AdQSM: A New Method for Estimating Above-Ground Biomass from TLS Point Clouds
Round 1
Reviewer 1 Report
This is a great improvement on the previous version of this manuscript and you have achieved a lot in a relatively short time. Thank you for acting on my suggestion with regards to the validation data. Also, the difference between this paper and your recently published one is now clear. I do have some additional suggestions to improve this manuscript but nothing that will require major work.
Small issues:
- The in-text citations are often incorrect. I assume this is due to some reference manager, but please check this before submitting the final version.
- You repeatedly note the folder in which certain data was found (e.g. line 137). This should be removed from the text.
Larger issues:
- It is great that you have used this valuable field data to test your method. I just want to make sure that you have checked the data sharing agreement and, if it is suggested, asked the data owners to contribute to this manuscript. I know how much work was put into collecting this data!
- The manuscript has a large number of figures. I would suggest just keeping the key figures and combining others / putting them in supplementary. For example, the residuals in figure 3 could be in the supplementary, or could be included as small inset panels in figure 2. The same can be done for Figure 6, 9 and 12.
- Figure 4 and Figure 10 should have field estimates on the x-axis and model estimates on the y-axis, the tree indices are not helpful and these figures are difficult to interpret.
Author Response
Responses to the comments of Reviewer 1
Dear Reviewer,
Thank you for your insightful comments and constructive suggestions. Following your feedback, we have conducted thorough proofreading and editing of our manuscript to improve the readability and clarity of the paper. We have added a comparison test with TreeQSM and we further clarified the meaning of AdQSM. In the following, we address the individual comments and summarize the changes made in the revision.
Point 1: Small issues:
The in-text citations are often incorrect. I assume this is due to some reference manager, but please check this before submitting the final version.
You repeatedly note the folder in which certain data was found (e.g. line 137). This should be removed from the text.
Answer: Thanks for pointing out these issues, which have been resolved in the revision.
Point 2: Larger issues:
It is great that you have used this valuable field data to test your method. I just want to make sure that you have checked the data sharing agreement and, if it is suggested, asked the data owners to contribute to this manuscript. I know how much work was put into collecting this data!
Answer: Thank you for your comments on this in the early stage, which helped us to obtain this valuable dataset. This paper fully complies with the data usage policies and regulations, references the original paper that provided the dataset, and acknowledges the data owners. We have also contacted the authors who provided the dataset and invited them to participate in our future research.
Point 3:
The manuscript has a large number of figures. I would suggest just keeping the key figures and combining others / putting them in supplementary. For example, the residuals in figure 3 could be in the supplementary, or could be included as small inset panels in figure 2. The same can be done for Figure 6, 9 and 12.
Figure 4 and Figure 10 should have field estimates on the x-axis and model estimates on the y-axis, the tree indices are not helpful and these figures are difficult to interpret.
Answer: Thank you for the suggestion. In this revision, we have moved some pictures (not too relevant to the text, but can improve the readability) and formulas to "Appendix A " (e.g., Figure A1, Equation A1, and Equation A2).
Author Response File: 
Author Response.docx
Reviewer 2 Report
General Comments
The submitted manuscript is a substantial revision of an earlier version submitted to Remote Sensing. In this manuscript, the authors promote and test their AdQSM model which was developed and improved from the version already introduced in an earlier publication by the same authors. I appreciate the work that has been put in to improve this manuscript from the earlier submission. I think that the substantial changes to the manuscript significantly improved the overall relevance of the paper. The difference to the earlier published work are clearer and the addition of an actual validation of their model is very welcome. The results look very promising and the model could be indeed an interesting addition to the already existing models. I appreciate the suggested idea of a benchmark paper, where these models are compared to each other to deliver a guideline for the community for which models to use. For such a future manuscript I would suggest to largely increase the tree sample size and to cover multiple ecosystems (temperate, tropical, boreal etc.) to deliver a broader understanding where which model is better suited. However, I would suggest to add at least a direct comparison to treeQSM in this Paper, so that the reader already becomes an idea of how AdQSM compares to a well established model. As the authors have shown in the first submission of this manuscript that they are already very much used to use treeQSM I would argue that this would not be a lot of work to do. It would however greatly improve the relevance of this manuscript. A future benchmarking paper would still be very relevant, especially if the sample size could be increased and if the sample size covers multiple ecosystems (temperate, tropical, boreal etc.)
I am impressed of the substantial revision of the manuscript in such a short time. It seems that the whole manuscript was largely rewritten. However, the fast rewriting of the manuscript shows throughout the entire manuscript. Multiple arguments by the authors are somewhat poorly understandable. I would suggest a thorough proof reading and editing of the manuscript to
improve readability and understanding of the paper.
To sum up, the two major points which need to be addressed for me to consider this manuscript for publication are:
- Addition of a comparison of the AdQSM results with a well established QSM (e.g. treeQSM). This would deliver a better sense for the importance and significance of the introduced model.
- To the same point: It is not really clear why your model should be considered compared to other, well established QSMs. What are you doing differently and why is this better? Could you clarify on this a bit more? For this, a comparison to e.g. treeQSM could be valuable.
