Detection of Forest Tree Losses in Côte d’Ivoire Using Drone Aerial Images

Round 1

Reviewer 1 Report

Major comments:

• The topic is not new; in fact, it has been well understood in the scientific community that using UAV or drone-based imagery can be used for monitoring forest disturbances.
• The main work in this study is the estimation of tree height and tree crown! However, the main objective of the study is to assess the dynamics of deforestation and forest degradation. Therefore, I don’t think these two parameters (tree height and crown) are good enough to achieve the study goal.
• I think the Introduction chapter does not give a clear background of the problem to be addressed. For example, the authors should explain the advantages and disadvantages of different RS platforms (satellite, airborne, UAV and field data) .

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This study tries to build a deforestation monitoring system with drones. However, I feel the author might have not considered the limitation of UAV photogrammetry and thus provided an unconvincing presentation. More justifications are needed (motivation and method) to make this study fit the requirement of academic quality for publication. Below are some detailed suggestions.

Introduction: I feel the motivation of using drones here is not really convincing. For example, how much different is the Ivorian context from the Amazon, while there were so many studies utilising the satellites for deforestation monitoring? Besides, the survey area of drones is one of the major limitations to applying it to large areas. I suggest you can go deeper from the perspective of combining satellite systems with drones to utilise both strengths so that you can emphasise why you should use drones for this application.

L63: should be Sentinel-2

Table 1: some single words such as "crops" are broken into two lines. Better to fix it.

L149: 65% frontlap is too low. Many studies suggest at least 80% is needed to provide accurate results. Need to justify it or discuss this factor later.

L169-170: "texturization aims to create and improve the texture of the images"; I don't understand this sentence. The texturization is to create the texture for the 3D model. How does it improve the texture of your raster products? (The reference here is in French so can't read what does the context mean here.)

Geometric correction: It is also important that what parameters of the Brown optical model was optimised when processing the photo-alignment (see James, M.R., Robson, S., 2014. Mitigating systematic error in topographic models derived from UAV and ground-based image networks. Earth Surface Processes and Landforms 39, 1413–1420. https://doi.org/10.1002/esp.3609). However, since you're measuring tree height, the effect of optical model optimisation should be mitigated when calculating the canopy height model (see Tu, Y.-H., Phinn, S., Johansen, K., Robson, A., Wu, D., 2020. Optimising drone flight planning for measuring horticultural tree crop structure. ISPRS Journal of Photogrammetry and Remote Sensing 160, 83–96. https://doi.org/10.1016/j.isprsjprs.2019.12.006). These factors are worth discussing in the perspective of geometric correction rather than just geo-referencing the DSM.

DTM production: Do you have a reference for this method? I believe this makes you lose the details of the terrain. Did you do an accuracy assessment? If you don't have validation data, another suggestion is that you can compare the DTM generated from the three different dates.

Generation of DHM and DCM: In L207-208, you seem to treat these two as the same definition, but you apparently treat DCM and DHM as different items in equation 2. Please have clear definitions to differentiate the two models.

Tree crown area estimation: I don't see the purpose you did this, as you didn't use it to detect tree loss.

Tree loss detection: How did you determine the criteria of tree height and tree crown area? Besides, I see some potentially critical issues here. (1) your flight configuration was not optimal to provide accurate topographic results, so the tree height difference might be caused by uncertainty (the impact of low overlap already appears in Table 2). (2) Your method is limited on individual tree height, which you delineated manually. Considering how many trees are in a forest, it is probably not a viable solution for a national-scale forest preservation project. Why haven't you considered producing something like LAI for this purpose? In addition, you only used the data from Nov 2018 and Apr 2019. How about the data from Jan 2019? Why did you present it if you didn't use it? I guess the data on Jan can be a source of reference to validate the accuracy of your data if, for example, the tree loss occurred between Jan and Apr 2019.

3.2, 3.2, and 3.4 can be combined together. The DSM and DTM themselves here are meaningless, as you didn't have ground control points.

L320-L322: I can only tell the third peak in site 6, so this assumption is too speculated.

Table 3 and 4: I feel the information here are meaningless. Both the maximum and average height in the two dates are in the range of uncertainty, which means they are not statistically different, yet your method depends on the tree height difference. I suggest you derive the results in three categories that you used (tree loss, stable forest, and non-stable forest) so that you can probably see more differences in the tree loss category.

3.6 and 3.7: Again I can't tell how did you implement tree crwon area for tree loss detection.

4.1: This is more like an introduction than a discussion. If you want to discuss this deforestation process, you can try to relate the timing of each stage with the data acquisition strategy.

4.2: I believe the real tree loss area is probably more than what you delineated, which was the limitation of your method. You should discuss it rather than repeat your result here, and maybe combine it with 4.3.

4.3: Your overlap level definitely caused some problems with your results (Table 2). See my comment earlier and discuss it here.

4.4: Drone image photogrammetry can also suffer from cloud cover, so it's not really an advantage here. Maybe what you mean is the flexibility of operation to avoid cloudy days? Also, as I mentioned at the beginning, both systems have their own limitations so it's better to justify how they cooperate.

Discussion in general: I would argue that your proposed method needs lots of labour work so probably try to discuss something to address this issue. You mentioned lots of other auxiliary data in the conclusion, which is better in the discussion section.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

In the manuscript, the authors applied drones for early detection of tree losses. According to the protocol as shown in Figure 2, it seems using techniques in some professional tools. I should admit that it is beyond my knowledge. Therefore, this review is an outsider review.

(1) If the protocol shown in Figure 2 is using techniques in professional tools, the novelty of this work is the application of drone in data analysis of Côte d'Ivoire, and the value of this work should be enriched in detail.

(2) If possible, the language should be polished. For instance, Line 017 “Remote sensing forest monitoring methods, currently deployed, are not always adapted to the Ivorian context because of the high cloud cover, the diversity of shaded crops, and land clearing methods that are difficult to monitor.” is too long, and it can be changed to “Remote sensing forest monitoring methods, currently deployed, are not always adapted to the Ivorian context because of the high cloud cover, the diversity of shaded crops, and land clearing methods.” In addition, Line 023 “The method used is based on a detection analysis of tree losses in forest areas from a time series of aerial images acquired by drones from November 2018 to April 2019 on five (5) sites in the studied forest. Based on photogrammetric models and photo-interpretation, tree heights as well as tree crown sizes were estimated. Then, tree losses were detected based on the variation of tree heights during the study period.” Uses too many “based on”.

(3) Line 035: The key words are not professional since it is not suggested to use abbreviations and country or city names. Please correct it.

(4) Line 040: What is “FAO”. The full names of abbreviations should be written before the use. Please specify.

(5) Figure 1 shows 11 study location, while in the Abstract, there are 5 sites. Please specify the difference.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

This manuscript has been improved a lot, especially in the discussion section. Some minor revisions are needed. Below are some examples.

L74-75: What is "institutions in the South"? I guess some words are missing.

4.2: Figure 30 and the quality assessment of DTM should go to 3.2 as they are part of your result.

Conclusion: The second paragraph suits better in the discussion section (maybe integrated with 4.4).

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Thanks for the authors. My comments have been well addressed. 

Author Response

Thank you for reviewing our manuscript and thank you for this comment