<\/p>\n
SLAND,pi and SLAND,TRENDY are both based on pre-industrial land cover and, thus, their difference is due to model-intrinsic differences between BLUE and the DGVMs from the TRENDY project. Uncertainties (shaded areas in time series and whiskers in inset) indicate the minimum-to-maximum range across BLUE estimates of SLAND,trans, SLAND,pi, RSS, and LASC obtained by scaling BLUE carbon densities with individual DGVMs (dots indicate individual estimates; see Methods) and one standard deviation for TRENDY estimates. The DGVM simulations with observed transient environmental conditions used observation-based temperature, precipitation, and incoming surface radiation data at 0.5\u00d70.5 degree spatial resolution of the Climatic Research Unit (CRU) and Japanese Reanalysis (JRA; Friedlingstein et al., 2019; Harris et al., 2014). Annual time series of global atmospheric CO2 concentrations for 1700\u20132018 was derived from ice core data (before 1958; Joos and Spahni, 2008) merged with National Oceanic and Atmospheric Administration (NOAA) data (from 1958 onward; Dlugokencky and Tans, 2020). While HYDE agricultural areas are used in LUH2, the main difference lies in LUH2 additionally adding wood harvest from the Global Forest Resources Assessments of the FAO and sub-grid-scale (\u201cgross\u201d) transitions to capture shifting cultivation in the tropics. Fluxes from deforestation, changes in land cover, land use andmanagement practices (FLUC for simplicity) contributed to approximately 14\u2009% of anthropogenic CO2 emissions in 2009\u20132018.<\/p>\n<\/p>\n