Determining a Carbon Reference Level for a High-Forest-Low-Deforestation Country

Issue raised by reviewer

Adressed by

Another general spell check and review of editing is needed, as there are several spelling mistakes left and formatting is not consistent throughout. Footnotes seem to be in a different font than the main text, and tables are formatted inconsistently. 

After addressing the comments by both reviewers, including those on the format, we sent the manuscript to a professional language editor to check for remaining tipos.

Abstract: minor spelling and grammar mistakes, missing punctuation.

Reviser and spelling & grammar checked

Title page: as it is now, title page is too long and doesn’t fit on one page

Noted, now it should fit.

Main text:

L 50: provide reference for cited figures

Revised wording, added citation

L 89/90: provide the correct term for LULUCF

Done

L 292 – update cross reference to Section 0

Updated; section 2.5

L 348 “by far the most”

Revised

L 352-254: revise sentence

Revised

L 396: reference error

Fixed.

l.436 ff revise sentence

Done. “While the working definition of forests for REDD+ has not yet been officially adopted in political circles, the entire issue of multiple and conflicting forest definitions raises questions of legitimacy given that each definition serves only the specific needs of the policy process that is en vogue at a given time ^[63,64].”

L 539 revise sentence

Done, simplified to “Creating and improving the necessary data, information and infrastructure is costly and funding is limited. This section therefore aims at defining priorities for future working directions.”

L 547 “In addition” rather than “In brief”

I did indeed mean “in brief/in short”: this sentence summarizes and condenses what is detailed in the sentences above

Discussion: I recommend restructuring this section slightly, to provide a more organized approach. At present it looks like a loose collection of different topics relevant for discussion, which could benefit from a grouping by topic. For example, everything related to the results, uncertainties, data quality, missing data, consistency etc could be discussed in one section on “Uncertainties and limitations of approach and results” (or something like that) whereas “Political implications” and general developments related to the results, such as UNFCCC reporting vs results-based or policy choices/forest definition, political risks etc could be grouped into a second block.

The Discussion section is now restructured in two big blocks, 1. Uncertainties and limitations and 2. Policy implications

As a personal preference, I would use the bold text fragments only in the ‘Conclusions & Recommendations’ section to clearly highlight the important take-away messages by topic, and in the Discussion use subheadings and a running text instead.

With the re-structuring of the discussion section into two big blocks, bold text fragments have now disappeared from this section.

l.554 Forest loss mapping of all REDD+ activities: identifying the post-deforestation land uses would also facilitate assessment of the drivers, so this point could be moved down to here from l. 547

Agreed and sentence moved.

Reviewer 2

This is an interesting article. The topic of how to set up reference levels for forests that do not quite meet the traditional methods to monitor deforestation and degradation is of absolute need, and this study comes at a good timing given the need for countries to ramp up action in reducing their net emissions. I find that this article, however, has important issues that must be addressed before being considered for publication, and a set of assumptions that are not well explained nor supported by data.

Gratefully noted.

It is clear from the text that the authors understand that what they propose is not a generally accepted way of determining a deforestation baseline, as noted by the FCPF comment at the end of section 4. That, however, would not be an issue on itself when pursuing a scientific publication, depending on the framing of the study.

Noted and partially agreed.

This paper bases the baseline estimation in a socioeconomic model that predicts an increasing deforestation trend over time. This model, however, is not provided in the main text or in the supplementary materials, it is simply described. This is not enough information for reviewers and readers to review the model and fully understand its outcomes and limitations.

Agreed. The model is now made available on a data depository and will be available under a creative commons license and the link is provided in the manuscript.

The data is under embargo until publication. Reviewers can nevertheless access the model by sending an email to [email protected].

Furthermore, all models should be validated if data is available. In this case, the authors have enough data since 2000 to compare actual numbers with modeled numbers. Without this, how do we know if this model is accurate and predicting within acceptable ranges? 

Model validation is performed at several steps in the process. First, in every of the steps outlined in section 2.5, model outputs are compared with independent data; these validation steps are presented in the supplementary material.

Further, Figure 2 shows the validation of projected deforestation against 3 independent remotely sensed datasets over 3 time periods. This shows that model results are generally in the range of the (disagreeing) remotely sensed datasets.

That being said, we now added a sentence above table 3 describing in more detail deviations of model results from remotely sensed data.        

Lastly, political risk and deforestation/degradation drivers can be highly dynamic and should be expected to change 15 years from now, but it seems like this was not taking into consideration.

Political risk in the sense of political stability and repercussions on society are implicitly covered by SSP scenarios analysis which underlies the Monte Carlo Analysis. To better reflect this, a half sentence is added to the discussion section. 

Additional forest accounting issues arise in the text. In the Consistency sub-section of the Discussion section, authors seem to justify modeling because of the limited map-based resources to develop a baseline. A land-base approach to reference levels is not the only approach; a widely used and accepted alternative is the activity-based approach, in which tabular data is used instead of a map/imagery data.

Our understanding of the reviewers comment is that moving the remote sensing analysis from detecting changing land cover to land use would add benefit. We certainly agree with this shortcoming in our own remote sensing analysis and the discussion and conclusion makes suggestions in that sense.

We do, however, not see how any remote sensing analysis alone can make statements about the evolving magnitude of land cover change in the future, which is where the predictive power of a land use model comes into play.  

Also, crediting periods must start within 2 years of the end of the reference period to be considered realistic, not 5. Overall, I recommend that the authors review standard forest reference level methodologies available in the FCPF, GOFC-GOLD, and IPCC/UNFCCC websites.

