Mind the Gap: Reconciling tropical forest carbon flux estimates from Global Earth Observation and National Reporting

doi:10.21203/rs.3.rs-2568897/v1

Background

Different methods estimating the global anthropogenic land flux, which is dominated by forest-related activities, vary in magnitude and direction with respect to whether the land is a net source or sink. One reason for these variations is the extent to which methods consider land to be “managed”, thus contributing to the anthropogenic flux. Earth Observation (EO) datasets characterising spatio-temporal changes in land cover and carbon stocks provide an independent approach to flux estimations that can be compared against National Greenhouse Gas Inventories (NGHGIs) to support accurate and timely monitoring, reporting and verification capacity. Using Brazil as a primary case study, with additional analysis in Indonesia and Malaysia, we compare EO-based estimates of forest fluxes to NGHGIs.

Results

Between 2001 and 2020, the EO-derived estimates of all forest-related emissions and removals indicate that Brazil was a net sink of carbon (-0.2 GtCO₂yr^− 1), while Brazil’s NGHGI reports a net carbon source (+ 0.8 GtCO₂yr^− 1). After adjusting the Global EO estimate to accommodate the Brazilian NGHGI definition of managed forest and other assumptions used in the inventory’s methodology, the Global EO net flux became a source of + 0.6 GtCO₂yr^− 1, comparable to the NGHGI. Remaining discrepancies are largely due to differences in carbon removal factors and forest types considered in the two datasets. In Indonesia, Global EO and NGHGI net flux estimates were similar (+ 0.6 GtCO₂ yr^− 1), but in Malaysia, they differed in both magnitude and sign (NGHGI: -0.2 GtCO₂ yr^− 1; Global EO: +0.2 GtCO₂ yr^− 1). Spatially explicit datasets on forest types were not publicly available in either country, limiting the possibility of detailed analyses and adjustments.

Conclusions

By adjusting the Global EO dataset to improve comparability with carbon fluxes estimated for managed forests in the Brazilian NGHGI, initially diverging estimates were largely reconciled and remaining differences explained. Despite limited spatial data in Indonesia and Malaysia, our comparison indicated where differing approaches may explain uncertainties and inaccuracies. Our study highlights that comparing Global EO and NGHGIs is a useful exercise to improve both datasets towards greater accuracy and alignment, provided that sufficiently transparent and complete information is available for such detailed analyses.

Forests

CO2 flux

LULUCF

removal factors

transparency

Managed Land Proxy

inventories

earth observation

carbon budget

Forest carbon flux estimates in the context of the Paris Agreement

The Land Use, Land-Use Change and Forestry sector (LULUCF) accounted for approximately 15% of total global net anthropogenic CO₂ emissions between 2011 and 2020 (4.6±2.0 GtCO₂yr^-1) (Figure 1). The flux, estimated from three global bookkeeping models, is due predominantly to forest-related activities such as deforestation, afforestation, reforestation, and forest management [1]. With an uncertainty of around ±50%, LULUCF is the most uncertain term in the carbon budget [2]. Prominent differences exist in both the magnitude and sign of the net anthropogenic LULUCF CO₂ flux between global land-related datasets. Earlier studies found a large discrepancy between global bookkeeping models and National Greenhouse Gas Inventories (NGHGI) [3,4], and the Intergovernmental Panel on Climate Change (IPCC) 6^th Assessment Report (AR6) confirmed these discrepancies. The latest update to respective datasets shows a difference of 6.5 GtCO₂yr^-1between NGHGI and global bookkeeping models used by the research community. The NGHGIs reported a small net sink (about -1.7 ±1.0 GtCO₂ yr^-1 over 2001-2020) compared to the models reporting net emissions (about 4.8 ±2.4 GtCO₂ yr^-1 over 2001-2020) [5] (Figure 1). The FAOSTAT global net CO₂ flux trend is similar to NGHGIs. Both are based on country-reported data and FAOSTAT applies an IPCC tier-1 style (simple) methodology (Figure 1). The difference between total fluxes is about 2.7 GtCO₂yr^-1, and has largely been explained [6].

Meanwhile, a discrepancy of about 5 GtCO₂ yr^-1 is also found between Integrated Assessment Models (IAM) and NGHGIs for 2005 to 2015 [7]. This discrepancy is problematic for the Global Stocktake (GST), which evaluates global progress made towards reaching the goals of the Paris Agreement. Historical net LULUCF emissions estimates will be based on NGHGIs [7,8], while countries’ future mitigation actions, as pledged in their Nationally Determined Contribution (NDCs), will be assessed based on IAMs. In turn, IAMs estimate the mitigation needed to limit climate change to the Paris Agreement temperature thresholds.

Earth Observation (EO) data have been used to help understand and reconcile the gap at the global [7] and country scale [9] found between datasets used in the GST. They support climate policy by providing high spatial and temporal resolution data, thus identifying key carbon sources and sinks beyond afforestation and deforestation that are often not captured in NGHGIs (e.g. forest degradation) [10]. Global models based primarily on EO data, that incorporate spatially explicit, empirical observations to estimate carbon fluxes, may aid countries’ monitoring, reporting and verification (MRV) of their NGHGI, and potentially provide a benchmark to evaluate the land sink from other global non-EO models.

A recent Global EO-based model mapping average forest-related greenhouse gas (GHG) emissions and removals by Harris et al. (2021) [11] provided the first integrated, globally consistent and geospatially explicit assessment of forest carbon fluxes over the period 2001 to 2020 (Figure 1). The dataset used an inventory framework, following the same IPCC guidelines used by NGHGIs such that it encompasses many of the same processes that NGHGIs are encouraged to consider. However, globally the EO analysis found a large net sink (-6.7 GtCO₂ yr^-1) with a difference of about 5 GtCO₂ yr^-1 between it and the NGHGIs. This stark discrepancy was observed despite adjustments made to the Global EO to exclude fluxes in primary forests/non-intact forests (Figure 1), a proxy for unmanaged lands, which are not considered by NGHGIs [7]. Including EO data had the opposite effect from what was anticipated, rather than increasing confidence in LULUCF estimates, the new data increased the range of reported fluxes (Figure 1). Our study analyses this gap and proposes methods to reconcile estimates to increase confidence.

Mind the gap: Reconciling land-use flux estimates

By using observations alone, it is not possible to directly distinguish anthropogenic and non-anthropogenic fluxes as defined by IPCC (2010): (a) direct anthropogenic effects (e.g., deforestation and land management), (b) indirect anthropogenic effects (e.g. CO₂ fertilisation, anthropogenic climate change-induced temperature changes) and (c) natural effects (e.g., interannual climate variability). Anthropogenic and natural effects overlap in space and time. It is only possible with model-based assumptions to artificially separate the effects. As a pragmatic solution to trying to distinguish anthropogenic fluxes, the IPCC Good Practice Guidelines for Inventory methodologies developed the Managed Land Proxy, whereby NGHGIs consider all GHG fluxes on managed land to be anthropogenic [12,13]. Each country can define “Managed Land” in its own way, adding further complexity to any comparisons between datasets. Countries’ NGHGIs collectively consider a larger land area as “managed” compared to the assumptions applied in global models. Models consider reported harvested wood volume as coming from clearcutting, thus thinning and selective logging in less intensively managed lands are not modelled as part of “managed” land [7]. Additionally, each country has different capacities, so each inventory varies in transparency, consistency, comparability, completeness and accuracy (Supplementary Table 1) – the five UNFCCC ‘principles’ for reporting NGHGIs [14].

The main reason for the gap between NGHGIs and global non-EO models (bookkeeping and IAMs) has now broadly been explained, with approximately 80% of the discrepancy reconciled by considering the extent to which each method defines the forest sink as “managed” and thus anthropogenic [3,7]. Other factors certainly play a role in explaining the difference, including incomplete NGHGIs and a simplified representation of forest management in global models, but these factors likely partially compensate each other. Using an EO-derived map of intact and non-intact forest areas, Grassi et al. (2021) adjusted estimates from the IAMs by considering not only all direct anthropogenic effects but also indirect anthropogenic effects (identified by Dynamic Global Vegetation Models) occurring in non-intact forest areas. This adjustment expanded the definition of managed forests initially considered by the IAMs to better match that of the NGHGIs.

To date there has yet to be an equivalent detailed, spatially-explicit attempt to understand the differences between NGHGIs and Global EO data, as there has been between NGHGIs and global non-EO models. Such an analysis would (i) improve the credibility of forest-related anthropogenic fluxes when comparisons between datasets are made, (ii) demonstrate how EO data can be useful in producing and improving NGHGIs, and (iii) make EO a useful verification and evaluation tool for NGHGI, by reflecting what countries are currently measuring but also highlighting missed emissions and removals sources that should be measured, considering what current NDC targets include and how these could be made more ambitious. Ultimately, such an analysis will determine to what extent the gap between Global EO based estimates and NGHGIS can be reconciled.

The gap between NGHGIs and Global EO estimates is demonstrated clearly at the global scale (Figure 1), however, reasons for the discrepancy varies by country, with methodological differences in each NGHGI. While a global-scale analysis is certainly interesting, for the purposes of MRV, the discrepancies can only really be understood and resolved at the country-scale [9]. Brazil is the largest contributor to LULUCF emissions (~25% of global LULUCF emissions) [2,3,15], however, key biomes are also large gross sinks, and Amazonia makes up about 8% of the total global land sink [2,16]. Indonesia is the second largest contributor to LULUCF emissions, accounting for about 10% of global LULUCF emissions [2]. Brazil and Indonesia have pledged to protect and restore forests within their NDC and thus forests are critical to these countries to meet their climate targets [17].

Recently numerous peer-reviewed studies and high-profile analysis by news articles have highlighted the gap that exists between GHG fluxes estimates, and not just in the LULUCF sector [4,18]. An article published by the Washington Post found large differences between land-based flux estimates in Malaysia’s NGHGI and independent research papers [19]. With these motivations, our intention for this study was to compare forest-related flux estimates from NGHGIs from key case study countries, Brazil, Indonesia, and Malaysia with those from a Global EO-based analysis.

