Assessing the Predictive Power of Democratic Republic of Congo’s National Spaceborne Biomass Map over Independent Test Samples

Round 1

Reviewer 1 Report

Review of manuscript "Assessing the predictive power of Democratic Republic of Congo’s LiDAR-based biomass map over independent test samples"


In general, this is a quite well organized study with clear problem and methods. One big misleading issue for reader is, however, calling the map "LiDAR-based" as [18] clearly specifies, that only ALS transects were used and overall predictions were made using satellite data.


I would like to ask authors to carry out some little tests to estimate how could sample plot location error influence uncertainty in your validation experiment.

There is also not fuly clear of what were uncertainties in field plot area.

Are there any overlaps with lidar transects used in [18] and NFI sample plots?

In [18]  "We used 92 1-ha forest inventory plots (Supplementary Table S1) located in approximately 15% of LiDAR transects to develop the model."
:As the authors of this manuscript and [18] are partially the same I  assume that it would be  possible to compare also AGB distributions of NFI plots used in this manuscript and the field plots used in [18]. It is possible that some of the systematic differences can be described  by the distributions.


*  Detailed comments.  More can be found as annotations in pdf-file. *

: The title is misleading.
In [18] is:
(1) lidar random transects and
(2) MaxEnt produced estimates of AGB at 100 m (1-ha) spatial resolution from Landsat, ALOS PALSAR, and SRTM data.
: So you are not correct here and title is misleading.


R24-25
Why did R2  decrease when perturbed plots were removed?


R86-98
In title is "LiDAR-based". [18] tells that (1) lidar random transects and then (2) upscaling.
So, what are you investigating here - the lidar transects or the upscaled map for whole DCR?
Please check the entire manuscript to avoid any confusion.


R219-R220
In [18] was used a different AGB allometric model than in your study. Please analyse, how this influences your validation results and conclusions.


* Supplementary material *
Table S1. Relationships between field- and map-derived plots aboveground biomass by land cover class. Perturbed plots were removed from the analysis.
: Define the abbreviations in table heading or footnote.


Table S2. Mean biomass estimation (in Mg.ha-1) by stratum using field sample plots and the biomass map.
: Define the abbreviations in table heading or footnote.
:: Number of plots seems to be 470. Are perturbed plots excluded? Specify that in the table heading.

Comments for author File: Comments.pdf

Author Response

Please see attached.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript "Assessing the predictive power of Democratic Republic of Congo’s LiDAR-based biomass map over independent test samples" is submitted as a full research article and deals with a very interesting subject of great interest of the vegetation community, mainly those who are interested wih the retrieval of biomass. Although the subject and content fits well with the aim and goals of the targeted journal, I do not see any visual products that can be generated from the reported datasets. Although there is a strong field work component reported, some basic information about the biophysical parameters are also not reported. For instance, what was the plot size and which criteria. Is this size the same for the different vegetation types/physiognomies? What was the basic statistics coming from such measurements. Would it be possible to represent them over the landscape and also with the topograpic layer/information? 

Additionally, how was the remote sensing data acquired? Does it fit with the ground measurements or is there a delay between measurements? How it was processed? handled and extracted? What about the match between the field and remote sensing data? any possibility to show a 3D figure of the sample plot over different landscapes? interestingly, no description of the data variability such as number of species and min and max height, dbh, etc were not reported, as well as the care of dealing with giant trees for the extrapolation of the parameters from plot to hectare. 

how many  LIDAR metrics and features were extracted and related? over the landscape? L444 mentioned about comparison, but too weak in terms of figures and visual comparison, even with the landscape one (L501) that are also not shown. There is also no section regarding main conclusions. 

Therefore, I would recommend "major changes" and another review round given improvements are made in the revised version. The reproducible research aspect is weak and does not allow replication of the procedures. In summary, it seems more a conference paper for advertisement rather than a scientific paper.

Bellow some more specific comments:

L32: only REDD+? make in a broader context;

L36: there is a missing review about previous research trying to retrieve biophysical measurements over large areas and other mapping initiatives in tropical rain forest environments worldwide;

L53: please add other initiaves besides REDD+;

L109: better to list the two goals in a kind of bulleted list order;

L112: bring the figure after calling it in the text, not before; besides that, please increase a bit the font size of the legend and graphic scale of 1B;

L122: reference?

