Early Detection of Southern Pine Beetle Attack by UAV-Collected Multispectral Imagery

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article considers the possibility of remote detection of SPB brood on affected trees every two weeks in the summer of 2022 and linked the stage of brood with visible symptoms of tree decline. DJI P4 MS images at the following wavelengths were used for research: blue (B): 450 nm ± 16 nm; green (G): 560 nm ± 16 nm; red (R): 650 nm ± 16 nm; red edge (RE): 730 nm ± 16 nm; and near-infrared (NIR): 840 nm ± 26 nm.

As a result of the research, it has been shown that the accuracy of detection is about 70-80 percent. The results of ground-based studies confirm the need for early detection of attacks, and the results of UAVs show that multispectral images can detect an early SPB attack.
A significant part of pine forests is located in Siberia. However, the introduction lacks information about similar studies in Siberian forests. This gap needs to be filled. Accordingly, it is necessary to revise the list of cited literature.

In general, the article is typical for this type of research. There is nothing particularly new in the article. The article can be published after the specified additions in the introduction and bibliography.

Author Response

Comment 1: "A significant part of pine forests is located in Siberia. However, the introduction lacks information about similar studies in Siberian forests. This gap needs to be filled. Accordingly, it is necessary to revise the list of cited literature."

Response 1: Thank you for pointing out this gap in our literature review. We aim to include remote sensing studies of early attack of tree-killing bark beetles, particularly in the genus Dendroctonus. However, much work has been done in early attack of Ips typographus in Norway spruce, so we choose to include reference to the body of work done in this particular system as well as those dealing with genus Dendroctonus. Your suggestion prompted us to conduct another thorough review of the bark beetle early attack / remote sensing literature. We found that we  previously overlooked Safonova et al. 2022, which investigated early attack of Ips typographus in Norway spruce in Bulgaria. We now include this reference in our introduction and in Supplemental Table S1. We did not find any studies of remote sensing and early attack of bark beetles in Siberia. 

Reviewer 2 Report

Comments and Suggestions for Authors

All comments will be found in the yellow comment box next to the text.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

I hope to re-do the linguistic proofreading, pay attention to the vocabulary of academic writing, use the passive voice in writing, and do not use the personal pronoun.

Author Response

Comment 1: "In academic writing, use the passive voice."

Response 1: We appreciate your commitment to the quality of scientific writing. However, we purposefully use active voice throughout our manuscript to increase the clarity of our work. We argue that the use of active voice is acceptable, and even preferable, to most readers.

Comment 2: (Line 222) "Why not at 100 or 50 height?"

Response 2: Operation of UAVs for research use in the United States is bounded by Federal Aviation Administration (FAA) rules. The maximum height we can fly is 120m according to FAA Part 107 rules. This height provides us with better line of sight above the tree line at distance in the forested landscape. Line of sight from pilot to UAV is required under FAA Part 107. We now state "flying at the maximum permitted altitude of 120 m".

Comment 3: (Line 232) "Please add the specifications of the computer used."

Response 3: We have now added the specifications at the end of section 2.4: "We conducted all analyses on a Intel Xeon E5-2637 v4 CPU with 256 GB of RAM and a NVIDIA Quadro P5000 GPU with 16 GB VRAM."

Response 4: (Line 267) "You did not need to appendix, as it is just one table."

Response 4: We chose to situate this table in an appendix since the table takes up a whole page, and could disrupt the flow of the main text.  In the Remote Sensing manuscript template, the description of the appendix section is as follows: "The appendix is an optional section that can contain details and data supplemental to the main text—for example, explanations of experimental details that would disrupt the flow of the main text but nonetheless remain crucial to understanding and reproducing the research shown; figures of replicates for experiments of which representative data is shown in the main text can be added here if brief, or as Supplementary data."

Comment 5: (Line 279) "It would be better for the results if the number of training samples was more than the test samples."

Response 5: We understand that sample size is an important feature of random forest analysis. As seen in Table 2, our sample size for our class of most interest, SPB green attack, is limited. Thus, we chose to use a 50/50 training/testing split so that even the least represented class would contain a large enough number of samples for the analysis. We did test 45/55 and 55/45 splits, and this did not impact our results.

