Abstract. Regional net carbon fluxes of terrestrial ecosystems could be estimated with either biogeochemistry models by assimilating surface carbon flux measurements or atmospheric CO₂ inversions by assimilating observations of atmospheric CO₂ concentrations. Here we combine the ecosystem biogeochemistry modeling and atmospheric CO₂ inverse modeling to investigate the magnitude and spatial distribution of the terrestrial ecosystem CO₂ sources and sinks. First, we constrain a terrestrial ecosystem model (TEM) at site level by assimilating the observed net ecosystem production (NEP) for various plant functional types. We find that the uncertainties of model parameters are reduced up to 90% and model predictability is greatly improved for all the plant functional types (coefficients of determination are enhanced up to 0.73). We then extrapolate the model to a global scale at a 0.5° × 0.5° resolution to estimate the large-scale terrestrial ecosystem CO₂ fluxes, which serve as prior for atmospheric CO₂ inversion. Second, we constrain the large-scale terrestrial CO₂ fluxes by assimilating the GLOBALVIEW-CO2 and mid-tropospheric CO₂ retrievals from the Atmospheric Infrared Sounder (AIRS) into an atmospheric transport model (GEOS-Chem). The transport inversion estimates that: (1) the annual terrestrial ecosystem carbon sink in 2003 is −2.47 Pg C yr⁻¹, which agrees reasonably well with the most recent inter-comparison studies of CO₂ inversions (−2.82 Pg C yr⁻¹); (2) North America temperate, Europe and Eurasia temperate regions act as major terrestrial carbon sinks; and (3) The posterior transport model is able to reasonably reproduce the atmospheric CO₂ concentrations, which are validated against Comprehensive Observation Network for TRace gases by AIrLiner (CONTRAIL) CO₂ concentration data. This study indicates that biogeochemistry modeling or atmospheric transport and inverse modeling alone might not be able to well quantify regional terrestrial carbon fluxes. However, combining the two modeling approaches and assimilating data of surface carbon flux as well as atmospheric CO₂ mixing ratios might significantly improve the quantification of terrestrial carbon fluxes.