Brazil’s NGHGI is regarded as one of the most complete inventories produced by Non-Annex 1 countries, with a high transparency in describing the methodology and making all datasets freely available. This makes Brazil an ideal case study to compare different datasets estimating gross forest-related emissions and removals, including methodological developments for assessing the role of satellite data. For Indonesia and Malaysia, the land-cover datasets used in the NGHGIs, are not publicly available for spatial analysis and download. Nevertheless, tabular summary statistics in the NGHGIs enabled some analysis.

Using Brazil as the primary case study but also applying the same principles as far as possible to other case studies, the aims of our work were to: (i) compare estimates of forest-related greenhouse gas (GHG) fluxes from a Global Earth Observation dataset and NGHGIs in three countries with large areas of tropical forest, (ii) outline potential reasons for differences between estimates, including the impact of different definitions of managed forest, and (iii) assess implications for future scientific research and for use of EO datasets in a policy context.

Flux datasets

(i) Global Earth Observation

The Global EO dataset used here [11] is a globally consistent framework that standardises key aspects of forest carbon inventories, including scope, definitions, assumptions and the level of transparency and completeness of the method or approach used (Table 1). The framework used the most recent IPCC Good Practice Guidelines [20] and applied NGHGI gain-loss methods to each Landsat pixel (~30x30 m) of forest (tree canopy >30% in 2000 or subsequent tree cover gain); the framework captured transitions to and from forest, as well as forest remaining forest, but did not capture non-forest-related land uses [11]. The framework did not differentiate between managed and unmanaged forests, but the spatially explicit data allows for fluxes to be delineated. The framework’s initial conditions were a global map of aboveground biomass density in the year 2000 [21] in conjunction with forest extent in 2000 [22] assigned to different forest types and ages, including planted forests [23]. Activity data were based on tree cover change [22], fires [24], and drivers of tree cover loss [25]. Emissions included all carbon pools (aboveground, belowground, dead wood, litter, soil organic carbon) and emissions from fires and peat drainage and burning. Removal factors came from a variety of sources and were applied using a “stratify and multiply” approach and included aboveground and belowground removals. Global geospatial data were used wherever possible (e.g. a wall-to-wall map of removal factors for young, naturally regenerating forests based on a literature review [26]), and IPCC Tier 1 defaults values representative for large-scale ecoregions were used otherwise (old-growth forest, old secondary forest and plantation) [20]. The final product was global maps of modelled average annual forest-related GHG emissions, removals and net flux from 2001 to 2019, but were updated to 2020 at the time of this analysis (V1.1).

Table 1 | The main driving sources in three flux datasets estimating forest-related greenhouse gas fluxes for Brazil. The three datasets are: Global Earth Observation-based (EO) dataset [11], Brazil’s National Greenhouse Gas Inventory (NGHGI) and an independent in-country inventory-style method SEEG (Sistema de Estimativa de Emissão de Gases de efeito estufa) [29]. Driving sources, as well as input datasets and assumptions made my datasets are summarised. *Method used depends on transition type e.g. deforestation is stock-difference, regrowth is gain-loss.

Driving source	Global EO	NGHGI-Brazil	SEEG-Brazil
Version/Update	V2	NC4	9.0
Time period	2001 to 2020 (whole period average)	1994 to 2016 (6-to-8-year intervals) Manually adjusted in this study to: 2001 to 2020	1990 to 2020 (annual)
IPCC Method	Gain-Loss	Gain-Loss* Stock-Difference*	Gain-Loss* Stock-Difference*
Activity data - Satellite used - Main driving dataset	- Landsat - Hansen et al. 2013	- Landsat &Resourcesat-1 - RADAMBRASIL, PROBIO	- Landsat - MapBiomas
Spatial resolution of land use dataset	~30m	~30m	~30m
Emission/removal factors	Emissions: wall-to-wall aboveground biomass and soil carbon supplemented by wall-to-wall driver of loss map. Removals: Various sources for different forest types.	Country-specific (~IPCC Tier 2/3 methodology) data based on field inventories and literature	Follows NGHGI
Forest types considered	- Primary Forest - Non-primary forest (old secondary forest) - Young secondary forest - Plantations (Forest plantations and tree crops) - (Mangroves – excluded in this analysis)	- (Non-managed forests) - Managed forest - Secondary forest - Plantations (Forest plantations only)	-Protected Forests* -Secondary forest -Plantation (Forest plantations only) *equivalent to managed old-growth forest in the NGHGI
Forest transitions considered (Land use and land use change)	- Forest remaining forest land - Deforestation - (Degradation) – considered retrospectively and non-spatially (Pearson et al. 2017) -Afforestation and reforestation	-Forest remaining forest land - Deforestation - Selective logging (Amazonia only) -Afforestation and reforestation	-Forest remaining forest land - Deforestation - Degradation by fire -Afforestation and reforestation

(ii) Brazil’s National Greenhouse Gas Inventory

In the fourth National Communication (NC4) [27], Brazil’s gross emissions and removals were reported for the periods 1994 to 2002, 2002 to 2010 and 2010 to 2016 (Table 1) and for the Amazon biome, also for 2002 to 2005 and 2005 to 2010. Brazil’s NGHGI spatially delineates which lands are unmanaged and managed and considers temporal changes in the Land Use and Land Cover (LULC) types using remote sensing data [28]. The LULC maps are combined with spatially explicit information on the carbon pools and associated fluxes based on field data, literature values and remote sensing. The information is used to produce matrices showing the carbon emissions and removals within managed LULC types and transitions.

Emissions and removals in non-managed lands are not reported in the flux estimates, nor are fluxes in areas that have been classified as secondary forests throughout the period of analysis areas. To use the IPCC terminology, no carbon fluxes are considered on “secondary forest lands remaining secondary forest” (Table 2). The same approach is applied to forest plantations remaining forest plantations. Processes occurring within these categories are assumed to be in equilibrium e.g., shifting cultivation and harvest. The key categories reported include emissions and removals in managed old-growth forests and emissions and removals in other land converted to secondary forest or forest plantations over the period analysed, i.e. very young secondary forests/plantations, for example a pasture in 2002 that was detected as secondary forest in 2010 (Table 2).

The data can be viewed in a user-friendly dashboard and are available freely to download for detailed analysis following registration (www.ccst.inpe.br/cn/), ensuring transparency. To compare the NGHGI with the Global EO estimate, we only considered emissions and removals occurring in forest-related categories from both datasets, namely, “forest land remaining forest land”, “forest land converted to other lands”, and “other land converted to forest land”. Forest classification in the NGHGI is the same used by the Food and Agriculture Organization (FAO) of the United Nations [29].

Table 2 | Consideration of removals in different forest types by three flux datasets to estimating forest fluxes. Forest types used in the Global Earth Observation (EO) dataset [11] and an independent estimate in Brazil – Sistema de Estimativas de Emissões e Remoções de Gases de Efeito Estufa (SEEG) have been aligned according to the IPCC LULUCF reporting categories [64] and are compared to the Brazilian National Greenhouse Gas Inventory (NGHGI). *Removals in this category are available but were not considered following a series of adjustments. ** Removals in this category are included by the Brazilian NGHGI and SEEG but tree crops are considered in the cropland category rather than forest land.

(iii) Other datasets: SEEG and FAOSTAT

While the primary aim of our study was to compare the GHG flux estimates from the Global EO and the NGHGI for completeness and to aid our understanding of associated uncertainties, we also included estimates from another in-country dataset in Brazil and the FAOSTAT.

Brazil’s System for Estimating Greenhouse Gas Emissions (Sistema de Estimativas de Emissões e Remoções de Gases de Efeito Estufa – SEEG) is independent of the NGHGI and aims to produce annual estimates of GHG emissions in Brazil [30]. SEEG is updated annually using the latest LULC time series products from the ‘MapBiomas’ project as the input activity data to quantify the GHG flux estimates for the LULUCF sector [31] (Table 1).

The SEEG dataset follows the approach of NGHGI and only considers emissions and removals associated with anthropogenic activities, considering conservation and indigenous lands to be protected and thus managed. A key difference is that the identification of the LULC type and annual changes in the SEEG is pixel-based, as opposed to the NGHGI which is in part based on expert visual interpretation assessment. The SEEG methodology broadly uses the same emission and removal factors as the NGHGI [32] but additionally considers fluxes in “secondary forest remaining secondary forest” (Table 2). The gross emissions and removals within the LULUCF sector can be downloaded freely from the SEEG platform and can be filtered by (i) Biome and State, (ii) Land use type, (iii) Land use change type and (iv) Land cover type (https://plataforma.seeg.eco.br/sectors/mudanca-de-uso-da-terra-e-floresta) [33]. To make the estimates from the SEEG comparable with the Global EO, we only considered CO₂ emissions and removals occurring in forest-related categories.

We also included estimates from FAOSTAT, which are not spatially explicit, but reported as countrywide numbers. FAOSTAT considers only the aboveground and belowground carbon pools and the approach is comparable to IPCC 2006 Tier 1 guidelines [34]. We extracted the average forest-related CO₂ flux for the period 2001 to 2020 by considering the categories ‘Forestland’ to represent ‘forest land remaining forest land’ and ‘Net Forest conversion’ to represent ‘forest land converted to other land’ [35].

Estimates of GHG fluxes from different datasets

The Global EO dataset provides the emission and removals estimates as a flux per hectare in a wall-to-wall map (~30m resolution). The data can be aggregated within different boundaries by considering the total area of the pixels and associated fluxes within. Neither the NGHGI nor the SEEG provide spatially explicit information on the emissions and removals, but the input LULC dataset for SEEG (MapBiomas) and the NGHGI are spatially explicit. These were used to compare spatial overlap of land considered as forest in Brazil between the various approaches, specifically “managed forest” by the NGHGI (Table 1). The NGHGI managed land maps enabled us to extract the gross emissions and removals of the Global EO by forest types.

