The use of global three-dimensional (3-D) models with satellite observations of CO<sub>2</sub> in inverse modeling studies is an area of growing importance for understanding Earth's carbon cycle. Here we use the GEOS-Chem model (version 8-02-01) CO<sub>2</sub> mode with multiple modifications in order to assess their impact on CO<sub>2</sub> forward simulations. Modifications include CO<sub>2</sub> surface emissions from shipping (~0.19 Pg C yr<sup>−1</sup>), 3-D spatially-distributed emissions from aviation (~0.16 Pg C yr<sup>−1</sup>), and 3-D chemical production of CO<sub>2</sub> (~1.05 Pg C yr<sup>−1</sup>). Although CO<sub>2</sub> chemical production from the oxidation of CO, CH<sub>4</sub> and other carbon gases is recognized as an important contribution to global CO<sub>2</sub>, it is typically accounted for by conversion from its precursors at the surface rather than in the free troposphere. We base our model 3-D spatial distribution of CO<sub>2</sub> chemical production on monthly-averaged loss rates of CO (a key precursor and intermediate in the oxidation of organic carbon) and apply an associated surface correction for inventories that have counted emissions of CO<sub>2</sub> precursors as CO<sub>2</sub>. We also explore the benefit of assimilating satellite observations of CO into GEOS-Chem to obtain an observation-based estimate of the CO<sub>2</sub> chemical source. The CO assimilation corrects for an underestimate of atmospheric CO abundances in the model, resulting in increases of as much as 24% in the chemical source during May–June 2006, and increasing the global annual estimate of CO<sub>2</sub> chemical production from 1.05 to 1.18 Pg C. Comparisons of model CO<sub>2</sub> with measurements are carried out in order to investigate the spatial and temporal distributions that result when these new sources are added. Inclusion of CO<sub>2</sub> emissions from shipping and aviation are shown to increase the global CO<sub>2</sub> latitudinal gradient by just over 0.10 ppm (~3%), while the inclusion of CO<sub>2</sub> chemical production (and the surface correction) is shown to decrease the latitudinal gradient by about 0.40 ppm (~10%) with a complex spatial structure generally resulting in decreased CO<sub>2</sub> over land and increased CO<sub>2</sub> over the oceans. Since these CO<sub>2</sub> emissions are omitted or misrepresented in most inverse modeling work to date, their implementation in forward simulations should lead to improved inverse modeling estimates of terrestrial biospheric fluxes.