the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 May 2024
Reviews and syntheses: One forest carbon model to rule them all? Utilizing ensembles of forest cover and biomass datasets to determine carbon budgets of the world’s forest ecosystems
Abstract. Understanding global forest carbon stocks is necessary to assess the world’s global carbon budget, with land cover change being estimated to contribute roughly 20 % of the emissions of greenhouse gases to the atmosphere. In the last decade or so, remote sensing has contributed estimates of above ground stocks of biomass – a key part of forest carbon stocks – with over twenty biomass maps available at pan-tropical and global scales. To further the understanding of forest carbon stocks, this research seeks to synthesize the findings of disparate data sources on: (i) forest cover, (ii) forest cover change, (iii) above ground biomass (AGB) / above ground carbon (AGC) stocks in forests. Satellite-derived forest cover and AGB estimates have substantial variability. In 2020, forests were estimated to cover between 22.6 million and 49.7 million km2 of the Earth’s land surface, thus ranging from 17.1 % to 37.6 % of total land cover. Likewise, examining forest cover change from available datasets, the estimated change in global forest cover between 2000 and 2020 was loss of approximately 88,734 to 124,184 km2 per year, combined with regrowth of forest cover of approximately 58,628 to 169,912 km2 per year. Combining that forest cover data with remotely sensed AGB estimates, total stocks of AGB for the year 2000 were estimated to be 325–697 Gt, while for the year 2020, the range was 401–580 Gt. The equivalent quantity of CO2 (i.e., CO2e) of that stock of forest biomass was therefore estimated to be 560 to 1,200 Gt for the year 2000, and 692–999 Gt for the year 2020. Our analysis found that the forest cover loss in tropics was the largest, at the rate of 1.4 % to 3.5 % net reduction between 2000 and 2020, whereas for the same time period, the temperate and boreal zones showed substantially lower forest cover loss (-2.5 % to 0.5 % and 1 % to 5.3 % respectively). This synthesis paper demonstrates that there is a fairly wide range of variability in estimates related to forest cover, forest cover change, and above ground biomass stocks, which are the main inputs for estimating forest carbon stocks and greenhouse gas emissions from land cover change.
- Preprint
(2775 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2024-1179', Anonymous Referee #1, 17 Jun 2024
General comments
This manuscript compares global and nearly-global datasets of land cover, forest cover and aboveground biomass and synthesizes information to understand their relevance. The work of collecting all these datasets and comparing them is per se tremendous and the analyses undertaken by the authors provide some interesting numbers. Nonetheless, the major issue with this manuscript is that they provide a long list of numbers. These numbers diverge and the conclusion is that there is not a single model, which was already known before. I would have appreciated if the authors had moved beyond synthesizing datasets and provided map producers with some guidelines of dos and don’ts from their perspective. In addition, there is no mention of the methods followed to synthesize data so I cannot judge whether these might have impacted the numbers that were derived from the maps. I also find the text superficially written (generic statements, varying numbers). Although, I believe that the manuscript has some scientific merit, it is currently not meeting the standards for publication in a peer-reviewed journal.
Below the authors can find a list of comments that they may consider should they pursue resubmission to another journal.
Specific comments
- The title seems to promise something that was not fulfilled by the article. What is a “forest carbon model” exactly?
- What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?
Lucas, R.M., Mitchell, A.L., Armston, J., 2015. Measurement of Forest Above-Ground Biomass Using Active and Passive Remote Sensing at Large (Subnational to Global) Scales. Current Forestry Reports 1, 162–177. https://doi.org/10.1007/s40725-015-0021-9
Rodríguez-Veiga, P., Wheeler, J., Louis, V., Tansey, K., Balzter, H., 2017. Quantifying forest biomass carbon stocks from space. Current Forestry Reports 3, 1–18. https://doi.org/10.1007/s40725-017-0052-5
- What is your definition of “forest”?
- What is your definition of “biomass”?
- 4 is not introduced in the text.
- 5. Using values published by FAO in their resources assessment could help as reference to understand it. However, not knowing how each of the land cover datasets was re-labelled to forest/non-forest AND how maps were resampled to 1 km does not help to understand the differences in this figure. Actually, comparing maps of land cover at 1 km is introducing a lot of uncertainty in the results (too coarse resolution).
- The comparative analysis in Section IV is not really addressing the results from the comparison but providing a high-level presentation of concepts and potential explanations. In particular, the paragraph on lines 376-387 must be reconsidered. If one embarks in a study like the one envisaged in the title, one should also provide the answers. To me, these answers were not provided.
- The Assumptions in Section IV should also be reconsidered as they very much reduce the importance and the merit of the study.
- The purpose of the Caveats in Section IV is unclear. It is currently a mixture of speculations, work already published by others (lines 435-445) and possible work that was not undertaken (Lines 456-461).
- The Implications in Section IV could profit from some re-writing as well. 1) The authors should not ask questions (lines 469-474) but provide some indications. 2) Lines 486 – 492 are not related to this study.
Technical corrections
- Line 17: please correct “to further the understanding”
- Line 45: “15 years or so” were “last decade or so” in the abstract, “dozen sources” here were “20 sources” in the abstract. Please be precise and consistent.
- Line 46. Use “published” instead of “developed” since a few datasets have not been published.
- Lines 45-51. Near-field remote sensing is missing here (ALS, TLS).
- Line 54. Studies do not “advocate” but try to emphasize what is new to existing knowledge. As time goes by, new research demonstrates what the previous studies did not or could not consider.
- Line 70. Previous studies should be referred to.
- Line 72. “Implications” were not mentioned in the abstract, why?
- Lines 75-96. Move to the Introduction
- Lines 101-108. Why did you create your own classification since there are already official classifications around such the FAO Global Ecological Zones dataset, which is the basis of the IPCC classes as well? Classifying zone based on temperature can be misleading. Looking at Figure 2, Tibet for example belongs to the boreal zone, which is incorrect.
- Line 114. What is the impact of not distinguishing “tree” and “forest” cover on your study?
- Lines 120 and 129. The Hansen dataset is not a land cover dataset.
- Line 130. The JAXA FNF dataset is not land cover.
- Line 134. What does “For the most part” mean?
- Line 138. I would strongly suggest to avoid the Liu and Xu datasets at all because of the completely different spatial resolution.
- Line 139. What “statistical analysis”?
- Line 142. Where is this number “16” coming from considering that before you mentioned 22 and 29 datasets?
- Line 143. “were resampled” how?
- Line 145. Are you seriously thinking that nearest neighbor does not alter the statistics of the maps when going from 30 m to 1000 m?
- Line 146. Why resampling to the Mollweide projection? To my knowledge most (all?) datasets were in lat/long projection. Why not keeping this one?
- Line 150. The focus on the pan-tropical zone was already mentioned.
- Line 151. What is the “first level of analysis”?
- Line 151. And here we have “19” which is a new number apparently.
- Line 152. Shouldn’t it be 2010 instead of 2019?
- Line 153. What does “did not skew the analysis” mean here?
- Line 164. What are “widely accepted methods?” This should be made clear.
- Line 169. And here we have “11” datasets. Plus 3 one line below makes 14, again a new number.
- Lines 169 and 171. You may want to say “combinations” rather than “permutations”.
- Line 171. “those data”?
- Line 172. What does “tuning into nuances in the data” mean?
- Lines 174-176. Is this relevant?
- Line 180. Figure 3. The flowchart is redundant since what is done here is purely zonal statistics.
- Lines 184-190. Totally unclear what is done here.
- Lines 193-194. Text is redundant.
- Line 202. What are the “two outliers”?
- Lines 233 and 234. Where are these rates of km2/year from?
- Lines 244, 246 and 247. Are these % values for the totals?
- Figure 6 and Line 261. The reference should be to Santoro et al., 2021, ESSD.
- Figure 7. Fonts on the x-axis are too small. By the way, the Geocarbon dataset is a blend of 3 datasets: Saatchi et al (year 2000), and Baccini et al (year 2007) for the tropics and Santoro et al (year 2010) for temperate and boreal zones. Associating Geocarbon to 2000 can be misleading.
- Line 275. Why “As to be expected?”
- Line 279. This belt contains temperate rainforest (southeast Australia and Tasmania)
- Lines 280-282. Already mentioned, redundant.
- Lines 284. Your delineation does not correspond to the delineation adopted by IPCC.
- Line 294. Figs. 8-11 do not add information compared to the presentation of results with histograms unless it is explained in the text what these figures add with respect to Fig. 7.
- Line 304. With “models” do you refer to “estimates” perhaps?
- Line 304. Please reconsider this statement. A disagreement of 100% for an AGB = 1 ton/ha is less relevant than a disagreement of 20% for an AGB of 400 tons/ha.
- Figure 11. Is a CV based on 3 numbers a reliable figure for the level of agreement between maps?
Citation: https://doi.org/10.5194/egusphere-2024-1179-RC1 -
AC1: 'Reply on RC1', Emil Cherrington, 10 Jul 2024
Dear reviewer # 1,
First off, thank you for accepting the review invitation and for taking the time to review our manuscript. Thank you also for the suggestions regarding revisions. On behalf of my co-authors, I would like to respond point-by-point to the feedback you provided. Below is the response to the initial summary you provided.
= = = = =
“This manuscript compares global and nearly-global datasets of land cover, forest cover and aboveground biomass and synthesizes information to understand their relevance. The work of collecting all these datasets and comparing them is per se tremendous and the analyses undertaken by the authors provide some interesting numbers.”
We agree. Per some of your later comments, we definitely consider that assembling, processing, and analyzing that data to extract key findings is the very spirit of “synthesis,” in line with BioGeosciences’ indication that “reviews and syntheses summarize the status of knowledge and outline future directions of research within the journal scope” (https://www.biogeosciences.net/about/manuscript_types.html). We have assembled and generated over 100GB of spatial data, which we interpreted in a consistent framework to provide insights on above ground biomass - and forest carbon by proxy.
“Nonetheless, the major issue with this manuscript is that they provide a long list of numbers.”
Per the scope of our study, which we state in lines 70-76, i.e., to “dig deeper into those datasets…” and “to characterize global forest carbon stocks.” We believe the manuscript does provide analysis to put the list of numbers in context.
“These numbers diverge and the conclusion is that there is not a single model, which was already known before.”
When you indicate that “[it] was already known before,” how was this already known before? We note that you later suggest that we review Lucas et al. (2015) and Rodriguez-Viega et al. (2017), per your question, “What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?”
We appreciate that you suggested that we also review and incorporate insights from Lucas et al. (2015) and Rodriguez-Veiga et al. (2017). In our perspective, both were insightful studies, and we plan to include them in our manuscript revision. While the impression left by your review is that both papers have essentially already addressed the issues we have addressed in our recent manuscript, please note that (1) the scope of our study differs substantially from the scope of both papers, and (2) in the intervening seven and nine years since the publication of both studies, the body of knowledge on above ground biomass has increased substantially. Please allow us to explain.
