Atmospheric CO2 inversion models overestimate northern extratropical land and ocean carbon uptake as assessed at background in-situ sites

Abstract Several aspects of the global biospheric carbon cycle are highly uncertain, including estimation of CO₂ budgets at hemispheric scale by top-down inverse modelling approach. We use CO₂ fluxes from a model intercomparison project (OCO-2_v10_MIP) that was conducted using long-term background station CO₂ and total-column CO₂ (XCO₂) observations for the period 2015–2020. Total (ocean + land) CO₂ flux and mean model-observed CO₂ differences against 50 background sites exhibit statistically significant correlations for the northern and southern extra-tropics (latitude > 30o). Using these correlations, we define emergent constraints to calculate “best estimate” total CO₂ fluxes. After correcting ocean fluxes and riverine carbon export, we estimate terrestrial CO₂ fluxes or land carbon stockchange of -1.54 ± 1.18, -0.04 ± 1.11 and − 0.13 ± 0.27 PgC yr⁻ ¹ in the northern extra-tropics, tropics, and southern extra-tropics, respectively. These fluxes agree well with those based on national inventories and land surface observations, but suggest that CO₂ inversion models generally overestimate northern extratropical uptake and tropical emissions.