During the last decade, remote sensing sounders have demonstrated their capability for monitoring atmospheric composition and pollution. The Infrared Atmospheric Sounding Interferometer (IASI) instrument delivers more than 1.3x106 spectra per day and provides global coverage twice a day. It is characterized by an apodized resolution of 0.5 cm-1 and a low radiometric noise. Concentrations and seasonal patterns for several atmospheric key species are routinely retrieved from the IASI spectra using inverse methods. Such processes rely on both observations and an a priori, which can lead to errors in the geophysical retrievals. Retrieving carbon dioxide (CO2) or methane (CH4) is challenging due to the sensitivity of infrared channels to both temperature and concentration variations. In addition, spectroscopic broadening effects (for example line-mixing) specific to these molecules add more complexity in the retrieval process. In this work we investigate the temporal variations of some atmospheric species by using only the L1C radiances. This method avoids the simulation of the complex spectroscopic effects and provides a new process to analyze the IASI dataset. Only the clear sky scenes are taken into account in this work. Clear sky scene are discriminated from cloudy scenes using the IASI L2 products. This work discusses the L1C radiances analysis for 5 complete years of IASI observations from January 2008 to December 2012, in CH4 and CO2 spectral ranges to study their temporal variability. Seasonal patterns calculated for different latitude bands are discussed for each molecule. These patterns highlight the typical seasonal variations of the species. Geographical distributions of mean radiance value are calculated for each month and year, the long term evolution as well as sporadic events are discussed.