- The review suggests that we should have focused on providing feedback to the “map makers” about increasing the accuracies of their products. We would like to mention that it was not the scope of our study. Our scope was to understand what the products of 16 studies on global-scale and tropical-scale biomass imply, from an applications perspective. Neither of the studies you mentioned focus on that topic, as their scopes were different. In terms of what was already known prior to our study, one particular study, Zhang et al. (2019) sought to synthesize the state of knowledge by documenting the various global, regional, and national above ground biomass studies [and accompanying datasets] which existed. While we found that study enlightening - as it introduces practitioners to the data that are available - we were left wanting to understand the implications of varying datasets. Or to put it another way, what do the various above ground biomass (AGB) datasets tell us about global and regional carbon budgets with respect specifically to AGB, which is a component of such budgets? If we should have stated that explicitly within our manuscript, we ask your indulgence and would be more than happy to state that in the manuscript’s revision. And regarding your sentiment that we are restating what is already known, we contend that previous studies have not analyzed and dug into the results of the 16 global scale and tropical scale studies. So, in that regard, we beg to differ. From our perspective, we saw a gap and put effort into answering a question that had not been answered before.
- The body of knowledge has increased substantially since the publication of Lucas et al. (2015) and Rodriguez-Viega et al. (2017), eight of the sixteen studies we have listed were published from 2020 onwards, meaning that Lucas et al. and Rodriguez-Viega et al. would therefore not have had access to these in their reviews. Half of the body of knowledge did not yet exist when the two studies you suggested were published, so their insights would therefore not have taken into account more recent knowledge. For instance, in trying to synthesize available knowledge on AGB stocks, Lucas et al. (2015) mainly based their findings on comparing only two AGB datasets, that of Saatchi et al. (2011) and Baccini et al. (2012). From the perspective of quantifying AGB stocks, we appreciate that Lucas et al. (2015) did attempt to convey to their readers information on AGB stocks, although they did not go so far as to compare the total stock estimates derived from Saatchi et al. and Baccini et al.
Along the lines of your suggestion, Rodriguez-Viega et al. (2017) does provide suggestions on how accuracies of the various AGB products can be improved, and in their assessment, they examined five tropical-scale and global products (from Saatchi et al. 2011, Baccini et al. 2012, Avitabile et al. 2016, Liu et al. 2015, and Hu et al. 2016), in addition to citing Ruesch & Gibbs (2008) and Kinderman et al. (2008). Their work, however, leaves out 11 other AGB products which have been released since 2017.
Please note that we are not criticizing the scopes of Lucas et al. (2015) and Rodriguez-Viega et al. (2017), merely pointing out where their reviews of existing AGB studies in turn left gaps which we sought to explore with our own study. Since Lucas et al. (2015) and Rodriguez-Viega et al. (2017) evaluated datasets available at the time of their publication, we found it valuable to extend that body of knowledge with the latest AGB datasets as well as the latest body of knowledge. We respectfully argue that filling the gap left by previous studies is a valuable endeavor and remains the main focus and driver for this manuscript.
“I would have appreciated if the authors had moved beyond synthesizing datasets and provided map producers with some guidelines of dos and don’ts from their perspective.”
This manuscript is a review / synthesis, and we believe we remain in scope of the journal's manuscript submission guidelines. Thanks to your feedback, we will further strengthen our statements in the manuscript lines 70-76. The scope of our study is characterizing AGB and forest carbon, we are evaluating the current state of models and datasets available and we think it was not our place to suggest how the studies we reviewed could generate better datasets. We are merely trying to characterize the similarities and differences, we are not trying to help them improve their respective accuracies. That, however, would be a great but a different paper.
“In addition, there is no mention of the methods followed to synthesize data so I cannot judge whether these might have impacted the numbers that were derived from the maps.”
The aim of the manuscript is to identify publicly available datasets, methods, and explore how they converge (or not) for a geographic region. Our manuscript does indeed outline methods we evaluated as part of this work. In fact, we would like to thank you for your feedback and comments on the methods section of our manuscript. For example, we appreciate your comments about the clarification on the resampling methods and the comments about Mollweide projection. It will help us further strengthen our methods section.
“I also find the text superficially written (generic statements, varying numbers).”
We appreciate this feedback, but we would have appreciated specific references to where we could strengthen the manuscript.
“Although, I believe that the manuscript has some scientific merit, it is currently not meeting the standards for publication in a peer-reviewed journal.”
While we appreciate your candor, we believe that a manuscript that examines the publicly available datasets and methods across the globe and comparing the results obtained from that amalgamation is indeed a key function of a review/synthesis manuscript. We have attempted to do precisely that. With all humility, we believe no other publication has looked exhaustively at the available datasets, methods and presented the results in a cohesive form before. That, we believe, is the merit of this review manuscript.
While we deeply appreciate your perspective, your preference that we should have taken a different direction in terms of what we analyzed, in our humble opinions, does not make our findings any less scientifically valid. We purposefully do not provide a critique of the outputs of the 16 studies. We merely analyzed each of them in order to provide our own stakeholders with an understanding of the implications of each dataset. In the domain of our work, we collaborate with users who would like to know how much biomass is present in their area of interest, whether that is a country, province, or a protected area. Having 51 individual datasets from 16 different sources can be confusing for research communities and downstream practitioners alike. We see our study as one way to parse the data and provide guidance to users on the characteristics of that data.
“Below the authors can find a list of comments that they may consider should they pursue resubmission to another journal.”
“Specific comments”
“The title seems to promise something that was not fulfilled by the article.”
We would have hoped that you - as well as the audience in general - could see that we are asking a rhetorical question. Our conclusion is that the models diverge, so there is no one model to rule them all, to borrow from JRR Tolkien’s “one ring to rule them all.” We remain open to a different recommendation.
“What is a “forest carbon model” exactly?”
Thank you for this comment. By “forest carbon models,” we were referring to models of forest carbon stocks, which we have indicated in lines 174-176, are directly related to the models of above ground biomass (AGB). We will integrate your feedback to clarify the manuscript accordingly.
“What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?”
Please see our extensive response above regarding both studies.
= = = = =
“What is your definition of “forest”?”
Thank you for your comment. For the purpose of this study, “forest” is defined as areas mapped as “forest” or “tree cover” in the respective input land cover data sources used. Tree cover is treated as a proxy for forest cover, as indicated in line 114. We will strengthen the wording to make sure we make it clear to the reader.
“What is your definition of “biomass”?”
Thank you for the suggestion. Since we are focusing on “aboveground biomass,” we will go the definition provided by the IPCC, as in “all living biomass above the soil including stem, stump, branches, bark, seeds, and foliage” (source: https://www.ipcc-nggip.iges.or.jp/public/gpglulucf/gpglulucf_files/Glossary_Acronyms_BasicInfo/Glossary.pdf).
“4 is not introduced in the text.”
Thank you for pointing out our oversight. We assume that you mean Figure 4. We should have introduced Figure 4 in the same line where we introduced Figure 5. We will remedy it in the revised version.
“5. Using values published by FAO in their resources assessment could help as reference to understand it. However, not knowing how each of the land cover datasets was re-labelled to forest/non-forest AND how maps were resampled to 1 km does not help to understand the differences in this figure.”
Thank you for your comment. We assume that you are referring to Figure 4. You later comment that we should have provided detail on the statement in lines 163-164 that “all the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods.” We were trying to economize in terms of our word count in the Methods section, but we would be happy to provide technical annexes indicating which classes were classified to “forest” (essentially the ones in the maps originally going as “forest” or “tree cover”), and which went to “non-forest” (i.e., the remaining classes). The differences in the figure are brought about by the reality that the various source land cover datasets differ in what they identify as forest cover / tree cover, and that is what we are seeking to show here. And for what it’s worth, other studies like Venter et al. 2022 (“Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover”) have taken a similar approach of evaluating the differences among various global land cover datasets.
“Actually, comparing maps of land cover at 1 km is introducing a lot of uncertainty in the results (too coarse resolution).”
Thank you for your comment. While some will be introduced in majority resampling 10m, 25m, 30m, 300m, 500m data to 1km, it is unlikely to be substantial. We are working on providing additional analyses showing how resampling to 1km changed estimates of overall forest cover.
“The comparative analysis in Section IV is not really addressing the results from the comparison but providing a high-level presentation of concepts and potential explanations.”
Thank you for that observation. In this subsection, we tried to provide a focused analysis on how certain datasets converged, while others did not, e.g. in lines 339-369. We explored differences regarding forest cover change (lines 339-342) and regarding the overall forest AGB budgets that could be extracted from the datasets (lines 350-362).
“In particular, the paragraph on lines 376-387 must be reconsidered.”
Thank you for the comment and allowing us to reexamine how we are stating that section. We are trying to explain that it is not straightforward to analyze forest AGB given the various potential combinations of data and the differences among the datasets themselves. Therefore, rather than trying to say one or a few specific datasets are “correct,” which we explain is difficult to ascertain because of the lack of corresponding data from destructive sampling (per guidance from Duncanson et al. 2020), we propose that one way to understand the world’s AGB budget is to utilize an ensemble of the various AGB estimates (like how ensembles are used in weather forecasting). So rather than reconsidering that paragraph, in absence of other data, we see this paragraph as presenting a key consideration. We will reword the section for adding this nuanced clarification.
“If one embarks in a study like the one envisaged in the title, one should also provide the answers. To me, these answers were not provided.”
Again, we appreciate the candor. As mentioned in one of our earlier responses, we saw the title as us asking a rhetorical question. Is there indeed only one forest carbon (AGB) model? The obvious takeaway from the title itself is expected to be “No, there isn’t,” but our study explores differences among the various AGB datasets. The fact that there are substantial differences among the various available AGB datasets, the manuscript attempts to answer the rhetorical question.
“The Assumptions in Section IV should also be reconsidered as they very much reduce the importance and the merit of the study.”
Thank you for your perspective on that. We thought that for the sake of transparency that it was useful to highlight some of the assumptions we made, lest our analysis be criticized for not including said assumptions. Also, we believe that addition of these assumptions does not reduce the importance or merit of the study. For instance, we discuss that we resampled the data to a common reference scale (1km). Just the act of resampling a few of the higher spatial resolution datasets does not drastically change the findings of the study. Alternatively, we could have resampled all of the data to the finest scale of reference (10m), but that would also have improperly represented the spatial precision of coarser datasets and made the data volume extremely large and unwieldy. We thought it was important to include our assumptions for the sake of completeness of a follow-on, and potentially even critical, article
“The purpose of the Caveats in Section IV is unclear. It is currently a mixture of speculations, work already published by others (lines 435-445) and possible work that was not undertaken (Lines 456-461).”
