Characterizing Water Composition with an Autonomous Robotic Team Employing Comprehensive In Situ Sensing, Hyperspectral Imaging, Machine Learning, and Conformal Prediction

Round 1

Reviewer 1 Report

The manuscript aims to construct remote sensing method for water quality using autonomous robotic team that consists of an uncrewed surface vessel and an unmanned aerial vehicle. The reviewer appreciates authors’ every effort drawn in the manuscript. Overall comments are as follows.

1. Optically non-active variables such as pH, conductivity, and ions were estimated in this study. It needs more discussions why optically non-active variables were used as target variables.  

2. In Section 2.3 Data Collection, UAV and USV collected data near solar noon. In general, reflectance was not collected for three hours around solar noon because of sunglint. Why was collection conducted at solar noon and how to treat the effect of sunglint?

3. In a random forest model, there are several parameters that need to be optimized. Based on line 230, this study optimized two parameters such as the number of trees and sampling fraction. How to optimize other parameters such as:
1) the minimum leaf size (the number of training samples that required to be present in a leaf node of a decision tree)
2) the maximum features (the number that random forest is allowed to try in individual tree)  

3) the maximum depth (the depth of the decision trees)

4) the minimum split size (the number of training samples that required to be present in a node of a decision tree in order for it to be split during training)

Results for the parameter optimization are needed to be mentioned in the manuscript.

4. 25 features were selected using permutation importance in each model. How many features are used in this study? The total list of features are needed to be mentioned in the manuscript. 

5. Various machine learning methods were used in this study, based on mention in Section 2.4. In Table2, there are single performance in each target variable. What kinds of model were used in Table 2? How to choose the best model? Modeling results including the optimized model parameters and performance are needed to be mentioned in the manuscript or supplementary material.

Minor points

- Line 191, “MLJ”, full name is needed.

- Line 197, “CART”, full name is needed.

- Line 450, What is “the German EnMap”? It needs description.

Overall, sentences are too long and difficult to understand. I recommend simplifying the sentences for better understanding.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The research presented in the paper entitled “Characterizing Water Composition with an Autonomous Robotic Team Employing Comprehensive In-Situ Sensing, Hyperspectral Imaging, Machine Learning, and Conformal Prediction” I find interesting and valuable. The authors carefully rethought each step of research and I see as a reader that the author's knowledge of the subject is high. The idea to use data obtained using both USV and UAV for me is very good. However, the manuscript is not written well taking into account scientific article requirements. For me, it is more a technical instruction than research with clear purposes. I find this research worth publishing but the content of the manuscript requires radical transformations. I suggest the editor allow this article for publication after major revision.

Below You can find my detailed comments.

Introduction:

1.      Too many references. In the beginning, You are presenting overall knowledge but in lines 24 and 32 You cite 5 publications. In my opinion that completely doesn’t match this simple statement and in the first paragraph (lines 22-23) two or three references are enough.

2.      There is no background of the presented research and some paragraphs e.g. lines 33-45 should be moved into the discussion section. You should completely reconstruct the Introduction firstly by adding some background (describing Your previous research, giving information about other research where UAV and USV data are compared, describing what is and how machine learning is working) and secondly by clearly indicating the main purpose of this research and secondary goals.

3.      Line 55-56 I suggest mentioning area limitations when using UAV despite e.g. high altitude aerial remote sensing.

4.      Line 64-66 I miss explaining what machine learning is, what data it requires, and what its advantages and disadvantages are.

5.      Line 67 – 69 In order to talk about the extension of previous research, it is necessary to constructively summarize what the research was about, what the conclusions were, and what the extension of this previous research will involve.

6.      Line 69-79 – it is a summary and should be placed at the end of the manuscript

Materials and methods:

1.      Line 91-92 this technical description doesn’t completely influence on research so should be removed

2.      Line 103-112 I like the way the authors describe parameters. I suggest to refer literature here.

3.      I suggest adding a photo of the USV and a similar description that Is done in Figure 1 for the UAV

4.      Figures 2 and 3: I need to praise authors for these figures that clearly and in essence explain processes.

5.      Line 166-167: I suggest that you indicate the exact daily dates here

6.      Lines 177-188: I need to appreciate the authors for paying attention to all (important) details.

Results:

1.      I don't understand why the erroneous raid of 11/23/2020 is mentioned here and why it was not carried out again, e.g. on the next day.

2.      Line 332: I suggest reconstructing the text because it is inconvenient for readers when figures appear in half of the sentence.

3.      Figures 7, 10, 13, 16: I suggest adding a scale bar.

4.      The results are well described and explained with pictures. I have no other comments, but please explain why the tests were not repeated in another season

Discussion:

This section should be completely corrected. First of all, there is no literature reference. It is like a research summary. The results should be discussed with Your previous research and also with similar research of other authors.  

Conclusion:

I suggest pointing here a few main conclusions. But generally, it is written fine. However, the conclusion in lines 453-454 doesn’t completely cover with research purpose presented in the introduction.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors' corrections significantly improved the quality and value of the article. The only thing that is missing is the study area description. It is only a condition that should be completed and this is why minor revision is needed. After this minor revision, I recommend the article entitled “Characterizing Water Composition with an Autonomous Robotic Team Employing Comprehensive In-Situ Sensing, Hyperspectral Imaging, Machine Learning, and Conformal Prediction” to be approved for publication.

The authors reorganized the text and now in the introduction give a solid research background and clearly describe what were the results of previous research and the goal of the presented research. The methodological and result sections are ordered which helps readers to understand the research. They are a bit long but when shortened they could be not as understandable as they are now. Also, the discussion section is improved and contains more references to the other research. The conclusions are really well put.

Below you can find some of my suggestions for minor corrections:

Article language: I do not feel competent to assess it but I think the English language should be improved e.g. I suppose line 32 “where” should be replaced with the word “whereas”

Line 47 – I think that there is no need to mention what kind of drones are classified as UAV Line 55 – 61 – It is worth to mention also following research: https://doi.org/10.1016/j.ecolind.2023.110103

Line 173 – the word workflow instead of pipeline would be better Line 281-292 – The Figure should be placed at the end of the paragraph, not in the middle of the sentence Line 441 – You are describing UAV and difficulties with manual data collection and then You are describing USV. I suggest adding a sentence that explains why you are starting to write about USV like: The use of USVs greatly accelerates collecting water quality data in the field.

Line 475-495 – It is worth comparing your results described in these two paragraph also compare with other research

Author Response

Please see the attachment

Author Response File: Author Response.pdf