- Thorough proof reading and editing work needed to improve readability and clearity of the paper.
Detailed Comments:
line 15: "... to model 18 trees ...": In the manuscript you mentioned 19 trees? In general: the amount of trees used in the various analysis is somewhat confusing. You mention many different values here. Could you please make it clearer throughout the entire document how many trees were actually used and why some trees were not used in certain experiments (I assume missing reference values)
lines 16-18: Not sure if such a detailed description of the trees is needed in the abstract. Statement within parenthesis
line 23: why surprisingly?
line 26: do you mean systematic deviations?
line 32: better in relation to what? I would be more specific.
table 1: Number of pots -> Number of plots
line 130: Riegl VZ-400 is not necessarily the best TLS. delete "best"
line 131: "... and the most advanced sampling technology" -> What do you mean by that?
line 132: Please rephrase this sentence. Language
line 133: "In situ destructive..." This sentence is rather a repetition. I would rephrase these two sentence to make a better one
lines 130-135: I would argue that when using this dataset a few further explanation on the acquisition and samples should be given in this paper as well and not only a reference to the original paper.
line 137: I would not talk about a folder of a dataset that you acquired from somewhere else. Rather talk about the dataset provided...
line 138: Maybe rather: "Tree species were identified by an experienced ..."
line 139: "Table 2 provides..."
line 141: "harvest were measured" -> before the tree was harvested
line 141: It is not the forestry tape that measured the DBH, the DBH was measured with a forestry tape. Also: do you have further information regarding the measurement device? Was it really just a tape, or a calipper? If the latter, give more precise information on it.
line 142: dito, see comment for line 141 but for the laser altimeter (tree height was measured with the laser altimeter)
table 2: Maybe add further information regarding the measured trees (DBH, tree height etc.) in this table
line 146: see comment for line 137
line 149: Are basic wood densities available for all the trees?
line 150: and other information: such as? or maybe leave this part out
line 165-166: Sentence not very clear. Please rephrase and clarify
Section 2.2.1 and 2.3: The whole section is in need thorough revision. It is not fully clear how the volume references were acquired? Am I correct, that the reference volumes were also based on models and diameter and length measurements? Not actual volume with e.g. water displacement method or tree weight measurement with scales? Also how were AGB reference values derived? You say that you were only able to compare AGB retrieval performance for 10 trees. However, you have volume information for more than 10 trees and you have wood density information as well. Why is the sample size for AGB so small? It should be made more clear how and why the sample selection was performed.
line 175: "The uncertainty introduced when measuring the volume of tree trunks, buttresses and branches were taken into account." -> How?
line 193: Sentence is not clear. Did you mean "... will reduce the accuracy of parameters such as tree volume..."?
Section 2.2.2: I would appreciate if the authors could also mention the differences of their model with other QSMs. What are they actually doing different to others that should improve performance?
line 220ff: So did I understood correctly, you had actuall destructive AGB reference values of 10 trees, the reference values for the remaining 19 trees were derived from wood volume and wood density estimates? I would make it more clear where the reference values come from and how they were acquired. This information is currently rather ambiguous
line 251-252: "... and the estimated value of DBH had a better fitting effect with the reference value." What do you mean by that? better than what?
line 254-255: see comment for line 251-252
line 271: "Table 4 showed..." -> "Table 4 shows..."
line 271: "... tree height estimated based on AdQSM." -> "... tree height estimates based on AdQSM"
line 272-274: I don't think it is necessary to repeat again the DBH range of the tree samples. The last part of the sentence is further not entirely clear. Do you mean 75.9% of the sample trees had a DBH deviation of less than 6.65 cm?
line 275-277: Same comment as for line 272-274.
line 285-286: Suggestion: Reference tree volumes ranged from 1.04089m3 (are so many digits after the decimal really worth it?) to 43.89407m3, and AdQSM estimated tree volumes ranged from 1.14695m3 to 56.2244m3
line 286: Figure 4 provides
line 291: Figure 5 shows
Figure 5: From the methods section it is not clear how you are able to derive the standard deviation of the QSM estimated values. I expect you ran the model multiple times and that there is also (similar to treeQSM) a random term in the model, hence multiple Volume realisations. This should be made more clear.
Line 303: Figure 6 shows
line 308: "There was no significant difference in the distribution range of the residuals as the reference tree volume increased" -> not sure this statement is supported by Figure 6. There you can clearly see an increase in residuals with increasing reference volume
line 310: Table 5 provides
line 311-313: similar comment as for line 272-274
line 317: Language. Please rephrase sentence. what is the visualization effect?
line 318: "estimating AGB" -> "AGB estimates"
line 326: Figure 8 shows
line 337: Figure 9 shows
line 345: Table 6 shows
line 406: Guyana trees provide no information about the volume: This information should be given in the description of the dataset. This is not clear to me from reading the dataset description.
lines 438-440: What do you mean by well-designed supervisory adjustment tools? Can you refer to an example or elaborate more what you mean by that?
Section 4.1: This section is mostly a repetition of the methodologies (in some points you actually made the methodologies a bit more clear here) and the results but it is not realy a discussion of your results (or only in small parts.)
