Retrieving Water Quality Parameters from Noisy-Label Data Based on Instance Selection

Round 1

Reviewer 1 Report

This manuscript proposes an improved method to identify noisy label instances in UAV hyperspectral water quality retrieval. It is somewhat innovative. However, the article lacks quantitative description and discussion, especially in the Conclusion part. And the improved methodological principle in the application of the method is not stated very clearly.
1.     There are few data samples in this paper, and it is somewhat insufficient to illustrate the effectiveness of the method proposed in the paper.

2.      It would be better to provide the water quality distribution inversion map in combination with the actual situation on site.

3.      Figure 1 lacks map elements, such as compass and scale. And would it be better to put Figure 1 in Section 2.1?

4.      Figure 3 also lacks a compass.

5.      The specific number of sample data needs to be explained in detail. Whether 28 samples and 8 samples mentioned in line 124 mean that there are only 36 sample data. As an intelligent algorithm, is the amount of measured data in the experiment too small?

6.      Is f(.) in Line 194 a typo?

7.      The method of noisy-label can be described in more detail, for example, what principle is used to establish these equations.

8.      It seems that the comparison of several methods using noisy-label in this manuscript has not been obviously improved. What is the theoretical basis for the use of Manhattan distance in Line 224?

9.      Why are the parameters in 3.1 set like this? For example, how does the k value and n_estimator come from?

10.   In Table 4, 5, 6 and Figure 5, please add the number of training samples and test samples and explain them in the manuscript.

11.   The content of all the current conclusions are all qualitative descriptions, which are not rigorous enough. And it is better not to put the arrangement of future work in Line 446-449, but to put it in a single paragraph.

12. Perhaps a discussion of the possible causes of the noise could be added to the Discussion Section.

Author Response

Dear Reviewer,

Thank you for your constructive comments for this article. Please see the attachment.

Best wishes,

Author

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript presented is interesting for the scope of the journal because it presents a procedure to study water quality variables from UAV images. It especially focuses on the study of turbidity in river systems where the fact that the waters have a continuous flow that gives them a more important variability than other systems such as lakes.

In my opinion the paper is well conceived but has some details that should be reviewed by the authors before publication. The sections of the manuscript are correct, although it is usually preferred that the results and discussion go separately in scientific papers. It is usual for them to be joined together when it is a case study. The most important deficiency is that the contents of the manuscript do not conform to the requirements indicated in the instructions to authors. In my subsequent comments I indicate section by section.

The abstract is longer than the journal recommendations and should be abbreviated to the required length.

The introduction should present citations from 1 in increasing order. It is noted that the first citation is number 3, so it is necessary to reorder.

In paragraph 85-102 there are numerous abbreviations that have not been defined the first time they appear in the text.

Prior to the contributions of the work, the objectives should be adequately presented.

Paragraph 104-107 cannot be presented in this form of wording, because it is a process that depends on the results of the work; it should be formulated something similar to "in case of noise problems, a methodology will be developed that ..." and it should not be the first contribution, because it depends on other previous aspects.

Regarding the methods, the figure of the study area should be improved, locating it on a map of China and indicating the geographical coordinates in some way, as in the next figure.

About the field measurements, it should be detailed with what kind of equipment the measurements are made. Some things, such as turbidity are obvious, but in reference to Chroma and COD more details should be provided, indicating make and model.

In the manuscript the distance to the sea of the measurement area in the river is not known, but I understand that it is quite far from the sea. Therefore, it does not make sense to indicate that there were breezes, which refer to a type of wind of a certain speed blowing from the sea to land or inverse in a coastal strip of a few tens of kilometers at most. Therefore, one should speak of light winds or as appropriate according to the speed they would have measured. Never breezes.

In the treatment of invalid pixels, they are called noisy; however it is likely that there are several types of problem for a pixel to be unsuitable for the study, for example because of the angle between the camera and the position of the sun. It is likely that the realization of the flights in the presented directions does not comply with the necessary difference with the sun angle to avoid reflections on the water. The authors could provide more information on the type of invalid pixels and even present some examples between valid and invalid.

In the results, the correlation coefficient between variables and wavelengths is presented; however, in the methodology it is stated that the coefficient of determination will be used. It should be specified which statistic is being used and also present in the graph where the significance level is, so that readers can see where it is and is not significant. These details should be made clear in the text of section 3.2 and in Figure 4.

Moreover, if they are using a correlation coefficient, it would be necessary to indicate whether the normality of the data has been studied and from that detail, what type of correlation has been used. Statistically, it is usual not to use extreme values in the analysis, as the authors do in some cases, for example in paragraph 358-359 in which they acknowledge the removal of extreme values. It is likely that a prior study of normality would already remove the outliers, before performing the various downstream processes.

Even though the authors present a section entitled results and discussion, the paper lacks a discussion in the academic sense indicated in the instructions to authors, so they should provide this section in the style of the journal.

References should be presented in the style indicated by the journal for authors.

Author Response

Dear Reviewer,

Thank you for your constructive comments for this article. Please see the attachment.

Best wishes,

Author

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is important for science. 

The work with VANT is relevant and innovative.

Drone Walter quality estimation is

importante to contribute with Walter Security.

the papear is important in tematic.

Sugestions 

Improve citations with 2022 publications.

Author Response

Dear Reviewer,

Thank you for your constructive comments for this article. Please see the attachment.

Best wishes,

Author

Author Response File: Author Response.pdf

Reviewer 4 Report

The article addresses the important problem of water quality assessment. Using new techniques and improving existing ones is the right approach. The authors presented an interesting solution that fits into the above assumptions.

I recommend the article for publication, but it needs some revisions.

Please clearly state the purpose of the article/research.

Unfortunately, the authors do not refer in any way to the limitations of UAV use. I am referring, among others, to meteorological conditions such as strong winds and precipitation.

The article should necessarily be expanded with a Discussion chapter, which is currently missing. The results obtained should be related to other studies using UAVs for water quality assessment.

References

The literature record does not conform to MDPI standards.

Author Response

Dear Reviewer,

Thank you for your constructive comments for this article. Please see the attachment.

Best wishes,

Author

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I think the manuscript has been improved to warrant publication in Remote Sensing

Author Response

Dear Reviewer,

Thank you for your suggestions. We have revised the format and reference order in the minor revision progress. Thank you again for your constructive suggestions.

Kind regards,

 Authors

Reviewer 2 Report

The authors have incorporated the proposed modifications in the new revised version of the manuscript. The academic content is correct. However, there are still some stylistic and formatting errors, which should be corrected in the final version. 

First, it has already been pointed out that the citation system used by the journal requires that citations be numbered in consecutive order. The authors should review again the introduction, where citation 11 has been placed before citation 6, so citation 11 should be 6 and renumber the following citations again. Also, citations 41 and 43 have been placed on line 51, and citation number 42 on line 52.

It has also been observed that the tables have been numbered with the Roman numeral system, when the journal uses Arabic numerals. This detail will presumably be corrected in the editing process.

Author Response

Dear Reviewer,

Thank you for your suggestions. We have revised the format and reference order in the minor revision progress.  The format in the Table is changed to Palatino Linotype according to the template. And the references are now in the correct order following your kind explanation. Thank you again for your constructive suggestions.

Kind regards,

 Authors