CO2 fluxes were measured continuously for twelve months (2003) using eddy covariance technique at canopy layer in a dominant subtropical forest in South China. Our results showed that daytime maximum CO2 fluxes of the whole ecosystem varied from −15 to −20 µmol m−2 s −1 5 . The peaks of CO2 fluxes appeared earlier than the peaks of solar radiation. Contribution of CO2 fluxes in a subtropical forest in the dry season was 53% of the annual total from the whole forest ecosystem. Daytime CO2 fluxes were very large in October, November and December, which was therefore an important stage for uptake of CO2 by the forest ecosystem from the atmosphere. 10 Using the estimates of biomass, soil carbon and parameters of leaf photosynthesis from other studies at the same forest, we ran a process-based model, CBM (stands for CSIRO Biosphere Model) for this site, and compared the predicted fluxes of CO2 with measurements. We obtained reasonable agreement. The mean difference between the simulated and measured daytime CO2 fluxes from the year-round (8249 records) was −0.2 µmol m−2 s −1 15 and implied well within measurement accuracy. Based on estimates of forest ecosystem respiration, NEE was calculated −242 and −276 gCm−2 year−1 for measured and modelled, respectively. In previous study, NPP for this forest stand was 694 gCm−2 year−1 during 2003/04 and litterfall was 424 gCm−2 year−1 . We therefore calculated NEE as −270 gCm−2 year−1 and very sim20 ilar to the values obtained by measured and modelled CO2 fluxes in this study