Given the shorter period covered by the Brazilian NGHGI (2002 to 2016) (Table 1), we modified the estimates of the NGHGI to account for the temporal differences. This was important given the high emissions associated with forest loss in early 2000s [33]. Using the annual SEEG data, we calculated the fractional difference between average gross fluxes over 2001 to 2020 (the period available for the Global EO) and 2002 to 2016 (the period available for the NGHGI). We then multiplied the fractional difference by the 2002 to 2016 NGHGI estimates, both at the biome and country scale, such that the approaches and the estimates were comparable over a common time-period (Supplementary Table 2).

Adjustments to the Global EO forest-related flux to make it comparable to the other approaches in Brazil

The Global EO data does not differentiate between managed and unmanaged forests. To make the approaches more comparable, we analysed the impact of different adjustments to what might be considered “anthropogenic” forest-fluxes within the Global EO approach:

No adjustment – We extracted all Global EO gross emissions and removals occurring within the country boundary of Brazil and compared this estimate with the NGHGI and SEEG, which only consider emissions occurring on managed lands.
Adjustment 1 – We applied a similar approach to Grassi et al. (2021) [36]. Instead of using an intact-forest mask, we used a primary forest mask for the year 2001 as a proxy to exclude removals in non-managed lands [37]. Additionally, we considered all emissions in the territorial boundary of Brazil instead of only in the non-primary forests to include post-2001 forest cover losses and associated emissions in these areas.
Adjustment 2a – We applied the 2016 LULC map from the NGHGI to the Global EO map considering emissions and removals occurring on all managed lands as defined by the NGHGI (both forest and non-forest).
Adjustment 2b – We considered the gross emissions occurring on all managed lands (as Adjustment 2a). However, we only considered the gross removals in managed forests as defined by the NGHGI in 2016. The distinction allowed us to understand if there were any differences in the extent of forest land used by the NGHGI and Global EO and the impact this would have on the flux estimates.
Adjustment 2c – A final adjustment was made in addition to Adjustment 2b. It followed the approach used by the NGHGI of excluding removals in “secondary forest remaining secondary forest” or removals in “plantations remaining plantations” (Table 2). We excluded these categories, or those as closely matching these categories, in the SEEG and Global EO. In the SEEG dataset, we excluded removals occurring in “secondary forest remaining secondary forest”. In the Global EO, we considered old secondary forests (> 20 years old) to represent the IPCC category “secondary forest remaining secondary forest” and thus excluded removals in old secondary forest areas. We also excluded gross removals in plantation areas that did not experience any forest loss from 2001 to 2020, according to the dataset used by the Global EO [22], to represent “plantations remaining plantations”.

Table 3 | The main adjustments made to align the Global EO forest definitions with the Brazilian National Greenhouse Gas Inventory (NGHGI). Adjustments have been broken down into gross removals (blue) and gross emissions (yellow) in different forests type and how these are considered according to the IPCC categories. Ticks indicate that the flux was included in the adjustment, crosses indicate that the flux was excluded in the adjustment. Symbols proceeded by the letter ‘R’ refer to the removals flux component, symbols proceeded by the letter ‘E’ refer to the emissions flux component.

Evaluation of the removal factors and forest types

In addition to evaluating the discrepancy attributable to managed forest area, we considered differences in the areas of forest types and associated removal factors used in each approach. We identified the removal factors used for different forest types from the spatial dataset of the NGHGI [38]. The NGHGI uses peer-reviewed data available for individual biomes [28]. The SEEG uses the same removal factors as the NGHGI and so were not included in the comparison. We calculated the modal removal factor used by the Global EO dataset for each forest type (Table 1) to be compared with the NGHGI biome-specific removal factors.

The forest types used by the NGHGI (Table 1) are available spatially explicitly for the years considered in the NGHGI [38]. The forest types used by the SEEG are from the MapBiomas dataset [39]. MapBiomas provides pre-processed annual data on deforestation and secondary vegetation, which can be combined with their annual LULC map to obtain a forest type map in 2020. The forest types used by the Global EO were sourced from the various datasets, and included old-growth forest, old secondary forest (> 20 years old), young secondary forest (<20 years) and plantations (Table 1) [11].

Assessing comparability of fluxes with high uncertainty

The LULUCF sector is the most uncertain component of the Global Carbon Budget (~50% for globally aggregated data) [2]. In the tropics, the Global EO method estimated a propagated standard deviation of 45% and 110% for the gross emissions and removals, respectively [11]. The high removals uncertainty was driven by a high uncertainty in Tier 1 removal factors for some tropical ecozones [11]. Brazil’s NGHGI reports an uncertainty of 32% for their estimate of LULUCF CO₂ emissions and removals [28]. Given the high levels of uncertainty, using the traditional statistical method of determining if error bars overlap to assess the significance of the results would not be a very useful metric, as the estimates and their associated uncertainties would always overlap. Comparing the results from multiple approaches, e.g., SEEG and FAOSTAT, provides a clearer understanding of the relative uncertainty.

Application of the analysis in other case study regions

A detailed analysis, like the Brazilian case study, was not possible for Indonesia and Malaysia as the LULC maps are not publicly available for download. The uncertainty of Indonesian and Malaysian NGHGI is about 14% and 7%, respectively, when excluding the LULUCF sector. These values increase to 20% and 66% for Indonesia and Malaysia, respectively when LULUCF is included. LULUCF is therefore a key source of uncertainty in these two countries [40,41].

Indonesia’s most recent NGHGI in the third Biennial Update Report (BUR3) includes annual data from 2000 to 2019 [40]. Given that the Global EO is only available as an average value for 2001 to 2020, we extended the emission and removals value of Indonesia’s BUR3 for 2019 to represent 2020 and then took the 2001 to 2020 average for a more equal comparison. Data for Malaysia is from their BUR3, covering the period 2001 to 2016 [41]. We extended the data from 2016 to represent 2017 to 2020 and took the 2001 to 2020 average. Key tree crops such as oil palm and acacia are considered agricultural croplands in the NGHGI of both countries; we excluded these tree crop plantations from the net flux estimate of the Global EO dataset in our comparison (similar to Adjustment 2c made in Brazil). We excluded fluxes within the categories “other land converted to cropland” and “cropland remaining cropland” in the NGHGIs, as well as other categories not linked to forests (Supplementary Table 4).

Unlike Brazil, Indonesia and Malaysia consider all lands within their country boundary to be managed. Thus, Global EO emissions and removals within the entire boundary of the respective countries can be considered in comparison with the NGHGI. Both Indonesia [40] and Malaysia [41] report emissions and removals in key IPCC categories similar to those outlined in Table 2.

Making the extent of managed land more comparable for flux estimates in Brazil (Adjustments 1, 2a, 2b)

Assuming all forest gross emissions and removals within the country boundary of Brazil were considered as managed (No adjustment), the Global EO net flux would be -0.2 GtCO₂yr^-1, a small sink from 2001 to 2020 (Figure 2: 1^st bar). Over the same period, the other flux datasets, the NGHGI, SEEG and FAOSTAT, report a net source of 0.8 GtCO₂yr^-1, 0.6 GtCO₂yr^-1, 0.7 GtCO₂yr^-1, respectively (Figure 2: bars 5 to 7). The extent of the managed land definition in the Global EO dataset was adjusted to make it more comparable with the NGHGI and other flux datasets, and the forest flux was recalculated (Figure 2). When considering gross emissions on all lands in the country boundary and only removals occurring in non-primary forest lands in the Global EO dataset (Adjustment 1) (Figure 3a), the forests become a net source of 0.5 GtCO₂yr^-1 (Figure 2: 2^nd bar). While the Global EO net flux is only about a third lower than the NGHGI, the gross emissions and removals remain higher. The differences likely arise because the spatial extent of these forest types is not similar in Brazil (Figure 3a and c). Spatially, the non-primary forest lands only overlap with 11% of managed forests in the NGHGI (Supplementary Table 3). When aggregated the total “non-primary forest” area (507Mha) is very similar to the NGHGI “managed forest” area (397 to 484 Mha), emphasising the importance of spatial analysis when considering proxies for managed forests (Supplementary Figure 1).

Applying the NGHGI spatial mask of all managed land (Figure 3b) to the Global EO flux dataset yielded a net forest flux of 0.1 GtCO₂yr^-1, making it a small source (Figure 2: 3^rd bar) (Adjustment 2a). Gross emissions are very similar to other datasets, with the estimate from the Global EO being 0.13 GtCO₂yr^-1 smaller compared to the NGHGI. Gross removals remain ~0.6 GtCO₂yr^-1 higher than the NGHGI. This suggests that deforestation areas and emissions factors are similar in Brazil’s NGHGI and the Global EO, and that most of the difference is due to differences in forest classification and removal factors applied in managed forests.

With an additional Global EO data adjustment to consider removals occurring only within “managed forest” as defined by the NGHGI but still include emissions from all managed land (Figure 3c), the net flux was a source of 0.4 GtCO₂yr^-1(Figure 2: 4^th bar) (Adjustment 2b). Despite the same definition of managed forest, the gross removals flux of the Global EO data is two fifths larger than the NGHGI.

Dissaggregating Adjustment 2b at the biome scale, differences between the flux datasets become more apparent (Figure 2 and Figure 4). In Amazonia, the biome with the biggest contribution to the net flux (69% to 78%), the gross emissions in the Global EO (0.8 GtCO₂yr^-1) are very similar to the NGHGI (0.9 GtCO₂yr^-1). However, gross removals are 1.5 times lower in the NGHGI. In the Cerrado and Atlantic Forest, which combined make up approximately 15.7% to 34% of the net flux, gross removals are up to two and three times higher, respectively, in the Global EO compared to the NGHGI (Figure 4). Given that these comparisons only included gross removals in managed forests for the Global EO data, the remaining differences observed at the biome and country scale cannot be reconciled by only considering the extent of forest cover and whether it is managed forest or not (Figure 2 and Figure 4).

We found considerable spatial differences when comparing the extent of forest cover in the Global EO and NGHGI datasets (Figure 5a and Figure 5b). The largest differences are outside of the humid forest-dominated Amazon biome. In the Atlantic Forest, the other key humid forest biome, only 35% of the total potential forest area (the areas considered as forest in either or both datasets) was classified as forest cover in both datasets. Across the other four biomes, which are not dominated by humid-forest cover, the area considered to be forest is higher in the NGHGI. These regions were not considered forest in the Global EO study as their tree canopy cover per pixel was less than 30%. The NGHGI does not define such a threshold and considers the FAO classification [27,28].