Thank you for your comment. We believe we are putting our findings in the broader context of other published works. We are of the perspective that the Discussion is where one indeed contemplates the findings in a broader context.
“The Implications in Section IV could profit from some re-writing as well.”
Thank you, we appreciate this comment. For clarity, we will attempt to rewrite the single “implications” paragraph.
“1) The authors should not ask questions (lines 469-474) but provide some indications."
Thank you for the comment. The lines listed in these comments are in the discussions section, and the questions are pertaining to the comparisons with other published works. Per our earlier comment, we again thought it was appropriate to include our comments and speculations within the Discussion section of the manuscript.
“2) Lines 486 – 492 are not related to this study.”
While we appreciate the feedback, we consider that understanding standing aboveground biomass stocks to be an important part of understanding overall carbon budgets. In terms of a future research direction (the title of this section of the Discussion), we considered that a future study or studies could consider how our findings contrast with what is estimated by the respective FREL reports. Of course, that was not within the scope of the current study, but we think it would provide additional insights.
= = = = =
“Technical corrections”
"Line 17: please correct “to further the understanding”
Thank you. We will reword to better reflect our statement.
"Line 45: “15 years or so” were “last decade or so” in the abstract, “dozen sources” here were “20 sources” in the abstract. Please be precise and consistent."
Thank you for this comment. In our revision, we will rephrase these statements to be more precise and consistent.
"Line 46. Use “published” instead of “developed” since a few datasets have not been published."
Thank you for the comment. While most of the datasets are published, a couple are available online and do not have a citation to attribute to. So, instead of a blanket “published”, we chose to mention these datasets are available. We will add a clarifying statement to make sure this point is not lost. Just to clarify, we have cited all the available, published literature that we could find and we will attribute those not “peer reviewed and published”, we will add a clarifying statement. The data from the 16 global and pan-tropical studies we sourced are all public (they are all on the Internet), but indeed, a few of them have not been published in peer-reviewed papers. Hence, we thought mentioning that the data are available (i.e., have been developed) is less confusing than to say “published,” because it might give the impression that we meant, “published in peer-reviewed papers.”
“Lines 45-51. Near-field remote sensing is missing here (ALS, TLS).”
Thank you for the suggestion. We will be happy to include some text on this, likely drawing from Duncanson et al. (2020), the reference to which we should also have included in line 49, but which somehow we missed.
“Line 54. Studies do not “advocate” but try to emphasize what is new to existing knowledge. As time goes by, new research demonstrates what the previous studies did not or could not consider.”
Thank you, this comment made us realize that we should indeed change our wording to clarify. However, especially in light of multiple, recent differing estimates of AGB, with all due respect to the various authors, we have seen quite a few papers that seem to promote only their own dataset’s estimates. So in that sense, it does seem that such studies are advocating for their estimates above other ones. That was also one of the drivers of conducting this study. Or more specifically, one area in which we address this is in lines 456-461.
“Line 70. Previous studies should be referred to.”
Thank you for pointing that out. We should have explicitly mentioned Avitabile et al. (2016), Santoro et al. (2021), Santoro et al. (2023), and Zhang et al. (2019), and we will also include Lucas et al. (2015) and Rodriguez-Viega et al. (2017), which you suggested.
“Line 72. “Implications” were not mentioned in the abstract, why?”
Thank you for your comment. We will reflect on our abstract to include some inferences about implications.
“Lines 75-96. Move to the Introduction”
Thank you for this great suggestion. We will be happy to do so in the revision.
"Lines 101-108. Why did you create your own classification since there are already official classifications around such the FAO Global Ecological Zones dataset, which is the basis of the IPCC classes as well? Classifying zone based on temperature can be misleading. Looking at Figure 2, Tibet for example belongs to the boreal zone, which is incorrect."
Thank you for your comment. Please note that we did not create our own classification scheme. We merely applied the existing IPCC classification scheme to the most recent WorldClim precipitation and temperature data, and elevation data. And we would like to note that “classifying zone based on temperature can be misleading,” temperature data are indeed an input to the IPCC’s classification scheme. One reason for doing so, from our assessment, was that the IPCC climate classification maps (of which there are various sources) were based on older precipitation and temperature data, so we wanted to have an updated set of boundaries. As we have also mentioned in our Introduction and Methods sections, we focus on differentiating the data by the main climate zones mainly to follow on studies like Trumper et al. (2009) which presented data on biomass stocks as a function of climate type. Since a few of the datasets (e.g. Avitabile et al. 2016, Baccini et al. 2012, Dubayah et al. 2023, Saatchi et al. 2011) are mainly confined to the tropics, having a climate zone-based frame of reference was also useful. We will be careful to insert this fine point in the manuscript to make sure the next reader does not ponder on this topic.
"Line 114. What is the impact of not distinguishing “tree” and “forest” cover on your study?"
Thank you. We would like to point out that we do indeed discuss that in the caveats section, specifically lines 399-407.
“Lines 120 and 129. The Hansen dataset is not a land cover dataset.”
Thank you for your comment. We would like to point out that Hansen et al. did indeed release land cover datasets for 2000 and 2020 (published as Potapov et al. 2022), but we reclassified their percent tree canopy cover data to extract forest cover.
“Line 130. The JAXA FNF dataset is not land cover.”
Thank you for your comment. For reference, JAXA refers to the dataset as a FNF (“forest / non-forest”) dataset. One could certainly argue that a “forest cover” dataset is a type of land cover dataset, as it depends on one’s definition of “land cover.” The dataset does represent more than just forest, as the first version of the dataset (covering 2007-2010, and 2015-2017) has three classes: forest, non-forest, and water, and all three are various land cover classes. The second version of the data (covering 2017-2023), which we also used in this study has four land cover classes: dense forest, non-dense forest, non-forest, and water. These are detailed in https://www.eorc.jaxa.jp/ALOS/en/dataset/pdf/DatasetDescription_PALSAR_Mosaic_FNF_revO.pdf and https://www.eorc.jaxa.jp/ALOS/en/dataset/pdf/DatasetDescription_PALSAR2_FNF_v200a.pdf.
“Line 134. What does “For the most part” mean?”
We indicated “for the most part,” because as indicated in lines 126-132, one can see that the ESA CCI-Land Cover and MODIS land cover products were of a fairly coarse spatial resolution (300-500m), relative to the other datasets, whose spatial resolutions ranged from 10m to 30m. We did not state that because we thought it was evident from the previous bulleted list.
“Line 138. I would strongly suggest to avoid the Liu and Xu datasets at all because of the completely different spatial resolution.”
We appreciate the suggestion, but the objective of our study to quantify the AGB estimated in all of the existing global and pan-tropical AGB datasets, hence we considered the Liu et al. (2015) and Xu et al. (2021) datasets even though their spatial resolutions were coarser than the other ones. The Kindermann et al. (2008) dataset was also coarser than the other datasets, at a 50 km x 50 km spatial resolution.
“Line 139. What “statistical analysis”?”
Thank you for your comment. Please note on lines 181-190 where we describe the statistical analysis performed for this study.
“Line 142. Where is this number “16” coming from considering that before you mentioned 22 and 29 datasets?”
Thank you for this comment. Between lines 134 and line 142, we account for the various numbers of datasets (51 total) and sources (16 total), but we could a better job explaining, and perhaps we should rewrite the section for clarity. Table 1 illustrates the 16 source studies and the 51 datasets available between 2000 and 2022. As indicated in lines 153-154, from the total 51 datasets, we were able to down-select to 19, representing 11 for circa 2000, 5 (not 6) for circa 2010, and 3 for circa 2020. We will make the clarifications in the next version.
“Line 143. “were resampled” how?”
Please see the next set of responses.
“Line 145. Are you seriously thinking that nearest neighbor does not alter the statistics of the maps when going from 30 m to 1000 m?”
Thank you for your comment. We did not correctly spell out the full resampling process used. The land cover datasets were resampled using majority resampling, since the data are categorical and not quantitative. We did indeed use NN resampling on the AGB data. We can rerun our analyses using bilinear interpolation or cubic convolution, but we also think it would be illustrative to do a comparison of the resampling techniques and comment on how much the values change with each technique.
“Line 146. Why resampling to the Mollweide projection? To my knowledge most (all?) datasets were in lat/long projection. Why not keeping this one?”
Thank you for the comment. We believe it was appropriate to use an equal area projection system like Mollweide so as to ensure that areas at the poles (boreal zones) were not overrepresented. We will provide explanatory text about how using equal area projections helps to minimize certain geographic distortions.
“Line 150. The focus on the pan-tropical zone was already mentioned.”
Thank you for the suggestion. In the revision, we will remove that text.
“Line 151. What is the “first level of analysis”?”
The first level of analysis refers to our first filtering of the data.
“Line 151. And here we have “19” which is a new number apparently.”
Please see our earlier response to your question regarding the 16 data sources, but the 19 AGB datasets / maps represents 11 AGB datasets for circa 2000, 5 for circa 2010, and 3 for circa 2020. We hope our rewording of that section will clarify such confusion for the next reader.
“Line 152. Shouldn’t it be 2010 instead of 2019?”
Thank you for this comment. We meant that in the analysis which translated into Figures 10 and 11, we did not use the ESA CCI-Biomass data for 2017, 2018, and 2019. For circa 2020, we only incorporated the CCI-Biomass data for 2020, along with the data from Dubayah et al. (2023) and Xu et al. (2021). We will add a clarification in the revision.
“Line 153. What does “did not skew the analysis” mean here?”
Thank you for the comment. If we generate averages based on multiple AGB datasets but include multiple maps from the same source (e.g., including all 12 Liu et al. 2015 datasets, or all 19 Xu et al. 2021 datasets, or the multiple Santoro et al. 2023 datasets), the consequence will be that the averages will be skewed toward the averages of those datasets.
“Line 164. What are “widely accepted methods?” This should be made clear.”
Thank you. In the revised version, we will provide the specific methods in a technical annex.
“Line 169. And here we have “11” datasets. Plus 3 one line below makes 14, again a new number.”
Thank you for the comment. As we have noted in our response to you regarding line 42, in a revision, we will make the references to the numbers of datasets and data sources utilized clearer.
“Lines 169 and 171. You may want to say “combinations” rather than “permutations”.”
Thank you. We concur.
“Line 171. “those data”?”
Thank you. We should have been specific to say “the AGB data,” which are referenced in the earlier sentence.
“Line 172. What does “tuning into nuances in the data” mean?”
Thank you. We could rephrase this to say “This allowed for an even finer characterization of the AGB data within the respective climate zones.”
“Lines 174-176. Is this relevant?”
Perhaps we could reword this. We were mainly trying to illustrate how the AGB data could easily be converted
“Line 180. Figure 3. The flowchart is redundant since what is done here is purely zonal statistics.”