Section 4.2: You give multiple suggestion on how to improve the accuracy of your model. However, as the explanation of the model in the methods section is quite short, the arguments in this part are somewhat less understandable as the reader does not fully understand how the model performs.
line 428: What do you mean by "artificial selection"
line 469: AdQSM
general: The naming of your model is for me still not clear from the manuscript. You explained it in the comments to my first review of the manuscript. Could you add this explanation to the manuscript as well (e.g. underline the a in accurate and d in detailed when introducing the name)?
Author Response
Responses to the comments of Reviewer 2
Dear Reviewer,
Thank you for your insightful comments and constructive suggestions. Following your feedback, we have conducted thorough proofreading and editing of our manuscript to improve the readability and clarity of the paper. We have added a comparison test with TreeQSM and we further clarified the meaning of AdQSM. In the following, we address the individual comments and summarize the changes made in the revision.
Point 1: To sum up, the two major points which need to be addressed for me to consider this manuscript for publication are:
- Addition of a comparison of the AdQSM results with a well established QSM (e.g. treeQSM). This would deliver a better sense for the importance and significance of the introduced model.
- To the same point: It is not really clear why your model should be considered compared to other, well established QSMs. What are you doing differently and why is this better? Could you clarify on this a bit more? For this, a comparison to e.g. treeQSM could be valuable.
Answer: Thank you for the suggestions. In the revised paper, we have added a subsection "2.4.2. Comparison on the accuracy of tree volume reconstruction with TreeQSM", where we report on the comparison of AdQSM and TreeQSM on estimating tree volumes. The experimental results are shown in "3.2.1. Tree volume". We compared the accuracy of AdQSM and TreeQSM in modeling the volume of 29 trees. The CV-RMSE of AdQSM and TreeQSM with reference volume were 22.62% and 23.20%. The CCCs between the volume estimates based on AdQSM and TreeQSM and the volume reference values are 0.97 and 0.96.
On the one hand, the experimental results in section 3.2.1 show that the accuracy of tree volume estimated by AdQSM is similar to that of TreeQSM, even slightly higher. This may be related to AdQSM's ability to model tree geometry with high precision. In the original publication of AdTree, the authors had compared it with TreeQSM, demonstrating the advantages of reconstructed tree geometry over TreeQSM.
On the other hand, AdQSM is more efficient than TreeQSM because it is implemented using the C++ programming language. This paper only makes rough statistics on the modeling time of AdQSM. In the Win10 64-bit operating system (Intel I7-8700 processor, 3.20GHz, and 16G RAM), AdQSM can quickly complete single tree modeling (within a few seconds), while TreeQSM takes a few minutes. AdQSM provides more possibilities for users to quickly model, integrate, or transplant to software platforms related to tree point cloud processing. However, TreeQSM is more complete than AdQSM in terms of user interaction, interface design, and data processing after modeling. In summary, AdQSM could be considered an addition to the already existing models.
We now have validated the proposed model and we are currently cleaning the code of AdQSM, data, and other related materials. We will make our source code publically available after the acceptance of the paper.
Point 2: line 15: "... to model 18 trees ...": In the manuscript you mentioned 19 trees? In general: the amount of trees used in the various analysis is somewhat confusing. You mention many different values here. Could you please make it clearer throughout the entire document how many trees were actually used and why some trees were not used in certain experiments (I assume missing reference values)
Answer: Thank you for pointing out this issue. We admit that there was an error in our previous version and we have corrected this information in the revised version. In the previous version, we lacked the tree volume reference value of Guyana (10 trees), so only 19 trees were used to test the accuracy of tree volume estimated by AdQSM. In the revision, we obtained the reference volume of 10 trees from Guyana, and we modeled 29 trees (18 tree species) to test the accuracy of AdQSM's estimation of DBH, tree height, and tree volume. Since we used the dry mass of trees as the reference value for AdQSM to calculate AGB, we only obtained the dry mass of 10 trees from Indonesia. Therefore, in terms of AGB verification, we counted the experimental data of AdQSM on these 10 trees.
Point 3: lines 16-18: Not sure if such a detailed description of the trees is needed in the abstract. Statement within parenthesis
Answer: We deleted the detailed description of trees in the "Abstract" section and moved this part to section 2.1.1.
Point 4: line 23: why surprisingly?
Answer: AdQSM showed satisfactory results in the original manuscript. In the revised version, we have added a comparative experiment between AdQSM and TreeQSM, and the experimental results in section 3.2.1 show that the accuracy of tree volume estimated by AdQSM is similar to that of TreeQSM, even slightly higher.
Point 5: line 26: do you mean systematic deviations?
Answer: Yes. We use "systematic deviations" instead of "system deviation" in the revised paper.
Point 6: line 32: better in relation to what? I would be more specific.
Answer: In the revised version, "The DBH and height of the harvested trees were used as reference values to test the accuracy of AdQSM's estimation of DBH and tree height."
Point 7: table 1: Number of pots -> Number of plot
Answer: We have fixed this typo.
