A consistent budgeting of terrestrial carbon fluxes

As the remaining carbon budget to limit global warming in line with the Paris Agreement is rapidly shrinking, accurate estimates of the emissions from land-use and land cover change (ELUC) and the terrestrial natural CO2 sinks (SLAND) are crucial. In current carbon budgeting approaches, the ELUC and SLAND estimates are conceptually not consistent, since they stem from two different model families that differ in how CO₂ fluxes are attributed to environmental or land-use changes. Consequently, anthropogenic and natural budget terms are not fully distinguished. ELUC is estimated by bookkeeping models, which typically use time-invariant carbon densities representing contemporary environmental conditions. They thus assume a steady environmental state and neglect changes in environmental conditions preceding or succeeding a land-use change event, e.g., denser growing forests in response to rising atmospheric CO2 concentrations, which emit more when cleared for agricultural land. SLAND is estimated by dynamic global vegetation models, which account for environmental changes but assume that the land cover distribution remained at its pre-industrial state. They thus include carbon sinks in forests that in reality were cleared decades ago. Here we suggest an approach for consistent budgeting of ELUC and SLAND by integrating the response of vegetation and soil carbon to environmental changes, derived from dynamic global vegetation models, into a spatially explicit bookkeeping model (BLUE). A set of dedicated simulations allows us to disentangle and re-attribute environmental and land-use components of the land-atmosphere CO2 exchange. Our results show that land is a cumulative net source of CO₂ since 1850, which contrasts current global carbon budgets indicating a net sink. The underlying reason is both a higher estimate of ELUC than previously suggested as well as a smaller land sink: The implementation of environmental changes increases global ELUC over time (14% compared to current estimates for 2012-2021) mainly due to increased emissions from deforestation and wood harvest, which are only partly offset by increased sinks through reforestation/afforestation and other regrowing vegetation. Our SLAND estimate calculated under actual land cover amounts to 3.0 GtC yr^-1 for 2012-2021, which is substantially lower both globally and regionally compared to estimates assuming pre-industrial land cover: we find a SLAND is smaller by 0.7 GtC yr^-1 in 2012-2021, i.e., 19% lower as compared to the conventional approach using pre-industrial land cover. The overestimate of SLAND under pre-industrial land cover is particularly pronounced in regions with strong ecosystem degradation, such as Southeast Asia, Brazil, and Equatorial Africa. The consistent estimation of terrestrial carbon fluxes is thus essential not only to provide a tangible estimate to monitor the progress of net-zero emission commitments and the remaining carbon budget, but also to highlight the need to protect remaining natural ecosystems for climate regulation. Our approach provides greater consistency with atmospheric inversions and provides a finer split of anthropogenic and natural fluxes useful for a direct comparison of global carbon cycle models to national greenhouse gas inventories.