Thank you for this comment. We humbly beg to differ. While zonal statistics might be familiar to specialists, we considered that it was extremely relevant to include this show figure as it explains how the various datasets were combined, and in what order, and how statistics were extracted from the data.
“Lines 184-190. Totally unclear what is done here.”
Thank you for the comment. As we indicate in the manuscript, we have done pixel-to-pixel comparisons among the various AGB and forest cover datasets, resulting in what is shown in Figure 4, and Figures 8-11.
“Lines 193-194. Text is redundant.”
Thank you for the comment. We merely thought that we should introduce what is included in the rest of the section.
“Line 202. What are the “two outliers”?”
Thank you for this comment. We should have been more specific and indicated that the two outliers - which we indeed mentioned in lines 199-201 - were the JAXA and MODIS datasets. The effects of those outliers can be seen in Figure 5, but one sees that otherwise, the other land cover datasets are closer in their estimates of overall global forest cover.
“Lines 233 and 234. Where are these rates of km2/year from?”
Thank you for the comment. These are derived from Table 2, specifically from taking the total deforestation (i.e. in boreal, temperate, and tropical), and subtracting the regrowth (in all three climate zones) from that. We will clarify in the revised manuscript.
“Lines 244, 246 and 247. Are these % values for the totals?”
Thank you for the comment. These too are based on Table 2. These are %s of change based on the initial forest areas (2000 for CCI-LC and UMD, and 2001 for MODIS).
“Figure 6 and Line 261. The reference should be to Santoro et al., 2021, ESSD.”
Thank you for the suggestion. The figure is also referenced in Santoro et al., 2023 (Figure 1-1 on page 16).
“Figure 7. Fonts on the x-axis are too small. By the way, the Geocarbon dataset is a blend of 3 datasets: Saatchi et al (year 2000), and Baccini et al (year 2007) for the tropics and Santoro et al (year 2010) for temperate and boreal zones. Associating Geocarbon to 2000 can be misleading.”
Thank you for the suggestion. We will increase the x-axis font. Thank you also for the clarification regarding the GeoCarbon data.
“Line 275. Why “As to be expected?””
Thank you. We should have included a statement or reference to previous studies indicating that there is more biomass and also more biological diversity [as indicated by tree species] nearer to the Equator. Perhaps we should rephrase this to “As has been indicated in earlier studies…”
“Line 279. This belt contains temperate rainforest (southeast Australia and Tasmania)”
Thank you for that observation.
“Lines 280-282. Already mentioned, redundant.”
We agree.
“Lines 284. Your delineation does not correspond to the delineation adopted by IPCC.”
Please see our statement above as to why our map might not match the earlier maps provided by the IPCC. We will include a footnote to indicate that because we have used different elevation, precipitation, and temperature input data that the delineation will look different.
“Line 294. Figs. 8-11 do not add information compared to the presentation of results with histograms unless it is explained in the text what these figures add with respect to Fig. 7.”
Thank you for your comment. We interpret your suggestion to mean that we should provide more explanations of Figures 8-11. We will do so in the revised manuscript.
“Line 304. With “models” do you refer to “estimates” perhaps?”
Thank you., We meant that the outputs of the various AGB models - i.e., the individual datasets evaluated - did not converge, but we can rephrase the sentence.
“Line 304. Please reconsider this statement. A disagreement of 100% for an AGB = 1 ton/ha is less relevant than a disagreement of 20% for an AGB of 400 tons/ha.”
Thank you. We see where you are going, and that’s precisely why we also estimated the Coefficient of Variation (CV), and not just the standard deviation (SD), with the CV being the SD normalized by the mean. Therefore, the CV is displaying the relative variation, and is not a measure of the absolute variation, like the SD would be. So since we used CV, we consider that our findings hold, in terms of the CV generally being higher in non-tropical and lower biomass areas.
“Figure 11. Is a CV based on 3 numbers a reliable figure for the level of agreement between maps?”
Thank you. We don’t see why it wouldn’t be. The CV is based on the standard deviation and the mean - which we also derived - and suggesting that one should not derive CV from only three datasets seems to us similar to suggesting that one cannot or should not derive standard deviations from a small sample. We also did so for c. 2010 based on five sources - but did not include this because we already had many figures - but what we mainly saw was that the means and the CVs for c. 2000 and c. 2020 essentially showed convergence in similar areas, indicating model agreement overall.
Citation: https://doi.org/10.5194/egusphere-2024-1179-AC1
-
RC2: 'Comment on egusphere-2024-1179', Anonymous Referee #2, 12 Aug 2024
I read Cherrington's paper aiming to use ensembles of forest cover and biomass datasets to determine the carbon budgets of the world's forest ecosystems. This is an interesting topic. However, the readability of the manuscript does not have an easy-to-follow flow. Many critical sections are missing, especially for data processing and methodologies.
Very little literature review in the Introduction. Only two papers or reports were cited in the Introduction although they are not directly related to the topic in this study. It would be hard to get the data and knowledge gaps if you did not know the research progress.
Methods section has little information for readers to follow the data processing. Please provide a brief introduction for each data used in this study, either in the Methods or supplementary materials. For example, what is the meaning of "All the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods"? (Page 8 Line 165). It is unclear how to use forest as a mask. It would be meaningless if the manuscript did not consider forest definitions in the data analyses. It is unclear how climate data was used. The AGB data processing is also not clear.
The discussion is unclear and weak. Substantial efforts are needed to discuss the reasons for the differences in forest cover and biomass trends. Please use subtitles to guide readers to understand the discussion. Please avoid unclear sentences such as " We should provide a caveat." (Page 22 Line 345). ABG should be AGB? (Page 355 Line 355).
Citation: https://doi.org/10.5194/egusphere-2024-1179-RC2 -
AC2: 'Reply on RC2', Emil Cherrington, 30 Aug 2024
Dear reviewer # 2,
First off, thank you for accepting the review invitation and for taking the time to review our manuscript. Thank you also for the suggestions regarding revisions. On behalf of my co-authors, I would like to respond point-by-point to the feedback you provided.
= = = = =
“This is an interesting topic. However, the readability of the manuscript does not have an easy-to-follow flow.”
Thank you for that observation. We agree with you that above ground biomass data - and particularly the distinctions between the various published datasets - is indeed an interesting topic. Regarding your observation readability, we would indeed like to make the manuscript easier to read. In line with one of your following suggestions, we consider that - in consultation with the Editor - we might need to move some of the text to Supplementary Materials to improve the manuscript’s flow.
“Many critical sections are missing, especially for data processing and methodologies.”
Thank you for this comment. We would like to point out that lines 75-190 go into a great detail about the specifics of the methods employed, including the data processing. If additional detail is required, we would appreciate it if you could indicate what specifically is missing regarding the data processing and the methodology.
“Very little literature review in the Introduction. Only two papers or reports were cited in the Introduction although they are not directly related to the topic in this study. It would be hard to get the data and knowledge gaps if you did not know the research progress.”
Thank you for those observations. Regarding the observation that the papers cited were not directly related to the topic of study, we would beg to differ, as Trumper et al. (2009), for instance, provides one of the earliest characterizations of the global biomass budget, which is what our study elaborates on. Additionally, our reference to the IPCC and the UNFCCC in turn seeks to provide context to the significance of biomass and forest carbon in terms of multilateral environmental agreements that countries are signatories to.
While we did indeed only cite a few papers in the short Introduction section (lines 34-72), the breadth of papers reviewed was mainly covered in the Discussion section (lines 314-512) where we tried to tie back the results to earlier studies. That said, we could certainly make changes to the Introduction and the Discussion to introduce some of the reviewed studies earlier in the paper.
“Methods section has little information for readers to follow the data processing.”
Thank you. Please see our earlier response regarding your similar comment on “data processing and methodologies.”
“Please provide a brief introduction for each data used in this study, either in the Methods or supplementary materials.”
That is a great suggestion. Our challenge has been that we reviewed 8 distinct sources of land cover data, as well as 16 sources of above ground biomass data, so describing those in detail in the Methods section would have likely made the manuscript even more lengthy. Given that our manuscript is already very long at 32 pages, we appreciate your suggestion regarding providing additional information on the input datasets in Supplementary Materials.
For example, what is the meaning of "All the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods"? (Page 8 Line 165).” It is unclear how to use forest as a mask.”
Thank you for that observation. In the subsection on “Characterizing forest AGB” (lines 166-176), we indicated how the forest masks (whose derivation was described in lines 161-164) were combined with the AGB data. Essentially, we extracted the AGB data using derived forest cover maps (i.e., “forest masks”), and that is shown graphically in Figure 3. If the descriptions provided in lines 166-176 are not clear, we can further elaborate in that section.
“It would be meaningless if the manuscript did not consider forest definitions in the data analyses. It is unclear how climate data was used. The AGB data processing is also not clear.”
We agree with you that it is important to provide descriptions of forest definitions, climate data, and processing of AGB. Specifically regarding the climate data and the AGB data processing, in lines 171-172, we highlight that we use a masking process, similar to the process used to generate the forest masks. We derived maps of the three climate zones, and we clipped the AGB data to those three maps to determine how much AGB falls within each climate zone. That processing is also depicted graphically in Figure 3. If additional descriptions of said processing are necessary, we can provide additional explanations. Regarding the forest definitions, we do address this in the Discussion section (lines 399-407) as a source of uncertainty. For the sake of not making the manuscript any longer than it already is, we did not consider elaborating on the details of the 8 land cover maps in the Methods section, but perhaps that could also go into a part of the Supplementary Materials.
“The discussion is unclear and weak. Substantial efforts are needed to discuss the reasons for the differences in forest cover and biomass trends.”
Thank you for that observation. In our perspective, we spent practically the entire Discussion (lines 312-512) section exploring the reasons why the biomass data (our principal focus) - and to a lesser extent the forest cover data - differ. And part of that analysis is merely pointing out - for the benefit of people who might not otherwise be able to do a deep dive into the 16 sources of the 51 individual biomass - what the differences are, and that is explored both in the Results (lines 192-311) and in the comparative analysis section (lines 318-395) of the Discussion.
Within the Caveats section of the Discussion (and specifically in lines 435-445), we do seek to address how different input datasets and methods might have contributed to differing biomass estimates, especially as we point out that the inputs and approaches are similar: “Studies such as Saatchi et al. (2011), Baccini et al. (2012), Hu et al. (2016), Spawn et al. (2020), Yang et al. (2020), and Xu et al. (2021) used similar methods and multispectral and spaceborne LiDAR data inputs, while Santoro et al. (2021) and 440 Santoro et al. (2023) used modeling approaches using radar data inputs. The data generated by both sets of approaches - which depended heavily on wall-to-wall remotely sensed data seemed to be more similar than than the products of the approaches taken by Kindermann et al. (2008) or Ruesch and Gibbs (2008)…” (lines 437-441).