Point 8: line 130: Riegl VZ-400 is not necessarily the best TLS. delete "best"
Answer: We have deleted "best".
Point 9: line 131: "... and the most advanced sampling technology" -> What do you mean by that?
Answer: Plot spatial design was designed to cover the expected area where the tree to be harvested will land: 30 × 50 m in Peru, 30 × 40 m in Indonesia and Guyana (Figure 2). We aligned the plot’s main axis (longest plot dimension) to the expected felling direction. We established the plot edges such that the harvested tree was located at a half distance over the plot shortest edge, and 5 to 10 m along the central main axis (parallel to the plot longest edge).
Figure 2. Plot spatial design in Peru (left) and Indonesia and Guyana (right). The plots were established around the harvested tree (the green object) and aligned towards the expected felling direction (yellow arrow). Coloured crossed dots indicate scan locations within the plot. The origin of the local coordinate system was located at the bottom left corner of the plot (bold red cross symbol).
The TLS scan locations within each plot followed a systematic spatial pattern within each plot (see Figure 2). It must be noted that the spatial sampling applied was optimized for plot scanning and not for individual trees, which could improve the data quality for target trees. We set up a total of eight scan locations in each Peruvian plot (left image, Figure 2), and 13 scan locations in the other plots (right image, Figure 2). Each scan had an angular resolution of 0.06°. To acquire a full hemispherical scan, two scans were taken at each scan location; one in an upright position (scanner rotation perpendicular to the ground) and the other in a tilted position (scanner parallel to the ground).
Supporting Information for the article “AGB of large tropical trees with T-Lidar”. Plot spatial design and tree selection are detailed in “Estimation of above- ground biomass of large tropical trees with terrestrial LiDAR”(Jose Gonzalez de Tanago).
In "2.1.1" of the revised manuscript, we have clarified this.
Point 10: line 132: Please rephrase this sentence. Language . line 133: "In situ destructive..." This sentence is rather a repetition. I would rephrase these two sentence to make a better one
Answer: Following your suggestion, we have rephrased this sentence.
Point 130-135: lines 130-135: I would argue that when using this dataset a few further explanation on the acquisition and samples should be given in this paper as well and not only a reference to the original paper.
Answer: We further explained the data collection and samples in the revised version, see "2.1.1", "2.2.1", and "2.3".
Point 12: line 137: I would not talk about a folder of a dataset that you acquired from soline 146: see comment for line 137 mewhere else. Rather talk about the dataset provided...
Answer: We have added a description of the data and samples in the "2.1.1", "2.2.1", "2.1.3", and "2.3" sections.
Point 13：line 138: Maybe rather: "Tree species were identified by an experienced ..."line 139: "Table 2 provides..."
Answer: In the revision, we solved these writing or grammatical issues.
Point 14：line 141: "harvest were measured" -> before the tree was harvested
Answer: This sentence has been rewritten as follows: "Some attributes (the DBH, tree height, the height of the living crown, and crown size of each tree) were measured before the tree was harvested."
Point 15：line 141:It is not the forestry tape that measured the DBH, the DBH was measured with a forestry tape. Also: do you have further information regarding the measurement device? Was it really just a tape, or a calipper? If the latter, give more precise information on it
line 142: dito, see comment for line 141 but for the laser altimeter (tree height was measured with the laser altimeter)
Answer: The DBH was measured with a forestry tape with a precision of 0.01m. The tree height was measured with the Nikon "Forestry-Pro" (Hayama, Japan) laser altimeter with a precision of 0.2m.
Point 16：table 2: Maybe add further information regarding the measured trees (DBH, tree height etc.) in this table
Answer: We have added a description of the measured trees in the "2.1.2" as follows: The maximum DBH was 127.6cm and the maximum tree height was 50.5m. The average DBH was 73.5 cm and the average tree height was 33.34m.
Point 17:line 149: Are basic wood densities available for all the trees?
Answer: All 10 trees have a corresponding average basic wood density (See Table 3). In section "2.3" we provide the average basic density of each tree participating in AGB estimation.
Point 18：line 150: and other information: such as? or maybe leave this part out
line 165-166: Sentence not very clear. Please rephrase and clarify
Section 2.2.1 and 2.3: The whole section is in need thorough revision. It is not fully clear how the volume references were acquired? Am I correct, that the reference volumes were also based on models and diameter and length measurements? Not actual volume with e.g. water displacement method or tree weight measurement with scales? Also how were AGB reference values derived? You say that you were only able to compare AGB retrieval performance for 10 trees. However, you have volume information for more than 10 trees and you have wood density information as well. Why is the sample size for AGB so small? It should be made more clear how and why the sample selection was performed
Answer: In the resubmitted manuscript, 2.1.3, 2.2.1, and 2.3 have been rewritten.
The reference volume of the 29 harvested sample trees was determined based on the tree geometric measurement values. For the geometric measurement method, please refer to section 2.1.3. The total volume of an individual tree is calculated as the sum of the trunk, larger branches (diameter> 10cm), and buttresses. After obtaining the tree geometry measurement data, the Smalian formula in Nogueira, Nelson, and Fearnside was used to estimate the volume of the trunk and branches, and the general prism volume formula was used to calculate the buttress volume. Detailed information can be found in Appendix A.
