Evaluation of the Forecast Impact of Assimilating High Density Satellite Radiances with NAVDAS-AR

It has long been the practice at operational NWP centers to substantially thin satellite observations, particularly radiances, before assimilation. The reasons for the aggressive thinning include computational considerations (to meet operational timing requirements), the desire for a diverse “mix” of observations, and to avoid the effects of correlated observation errors which are not accounted for in current operational DA schemes.

Recently, Bormann and Bauer (2010) and Bormann et al. (2010) have found that the spatial and interchannel observation error correlations are small for AMSU-A microwave temperature sounding channels and for AIRS and IASI long-wave temperature sensitive channels that peak in the mid-troposphere to stratosphere. At the recent 17th International TOVS Study Conference, P. Bauer presented results illustrating improvements to forecast skill with the ECMWF NWP system through the assimilation of higher density polar-orbiting and geostationary satellite radiances.

Based on these encouraging results, NRL is conducting a similar study where the density of polar-orbiting satellite radiances was increased by up to a factor 2, depending upon sensor. The assimilation tests are run using the operational 4D-Var assimilation system NAVDAS-AR (NRL Atmospheric Variational Data Assimilation System-Accelerated Representer) with the Navy Operational Global Atmospheric Prediction System (NOGAPS). The operational resolution is currently T319L42, with an inner loop resolution of T119L42 and model top near 0.4 hPa. The results of extended data assimilation tests will be presented. In addition to the standard forecast skill metrics, we will present observation impact results using the adjoint technique described in Langland and Baker (2004). We will use the observation impact to evaluate whether the additional observations increase or decrease the beneficial impact for each assimilated sensor and channel. In addition, we will assess whether the impacts are located primarily in sensitive regions, or whether the impact is more broadly distributed.