That said, if you have specific thoughts on how to make our Discussion stronger, we are open to suggestions.
“Please use subtitles to guide readers to understand the discussion.”
Thank you for that suggestion, but for that exact purpose, we did include subtitles in the Discussion to guide readers, e.g., Comparative analysis (line 319), Assumptions (line 397), Caveats (line 423), Implications (line 463), and Future research directions (line 476).
“Please avoid unclear sentences such as " We should provide a caveat." (Page 22 Line 345).”
Point taken. We will revise such sections in the manuscript, for clarity.
“ABG should be AGB? (Page 355 Line 355).”
You have an eagle’s eye. Thank you for that observation. We have also confirmed that line 355 is the only place in the manuscript with that error.
Citation: https://doi.org/10.5194/egusphere-2024-1179-AC2
-
AC2: 'Reply on RC2', Emil Cherrington, 30 Aug 2024
Status: closed
-
RC1: 'Comment on egusphere-2024-1179', Anonymous Referee #1, 17 Jun 2024
General comments
This manuscript compares global and nearly-global datasets of land cover, forest cover and aboveground biomass and synthesizes information to understand their relevance. The work of collecting all these datasets and comparing them is per se tremendous and the analyses undertaken by the authors provide some interesting numbers. Nonetheless, the major issue with this manuscript is that they provide a long list of numbers. These numbers diverge and the conclusion is that there is not a single model, which was already known before. I would have appreciated if the authors had moved beyond synthesizing datasets and provided map producers with some guidelines of dos and don’ts from their perspective. In addition, there is no mention of the methods followed to synthesize data so I cannot judge whether these might have impacted the numbers that were derived from the maps. I also find the text superficially written (generic statements, varying numbers). Although, I believe that the manuscript has some scientific merit, it is currently not meeting the standards for publication in a peer-reviewed journal.
Below the authors can find a list of comments that they may consider should they pursue resubmission to another journal.
Specific comments
- The title seems to promise something that was not fulfilled by the article. What is a “forest carbon model” exactly?
- What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?
Lucas, R.M., Mitchell, A.L., Armston, J., 2015. Measurement of Forest Above-Ground Biomass Using Active and Passive Remote Sensing at Large (Subnational to Global) Scales. Current Forestry Reports 1, 162–177. https://doi.org/10.1007/s40725-015-0021-9
Rodríguez-Veiga, P., Wheeler, J., Louis, V., Tansey, K., Balzter, H., 2017. Quantifying forest biomass carbon stocks from space. Current Forestry Reports 3, 1–18. https://doi.org/10.1007/s40725-017-0052-5
- What is your definition of “forest”?
- What is your definition of “biomass”?
- 4 is not introduced in the text.
- 5. Using values published by FAO in their resources assessment could help as reference to understand it. However, not knowing how each of the land cover datasets was re-labelled to forest/non-forest AND how maps were resampled to 1 km does not help to understand the differences in this figure. Actually, comparing maps of land cover at 1 km is introducing a lot of uncertainty in the results (too coarse resolution).
- The comparative analysis in Section IV is not really addressing the results from the comparison but providing a high-level presentation of concepts and potential explanations. In particular, the paragraph on lines 376-387 must be reconsidered. If one embarks in a study like the one envisaged in the title, one should also provide the answers. To me, these answers were not provided.
- The Assumptions in Section IV should also be reconsidered as they very much reduce the importance and the merit of the study.
- The purpose of the Caveats in Section IV is unclear. It is currently a mixture of speculations, work already published by others (lines 435-445) and possible work that was not undertaken (Lines 456-461).
- The Implications in Section IV could profit from some re-writing as well. 1) The authors should not ask questions (lines 469-474) but provide some indications. 2) Lines 486 – 492 are not related to this study.
Technical corrections
- Line 17: please correct “to further the understanding”
- Line 45: “15 years or so” were “last decade or so” in the abstract, “dozen sources” here were “20 sources” in the abstract. Please be precise and consistent.
- Line 46. Use “published” instead of “developed” since a few datasets have not been published.
- Lines 45-51. Near-field remote sensing is missing here (ALS, TLS).
- Line 54. Studies do not “advocate” but try to emphasize what is new to existing knowledge. As time goes by, new research demonstrates what the previous studies did not or could not consider.
- Line 70. Previous studies should be referred to.
- Line 72. “Implications” were not mentioned in the abstract, why?
- Lines 75-96. Move to the Introduction
- Lines 101-108. Why did you create your own classification since there are already official classifications around such the FAO Global Ecological Zones dataset, which is the basis of the IPCC classes as well? Classifying zone based on temperature can be misleading. Looking at Figure 2, Tibet for example belongs to the boreal zone, which is incorrect.
- Line 114. What is the impact of not distinguishing “tree” and “forest” cover on your study?
- Lines 120 and 129. The Hansen dataset is not a land cover dataset.
- Line 130. The JAXA FNF dataset is not land cover.
- Line 134. What does “For the most part” mean?
- Line 138. I would strongly suggest to avoid the Liu and Xu datasets at all because of the completely different spatial resolution.
- Line 139. What “statistical analysis”?
- Line 142. Where is this number “16” coming from considering that before you mentioned 22 and 29 datasets?
- Line 143. “were resampled” how?
- Line 145. Are you seriously thinking that nearest neighbor does not alter the statistics of the maps when going from 30 m to 1000 m?
- Line 146. Why resampling to the Mollweide projection? To my knowledge most (all?) datasets were in lat/long projection. Why not keeping this one?
- Line 150. The focus on the pan-tropical zone was already mentioned.
- Line 151. What is the “first level of analysis”?
- Line 151. And here we have “19” which is a new number apparently.
- Line 152. Shouldn’t it be 2010 instead of 2019?
- Line 153. What does “did not skew the analysis” mean here?
- Line 164. What are “widely accepted methods?” This should be made clear.
- Line 169. And here we have “11” datasets. Plus 3 one line below makes 14, again a new number.
- Lines 169 and 171. You may want to say “combinations” rather than “permutations”.
- Line 171. “those data”?
- Line 172. What does “tuning into nuances in the data” mean?
- Lines 174-176. Is this relevant?
- Line 180. Figure 3. The flowchart is redundant since what is done here is purely zonal statistics.
- Lines 184-190. Totally unclear what is done here.
- Lines 193-194. Text is redundant.
- Line 202. What are the “two outliers”?
- Lines 233 and 234. Where are these rates of km2/year from?
- Lines 244, 246 and 247. Are these % values for the totals?
- Figure 6 and Line 261. The reference should be to Santoro et al., 2021, ESSD.
- Figure 7. Fonts on the x-axis are too small. By the way, the Geocarbon dataset is a blend of 3 datasets: Saatchi et al (year 2000), and Baccini et al (year 2007) for the tropics and Santoro et al (year 2010) for temperate and boreal zones. Associating Geocarbon to 2000 can be misleading.
- Line 275. Why “As to be expected?”
- Line 279. This belt contains temperate rainforest (southeast Australia and Tasmania)
- Lines 280-282. Already mentioned, redundant.
- Lines 284. Your delineation does not correspond to the delineation adopted by IPCC.
- Line 294. Figs. 8-11 do not add information compared to the presentation of results with histograms unless it is explained in the text what these figures add with respect to Fig. 7.
- Line 304. With “models” do you refer to “estimates” perhaps?
- Line 304. Please reconsider this statement. A disagreement of 100% for an AGB = 1 ton/ha is less relevant than a disagreement of 20% for an AGB of 400 tons/ha.
- Figure 11. Is a CV based on 3 numbers a reliable figure for the level of agreement between maps?
Citation: https://doi.org/10.5194/egusphere-2024-1179-RC1 -
AC1: 'Reply on RC1', Emil Cherrington, 10 Jul 2024
Dear reviewer # 1,
First off, thank you for accepting the review invitation and for taking the time to review our manuscript. Thank you also for the suggestions regarding revisions. On behalf of my co-authors, I would like to respond point-by-point to the feedback you provided. Below is the response to the initial summary you provided.
= = = = =
“This manuscript compares global and nearly-global datasets of land cover, forest cover and aboveground biomass and synthesizes information to understand their relevance. The work of collecting all these datasets and comparing them is per se tremendous and the analyses undertaken by the authors provide some interesting numbers.”
We agree. Per some of your later comments, we definitely consider that assembling, processing, and analyzing that data to extract key findings is the very spirit of “synthesis,” in line with BioGeosciences’ indication that “reviews and syntheses summarize the status of knowledge and outline future directions of research within the journal scope” (https://www.biogeosciences.net/about/manuscript_types.html). We have assembled and generated over 100GB of spatial data, which we interpreted in a consistent framework to provide insights on above ground biomass - and forest carbon by proxy.
“Nonetheless, the major issue with this manuscript is that they provide a long list of numbers.”
Per the scope of our study, which we state in lines 70-76, i.e., to “dig deeper into those datasets…” and “to characterize global forest carbon stocks.” We believe the manuscript does provide analysis to put the list of numbers in context.
“These numbers diverge and the conclusion is that there is not a single model, which was already known before.”
When you indicate that “[it] was already known before,” how was this already known before? We note that you later suggest that we review Lucas et al. (2015) and Rodriguez-Viega et al. (2017), per your question, “What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?”
We appreciate that you suggested that we also review and incorporate insights from Lucas et al. (2015) and Rodriguez-Veiga et al. (2017). In our perspective, both were insightful studies, and we plan to include them in our manuscript revision. While the impression left by your review is that both papers have essentially already addressed the issues we have addressed in our recent manuscript, please note that (1) the scope of our study differs substantially from the scope of both papers, and (2) in the intervening seven and nine years since the publication of both studies, the body of knowledge on above ground biomass has increased substantially. Please allow us to explain.
- The review suggests that we should have focused on providing feedback to the “map makers” about increasing the accuracies of their products. We would like to mention that it was not the scope of our study. Our scope was to understand what the products of 16 studies on global-scale and tropical-scale biomass imply, from an applications perspective. Neither of the studies you mentioned focus on that topic, as their scopes were different. In terms of what was already known prior to our study, one particular study, Zhang et al. (2019) sought to synthesize the state of knowledge by documenting the various global, regional, and national above ground biomass studies [and accompanying datasets] which existed. While we found that study enlightening - as it introduces practitioners to the data that are available - we were left wanting to understand the implications of varying datasets. Or to put it another way, what do the various above ground biomass (AGB) datasets tell us about global and regional carbon budgets with respect specifically to AGB, which is a component of such budgets? If we should have stated that explicitly within our manuscript, we ask your indulgence and would be more than happy to state that in the manuscript’s revision. And regarding your sentiment that we are restating what is already known, we contend that previous studies have not analyzed and dug into the results of the 16 global scale and tropical scale studies. So, in that regard, we beg to differ. From our perspective, we saw a gap and put effort into answering a question that had not been answered before.