Comment 6: (Lines 279-281) "Please explain more."

Response 6: Thank you for directing us to this lack of clarity. We are now more explicit about the 4 different sets of features that we tested in the feature tuning process. We have edited the sentence to read: "We tested four different sets of features: (a) using all 52 features, using the (b) top 15 and (c) top 40 features with the highest mean decrease in Gini index [71] resulting from the algorithm that included all 52 features, and (d) and only including features derived from RGB bands (red, green, and blue bands; effectively removing the benefits of multi-spectral imagery)."

Comment 7: (Lines 291-294) "Need more explanation."

Response 7: We direct the reader to literature on the balanced random forest method, and we have updated Table 2 to explictly include the sample size included in our balanced random forest algorithm. If you have specific questions about this methodology, we welcome that!

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

very nice work.

I would recommend a few more relevant recent references and little more about methodology.

Last, I didn't see any future research in the conclusions. I suggest add how can this research be used for future studies.

Best,

Author Response

Comment 1: "I would recommend a few more relevant recent references and a little more about methodology."

Response 1: We have included the most current research in remote sensing of bark beetle early attack (including one additional paper in response to reviewer 1), and we have elaborated on our methodology in response to other reviewer questions. If you have more specific feedback, we welcome that!

Comment 2: "I suggest add how can this research be used for future studies."

Response 2: We do in fact include ideas for future work in the last paragraph of our discussion section, and in the last sentence of our conclusion.

Reviewer 4 Report

Comments and Suggestions for Authors

The paper is very interesting and brings important results for pest management in southern pine crops

I have the following suggestions:

1) Describe the objectives of the paper more clearly, for example:

Objective 1: "Relate the color of the pine canopy with the pest cycle"

Objective 2: "Identify healthy, infested (with and without symptoms) and dead pine trees

2) In the Material and methods item:

- Mention the accuracy and method of GNSS positioning (commented on the accuracy of GNSS positioning only in the Discussion item). It is important to highlight that in highly vegetated areas the accuracy of GNSS positioning is degraded

- Further clarify the reason and use of a DTM in detecting and delimiting individual trees

- We believe that a detailed flowchart of the methodology would provide the reader with a better understanding. This could contain data collection (terrestrial and UAV), the respective processing techniques, the generated products and the classification and validation of the data

3) In the Conclusions and Abstract items

- It would be appropriate to include that "multispectral imagery allows for better classification of SPB early attacks than RGB imagery alone"

 

Author Response

Comment 1: "Describe the objectives of the paper more clearly."

Response 1: We believe our objectives are both clear and precise. We have edited the last paragraph of our introduction to make it clear that we are identifying our objectives here. The first sentence of last paragraph of introduction now reads: "In this study, we address two objectives: 1) link SPB brood stage to visible symptoms of tree decline on the ground and 2) identify early attack (or green attack) of SPB using UAV-mounted multispectral remote sensors."

Comment 2: "Mention the accuracy and method of GNSS positioning (commentd on the accuracy of GNSS positioning only in the Discussion item). It is important to highlight that in highly vegetated areas the accuracy of GNSS positioning is degraded."

Response 2: We agree that GNSS positioning is a critical part of UAV-based studies of forests. We now include more information about our GPS receiver in the methods: "We recorded GPS location using an Arrow 100 GNSS receiver (Eos Positioning Systems; Terrebonne, Quebec, Canada), which maintains sub-meter accuracy even in forested settings by using satellite based augentation systems (SBAS)."

Comment 3: "Further clarify the reason and use of a DTM in detecting and delimiting individual trees."

Response 3: We have elaborated on the tree detection process in section 2.6, and this is addressed in the new figure 3 as well.

Comment 4: "We believe that a flowchart of the methodology would provide the reader with a better understanding. This could contain data collection (terrestrial and UAV), the respective processing techniques, the generated products and the classification and validation of the data."

Response 4: We have now added a flowchart. See figure 3.

Comment 5: It would be appropriate to include that "multispectral imagery allows for better classification of SPB early attacks than RGB imagery alone"

Response 5: Thank you for this sentence! We have used this phrasing in our conclusion.