Relative contribution of different forest types to gross removals in Brazil

We found that most of the discrepancy was in the gross removals component of the forest flux. To explore potential reasons for this discrepancy, we disaggregated the removals flux according to forest types included in each flux dataset and the area occupied by each forest type in 2020 (Table 4). The difference in area occupied versus relative gross removals of respective forest types also varies across the six biomes (Supplementary Figure 3 and Supplementary Figure 4). The reasons for differences in both area and gross removals are interconnected and difficult to untangle.

Table 4 | The 2020 area of different forest types and associated carbon removal flux according to different datasets in Brazil. The carbon removal component is for the average annual removal flux for the period 2001 to 2020. For an explanation of different data sources see methods. For the National Greenhouse Gas Inventory (NGHGI) [28] only pixels in the managed forest area outlined in NC4 were used (green regions in Figure 3c). *As the NGHGI is only available up to 2016, numbers for the NGHGI have been adjusted by multiplying by the fractional difference in the area/gross removals of each forest type in 2016 and 2020 according to SEEG (MapBiomas) SEEG-Brazil [30]. **This value also includes selective logging areas and gross removals from Amazonia. *** These values are based on the Adjustment 2b – i.e., using the NGHGI managed forest lands to extract area and removals in the Global Earth Observation (EO) [11].

	Global EO***		NGHGI-Brazil		SEEG-Brazil
Forest type	Area (2020) (1000 ha)	Gross removals (GtCO₂ yr^-1)	Area (~2020)* (1000 ha)	Gross removals* (GtCO₂yr^-1)	Area (2020) (1000 Mha)	Gross removals (GtCO₂yr^-1)
Managed old-growth	195,954	-0.42	220,158	-0.32	217,702	-0.31
Secondary forest	23,998	-0.16	21,877**	-0.05**	8,547	-0.16
Plantation	6,779	-0.19	11,144	-0.08	6,311	-0.003
Other land	31,732	-	5,284	-	25,903	-
TOTAL	258,463	-0.77	258,463	-0.45	258,463	-0.48

According to the NGHGI, plantations occupy an area of approximately 11Mha and have a gross removals flux of -0.08GtCO₂ yr^-1. According to the Global EO estimate, plantations occupy about two-thirds of this area (6.8Mha) but estimated removals in plantations are more than double compared to the NGHGI estimate (-0.19 GtCO₂ yr^-1) (Table 4). This can partly be explained by the fact that the NGHGI and SEEG do not consider removals in “plantations remaining plantations” (Table 2). Furthermore, tree plantations such as rubber, acacia and oil palm are considered tree plantations in the Global EO dataset, whereas, in the NGHGI and SEEG, they are considered as cropland and are, therefore, not included in forest-related removals (Table 2). Given that we adjusted the Global EO gross removals to only consider “managed forest” areas according to the NGHGI (Adjustment 2b), the relative contribution to both the area and gross removals of tree crops to the final adjusted Global EO is small (Supplementary Figure 2).

The Secondary Forest area in the SEEG estimate (8,547Mha) is two-thirds smaller than the NGHGI (21,877Mha), yet gross removals are three times higher in SEEG (-0.16 Gt CO₂yr^-1), compared to the NGHGI (-0.05 Gt CO₂yr^-1). Secondary forest removals in the Global EO are also three times higher than in the NGHGI, despite occupying a similar total area (Table 4). The difference can partly be explained by the fact that the SEEG considers removals in “secondary forest remaining secondary forest” but the NGHGI does not (Table 2). Across the three datasets, the area of old-growth forest is approximately the same (~20Mha), and the gross removals are a third larger in the Global EO (-0.4 GtCO₂ yr^-1) compared to the NGHGI and SEEG (~0.3 GtCO₂ yr^-1) (Table 4), but within the realms of uncertainty of the NGHGI dataset (32% - see Methods).

Reconciling the difference in gross removals between flux datasets in Brazil (Adjustment 2c)

We made a final adjustment (Adjustment 2c) by excluding older areas of plantations and secondary forests from the Global EO flux data (Figure 6). This adjustment accounted for the fact that the NGHGI treats fluxes within these categories as net zero (see Methods, Table 2). Considering this final adjustment, the difference between the NGHGI and Global EO/SEEG is halved compared to Adjustment 2b. The NGHGI net flux remains higher (0.81 Gt CO₂yr^-1) but only by 25% and 10% for the Global EO (0.61 Gt CO₂yr^-1) and SEEG (0.72 Gt CO₂yr^-1), respectively (Figure 6).

Explaining the remaining discrepancy: removal factors and forest type

There are a few reasons why a gap between the Global EO and other flux datasets remains after Adjustment 2c despite considering the same areas of managed forests and associated fluxes; these may be linked to the fact that each flux dataset used different (i) removal factors and (ii) spatial extents of forest type. By comparing the removals factors and forest types used in the various datasets, we can explore the remaining discrepancies without changing the methodology used by the flux datasets, which would go beyond the scope of this study.

Looking at the differences in removal factors used for old-growth forests across the Brazilian biomes, which make up 55-70% of the gross removals flux (Table 4), we see that IPCC Tier 1 removal factors used by Global EO dataset are larger compared to the NGHGI (Table 5). In Amazonia, the removal factor used by the Global EO is a fifth higher than the NGHGI removal factor (Table 5). This helps to explain the higher gross removals in this forest type across the country (Table 4) and regionally, e.g., in Amazonia (Supplementary Figure 4). In the other biomes, the removal factors are higher, by between 20-850% (Table 5). However, the relative contribution of old-growth forest carbon removal to the total gross removals is considerably less than Amazonia (Supplementary Figure 4).

Table 5 | Old-growth Forest removal factors used by different flux datasets and disaggregated by biomes in Brazil. For all datasets, values include above and below ground carbon. The National Greenhouse Gas Inventory (NGHGI) and SEEG) [30] applied a single, biome-specific, removal factor for each biome. The calculated modal removal factor in each biome for the Global Earth Observation (EO) dataset [11] (see Methods) is shown. Values in brackets indicate old-growth removal factors also used but where the associated ecozone did not account for a large area within the biome. In the Pampa biome the Global EO did not detect any old-growth (primary forest) areas and only old secondary forest (Old SF) regions were identified.

Biome	NGHGI and SEEG (Mg C ha^-1yr^-1 )	Global EO (Mg C ha^-1yr^-1)	Percentage Difference (Global EO/NGHGI)
Amazonia	0.48	0.59 (0.24)	+ 23%
Atlantic Forest	0.44	0.59 (0.24)	+ 34%
Cerrado	0.2	0.24 (0.59)	+ 20%
Pantanal	0.2	0.24	+ 20%
Pampa	0.44	NA (0.59 = Old SF)	+ 34%
Caatinga	0.1	0.95	+ 850%

The removal factors in forest plantations are generally larger in the Global EO method, compared to the NGHGI, with a range of 9.5% to 59% (Table 6). In young secondary forests, those regrowing for less than 20 years, the Global EO average removal factors are also generally larger, between 29% and 543% (Table 6).

The NGHGI does not include removals in ‘secondary forest remaining secondary forest’, and only considers removals in ‘other land converted to forest land’, thus it only uses removal factors applicable for younger (<20 years) secondary forest. The distinction between old and young secondary forest regrowth factors is, however, made by the Global EO product. All forested regions not identified as primary forests, having tree cover gain, plantations, or mangroves in 2000 are classified as ‘old secondary forests’ (>20 years). The appropriate removal factors are then used according to the updated IPCC guidelines [42]. The average removal factors of old secondary forest in the Global EO are lower than the NGHGI young secondary forests removal factors, but higher than old-growth forests ones (Table 6).

Table 6 | Plantation and Secondary Forest removal factors used by different flux datasets and disaggregated by Brazilian biomes. For all datasets, values include above and below ground carbon. Secondary forests are further disaggregated by forest-age classes. This shows the average removal value for each biome, which in the National Greenhouse Gas Inventory (NGHGI) varies according to the type of land use transition taking place, and in the Global Earth Observation (EO) [11] varies depending on the type of forest plantation. Values in brackets denote the range of possible removal factors applied by each method in the respective biomes. The calculated modal removal factor in each biome for the Global EO dataset (see methods) is shown. Values in brackets indicate removal factors also used but where the associated ecozone did not account for a large area within the biome. Note the NGHGI does not distinguish between old and young secondary forest and only applies a single removal factor.

Biome	NGHGI (Mg C ha^-1yr^-1)	Global EO (Mg C ha^-1yr^-1)	Percentage Difference (Global EO/NGHGI)
Plantations
Amazonia	12.5 (8.3 to 12.66)	11.2 (4.72 to 20.3)	- 10.4%
Atlantic Forest	11.6 (10.5 to 12.6)	12.7 (4.72 to 20.3)	+ 9.5%
Cerrado	12.6 (11.1 to: 12.6)	16.6 (4.72 to 20.3)	+ 31.7%
Pantanal	12.7 (12.7 to 12.7)	20.2 (4.72 to 20.3)	+ 59.1%
Pampa	11.0 (11.0 to 11.0)	12.9 (4.72 to 20.3)	+17.3%
Caatinga	12.6 (12.1 to 12.7)	17.0 (4.72 to 20.3)	+ 34.9%
Young secondary forests (<20 years)
Amazonia	3.1 (0.6 to 5.2)	6.4 (3.9 to 8.0)	+ 106%
Atlantic Forest	1.7 (1.7 to 1.7)	4.5 (2.0 to 7.6)	+ 165%
Cerrado	2.7 (0.6 to 4.7)	4.3 (2.5 to 7.3)	+ 59%
Pantanal	2.8 (0.6 to 4.7)	3.6 (2.3 to 5.5)	+ 29%
Pampa	3.2 (0.6 to 4.7)	2.8 (1.6 to 5.1)	- 13%
Caatinga	0.7 (0.6 to 1.0)	4.5 (2.5 to 7.2)	+ 543%
Old secondary forests (> 20 years)
Amazonia	-	1.36 (1.60)	-
Atlantic Forest	-	1.36 (1.60)	-
Cerrado	-	1.60 (1.36)	-
Pantanal	-	1.60	-
Pampa	-	0.59	-
Caatinga	-	1.60	-

The other remaining discrepancy, related to the spatial extent of forest types assigned to specific pixels across the three flux datasets, will also dictate the type of removal factors applied. At the country-scale, we quantified differences in the classification of forest type between the datasets (Figure 7). The most noticeable difference is in the classification of managed old-growth forests in the NGHGI and SEEG and ‘secondary forest remaining secondary forest’ in the Global EO. Here, 65% and 85% of the pixels classified as Secondary Forest (Forest land remaining forest land, FL -> FL) by the Global EO method were classified as managed old-growth forest by the NGHGI and the SEEG, respectively (Figure 7). Of the forest-cover types, the classification of managed old-growth (FL -> FL), plantations remaining plantations (FL -> FL) and other land converted to forest land (OL -> FL) were the most consistent across the three datasets, with up to 98% consistency between the Global EO and the NGHGI.