We only obtained the dry mass of 10 trees from Indonesia that were dried in the oven, and 19 trees from Peru and Guyana lacked the dry mass values. Therefore, this paper uses the dry mass of 10 trees from Indonesia to verify the accuracy of AGB estimation based on the AdQSM reconstruction of tree volume. We multiply the estimated volume of 10 individual trees by the average basic wood density (ρ) to calculate the individual AGB. The average basic wood density (g/cm3) of a tree comes from the different tree components of the destructive measurement
Point 19: line 175: "The uncertainty introduced when measuring the volume of tree trunks, buttresses and branches were taken into account." -> How?
Answer: As stated by Berger, McRoberts, and Fearnside, any misrepresentation of the trunk and branch volumes by Smalian approximation and any measurement errors taken were considered negligible and ignored. Furthermore, it is assumed that the sum of all cylinders represents the real tree volume without error, and thus the measurement of wood volume has no error.
Point 20: line 193: Sentence is not clear. Did you mean "... will reduce the accuracy of parameters such as tree volume..."?
Section 2.2.2: I would appreciate if the authors could also mention the differences of their model with other QSMs. What are they actually doing different to others that should improve performance?
Answer: AdQSM uses cylinders to fit the geometry of trees. However, compared with the trunk point clouds, the buttress point clouds are abrupt and non-linear. There may also be a lot of noise points in the point clouds at the bottom of the stem, which will eventually cause the radius of the entire tree cylinder model to be inaccurate. These factors will reduce the accuracy of the model to estimate tree attributes. These issues were not considered in the original modeling method. Deviations in the radius of the tree bottom will lose the accuracy of attributes such as tree volume or DBH.
In the revision, we added a new comparative experiment between AdQSM and TreeQSM, and in "4 Discussions" we provide analysis of the differences between AdQSM and TreeQSM.
Point 21: line 220ff: So did I understood correctly, you had actuall destructive AGB reference values of 10 trees, the reference values for the remaining 19 trees were derived from wood volume and wood density estimates? I would make it more clear where the reference values come from and how they were acquired. This information is currently rather ambiguous
Answer: In the revised manuscript, we reconstructed 29 trees to test AdQSM's accuracy in estimating DBH, tree height, and tree volume.
However, we only had reference values for the dry mass of 10 trees from Indonesia and lacked the dry mass of the other 19 trees. Therefore, we used 10 trees to test the accuracy of AdQSM's estimation of AGB. We multiply the volume of the reconstructed 10 trees by the average basic density from destructive sampling as the AdQSM AGB estimate.
Point 22: line 251-252: "... and the estimated value of DBH had a better fitting effect with the reference value." What do you mean by that? better than what? line 254-255: see comment for line 251-252
Answer: We have rewritten these two sentences. We take DBH and tree height from destructive sampling as reference values. The main purpose of “3.1” section is to test the accuracy of estimating DBH and tree height based on AdQSM.
Estimating tree volume is one of the core functions of the QSM method. We have added a comparative experiment between AdQSM and TreeQSM in "3.2.1" of the revised manuscript.
Point 23: line 271: "Table 4 showed..." -> "Table 4 shows..."
line 271: "... tree height estimated based on AdQSM." -> "... tree height estimates based on AdQSM"
Answer23: This type of issues were thoroughly checked and revised in the revised manuscript.
Point 24: line 272-274: I don't think it is necessary to repeat again the DBH range of the tree samples. The last part of the sentence is further not entirely clear. Do you mean 75.9% of the sample trees had a DBH deviation of less than 6.65 cm?
line 275-277: Same comment as for line 272-274.
Answer24: We have deleted the duplicate information of the sample trees. For example: The DBH deviation of 75.9% of the sample trees was less than 6.65cm. The tree height deviation of 75.9% of the sample trees was less than 1.93m.
Point 25: Suggestion: Reference tree volumes ranged from 1.04089m3 (are so many digits after the decimal really worth it?) to 43.89407m3, and AdQSM estimated tree volumes ranged from 1.14695m3 to 56.2244m3
Answer25: We used more understandable expression and keep 3 decimal places.
Point 26: line 286: Figure 4 provides
line 291: Figure 5 shows
Answer26: These writing issues were resolved in the revised manuscript.
Point 27: From the methods section it is not clear how you are able to derive the standard deviation of the QSM estimated values. I expect you ran the model multiple times and that there is also (similar to treeQSM) a random term in the model, hence multiple Volume realisations. This should be made more clear.
Answer27: In "2.4.2", we have added a comparative experiment. In the TreeQSM modeling process, to obtain the optimal parameters, nine trees were randomly selected and modeled 10 times for each tree. Then relatively stable parameters (PatchDiam) were obtained, and finally, the volumes of the trees can be computed. AdQSM works in the same way.
