Estimation of the Three-Dimension Green Volume Based on UAV RGB Images: A Case Study in YueYaTan Park in Kunming, China

Round 1

Reviewer 1 Report

In my opinion, this paper is very interesting for the approach to the assessment of the three-dimension green volume (3DGV). The authors propose the mean of the neighboring pixels algorithm to estimate the 3DGV. The results of the work showed a good performance of the proposed method. The paper is comprehensive and rigorous.

I have not some specific comments about the paper, but I suggest evaluating the procedure proposed in some fields research using the RGB data from UAV (as Qi et al., 2021; Altieri et al., 2022; Vinci et al., 2023). These papers suggested, the evaluation was conducted on field data and not in urban areas, but in Vinci et al., 2023, the analysis was conducted directly on the point cloud. As assessed in the Discussion section, the DSM and the DTM could be affected by some errors, thus could be interesting to test the procedure on urban green space.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The article is well written with high scientific merits. I suggest minor revision.

Some specific comments:

1. Add authority in botanical names presented in Table 3

2. Urban green space (UGS), use as UGS after first-time use throughout the MS

3.  In the Results section delete lines 360-362 (the first para). also suggested explaining the results first and citing the figure in the bracket

4. Rewrite the results sections following the already published article in 'Forests'

 

Author Response

Please see the attachment.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

 

forests-2256883-peer-review-v1

 

The manuscript “Estimation of the three-dimension green volume based on UAV RGB images: A case study in YueYaTan Park of Kunming, China” addresses an interesting and up-to-date subject, which adhere to Forests journal policies.

 

The manuscript tackles an interesting topic, related to the evaluation of trees and biomass metrics by means of UAV photogrammetry, field surveys and spatial analysis. The work is comprehensive and well written. Medium improvements recommended:

 

-        R127 I think it is “Study area”

-        At the mentioned field survey, add the remark GNSS or DGPS system for the used instrument ZHD V200

-        You mention 12 GCPs for the georeferencing accuracy assessment. It is not clear what accuracy you obtained or what was the RMSE

-        The DSM and DTM outputs are very important, as the difference between them is the vegetation height (the backbone of your research). Please further detail the process of obtaining the DTM, was it fully automated from Pix4D, or did you also clean manually the point cloud

 

 

Author Response

Please see the attachment.

 

Author Response File: Author Response.pdf

Reviewer 4 Report

The paper deals with estimation of the three-dimensional green volume based on UAV RGB images. The paper deals with a very topical issue and advanced methods are used for data processing. As a whole I have to evaluate it as very successful, comprehensive and without serious errors. The literature search is very detailed, the data processing methodology is written in a clear and understandable way and all figures and graphs are easy to understand. Only the discussion is somewhat shorter.

Nevertheless, I have a few comments on the abstract, methodology and discussion:

I found several typos in the Abstract (line 19: Mode instead of Model, line 21: Sentence is starting with And).

Within the vegetation classification, Random Forest was used to extract vegetation and non-vegetation pixels. It is not clear from the description whether all 5 vegetation indices (EXG, NGRDI, NGBDI, RGRI, VDVI) as well as texture features entered the model as independent variables. Although this approach is certainly significantly more accurate than simple per-pixel classification of the source RGB orthophoto, my opinion is that even a simple supervised classification (e.g. Maximum Likelihood Classification) could achieve a similar result, as the goal was only to extract vegetation, not e.g. tree species. If I understood correctly, it would be useful to explicitly specify that all indexes and texture features entered the RF classification.

I have no comment on the 3DGV estimation methods themselves. However all methods are based on the CHM, which is derived from the difference between the DSM and the DTM. I did not find anywhere the number of so-called ground points from which the DTM was created. In the case of a dense canopy, the number of ground points from RGB data from the UAV can be very small and lead to inaccurate results. Please add the average point cloud density for calculating the DSM and DTM. These sources of error should also be mentioned in the Discussion. 

My last comment concerns the appropriateness of using UAV data for 3DGV estimation. The definition of 3DGV already shows that: "The 3DGV is a volume occupied by the stems and leaves of growing plants". In the case of UAV data, only the canopy surfaces are captured and information on the stem and lower parts of the canopy is missing. Although the volume of these parts may be significantly smaller in the overall estimate (and possibly a source of error), it is also worth mentioning in the discussion. 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf