The analysis and forecast domain of the real-time experimental Rapid Refresh (RR) system (scheduled to replace the NCEP operational Rapid Update Cycle in 2010) covers all of North America, a significant expansion compared to the CONUS domain coverage of the Rapid Update Cycle (RUC). Associated with this domain expansion is the necessity to include assimilation of satellite radiance data in the RR 1-h data assimilation cycle. This is being accomplished for the RR through use of the Gridpoint Statistical Interpolation (GSI) package. GSI has been developed by NCEP in cooperation with other institutions and already applied successfully in both global and regional operational systems. Regional satellite radiance assimilation is not a fully developed field and there are several problems related to the availability, accuracy, and use of satellite radiance data. One of the major problems is the bias in satellite radiances and the related bias correction should take into account the most important sources of biases. These sources of bias include small systematic changes in satellite orbits, instrument biases (such as calibration problems), and inaccuracies in instrument spectral functions, etc. Preprocessing can also cause biases through non-complete cleaning of radiances and surface effects may also contribute to biases. An additional source of bias is the radiative transfer model and the use of simplifying assumptions about the transfer of radiation. Because uncorrected biases can ruin the numerical forecasts, bias correction is a very important step in satellite data assimilation. It is worth noting that background field biases also play a role in radiance increments. In this presentation, we will provide a short description of satellite data types used by the RR system, including information about the number of available radiance observations for a one hour data assimilation cycle. A short review will then be given concerning the performance of the Community Radiative Transfer Model (CRTM) in the RR version of GSI. The regression based bias correction techniques used in the RR (based on the method used at NCEP) will then be discussed. Finally, the impact of satellite radiance data on RR analyses and forecasts (as revealed by retrospective assimilation experiments) will be presented.
This publication was presented at the following: