical year for the catchments studied? Were the sampling dates representative of both baseflow and high-flow (storm) conditions?
This context is essential to assess the representativeness of the dataset and the grounds forany conclusions regarding bacterial loading dynamics. I recommend the authors include at least a summary of 2023 rainfall and river discharge patterns (e.g., from gauging stations), and compare them to long-term values.

The authors rely on daily precipitation sums in their analysis. However, this temporal generalization may obscure critical differences in rainfall characteristics that influence bacterial allocation. For example, a short, high-intensity storm (which tends to generate more surface runoff and associated bacterial transport) can have the same daily total as a long-duration, low-intensity rain event (which promotes infiltration and less overland flow).
I recommend that the authors discuss this limitation and, where possible, consider incorporating sub-daily rainfall data (e.g., hourly) or at least characterize event types (e.g., intensity–duration–frequency analysis) to better link precipitation dynamics with microbial loading processes.

 

In line 260 it should be: from Sowah et al. [40].

 

The methodology for estimating bacterial inputs from discharges is unclear. While the authors state that daily dry weather flow (DWF) volumes were used to estimate bacterial input (lines 264–267), they do not describe how these flows were calculated and converted into bacterial loads within the model.

In lines 426-427 the comparison between simulated and observed bacterial loads is presented as a mean value at each site. However, using mean values alone is not a sufficient or meaningful approach for evaluating the performance of a model that simulates dynamic, event-driven processes such as bacterial transport.
I recommend that the authors instead compare the model outputs directly against observed values at the matching time points (i.e., the dates when samples were collected). This would provide a more quantitative and informative assessment of model accuracy and bias.
If such a direct comparison is not feasible due to temporal mismatches or model time-step limitations, the authors should at least discuss this limitation and avoid relying solely on summary statistics like means.

 

Figures 5-8 needs revision. The fonts are to small and illegible. In figures 5 and 6 the lines are not explained in the capture. If they are trend lines, please provide an equation, R² and p value.

 

The discussion section is the weakest part of the paper. Up to the line 622 it is rather introduction that discussion.

The latest part is packed with the vague statements, like:  "Land use was identified as a predictor variable for the summer period". What specific land use types? How strong was the correlation? Was this statistically significant? What variables were tested?

Ungrounded statements, like: the jump from “livestock are more likely to be grazing” to “more likely to encounter watercourses for drinking” to “adding to bacterial reservoir” is overly speculative and not supported with evidence. Livestock presence doesn't automatically imply direct watercourse interaction unless field access, fencing, and farm management practices are known.

The reasoning presented regarding land use as a predictor variable for bacterial concentrations in summer lacks clarity and proof. While it's plausible that grazing livestock may contribute to increased bacterial loads, the connection drawn between land use and bacterial presence is speculative and insufficiently supported.

In lines 640-643 the authors state that faecal bacteria accumulate during dry weather and are washed into watercourses during intense rainfall, leading to high bacterial loads. While this mechanism is well established in the literature, no evidence is provided from the current study to demonstrate this effect.
If the authors did not explicitly analyze rainfall-runoff events following dry periods, this statement should be framed more cautiously as a hypothetical mechanism rather than a demonstrated result of their study.

In conclusion:

The discussion raises several important points regarding the challenges of simulating bacterial dynamics in catchments using SWAT. The recognition of data limitations, spatial heterogeneity, and the importance of livestock and rainfall inputs is welcome, if not required.
Furthermore, the discussion often lacks rigorous linkage between claims and the study’s actual findings. Many mechanisms are stated without supporting data from this study. Terms such as “reasonable agreement” and “model responded similarly” should be backed with quantitative metrics.
Furthermore, speculative statements should be clearly marked as hypotheses unless supported by data.

 

Author Response

Thank you for taking the time to review the paper and for your very useful comments. We have addressed the comments below, outlining the changes and inclusions to the manuscript that have been made, and we hope that they are to your satisfaction.

 

Comment. In the lines 57-60 the authors state that catchment models have had "less frequent application in the routine assessment of bacterial loadings to waterbodies." This generalization does not fully reflect the existing work in this area. Numerous studies and models, such as SWAT, HSPF, and others, have been applied to model microbial or fecal indicator organisms (e.g., E. coli, fecal coliforms) in catchments, particularly for water quality and public health assessments.

Response. Thank you for this comment. We agree that the text needs clarification and have modified accordingly.

Comment. In the lines 91-95 authors provide a general description of the SWAT model, referencing its ability to simulate a wide range of hydrological and transport processes. However, it remains unclear whether all these components, including soil percolation, lateral flow, and groundwater discharge, were actually used in the current study. The focus appears to be limited to precipitation, land cover, and cattle presence, which raises concerns about the modelling for fecal bacteria transport.

