Mapping Tree Mortality Caused by Siberian Silkmoth Outbreak Using Sentinel-2 Remote Sensing Data

Round 1

Reviewer 1 Report

The manuscript titled "Mapping Tree Mortality Caused by Siberian Silkmoth Outbreak Using Sentinel-2 Remote Sensing Data" discusses the application of remote sensing techniques to assess tree mortality resulting from a Siberian silkmoth outbreak in the Irbey region of the Krasnoyarsk Krai during the years 2018-2020. The study employs Sentinel-2/MSI satellite imagery and field measurements to create regression models and spectral indices for estimating the percentage of dead trees in this forest disturbance. The manuscript emphasizes the practical significance of such assessments for forest management decisions. In this review report, I provide a comprehensive evaluation of the manuscript, including its strengths and areas for improvement.

The manuscript addresses a critical environmental issue concerning the impact of Siberian silkmoth outbreaks on coniferous forests. The study's focus on using satellite imagery to assess tree mortality is highly relevant and offers valuable insights for forest management and ecological conservation. The methods section is well-detailed and provides a clear description of the research approach. The use of concentric circles in field studies, measurement of various forest stand elements, and the assessment of ecological parameters add credibility to the research. Additionally, the selection of Sentinel-2 imagery and spectral indices for analysis is explained thoroughly. The results section effectively presents the findings of the study. The correlation between the spectral indices and field measurements of tree mortality is highlighted, with the NBR index showing the highest correlation. The tree mortality map and the distribution of tree mortality rates are illustrated clearly in figures. These results provide a substantial basis for the study's conclusions. The discussion section provides valuable insights into the interpretation of the results. The authors appropriately compare their findings to previous research and discuss the significance of using NIR/SWIR-based spectral indices for assessing coniferous defoliators' impact. The discussion of the timing of satellite imagery and its correlation with tree mortality is particularly insightful. The manuscript is well-structured, with a logical flow from the abstract to the discussion. The figures, tables, and equations are adequately labeled and enhance the reader's understanding. The language used is technical but clear, catering to the intended scientific audience.

Nevertheless some points must be addressed. In particular.

1. In the abstract, it would be beneficial to provide a concise summary of the key findings and their implications. This will give readers a quick overview of the manuscript's main contributions.

2. The introduction could benefit from more context regarding the broader ecological implications of Siberian silkmoth outbreaks and the need for accurate tree mortality assessments.

3. While the discussion effectively highlights the choice of NIR/SWIR-based indices for assessing coniferous defoliators, it would be valuable to discuss the limitations of the study and potential avenues for future research.

4. The manuscript would be improved by providing a brief section on data limitations and potential sources of error in the methodology.

Some minor points should be checked.

Author Response

Dear Reviewer,

We are grateful for the review of our work and valuable comments. We tried to fix them, we give a detailed description below.

All corrections are highlight by green in the text.

1. In the abstract, it would be beneficial to provide a concise summary of the key findings and their implications. This will give readers a quick overview of the manuscript's main contributions.

We have added this part to the Abstract, lines 20-25.

2. The introduction could benefit from more context regarding the broader ecological implications of Siberian silkmoth outbreaks and the need for accurate tree mortality assessments.

We have added this part to the Introduction, lines 30-60, 72-75.

3. While the discussion effectively highlights the choice of NIR/SWIR-based indices for assessing coniferous defoliators, it would be valuable to discuss the limitations of the study and potential avenues for future research.
4. The manuscript would be improved by providing a brief section on data limitations and potential sources of error in the methodology.

We have added this two parts to the Discussion, lines 366-378.

Reviewer 2 Report

Dear Authors,

    The manuscript entitled "Mapping tree mortality caused by Siberian silkmoth outbreak using Sentinel-2 remote sensing data" addresses the use of remotely located data for the detection of trees attacked by forest pest (Siberian silkmoth).

     The writing of the article is adequate but minor adjustments are needed, as presented in the comments of the digital file. The presentation of the problem and justifications for the study are present. And the objective of the study has been presented.

     However, important adjustments are needed in the methodology and especially in the results. I present my suggestions to be implemented in the article before the manuscript is approved.

1) Abstract: highlight which parameters correlate with mortality and report the accuracy obtained;

2) Detail in figure 1 (general and detail) the Irbeysky District with the distribution of the plots.

3) Line 105: "missing data is restored by extrapolation.". Detail the method used. Extrapolation based on the regression model?

4) Chapter 2.2.1: the variables were selected taking into account which bibliographical references? Please cite.

