High-resolution estimates of vegetation and soil carbon densities for regional and global carbon budgets

Mitigating global climate change requires massive greenhouse gas emission reductions and carbon removal efforts. Although terrestrial ecosystems store large amounts of carbon, land-use change has substantially diminished these stocks in many regions. However, a consistent, high-resolution approach to quantify the differences between actual and potential carbon stocks in vegetation and soils – the terrestrial carbon deficit – remains elusive, limiting the evaluation of global climate models. In particular the high spatial heterogeneity of vegetation and soil organic carbon stocks at the ecosystem level introduces major uncertainty into common methods for estimating land-use change carbon fluxes, propagating uncertainties into national, regional and global carbon budgets.

Here, we generate spatially explicit maps of vegetation and soil carbon stocks for ten ecosystem types by combining a machine-learning algorithm with semi-empirical observations and simulations of global dynamic vegetation models (DGVMs). Our results show that commonly used default carbon values substantially underestimate the heterogeneity of carbon within ecosystems. By integrating our spatially explicit carbon data into the bookkeeping of land-use emissions model BLUE – one of the models underlying the net land-use change flux estimates of the annual Global Carbon Budget of the Global Carbon Project –, we find that global estimates of the net land-use change flux for 1960–2023 are 3–14% lower than estimates relying on default values from the literature. The estimates further reveal in several regions pronounced differences of more than 20%, highlighting the value of spatially explicit carbon data for accurate national and sub-national net land-use change flux assessments. Improving this accuracy reduces the uncertainty in net land-use change flux estimates and in land-based carbon mitigation potential calculations, which both are fundamental for informing political decision-making to achieve carbon neutrality and global climate targets.

Across ecosystems, we quantify the terrestrial carbon deficit to be 344 (251–393) PgC, equivalent to a 24% depletion, predominantly driven by pasture expansion (30%), cropland expansion (24%), and forest management (23%). We reveal that dynamic global vegetation models (DGVMs) underestimate the terrestrial carbon deficit by 37% on average (range: 2%–58%), highlighting critical limitations. Our findings support assessments of anthropogenic impacts on ecosystems and help constrain global climate models to better evaluate nature-based solutions and climate mitigation policies.