An Advanced Terrain Vegetation Signal Detection Approach for Forest Structural Parameters Estimation Using ICESat-2 Data

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I am glad to review this manuscript. The authors employs deep learning network to correct the terrain and remove noise from the original point cloud, thereby calcualting forest height and coverage. The paper is innovative, but there are areas for improvement.

Materials and Methods

Some paragraphs are densely packed with technical jargon and complex sentences. Breaking these into simpler, more digestible sentences could help improve readability for a broader audience, including those less familiar with the specific remote sensing technologies or statistical methods used.

Here is an example.

"In order to obtain more types of accurate forest structural parameters, previous studies have utilized inversion models that integrate spaceborne LiDAR footprint data with multi-source data."

How about "Previous research has combined spaceborne LiDAR data with other types of data to more accurately measure different aspects of forest structure."

Still a lot of sentences are hard to read (very tired to read). Please rewrite those long sentences.

 

Results

Regrettably, I did not see why you chose to use deep learning methods to re-estimate forest height and cover, although I am aware that the official products ATL08 perform poorly in areas with terrain. Indeed, you compared your method with airborne LiDAR data, but where lies the advantage of your method over ICESat-2 ATL08 data? You should also compare your results with ATL08 to state where exactly ATL08 falls short. Therefore, you should also strengthen the introduction section about this.

 

Minor

Page 2 Line 89: Martin compared…… is not a correct citation.

Page 3 Line 126 CV is the first time to appear in the text. Please use full name, as well as PRIF. The definition you do in the abstract does not count. You should define them again in the main text.

Comments on the Quality of English Language

avoid use long sentences. 

Author Response

We feel great thanks for your professional review work on our article. With your valuable feedback, we were able to improve the quality of our manuscript. We highlight the reasons for using deep learning methods to predict forest structure parameters and modify the construction of sentences to improve readability.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This study appears to be about using various deep learning techniques to process spaceborne LiDAR, using an approach that considers the spatial distribution characteristics of the profile raster images of footprints from ICE-Sat2 data.  The ICE-Sat missions were the benchmark system for measuring ice sheet mass balance over polar regions, clouds and aerosols, and then land topography and vegetation characteristics.  Specifically, the idea of transforming spaceborne LiDAR data into profile raster images of footprints (PRIF) to serve as research targets and considering the spatial distribution characteristics is presented as an important idea so as to not lose this spatial information.  However, there was not a clear comparison in the study of results from using the PRIF and spatial distribution characteristics to results if a PRIF was not used as a research target to test the assert that using the PRIF this way was superior.  The comparison of results was between algorithms using a standardized PRIF.   The various algorithms in the proposed framework are existing algorithms. It is not clear why these three specific deep learning algorithms were chosen (L154) and expected to be superior. 

L612-626:  this paragraph about a future satellite is not related to anything else in this manuscript and needs to be deleted.

Comments on the Quality of English Language

L612-626:  this paragraph about a future satellite is not related to anything else in this manuscript and needs to be deleted.

Minor grammar issues:

L20 – “detected” should be detecting

L 38: “to response and mitigate” should be “to respond to and mitigate”

L40: “ to response climate change” should be “to respond to climate change”

Look throughout manuscript and correct others like these, I have not listed all of these minor issues.

Author Response

We feel great thanks for your professional review work on our article. With your valuable feedback, we were able to improve the quality of our manuscript. The revisions can be seen in the attaches.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Manuscript ID: remotesensing-2907984

Title: An advanced terrain-vegetation signal detection approach for forest structural parameters estimation using ICESat-2 data

Authors: Yifan Li, Xin Shen, Lin Cao

Recommendation: Major review

 

The present manuscript presents a new approach for ICESat-2 data analysis in detecting forest structural parameters, by transforming it into profile raster images of footprints (PRIF). Terrain-vegetation signal detection and forest structural parameters are performed through the proposed TSNN framework integrating CV, OPTICS and deep learning algorithms (CNN, ResNet50,EfficientNetB3).

 

Overall the manuscript is well structured, of interest for the readers of the Journal, nevertheless there are missing points that need to be addressed and the quality of English language needs to be improved.

 

Please refer to the following suggestions and comments: 

 

1. Please clearly state in the Introduction section the research questions of the performed investigations, that are addressed in the study and answered in the Conclusions section.

2. Why are the ALS derived products only 2 m resolution, when the point density is 9 points per square meter?

3. The PRIF are based on 30 m width segments? Which is the footprint on the ground for the profiles in Figures 3, 7, 9? Please explain in more detail..

