In view of the importance of studying a global carbon cycle and inadequacy of a groundbased network of observations of the atmospheric carbon dioxide (CO 2 ) concentrations and its fluxes, the feasibility of remote sensing of the atmospheric CO 2 concentration obtained from the satellite high-spectral resolution IR-sounder data is studied. In reference to the AIRS infrared-sounder measurements (the EOS Aqua satellite) the satellite data informativeness is analyzed, and a subset of most sensitive channels in respect to the CO 2 concentration variations is selected. A method of retrieving the carbon dioxide concentration X CO 2 averaged over height from the AIRS data (in the middle and upper troposphere) is suggested and tested. The method is based on a numerical solution of an inverse problem. The comparison of satellite estimates of X CO 2 averaged over a month with the aircraft (in situ) data on observations over the areas of boreal forests (the Novosibirsk region) and ecosystems (the region of Surgut) for 10 months of 2003 confirms the possibility to trace a seasonal trend of X CO 2 with an error not worse than 1%.