5) Is the number of plots (37) sufficient for the study? What percentage was used to adjust the regression and what percentage of samples was used for validation?

6) Line 152: "The boundaries and area of the stands damaged by the Siberian silkmoth were detected by Forest cover loss map from Global Forest Watch dataset". Is the methodology used by Global Forest Watch capable of differentiating vegetation suppression from attacked areas? Please explain this process of delimiting attacked areas.

7) Please provide the data repositories for the Global Forest Watch and Hansen GFW datasets.

8) Line 167: Was the Copernicus repository used for downloading?

9) Line 174 to 184: according to ESA:

Level-1C processing includes radiometric and geometric corrections including ortho-rectification, add of the radiometric offset and spatial registration on a global reference system with sub-pixel accuracy. This is reflectance at the top of the atmosphere (TOA).

The Level-2A processing includes a Scene Classification and an Atmospheric Correction applied to Top-Of-Atmosphere (TOA) Level-1C orthoimage productsLevel-2A main output is an orthoimage atmospherically corrected, Surface Reflectance product.

Please be aware that "Surface Reflectance (SR)" is a new term that has been introduced to replace the former one: "Bottom of Atmosphere (BOA) reflectance."

Note that the images already have reflectance values. In the case of level 1C images, it is advisable to apply the correction with the Sen2cor algorithm, which is the appropriate application for sentinel images. Formulas 1 and 2 transform the DN values in the JPG format images, which are in integers, into reflectance values. For QUANTIFICATION_VALUEi is equal to 10000;

10) Was it necessary to resample the SWIR bands in order to calculate the indices? Please detail the resampling technique used.

11) Present the hardware and software resources used in the study;

12) Table 3: one of the great differentials of the MSI sensor is the four red edge bands. These bands were specially developed for the study of plant health. You didn't use any indices based on the red edge bands, let alone indices such as FAPAR. Please justify not using the red edge bands. NDVI has well-known saturation problems in areas with denser forest cover.

13) Line 210: is linear regression sufficient to explain the variable?

14) In the legend to Figure 5, show the areas (ha) for each class.

15) Line 267: "So, indices related to vegetation water content is better for assessing the impact of coniferous defoliators on forest stands than that related to canopy chlorophyll content and structure." Swir has a problem with the background. Please review this statement. Does the study carried out allow this statement to be made?

16) With regard to the background, the indices used usually have problems with greater soil exposure. Is this not affecting the results of your study?

17) To be statistically significant, it would be necessary to validate the map shown in figure 5 by cross-referencing it with some samples not used in the regression adjustment. Without this, it is not possible to assess the effectiveness of the study.

18) An interesting alternative would be to calculate FAPAR (fraction of photosynthetically active radiation absorbed by the canopy), LAI, Canopy chlorophyll content (CCC) and Canopy water content (CWC).

Best regards,

Comments for author File: Comments.pdf

Carefully review the use of punctuation, especially commas. Pay attention to the use of long paragraphs as they tend to confuse the reader. Check the agreement of verbs. When acronyms are presented for the first time in the text, their meaning should be given.

Author Response

Dear Reviewer, we suggest you to get acquainted with our corrections in the attached file

Author Response File: Author Response.pdf

Reviewer 3 Report

The purpose of the paper I am reviewing is to study the possibilities of using modern satellite imagery of medium spatial resolution to estimate the percentage of dead trees.

Authors proposed tu use imagery from the Sentinel-2/MSI sensor was used to calculate a number of spectral indices for images received before and after the Siberian silkmoth atack. Remote forest disturbances assessment performed after Siberian silkmoth a, makes it possible to assess the potential for natural regeneration, estimate forest fire danger and reveal the need to implement forest management practices. The need to conduct such research is justified and interesting from a scientific point of view.

In my opinion, the most important thing should be the methodology of the research, which should indicate the innovativeness of the proposed method, algorithms and action plan. This is definitely what's missing at paper.

Chapter 2.4. Methods in its description: line 186-206 contains facts and characteristics of images that have their sources in the literature provided in this chapter.

Lines 208-216 could be called a short fragment of the methodology. But in your opinion this is not enough.

This chapter should be carefully prepared.

Then, in accordance with the methodology, the research results should be presented.

Author Response

Dear Reviewer,

We are grateful for the review of our work and valuable comments. We tried to fix them, we give a more detailed description below.

Chapter 2.4. Methods in its description: line 186-206 contains facts and characteristics of images that have their sources in the literature provided in this chapter.

We have moved this part of the text to the Introduction (to lines 83-102).

Lines 208-216 could be called a short fragment of the methodology. But in your opinion this is not enough.

