Making Satellite Retrievals Compatible with NWP Background Fields for NWP Data Assimilation

Numerical Weather Prediction (NWP) centers assimilate a small subset of clear sky channel radiances from the entire spectrum of radiance measured by hyperspectral sounding instruments. The reason for this traces back to the assimilation of the low spectral resolution radiance data observed within the small number of spectral channels provided by prior generation sounding radiometers. Thus, it was more efficient to assimilate a small number of spectral channel radiances rather than a complete vertical profile of retrieved temperature and humidity values. Most important, the assimilation of radiance, rather than the retrieved vertical profile variables, avoided vertical aliasing of the model background fields into which they were to be assimilated. This vertical aliasing was caused by the low vertical resolution of the profile retrievals from prior generation sounding radiances.

However, today's hyperspectral resolution satellite sounding instruments provide a spectrum of thousands of spectral channel radiances with high spectral resolution. These spectra can provide profile retrievals with relatively high vertical resolution for cloudy as well as clear sky conditions over both land and sea. A fast, practical technique has been developed for handing the problems with vertical aliasing that plagued earlier attempts to assimilate satellite profile retrievals. This technique includes: (1) a dual-regression retrieval method capable of inverting radiances into profiles quickly, (2) a simple method for removing aliasing caused by vertical resolution differences between the retrieved profile and the model background grid into which it is to be assimilated, and (3) the use of a fast principle component based radiative transfer model. Here, we describe the technique and present results that demonstrate the improvement of NWP model background fields due to the assimilation of the satellite hyperspectral sounding information.