Considerations for Assessing Functional Forest Diversity in High-Dimensional Trait Space Derived from Drone-Based Lidar

Round 1

Reviewer 1 Report

The topic of this manuscript is timely and interested to remote sensing readers. Overall, it is well written, but  I do have some concerns.

1. The authors mentioned in the introduction about the importance of quantifying traits of individual trees (line 67-69) and stated that "Our work aims to contribute to the assessment of functional diversity by providing new insights in assessing detailed ecologically meaningful traits at the level of individual trees. " (line 129-131) and in the conclusions (line 612-614). However, in the methods and results, it is not clear how did the authors obtain individual tree information and derive traits at individual tree level? It seems the authors focus more on the area-based approach.

2. The authors stated the kernel size as one main finding, but it is not clear how did the authors analyze using their own data? or it was only a recommendation from Blonder et al. 2014 (line 541-552)?

3. The authors highlighted collected data should have a pixel spatial resolution capable of capturing required spatial details of the targeted ecological unit (line 553-555; line 592-593), and they mentioned in the methods (line 275-284) that they would assess the impact of spatial scale and grid resolution on trait retrieval and tested grid cell sizes from 0.5 up to 5 m. However, I did not see the results comparing and assessing the effects of different grid sizes, which also made the conclusion a bit weak.

Some more specific comments:

1. The first paragraph of the introduction mainly talked about EBVs, but it seems its relevance is not very close to the scope of this paper.

2. Line 51: typo, two periods.

3. Line 84-88, the authors did not fully illustrate the two potentials of ULS, (1) local-to-regional landscape scales (the study area is about 1ha stand/plot scale), (2) link TLS observation to ALS or space-borne LiDAR and reveal the impact of point density and observation angles on morphological trait retrieval. If it is a general background of ULS, then the relevant literature is needed.

4. Line 123-125, PCA has been used in the hypervolume-based functional diversity estimation, e.g., Zhao et al., 2021. (Zhao, Y., Sun, Y., Lu, X., et al., 2021. Hyperspectral retrieval of leaf physiological traits and their links to ecosystem productivity in grassland monocultures. Ecol. Indic. 122, 107267.)

5. Figure 1: CHM resolution 0.5m?

6. Line 227-228, how many meters each vertical layer is?

7. Line 290 and Appendix Table B6-B11, I did not get the point why the authors did logarithmic and power-law regression, what’s the results or conclusion? If it is not necessary, I would suggest deleting these tables.

8. I felt that Table B5 could be more useful to see the consistency between TLS and ULS traits, if combined with scatter plots.

9. Line 343-345 and figure 7: I would generally prefer separate maps for each functional diversity metric. The composite-color maps look nice but are difficult to read.

10. Line 373-378 repeat as the methods, should delete.

11. Line 418-420, the point number changed from 9-225, what’s the increasing step? Or it might be clear to mark the point of the studied point numbers in Figure 6

12. Figure 7, separate maps of functional diversity metrics are needed. The RGB maps could not read what’s the range for each FD metric, which also makes it challenging to interpret relevant results and discussion.

13. Figure 7,  The "general consistency" between different FD maps derived from the three approaches are merely visual. It is worth showing pixel-wise differences between the maps.

14. Line 478-482, "Jiang et al. [77] used the Beer law to estimate leaf area index from aerial full-waveform lidar data, which offers future avenues for our follow-up research", the linkage with this study is not clear.

15. 4.3 mainly discussed increasing the number of traits, while the authors titled this section as "trait probability density".

16. Figure 4 and Line 524-528, it’s unclear why FDiv would turn to 0 when more than 3 traits were used? Also more detailed information about how FRic, FEve, FDiv were calculated in the methods is needed.

17. Appendix figure B1, B2, the axis labels are too small to read.

18. The manuscript named the first trait as Canopy Height (CH), but in the Appendix the authors used CHM, e.g., Figure B3, B4, Table B1,B2

19. Figure B7-B10, the labels are wrong.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Keywords should not repeat words from the title as drone, lidar, functional diversity.

Table 1: An additional column with list od ecosystem services adressed by particular traits would be interesting for reader.

Line 227: Formulas 1, 2, 3 - D ENL: description how vertical layers (number and height of the layers) were distinguished/predefined is missing.

Line 242: Standard deviation in absolute or relative terms? Although after standardization it doesn't matter..

Formula 5: What is the logic of including the dz parameter? How is the layer thickness related to the number of points/returns? In which units it is expressed?

Line 357: Canopy height is introduced as CH in line 223, here CHM. Please be consistent.

Figure 3: The labels of the axes should be identical with abbreviations introduced in the text.

Line 435: Do high F Ric values really correspond to tall trees on an area? I do not think so.