The differences in removal factors and forest types used by each flux dataset highlights the individual approaches, boundary conditions and methodological priorities. It is therefore not possible to reconcile the flux further, without entirely re-doing the work of each separate flux dataset. The remaining differences in the flux between the Global EO (0.61 Gt CO₂yr^-1), NGHGI (0.81 Gt CO₂yr^-1) and SEEG (0.72 Gt CO₂yr^-1) (Figure 6) give a measure of the uncertainty of gross and net flux estimates for forests in Brazil.

Differences in the key carbon pools with implications for gross emissions in Brazil

Differences in the major carbon pools and how the various datasets handle deforestation and degradation will influence the emission factors and help to explain some of the observed differences in gross emissions.

The Brazilian NGHGI, SEEG and Global EO consider all five carbon pools: aboveground carbon, belowground carbon, dead wood, litter, and soil organic carbon, but apply different assumptions. The Global EO applies IPCC default ratios to estimate belowground carbon from aboveground carbon, default values for deadwood and litter, and a global map of soil organic carbon in mineral soils. Where available, the NGHGI uses biome specific conversion ratios for the belowground carbon and specific values for deadwood and litter and otherwise uses IPCC default values.

For estimating Aboveground Carbon (AGC), the Global EO uses a pixel-based, remote sensing approach [16] and hence the spatial variability of AGC estimates is greater compared to the NGHGI (Figure 8). The NGHGI prioritises using structural field-derived data from various in-country inventories across the six biomes and literature derived estimates [28]. In both Amazonia and the Atlantic Forest, the mean old-growth forest AGC values for the Global EO are similar to the NGHGI (Figure 8), which may partly be because both flux datasets rely on field-based calibrations applying similar allometric equations. The estimates diverge more in the other biomes; however, the interquartile ranges (IQR) generally overlap.

There are differences in AGC between the Global EO classification of old-growth forests and old secondary forests. In Amazonia and Cerrado, the IQR of the old-growth and old secondary forests do not overlap, suggesting differences in the forest aboveground carbon dynamics between these two classifications. The NGHGI does not distinguish between old-growth forests and old secondary forests. It is therefore not clear if the NGHGI forest land AGC estimates are derived from old-growth forests or old secondary forests (Figure 8).

The method to determine deforestation and degradation varies across the flux datasets and influences the gross emissions. The Global EO uses a remotely sensed dataset to identify forest cover loss and includes losses associated with stand-replacing disturbances such as fire and logging. Smaller scale degradation events (< 30m) may go undetected and, therefore, unquantified by the medium resolution satellite observations [11]. The representation of degradation in the Global EO dataset is, therefore, incomplete. The NGHGI only considered degradation via selective logging, and only in Amazonia, thus potentially explaining why the Global EO estimate for gross emissions is lower than the NGHGI. The SEEG does provide an estimate of emissions by fire unrelated to deforestation in all biomes, however these are currently an additional dataset, and are not included in this analysis [32].

Comparing flux estimates of different datasets in South-East Asia

The net flux and associated gross emissions and removals for Indonesia are remarkably similar in the Global EO (0.55 GtCO2e yr^-1) and the NGHGI (0.57 GtCO2e yr^-1) dataset (Figure 9) [11]. Like the NGHGI, the Global EO study also includes emissions from peat fires and peat decomposition, key emission sources in Indonesia. The relative contribution of different sources to the gross emissions estimates is, however, different for the two datasets. Both the Global EO and the Indonesian NGHGI applied IPCC Tier 1-style methodology, giving some confidence to the observed similarity in the gross removals component. Further understanding the reasons for the similarity was difficult to attribute, given the lack of transparency and detail in the methodology and reporting of the NGHGI. For example, there needed to be more information on the removal factors applied to the natural forests (old-growth or secondary forests). Comparing the area of land type shows there are some discrepancies between the two datasets; for example, the area of natural forest cover in 2020 (old-growth and secondary forests) in the Global EO dataset is a third larger than the Indonesian NGHGI natural forest area in 2019 (Table 7).

For Malaysia, the estimated net fluxes from the NGHGI (-0.2 GtCO2e yr^-1) and Global EO (0.2 GtCO2e yr^-1) are not only different in the magnitude (size) but also the sign (source or sink), despite a similar area of ‘forest land remaining forest land’ identified by both datasets (Table 7). As with the NGHGI of Indonesia, the NGHGI of Malaysia lacked detail to determine the cause of the difference.

One potential reason for the difference between the gross removal estimates in Malaysia may be linked to the high removal factor that is being applied according to the NGHGI to all forest lands remaining forest lands. For example, a removal factor of 4.37 MgC ha^-1 yr^-1 is applied to inland forests, and no distinction between secondary and old-growth forests is made [41]. The intact forest removal factor is about twelve and three times higher than the IPCC default values for old-growth and old secondary Asian tropical rainforests, respectively [42]. IPCC default values are estimated at 0.35 MgC ha^-1 yr^-1 (95% confidence interval: 0.05 to 0.65 MgC ha^-1 yr^-1) for old-growth forests and 1.35 MgC ha^-1 yr^-1 (95% confidence interval: -0.8 to 3.5 MgC ha^-1 yr^-1) for old secondary forests [42], and are used by the Global EO-based dataset (Table 1). The value used by the NGHGI is more representative of young secondary forest regrowth rates in this region [26,42]. The estimated gross removals flux of the NGHGI is, therefore, implausible [6].

Table 7 | Area of forested and non-forest lands in Indonesia and Malaysia estimated by different flux datasets. The datasets are the Global Earth Observation (EO) dataset and the National Greenhouse Gas Inventories (NGHGI). The data for the Global EO are for the year 2020 [11], for the Indonesian NGHGI for 2019 [40], and for the Malaysian NGHGI for 2016 [41]. The years for the NGHGI are the most up to data values as submitted to the UNFCCC. Units are in 1000 ha. Mangroves have been excluded from all flux datasets. The Indonesian NGHGI data includes data on the so-called “Area Penggunaan Lain” – APL areas which are largely considered non-forested but include some forested lands. * For the purpose of this study all croplands are assumed to be tree plantations.

Land type	Global EO (1000 ha)	NGHGI (1000 ha)
Indonesia
Forested lands	Old-growth forest: 81,126.7 Old secondary forest (>20 years): 38,388.4 Young secondary forest: 1,970.3 “Forest land remaining forest land” (old-growth + old secondary) subtotal: 119,515.1 Natural forest subtotal: 121,485.5	Primary forest (including swamp): 45,267.3 Secondary forest (including swamp): 40,825.2 Natural forest subtotal: 86,092.5
Plantations	Forest plantation: 358.7 Tree plantation: 26,127.6 Subtotal: 26,486.3	Forest plantation: 5109.4 Tree plantation: 18,007.9 Subtotal: 23,117.3
Non-forested land	Other land: 40,890.0	Other land: 75,629.9
Malaysia
Forested lands	Old-growth forest: 12,4483.4 Old secondary forest (>20 years): 4,595.9 Young secondary forest: 361.9 “Forest land remaining forest land” (old-growth + old secondary) subtotal: 17,079.3 Natural forest subtotal: 17,441.2	Forest land remaining forest land: 17,661.7
Plantations	Forest plantation: 0.0 Tree plantation: 10,613.8	Cropland remaining cropland: 7,892.9 Land converted to cropland: 0.0
Non-forested land	Other land: 4,922.5	Settlement remaining settlement: 4,262.8 Forest land converted to settlement: 148.0 Settlement total: 4,410.8 Grassland remaining grassland: 1,000.0

The large difference between the gross emissions in Malaysia may be linked to the fact that only emissions associated with “forest land converted to settlement” are reported in the Malaysian NGHGI (Supplementary Table 4). The Global EO finds emissions associated with other transitions, including “forest land converted to cropland”, and commodity driven agriculture made up 90% of gross emissions between 2001-2020 [43]. The fact that the Malaysian NGHGI suggests no emissions from other land transitions, especially to commodity croplands, such as oil palm is therefore surprising. There may be assumptions made by the NGHGI such as not reporting gross fluxes in the harvest in tree crop areas, as these may be assumed to regrow later, such that the net flux could be assumed to be near zero.

We frame the following discussion around the five UNFCCC reporting principles included in the IPCC Good Practice Guidelines, which are: Comparability, Transparency, Consistency, Accuracy and Completeness.

Comparability: Aligning classifications of “managed” forest

Our analysis highlights the influence of the managed land definition on estimates of gross emissions and removals from forest related fluxes (Fig. 2). Brazil’s NGHGI considers a large area of conservation and indigenous lands to be under anthropogenic influence and, therefore, “managed” (Fig. 3b and c). These areas are typically not highly degraded on scales detectable by moderate resolution satellite data such as Landsat, and so were classified as old-growth (primary) forests by the dataset used by the Global EO method used in this analysis [11, 37]. Using a global mask of non-primary forest to represent managed land was, therefore, not the best approach to accurately comparing Global EO data with the NGHGI in Brazil [4]. A previous study reconciling country-scale NGHGIs and Bookkeeping models found similar results when using a global intact-forest mask in Brazil [9].