Response. The SWAT model incorporates modelling of soil percolation, lateral flow, and groundwater discharge and text has been modified to make this clearer. For example, we have added the fact that soil properties are taken into account by the model. We have made clear that bacterial transport via groundwater is not captured in the model (lines 115 to 118) but is modelled in the top 10 cm of soil and can enter the stream network via runoff or sediment transport.

Comment. In modeling fecal bacteria, the transport pathways are critical. Bacteria can move via: Direct deposition to streams (e.g., by grazing cattle), surface runoff during rainfall events, subsurface flows (though this pathway is less common and highly dependent on soil characteristics and calibration), resuspension from streambed sediments (which SWAT can only partially simulate).

Response. The fate of bacteria in the SWAT model is covered in the Introduction and we have worked on this section to make it more clear to the reader.

Comment. In lines 121-124 the authors present the sparsity of bacterial monitoring data and its seasonal/rainfall-driven variability as a primary challenge (Challenge 1). While this is certainly true, it is also a widely recognized and well-documented issue in catchment hydrology and microbial water quality studies. For example, multiple studies (e.g., by Kay et al., 2008; Gannon et al., 2005; Whitehead et al., 2009) have demonstrated the episodic nature of bacterial contamination, especially during storm events.

Response. Thank you for the comment. We have included information in the introduction on the known episodic nature of bacterial contamination. The authors feel that this still represents a major challenge especially as the timing of known spill events from the sewerage system in the UK is poor. We have clarified the challenge to specify the study area as the nature of these events will be specific to catchments.

Comment. In lines 163-165 the authors state that 12 water samples were collected in six months in 2023. However, there is no accompanying information about the hydro-meteorological context of that year. Was 2023 a hydrologically typical year for the catchments studied? Were the sampling dates representative of both baseflow and high-flow (storm) conditions?
This context is essential to assess the representativeness of the dataset and the grounds for any conclusions regarding bacterial loading dynamics. I recommend the authors include at least a summary of 2023 rainfall and river discharge patterns (e.g., from gauging stations), and compare them to long-term values.

Response. Thank you for the comment. We have included 2 figures detailing the summary information regarding the streamflow for 2023 in the Supplementary Materials. We did not consider it necessary to include the figures in the paper as this provides context and we have included summary text in section 2.2. Furthermore, we have made it clear that the study represents a pilot study to highlight and explore the issues and not a comprehensive study.

Comment. The authors rely on daily precipitation sums in their analysis. However, this temporal generalization may obscure critical differences in rainfall characteristics that influence bacterial allocation. For example, a short, high-intensity storm (which tends to generate more surface runoff and associated bacterial transport) can have the same daily total as a long-duration, low-intensity rain event (which promotes infiltration and less overland flow).
I recommend that the authors discuss this limitation and, where possible, consider incorporating sub-daily rainfall data (e.g., hourly) or at least characterize event types (e.g., intensity–duration–frequency analysis) to better link precipitation dynamics with microbial loading processes.

Response. Thank you for your consideration on this point. We agree that use of daily precipitation can mask intense rainfall events that may result in higher-than-normal transport of bacteria via surface runoff. However, the modelling approach is a balance with all the input data sources, and in this case whilst SWAT can run at a sub-daily time step, we feel doing so would mask the other large uncertainties around the other sources of bacterial input such as direct discharge via STW and CSOs, livestock grazing times, locations, slurry production and application, and the direct access to water courses. Moreover, bacterial decay modelling in SWAT operates at a daily time step and running SWAT with daily rainfall events is not uncommon for such a type of study. We have added text in the discussion around this limitation.

Comment. In line 260 it should be: from Sowah et al. [40].

Response. Thank you for spotting this. Change made.

Comment. The methodology for estimating bacterial inputs from discharges is unclear. While the authors state that daily dry weather flow (DWF) volumes were used to estimate bacterial input (lines 264–267), they do not describe how these flows were calculated and converted into bacterial loads within the model.

Response. We have made the method of estimating the bacterial loads for point sources clearer including for CSOs. This includes the addition of a reference for the flow rates used.

Comment. In lines 426-427 the comparison between simulated and observed bacterial loads is presented as a mean value at each site. However, using mean values alone is not a sufficient or meaningful approach for evaluating the performance of a model that simulates dynamic, event-driven processes such as bacterial transport.
I recommend that the authors instead compare the model outputs directly against observed values at the matching time points (i.e., the dates when samples were collected). This would provide a more quantitative and informative assessment of model accuracy and bias.
If such a direct comparison is not feasible due to temporal mismatches or model time-step limitations, the authors should at least discuss this limitation and avoid relying solely on summary statistics like means.