Point 28: Line 303: Figure 6 shows
line 308: "There was no significant difference in the distribution range of the residuals as the reference tree volume increased" -> not sure this statement is supported by Figure 6. There you can clearly see an increase in residuals with increasing reference volume
line 310: Table 5 provides
line 311-313: similar comment as for line 272-274
line 326: Figure 8 shows
line 337: Figure 9 shows
line 345: Table 6 shows
line 318: "estimating AGB" -> "AGB estimates"
Answer28: The revised manuscript further modified the description of the experimental results and corrected the writing problem.
Point 29:line 317: Language. Please rephrase sentence. what is the visualization effect?
Answer29: We rewrote these two sentences as follows: AdQSM can model branches in detail (Figure 7). The detailed branch structure can improve the accuracy of AGB estimation.
Point 30: line 406: Guyana trees provide no information about the volume: This information should be given in the description of the dataset. This is not clear to me from reading the dataset description.
Answer 30: In the revision, we finally obtained all the reference volumes of 29 trees. Therefore, we used 29 trees to test the accuracy of AdQSM's estimation of tree volume.
Point 31: lines 438-440: What do you mean by well-designed supervisory adjustment tools? Can you refer to an example or elaborate more what you mean by that?
Answer 31: We provide a (PatchDiam) parameter setting tool similar to TreeQSM, allowing users to set parameters before modelling, improving modeling accuracy.
Point 32: Section 4.1: This section is mostly a repetition of the methodologies (in some points you actually made the methodologies a bit more clear here) and the results but it is not realy a discussion of your results (or only in small parts.)
Section 4.2: You give multiple suggestion on how to improve the accuracy of your model. However, as the explanation of the model in the methods section is quite short, the arguments in this part are somewhat less understandable as the reader does not fully understand how the model performs
Answer 32: We have added the discussion and analysis of model comparison experiments in Section "4.1 Discussion ". We have deleted repeated descriptions of methods or experimental results. The "4.2" part mainly describes the error source of the model and the measures to reduce the error. In the revised manuscript, we try our best to improve the readability of this part.
Point 33: line 428: What do you mean by "artificial selection"
Answer 33: During the modeling process, manually selecting the point clouds to fit the initial cylinder is the biggest source of error.
Point 34: line 469: AdQSM general: The naming of your model is for me still not clear from the manuscript. You explained it in the comments to my first review of the manuscript. Could you add this explanation to the manuscript as well (e.g. underline the a in accurate and d in detailed when introducing the name)?
Answer 34: Ok. When we first used "AdQSM" in the manuscript, we made further explanations and annotations, as follows: This paper introduces an accurate and detailed quantitative structure model (AdQSM), which can estimate the AGB of large tropical trees.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Dear authors
Thank you for considering my suggestions to improve this manuscript. I am glad to see how it improved over these iterations and I am happy to suggest the manuscript for submission after just a few minor corrections:
line 18: ".. the accuracy of diameter at breast height (DBH),..."
line 38: "... and testing of allometric equations." -> There is not one single allometric equation.
line 229: formating -> 2.4.1 to a new line
Figure 5: color selection could be improved to increase readability
Figure 6: Maybe add an annotation directly to the figures to clarify which belongs to TreeQSM and which to AdQSM.
line 339: TreeQSM can model branches as well. You could have compared these as well, but this is not really necessary. But maybe make it clear that this is also possible with TreeQSM.
Figure 8: Caption is on a different page than the figure
line 405: Did you mean less than 0.5Mg? 5 Mg is quite huge.
line 435 - 437: Language. Suggestion: AdQSM and TreeQSM differ in terms of programming language used for the implementation. While TreeQSM was implemented in Matlab, AdQSM was developed in C++, therefore showing a performance advantage over TreeQSM.
line 440-441: "AdQSM provides more possibilities for users to quickly model, integrate, or transplant to software platforms related to tree point cloud processing" -> this is unclear to me. What do you mean by "transplant to software platforms related to tree point cloud processing"? Do you mean transfer? But why is that easier than in TreeQSM?
line 513: SimpleTree. And do you mean PypeTree?
Author Response
Responses to the comments of Reviewer 2
Dear Reviewer,
Thank you for your insightful comments and constructive suggestions. Following your feedback. We have solved the problem of typos, pictures and grammar. In the following, we address the individual comments and summarize the changes made in the revision.
Point 1: line 18: ".. the accuracy of diameter at breast height (DBH),..."
Answer: We corrected this information in the revised version.
Point 2:line 38: "... and testing of allometric equations." -> There is not one single allometric equation.
Answer: We revised the sentence. “This paper provides not only a new QSM method for estimating AGB based on TLS point clouds but also the potential for further development and testing of allometric equations”.
Point 3:line 229: formating -> 2.4.1 to a new line
Answer: We have moved this title to the next line
Point 4:Figure 5: color selection could be improved to increase readability
Answer: We re-selected the color. The orange-redgrey band represented the 95% confidence interval of the regression.
Point 5:Figure 6: Maybe add an annotation directly to the figures to clarify which belongs to TreeQSM and which to AdQSM.
Answer: We have added annotation to Figure 8 to clarify TreeQSM and AdQSM.
Point6:line 339: TreeQSM can model branches as well. You could have compared these as well, but this is not really necessary. But maybe make it clear that this is also possible with TreeQSM.