- The body of knowledge has increased substantially since the publication of Lucas et al. (2015) and Rodriguez-Viega et al. (2017), eight of the sixteen studies we have listed were published from 2020 onwards, meaning that Lucas et al. and Rodriguez-Viega et al. would therefore not have had access to these in their reviews. Half of the body of knowledge did not yet exist when the two studies you suggested were published, so their insights would therefore not have taken into account more recent knowledge. For instance, in trying to synthesize available knowledge on AGB stocks, Lucas et al. (2015) mainly based their findings on comparing only two AGB datasets, that of Saatchi et al. (2011) and Baccini et al. (2012). From the perspective of quantifying AGB stocks, we appreciate that Lucas et al. (2015) did attempt to convey to their readers information on AGB stocks, although they did not go so far as to compare the total stock estimates derived from Saatchi et al. and Baccini et al.
Along the lines of your suggestion, Rodriguez-Viega et al. (2017) does provide suggestions on how accuracies of the various AGB products can be improved, and in their assessment, they examined five tropical-scale and global products (from Saatchi et al. 2011, Baccini et al. 2012, Avitabile et al. 2016, Liu et al. 2015, and Hu et al. 2016), in addition to citing Ruesch & Gibbs (2008) and Kinderman et al. (2008). Their work, however, leaves out 11 other AGB products which have been released since 2017.
Please note that we are not criticizing the scopes of Lucas et al. (2015) and Rodriguez-Viega et al. (2017), merely pointing out where their reviews of existing AGB studies in turn left gaps which we sought to explore with our own study. Since Lucas et al. (2015) and Rodriguez-Viega et al. (2017) evaluated datasets available at the time of their publication, we found it valuable to extend that body of knowledge with the latest AGB datasets as well as the latest body of knowledge. We respectfully argue that filling the gap left by previous studies is a valuable endeavor and remains the main focus and driver for this manuscript.
“I would have appreciated if the authors had moved beyond synthesizing datasets and provided map producers with some guidelines of dos and don’ts from their perspective.”
This manuscript is a review / synthesis, and we believe we remain in scope of the journal's manuscript submission guidelines. Thanks to your feedback, we will further strengthen our statements in the manuscript lines 70-76. The scope of our study is characterizing AGB and forest carbon, we are evaluating the current state of models and datasets available and we think it was not our place to suggest how the studies we reviewed could generate better datasets. We are merely trying to characterize the similarities and differences, we are not trying to help them improve their respective accuracies. That, however, would be a great but a different paper.
“In addition, there is no mention of the methods followed to synthesize data so I cannot judge whether these might have impacted the numbers that were derived from the maps.”
The aim of the manuscript is to identify publicly available datasets, methods, and explore how they converge (or not) for a geographic region. Our manuscript does indeed outline methods we evaluated as part of this work. In fact, we would like to thank you for your feedback and comments on the methods section of our manuscript. For example, we appreciate your comments about the clarification on the resampling methods and the comments about Mollweide projection. It will help us further strengthen our methods section.
“I also find the text superficially written (generic statements, varying numbers).”
We appreciate this feedback, but we would have appreciated specific references to where we could strengthen the manuscript.
“Although, I believe that the manuscript has some scientific merit, it is currently not meeting the standards for publication in a peer-reviewed journal.”
While we appreciate your candor, we believe that a manuscript that examines the publicly available datasets and methods across the globe and comparing the results obtained from that amalgamation is indeed a key function of a review/synthesis manuscript. We have attempted to do precisely that. With all humility, we believe no other publication has looked exhaustively at the available datasets, methods and presented the results in a cohesive form before. That, we believe, is the merit of this review manuscript.
While we deeply appreciate your perspective, your preference that we should have taken a different direction in terms of what we analyzed, in our humble opinions, does not make our findings any less scientifically valid. We purposefully do not provide a critique of the outputs of the 16 studies. We merely analyzed each of them in order to provide our own stakeholders with an understanding of the implications of each dataset. In the domain of our work, we collaborate with users who would like to know how much biomass is present in their area of interest, whether that is a country, province, or a protected area. Having 51 individual datasets from 16 different sources can be confusing for research communities and downstream practitioners alike. We see our study as one way to parse the data and provide guidance to users on the characteristics of that data.
“Below the authors can find a list of comments that they may consider should they pursue resubmission to another journal.”
“Specific comments”
“The title seems to promise something that was not fulfilled by the article.”
We would have hoped that you - as well as the audience in general - could see that we are asking a rhetorical question. Our conclusion is that the models diverge, so there is no one model to rule them all, to borrow from JRR Tolkien’s “one ring to rule them all.” We remain open to a different recommendation.
“What is a “forest carbon model” exactly?”
Thank you for this comment. By “forest carbon models,” we were referring to models of forest carbon stocks, which we have indicated in lines 174-176, are directly related to the models of above ground biomass (AGB). We will integrate your feedback to clarify the manuscript accordingly.
“What is unique to this paper that has not been already discussed in similar review papers such as Lucas et al., and Rodriguez-Vega et al.?”
Please see our extensive response above regarding both studies.
= = = = =
“What is your definition of “forest”?”
Thank you for your comment. For the purpose of this study, “forest” is defined as areas mapped as “forest” or “tree cover” in the respective input land cover data sources used. Tree cover is treated as a proxy for forest cover, as indicated in line 114. We will strengthen the wording to make sure we make it clear to the reader.
“What is your definition of “biomass”?”
Thank you for the suggestion. Since we are focusing on “aboveground biomass,” we will go the definition provided by the IPCC, as in “all living biomass above the soil including stem, stump, branches, bark, seeds, and foliage” (source: https://www.ipcc-nggip.iges.or.jp/public/gpglulucf/gpglulucf_files/Glossary_Acronyms_BasicInfo/Glossary.pdf).
“4 is not introduced in the text.”
Thank you for pointing out our oversight. We assume that you mean Figure 4. We should have introduced Figure 4 in the same line where we introduced Figure 5. We will remedy it in the revised version.
“5. Using values published by FAO in their resources assessment could help as reference to understand it. However, not knowing how each of the land cover datasets was re-labelled to forest/non-forest AND how maps were resampled to 1 km does not help to understand the differences in this figure.”
Thank you for your comment. We assume that you are referring to Figure 4. You later comment that we should have provided detail on the statement in lines 163-164 that “all the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods.” We were trying to economize in terms of our word count in the Methods section, but we would be happy to provide technical annexes indicating which classes were classified to “forest” (essentially the ones in the maps originally going as “forest” or “tree cover”), and which went to “non-forest” (i.e., the remaining classes). The differences in the figure are brought about by the reality that the various source land cover datasets differ in what they identify as forest cover / tree cover, and that is what we are seeking to show here. And for what it’s worth, other studies like Venter et al. 2022 (“Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover”) have taken a similar approach of evaluating the differences among various global land cover datasets.
“Actually, comparing maps of land cover at 1 km is introducing a lot of uncertainty in the results (too coarse resolution).”
Thank you for your comment. While some will be introduced in majority resampling 10m, 25m, 30m, 300m, 500m data to 1km, it is unlikely to be substantial. We are working on providing additional analyses showing how resampling to 1km changed estimates of overall forest cover.
“The comparative analysis in Section IV is not really addressing the results from the comparison but providing a high-level presentation of concepts and potential explanations.”
Thank you for that observation. In this subsection, we tried to provide a focused analysis on how certain datasets converged, while others did not, e.g. in lines 339-369. We explored differences regarding forest cover change (lines 339-342) and regarding the overall forest AGB budgets that could be extracted from the datasets (lines 350-362).
“In particular, the paragraph on lines 376-387 must be reconsidered.”
Thank you for the comment and allowing us to reexamine how we are stating that section. We are trying to explain that it is not straightforward to analyze forest AGB given the various potential combinations of data and the differences among the datasets themselves. Therefore, rather than trying to say one or a few specific datasets are “correct,” which we explain is difficult to ascertain because of the lack of corresponding data from destructive sampling (per guidance from Duncanson et al. 2020), we propose that one way to understand the world’s AGB budget is to utilize an ensemble of the various AGB estimates (like how ensembles are used in weather forecasting). So rather than reconsidering that paragraph, in absence of other data, we see this paragraph as presenting a key consideration. We will reword the section for adding this nuanced clarification.
“If one embarks in a study like the one envisaged in the title, one should also provide the answers. To me, these answers were not provided.”
Again, we appreciate the candor. As mentioned in one of our earlier responses, we saw the title as us asking a rhetorical question. Is there indeed only one forest carbon (AGB) model? The obvious takeaway from the title itself is expected to be “No, there isn’t,” but our study explores differences among the various AGB datasets. The fact that there are substantial differences among the various available AGB datasets, the manuscript attempts to answer the rhetorical question.
“The Assumptions in Section IV should also be reconsidered as they very much reduce the importance and the merit of the study.”
Thank you for your perspective on that. We thought that for the sake of transparency that it was useful to highlight some of the assumptions we made, lest our analysis be criticized for not including said assumptions. Also, we believe that addition of these assumptions does not reduce the importance or merit of the study. For instance, we discuss that we resampled the data to a common reference scale (1km). Just the act of resampling a few of the higher spatial resolution datasets does not drastically change the findings of the study. Alternatively, we could have resampled all of the data to the finest scale of reference (10m), but that would also have improperly represented the spatial precision of coarser datasets and made the data volume extremely large and unwieldy. We thought it was important to include our assumptions for the sake of completeness of a follow-on, and potentially even critical, article
“The purpose of the Caveats in Section IV is unclear. It is currently a mixture of speculations, work already published by others (lines 435-445) and possible work that was not undertaken (Lines 456-461).”
Thank you for your comment. We believe we are putting our findings in the broader context of other published works. We are of the perspective that the Discussion is where one indeed contemplates the findings in a broader context.
“The Implications in Section IV could profit from some re-writing as well.”
Thank you, we appreciate this comment. For clarity, we will attempt to rewrite the single “implications” paragraph.
“1) The authors should not ask questions (lines 469-474) but provide some indications."
Thank you for the comment. The lines listed in these comments are in the discussions section, and the questions are pertaining to the comparisons with other published works. Per our earlier comment, we again thought it was appropriate to include our comments and speculations within the Discussion section of the manuscript.
“2) Lines 486 – 492 are not related to this study.”
While we appreciate the feedback, we consider that understanding standing aboveground biomass stocks to be an important part of understanding overall carbon budgets. In terms of a future research direction (the title of this section of the Discussion), we considered that a future study or studies could consider how our findings contrast with what is estimated by the respective FREL reports. Of course, that was not within the scope of the current study, but we think it would provide additional insights.
= = = = =
“Technical corrections”
"Line 17: please correct “to further the understanding”
Thank you. We will reword to better reflect our statement.
