Extraction of Spectral Information from Airborne 3D Data for Assessment of Tree Species Proportions
Round 1
Reviewer 1 Report
This is a solid piece for work and a useful development of lidar methods, in this case the use of multispectral lidar to detect species in addition to the structural metrics. Here aerial image-based analyses were compared with multi-spectral lidar using well measured forests stands as reference measures.
The design was robust and well implemented, thorough statistical analyses were used which were well described. I was interested in the fact that measures across contrasting seasons improved accuracy.
A useful methods paper that should be published and continues the development off lidar based methods in commercial forestry and terrestrial ecology.
Some expression could be improved.
L 252-253 this isn’t a sentence “This, since these proportions are fractions, i.e. having continuous values within the finite interval, and shows non-linear dependencies to the metrics.”
Consider something like “Since these proportions are fractions, i.e. they have continuous values within a finite interval, they show non-linear dependencies relative to the metrics used.” You could specify which metrics you mean.
L264 should that be “… made using a majority of votes”? I am not familiar enough with Random Forests but I assume this text means the selected (most parsimonious?) tree is based on the number of ‘votes’ it receives?
L269-270 “The predictions were made…”, plural as there are more than one prediction.
Results and Discussion read well.
Author Response
Dear Reviewer 1,
Please find the attached PDF file with our point-by-point response to your comments.
Best regards /Johan (corresponding author)
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This study is about the estimates of tree species-specific proportions from aerial data, using different point-cloud colorizations.
The paper is written quite well, but some changes are required, especially for the introduction and Methods sections.
Introduction
up to line 76: The first part of the introduction is well written and clear, but I found it quite out of date. All your references are quite old, the most recent one is dated 2017. I understand that your subject is consolidated and milestones were defined around 2010, but the topic is constantly evolving and new methods have developed in the last five years, especially regarding data processing. I suggest you to revise a bit this part (even shortening it), with a new literature review that includes both old milestones and brand-new methods.
Lines 77-84: some references are required. I suggest you have a look at doi: 10.3390/rs12061001, https://doi.org/10.1007/s40725-019-00087-2
Lines 84-86: This is the first time that you mention the issue of point-clouds coloring, that will be the main topic of your paper. Therefore, I think it needs more prominence, including also (if any) some literature.
Materials:
Figure 1: because of some formatting problems, I cannot see the whole figure and the north arrow is cut out. May you specify also the size of your study area?
Just for my curiosity, is there any reason why you performed 2014 and 2016 flights not in the same month of the year? Could this affect or improve your results? Moreover, lidar acquisition was performed once again in July.
Some more info on the two cameras would be appreciated (i.e. focal length, image size, view angle,..)
GSD is not defined the first time it appears within the text
I would rephrase lines 144-146 because technically you are using UC2016 image dataset for generating the DSM to be used for comparison purposes.
Methods:
Cells that do not satisfy conditions in line 211 are removed from the analysis?
I was quite confused after the bullet item with all the groups. Probably a summarizing table would be more informative.
Results (is section 4, not 3):
Table 2 would be more suited in the modeling section, and you have to explain the parameters used in the models. In addition, I did not understand how you selected the four models in the table. Are they the best models resulting from the random forest regression? Have you used the same model for all the different coloring methods? RMSE values reported in the table include all coloring methods and all groups for the specific dataset? Moreover, I did not get how the different group metrics enter the model.
Figure 3: coloring methods names used in this figure differ from the ones used in table 1. Please, homogenize.
While reading your paper, I was wondering why there is a need for all those groups. If your aim is understanding how different point cloud coloring can affect the estimates, probably it would be better only to focus only on a few metrics and analyze and deepen the discussion on point cloud coloring variations. Anyway, your discussion is good and covers not only colorizations. I remain a little skeptical about your title and aim of your study.
Author Response
Dear Reviewer 2,
Please find the attached PDF file with our point-by-point response to your comments.
Best regards /Johan (corresponding author)
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Dear Authors,
thanks for having addressed all my previous comments.
I do only have the last comment.
I agree with you that the new title better fits your whole work and, as said in your replies "...the study is also very much about spectral (and spatial) metrics and their significance for tree species estimation. We have also used data from two different seasons separately and in combination (leaf-on vs. leaf-off). Last but not least, a coloured DSM product available from the Swedish National Land Survey is included in the comparison as well as a multi-spectral lidar data set." So why not include also this in your introduction, when you describe the aim of your work?
In my opinion, by adding that you may give the right importance to your paper and also the reader can understand better what he is expecting to find in your paper.
Author Response
Dear Reviewer 2,
Please find the attached PDF file with our response to your last comment.
Best regards /Johan (corresponding author)
Author Response File: Author Response.pdf
