Land use shapes the relationship between tree cover and carbon stocks in the tropics

Tree cover (TC) and biomass carbon stocks (CS) are key parameters for characterizing the states and dynamics of tropical ecosystems. Despite the presence of several datasets with high spatial resolution, differences among data products prevail and systemic inter-relations between TC and CS remain poorly quantified. Further, the role of land use in explaining disagreements among datasets remains largely unexplored. Here, by combining established spatially-explicit estimates of TC and CS over contemporary timescales, we analyse uncertainties between these two ecosystem parameters across the global tropics (~ 23.4°N to 23.4°S). We quantify land use effects by contrasting actual and potential (ie. in the hypothetical absence of land use) states of vegetation and by correlating TC and CS changes with land use intensity. Our results show that land use strongly alters both TC and CS, with disproportionate impacts on CS and large variations across tropical ecozones. Differences between potential and actual vegetation CS remain above 50% across the tropics except for rainforests (34%). Differences within corresponding TC estimates are more variable, and higher among sparsely-vegetated landscapes (81% for shrublands), highlighting the overwhelming extent of land use impacts. Cross-comparisons across available TC and CS datasets reveal large spatial disagreements. More than a third of all identified co-located TC and CS change datasets show disagreements in the direction of change (Gain vs Loss), and these divergences persist as a function of land use intensities. Our results provide a characterization of the prevailing uncertainty structures of input datasets and the spatial patterns of land use-induced disturbances at the pixel and ecozone-levels. This assumes added significance in light of the stock-taking exercises envisaged as part of the Paris Agreement, the advancement of terrestrial carbon modelling initiatives as well as emerging, novel remote sensing products.