Microwave observations from AMSU-A & AMSU-B (or MHS) instruments have been widely used in Numerical Weather Prediction (NWP) to improve the initial conditions for short-range forecasts. Although the information content of these observations is important, their use in NWP models is still insufficient. Indeed, many issues are still to be addressed, in particular, the assimilation of observations over land surfaces. This requires significant work on the surface representation in the models (emissivity and temperature). The surface emissivity exhibits a high and complex variability whose effect may be important for the forward model simulations. This translates in the inability of the forward model to produce realistic simulations of surface sensitive channels and therefore to reject many sounding channels during the data assimilation process. In this talk, new methods for estimating the emissivity and/or surface temperature will be presented. We will review the impact of these methods during the assimilation process, especially on the performance of the radiative transfer model RTTOV. we will show how the use of an improved description of the land surface emissivity allows the assimilation system to take advantage of the information content of AMSU temperature and humidity sounding channels over land.