Editing IPCC:AR6/WGIII/Chapter-7 (section)

==== 7.2.2.2 Why Do Various Methods Deliver Difference in Results? ====

<div id="h3-2-siblings" class="h3-siblings"></div>

The processes responsible for fluxes from land have been divided into three categories ( [[#IPCC--2006|IPCC 2006]] , 2010): (i) the ''direct human-induced effects'' due to changing land cover and land management; (ii) the ''indirect human-induced effects'' due to anthropogenic environmental change, such as climate change, CO 2 fertilisation, nitrogen deposition, and so on; and (iii) ''natural effects,'' including climate variability and a background natural disturbance regime (e.g., wildfires, windthrows, diseases or insect outbreaks).

Global models estimate the anthropogenic land CO 2 flux considering only the impact of direct effects, and only those areas that were subject to intense and direct management such as clear-cut harvest. It is important to note, that DGVMs also estimate the non-anthropogenic land CO 2 flux (Land Sink) that results from indirect and natural effects (Table 7.1). In contrast, estimates of the anthropogenic land CO 2 flux in NGHGIs (LULUCF) include the impact of direct effects and, in most cases, of indirect effects on a much greater area considered ‘managed’ than global models ( [[#Grassi--2021|Grassi et al. 2021]] ).

The approach used by countries follows the IPCC methodological guidance for NGHGIs ( [[#IPCC--2006|IPCC 2006]] , 2019). Since separating direct, indirect and natural effects on the land CO 2 sink is impossible with direct observation such as national forest inventories ( [[#IPCC--2010|IPCC 2010]] ), upon which most NGHGIs are based, the IPCC adopted the ‘managed land’ concept as a pragmatic proxy to facilitate NGHGI reporting. Anthropogenic land GHG fluxes (direct and indirect effects) are defined as all those occurring on managed land, that is, where human interventions and practices have been applied to perform production, ecological or social functions ( [[#IPCC--2006|IPCC 2006]] , 2019). GHG fluxes from unmanaged land are not reported in NGHGIs because they are assumed to be non-anthropogenic. Countries report NGHGI data with a range of methodologies, resolution and completeness, dependent on capacity and available data, consistent with IPCC guidelines ( [[#IPCC--2006|IPCC 2006]] , 2019) and subject to an international review or assessment processes.

The FAOSTAT approach is conceptually similar to NGHGIs. FAOSTAT data on forests are based on country reports to FAO-FRA 2020 ( [[#FAO--2020a|FAO 2020a]] ), and include changes in biomass carbon stock in ‘forest land’ and ‘net forest conversions’ in five-year intervals. ‘Forest land’ may include unmanaged natural forest, leading to possible overall overestimation of anthropogenic fluxes for both sources and sinks, though emissions from deforestation are likely underestimated ( [[#Tubiello--2020|Tubiello et al. 2020]] ). FAOSTAT also estimate emissions from forest fires and other land uses (organic soils), following IPCC methods ( [[#Prosperi--2020|Prosperi et al. 2020]] ). The FAO-FRA 2020 ( [[#FAO--2020b|FAO 2020b]] ) update leads to estimates of larger sinks in Russia since 1991, and in China and the USA from 2011, and larger deforestation emissions in Brazil and smaller in Indonesia than FRA 2015 ( [[#FAO--2015|FAO 2015]] ; [[#Tubiello--2020|Tubiello et al. 2020]] ).

The bookkeeping models by [[#Houghton--2017|Houghton and Nassikas (2017)]] , [[#Hansis--2015|Hansis et al. (2015)]] , and [[#Gasser--2020|Gasser et al. (2020)]] and the DGVMs used in the Global Carbon Budget ( [[#Friedlingstein--2020|Friedlingstein et al. 2020]] ) use either the LUH2 dataset ( [[#Hurtt--2020|Hurtt et al. 2020]] ), HYDE ( [[#Goldewijk--2017|Goldewijk et al. 2017]] ), FRA 2015 ( [[#FAO--2015|FAO 2015]] ) or a combination. The LUH2 dataset includes a new wood harvest reconstruction, new representation of shifting cultivation, crop rotations, and management information including irrigation and fertiliser application. The area of forest subject to harvest in LUH2 is much less than the area of forest considered ‘managed’ in the NGHGIs ( [[#Grassi--2018|Grassi et al. 2018]] ). The model datasets do not yet include the FAO FRA 2020 update ( [[#FAO--2020a|FAO 2020a]] ). The DGVMs consider CO 2 fertilisation effects on forest growth that are sometimes confirmed from the ground-based forest inventory networks ( [[#Nabuurs--2013|Nabuurs et al. 2013]] ) and sometimes not at all ( [[#van%20der%20Sleen--2015|van der Sleen et al. 2015]] ).