For Brazil, it is possible to directly apply the same mask of Managed Land as considered by the NGHGI to the Global EO, as the spatially explicit dataset of LULC in the NGHGI is made freely available. However, the spatial extent of what is considered managed land by other countries’ NGHGIs is not always clear and limits detailed comparisons, while introducing contortions on the extent of forest considered managed.

In our study, a key source for the differences between the estimates of the flux datasets was the extent to which different forest categories were included in relation to the IPCC categories (Table 2). Given the differences in scope, purpose, and capacity of the respective flux datasets, each included the various types of forest transitions in unique ways. High granularity of the Global EO data, in terms of its components and spatial data, gave it flexibility such that adjustments could be made to reflect NGHGIs approaches and definitions, thus making it a useful tool for MRV purposes. Interpretation and definitions of forest type vary between approaches and datasets [10]. In our analysis, the category “old secondary forest” in the Global EO flux dataset was most akin to “secondary forest land remaining secondary forest” in the NGHGI and was removed in adjustments (Fig. 6), as this category is not considered a flux source by Brazil’s NGHGI (Table 2). A future improvement to the Brazilian NGHGI may consider including gross emissions and removals in sub-categories of managed forests (FL -> FL) such as secondary forests and plantations, including respective removal factors, for a complete representation of forest-related fluxes.

As countries’ capacity to report increases, a further consideration may be for countries to report the gross emission and removals not only on managed lands but also on unmanaged lands [44]. The UNFCCC encourages countries to report GHG fluxes on unmanaged lands, however it is not a requirement and, few countries report on unmanaged lands. Reporting all emissions and removals would provide a complete picture of the fluxes occurring on the land [44] and be more directly comparable with other independent flux datasets, such as the Global EO dataset used in this study.

Transparency and consistency: methods, definitions, and data sources

There is an increasing call for the research community to contribute more to emerging MRV needs under the Paris Agreement [7, 45, 46]. For this engagement to be effective, transparent reporting and free availability of methods and datasets to view and access is required by the research community and the national inventory community [10]. As defined by the Katowice Rulebook, under the Paris Agreement and the ‘transparency framework’, countries will be required to provide a Biennial Transparency Report (BTR) from 2024 [47]. The BTR will include detailed information on the methods used, and a common reporting table to improve consistency and transparency among all NGHGIs [10, 20].

Transparency in the methodological approaches used in each of Brazil’s datasets made it possible to distinguish methodological and definitional difference such as forest categories. This transparent information enabled adjustments to be made to improve comparability and thus credibility of the estimates. Unfortunately, such a detailed analysis is not possible in other countries, such as Indonesia and Malaysia, where digitised versions of the LULC maps were unavailable for analysis. All the NGHGIs considered in our study clearly stated that tree crops such as palm oil and rubber are classed as perennial agricultural crops. Associated emissions and removals are, therefore, not forest-related (Table 1, Table 2). The Global EO included tree crops broadly within their definition of plantations (Table 1, Table 2) [23]. In this case, high transparency in the definitions used by each approach enabled appropriate adjustments to be made to the EO dataset for a suitable comparison.

Variations in definitions are not limited to our analysis and vary between global models and other internationally applied flux datasets such as the FAO [10], emphasising the need for transparency in definitions and delineating which IPCC categories are included. Future updates to the Global EO flux dataset [11] could include providing spatial disaggregation of the gross emissions and removals by the different forest types. The disaggregation would allow countries to make necessary adjustments to include only the processes and forest types included in their NGHGI if using this type of data for MRV.

Countries could consider clearly documenting the emissions and removal factors used in their submissions to the UNFCCC for greater transparency, this includes Indonesia explored in this study. Future improvements to the Global EO dataset could include providing annual estimates of emissions and removals from forests. This would improve the consistency and enable a more representative comparison, covering only the years included by the NGHGI.

Accuracy and completeness: Differences in forest types and removal factors – expanding definitions.

For completeness, there is a need to consider, as far as possible, all forest types, which might be sources or sinks. Due to the spatial and temporal heterogeneity of forest types, and associated fluxes, EO data can help with capturing this heterogeneity and may improve accuracy.

Our analysis has shown spatial differences in the forest cover between the Brazilian NGHGI and Global EO approach, especially outside the Amazonia biome, with divergences in the savannah and dry forest biomes (Fig. 5). These discrepancies explain some of the differences between the fluxes (Fig. 2) and may be an important consideration for carrying out similar analysis elsewhere. In increasingly forest fragmented regions such as in the Cerrado, it is important to accurately distinguish forested areas from other naturally existing vegetation, such as other woody vegetation, shrublands and natural grasslands, to track transitions for a complete representation. Such distinctions can be aided by visual interpretation and knowledge of the areas by experts consulted during the NGHGIs, thus highlighting the potential for NGHGIs to refine methods used by global EO datasets.

Across the biomes, fire and logging disturbances can cause forest degradation that often occurs at smaller-scales, and thus goes undetected by moderate resolution (> 30m) remote sensing products [48]. The relative impact of forest degradation on gross emissions, and subsequent recovery, is currently poorly constrained in the Global EO dataset, NGHGIs, and global models [2, 11, 27, 33]. As a result, motivation in NDC pledges to tackle forest degradation has trailed behind limiting deforestation [49]. Individual studies focusing on forest degradation have quantified and shown that degradation is a sizeable contribution to forest carbon emissions [50–52], but fewer studies have focused on the recovery of such degraded forests [53–55]. There is a need to expand the current definitions of forest types to accurately represent degraded forests, to avoid underestimating key sources of carbon emissions and miss potential opportunities to protect and increase the carbon sink in recovering degraded forests. However, this raises the question of how far NGHGIs should go to represent the national forest carbon cycle completely, which is not their intended purpose. Instead, independent EO data may be a useful tool to provide spatial, highly granular information that is not considered by NGHGIs and fill information gaps.

Our analysis highlighted differences, in the removal factors used by different datasets, increasing the uncertainty on what the most accurate removal factor to use for each forest type is. Per unit area, the difference was particularly evident in young secondary forests in Brazil (“other land converted to secondary forests”) (Table 6). However, the total removals associated with these young secondary forests are modest (Table 4), so the overall impact on the total gross removals is small. The difference between the old-growth forest removal factors per unit area was less, around ~ 25%, but given the large area extent of old-growth forest, the difference in the total removals was considerable (Table 5). Accurate and representative removal factors for all forest types [42, 45, 56] are essential for credibility, as many countries and organisations aim to reduce their net carbon emissions through forest conservation and restoration. Inaccurate removal factors risk overestimating sinks and could lead to reduced ambition in reducing fossil fuel emissions. In the Malaysian NGHGI, we found the removal factor used for inland “old-growth” forests to be more representative of young secondary forests, suggesting there may be inaccuracies that warrant further analysis. Remote sensing studies that analyse changes in AGC in space [56] and/or time [51, 52] may provide useful information to improve the accuracy of removal factors thus providing Tier-2 or Tier-3 information which is crucial for countries in which LULUCF emissions are high, such as the ones analysed here.

There remain differences between the flux datasets regarding the forest classification assigned to a given pixel in space and time, which has implications for AGC and the emission/removal factor applied [57]. The impact of differences in forest type was most noticeable in the classification of “old secondary forest” in the Global EO dataset, of which many regions were classified as managed old-growth forests by the Brazilian NGHGI and SEEG (Fig. 7). Future work identifying forests according to the type [25] and intensity [58] of disturbance may be a useful starting point for more disaggregated classification and would expand the current definitions of forest that are being considered by all flux datasets in this study and as well as others such as the FAO [59]. Expanding the representation of forest demographics beyond simply < 20 and > 20 years would also enable more accurate removal factors to be applied. Where feasible, there may be scope to align the definitions of forest type used by different flux datasets. This would enable a more credible comparison and accurate assessment [10, 45, 57, 60]. Nevertheless, there are likely to be difficulties in negotiating definitions among different approaches.

Independent data, such as the Global EO method analysed here, have the potential to provide key sources of independent information for the MRV of NGHGIs [10]. Conversely, information from NGHGIs may be useful to refine the methods and data used in the Global EO approach, for example by using region-specific removal factors. Using Brazil as a primary case study, we could reconcile the difference between GHG fluxes from a Global EO dataset and the NGHGI. With no adjustment to the Global EO dataset, considering all forest fluxes in the country boundary, the difference between the Global EO flux (-0.2 GtCO₂yr^− 1) and the NGHGI flux (0.8 GtCO₂yr^− 1) from 2001 to 2020 was 1.0 GtCO₂yr^− 1 and estimates were opposite in sign. Adjusting the Global EO flux by applying the Brazilian NGHGI definition of managed forest area and only considering the assumptions in their inventory (in terms of land transition and flux categories) resulted in the adjusted Global EO net flux to be 0.6 GtCO₂yr^− 1, now just 0.2 GtCO₂yr^− 1 lower compared to the NGHGI. The analysis was made possible due to the availability of the data and transparency in methodologies from all flux datasets.

In the other two case study countries, Indonesia and Malaysia, the analysis was limited to the tabular data provided by the respective NGHGIs as no spatially explicit datasets on forest cover types were publicly available. Limited transparency in the emission and removal factors used also restricted detailed analysis. However, the comparison indicated where there might be uncertainties, inaccuracies, and areas for improvement in NGHGIs and Global EO datasets.

The adjustments as outlined in this study can also be used to aid comparisons with and benchmarking of other approaches, such as global models. Below we outline some key lessons learned in this study:

Full transparency is crucial for effective comparison between NGHGI and EO. This includes ensuring transparency in the methodology and open access to the data used by all approaches.
Effective comparisons require more clarity in the use of the Managed Land Proxy and IPCC forest categories. Making the appropriate adjustments to account for differences between datasets becomes increasingly important when using independent flux datasets for MRV at (sub)-country scales. For completeness, NGHGIs could voluntarily consider providing land area and associated fluxes on unmanaged land.
Further disaggregation in the forest types and associated fluxes in datasets will facilitate greater comparability and completeness. Independent datasets could disaggregate the fluxes by forest types to allow countries to make adjustments, only comparing the processes and forest types in their NGHGI. All flux datasets could expand the forest cover types and associated emission and removal factors, specifically to consider disturbance type and intensity and associated recovery.