Response. Thank you for this comment. The use of a mean value for the site was considered due to the high level of variability observed in the in situ data, an observation seen by others (e.g. Coffey et al 2010) and in keeping with many other studies using SWAT to model nutrient and bacterial loads where monthly mean values are used and even visual comparisons (Ferreira et al 2023). With large uncertainly around diffuse and point source inputs we would not expect a good correlation at the daily time period. To assist the reader, we have added statistics looking at the direct comparison between observed and modelled data but would caution on its use due to the large uncertainty around many of the inputs. We have added text to the discussion on this to make it clearer to the reader.

Comment. Figures 5-8 needs revision. The fonts are to small and illegible. In figures 5 and 6 the lines are not explained in the capture. If they are trend lines, please provide an equation, R² and p value.

Response. Thanks for this comment. The figure captions for Figures 5 and 6 now include an explanation for the reference line showing exact parity between in situ and modelled values. The figures have all been improved and supplied as vector graphics to make them clearer.

Comment. The discussion section is the weakest part of the paper. Up to the line 622 it is rather introduction that discussion.

Response. We thank the reviewer for this constructive comment and have incorporated much of the first part of the discussion into the Introduction and reframed the beginning of the Discussion section accordingly.

Comment. The latest part is packed with the vague statements, like:  "Land use was identified as a predictor variable for the summer period". What specific land use types? How strong was the correlation? Was this statistically significant? What variables were tested?

Response. Thank you for the comment. We have revised where appropriate to include information about significance in the discussion without unduly repeating from the results section.

Comment. Ungrounded statements, like: the jump from “livestock are more likely to be grazing” to “more likely to encounter watercourses for drinking” to “adding to bacterial reservoir” is overly speculative and not supported with evidence. Livestock presence doesn't automatically imply direct watercourse interaction unless field access, fencing, and farm management practices are known.

Response. Summary statistics in the discussion were included to raise this point concerning livestock access to watercourses - see line 675. The paper highlights the fact that we do not have this information.

Comment. The reasoning presented regarding land use as a predictor variable for bacterial concentrations in summer lacks clarity and proof. While it's plausible that grazing livestock may contribute to increased bacterial loads, the connection drawn between land use and bacterial presence is speculative and insufficiently supported.
Response. In our study we have not indicated a direct association between a specific landuse and the bacterial loads. The modelling variable represents a set of principal components of the hydrological response units of which landuse is a part. We have modified this in the discussion.

Comment. In lines 640-643 the authors state that faecal bacteria accumulate during dry weather and are washed into watercourses during intense rainfall, leading to high bacterial loads. While this mechanism is well established in the literature, no evidence is provided from the current study to demonstrate this effect. If the authors did not explicitly analyze rainfall-runoff events following dry periods, this statement should be framed more cautiously as a hypothetical mechanism rather than a demonstrated result of their study.
Response. Thank you for this comment. During this study, we did not explicitly analyse rainfall-runoff events following dry periods, as these are difficult to predict and the capacity to sample watercourses in response to meteorological events was not within the scope of the project.  We have made it clear that the objectives of the study were to compare modelled outputs with observational data, based on routine monitoring, and acknowledged that routine observations may not capture ‘peak events’.  We have added a sentence to the discussion to highlight that these observed peaks after long dry periods are well-documented in the literature, and for this reason we do not feel that it is necessary to include additional references here.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

1- The flowchart of the methodology missed several boxes. Start by "identification of the research gap by studying the literature."

2- The last box should be a "diamond shape" with two branches, "Yes and " No". If "No", then you need to go back for calibration, and if "Yes", go for the next step (box), "Analysis of results and draw conclusions."

3- Unify the writing style of the references (check how the title of each reference is written and unify the style; check the journal guidelines).

Author Response

Thank you for your time reviewing the manuscript and providing useful feedback.

• The flowchart of the methodology missed several boxes. Start by "identification of the research gap by studying the literature."

Response: Thank you for this suggestion for an addition to the flowchart. The inception of the project was based on a pilot study to determine spatial and temporal characteristics that would be needed for an effective monitoring programme of inputs to an estuary dedicated as an official shellfish harvesting area, so we do not feel that it is necessary to include “identification of the research gap...” in this flow chart. 

• The last box should be a "diamond shape" with two branches, "Yes and " No". If "No", then you need to go back for calibration, and if "Yes", go for the next step (box), "Analysis of results and draw conclusions."

Response: We have added details on the flow chart concerning model calibration which occurs during the ‘catchment model generation’ phase. The flow chart accurately describes the methodology employed. Please note that calibration of the SWAT model involves a highly iterative, multiparameter process that can found in several papers, and we have cited a reference for the reader in the flow chart.

3- Unify the writing style of the references (check how the title of each reference is written and unify the style; check the journal guidelines).

