Exploring the Potential of UAV-Based Hyperspectral Imagery on Pine Wilt Disease Detection: Influence of Spatio-Temporal Scales

Round 1

Reviewer 1 Report

Dear Authors,

All in all, the paper is very interesting and describe a current problem with state-of-the-art techniques and methods. However, it lacks in some conceptual parts, see below:

The large change in the significance values by Kruskal-Wallis tests involving coarse resolution (mostly 12 cm resolution data and in some cases the 10 cm version also) versions is very interesting. Please discuss it with more detail what can be the reason for that.

It would be helpful to include a graph about Kappa-values as well in the same style like with Accuracy values in Figure 5.

Conclusions should be more concise, please edit it. Current text between L445 and L460 should be in the Discussion.

 

Some specific mistakes or typos:

L20-21 The last sentence of the abstract is questionable, it would sound more scientifical if you change it to something like "This study can be useful/important to improve the efficiency of...".

L81 resampling

L93-94 Is it correct that the infection spans only a few days? Edit please if needed.

Figure 1. On the close-up image of sample area it would be helpful to indicate the 8 plots mentioned in the text below the image.

L141 we

L160 Please give more information about the physical and chemical characteristics of the calibration panel, is it a Spectralon panel or regular plastic? Secondly, square meter or change "2" to upper index please.

L161 Russia

L165 The corresponding... and add extra space between values and their units

L173 "... was used to convert the units."

L175 ENVI reference is Harris Geospatial, USA or United States of America, please edit the text.

L186 Same as in the previous for the country.

L205 Same as in the previous for the country.

L210 What is the scientific reason to set p-value at 0.05? Please add reference to explain your decision.

Table 4-5 should be repositioned in the text closer to Table 6 and corresponding explanation presented in the 3.3 subchapter (which should be in italic as all other subchapter title)

Tables 7-8-9 Please change the SP values from flying height to resolution to be uniform along the paper and with the text.

Table 7. Please change in the bottom part "Overall accuracy" to "OA (%)" for uniformity.

Table 8. Extra 0 in 250 m version's OA value, delete

Figure 5. Please include a higher resolution image.

Figure 6. modify it please for higher resolution, it is nearly unreadable in its current form 

L416-417 What is your reason to claim this? Please support it with explanation and examples.

L418-421 Random Forest is not deep-learning however it is still a machine learning algorithm so the statement presented here also true for RF as well. If it is already proved that DL can also work, please give some details about your decision to choose RF instead of DL in the paragraph starting from L219.

L438 Please add numbered references also for mentioned papers.

L462-463 Time will decide...

Author Response

1) The large change in the significance values by Kruskal-Wallis tests involving coarse resolution (mostly 12 cm resolution data and in some cases the 10 cm version also) versions is very interesting. Please discuss it with more detail what can be the reason for that.

Response: Thank you very much for appreciating our work and sending us a bunch of comments, which are quite helpful for us to improve the manuscript. The Coarse resolution increases the probability of misclassification, which also affects the significance values by Kruskal-Wallis tests.

2)It would be helpful to include a graph about Kappa-values as well in the same style like with Accuracy values in Figure 5.

Response:Thanks for your suggestion. We have added the Figure 6 about Kappa- values.

3)Conclusions should be more concise, please edit it. Current text between L445 and L460 should be in the Discussion.

Response:Thanks for your suggestion. We have concised the conclusions and adjust the text between L445 and L460.

Some specific mistakes or typos:

L20-21 The last sentence of the abstract is questionable, it would sound more scientifical if you change it to something like "This study can be useful/important to improve the efficiency of..."

Response:Thanks for your suggestion. We have rewritten the sentence.

L81 resampling

Response:Thanks for your suggestion. We have rewritten the word.

L93-94 Is it correct that the infection spans only a few days? Edit please if needed.

Response:Thanks for pointing this. We have changed the July to October.

Figure 1. On the close-up image of sample area it would be helpful to indicate the 8 plots mentioned in the text below the image.

Response:Thanks for your suggestion. We have added the 8 plots in Figure 1.

L141 we

Response:Thanks for pointing this. We have changed the We to we.

L160 Please give more information about the physical and chemical characteristics of the calibration panel, is it a Spectralon panel or regular plastic? Secondly, square meter or change "2" to upper index please.

Response:Thanks for your suggestion. The calibration panel was made by Barium sulfate.

