A comprehensive global 3D model of δ 18 O in atmospheric CO 2 I have built the first comprehensive global 3D model to simulate δ 18 O in atmospheric CO 2 . The aim is to disentangle this isotope signal which offers the potential to deter- mine the CO 2 gross fluxes of the terrestrial biosphere. Model surface processes have been validated against regional experiments while the integrated atmospheric signal has been investigated comparing with global atmospheric observations. Modelled surface processes compare well with observed data over the whole latitude range except for high northern latitudes where the experimental data are very sparse. The water isotopes of rain are well reproduced but the soil signal is too attenuated in the model. The model further simulates very well the seasonal cycle of atmospheric CO 2 .For δ 18 O–CO 2 ,it precedes the measured cycle by two month. This discrepancy could not be resolved with our current knowledge of δ 18 O–CO 2 . The modelled amplitude is very sensitive to different parameterisations, standard values result in only 2 / 3 of the observed ampli- tude. Concerning the meridional gradient in δ 18 O–CO 2 , I could show that about one half of the signal comes from the covariance of δ 18 O–CO 2 sources with atmospheric transport and the other half from the imbalance of δ 18 O–CO 2 isofluxes between the northern and southern hemisphere.