Further, the DGVMs and bookkeeping models do not include a wide range of practices which are implicitly covered by the inventories; for example: forest dynamics ( [[#Pugh--2019|Pugh et al. 2019]] ; [[#Le%20Noë--2020|Le Noë et al. 2020]] ), forest management including wood harvest (Nabuurs, et al. 2013; [[#Arneth--2017|Arneth et al. 2017]] ), agricultural and grassland practices ( [[#Pugh--2015|Pugh et al. 2015]] ; [[#Sanderman--2017|Sanderman et al. 2017]] ; [[#Pongratz--2018|Pongratz et al. 2018]] ); or, for example, fire management ( [[#Andela--2017|Andela et al. 2017]] ; [[#Arora--2018|Arora and Melton 2018]] ).

Increasingly, higher emissions estimates are expected from DGVMs compared to bookkeeping models, because DGVMs include a loss of additional sink capacity of 3.3 ± 1.1 GtCO 2 yr –1 on average over 2009–2018, which is increasing with larger climate and CO 2 impacts ( [[#Friedlingstein--2020|Friedlingstein et al. 2020]] ). This arises because the DGVM methodological setup requires a reference simulation including climate and environmental changes but without any land-use change such as deforestation, so DGVMs implicitly include the sink capacity forests would have developed in response to environmental changes on areas that in reality have been cleared ( [[#Gitz--2003|Gitz and Ciais 2003]] ; [[#Pongratz--2014|Pongratz et al. 2014]] ) (IPCC AR6 WGI Chapter 5).

Carbon emissions from peat burning have been estimated based on the Global Fire Emission Database (GFED4s; [[#van%20der%20Werf--2017|van der Werf et al. 2017]] ). These were included in the bookkeeping model estimates and added 2.0 GtC over 1960–2019 (e.g., causing the peak in South-East Asia in 1998) (Figure 7.5). Within the Global Carbon Budget ( [[#Friedlingstein--2020|Friedlingstein et al. 2020]] ), peat drainage from agriculture accounted for an additional 8.6 GtC from 1960–2019 according to FAOSTAT ( [[#Conchedda--2020|Conchedda and Tubiello, 2020]] ) used by two of the bookkeeping models ( [[#Hansis--2015|Hansis et al. 2015]] ; [[#Gasser--2020|Gasser et al. 2020]] ).

Remote-sensing products provide valuable spatial and temporal land-use and biomass data globally (including in remote areas), at potentially high spatial and temporal resolutions, that can be used to calculate CO 2 fluxes, but have mostly been applied only to forests at the global or even regional scale. While such data can strongly support monitoring reporting and verification, estimates of forest carbon fluxes directly from Earth Observation (EO) data vary considerably in both their magnitude and sign (i.e., whether forests are a net source or sink of carbon). For the period 2005–2017, net tropical forest carbon fluxes were estimated as –0.4 GtCO 2 yr –1 ( [[#Fan--2019|Fan et al. 2019]] ); 0.58 GtCO 2 yr –1 ( [[#Grace--2014|Grace et al. 2014]] ); 1.6 GtCO 2 yr –1 ( [[#Baccini--2017|Baccini et al. 2017]] ) and 2.87 GtCO 2 yr –1 ( [[#Achard--2014|Achard et al. 2014]] ). Differences can in part be explained by spatial resolution of the datasets, the definition of ‘forest’ and the inclusion of processes and methods used to determine degradation and growth in intact and secondary forests, or the changes in algorithm over time ( [[#Palahí--2021|Palahí et al. 2021]] ). A recent global study integrated ground observations and remote sensing data to map forest-related GHG emissions and removals at a high spatial resolution (30 m spatial scale), although it only provides an average estimate of annual carbon loss over 2001–2019 ( [[#Harris--2021|Harris et al. 2021]] ). The estimated net global forest carbon sink globally was –7.66 GtCO 2 yr −1 , being –1.7 GtCO 2 yr −1 in the tropics only.

Remote sensing products can help to attribute changes to anthropogenic activity or natural inter-annual climate variability ( [[#Fan--2019|Fan et al. 2019]] ; [[#Wigneron--2020|Wigneron et al. 2020]] ). Products with higher spatial resolution make it easier to determine forest and carbon dynamics in relatively small-sized managed forests (e.g., Y. [[#Wang--2020|Wang et al. 2020]] ; [[#Heinrich--2021|Heinrich et al. 2021]] ; [[#Reiche--2021|Reiche et al. 2021]] ). For example, secondary forest regrowth in the Brazilian Amazon offset 9 to 14% of gross emissions due to deforestation 1 ( [[#Aragão--2018|Aragão et al. 2018]] ; [[#Silva%20Junior--2021|Silva Junior et al. 2021]] ). Yet disturbances such as fire and repeated deforestation cycles due to shifting cultivation over the period 1985 to 2017, were found to reduce the regrowth rates of secondary forests by 8 to 55% depending on the climate region of regrowth ( [[#Heinrich--2021|Heinrich et al. 2021]] ).

<div id="7.2.2.3" class="h3-container"></div>

<span id="implications-of-differences-in-afolu-co-2-fluxes-between-global-models-and-national-greenhouse-gas-inventories-nghgis-and-reconciliation"></span>