L161 Russia

Response:Thanks for your suggestion. We have changed the Russian to Russia.

L165 The corresponding... and add extra space between values and their units

Response:Thanks for your suggestion. We have added the space.

L173 "... was used to convert the units."

Response:Thanks for your suggestion. We have rewritten the sentence.

L175 ENVI reference is Harris Geospatial, USA or United States of America, please edit the text.

Response:Thanks for your suggestion. We have rewritten the reference.

L186 Same as in the previous for the country.

L205 Same as in the previous for the country.

Response:Thanks for your suggestion. We have changed the country to USA.

L210 What is the scientific reason to set p-value at 0.05? Please add reference to explain your decision.0.05

Response:Thanks for pointing this. We have added the reference.

Table 4-5 should be repositioned in the text closer to Table 6 and corresponding explanation presented in the 3.3 subchapter (which should be in italic as all other subchapter title)

Response:Thanks for pointing this. We have changed the position of Table 4-5. And the title of 3.3 was adjusted.

Tables 7-8-9 Please change the SP values from flying height to resolution to be uniform along the paper and with the text.

Response:Thanks for your suggestion. We have changed the flying height to SP.

Table 7. Please change in the bottom part "Overall accuracy" to "OA (%)" for uniformity.

Response:Thanks for your suggestion. The Overall accuracy here was explain the OA.

Table 8. Extra 0 in 250 m version's OA value, delete

Response:Thanks for pointing this. We have deleted the “0”.

Figure 5. Please include a higher resolution image.

Response:Thanks for your suggestion. We have inporved the resolution of Figure 5.

Figure 6. modify it please for higher resolution, it is nearly unreadable in its current form 

Response:Thanks for your suggestion. We have inporved the resolution of Figure 6.

L416-417 What is your reason to claim this? Please support it with explanation and examples.

Response:Thanks for pointing this. We have deleted this sentence.

L418-421 Random Forest is not deep-learning however it is still a machine learning algorithm so the statement presented here also true for RF as well. If it is already proved that DL can also work, please give some details about your decision to choose RF instead of DL in the paragraph starting from L219.

Response:Thanks for your suggestion. We are studying the application of DL in PWD detection, which will be described in our another paper.

L438 Please add numbered references also for mentioned papers.

Response:Thanks for your suggestion. We have added the reference.

Reviewer 2 Report

This manuscript examines the impact of spatiotemporal scales on PWD detection in a monoculture Masson pine plantation from UAV-based hyperspectral imaging. This is an interesting study but the following concerns should be addressed before I recommend it for publication.

 

1) In the introductory section, more details are needed on the impact of pine nematode disease on coniferous forests in China. Also, it seems like there is not much literature on recent advances in vegetation disease.

2)  many scholars have already done similar modelling experiments, and this study did this work using an RF model which is not very new. It would be interesting to see the authors investigate The optimal vegetation index of tree ecology with the PWD and spectral response changes.

3) In the methods section, a more detailed description of the UAV hyperspectral imaging techniques used and an explanation of how the data were processed and analyzed is needed.

4) Figure 5 and 6 are difficult to read and please make sure all figures are produced in high quality.

5) In subsection 3.4 “user's accuracy (AU)” is expressed differently before and after, please clarify whether it is AU or UA.

Author Response

1) In the introductory section, more details are needed on the impact of pine nematode disease on coniferous forests in China. Also, it seems like there is not much literature on recent advances in vegetation disease.

Response: Thank you very much for appreciating our work and sending us a bunch of comments, which are quite helpful for us to improve the manuscript. We have added the literature about recent advances in vegetation disease. Please check the details of modification in revised manuscript.

2) many scholars have already done similar modelling experiments, and this study did this work using an RF model which is not very new. It would be interesting to see the authors investigate The optimal vegetation index of tree ecology with the PWD and spectral response changes.

Response: Thank you very much for appreciating our work. This paper focuses on the influence of spatial resolution on PWD detection, and the modelling experiment is not the key research part. So, we did the work by RF model, which is one of the most commonly used model. In our other studies, we focus on the spectral response to PWD and the determination of optimal vegetation index.

3)In the methods section, a more detailed description of the UAV hyperspectral imaging techniques used and an explanation of how the data were processed and analyzed is needed.

Response: Thanks for your suggestion. We have added the information about the UAV hyperspectral imaging techniques and the data process. Please check the details of modification in revised manuscript.

4）Figure 5 and 6 are difficult to read and please make sure all figures are produced in high quality.