This chapter should be carefully prepared.

In my opinion, the most important thing should be the methodology of the research, which should indicate the innovativeness of the proposed method, algorithms and action plan. This is definitely what's missing at paper.

Then, in accordance with the methodology, the research results should be presented.

We have expanded the Methods and Results chapters by adding a more detailed description, all corrections are highlight by green in these parts of paper.

Round 2

Reviewer 2 Report

Dear Authors,

  After a thorough reading of the second version of the manuscript "Mapping tree mortality caused by Siberian silkmoth outbreak using Sentinel-2 remote sensing data", I realised that you have made significant changes to the text. The wording has been improved and important information has been added to the methodology and results.

   Most of my questions have been answered. However, I would ask for your attention in the final version of the document:

     1) )Is the number of plots (37) sufficient for the study? What percentage was used to adjust the regression and what percentage of samples was used for validation? You responded by justifying that the number and distribution of samples was largely due to the difficulty of accessing the infestation sites. It would be interesting for you to add the text of the answer in the final version: "А relatively low amount of measurements is related to the extreme difficulty of collecting data in the field. There are no roads that can be travelled by car in summer that reach the edge of the outbreak, so it took about 5 hours to walk to the outbreak and back every day. There is a significant fall of dead trees in the outbreak territory, so travelling there is extremely difficult, traumatic and takes a long time. Another reason is security issues. This place is quite far from settlements and is a habitat for wild animals, including bears"

     2) ) Was it necessary to resample the SWIR bands to calculate the indices? Please describe in detail the resampling technique used.  In this case, did you perform map algebra in Arcmap? If so, it automatically resamples from 20 to 10 metres using the Nearest Neighbours algorithm.

     3)Table 3: one of the great differentials of the MSI sensor is the four red edge bands. These bands were specially developed to study plant health. You haven't used any indices based on the red edge bands, let alone indices such as FAPAR. Please justify not using the red edge bands. NDVI has well-known saturation problems in areas with denser forest cover. Please consider presenting this idea as a recommendation for future studies.

      4) An interesting alternative would be to calculate FAPAR (fraction of photosynthetically active radiation absorbed by the canopy), LAI, canopy chlorophyll content (CCC) and canopy water content (CWC). Please consider presenting this as a recommendation for future studies.

       I conclude my review and thank you for sending the cover letter.

Sincerely,

Few adjustments to the score are needed. Please proofread the final version thoroughly.

Author Response

Dear Reviewer,

We cordially thank you for your valuable comments ones again.
We also believe that working on the previous comments helped us improve the manuscript.

1)Is the number of plots (37) sufficient for the study? What percentage was used to adjust the regression and what percentage of samples was used for validation?  You responded by justifying that the number and distribution of samples was largely due to the difficulty of accessing the infestation sites. It would be interesting for you to add the text of the answer in the final version: "А relatively low amount of measurements is related to the extreme difficulty of collecting data in the field. There are no roads that can be travelled by car in summer that reach the edge of the outbreak, so it took about 5 hours to walk to the outbreak and back every day. There is a significant fall of dead trees in the outbreak territory, so travelling there is extremely difficult, traumatic and takes a long time. Another reason is security issues. This place is quite far from settlements and is a habitat for wild animals, including bears".

We have added a slightly shortened text in lines 195-199.

2)Was it necessary to resample the SWIR bands to calculate the indices? Please describe in detail the resampling technique used.  In this case, did you perform map algebra in Arcmap? If so, it automatically resamples from 20 to 10 meters using the Nearest Neighbors algorithm.

Yes, we used the Arcmap map algebra. Thank you for this explanation.

3)Table 3: one of the great differentials of the MSI sensor is the four red edge bands. These bands were specially developed to study plant health. You haven't used any indices based on the red edge bands, let alone indices such as FAPAR. Please justify not using the red edge bands. NDVI has well-known saturation problems in areas with denser forest cover. Please consider presenting this idea as a recommendation for future studies.

We added this recommendation in lines 376-381 of Discussion.

4) An interesting alternative would be to calculate FAPAR (fraction of photosynthetically active radiation absorbed by the canopy), LAI, canopy chlorophyll content (CCC) and canopy water content (CWC). Please consider presenting this as a recommendation for future studies.

We added this recommendation in lines 376-381 of Discussion.

Reviewer 3 Report

The authors responded correctly to the comments. All have been included.

Author Response

Dear Reviewer,

We cordially thank you for your valuable comments ones again.
We also believe that working on the previous comments helped us improve the manuscript.