Remote Detection of Cyanobacterial Blooms and Chlorophyll-a Analysis in a Eutrophic Reservoir Using Sentinel-2

Round 1

Reviewer 1 Report

In this study, correlation and regression analyses between chlorophyll-a/cyanobacteria concentrations and multispectral bands/indices are performed. The overall objective of the study is not clearly stated; however, results suggest authors are interested in obtaining predictive models to obtain cyanobacteria concentrations using Sentinel-2 satellite imagery. There are several aspects that should be addressed by the authors before accepting this manuscript for publication in Sustainability:

1. The main problem of this study is a poor predictive analysis execution which hinders any relevant contribution to the area of algal bloom monitoring. Particularly, no validation was performed and no confidence and prediction bands from regression analysis were reported. Therefore, it is not correct to talk about “predictive accuracy” when discussing the results. Why the authors did not use the upstream data for training purposes (i.e., regression analysis), and the downstream data to validate the trained models, or vice versa? Why the upstream/downstream datasets were not merged for the correlation analysis? Why confidence/prediction bands from the regression analysis are not reported?
2. The introduction does not clearly indicate what is the main motivation of the study. There is a wide number of previous studies using satellite imagery for algal bloom monitoring. The authors should clearly mention what advantages are expected from Sentinel-2 compared to previous open access satellite imagery. Also, previous studies using Sentinel-2 for algal bloom monitoring must be presented, and the authors should clearly mention what is the methodological contribution of this study to cyanobacteria detection using Sentinel-2 imagery and what they hypothesize. The methodology must be updated to reflect this and point 1.
3. Many elements in the discussion and conclusion sections of the manuscript are not substantiated by the study results. Additionally, several sentences in those sections are repeated or are very similar to sentences in the introduction. I am particularly critical of how the authors explained most of the resulting poor correlations/determination coefficients. Most of the explanations are based on other water constituents that may interfere with the spectral information. However, no field data is reported to support them, making most of the discussion very vague or speculative. If authors have additional water quality information, they should report it and include it in the analyses.

Please, find below additional specific observations intended to improve the overall quality of the manuscript:

• Lines 38-41: Sentences sound very repetitive. Please, improve the writing.
• Line 52: Height is not a water quality indicator (are you referring to secchi disk depth or similar?)
• Lines 58-59: Add reference here. Is this a contribution from your study compared to previous applications? If it is, mention it and explain it very clearly in the materials and methods section.
• Line 69: What is a “good” water security?
• Line 70: there is an error in “attain achieve”
• Line 71: “guarantee” is not the right word to use here.
• Lines 80-81: do you mean correlation analysis was performed?
• Line 89: “by correlating” --> “by performing a correlation analysis between…”
• Line 104: Hm --> hm
• Figure 1: Use a proper color code for land use. Why is Galicia and Pontevedra relevant on the map? All words must be in English (e.g., kilometers, no “kilometros”). What is the entrance/exit of the reservoir? The overall quality of the figure is very poor, it should be improved.
• In section 2.1 information such as the number of samples, frequency, coordinates of sampling points and the field and laboratory analysis methods must be reported (how were the samples collected and how the water quality variables were analyzed in the laboratory?).
• Lines 126-129: how was this performed?
• Line 137: What was the orthophoto used for?
• Lines 141-142: Is this part of the anomalies’ detection procedure?
• Lines 160-161: not clear, please rewrite.
• Lines 159-167: no explanation about gNDVI was done.
• Line 171: what were the criteria to select the type of regression equation?
• Section 2.4 must include a formal predictive analysis. Also, the statistical significance of the correlations must be reported.
• Section 3.1: data collected must thoroughly be described.
• Table 2 is not mentioned in the manuscript text and their results are not previously mentioned in the methodology. Table 2 and Table 4 should be replaced by dispersion plots indicating the confidence/predictive bands, determination coefficients and equations.
• Table 3 should include band names and wavelengths.
• Pearson correlation is normally expressed as “r” and coefficient of determination as R².
• Lines 197-199: not clear. Please, rewrite.
• Line 214: R_U value might be a typo since it is negative (opposite to the value reported in Table 3).
• Line 234: and --> to.
• Line 303: typo in 0 to 223,000 cells/mL (it is not consistent with Figure 3).
• Line 362: logarithm --> algorithm
• Appendix A: organize dates in ascending order.

Author Response

Dear reviewer 1,

Please, see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The studied topic is interesting and potential useful for large scale chla monitoring. I have bellowing comments.

• Formula in table 2 can be further improved.
• How the cyanobacteria concentration is analyzed in laboratory ?
• Figure 1. leyendà
• The reservoir is only 1.5km length and 1 km width. For sentinel-2 level-1C, how is the grid of the data product of sentinel-2 spectral band? How it can identify the upstream and downstream of the studied reservoir. Please explain more.
• I did not see how the established equations can be used for prediction of cyanobacterial blooms, as indicated in conclusions.
• How the spatial range of NDWI, NSMI, and Toming’s index are used in those regression equations. please explain.

Author Response

Dear reviewer 1,

Please, see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The presentation of the results still needs to be improved, especially the regression analysis. The choice of linear and second-order polynomials seems arbitrary. Authors need to justify this choice, or perform a formal model selection procedure. The new figures have a poor quality, and the bi-directional error bars are nor clearly described (how did you estimate the errors?). The figures must include the confidence bands from the regression analysis.

I find your explanation for separating the data problematic. You are finding different equations for the two different locations, so what criteria are you using to know which equation to apply to estimate concentration values in the entire reservoir? In future applications, how to know what equation to use elsewhere if a separated equation is fitted for each sampling site? This data separation must be substantiated in the manuscript.

The main objective is very vague (e.g. “an analysis of the possibility”). If you are performing a correlation/regression analysis, what you are pursuing is to explore or establish relationships between whatever explanatory variables you consider important and the cyanobacteria/chl-a concentrations. The objective must be stated clearly.

Authors said they reported additional field data in the appendix, but I could not find that information in the revised version.

I do not see which parts you removed from the discussion that are not related to the main findings.

Change in line 21 is problematic….correlating with what?

Change in line 22 from cyanobacteria to correlation does not make sense (what is correlation composition?)

In line 418: analyze -> analyzing

Please report the coordinates as latitude, longitude, not UTM.

Author Response

Dear Reviewer 1,

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

I recommend to publish the paper.

Author Response

Dear Revierwer 2,

Once again we would like to thank you for your collaboration in improving our manuscript. We regret the confusion about the submission of the response letter in the previous review and we thank you for your feedback in order to publish our work.

Best regards,

the authors

Round 3

Reviewer 1 Report

The authors did not satisfactorily address my comments.

The major issue is still the regression analysis, which has not been properly presented or substantiated. Reporting confidence limits is not an advanced statistical procedure but a conventional one! The figures are still low quality...

Also, it is problematic that some comments that were addressed in the previous round were retracted by the authors themselves in this new round (e.g., error bars, supplementary data). As a result, the overall work is less reliable.