"Line 45: “15 years or so” were “last decade or so” in the abstract, “dozen sources” here were “20 sources” in the abstract. Please be precise and consistent."
Thank you for this comment. In our revision, we will rephrase these statements to be more precise and consistent.
"Line 46. Use “published” instead of “developed” since a few datasets have not been published."
Thank you for the comment. While most of the datasets are published, a couple are available online and do not have a citation to attribute to. So, instead of a blanket “published”, we chose to mention these datasets are available. We will add a clarifying statement to make sure this point is not lost. Just to clarify, we have cited all the available, published literature that we could find and we will attribute those not “peer reviewed and published”, we will add a clarifying statement. The data from the 16 global and pan-tropical studies we sourced are all public (they are all on the Internet), but indeed, a few of them have not been published in peer-reviewed papers. Hence, we thought mentioning that the data are available (i.e., have been developed) is less confusing than to say “published,” because it might give the impression that we meant, “published in peer-reviewed papers.”
“Lines 45-51. Near-field remote sensing is missing here (ALS, TLS).”
Thank you for the suggestion. We will be happy to include some text on this, likely drawing from Duncanson et al. (2020), the reference to which we should also have included in line 49, but which somehow we missed.
“Line 54. Studies do not “advocate” but try to emphasize what is new to existing knowledge. As time goes by, new research demonstrates what the previous studies did not or could not consider.”
Thank you, this comment made us realize that we should indeed change our wording to clarify. However, especially in light of multiple, recent differing estimates of AGB, with all due respect to the various authors, we have seen quite a few papers that seem to promote only their own dataset’s estimates. So in that sense, it does seem that such studies are advocating for their estimates above other ones. That was also one of the drivers of conducting this study. Or more specifically, one area in which we address this is in lines 456-461.
“Line 70. Previous studies should be referred to.”
Thank you for pointing that out. We should have explicitly mentioned Avitabile et al. (2016), Santoro et al. (2021), Santoro et al. (2023), and Zhang et al. (2019), and we will also include Lucas et al. (2015) and Rodriguez-Viega et al. (2017), which you suggested.
“Line 72. “Implications” were not mentioned in the abstract, why?”
Thank you for your comment. We will reflect on our abstract to include some inferences about implications.
“Lines 75-96. Move to the Introduction”
Thank you for this great suggestion. We will be happy to do so in the revision.
"Lines 101-108. Why did you create your own classification since there are already official classifications around such the FAO Global Ecological Zones dataset, which is the basis of the IPCC classes as well? Classifying zone based on temperature can be misleading. Looking at Figure 2, Tibet for example belongs to the boreal zone, which is incorrect."
Thank you for your comment. Please note that we did not create our own classification scheme. We merely applied the existing IPCC classification scheme to the most recent WorldClim precipitation and temperature data, and elevation data. And we would like to note that “classifying zone based on temperature can be misleading,” temperature data are indeed an input to the IPCC’s classification scheme. One reason for doing so, from our assessment, was that the IPCC climate classification maps (of which there are various sources) were based on older precipitation and temperature data, so we wanted to have an updated set of boundaries. As we have also mentioned in our Introduction and Methods sections, we focus on differentiating the data by the main climate zones mainly to follow on studies like Trumper et al. (2009) which presented data on biomass stocks as a function of climate type. Since a few of the datasets (e.g. Avitabile et al. 2016, Baccini et al. 2012, Dubayah et al. 2023, Saatchi et al. 2011) are mainly confined to the tropics, having a climate zone-based frame of reference was also useful. We will be careful to insert this fine point in the manuscript to make sure the next reader does not ponder on this topic.
"Line 114. What is the impact of not distinguishing “tree” and “forest” cover on your study?"
Thank you. We would like to point out that we do indeed discuss that in the caveats section, specifically lines 399-407.
“Lines 120 and 129. The Hansen dataset is not a land cover dataset.”
Thank you for your comment. We would like to point out that Hansen et al. did indeed release land cover datasets for 2000 and 2020 (published as Potapov et al. 2022), but we reclassified their percent tree canopy cover data to extract forest cover.
“Line 130. The JAXA FNF dataset is not land cover.”
Thank you for your comment. For reference, JAXA refers to the dataset as a FNF (“forest / non-forest”) dataset. One could certainly argue that a “forest cover” dataset is a type of land cover dataset, as it depends on one’s definition of “land cover.” The dataset does represent more than just forest, as the first version of the dataset (covering 2007-2010, and 2015-2017) has three classes: forest, non-forest, and water, and all three are various land cover classes. The second version of the data (covering 2017-2023), which we also used in this study has four land cover classes: dense forest, non-dense forest, non-forest, and water. These are detailed in https://www.eorc.jaxa.jp/ALOS/en/dataset/pdf/DatasetDescription_PALSAR_Mosaic_FNF_revO.pdf and https://www.eorc.jaxa.jp/ALOS/en/dataset/pdf/DatasetDescription_PALSAR2_FNF_v200a.pdf.
“Line 134. What does “For the most part” mean?”
We indicated “for the most part,” because as indicated in lines 126-132, one can see that the ESA CCI-Land Cover and MODIS land cover products were of a fairly coarse spatial resolution (300-500m), relative to the other datasets, whose spatial resolutions ranged from 10m to 30m. We did not state that because we thought it was evident from the previous bulleted list.
“Line 138. I would strongly suggest to avoid the Liu and Xu datasets at all because of the completely different spatial resolution.”
We appreciate the suggestion, but the objective of our study to quantify the AGB estimated in all of the existing global and pan-tropical AGB datasets, hence we considered the Liu et al. (2015) and Xu et al. (2021) datasets even though their spatial resolutions were coarser than the other ones. The Kindermann et al. (2008) dataset was also coarser than the other datasets, at a 50 km x 50 km spatial resolution.
“Line 139. What “statistical analysis”?”
Thank you for your comment. Please note on lines 181-190 where we describe the statistical analysis performed for this study.
“Line 142. Where is this number “16” coming from considering that before you mentioned 22 and 29 datasets?”
Thank you for this comment. Between lines 134 and line 142, we account for the various numbers of datasets (51 total) and sources (16 total), but we could a better job explaining, and perhaps we should rewrite the section for clarity. Table 1 illustrates the 16 source studies and the 51 datasets available between 2000 and 2022. As indicated in lines 153-154, from the total 51 datasets, we were able to down-select to 19, representing 11 for circa 2000, 5 (not 6) for circa 2010, and 3 for circa 2020. We will make the clarifications in the next version.
“Line 143. “were resampled” how?”
Please see the next set of responses.
“Line 145. Are you seriously thinking that nearest neighbor does not alter the statistics of the maps when going from 30 m to 1000 m?”
Thank you for your comment. We did not correctly spell out the full resampling process used. The land cover datasets were resampled using majority resampling, since the data are categorical and not quantitative. We did indeed use NN resampling on the AGB data. We can rerun our analyses using bilinear interpolation or cubic convolution, but we also think it would be illustrative to do a comparison of the resampling techniques and comment on how much the values change with each technique.
“Line 146. Why resampling to the Mollweide projection? To my knowledge most (all?) datasets were in lat/long projection. Why not keeping this one?”
Thank you for the comment. We believe it was appropriate to use an equal area projection system like Mollweide so as to ensure that areas at the poles (boreal zones) were not overrepresented. We will provide explanatory text about how using equal area projections helps to minimize certain geographic distortions.
“Line 150. The focus on the pan-tropical zone was already mentioned.”
Thank you for the suggestion. In the revision, we will remove that text.
“Line 151. What is the “first level of analysis”?”
The first level of analysis refers to our first filtering of the data.
“Line 151. And here we have “19” which is a new number apparently.”
Please see our earlier response to your question regarding the 16 data sources, but the 19 AGB datasets / maps represents 11 AGB datasets for circa 2000, 5 for circa 2010, and 3 for circa 2020. We hope our rewording of that section will clarify such confusion for the next reader.
“Line 152. Shouldn’t it be 2010 instead of 2019?”
Thank you for this comment. We meant that in the analysis which translated into Figures 10 and 11, we did not use the ESA CCI-Biomass data for 2017, 2018, and 2019. For circa 2020, we only incorporated the CCI-Biomass data for 2020, along with the data from Dubayah et al. (2023) and Xu et al. (2021). We will add a clarification in the revision.
“Line 153. What does “did not skew the analysis” mean here?”
Thank you for the comment. If we generate averages based on multiple AGB datasets but include multiple maps from the same source (e.g., including all 12 Liu et al. 2015 datasets, or all 19 Xu et al. 2021 datasets, or the multiple Santoro et al. 2023 datasets), the consequence will be that the averages will be skewed toward the averages of those datasets.
“Line 164. What are “widely accepted methods?” This should be made clear.”
Thank you. In the revised version, we will provide the specific methods in a technical annex.
“Line 169. And here we have “11” datasets. Plus 3 one line below makes 14, again a new number.”
Thank you for the comment. As we have noted in our response to you regarding line 42, in a revision, we will make the references to the numbers of datasets and data sources utilized clearer.
“Lines 169 and 171. You may want to say “combinations” rather than “permutations”.”
Thank you. We concur.
“Line 171. “those data”?”
Thank you. We should have been specific to say “the AGB data,” which are referenced in the earlier sentence.
“Line 172. What does “tuning into nuances in the data” mean?”
Thank you. We could rephrase this to say “This allowed for an even finer characterization of the AGB data within the respective climate zones.”
“Lines 174-176. Is this relevant?”
Perhaps we could reword this. We were mainly trying to illustrate how the AGB data could easily be converted
“Line 180. Figure 3. The flowchart is redundant since what is done here is purely zonal statistics.”
Thank you for this comment. We humbly beg to differ. While zonal statistics might be familiar to specialists, we considered that it was extremely relevant to include this show figure as it explains how the various datasets were combined, and in what order, and how statistics were extracted from the data.
“Lines 184-190. Totally unclear what is done here.”
Thank you for the comment. As we indicate in the manuscript, we have done pixel-to-pixel comparisons among the various AGB and forest cover datasets, resulting in what is shown in Figure 4, and Figures 8-11.
“Lines 193-194. Text is redundant.”
Thank you for the comment. We merely thought that we should introduce what is included in the rest of the section.
“Line 202. What are the “two outliers”?”
Thank you for this comment. We should have been more specific and indicated that the two outliers - which we indeed mentioned in lines 199-201 - were the JAXA and MODIS datasets. The effects of those outliers can be seen in Figure 5, but one sees that otherwise, the other land cover datasets are closer in their estimates of overall global forest cover.
“Lines 233 and 234. Where are these rates of km2/year from?”
Thank you for the comment. These are derived from Table 2, specifically from taking the total deforestation (i.e. in boreal, temperate, and tropical), and subtracting the regrowth (in all three climate zones) from that. We will clarify in the revised manuscript.
