Mapping the National Seagrass Extent in Seychelles Using PlanetScope NICFI Data

Round 1

Reviewer 1 Report

Seagrasses sequestrate large amounts of CO2. Seagrass mapping is crucial in order to reduce  their loss. This study is aimed at mapping seagrasses within the exclusive marine economic zone of Seychelles. This was carried out using the NICFI Basemap composite satellite images with a high resolution of approximately 5 m.

The novelty, significance of content, and practical value of the paper are mentioned and proven by the authors. The research design is appropriate, and the method is adequately described. The results are clearly presented, and possible limitations of this study are discussed in detail.

I recommend accepting this manuscript in its current form.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Overall Assessment

This paper used the NICFI basemaps as a data source and a random forest algorithm in GEE platform to extract the extent of seagrass meadows in the Seychelles region. The authors have mapped seagrass distribution in the Seychelles region for the first time using NICFI data. The study is very interesting and important to protect seagrass. However, in this study, the methodology is not innovative enough, the classification model is very old and the methods. The result section is very simple. So, I do not consider this paper to be of a standard appropriate for the journal of “Remote sensing”.

1. Introduction

Point 1: Page 2, Line 48-50 This paragraph is proposed to be merged with the previous paragraph.

Point 2: Page 2, Line 66-73 This paragraph lacks a detailed description of the literature on GEE-based seagrass meadow classification, such as the magnitude of classification accuracy.

Point 3: Page 3, Line 100-102 “Thus, this paper presents the first study…” This paper seems to me to have done only one study, and "first" is poorly worded.

Materials and Methods

Point 4: Page 3, 109-110 “Previous studies have mapped seagrasses in these shallow waters” What are the similarities and differences between the distribution maps of seagrass meadows produced by previous studies and those produced by the authors.

Point 5: Page 7, 238-243 The use of only one classification algorithm is not convincing and it is recommended that additional classification algorithms be added.

Results

Point 6: Page 7, Line253-254 “In the two other regions, the producer’s accuracy of the seagrass class is greater than the user’s accuracy by at least 5%” What is the significance of producer’s accuracy being higher than the user’s accuracy.

Discussion

Point 7: The discussion was overloaded with narratives about the deficient parts of this article, which overshadowed the innovative parts of this paper.

 The  English of this manuscript is very good, and only need minor editing the grammar,spelling and sentence structure.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper introduces an application of Random Forest using NICFI Basemap data. The analyses made are interesting; however, certain aspects within the introduction and methodology sections need to be aligned for a better comprehension of the article's significance and the methodology employed. For instance, in the methodology section, it is necessary to distinctly outline the purpose of each dataset and analysis conducted. Here are some suggestions and recommendations:

1.      The article's introduction is comprehensive and addresses key points, including knowledge gaps. However, there appears to be some overlap between certain details that would be better suited for the study area subsection. Additionally, it would be important to underscore the importance of the NICFI Basemap, highlighting its relevance and potential applicability, considering the aims of the paper.

2.      Figure 2 would be more appropriately positioned within the study area subsection.

3.      Compiling a table summarizing the attributes of the NICFI images alongside other images used in the analysis could provide valuable context.

4.      Streamlining the methodology structure by consolidating the description of the used database in one subsection and the analysis techniques in another would enhance clarity.

5.      Please include all relevant equations, including those for spectral indices, in the text.

6.      Consider making the GEE script available as supplementary material of the paper.

7.      Please, incorporate spectral indices boxplots in Figure 4.

8.      Was there a classification test conducted by excluding bands and indices with significant variation as indicated in the boxplot? It would be interesting to see at least the difference between using or not the spectral indices in the random forest analyses.

9.      Please provide a more comprehensive explanation of the Random Forest classification method and accuracy evaluation, including the specific formulas used.

10.  Are the achieved accuracy results consistent with the anticipated outcome? Could a comparative analysis be undertaken by contrasting them with a classification process involving Sentinel-2 imagery?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

All my questions and comments, the authors both provided the careful responses. I recommend to accept the paper on current condition.

The quality of English language is very well in the revision.

Reviewer 3 Report

The authors have responded comprehensively to all the queries posed. Moreover, the supplementary results presented have enriched the work, making it both more captivating and transparent.