Carbon monoxide CO is a key trace gas in the troposphere. Its concentration varies from 50 ppb at the surface in remote regions of the southern hemisphere to over 300 ppb in polluted regions. In megacities, the CO concentrations often reach several ppm and cause health problems. The combustion of fossil fuel and biofuel and vegetation (forest, savannah, grassland, agricultural wastes) produce large quantities of CO. The oxidation in the atmosphere of methane and other organic compounds results in the production of large amount of CO as well. Globally, the reaction of CO with the hydroxyl radical OH is the main sink for both CO and OH. CO lifetime varies from 1 to 5 months, depending on the latitude and season mostly. As OH is the main detergent of the atmosphere, the amount of CO in the troposphere can indirectly impact the lifetime of greenhouses such as CH4, HFC, HCFC. In NOx-rich environment and in the presence of ultra-violet radiation, CO can also be a precursor of tropospheric ozone.

Transportation represents 90% of the total CO sources in urban areas whereas in the tropics, seasonal biomass burning is the main source of CO. The assessment of CO global budget is complicated due to the fact that its sources and sinks are quite variable in both space and time. Atmospheric Chemistry and Transport Models (CTM) together are used to simulate the global CO budget. Numerical studies have shown that the amount of tropospheric CO has doubled since the pre-industrial era. For the last 20 years, an international network of surface stations led by NOAA/CMDL has provided weekly measurements of CO mostly in remote locations. Several field campaigns have also taken place on various continents and have enabled scientists to study more specific pollution and transport processes. In December 1999, the MOPITT instrument, an infra-red gas correlation radiometer, was launched onboard the NASA/Terra platform. The retrievals of CO from the radiances measurements provide the first global and ongoing coverage of CO distribution in the troposphere. The conjunction of the improvement in the areas of air composition modeling and monitoring have made it possible to successfully implement inverse modeling techniques mostly used at the time for the study of the carbon cycle.

I have combined the information contained in a 3D global CTM and in observations to reduce the uncertainty on CO anthropogenic sources. Using in situ and remote sensing data in a recursive time-dependent synthesis Bayesian inversion algorithm, I have calculated new estimates of CO sources for 12 months. The inversion was able to better retrieve the seasonal cycle for biomass burning emissions in the tropical and boreal regions, improving the agreement with other studies based on remote sensing measurements. It also identified a large under-estimation of CO emissions in Asia. With the current CMDL network of stations, the problem of optimizing CO sources is regionally under-constrained. Uncertainties in the emissions over the Southern Hemisphere continents could only be significantly reduced when using the satellite data. The optimized annual and global budget of CO obtained in the various inversion configurations converges towards the upper end of the IPCC range for CO total source: 1800-2700 TgCO/yr. Sensitivity studies have shown the dependence of the inverse problem solution on several hypotheses and on simplifications made to decrease the amount of computations required for the inversion.