Response: All references have now been checked and modified in accordance with journal requirements.

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors, 

Congratulations for your work. I hope that the suggestion of all reviewers made the manuscript even more valuable. 

Sincerely,

 

Author Response

Response: Thank you for your time reviewing the manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have made notable improvements to the manuscript. However, several important issues remain that still need to be addressed to ensure scientific soundness.

There is a mismatch between the number of in-text citations and the references listed. Specifically, there are 65 references cited in the main manuscript, but 70 entries in the References section. This suggests that there are five references listed but never cited in the text. I recommend the authors carefully cross-check all in-text citations and reference list entries to ensure consistency.

At present, the still limited use of citations in the Discussion weakens the interpretive depth of the manuscript. I recommend the authors revisit the Discussion and incorporate more critical engagement with past studies.

While a strong engagement with the literature is appreciated, the manuscript tends to overuse citations in the Introduction section, with some sentences followed by four or more references. In several cases, this seems unnecessary.

The current Discussion section focuses heavily on the SWAT model’s performance and setup. While some reflection on model reliability is warranted, the section should show a stronger focus on the environmental implications of the obtained results.

In the Introduction, the authors highlight three key challenges or assumptions that frame the study. However, the Discussion section does not return to these points to assess whether they were supported by the modelling results. For a more coherent narrative and stronger scientific argument, I recommend the authors explicitly revisit each of these initial assumptions or hypotheses in the Discussion.

 

 

Author Response

Thank you for your time reviewing the manuscript and your helpful comments.

The authors have made notable improvements to the manuscript. However, several important issues remain that still need to be addressed to ensure scientific soundness.

There is a mismatch between the number of in-text citations and the references listed. Specifically, there are 65 references cited in the main manuscript, but 70 entries in the References section. This suggests that there are five references listed but never cited in the text. I recommend the authors carefully cross-check all in-text citations and reference list entries to ensure consistency.

Response: The discrepancy between the number of references in the main text and the number cited is due to citation of references in the Supplementary Material, as per journal requirements. We have checked the references thoroughly and, to prevent future confusion, the references in the Supplementary Materials are also directly referenced in the main manuscript.

At present, the still limited use of citations in the Discussion weakens the interpretive depth of the manuscript. I recommend the authors revisit the Discussion and incorporate more critical engagement with past studies.

Response: The discussion has been reworked and engagement with past studies has been strengthened.

While a strong engagement with the literature is appreciated, the manuscript tends to overuse citations in the Introduction section, with some sentences followed by four or more references. In several cases, this seems unnecessary.

Response: Thank you for the comment. The points in the introduction where this occurs refer to a wide range of studies employing catchment modelling, and specifically SWAT modelling, to various problems and issues. Providing a wide range of examples shows the utility and widespread use of the models and enables the reader to further explore this area. We do not feel these are unnecessary and are in common with many other journal papers.

The current Discussion section focuses heavily on the SWAT model’s performance and setup. While some reflection on model reliability is warranted, the section should show a stronger focus on the environmental implications of the obtained results.

Response: The central aspect of the paper is to highlight the problem with having insufficient monitoring data, and ill-defined input catchment data, when attempting to produce a model that responds faithfully. For this reason, there is a key focus on models, and specifically SWAT, used in this study. It would be highly speculative to map the outcomes of this work to specific environmental implications, and we feel that we have already addressed this more generally in the discussion. However, we have added more detail regarding environmental factors, infrastructure and activities, considering both spatial and temporal factors, that may influence model performance, and underpin recommendations for improved monitoring design.

In the Introduction, the authors highlight three key challenges or assumptions that frame the study. However, the Discussion section does not return to these points to assess whether they were supported by the modelling results. For a more coherent narrative and stronger scientific argument, I recommend the authors explicitly revisit each of these initial assumptions or hypotheses in the Discussion.

Response: Thank you for this suggestion. The challenges have been revisited in the discussion (lines 644-649 – challenge 1; lines 728-736 – challenge 2; lines 701-707 – challenge 3).

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

While a strong engagement with the literature is appreciated, the manuscript tends to overuse citations in the Introduction section, with some sentences followed by four or more references. In several cases, this seems unnecessary. Please reduce citations in the introduction section.

Author Response

While a strong engagement with the literature is appreciated, the manuscript tends to overuse citations in the Introduction section, with some sentences followed by four or more references. In several cases, this seems unnecessary. Please reduce citations in the introduction section.

Response: Thank you for your engagement with the manuscript to make it a better paper. The Introduction has 33 citations and we have identified 3 places (lines 74 to 80; line 97 and lines 107 to 112) where multiple citations (examples) are present and have reduced these in number (removing 10 citations). We have been careful to choose the most appropriate citations for the reader.