“Lines 244, 246 and 247. Are these % values for the totals?”
Thank you for the comment. These too are based on Table 2. These are %s of change based on the initial forest areas (2000 for CCI-LC and UMD, and 2001 for MODIS).
“Figure 6 and Line 261. The reference should be to Santoro et al., 2021, ESSD.”
Thank you for the suggestion. The figure is also referenced in Santoro et al., 2023 (Figure 1-1 on page 16).
“Figure 7. Fonts on the x-axis are too small. By the way, the Geocarbon dataset is a blend of 3 datasets: Saatchi et al (year 2000), and Baccini et al (year 2007) for the tropics and Santoro et al (year 2010) for temperate and boreal zones. Associating Geocarbon to 2000 can be misleading.”
Thank you for the suggestion. We will increase the x-axis font. Thank you also for the clarification regarding the GeoCarbon data.
“Line 275. Why “As to be expected?””
Thank you. We should have included a statement or reference to previous studies indicating that there is more biomass and also more biological diversity [as indicated by tree species] nearer to the Equator. Perhaps we should rephrase this to “As has been indicated in earlier studies…”
“Line 279. This belt contains temperate rainforest (southeast Australia and Tasmania)”
Thank you for that observation.
“Lines 280-282. Already mentioned, redundant.”
We agree.
“Lines 284. Your delineation does not correspond to the delineation adopted by IPCC.”
Please see our statement above as to why our map might not match the earlier maps provided by the IPCC. We will include a footnote to indicate that because we have used different elevation, precipitation, and temperature input data that the delineation will look different.
“Line 294. Figs. 8-11 do not add information compared to the presentation of results with histograms unless it is explained in the text what these figures add with respect to Fig. 7.”
Thank you for your comment. We interpret your suggestion to mean that we should provide more explanations of Figures 8-11. We will do so in the revised manuscript.
“Line 304. With “models” do you refer to “estimates” perhaps?”
Thank you., We meant that the outputs of the various AGB models - i.e., the individual datasets evaluated - did not converge, but we can rephrase the sentence.
“Line 304. Please reconsider this statement. A disagreement of 100% for an AGB = 1 ton/ha is less relevant than a disagreement of 20% for an AGB of 400 tons/ha.”
Thank you. We see where you are going, and that’s precisely why we also estimated the Coefficient of Variation (CV), and not just the standard deviation (SD), with the CV being the SD normalized by the mean. Therefore, the CV is displaying the relative variation, and is not a measure of the absolute variation, like the SD would be. So since we used CV, we consider that our findings hold, in terms of the CV generally being higher in non-tropical and lower biomass areas.
“Figure 11. Is a CV based on 3 numbers a reliable figure for the level of agreement between maps?”
Thank you. We don’t see why it wouldn’t be. The CV is based on the standard deviation and the mean - which we also derived - and suggesting that one should not derive CV from only three datasets seems to us similar to suggesting that one cannot or should not derive standard deviations from a small sample. We also did so for c. 2010 based on five sources - but did not include this because we already had many figures - but what we mainly saw was that the means and the CVs for c. 2000 and c. 2020 essentially showed convergence in similar areas, indicating model agreement overall.
Citation: https://doi.org/10.5194/egusphere-2024-1179-AC1
-
RC2: 'Comment on egusphere-2024-1179', Anonymous Referee #2, 12 Aug 2024
I read Cherrington's paper aiming to use ensembles of forest cover and biomass datasets to determine the carbon budgets of the world's forest ecosystems. This is an interesting topic. However, the readability of the manuscript does not have an easy-to-follow flow. Many critical sections are missing, especially for data processing and methodologies.
Very little literature review in the Introduction. Only two papers or reports were cited in the Introduction although they are not directly related to the topic in this study. It would be hard to get the data and knowledge gaps if you did not know the research progress.
Methods section has little information for readers to follow the data processing. Please provide a brief introduction for each data used in this study, either in the Methods or supplementary materials. For example, what is the meaning of "All the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods"? (Page 8 Line 165). It is unclear how to use forest as a mask. It would be meaningless if the manuscript did not consider forest definitions in the data analyses. It is unclear how climate data was used. The AGB data processing is also not clear.
The discussion is unclear and weak. Substantial efforts are needed to discuss the reasons for the differences in forest cover and biomass trends. Please use subtitles to guide readers to understand the discussion. Please avoid unclear sentences such as " We should provide a caveat." (Page 22 Line 345). ABG should be AGB? (Page 355 Line 355).
Citation: https://doi.org/10.5194/egusphere-2024-1179-RC2 -
AC2: 'Reply on RC2', Emil Cherrington, 30 Aug 2024
Dear reviewer # 2,
First off, thank you for accepting the review invitation and for taking the time to review our manuscript. Thank you also for the suggestions regarding revisions. On behalf of my co-authors, I would like to respond point-by-point to the feedback you provided.
= = = = =
“This is an interesting topic. However, the readability of the manuscript does not have an easy-to-follow flow.”
Thank you for that observation. We agree with you that above ground biomass data - and particularly the distinctions between the various published datasets - is indeed an interesting topic. Regarding your observation readability, we would indeed like to make the manuscript easier to read. In line with one of your following suggestions, we consider that - in consultation with the Editor - we might need to move some of the text to Supplementary Materials to improve the manuscript’s flow.
“Many critical sections are missing, especially for data processing and methodologies.”
Thank you for this comment. We would like to point out that lines 75-190 go into a great detail about the specifics of the methods employed, including the data processing. If additional detail is required, we would appreciate it if you could indicate what specifically is missing regarding the data processing and the methodology.
“Very little literature review in the Introduction. Only two papers or reports were cited in the Introduction although they are not directly related to the topic in this study. It would be hard to get the data and knowledge gaps if you did not know the research progress.”
Thank you for those observations. Regarding the observation that the papers cited were not directly related to the topic of study, we would beg to differ, as Trumper et al. (2009), for instance, provides one of the earliest characterizations of the global biomass budget, which is what our study elaborates on. Additionally, our reference to the IPCC and the UNFCCC in turn seeks to provide context to the significance of biomass and forest carbon in terms of multilateral environmental agreements that countries are signatories to.
While we did indeed only cite a few papers in the short Introduction section (lines 34-72), the breadth of papers reviewed was mainly covered in the Discussion section (lines 314-512) where we tried to tie back the results to earlier studies. That said, we could certainly make changes to the Introduction and the Discussion to introduce some of the reviewed studies earlier in the paper.
“Methods section has little information for readers to follow the data processing.”
Thank you. Please see our earlier response regarding your similar comment on “data processing and methodologies.”
“Please provide a brief introduction for each data used in this study, either in the Methods or supplementary materials.”
That is a great suggestion. Our challenge has been that we reviewed 8 distinct sources of land cover data, as well as 16 sources of above ground biomass data, so describing those in detail in the Methods section would have likely made the manuscript even more lengthy. Given that our manuscript is already very long at 32 pages, we appreciate your suggestion regarding providing additional information on the input datasets in Supplementary Materials.
For example, what is the meaning of "All the land cover datasets were reclassified into forest / non-forest data based on widely accepted methods"? (Page 8 Line 165).” It is unclear how to use forest as a mask.”
Thank you for that observation. In the subsection on “Characterizing forest AGB” (lines 166-176), we indicated how the forest masks (whose derivation was described in lines 161-164) were combined with the AGB data. Essentially, we extracted the AGB data using derived forest cover maps (i.e., “forest masks”), and that is shown graphically in Figure 3. If the descriptions provided in lines 166-176 are not clear, we can further elaborate in that section.
“It would be meaningless if the manuscript did not consider forest definitions in the data analyses. It is unclear how climate data was used. The AGB data processing is also not clear.”
We agree with you that it is important to provide descriptions of forest definitions, climate data, and processing of AGB. Specifically regarding the climate data and the AGB data processing, in lines 171-172, we highlight that we use a masking process, similar to the process used to generate the forest masks. We derived maps of the three climate zones, and we clipped the AGB data to those three maps to determine how much AGB falls within each climate zone. That processing is also depicted graphically in Figure 3. If additional descriptions of said processing are necessary, we can provide additional explanations. Regarding the forest definitions, we do address this in the Discussion section (lines 399-407) as a source of uncertainty. For the sake of not making the manuscript any longer than it already is, we did not consider elaborating on the details of the 8 land cover maps in the Methods section, but perhaps that could also go into a part of the Supplementary Materials.
“The discussion is unclear and weak. Substantial efforts are needed to discuss the reasons for the differences in forest cover and biomass trends.”
Thank you for that observation. In our perspective, we spent practically the entire Discussion (lines 312-512) section exploring the reasons why the biomass data (our principal focus) - and to a lesser extent the forest cover data - differ. And part of that analysis is merely pointing out - for the benefit of people who might not otherwise be able to do a deep dive into the 16 sources of the 51 individual biomass - what the differences are, and that is explored both in the Results (lines 192-311) and in the comparative analysis section (lines 318-395) of the Discussion.
Within the Caveats section of the Discussion (and specifically in lines 435-445), we do seek to address how different input datasets and methods might have contributed to differing biomass estimates, especially as we point out that the inputs and approaches are similar: “Studies such as Saatchi et al. (2011), Baccini et al. (2012), Hu et al. (2016), Spawn et al. (2020), Yang et al. (2020), and Xu et al. (2021) used similar methods and multispectral and spaceborne LiDAR data inputs, while Santoro et al. (2021) and 440 Santoro et al. (2023) used modeling approaches using radar data inputs. The data generated by both sets of approaches - which depended heavily on wall-to-wall remotely sensed data seemed to be more similar than than the products of the approaches taken by Kindermann et al. (2008) or Ruesch and Gibbs (2008)…” (lines 437-441).
That said, if you have specific thoughts on how to make our Discussion stronger, we are open to suggestions.
“Please use subtitles to guide readers to understand the discussion.”
Thank you for that suggestion, but for that exact purpose, we did include subtitles in the Discussion to guide readers, e.g., Comparative analysis (line 319), Assumptions (line 397), Caveats (line 423), Implications (line 463), and Future research directions (line 476).
“Please avoid unclear sentences such as " We should provide a caveat." (Page 22 Line 345).”
Point taken. We will revise such sections in the manuscript, for clarity.
“ABG should be AGB? (Page 355 Line 355).”
You have an eagle’s eye. Thank you for that observation. We have also confirmed that line 355 is the only place in the manuscript with that error.
Citation: https://doi.org/10.5194/egusphere-2024-1179-AC2
-
AC2: 'Reply on RC2', Emil Cherrington, 30 Aug 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 557 | 124 | 28 | 709 | 14 | 17 |
- HTML: 557
- PDF: 124
- XML: 28
- Total: 709
- BibTeX: 14
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1