Analysis of Spectral Index Interrelationships for Vegetation Condition Assessment on the Example of Wetlands in Volyn Polissya, Ukraine

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I do like this paper and it is a useful and novel approach to peatland monitoring. The models are clearly presented and the whole document reads well.

I think the paper would benefit from a section at the end of the Discussion or in the Conclusions, to provide a plain English summary of what this means for the practitioner on the ground working to conserve the peat bogs and related landscape. At present this is written by specialists to be read by specialists with knowledge of remote sensing and modelling. It deserves a wider readership.

Minor points are  - please don't start a sentence with a number - e.g., line 158 '57.......' Write as 'Fifty-seven .....'  

At the end of the Discussion please provide a short note to consider any shortcomings of the applied approach.

 

Author Response

Comments 1: I think the paper would benefit from a section at the end of the Discussion or in the Conclusions, to provide a plain English summary of what this means for the practitioner on the ground working to conserve the peat bogs and related landscape. At present this is written by specialists to be read by specialists with knowledge of remote sensing and modelling. It deserves a wider readership.

Response 1: We agree with this comment. We tried to restructure the Discussion and Recommendations sections to make it more understandable for the general public

Comments 2: Minor points are  - please don't start a sentence with a number - e.g., line 158 '57.......' Write as 'Fifty-seven .....'

Response 2: Agree.

Comments 3: At the end of the Discussion please provide a short note to consider any shortcomings of the applied approach.

Response 2: Agree. We tried to fix this.

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

The manuscript describes in an accessible way both the method of conducting and the results of multi-index analyses, and demonstrates their usefulness in monitoring changes occurring in naturally valuable areas and in planning strategies for their protection in the future.

My comments concern several issues that, in my opinion, could improve the manuscript.

1. I would suggest dividing section 2.1 Study region and datasets into two separate parts, mainly because in a uniform text there is a sudden change of topic and it is inconsistent.
2. It seems to me that the Methodology subsection should also be removed and the information contained in it should be combined with the next section (lines 194-196). The scheme (Fig. 2) can be found at the end of section 2.1 or, if it is divided - 2.2.
3. I believe that the Authors should combine the two chapters Results and Discussion into one. In its current form, the Discussion chapter is, in my opinion, a summary of the results. There is no confrontation with the research results or opinions of other scientists, there are no citations (which can be found e.g. in the Results chapter: e.g. line 414). The discussion must support the possibility of drawing conclusions and these cannot be conclusions based solely on the Authors' own observations. It seems to me that some of the text in 2.2.1 Vegetation Indexes could be used in the Discussion. In this part, the authors describe all the indexes and their possible uses in great detail. In the Discussion, the Authors should indicate examples of research results known from the literature or examples of the use of such indexes by other scientists.
4. I think that especially the part of the text from line 534 could be presented as Recommendations

I hope that once the appropriate corrections have been made, the article will be made available to a wider audience.

Author Response

Comments 1: I would suggest dividing section 2.1 Study region and datasets into two separate parts, mainly because in a uniform text there is a sudden change of topic and it is inconsistent.

Response 1: We agree with this comment. Study region and datasets is divided to separate subparts.

Comments 2: It seems to me that the Methodology subsection should also be removed and the information contained in it should be combined with the next section (lines 194-196). The scheme (Fig. 2) can be found at the end of section 2.1 or, if it is divided - 2.2.

Response 2: Agree. We decided to leave the methodology section as a general part of our methodology, but your opinion on placing a figure summarizing the methodology proposed in the article at the end of the section is reasonable.

Comments 3: believe that the Authors should combine the two chapters Results and Discussion into one. In its current form, the Discussion chapter is, in my opinion, a summary of the results. There is no confrontation with the research results or opinions of other scientists, there are no citations (which can be found e.g. in the Results chapter: e.g. line 414). The discussion must support the possibility of drawing conclusions and these cannot be conclusions based solely on the Authors' own observations. It seems to me that some of the text in 2.2.1 Vegetation Indexes could be used in the Discussion. In this part, the authors describe all the indexes and their possible uses in great detail. In the Discussion, the Authors should indicate examples of research results known from the literature or examples of the use of such indexes by other scientists.

Response 3: Agree. We tried to fix this.

Comments 4: I think that especially the part of the text from line 534 could be presented as Recommendations

Response 4: Agree. We have added the Recommendations section

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents an analysis of temporal and spatial trends in vegetation indices derived from remote imagery of the Cheremsky Nature Reserve in Ukraine.  This is a pristine nature reserve and is quite important for its wetlands and biodiversity.  The authors analyzed 8 years of SAR imagery and computed NDVI, EVI, SAVI, MSAVI, GNDVI, NDRE, and NDWI.