As more information and datasets become available, including EO, countries’ requirements and capacity to report is expected to increase in transparency, comparability, consistency, completeness, and accuracy. Additionally, region specific information compiled for use in NGHGIs can also be used to improve local applicability of Global EO assessments. Given that numerous pledges and commitments have been made by business organisations, cities, nations, and international agreements to protect and restore forests, the need to monitor, report and especially verify estimates becomes increasingly important. If this is not achieved in accurate and credible ways, we risk mis-representing the carbon fluxes of the world’s forests, one of our most important allies on land, to tackle the climate emergency.

Ethics approval and consent to participate: Not applicable

Consent for Publication: All authors consent to publication in CBM

Availability of data and materials: All data used in this research is readily available from the sources that are fully referenced in the text and bibliography. The code used to extract the data from its sources will be made available upon acceptance of the manuscript.

Competing interests: the authors declare no competing interests.

Funding: V.H.A.H. was supported by a NERC GW4+ Doctoral Training Partnership studentship from the Natural Environment Research Council (NE/L002434/1). V.H.A.H., T.R. and S.S. were supported by the RECCAP2 project which is part of the ESA Climate Change Initiative (contract no. 4000123002/18/I-NB) and the H2020 European Institute of Innovation and Technology (4C; grant no. 821003).

Author contributions: V.H.A.H and J.H designed the concept and methodological process with input from G.G., M.H., N.H., D.G., S.S, T.H., L.E.O.A.. V.H.A.H carried out the data analysis and wrote the manuscript draft. D.G provided access to the datasets and both D.G and N.H. provided expert advice on the Global Earth Observation dataset used. Coding help was given by D.G. and T.M.R. Expert information on the use of the SEEG and Brazil’s NGHGI datasets were provided by J.Z.S.., B.Z. and R.C.. All authors discussed results and provided input during the preparation of the manuscript.

Acknowledgments: We would like to thank the authors of the datasets used in this study for providing open access to their datasets where possible. The views expressed are purely those of the writers and may not under any circumstances be regarded as stating an official position of the European Commission.

