Advanced infrared sounders will provide thousands of radiance data at every observation location. The number of individual pieces of information is not usable in an operational NWP context, and we have investigated the possibilities of choosing an ''optimal'' subset of data. These issues have been addressed in the context of optimal linear estimation theory, using simulated data. Several methods have been tried to select a set of the most useful channels for each individual atmospheric profile. An iterative method selecting sequentially the channels with largest information content was demonstrated to always produce the best results, but at a relatively large cost. To test the robustness of this iterative method, a variant has been tried. It consists in building a mean channel selection aimed at optimizing the results over the whole profile database, and then applying to each profile this "constant" selection. Results show that this "constant" iterative method is very promising. The practical advantage of this method for operational purposes is that the same set of channels can be used for various atmospheric profiles. This approach has been applied to simulated AIRS data. Results show that, although our "constant" channel selection is significantly different from the one used by NESDIS for the distribution of AIRS data to NWP centres, results in terms of retrieval accuracy are very similar. These channel selection studies apply to clear-sky conditions and have to be extended to cloudy conditions. In order to address the question of cloudiness in sensitive areas, ten FASTEX (Fronts and Atlantic Storm-Track Experiment) cases in February 1997 have been studied. Sensitive areas are regions where small errors in the initial conditions can lead to strong forecast errors. In every sounder pixel, cloud parameters were deduced from the AVHRR (Advanced Very High Resolution Radiometer) observations with the MAIA (Mask AVHRR for ATOVS Inversion). Results show that most of the sensitive area is covered with clouds. High clouds are mostly located in the southern part of the sensitive area, while low clouds predominantly affect the northern part. It is hoped that an appropriate channel selection can extract most information about the atmospheric profile above low clouds in this northern part of the sensitive area, leading to a substantial improvement in the forecast.Finally, the assimilation of advanced sounder data over land is being prepared. Spectrally varying emissivities are computed for each soil type used as climatological input to the NWP model. These emissivities are then validated by comparing them for selected spectral bands to MODIS data. This preparation for advanced sounder data has been performed using mainly AIRS and IASI simulated data. As AIRS data become available, we intend to start monitoring the radiance data and revisit the subject with real statistics.