Two simulations of the seasonal variation of the global atmospheric CO2 distribution are obtained by combining an atmospheric transport model, two parameterizations of soil heterotrophic respiration (SHR), and a mechanistic model of carbon assimilation in the biosphere (CARAIB) that estimates the net primary production (NPP) of continental vegetation. The steady state hypothesis of the biosphere allows the spatial distribution and the global content of the soil carbon to be expressed as a function of the root fractions of soil respiration under forested and herbaceous vegetation covers. The sensitivity of the modeled CO2 signal to the wind field does not exceed the observed interannual variability. The influence of the various vegetation zones is quantified by the Fourier analysis of the modeled atmospheric signal. In the northern hemisphere, the temperate ecosystems dominate the seasonal atmospheric signal of the extratropical latitudes. The ecosystems of the tropical northern zone determine the local signal, while the southern tropical ecosystems influence largely the signal in the whole southern hemisphere. The results give credence to the mechanistic modeling of NPP since the simulated atmospheric signal is comparable with that obtained with normalized difference vegetation index (NDVI) based diagnostic models coupled with a parameterization of SHR fitted to optimize the atmospheric signal. ¿ American Geophysical Union 1996