Persistence of the Global Carbon Budget Imbalance

Despite major technical advances in the estimation of all fluxes in the carbon cycle, including emissions from fossil and land use sources and sinks to land and oceans, mathematical imbalances continue to arise when these fluxes are combined. The global carbon budget imbalance (BIM) – a measure of the annual emissions and/or sinks of CO₂ that cannot be accounted for – has become a key evaluation metric for the global carbon budget. Here we show that the BIM has not significantly reduced during 2011 to 2023 despite the introduction of a large number of processes and improvements in model, after an initial improvement between 2009 and 2011 likely due to the increase in the number of land models from 5 to 9. We assessed the potential to reduce the BIM by selecting different combinations of estimates from available models of each global carbon flux, systematically considering 7,864 combinations of land use estimates from 19 bookkeeping and dynamic global vegetation models (DGVMs), land sink estimates from 17 DGVMs, and ocean sink estimates from 19 data-products and global ocean models. We show that the multi-model average generally performs well compared to individual models. A selected ensemble of best performing combinations reduced the BIM by one third (from 0.68 GtC/yr to 0.45 GtC/yr), but the selected models were dependent on the time-period analysed. The selection of land sink models shows greatest potential for BIM reductions, with selection rates particularly impacted by a model’s representation of gross primary productivity and heterotrophic respiration response to climate variability. Bookkeeping estimates for land-use change and data-based estimates for the ocean CO₂ sink were not selected preferentially over process models. Despite the potential for BIM reduction through model selection, its absolute value remains substantial, suggesting that processes are missed or incompletely represented incompletely.