Comparison of uncertainties in land-use change fluxes from bookkeeping model parameterisation

However, HI850 does end with about 12 % more primary land in 2014 than the LUH2 dataset. In all cases, the amount of primary land is larger in BLUE than in the original LUH2 dataset, at the cost of other land-cover types. Overall, this means that the total amount of net LULCC flux will be underestimated in BLUE, the most in the HI850 experiment. Both models rely on observation-based estimates for their parameterisations and forcing datasets,and the choices on spatial and plant functional type representation, starting year and other aspects are well justified cash flow in both models. However, these multiple differences add to uncertainty in FLUC estimates and make it difficult to attribute differences in FLUC and their trends to specific aspects of the FLUC calculation. Figure A1Global areas of the four BLUE land-cover types (primary land, secondary land, crop and pasture) based on the aggregated LUH2 input data (a, b) and their temporal net change (c, d).

Accounting Methods: Definition, Types, and Example

• In the Hurtt et al. (2011) sensitivity study based on the LUH1 dataset (Chini et al., 2014), the authors analysed over 1600 simulations with respect to model “factors” like the simulation start date, the choice of historical and future agricultural land-use and wood harvest scenarios, and inclusion of shifting cultivation.
• Europe shows 7 % higher cumulative FLUC for SBL-Net than SBL, likely because of the importance of subpixel post-abandonment recovery and re-/afforestation dynamics in Europe (Bayer et al., 2017; Fuchs et al., 2015).
• The lower FLUC from clearing to agriculture for SHNFull in most grid cells is linked with the lower vegetation and soil C densities for most forest PFTs (Table A2).
• Many of the DGVMs also utilize the LUH2 dataset to prescribethe gridded historical land-use and land-use changes used by those models.The availability of LUH2 has facilitated the development of morecomprehensive representations of land-use change in DGVMs.
• When agricultural area changes could not be matched by changes in forest imposed by HN2017, they were implemented as transitions with natural grassland.
• Harvest on secondary land does not produce a net flux to the atmosphere if considered over a long time-period (total source is equivalent to total sink).

This approach dampens the IAV of deforestation fires emissions, because it does not account for climate-driven variations in combustion completeness, or potential time lags between the clearance and actual burning of forest biomass28. The differences between SBL-Net and each of the factorial simulations (bottom panel of Fig. 4a) shows that C densities and allocation rules are the dominant factors not just for global FLUC, but also in most regions, and lead to lower RMSDHN-BLUE, compared to SBL-Net (Fig. 4b). Using HN2017 allocation fractions to pools for harvest and clearing results in lower cumulative FLUC everywhere (SHNAlloc) and decreases the RMSDHN-BLUE at global scale and in all regions but NSA and SSA. Altering C densities (SHNCdens) has contrasting effects in cumulative FLUC between regions, increasing cumulative FLUC in 3 out of 18 regions. Strong reductions in RMSDHN-BLUE for SHNFull are found in BRA, RUS, CHN and SAS (top panel in Fig. 4b), explained by RMSDHN-BLUE reductions by changing the C densities in vegetation and soil pools (SHNCdens) and allocation fractions. In SEAS, cumulative FLUC is reduced when using HN2017 parameters (SHNFull) but with a higher RMSDHN-BLUE.

1 Land-Use Harmonization 2 dataset

• In order to attribute differences in FLUC between the two models to specific aspects from Table 1, we perform a set of factorial simulationswith BLUE (see Table 2), in which we replace the BLUE parameters with thosefrom HN2017 (see also schematic in Fig. 1).
• Scenarios with reduced radiative forcing due to increased mitigation action (RCP3.4) produce increased cumulative net LULCC fluxes over the 21st century, since fossil fuel emissions are substituted partly by energy from biofuel (Hurtt et al., 2020).
• For deforestation into agricultural land, intact forests were given a high priority to be cleared, reflecting the expansion of agricultural land in temperate regions over the history, and being consistent with the current-day agricultural expansion in the tropics.
• We conduct 39 historical (from 850, 1700 or 1850 until 2015) and 12 future (2015–2100) simulations to quantify the relative importance of the uncertainty in the LULCC dataset on the historical net LULCC flux with respect to other common uncertainties.
• Table 1Naming of the main experiments based on LUH2 scenarios with low, baseline and high LULCC and three different starting years.
• A reduction of the cumulative net LULCC flux in the IC and Trans experiments initialised in 1700 or 1850 is both due to reduced contribution from harvest and pasture (only IC) and the opposite ordering of LULCC experiment in crop and abandonment contributions.
• The two bookkeepingmodels used in the global carbon budgets may differ in their FLUC estimates due to differences in the forcing data and differences in model structure, parameterisation and in how certain processes are represented.