Response:Thanks for your suggestion. We have improved the quality of Figure 5 and 6.

5）In subsection 3.4 “user's accuracy (AU)” is expressed differently before and after, please clarify whether it is AU or UA.

Response: Thanks for pointing this. The“user's accuracy (AU)” is UA.

Reviewer 3 Report

Lines 92-93: "from the end of June to the beginning of July each year" doesn't seem right. Please check or reformulate.

Line 138: "15:00 p.m." is "3:00 p.m."?

Line 141: is it "We" - should it be lowercase?

Line 177: it is known what "std dev" is but it is not usually written in this form in the text (if it is not explained before)

Author Response

Lines 92-93: "from the end of June to the beginning of July each year" doesn't seem right. Please check or reformulate.

Response:Thank you very much for appreciating our work and sending us a bunch of comments, which are quite helpful for us to improve the manuscript. We have changed the July to October.

Line 138: "15:00 p.m." is "3:00 p.m."?

Response:Thanks for pointing this. We have changed the 15 to 3.

Line 141: is it "We" - should it be lowercase?

Response:Thanks for pointing this. We have changed the We to we.

Line 177: it is known what "std dev" is but it is not usually written in this form in the text (if it is not explained before)

Response:Thanks for your suggestion. We have changed the std dev to Standard deviation.

Reviewer 4 Report

The reviewed article contains an original PWD detection methodology. The topic of the paper is highly topical, especially in countries that are directly affected by PWD. A similar topic was also the subject of the article "Run Yu, Langning Huo, Huaguo Huang, Yuan Yuan, Bingtao Gao, Yujie Liu, Linfeng Yu, Haonan Li, Liyuan Yang, Lili Ren, and Youqing Luo: Early detection of pine wilt disease tree candidates using time-series of spectral signatures. Queue. Plant Sci. 13:1000093.doi: 10.3389/fpls.2022.1000093" (Run Yu) in Attachment of Review.

Unlike the reviewed article, it does not contain research on the impact of spatial scale on PWD detection.

The article under review lacks information on the extent to which it can contribute to earlier detection of PWD compared to the procedures and methods known so far. The contribution of the paper would be to compare the results of the research with the results of the mentioned article "Run Yu". It would also be interesting if the authors made a qualified estimate of time savings, costs and saved losses of wood if their research results were translated into practice.

 Why did the authors not use LiDAR system in the presented methodology, which could have been part of the UAV. On the basis of published research results from the subject area, a combination of hyperspectral data and LiDAR system  data can achieve higher accuracy of the achieved results.

 

 

 

 

Comments for author File: Comments.pdf

Author Response

The reviewed article contains an original PWD detection methodology. The topic of the paper is highly topical, especially in countries that are directly affected by PWD. A similar topic was also the subject of the article "Run Yu, Langning Huo, Huaguo Huang, Yuan Yuan, Bingtao Gao, Yujie Liu, Linfeng Yu, Haonan Li, Liyuan Yang, Lili Ren, and Youqing Luo: Early detection of pine wilt disease tree candidates using time-series of spectral signatures. Queue. Plant Sci. 13:1000093.doi: 10.3389/fpls.2022.1000093" (Run Yu) in Attachment of Review.

Unlike the reviewed article, it does not contain research on the impact of spatial scale on PWD detection.

The article under review lacks information on the extent to which it can contribute to earlier detection of PWD compared to the procedures and methods known so far. The contribution of the paper would be to compare the results of the research with the results of the mentioned article "Run Yu". It would also be interesting if the authors made a qualified estimate of time savings, costs and saved losses of wood if their research results were translated into practice.

Response: Thank you very much for appreciating our work and sending us a bunch of comments, which are quite helpful for us to improve the manuscript. We are currently conducting predictive studies that can be effectively utilized in PWD surveillance.

 Why did the authors not use LiDAR system in the presented methodology, which could have been part of the UAV. On the basis of published research results from the subject area, a combination of hyperspectral data and LiDAR system  data can achieve higher accuracy of the achieved results.

Response: Thanks for pointing this. LiDAR mainly provides morphological parameters. Considering that the morphological changes of pine trees are very small in the early stage of PWD occurrence, spectral feature changes are more important for PWD monitoring. Therefore, this study did not incorporate LiDAR data. In further study, we will consider combining LiDAR with hyperspectral data for PWD monitoring in a multi-layered structure of pine trees.