4. What is the performance of the proposed TSNN method when changing the ratio 7:3 between training and test datasets? Which are the best values for the accuracy metrics when using a minimum amount of labeled dataset for training?

5. Section 2.7.2 ResNet50 starts with the first paragraph about CNN

Minor comments:

Line 111: were always limited

Lines 126, 130-133, 89-92, 215-218: rephrase sentences

Line 152: combining

Line 231: where are these x_max and y_max?

In the equations the indices for ‘max’

Line 310: model training

Line 312: to normalize

Figure 4 (a), (b) labels too large: decrease font size

Generally the font size for the equations too large

Line 375: comprises 49 convolutional

Figure 5 label: use present tense

Figure 7: a) Visualisation of original SATL03 footprint data

Figures 9, 10, 11: labels (a), (b) .. outside graphs

Line 581: rephrase

 

Comments on the Quality of English Language

The quality of English presentation needs to be improved.

Author Response

We feel great thanks for your professional review work on our article. With your valuable feedback, we were able to improve the quality of our manuscript. The revisions can be seen in the attaches.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

 

 Abstract

Please remove the sentence “Computer Vision (CV) technology can effectively cluster and interpret regulatory patterns in accessible images of targets, 14 enhancing efficiency and intelligence.” If the sentence is important, you should be precise.

Please avoid using (we, our, and us), use the passive voice. Please check the paper.

The abbreviation “profile raster images of footprints (PRIF)” should be written as “Profile Raster Images of Footprints (PRIF)”, please check all abbreviations in the paper.

“to serve as research target” is not an acceptable reason. Please cancel it.

The expression “footprints profile” is not clear in the abstract, the reader will be forced to throw the paper or go inside it to understand it. The role of the abstract is to summarize the paper and facilitate the reader to know if this paper is what he is looking for.

You said: “The findings highlight the significant potential of the TSNN in estimating diverse and accurate forest structural parameters using PRIF of spaceborne LiDAR footprint data.” This sentence is not logical and vague.  What are the structure parameters, do you mean the classification of a forest point cloud? 

The abstract is long and not precise, please rewrite it.

The paper title matches the abstract; I will read the paper and see.

 Introduction

This section represents a mixture of introduction and related works. Even though the paper topic is a hot research spot and having a particular importance, there is no cited references 2024 such as :

Tarsha Kurdi, F., Lewandowicz, E., Shan, J., Gharineiat, Z. 2024. Three-dimensional modeling and visualization of single tree LiDAR point cloud using matrixial form. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 17, pp. 3010-3022, 2024, doi: 10.1109/JSTARS.2024.3349549. 

Binbin Xiang, Maciej Wielgosz, Theodora Kontogianni, Torben Peters, Stefano Puliti, Rasmus Astrup, Konrad Schindler, Automated forest inventory: Analysis of high-density airborne LiDAR point clouds with 3D deep learning, Remote Sensing of Environment, Vol 305, 2024, 114078, https://doi.org/10.1016/j.rse.2024.114078.

 Line 80: what is the point density measured by “ICESat-2”?

Line 89: “Martin compared the accuracy” Please follow the author's guideline rules for reference citation.

Line 98: the word “processing” is a general term, please replace it by an accurate expression. Please check all the text.

Please highlight the novelty and the contribution in the paper.

Materials and Methods

Please don’t put two section titles consecutively, you must add a transition paragraph between them, please check all the paper.

 Study area

Table 1 is important, please explain the meaning of “Along-track photon spacing”.  Also, what do you mean by “Number of beams”?

Please redefine the abbreviations used in Figure 2 under it, please check all figures.

Tables 1 and 2 should mention the data accuracy values.

Till now the Forest's structural parameters are unknown.

Do you mean by “range of signal footprints” the 3D LiDAR point cloud measured by satellite?

Lines 215 to 224: you cite a list of operations and you consider them as well-known, you should explain them or add references. Moreover, you should define variables and parameters used in them and present the result of the application of each of them.

The added figure resolutions are weak, please improve them.

Conclusion

Please discuss the limitations of the suggested approach.

 

 

 

Comments on the Quality of English Language

 Minor editing of the English language is required.

Author Response

We feel great thanks for your professional review work on our article. With your valuable feedback, we were able to improve the quality of our manuscript. The revisions can be seen in the attaches.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I am satisfied with the response. I think this manuscript is ready to be published.

Reviewer 4 Report

Comments and Suggestions for Authors

The paper looks much better.

Comments on the Quality of English Language

 Minor editing of the English language may be required.