Answer: The current work has verified the accuracy of AdQSM in estimating branch volume, and it has not been compared with that of tree QSM. In the future, we will use the existing QSM methods in more sample trees and more research sites. The volume accuracy of branches (or more attributes) will be compared.
Point 7:Figure 8: Caption is on a different page than the figure
Answer: In the revised version, the figure and the caption are on the same page (13 of 23).
Point 8:line 405: Did you mean less than 0.5Mg? 5 Mg is quite huge.
Answer: Thank you very much for your comments. "0.5mg" should be used. The AGB deviation of 60% of the sample trees was less than 0.5Mg.
Point 9:line 435 - 437: Language. Suggestion: AdQSM and TreeQSM differ in terms of programming language used for the implementation. While TreeQSM was implemented in Matlab, AdQSM was developed in C++, therefore showing a performance advantage over TreeQSM.
Answer: In the revised version, we have accepted the above suggestions.
Point 10:line 440-441: "AdQSM provides more possibilities for users to quickly model, integrate, or transplant to software platforms related to tree point cloud processing" -> this is unclear to me. What do you mean by "transplant to software platforms related to tree point cloud processing"? Do you mean transfer? But why is that easier than in TreeQSM?
Answer: The input data of AdQSM is point clouds of a single tree. In some cases, it is necessary to directly perform three-dimensional reconstruction of multiple trees. This will involve knowledge in areas such as tree segmentation and recognition. Many software used to process forest point clouds (such as ENVI LiDAR and LiDAR360) are developed based on C++, and AdQSM can be more easily integrated into these applications. Of course, the discussion on this point may not be important for the research purposes of this article, so we removed "transplant to software platforms related to tree point cloud processing".
Point 11: line 513: SimpleTree. And do you mean PypeTree?
Answer: We corrected this information in the revised version.
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This is an interesting paper and it is great to have new methods to fit QSMs to terrestrial LiDAR data. However, there are a few substantial problems with the analysis at the moment.
- The written English could do with improvement. This doesn’t hinder the readers understanding of the manuscript.
- There is no link to the code, so I can’t test the method. I have some LiDAR data and TreeQSM installed and I would like to make this test.
- This paper has a lot of overlap with your previous one [reference 37] and many of the figures are identical. Please reduce the overlap and focus on validating your method as suggested in points 4 and 5 below.
- AGB validation: The current test of AdQSM against AGB derived using allometric equations is a bit strange. The point of reconstructing the whole tree it to get more accurate estimates than those possible with allometry. I suggest using some benchmark AGB data. Your citation [29] provides a link to freely download the TLS data http://data.auscover.org.au/xwiki/bin/view/Product+pages/Rushworth+TLS+tree+models and the AGB data is available on request: The forest inventory and destructive sampling data are owned by The University of Melbourne, Department of Forest and Ecosystem Science, Water St., Creswick, Victoria 3363, Australia (Contact either Simon Murphy or ChrisWeston:[email protected]; [email protected]).
- Branch structure validation: You claim that AdQSM represents branches more accurately than TreeQSM (lines 359-362), but there is currently no test of this in the paper. I appreciate this is a very difficult thing to test, but one of the papers you cite [18] does exactly this. You could ask the authors to share data (I know they are willing to share) and test your model.
More data to test the model：http://lucid.wur.nl/datasets/terrestrial-lidar-of-tropical-forests
Reviewer 2 Report
Hello.
The paper, as I understand it, presents and performs a test application of "A New Method" but it does not develop "A New Method" as one would expect reading the title.
Beside this critics, the presentation is technically acceptable.
Minor English corrections could be considered (but many published papers are much worse). Two examples are included as follows.
Page2:
1. Line 42: Is 'differential' correct?
2. Line 63: What is the remote sensing estimation? Already introduced?
3. Line 65: What is allometric growth?
Page 7:
4. Line 231: Double reference [33]?
Regards
Reviewer 3 Report
The authors claim to introduce a new Quantitative Structure Model (QSM) for the extraction of Above Ground Biomass.They compared their results of their model with results based on allometric equations and based on results from the well established QSM model introduced by Pasi Raumonen (TreeQSM). While this comparison showed good correspondance to both allometric and TreeQSM extracted values, the authors do not present a form of validation of their method.
Even though, the authors claim that they introduce a new QSM model to extract AGB of trees, most of the work presented in this submission has already been published in an earlier published work by the same authors (Fan, G., Nan, L., Chen, F., Dong, Y., Wang, Z., Li, H., & Chen, D. (2020). A new quantitative approach to tree attributes estimation based on LiDAR point clouds. Remote Sensing, 12(11), 1–20. https://doi.org/10.3390/rs12111779). In fact, some results presented in this work are exactly
the same as published in their earlier work (i.e. Comparison of derived DBH and Tree height to field measurements). It is not clear what has been performed in the earlier publication and what was added in this new submission. From my point of understanding, the only added value, was the multiplication of the results stated in Fan et al. 2020 with the basic wood density to extract AGB (which derivation is not clear to the reader. Where did the densities come from? Was the density value the same for each tree? etc.). The multiplication of earlier published results with a factor does not qualify for a new publication in my opinion.