Good point, we made relevant note of this in the discussion section.

Reviewer 3

The paper addresses a very important issue but it needs substantial revision before considering for publication. This seems more of a report submitted to UNFCCC. Globally, 39 countries have already submitted FREL/FRLs to the UNFCCC for technical assessment. Some of them have already claimed for upward adjustment of FREL/FRL. Therefore, this paper in not a unique paper and is similar to some of the reports. Please write this as a paper following standard guidelines of the journal. Moreover, following specific comments may help you to refine the paper.

We argue that the methodology presented in the paper differs notably in two aspects from adjusted reference levels submitted by countries, as laid out in the introduction:

First, the degree of detail in which the agricultural sector as driver is represented; and second the transparency of underlying assumptions.   

We partially agree that the paper manuscript at some places appears too much like a technical report and made several changes to the manuscript, as detailed below and in L 92 ff in the manuscript.

Title Change title, it does not reflect the aim of the paper, which is largely based on Cameroon even at sub-national level.

The reason for the indeed generalized title is that the methodology for adjusting the reference level can be applied to any country or jurisdiction, provided that the required data is available. The introduction is framed around general issues of reference level setting and the issues raised in the discussion section are the same that are faced by many other jurisdictions. Southern Cameroon merely serves as a case study to showcasing the approach since the authors happen to work there.

See as a random example, Melo et al. (2018): Striking divergences in Earth Observation products may limit their use for REDD+. Environmental Research Letters, 13(10), where the authors use Guinea Bissau, a country similar in size to our AoI, as a case study to draw conclusions applicable to the wider geographies.

Material and method The logic behind the upward adjustment “…upward adjustment of the reference level using a land use model of the agriculture sectors where demand for food and feed drives land use change” is not enough. There could be several critical factors for upward adjustment. Need to discuss possible factors, in case of Cameroon, and provide reasons why you choose only the food related factor.

An adjustment based on the (non-industrial) agriculture sector is justified by analysis of the drivers of deforestation. Both on regional level (see Tyukavina et al.)  and in Southern Cameroon (see section 3.1 of the manuscript) expansion of non-industrial agriculture has been identified as major driver of deforestation, by far outweighing other factors such as agro-industries and urban sprawl.

Adequate justification and references are made in the paper in the introduction (lines 125 ff) and in section 2.5 (lines 188 ff)

Tyukavina et al. (2018). Congo Basin forest loss dominated by increasing smallholder clearing. Science Advances, 4(11), eaat2993. doi:10.1126/sciadv.aat2993 

Can you please suggest why you used land cover based definition not land use based? In doing so, how can you make sure that areas with temporary forest-cover loss within the area deforested is not included. Usually, a land-cover based assessment of deforestation overestimate emission reductions associated with reducing deforestation.

Good point raised as temporary forest cover loss (or temporary forest gain) is a relevant factor in a shifting cultivation landscape. The methodology follows a differentiated approach.  

The remote sensing component (section 2.3) indeed maps land cover, not land use which is the somewhat authoritative and best available land cover information available for Southern Cameroon. The shortcomings of mapping land cover are outlined in the discussion section (now also enriched) and concrete suggestions towards moving to land use based definitions are made in the conclusions section.

On the other hand, the land use model de-composes land cover into land use components (e.g. arable land into cropland and fallow and cropland further into individual crops) using agricultural statistics.        

Section 2.2: Discuss what activities, pools and GHGs are considered in your study and why?

Agreed and section 2.2 adjusted

Tables and Figures: They should be self-explanatory. They are hard to follow. Refine all of them. The first sentence of Figure 1 is funny.

Agreed. We changed all captions to make them standalone moved from a “What?”-logic to a “So-what?” logic.

Samples size: Please discuss the sample size? Suggest why they are true representative of total population? What you mean by N°  in Table 2.

Emission factor:  the appropriate sample design strategies have been adopted to ensure the sample size is representative of the total population. For the NFI this is rapidly outlined in section 2.4 and in more detail in a FAO report (see Abena et al in the literature section). That being said, in the conclusions section, we indeed make the recommendation that the sample grid be intensified for the next NFI.

Identification of the drivers of deforestation was also done sample-based; here, our recommendation is “using radar sensors such as Sentinel 1 [106,107,110] and the definition of cut-off dates[111,112] given the often rapid consequential transitions between various land uses” for future assessments, rather than the sample-based approach. That being said, a (densified) sample network will still be needed for validation purposes.

Results What are your key findings? What is the percent of adjustment factors in different time horizons?

Agreed, adjustments have been made in the text, the beginning of section 3.3 is re-written.

Why emission factors of transition of forests to fallow land is lower than annual crop, and why grassland is even more?

To clarify this, we adopted the text to “Emission factors are developed from the difference in above- and below-ground biomass between forest and the land use after clearing…”

Conclusion Write a separate conclusion section

We added more elements of conclusion to the first two paragraphs of the conclusions & recommendations section.

Given the length of the article (29 pages incl literature) we refrained from creating a separate conclusions section. Unfortunately, overly long articles are simply not read.

Some highly relevant literature is missing, see FAO and UNFCCC websites. Include them in the introduction and discussion sections, where appropriate. One of them is here:

Maraseni, T.N., Reardon-Smith, Kathryn (2019) Meeting National Emissions Reduction Obligations: A Case Study of Australia. Energies, 12 (438). https://doi.org/10.3390/en12030438, PP1-13

Agreed and adapted.