Estimation of Suspended Sediment Concentration along the Lower Brazos River Using Satellite Imagery and Machine Learning

Round 1

Reviewer 1 Report

1、Lines 52 to 59 mention the advantages of using satellite imagery, which aligns with the needs of this study. It is suggested to enhance persuasiveness by adding relevant literature on satellite imagery usage in this section.

2、Line 82 mentions the use of machine learning models to counteract the shortcomings of satellite imagery in estimating SSC, which is consistent with this study. Therefore, it is recommended to add more literature support here, such as citing previous relevant studies.

 3、A scale and a north arrow are suggested to be added to Figure 1.

 4、Figure 2 displays SSC data for USGS-8116650-Past Studies, but it is not mentioned in the text above. What is the purpose of displaying this?

 5、There is significant whitespace on the left side of Figure 3, which could be indented for better visual appeal.

 6、Although Figure 4 is well-drawn, the information conveyed by the numbers 1, 2, 3, etc., in the figure is not clear enough. Further explanation in the figure or its caption is recommended.

 7、Line 190 mentions grouping bands together, is it for combination? It's a bit confusing.

 8、Section 2.4.1 describes the preparation of satellite data, but it's a bit lengthy. It is suggested to streamline it.

 9、In line 415, "the best performing," is "best" accurately expressed? Generally, "best" should indicate a clear direction, but all results are shown in the table.

 10、In section 3.1.1, the parameter mentioned in the correlation analysis is R2, but in the table, it's labeled as R. It is suggested to unify them for better integration of text and figures.

 11、In line 468, "One of the best performing models," it would be better to mention which one.

 12、The discussion section could reiterate the purpose and advantages of the study.

Author Response

Responses to comments are bolded. Your comments that were incorporated in the paper are shown in the manuscript submission and highlighted yellow. The cover letter submitted also describes the tracked changes for reviewers.

1、Lines 52 to 59 mention the advantages of using satellite imagery, which aligns with the needs of this study. It is suggested to enhance persuasiveness by adding relevant literature on satellite imagery usage in this section.

References added to help enhance the persuasiveness of this section.

2、Line 82 mentions the use of machine learning models to counteract the shortcomings of satellite imagery in estimating SSC, which is consistent with this study. Therefore, it is recommended to add more literature support here, such as citing previous relevant studies.

References added to support this statement.

 3、A scale and a north arrow are suggested to be added to Figure 1.

Scale and north arrow added to figure.

 4、Figure 2 displays SSC data for USGS-8116650-Past Studies, but it is not mentioned in the text above. What is the purpose of displaying this?

Reference in the text was added for this callout from the graph.

 5、There is significant whitespace on the left side of Figure 3, which could be indented for better visual appeal.

Figure intended and enlarged to improve visual appeal.

 6、Although Figure 4 is well-drawn, the information conveyed by the numbers 1, 2, 3, etc., in the figure is not clear enough. Further explanation in the figure or its caption is recommended.

Additional text added to the figure caption to help explain the figure.

 7、Line 190 mentions grouping bands together, is it for combination? It's a bit confusing.

Text modified and added to clarify the meaning of the groupings.

 8、Section 2.4.1 describes the preparation of satellite data, but it's a bit lengthy. It is suggested to streamline it.

Thank you for the comment. This text was reduced significantly for this version before submitting to review.

 9、In line 415, "the best performing," is "best" accurately expressed? Generally, "best" should indicate a clear direction, but all results are shown in the table.

Table 4 documents the different hyperparameters used in the grid search. The text before this table was slightly modified. The best performing hyperparameters are the ones that resulted in the lowest RSME for the respective inputs and model algorithm.

 10、In section 3.1.1, the parameter mentioned in the correlation analysis is R2, but in the table, it's labeled as R. It is suggested to unify them for better integration of text and figures.

R is used in the tables because it is used for the Taylor Diagram. R2 is also referenced in the text since it is the more common metric used in research studies during states of a model’s performance.

 11、In line 468, "One of the best performing models," it would be better to mention which one.

The text was updated to clarify the model used in the case study.