Further division by LULCC activity is discussed in the following and shown in the Supplement (see Fig. S1). Cumulative net LULCC flux estimates are most sensitive to harvest uncertainties, mainly over northern Europe, northern Asia and south-eastern Asia (China and north-eastern India). Components of the cumulative net LULCC flux due to uncertainty of crop expansion and abandonment follow the pattern of shifting cultivation in the tropics, which means that the sensitivity to uncertainties in abandonment and crops is balanced with the opposite sign.

• For easier comparison of direct model output, we do not include these post-processing steps.
• For each of the different scenarios, no further uncertainty ranges are provided, but the set of scenarios is used to explore the impact of past LULCC uncertainties on the future net LULCC flux.
• A second notable difference in temporal dynamics can be observed in the 2000s, as has been shown by Bastos et al. (2020).
• In contrast, ELUC is impacted by both direct human management actions, and environmental changes and variations.
• If we ascribe half of this covariance to the contribution by ELUC, then managed land contributes to ~40% of the IAV of Snet.
• From 2014 onwards, each of the three historical simulations is continued with four different scenarios of future LULCC.

1 Model characteristics and datasets used

• This not only extended the LUHdatasets beyond their final year (2005 for LUH1 and 2015 for LUH2) but alsoupdated the LUH data for years that were not previously based on FAO inputs(i.e., years for which data were previously extrapolated but for which FAOdata were now available).
• Bookkeeping models help to separate direct management and environmental effects in terrestrial carbon accounting10, but their results are not directly comparable with observations due to the static nature of the applied functions.
• In the global carbon budgets since 2017 (Friedlingstein et al., 2019; Le Quéré et al., 2018a, b), FLUC estimates forrecent decades are taken as the mean of the estimates of two BK models, theone from Houghton and Nassikas (2017) (HN2017) and the BLUE model described in Hansis et al. (2015).
• The various ELUC flux components, over a long term, are exposed to global environmental changes, and on yearly to multi-decadal scales, are subject to temporal variations driven by factors including climate variation and dynamics of human decisions leading to land-use conversion11.
• Figure 5Mapped differences (LUH2-GCB2019 – LUH2 v2h) in fractions of each0.25∘ grid cell for cropland, grazing land, secondary forest, andprimary forest in 2000, 2009, and 2015.
• Furthermore, the IRS requires taxpayers to choose an accounting method that accurately reflects their income and to be consistent in their choice of accounting method from year to year.

Nevertheless, the results presented here provide a reference for comparisons with the upcoming CMIP6 model simulations. It should be noted, however, that specific transitions and prevalence of specific PFTs in certain regions prohibits generalising this statement. Together with the larger land-use dynamics which stem from BLUE representing gross transitions and its usage of LUH2v2.1 as LUC forcing, these Food Truck Accounting changes lead to overall higher carbon losses that have a faster decay.

Data availability

Estimating FLUC bookkeeping model accurately in space and in time remains, however, challenging, due to multiple sources of uncertainty in the calculation of these fluxes. This uncertainty, in turn, is propagated to global and regional carbon budget estimates, hindering the compilation of a consistent carbon budget and preventing us from constraining other terms, such as the natural land sink. Uncertainties in FLUC estimates arise from many different sources, including differences in model structure (e.g. process based vs. bookkeeping) and model parameterisation.

What Is an Accounting Method?

Similarly,LUH2-GCB2019 also reduced the area of grazing land in Brazil between theyears 1990 and 2010 when compared with the LUH2 v2h grazing land area. The LUH2-GCB2019 dataset shares many of the same properties as itspredecessors, LUH2 v2h, LUH2-GCB2017, and LUH2-GCB2018, and is identical tothose datasets for all years up to 1950. Figure 2a shows the global area ofthe five aggregated land-use states represented by LUH2-GCB2019 for theyears 1950–2019. Global cropland area increased from 12.2×106 km2 in 1950 to 15.9×106 km2 in 2015 (for LUH2 v2h)or 16.1×106 km2 in 2015 (for LUH2-GCB2019) and16.8×106 km2 in 2019 (for LUH2-GCB2019). Global grazingland area increased from 26.1×106 km2 in 1950, peaked inthe year 2000 at 33.2×106 km2 (for LUH2 v2h) or peakedin 2001 at 33.1×106 km2 (for LUH2-GCB2019), and thendecreased to 32.8×106 km2 in 2015 (for LUH2 v2h) or32.7×106 km2 in 2015 (for LUH2-GCB2019) and32.6×106 km2 in 2019 (for LUH2-GCB2019). Cropland,grazing land, and urban land areas in LUH2-GCB2019 were identical to thoseareas provided by the HYDE dataset at global, regional, and 0.25∘spatial scales (by design).