159:  I don’t understand why the pod grass communities could be the reason the Cheremsky marsh was not drained. 

The study aims to select the most suitable combination of vegetation spectral indices obtained from cloud=free, harmonized Sentinel-2 surface reflectance data in the visible to short-wave infrared range.  What is the band width? 

Results in Fig 3 do not demonstrate convincing trends.  The NDVI, SAVI, and GNDVI indicates are almost identical in pattern, and if you removed 2022 and 2023 I doubt that the trends would be negative.  The positive trend NDWI is almost the mirror image of the NDVI, SAVI, and GNDVI images, but it also would have no trend if the 2022 and 2023 data were removed.  NDWI is calculated like the NDVI but with the substitution of SWIR for Red reflectance. The most interesting question to me is what happened in 2022 and 23?  Is this real or some artifact of the data?   Were these drought years?  Since NDWI is used to assess the vegetation’s water content, as the authors state: This requires further verification and research. 

Beyond the vegetative indices, I was confused by how these variables were used in the PCA and K-Means clustering.  Fig 6 is informative.   I think the Kmeans are used to inform the optimal number of clusters.  Then Fig 6 must be a map of the different clusters across space, but is is really difficult to see the difference between the 2 green colors.  Is the difference in color meant to show the distance between clusters?  Is that correct?  This could all use more clarification.  Expand the figure legends to concisely include information about how to interpret the figure.  Is Fig 6 a map of the means of the 3 PCA clusters?  Does it change through time? 

Can you provide a bit more explanation about what is mean by: Permutation Importance, also known as "Mean Decrease Accuracy" in the context of Random Forest, is an alternative method of evaluating feature importance that tries to circumvent some of the limitations of embedded importance [41,42].”  PFI is a method to determine which machine learning model features contribute the most to predictions.  Fine, but what predictions are you explaining?  

429: What is meant by ‘the NDVI index that takes into account vegetation’?  NDVI is a vegetation index.  I think this needs to be reworded.

I though the discussion was quite interesting and made a good case for the importance of the methodology. 

Author Response

Comments 1:. 159:  I don’t understand why the pod grass communities could be the reason the Cheremsky marsh was not drained.

Response 1: Agree. For a detailed understanding of the research object, an extended description has been added.

Comments 2: The study aims to select the most suitable combination of vegetation spectral indices obtained from cloud=free, harmonized Sentinel-2 surface reflectance data in the visible to short-wave infrared range.  What is the band width?

Response 2: Agree. We have supplemented Table 2.

Comments 3: Results in Fig 3 do not demonstrate convincing trends.  The NDVI, SAVI, and GNDVI indicates are almost identical in pattern, and if you removed 2022 and 2023 I doubt that the trends would be negative.  The positive trend NDWI is almost the mirror image of the NDVI, SAVI, and GNDVI images, but it also would have no trend if the 2022 and 2023 data were removed.  NDWI is calculated like the NDVI but with the substitution of SWIR for Red reflectance. The most interesting question to me is what happened in 2022 and 23?  Is this real or some artifact of the data?   Were these drought years?  Since NDWI is used to assess the vegetation’s water content, as the authors state: This requires further verification and research.

Response 3: Agree. We have added additional explanations in the text.

Comments 4: Beyond the vegetative indices, I was confused by how these variables were used in the PCA and K-Means clustering.  Fig 6 is informative.   I think the Kmeans are used to inform the optimal number of clusters.  Then Fig 6 must be a map of the different clusters across space, but is is really difficult to see the difference between the 2 green colors.  Is the difference in color meant to show the distance between clusters?  Is that correct?  This could all use more clarification.  Expand the figure legends to concisely include information about how to interpret the figure.  Is Fig 6 a map of the means of the 3 PCA clusters?  Does it change through time?

Response 4: Agree. We have updated Figure 6 to add the classification for 2022 and a graph of the distribution of points between classes by year

Comments 5: Can you provide a bit more explanation about what is mean by: Permutation Importance, also known as "Mean Decrease Accuracy" in the context of Random Forest, is an alternative method of evaluating feature importance that tries to circumvent some of the limitations of embedded importance [41,42].”  PFI is a method to determine which machine learning model features contribute the most to predictions.  Fine, but what predictions are you explaining?

Response 5: Agree. We have added additional explanations in the text.

Comments 6: 429: What is meant by ‘the NDVI index that takes into account vegetation’?  NDVI is a vegetation index.  I think this needs to be reworded.

Response 6: Agree, the sentence is rewritten

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

This is a nice paper and is ready to go.