Fine-Resolution Forest Height Estimation by Integrating ICESat-2 and Landsat 8 OLI Data with a Spatial Downscaling Method for Aboveground Biomass Quantification
Round 1
Reviewer 1 Report
The authors develop a methodology to estimate the height of forest canopies with a high spatial resolution (15 m) based on the fusion of two remote sensing sources using a random forests down-scaling model. The product developed is relevant because of the urgent need to estimate the amount and dynamic of forest carbon stocks for adaptation and mitigation measurements. The study make an approach to produce a spatial continuous aboveground biomass map of forests that is a crucial step for the estimation of forest carbon stocks. The fine resolution achieved is applicable for regional forest management and conservation strategies. The area selected for the application of the product is a region of Northwest China that combine three main types of forests: a natural mixed conifer forests, a secondary deciduous broadleaf forest and forest plantations. The region has a high potential for the validation of the model developed due to the diversity of forest formations in the area.
Landsat 8 and ICESat-2 synergies has being used in the last years in remote sensing for quantifying vegetation cover, land uses and biomass. The fine spatial resolution of the canopy height modelled is the main contribution of the paper.
Taking into account the fine spatial resolution of the modelled canopy height achieved, the estimation of the AGB is very coarse because the authors used a single allometric equation for all the forest types in the rich study area. The allometric equation was develop previously by Huang et al (2019). If the specific goal of the study is to develop a FINE-RESOLUTION estimation of AGB, it should be used at least allometric models for the three main forest types. This could enrich the validation and interpretation of the results obtained. If the authors maintain the goals of the study (fine resolution AGB estimation), they should re-define the single allometric model for calculating AGB to a more detailed one.
Specific comments
Title: “Fine-resolution forest aboveground biomass estimation by integrating ICESat-2 and Landsat 8 OLI data with a spatial downscaling method”. I would propose to change the title to “Fine-resolution forest height estimation by integrating ICESat-2 and Landsat 8 OLI data with a spatial downscaling method for aboveground biomass quantification” more adjusted to the product developed in the paper.
Abstract
Sentence 23-24: “using an allometric model according to the FOREST TYPE and height” is confusing because seems that the allometric model used vary by forest type, which is not the case in the paper because a single allometric model is used for all forest types. Please rephrase the sentence.
Introduction
Line 117: Please be consistent with terminology used in the text. Do not use the term “reference data” for the forest height derived or extracted from ICE- 16 Sat-2 as in Line 117. Reference data are the described in the 4.4. section: CFAM, GFAM and GEOCARBORN data.
Line 109: change “fine forest AGB estimation” to “fine spatial resolution forest AGB estimation”.
Line 121-122 “our main contribution is developing a new feasible method to estimate spatial forest AGB with fine details” fine details is a very general consideration, “fine resolution (15 m)” could be more specific. The main contribution of the authors is to develop a methodology for estimating “forest height with fine spatial resolution”, but not the AGB. the AGB calculation is performed by using a single previously developed allometric equation.
2. Data and methods
Figure 1: the Flowchart figure is clear with the exception of the “30 m forest height product” on the top. Please improve with arrows its origin. It is not clear where this product come from. Is it not explained in the method section.
2.2 Variables derived from the Variables derived from the Landsat 8 OLI images.
Sentence 195 is not appropriated. For AGB modelling, should be selected independent variables derived from the optical images that help or complement to estimate spatially continuous the vertical structure of the canopy but not their HEALTH LEVELS like pointed out in line 195. Band reflectance in the optical domain is considering a fundamental variable for estimating horizontal vegetation structure such as vegetation canopy cover or forests types’ identification [33].
2.3. Spatially continuous forest height modelling by Random Forests
Line 216: As mentioned before, it is not clear how the “30 m forest height products” mentioned in line 216 is derived.
In line 260 the authors spoke about “REAL canopy height derived from ICESat-2”, this is confusing if the canopy height data are derived could not be considered real data. Please erase the adjective real. The same consideration for the line 263: “? represents the mean of all REAL canopy height values”.
2.4 Allometric model for AGB estimation
Sentence lines 268-269: “the forest height occupied approximately 87% of stem biomass” is not clear. Please rephrase it.
Given the three forests types of the Jilin area cited in the study area section are:
· Natural forest of mixed coniferous especies
· Secondary forests of mixed deciduous especies
· Plantations of two species: one deciduous coniferous species (larix) and one broadleaf species (populus).
The authors have used a general allometric equation for all forest types, considering the “a” and “b” coefficients given in reference 45 for mixed forests in Northeast China. This generalisation implies that all forests species have the same wood density that is not true. Having specified in the reference 55 the “a” and “b” coefficients for Northeast China forests separately for evergreen conifer forest (ECF), that could be associated with the natural forests and deciduous broadleaf forest (DBF) for the secondary forests, adding an equation for plantations it is a pity to use in the study a generalized equation.
3. Study area
· Line 282 substitute please 1.87x105 km2 for 187.400 km2 to help the reader.
· Figure 3: Please change the Jillin Province map that give not information about the forest cover, for a forest map of the province. If the map would show the surface of the three main forest types (natural, secondary and planted) would be very informative.
· Units with superscripts letters have to be changed (such as km2 by km-2)
· Forest species names should be written in italic characters.
4.1. Importance evaluation of variables
In line 318-19 the authors refer to a “spectrum band”: Could you give a precise description of the spectrum band you are speaking of?
Line 319-322: “The main reason for this result is that the fine resolution (e.g., 15 m) of spectrum reinforces observation details and is suitable to capture the reflectance contributions for forests, which might be confused by grass in moderate resolution images” the sentence is not clear. The methodology used in the paper, discard vegetation heights under 2 m, therefore grass contamination is not possible. Please make it more clear.