Line 466: CHM vs CH.. please check and correct in the whole manuscript.

Lines 544-546: There is a duplicity with the lines 339-341.

Conlusions: The text sounds more like an abstract. I recommend move section 4.6 to the conclusions and emphasize it more.

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Thank you for submitting this interesting and timely article. This study investigate methods to better understand functional traits (richness, evenness, and diversity) using trait probability density calculated using kernel density estimator, support vector machine, and principle components analysis. The manuscript is clear, logical, and well-written and the material contained there-in is relative to expanding the utility of lidar remote sensing to ecosystem studies. Overall, I have only minor suggestions for the improvement of this manuscript.

In the assessment of computational time (lines 418-420), it is unclear whether these times were derived from a single test or from multiple iterations. I suggest the authors perform multiple iterations (n=10) to produce the KDE and SVM TPD and report back the median computation time along with the standard deviation.

Additionally, in lines 275-284, the authors state they performed the analyses at multiple grid cell sizes, ranging from 0.5-5 m. However, I did not see results presented as to whether the grid size affected interpretability, particularly for canopy height and ²D ENL, which have previously been shown to be different dependent on grid size in a high diversity tropical system (Swanson, A.C. and J.F. Weishampel. 2019. Scaling of lidar-derived metrics across a Mesomerican landscape. International Journal of Remote Sensing. 40(18): 1-27).  

I also suggest the following minor edits for the manuscript:

Line 65. Missing word. I believe it should read "such as canopy elements"

Line 152. Clarify "it". The way it currently reads, "it" refers to data collection, when I believe the authors are referring to the study site.

Line 162. Briefly summarize the sampling pattern described by Wilkes et al.

Line 169. It may be relevant to include the overall flight altitude.

Lines 223, 225, 237, 241, 244, 251, 258, and 262. Please use the full names of the attributes followed by the abbreviations the first time attributes are presented in the text.

Line 300. I think "axis" should be plural.

Line 369. The authors mention supplementary materials B1, but there is no mention of supplementary materials A. I suggest that Table A1 and Table 1 be merged, as they contain overlapping information.

Lines 370-372. Please refer to the appropriate supplemental information relating to pair plots, correlation analyses, etc.

Lines 418-420. This sentence is unclear. I think it should read "Additionally, we compared the computation time to produce KDE and SVM TPD outputs when increasing the number of data points in steps of 9 (from 25 to 225), which simulates common kernel sized while keeping the number of dimensions at 4."

Line 448. The word "to" is incorrect in "because PCA does to change..."

Line 453. Briefly restate the five criteria, instead of pointing back to section 2.3.

Line 466 and throughout the rest of the manuscript. Use the same abbreviations. Starting here and continuing through the supplemental information, the authors changed CH to CHM.

Line 470. "advantage" should be plural as multiple advantages are listed

Line 470. "less" should be "fewer"

Line 471. The authors suggest that ¹D ENL, which is based on the Shannon-Wiener Diversity Index, is less intuitive that ²D ENL. That may be addressed by using the Jost Diversity Index, which is based on the Shannon-Wiener Index, but is more interpretable (Jost, L. 2006. Entropy and diversity. Oikos. 363-375.)

Line 481. "the Beer law" should be "Beer's law"

Line 491. Use consistent wording for PCA (versus PC analyses).

Lines 493-494. The word "future" is used twice in this sentence.

Line 507. "PGP" is not defined in the paper. I think this refers to PGR.

Lines 513 and 514 do not seem to be tied to the rest of the paragraph. I suggest rewording this sentence to connect it to the presented results.

Line 525. Reword to "most likely has a low evenness"

Lines 545 and 546. Add "i.e.," to the parentheticals as in line 545, the 10 preceding the parenthesis can be misinterpreted as indicating the equation should be multiplied by 10.

Line 548. Change "5 < traits" to "> 5 traits" to make this sentence clearer.

Line 563. Change "explanation" to "explanatory"

Line 585. This is a run-on sentence and should be divided into at least two sentences.

Throughout the supplemental tables, the authors should use a consistent number of significant figures and consistent abbreviations (e.g., CH vs CHM). I also suggest that the authors consider remaking several of their images more accessible for colorblind audiences. 

Figure 4 is missing the y-axis label and might be more interpretable if split into 3 figures, one for FRic, one for FEve, and one for FDiv.

Finally, several citations are either incomplete or improperly formatted (extra information, incorrect capitalization, additional links, team names formatted as last name, first initial). The ones I found were 6, 13, 17, 19, 24, 28, 29, 30, 31, 34, 36, 39, 40, 41, 43, 44, 47, 50, 58, 60, 64, 65, 66, 70, 71, 73, 76, 78, 81, 86.

Author Response

Please see the attachment

Author Response File: Author Response.pdf