Understanding Global CO₂ Fluxes and Concentrations using Multi-Model Simulations and Satellite Observations

Accurate quantification of the global carbon cycle is essential for projecting future climate change, yet significant uncertainties remain in partitioning regional land and ocean CO₂ fluxes. This study presents a comprehensive evaluation of XCO₂ retrievals from four major satellite missions, such as GOSAT, GOSAT-2, OCO-2 (v11), and OCO-3 (v11), against four Global Carbon Project (GCP) atmospheric transport models (MIROC4-ACTM, COLA, NISMON, and GCASv2). We employ a harmonised approach utilising averaging kernel convolution and data-driven bias correction for the year 2020 to facilitate a consistent model-satellite inter-comparison. Our results demonstrate that while the transport models generally reproduce global and seasonal CO₂ distributions, significant regional biases persist, notably over boreal and tropical land areas where discrepancies often exceed ±2 ppm. We identify that structural differences between satellite observations are primarily attributed to distinct sampling patterns. Specifically, comparisons between the sun-synchronous OCO-2 and the ISS-mounted OCO-3 reveal systematic differences driven by OCO-3's wider range of Local Solar Hour (LSH) sampling. This sampling captures diurnal CO₂ variability that is not fully resolved by current transport models, particularly in regions with strong diurnal cycling. Furthermore, multi-model flux analyses for the 2016–2019 period highlight that the largest uncertainties in surface fluxes occur over the high-latitude oceans and tropical land regions. These flux uncertainties correlate strongly with the observed model-satellite mismatches, underscoring the need for improved representation of diurnal cycles, vertical transport, and surface exchanges in atmospheric inversion systems. This integrated assessment provides crucial diagnostics for advancing the fidelity of global carbon cycle monitoring and modelling.