In line 334 the authors state that in the lower height range (0-5m) the NIR band, that present the second importance score, is because is related to vegetation health. How the authors explain the relation between canopy height and vegetation health?
Do correspond the forests with canopy height in the range 0-5 m to plantations?
4.2. Model performance evaluation
Figure 6: Please the title of x-axe in the Figure 6 in confusing, using the same description as in line 352 change “Estimated height” for “RF-based Height” to help the reader.
4.3. Spatial forest AGB estimation
In line 378 of figure 7, please change the word “forest height” for “RF-based forest height” to help the reader.
In Figure 8 change please in the legend the units from “mg/ha” to “Mg/ha”.
Line 371: please give the real range of AGB: “minimum AGB value-354.56 Mg/ha”, and the 25 and 75 % percentile values or 5-95 % percentiles.
Please compare your estimated ranges and percentiles with data in the literature for example with the obtained in reference 45 among others. Are there in the same range for the same forest types?
4.4. Accuracy assessment of forest AGB map
In Figure 9 Please, change de y axis scale to 50-250 Mg/ha in all 9 figures to be able to visualize the 1:1 line. The same for figure 10, set please the upper y-axe limit to 180 like in the x-axe. Moreover the units of AGB are missing in both figures in the caption. You could perhaps change in the x axe “Ours” for “Our study”.
Even though the regression coefficients with the three reference datasets selected are high, the data of the present study in Figure 9 overestimate the AGB values in all cases compare with the 3 reference datasets, especially in the upper AGB range. Please explain in the text possible reasons.
The paper is well written. Introduction and method sections are clear and good explained. References are relevant and actual. The paper is good strutured.
Author Response
Dear reviewer,
Thank you for giving us an opportunity to revise our manuscript. We appreciate you very much for your professional comments and constructive suggestions on our manuscript entitled “Fine-resolution forest aboveground biomass estimation by integrating ICESat-2 and Landsat 8 OLI data with a spatial downscaling method”. (ID: forests-2436828). These comments are all valuable and helpful for improving our article. All the authors have seriously discussed about these comments. According to your comments, we have tried the best to modify our manuscript to meet with the requirements of the journal. Moreover, we have revised the grammar problems and checked the draft text with the support of a professional editing service (http://www.kgsupport.com/). In this revised version, changes to our manuscript within the document were all highlighted by using the "Track Changes" function in Microsoft Word. Point-by-point responses to the comments are listed in the attachment.
We have tried our best to improve the manuscript and made some changes to it. These changes will not influence the content and framework of the paper. We appreciate for your warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Sincerely,
Xudong Hu
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China
Email: [email protected]
Author Response File: 
Author Response.pdf
Reviewer 2 Report
General Comments:
1. The Jilin province comprise many other land uses than forest. However, the assessment was made throughout the province. Suggesting to delineate the forest cover first before conducting the AGB assessment.
2. Wondering why the Landsat image variables were used to estimate height rather than directly to AGB. Suggesting to incorporating all datasets both optical and ICESat. as predictor variables and use reference datasets for training and validation.
3. Since the study does not include ground data, it is worth to include some summary of the profile of forests within Jilin province. Some pictures will help.
4. ICESat data cover mostly in west part of Jilin, but literally there is not much forest within this area. Any justification on this?
5. Should include the original data of Landsat image over the study area as well, in natural or false color composite.
6. Training datasets used for modelling is not elaborated clearly. Figure 6 what does the "Estimated height" is referred to, and how it was obtained.
Specific Comments:
Figure 1: 15m forest height, 30m forest height products - what are these referred to?
Figure 5: Consider revising the graph format, difficult to understand what does it means, not representative.
L134: forest type? where this is explained.
L275: power function or coefficient?
L282: 1.87 X 105 km2 - revise the numbering format and the unit.
L290: Species name should be written in Capital for genus and italics for species name
L313: Why relate image variables to height? instead of AGB. Refer General Comments
Read the manuscript once again as there are some English need to be improvised.
Author Response
Dear reviewer,
Thank you for giving us an opportunity to revise our manuscript. We appreciate you very much for your professional comments and constructive suggestions on our manuscript entitled “Fine-resolution forest aboveground biomass estimation by integrating ICESat-2 and Landsat 8 OLI data with a spatial downscaling method”. (ID: forests-2436828). These comments are all valuable and helpful for improving our article. All the authors have seriously discussed about these comments. According to your comments, we have tried the best to modify our manuscript to meet with the requirements of the journal. Moreover, we have revised the grammar problems and checked the draft text with the support of a professional editing service (http://www.kgsupport.com/). In this revised version, changes to our manuscript within the document were all highlighted by using the "Track Changes" function in Microsoft Word. Point-by-point responses to the comments are listed in the attachment.
We have tried our best to improve the manuscript and made some changes to it. These changes will not influence the content and framework of the paper. We appreciate for your warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Sincerely,
Xudong Hu
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China
Email: [email protected]
Author Response File: 
Author Response.docx
Round 2
Reviewer 1 Report
The revised manuscript has improved considerably making the results of the study much more clear, applicable and attractive.
The authors have re-calculate the main results of the study rapidly, making the conclusions more robust. The new splinting in four allometric models increase the fine resolution of the results obtained.
I see the paper now prepare for publication. I wish the authors all the best for their future work because the monitoring of aboveground carbon stocks in forests is a crucial challenge for forest management.
Reviewer 2 Report
No further comments
