Understanding and predicting the interannual variability (IAV) of the global terrestrial carbon cycle

During the recent half-century, anthropogenic activities have led to a significant amount of CO2 emission from fossil fuel combustion and land-use changes. The global terrestrial ecosystems serve as a natural sink to mitigate the rapid growth in atmospheric CO2 and climate warming. However, terrestrial carbon flux shows large interannual variability (IAV), mostly driven by climate variability (e.g., temperature, precipitation, solar radiation). Currently it still remains elusive (1) which region (e.g., humid vs semi-arid) contributes most to the global IAV; and (2) how each individual component carbon flux (e.g., gross primary production GPP, ecosystem respiration Reco, or fires) of net biosphere exchange (NBE) is affected by climate variability, which hinders reliable prediction of the fate of terrestrial carbon sink in future climate projections. To address these questions, we first quantified the regional contribution to the IAV signal of global NBE using several atmospheric inversion datasets. We found that tropical (25S-25N) and humid (Aridity Index, AI>0.65) region has the most contribution, with tropical grasslands/savannas showing a larger or equal contribution than tropical evergreen broadleaf forests. Next, based on a causal inference framework, we inferred the causal strengths between carbon fluxes and environmental drivers for each key region. We found dominant control of water availability over tropical grasslands/savannas and inconsistent environmental control over tropical evergreen broadleaf forests (e.g., Amazon vs Congo). Finally, we evaluated the simulations from Earth System Models (ESMs) in the TRENDY project and found discrepancies in detected causal effects and inferred causal strengths compared to the observation-based analyses, which offers insights for future improvement for ESMs.