L125: reference? main characteristics (please describe);

L132: give the area of the plot;

L142: it means Fig1B? (please add the mention)

L150: cite number;

L159: please cite the area of the sample plot and strategy;

L228: please detail a bit more;

L230: cite author´s name;

L237: you assume that or you evaluate it? just change the sentence;

L253: please detail better all the LIDAR procedures and handling (starting from the adquisition, data handling and processing);

L282: please detail the procedure to allow replication;

L328: please consider adding a methodological flowchart; add a mention later on to recommend "spatial correlation" for future studies;

L330: please present the data (for small subsets) before starting with the analysis;

L351, L373, ...: which map?

L499: regardign to the forest management aspect, would be such maps useful for local assessment (and more refined) sampling for more detailend ground true information? and sustainable forest management practices? and what about changes over time for monitoring initatives? 

L517: please add a mention of the area of that vegetation type;

L572: please add a conclusion section;

Author Response

Please see attached.

Author Response File: Author Response.docx

Reviewer 3 Report

The paper entitled “Assessing the predictive power of Democratic Republic of Congo’s LiDAR-based biomass map over independent test samples” reflects the development of applied research, the topic is interesting and the manuscript has an approach innovative. Yet, some small issues should be addressed. Thus, minor changes are recommended.

Comments

1) Along the text – perturbed or disturbed?

2) Along the text – biomass or above ground biomass. Please standardise.

3) Line 48 – height or eight?

4) Line 61 – Northern countries or Northern hemisphere countries?

5) Figure 1 – would better placed in 2.1 Study area

6) Figure 1 – If the names of the provinces were added to Figure 1b it would improve the understanding of the study.

7) Line 180 – (see next section) or (see 2.2.3)?

8) Line 283 – how many are the unperturbed plots?

9) Line 298 – above ground biomass was measured or estimated? From the text it seems that allometric functions were used to estimate above ground biomass.

10) Figure 4 – Please do not use acronyms in the figure legend.

11) Lines 414-416 – please revise English.

12) Line 517 – which authors?

Author Response

Please see attached.

Author Response File: Author Response.docx

Reviewer 4 Report

Review of the manuscript entitled "Assessing the predictive power of Democratic Republic of Congo’s LiDAR-based biomass map over independent test samples”

General comments:

Indeed, our understanding of forests' role in the global carbon cycle cannot be overestimated. Over the years, through effective policies, forest management has improved how we measure biomass (or carbon storage) and its variability at different spatial-temporal scales. Traditionally, one way to estimate biomass (and subsequently carbon) is the destructive method (i.e., cut a tree down and weigh it). This method is highly discouraged as well as impractical to scale up spatial-temporally. In recent times, researchers can now generate global inventories of forest biomass using remote sensing data from different sources (e.g., optical, radar, microwave, lidar, etc.). Researchers link allometric equations with remote sensing information of tree measurements (such as height, stem diameter, canopy metrics, etc.).

Despite this promising venture, according to the authors, an independent verification of LiDAR-based biomass modelling in Congo Basin using ground truth data is lacking in current literature. According to the authors, with availability of NFI data, this study current study was carried out. The Congo Basin (with unique bioclimate and geography) was used as a test bed. To validate publicaly available biomass estimates, the authors used ground measurement from a field campaign (i.e., ground measurements acquired from 2017-2018).

This manuscript contributes to the biomass estimation community.  I found this study to be very useful, relevant and interesting to the general journal readership. The article is easy to read (i.e., well written and organized). The problem is stated. I knowledge the limitations and challenges the study area presented (i.e., local conflicts, etc) and the institution sharing this data. The think the sampling plot of 477 is statistically adequate for the analysis of the current results.

The results are complemented with significant figures to help in results visualization, however, I found some key areas in the manuscript that were not adequately described or missing. This raises a couple of questions, especially about some aspects of the analysis. I want the authors to address my comments in their responses. The manuscript will be suitable for publication after the authors have addressed the following comments and questions.

