A practical method to infer CO2 sources, sinks, and fluxes from measured mean concentration profiles is proposed and field-tested in a uniform pine forest using eddy covariance measurements. The proposed method computes the velocity statistics from a second-order closure model and uses these statistics to infer profiles of scalar fluxes using a scalar-flux budget. The model input requirements are leaf area density profile, mean shear stress at the canopy top, and measured concentration profiles within and just above the canopy. In contrast to the localized near-field (LNF) theory the model does not assume zero vertical velocity skewness, negligible advective effects, and local vertical homogeneity in the near-field concentration. The model results compared well with eddy covariance CO2 flux measurements inside the canopy and reproduced qualitatively much of the physiologically known daytime evolution of the CO2 source-sink profile. ¿ 1999 American Geophysical Union