The mere comparison of the "introduced" model with other models (allometric equations, TreeQSM) is a mere model comparison and not
realy a form of validation. I would strongly suggest the authors to add an actual validation of their method to convince the reader that you actually have developed a new and improved QSM model.
Considering this, I do not suggest this manuscript for publication. Following some further detailed comments.
Line 12-13: Acutally, if put in that way, I would expect that this is actually the first time this model is presented. However, the methodologies to extract wood volume have already been published in Fan et al. earlier this year. The only difference here is, that the final tree parameter is not wood volume but AGB. However, this is only a factor of wood density. The goal of Quantitative Structure models is first and foremost to extract wood volume. Therefore AdQSM should have been introduced first in Fan et al. 2020
L18 (and in general): There is not one allometric equation. Therefore get rid of "the" whenever you are talking about allometric equations
L22-24: Make two sentences. This is too complicated.
L26-27: What does that tell us? I have no idea, wether AdQSM is now better or worse than TreeQSM. I just know that they correlate better than compared to allometry based estimates.
L27-29: This method is not new. The method has already been greatly introduced in Fan et al. 2020a earlier this year. The only difference I see here, is that the results of Fan et al. 2020a have been multiplied by the wood density. This, in my opinion, does not qualify for a new method.
L34: Is this "real" necessary?
L52-54: Here I would expect a citation
L60-61: Complicated Sentence. Do not fully understand. Please rephrase
L65: detecting: change wording? deriving, estimating, extracting?
L66-68: This is somewhat unclear. Please clarify
L77: Oak and eucalyptus trees are probably quite different to your trees. This comparison might be worthwhile in the Discussion
L82-83: Not clear where the derived volume comes from (derived from what?) and what the reference volume is. Please clarify
L87: were -> showed
L105: is it really an improvement if you just multiply the results in Fan et al [37] with the basic wood density?
L106: Again: is this "real" necessary?
L111-113: I am just not sure what this comparison actually tells us. It is just a model comparison, but it gives us no real information on the accuracy of your new approach. Validation with destructively measured trees would be necessary to convince, that your model is actually better than treeQSM
L118-120: Are you actually verifying your models?
L140-141: This sentence is not really clear. Please clarify and rephrase
L142: refer -> can also refer to
L146: This paper -> We
L147: please also add latin name
All Figures in general: The figures are quite low in quality. Please improve figure quality (resolution)
L176: other operations: What other operations. Be more specific
L186-187: Incomplete sentence. Please rephrase. And how did you decide, the point cloud was incomplete?
L202: Please give some information regarding the basic wood density. Where was it derived from. What value did it have? Did you vary the density for different trees?
L209: What does the Ad in AdQSM stand for? Advance? clarify this somewhere.
L210: based on the convex hull -> based on a convex hull
L215-216: It might be worth to add further explanations on the difference between your model and the one from Pasi (treeQSM)
L228-252: Please add some information regarding the tree QSM parameterisation. The results from treeQSM can vary substantially depending on the used parameters. It would be good to know, what parameters you used.
L245-246: How did you quantify this?
L248: Yes, but this does not mean, that AdQSM is more accurate. It just detects more branches
Figure 7: Wouldn't it make more sense to combine these two graphics into one single figure? What is the added value of Figure 7a, compared to 7b?
L312-314: extracted from what? Using AdQSM? Please be more specific
Figure 11: Exactly the same result as presented already in Fan et al [37], with added error bars. No new results!
L325-326 and throughout entire manuscript: Rather state the comparison, so comparison between allometric/AdQSM and comparison between treeQSM/AdQSM rather than stating first and second experiment
Figure 13: This figure is hard to interpret. Also, what does it really tell us? Is this Figure actually necessary?
L343: Please give some examples for general ecological characteristics
L347-348: I am not sure, what this statement means? Your goal is not to model TLS point clouds, but rather to extract AGB from TLS point clouds, no?
L353-356: These two sentences are not clear. Please rephrase and clarify.
L362-363: Please elaborate what this means? This is not really a validation of your method. So what do we know from this comparison?
L371: changed -> changes?
L375-376: What do you mean by "better reference values"? Please clarify and rephrase. Also better compared to what?
L385: detect -> estimate
L385-387: It is not clear what was done in previous work, and what is actually new here. As far as I understood, the only new thing here, is that you multiplied the volume derived in [37] with the basic wood density to extract AGB. This is not really a new method! This is just a simple conversion from volume to weight (volume*density)
L389-390: What do you mean by hidden tree properties?
L391: Again: Make some examples for ecological characteristics
L398-401: This is also true for TreeQSM
L407-409: It could have been nice, to add this information (including a graph) into the results section.
L411-412: Although the AGB estimated by AdQSM has some reference values: What do you mean by that? What are these reference values? Do you mean the ones estimated by
Allometries nad TreeQSM? Be more specific
L413-415: But there are methodologies available for that. See the work by Di Wang or Matheus B. Vicari
L424: photometric: photogrammetric?
