The potential impact of SSMIS on numerical weather prediction (NWP) in the middle atmosphere is great, but we must first understand (1) errors in observed brightness temperatures, (2) errors in the radiative transfer model, and (3) errors in the model forecast (background). SSMIS radiances from the upper atmospheric sounding channels have been validated using ground-based lidar observations as ground truth, coupled with radiative transfer modeling. CRTM has been validated against an independent line-by-line model. Subsequently the fast model, together with SSMIS radiances, was used to investigate errors in ECMWF analyses in the altitude range of 45-80 km. Significant model biases are indicative of the potential of SSMIS to improve the representation of the upper stratosphere and lower mesosphere in NWP models through data assimilation.