The CO columns retrieved by the Measurement of Pollution in the Troposphere (MOPITT) satellite instrument between May 2000 and April 2001 are used together with the Intermediate Model for the Annual and Global Evolution of Species (IMAGES) global chemistry transport model and its adjoint to provide top-down estimates for anthropogenic, biomass burning, and biogenic CO emissions on the global scale, as well as for the biogenic volatile organic compounds (VOC) fluxes, whose oxidation constitutes a major indirect CO source. For this purpose, the big region and grid-based Bayesian inversion methods are presented and compared. In the former setup, the monthly emissions over large geographical regions are quantified. In the grid-based setup, the fluxes are optimized at the spatial resolution of the model and on a monthly basis. Source-specific spatiotemporal correlations among errors on the prior emissions are introduced in order to better constrain the inversion problem. Both inversion techniques bring the model columns much closer to the measurements at all latitudes, but the grid-based analysis achieves a higher reduction of the overall model/data bias. Further comparisons with observed mixing ratios at NOAA Climate Monitoring and Diagnostics Laboratory and Global Atmosphere Watch sites, as well as with airborne measurements are also presented. The inferred emission estimates are weakly dependent on the prior errors and correlations. Our best estimate for the global CO source amounts to 2900 Tg CO/yr in both inversion approaches, about 5% higher than the prior. The global anthropogenic emission estimate is 18% larger than the prior, with the biggest increase for east Asia and a substantial decrease in south Asia. The vegetation fire emission estimates decrease as well, from the prior 467 Tg CO/yr to 450 Tg CO/yr in the grid-based solution and 434 Tg CO/yr in the monthly big region setup, mainly due to a significant reduction of African savanna fire emissions. The biogenic CO/VOC flux estimates are found to be enhanced by about 15% on the global scale. The most significant error reductions concern the biogenic emissions in the tropics, the Asian anthropogenic emissions, and the vegetation fire source over Africa. Our inversion results are further compared with previously reported emission estimates.