A model-based three-dimensional (3-D) climatology of atmospheric CO2 concentrations has been constructed for the analysis of satellite observations, as a priori information in retrieval calculations, and for preliminary evaluation of remote sensing products. The locations of ground-based instruments and the coverage of aircraft in situ measurements are limited and do not represent the full atmospheric column, which is a primary requirement for the validation of satellite data. To address this problem, we have developed a method for constructing a 3-D CO2 climatology from the surface up to approximately 30 km by combining information from in situ measurements and several transport models. The model-simulated CO2 concentrations have been generated in the framework of the TransCom satellite experiment. The spatial and temporal biases of the transport-model-derived data set have been corrected using in situ CO2 measurements in the troposphere and in situ profiles of the mean age of air in the stratosphere. The constructed multimodel mean CO2 climatology represents the seasonal cycle and the inter-hemispheric gradient better than each transport model. Our approach performs well near the surface and in regions where the observational network is relatively dense. The column-mean CO2 of the constructed climatology was reduced by ∼1 ppm from that of a single transport models, consistent with model validation against measurements of the CO2 total column.