Authors’ information: Not applicable

Hansis E, Steven D, Pongratz J. Relevance of methodological choices for accounting of land use change carbon fluxes. Global Biogeochem Cycles [Internet]. 2015;29(8):1230–46. Available from: https://doi.org/10.1002/2014GB004997
Friedlingstein P, Jones MW, O’Sullivan M, Andrew RM, Bakker DCE, Hauck J, et al. Global Carbon Budget 2021. Earth Syst Sci Data [Internet]. 2022 Apr 26;14(4):1917–2005. Available from: https://essd.copernicus.org/articles/14/1917/2022/
Grassi G, House J, Kurz WA, Cescatti A, Houghton RA, Peters GP, et al. Reconciling global-model estimates and country reporting of anthropogenic forest CO2 sinks. Nat Clim Chang. 2018;8(10):914–20.
Grassi G, Schwingshackl C, Gasser T, Houghton RA, Sitch S, Josep G. Mapping land-use fluxes for 2001-2020 from global models to national inventories [Preprint]. Earth Syst Sci Data. 2022;(August):1–42.
Friedlingstein P, O’Sullivan M, Jones MW, Andrew RM, Gregor L, Hauck J, et al. Global Carbon Budget 2022. Earth Syst Sci Data [Internet]. 2022 Nov 11;14(11):4811–900. Available from: https://essd.copernicus.org/articles/14/4811/2022/
Grassi G, Conchedda G, Federici S, Abad Viñas R, Korosuo A, Melo J, et al. Carbon fluxes from land 2000-2020: bringing clarity on countries’ reporting. Earth Syst Sci Data Discuss [Internet]. 2022;2022(March):1–49. Available from: https://essd.copernicus.org/preprints/essd-2022-104/%0Ahttps://essd.copernicus.org/preprints/essd-2022-104/essd-2022-104.pdf
Grassi G, Stehfest E, Rogelj J, van Vuuren D, Cescatti A, House J, et al. Critical adjustment of land mitigation pathways for assessing countries’ climate progress. Nat Clim Chang [Internet]. 2021 May 26;11(5):425–34. Available from: http://www.nature.com/articles/s41558-021-01033-6
UNFCCC. Synthesis report for the technical assessment component of the first global stocktake. Vol. 7. Bonn; 2022.
Schwingshackl C, Obermeier WA, Bultan S, Kurz WA, Sitch S, Pongratz J. Differences in land-based mitigation estimates reconciled by separating natural and land-use CO2 fluxes at the country level Graphical. One Earth [Internet]. 2022;5(12):1367–76. Available from: https://doi.org/10.1016/j.oneear.2022.11.009
Perugini L, Pellis G, Grassi G, Ciais P, Dolman H, House JI, et al. Emerging reporting and verification needs under the Paris Agreement: How can the research community effectively contribute? Environ Sci Policy [Internet]. 2021;122(April):116–26. Available from: https://doi.org/10.1016/j.envsci.2021.04.012
Harris NL, Gibbs DA, Baccini A, Birdsey RA, de Bruin S, Farina M, et al. Global maps of twenty-first century forest carbon fluxes. Nat Clim Chang [Internet]. 2021 Mar 21;11(3):234–40. Available from: http://dx.doi.org/10.1038/s41558-020-00976-6
IPCC. Revisiting the use of managed land as a proxy for estimating national anthropogenic emissions and removals. Eggleston H., Srivastava N, Tanabe K, Bassansuren J, editors. Metting Report. INPE, São José dos Campos: IGES, Japan 2010; 2010. 56 p.
IPCC. Chapter 4 Forest Land. In: Eggleston H.S, Buendia L, Miwa K, Ngara T, Tanabe K, editors. IPCC Guidelines for National Greenhouse Gas Inventories [Internet]. 2006. p. 1–29. Available from: http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/4_Volume4/V4_04_Ch4_Forest_Land.pdf
UNFCCC/COP5. Review of the Implementation of commitments and other provisions of the Convention. Bonn; 2000.
Federici S, Grassi G, Nancy Harris, Lee D, Neeff T, Penman J, et al. GHG fluxes from forests: An assessment of national GHG estimates and independent research in the context of the Paris Agreement. 2017.
Hubau W, Lewis SL, Phillips OL, Affum-Baffoe K, Beeckman H, Cuní-Sanchez A, et al. Asynchronous carbon sink saturation in African and Amazonian tropical forests. Nature [Internet]. 2020 Mar 4;579(7797):80–7. Available from: http://www.nature.com/articles/s41586-020-2035-0
Grassi G, House J, Dentener F, Federici S, Den Elzen M, Penman J. The key role of forests in meeting climate targets requires science for credible mitigation. Nat Clim Chang. 2017;7(3):220–6.
Deng Z, Ciais P, Tzompa-Sosa ZA, Saunois M, Qiu C, Tan C, et al. Comparing national greenhouse gas budgets reported in UNFCCC inventories against atmospheric inversions. Earth Syst Sci Data. 2022;14(4):1639–75.
Mooney C, Eilperin J, Butler D, Muyskens J, Narayanswamy A, Ahmed N. Countries’ climate pledges built on flawed data, Post investigation finds [Internet]. The Washington Post. 2021 [cited 2022 Nov 3]. Available from: https://www.washingtonpost.com/climate-environment/interactive/2021/greenhouse-gas-emissions-pledges-data/
IPCC. 2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories Volume 4: Agriculture, Forestry, and Other Land Use [Internet]. Vol. 4. 2019. Available from: http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/4_Volume4/V4_04_Ch4_Forest_Land.pdf
Global Forest Watch. Aboveground Live Woody Biomass Density [Internet]. 2022 [cited 2022 Aug 5]. Available from: https://data.globalforestwatch.org/datasets/gfw::aboveground-live-woody-biomass-density/about
Hansen MC, Potapov P V, Moore R, Hancher MC, Turubanova S a, Tyukavina A, et al. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science (80- ). 2013;850(November):2011–4.
Harris N, Goldman ED, Gibbes S. Spatial Database of Planted Trees (SDPT Version 1.0) | World Resources Institute. Tech Note [Internet]. 2019;(February):1–36. Available from: https://www.wri.org/publication/planted-trees
Giglio L, Boschetti L, Roy DP, Humber ML, Justice CO. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sens Environ [Internet]. 2018;217(March):72–85. Available from: https://doi.org/10.1016/j.rse.2018.08.005
Curtis PG, Slay CM, Harris NL, Tyukavina A, Hansen MC. Classifying drivers of global forest loss. Science (80- ). 2018;361(6407):1108–11.
Cook-Patton SC, Leavitt SM, Gibbs D, Harris NL, Lister K, Anderson-Teixeira KJ, et al. Mapping carbon accumulation potential from global natural forest regrowth. Nature [Internet]. 2020 Sep 24;585(7826):545–50. Available from: http://www.nature.com/articles/s41586-020-2686-x
MCTI. Fourth National Communication of Brazil to the United Nations Framework Convention on Climate Change. 2020.
MCTI. Third National Communication of Brazil to the United Nations Framework Convention on Climate Change - Volume III. Ministry of Science T and I, editor. Brasília; 2016.
FAO. The State of the World’s Forests 2022. Forest pathways for green recovery and building inclusive, resilient and sustainable economies. [Internet]. Fao. 2022. 166 p. Available from: https://doi.org/10.4060/cb9360en
de Azevedo TR, Costa Junior C, Brandão Junior A, Cremer M dos S, Piatto M, Tsai DS, et al. SEEG initiative estimates of Brazilian greenhouse gas emissions from 1970 to 2015. Sci Data [Internet]. 2018 Dec 29;5(1):180045. Available from: http://www.nature.com/articles/sdata201845
Souza CM, Shimbo JZ, Rosa MR, Parente LL, Alencar AA, Rudorff BFT, et al. Reconstructing three decades of land use and land cover changes in brazilian biomes with landsat archive and earth engine. Remote Sens. 2020;12(17).
Shimbo JZ, Souza E, Silva C, Zimbres B, Alencar A, Brandão-Jr. A, et al. Nota Metodológica SEEG 9 Setor Mudança de Uso da Terra e Florestas [Internet]. 2021. Available from: https://seeg-br.s3.amazonaws.com/2019-v7.0/notas-metodologicas/SEEG8_NOTA_METODOLOGICA_MUT_2020.11.03.pdf
SEEG. SEEG v9 Mudança de Uso da Terra e Florestas [Internet]. 2022. Available from: https://plataforma.seeg.eco.br/sectors/mudanca-de-uso-da-terra-e-floresta
Tubiello FN, Conchedda G, Wanner N, Federici S, Rossi S, Grassi G. Carbon emissions and removals from forests: new estimates, 1990–2020. Earth Syst Sci Data [Internet]. 2021 Apr 26;13(4):1681–91. Available from: https://essd.copernicus.org/articles/13/1681/2021/
FAO. Emissions from agriculture and forest land. Global, regional and country trends 1990–2019 [dataset] [Internet]. Rome; 2022. Available from: https://www.fao.org/faostat/en/#data
Nabuurs GJ, Mrabet R, Abu Hatab A, Bustamante M, Clark H, Havlik P, et al. Agriculture, Forestry and Other Land Uses (AFOLU). In: Shukla P., Skea J, Slade R, Al Khourdajie A, van Diemen R, McCollum D, et al., editors. IPCC, 2022: Climate Change 2022: Mitigation of Climate Change Contribution of Working Group III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and New York, NY, USA; 2022.
Turubanova S, Potapov P V., Tyukavina A, Hansen MC. Ongoing primary forest loss in Brazil, Democratic Republic of the Congo, and Indonesia. Environ Res Lett. 2018;13(7).
SIRENE (Sistema de REgistro Nacional de Emissões. Dados de Uso da Terra, Mudança de Uso da Terra e Florestas [dataset] [Internet]. 2021. Available from: http://www.ccst.inpe.br/cn/
Mapbiomas Brasil. MapBiomas Project - Collection 6 (1985-2020) of Annual Series of Land use and Land Cover maps of Brazil [Internet]. 2021 [cited 2022 May 16]. Available from: https://mapbiomas.org/colecoes-mapbiomas-1?cama_set_language=pt-BR
Ministry of Natural Resources and Environment. Indonesia third biennial update report. Vol. 1. Jakarta; 2021.
Ministry of Environment and Water. Malaysia Third Biennial Update Report To The UNFCCC [Internet]. 2020. Available from: https://unfccc.int/sites/default/files/resource/MALAYSIA_BUR3-UNFCCC_Submission.pdf
Requena Suarez D, Rozendaal DMA, De Sy V, Phillips OL, Alvarez-Dávila E, Anderson-Teixeira K, et al. Estimating aboveground net biomass change for tropical and subtropical forests: Refinement of IPCC default rates using forest plot data. Glob Chang Biol. 2019;25(11):3609–24.
Global Forest Watch. Forest-related greenhouse gas emissions by driver [Internet]. 2022 [cited 2022 Aug 4]. Available from: www.globalforestwatch.org
Houghton RA. Terrestrial fluxes of carbon in GCP carbon budgets. Glob Chang Biol. 2020;26(5):3006–14.
Romijn E, De Sy V, Herold M, Böttcher H, Roman-Cuesta RM, Fritz S, et al. Independent data for transparent monitoring of greenhouse gas emissions from the land use sector – What do stakeholders think and need? Environ Sci Policy. 2018;85(March):101–12.
Bustamante MMC, Silva JSO, Cantinho RZ, Shimbo JZ, Oliveira PVC, Santos MMO, et al. Engagement of scientific community and transparency in C accounting: The Brazilian case for anthropogenic greenhouse gas emissions from land use, land-use change and forestry. Environ Res Lett. 2018 May 1;13(5).
UNFCCC. Dec. 19/CMA.1 Matters relating to Article 14 of the Paris Agreement and paragraphs 99–101 of decision 1/CP.21. Vol. 03302, Report of the Conference of the Parties serving as the meeting of the Parties to the Paris Agreement on the third part of its first session, held in Katowice from 2 to 15 December 2018. 2018.
Gao Y, Skutsch M, Paneque-Gálvez J, Ghilardi A. Remote sensing of forest degradation: a review. Environ Res Lett. 2020;15(10):2000–10.
Silva Junior CHL, Carvalho NS, Pessôa ACM, Reis JBC, Pontes-Lopes A, Doblas J, et al. Amazonian forest degradation must be incorporated into the COP26 agenda. Nat Geosci [Internet]. 2021;14(9):634–5. Available from: http://dx.doi.org/10.1038/s41561-021-00823-z
Maxwell SL, Evans T, Watson JEM, Morel A, Grantham H, Duncan A, et al. Degradation and forgone removals increase the carbon impact of intact forest loss by 626%. Sci Adv. 2019;5(10).
Qin Y, Xiao X, Wigneron J pierre, Ciais P, Brandt M, Fan L, et al. Carbon loss from forest degradation exceeds that from deforestation in the Brazilian Amazon. Nat Clim Chang [Internet]. 2021 May 29;11(5):442–8. Available from: http://dx.doi.org/10.1038/s41558-021-01026-5
Baccini A, Walker W, Carvalho L, Farina M, Sulla-Menashe D, Houghton RA. Tropical forests are a net carbon source based on aboveground measurements of gain and loss. Science (80- ). 2017;358(6360):230–4.
Philipson CD, Cutler MEJ, Brodric PG, Asne GP, Boy DS, Costa PM, et al. Active restoration accelerates the carbon recovery of human-modified tropical forests. Science (80- ). 2020;369(6505):838–41.
Rappaport DI, Morton DC, Longo M, Keller M, Dubayah R, Dos-Santos MN. Quantifying long-term changes in carbon stocks and forest structure from Amazon forest degradation. Environ Res Lett. 2018;13(6).
Requena Suarez D, Rozendaal DMA, De Sy V, Gibbs DA, Harris NL, Sexton JO, et al. Variation in aboveground biomass in forests and woodlands in Tanzania along gradients in environmental conditions and human use. Environ Res Lett. 2021;16(4).
Heinrich VHA, Dalagnol R, Cassol HLG, Rosan TM, de Almeida CT, Silva Junior CHL, et al. Large carbon sink potential of secondary forests in the Brazilian Amazon to mitigate climate change. Nat Commun [Internet]. 2021 Dec 19;12(1):1785. Available from: http://dx.doi.org/10.1038/s41467-021-22050-1
Federici S, Penman J, Wolosin M. GHG fluxes from forests : An assessment of national reporting and independent science in the context of the Paris Agreement. 2017;1–42. Available from: http://www.climateandlandusealliance.org/reports/ghg-fluxes-from-forests/?utm_source=CLUA+Quarterly+Email+Newsletter&utm_campaign=41584a255a-GHG_fluxes_paper_06_14_2016&utm_medium=email&utm_term=0_b2faff125f-41584a255a-38867569
Vancutsem C, Achard F, Pekel JF, Vieilledent G, Carboni S, Simonetti D, et al. Long-term (1990–2019) monitoring of forest cover changes in the humid tropics. Sci Adv [Internet]. 2021 Mar 5;7(10):eabe1603. Available from: https://advances.sciencemag.org/lookup/doi/10.1126/sciadv.abe1603
FAO. Global Forest Resources Assessment. Rome; 2020.
Pongratz J, Schwingshackl C, Bultan S, Obermeier W, Havermann F, Guo S. Land Use Effects on Climate: Current State, Recent Progress, and Emerging Topics. Curr Clim Chang Reports [Internet]. 2021;7(4):99–120. Available from: https://doi.org/10.1007/s40641-021-00178-y
Houghton RA, Nassikas AA. Global and regional fluxes of carbon from land use and land cover change 1850–2015. Global Biogeochem Cycles. 2017;31(3):456–72.
Gasser T, Crepin L, Quilcaille Y, Houghton RA, Ciais P, Obersteiner M. Historical CO2 emissions from land use and land cover change and their uncertainty. Biogeosciences. 2020;17(15):4075–101.
Prosperi P, Bloise M, Tubiello FN, Conchedda G, Rossi S, Boschetti L, et al. New estimates of greenhouse gas emissions from biomass burning and peat fires using MODIS Collection 6 burned areas. Clim Change. 2020;161(3):415–32.
IPCC. Good practice guidance for land use, land-use change and forestry. Penman J, Gytarsky M, Hiraishi T, Krug T, Kruger D, Pipatti R, et al., editors. Kanagawa, Japan: Institute for Global Environmental Strategies for Intergovernmental Panel on Climate Change; 2003.

No competing interests reported.

heinrichetalsupplementaryInfoCBM.pdf

Mind the Gap: Reconciling tropical forest carbon flux estimates from Global Earth Observation and National Reporting

Status:

Journal Publication

Version 1

Abstract

Background

Results

Conclusions

Figures

Background

Forest carbon flux estimates in the context of the Paris Agreement

Mind the gap: Reconciling land-use flux estimates

Methodology

Flux datasets

(i) Global Earth Observation

(ii) Brazil’s National Greenhouse Gas Inventory

(iii) Other datasets: SEEG and FAOSTAT

Estimates of GHG fluxes from different datasets

Adjustments to the Global EO forest-related flux to make it comparable to the other approaches in Brazil

Evaluation of the removal factors and forest types

Assessing comparability of fluxes with high uncertainty

Application of the analysis in other case study regions

Results

Making the extent of managed land more comparable for flux estimates in Brazil (Adjustments 1, 2a, 2b)

Relative contribution of different forest types to gross removals in Brazil

Reconciling the difference in gross removals between flux datasets in Brazil (Adjustment 2c)

Explaining the remaining discrepancy: removal factors and forest type

Differences in the key carbon pools with implications for gross emissions in Brazil

Comparing flux estimates of different datasets in South-East Asia

Discussion

Comparability: Aligning classifications of “managed” forest

Transparency and consistency: methods, definitions, and data sources

Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1