 12、The discussion section could reiterate the purpose and advantages of the study.

Text reiterating the purpose of the paper was added to the discussion section.

Author Response File: Author Response.pdf

Reviewer 2 Report

The objective of this study was to develop a method of estimating suspended sediment concentrations along the Lower Brazos River using satellite imagery.  This method, if successful, would provide an alternative to the use of relationships with measured volumes of flow in the river.

Minor problems

Figure 7

                  Labels needed on x axes.

Line 243

                  Equation number should be (11)

Figure 9

                  Labels needed on x axes.

It was not clear to me how the proposed models compared with the historic use of regressions of measured SSC versus Daily mean flow.  I am sure it is buried somewhere, but I would like to see plots of predicted versus measured SSC for the same test data points using the regressions of measured SSC versus Daily mean flow and the predicted values with the best model using remote sensing.  I would like to know how the two approaches compare.

Author Response

Responses to comments are bolded. Your comments that were incorporated in the paper are shown in the manuscript submission and highlighted light blue. The cover letter submitted also describes the tracked changes for reviewers.

Figure 7

                  Labels needed on x axes.

Labels of the band group is provided on the x axis for the normalized surface reflectance values. These reflectance values do not have units.

Line 243

                  Equation number should be (11)

Updated.

Figure 9

                  Labels needed on x axes.

Figure 9 is a plot of the river sections called out in Figure 1. So no x labels were assumed to be needed.

It was not clear to me how the proposed models compared with the historic use of regressions of measured SSC versus Daily mean flow.  I am sure it is buried somewhere, but I would like to see plots of predicted versus measured SSC for the same test data points using the regressions of measured SSC versus Daily mean flow and the predicted values with the best model using remote sensing.  I would like to know how the two approaches compare.

Figure 8 shows the comparison of all the models developed in the study to the same test data. The regression of measured SSC and daily mean flow is shown by the black circle in this plot. This plot shows a cluster of models developed have slightly better coefficients of regression than the regression of measured SSC and daily mean flow. Also, below are the graphs requested for reference for review.

Exponential – test:

Rating curve – test:

Author Response File: Author Response.pdf

Reviewer 3 Report

The presented work is very interesting and deals with an important topic. The content is well-presented and includes most elements of well-defined research. I suggest smaller changes to the Authors.

The paper considers two types of data, physical data, attained by standard measurements, and data attained from satellites. Can you provide a brief explanation of the measured data (both flow and sediment), how was it measured (equipment and sample analysis), and where (predefined locations in given cross sections, verticals, depths, or ?)? The same should be added for the presented case study. 

Section Modeling techniques should be complemented with a couple of sentences explaining the following subsections. 

The Manuscript should be expanded with additional referencing (e.g. for the equations given in section 2.5., for the statement "Stochastic gradient descent was used for training of ANNs", etc.)

Author Response

Responses to comments are bolded. Your comments that were incorporated in the paper are shown in the manuscript submission and highlighted green. The cover letter submitted also describes the tracked changes for reviewers.

The presented work is very interesting and deals with an important topic. The content is well-presented and includes most elements of well-defined research. I suggest smaller changes to the Authors.

The paper considers two types of data, physical data, attained by standard measurements, and data attained from satellites. Can you provide a brief explanation of the measured data (both flow and sediment), how was it measured (equipment and sample analysis), and where (predefined locations in given cross sections, verticals, depths, or ?)? The same should be added for the presented case study. 

Some additional text was added for details on the suspended sediment concentration methods used by the USGS and TCEQ agencies. Flow data collected by USGS could be through several different methods and no direct mention from a reference for the exact method used at the gauging location in this study was found. The case study did not use any new flow or suspended sediment concentration data that was not obtained from the TCEQ or USGS gages.

Section Modeling techniques should be complemented with a couple of sentences explaining the following subsections. 

Introductory text was included in this section to preface the subsections.

The Manuscript should be expanded with additional referencing (e.g. for the equations given in section 2.5., for the statement "Stochastic gradient descent was used for training of ANNs", etc.)

Reference callouts provided in locations in section 2.5.

Author Response File: Author Response.pdf