Major comments:

1.       The main concern I have about the paper is the biomass products used. It is unclear to me if the biomass data is a product of GEDI products collections. I did not find much information about this product (except in lines 83-95, 240-242)and how the data curators processed it. It is very important for the authors to provide readers about how the LiDAR data were processed (because it ends up being such a critical step in your study). Is the biomass data used in this study publicly available? If yes, is it part of GEDA products? Authors need to clarify and provide information where necessary, about (i) when the LiDAR was acquired, sources, (ii) how the processing of the LiDAR data was achieved, etc. (iii) How were gaps in the data handled? This is important, hence the authors' study of uncertainty metrics. Other associated issues include quality flags and model inputs (including scaling and choice of algorithm, etc.).

2.       No citation on the LiDAR equipment used by the data curators was provided. What is the sensitivity of the processing approach adopted to process the LiDAR data? Did the data curators mention it? If so, how will that impact your results? I would guess that it will be quite highly sensitive to say, for example, vegetation height used for processing (especially as the definition of max vegetation height varies across literature: see Dubayah et al 2021, Neigh et al. 2013, and Yavasli et al. 2016). I am not convinced about the LiDAR data presentation in the methods section.

3.       Alternatively, the authors do not describe the tools, libraries, or code the data curators used to create the BIOMASS package. I think it is the author's responsibility to make the reader feel confident about the paper's methodology.

4.       . Please, can you provide briefly any theoretical explanation of the H:D relationship?

5.       Please provide a workflow chart. It will help readers to understand at glance how the work was carried prior to reading.

Minor comments

Line 48: height?? Do you mean “eight” fold??

Line 61: “Northern countries”?? Do you mean “Northern Hemisphere countries”

I suggest the authors re-write long sentences. Also, in many parts of the manuscript, the authors made few speculations (lines 455-560). I suggest the authors focus on what was done in the study.

Line 582: Acknowledgments: Was there a reason why the authors did not acknowledge the LiDAR data team for making available their datasets for this research? I am curious.

References:

Is there a reason why the following citation was not cited?

Dubayah, R.O., J. Armston, J.R. Kellner, L. Duncanson, S.P. Healey, P.L. Patterson, S. Hancock, H. Tang, M.A. Hofton, J.B. Blair, and S.B. Luthcke. 2021. GEDI L4A Footprint Level Aboveground Biomass Density, Version 1. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/1907

Cite the following:

https://doi.org/10.1038/s41558-020-00976-6

 https://doi.org/10.1016/j.rse.2012.02.023

https://doi.org/10.1016/j.rse.2013.06.019

https://doi.org/10.1016/j.rsase.2016.11.004

Author Response

Please see attached.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thank you for exchaustive explanations.

Please check

Table S1. Biophysical parameters of forest samples plots from the National Forest Inventory. N stands for the number of trees,

: for what area?  Please update.

Table S2.

: There seem to be points instead of commas.

Author Response

Please see attached.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript "Assessing the predictive power of Democratic Republic of Congo’s national spaceborne biomass map over independent test samples" is a revision of a previous manuscript submitted to the esteemed Remote Sensing journal. After reading the rebuttal letter and the revised version of the manuscript I confirm that the majority of the comments were implemented and that the changes now reflect better what has been done and not done in the research. Several sentences and in different parts of the manuscript were changes and allow now a better overview of the methodological procedures made by the authors and also those steps coming from others. The assessment of the DRC’s national spaceborne AGB map using field inventory data from the first country’s NFI brought some perspectives and insights for the forthcoming large scale mapping initiatives. I saw now the content and contribution with another perspective and confirm that it is suitable for publication now. 

However, I saw some very minor issues that can be corrected during proofreading such as:

i) description of the abbreviations such as L243: GEDI, BIOMASS; L248: LIDAR and so on;

ii) L261: addition of a short sentence describing the importance of such independent validation assessment in large scale mapping initiatives;

iii)  adding author name before the reference number such as L318/421: Réjou‐Méchain et al. [20]; L328/862: Xu et al. [18], etc;

iv) in the discussion adding what woud be the required standards and requirements in such independent validation initiatives in order to allow the adoption of the AGB map in other environmental initiatives;

In the second read of the revised manuscript, I found some isolated sentences that requires a final and carefull proofreading of the manuscript. In summary, I appreciate the content and the efforts of the authors to clarify their strategy and research outcomes.

Author Response

Please see attached.

Author Response File: Author Response.docx