Satellite radiance data is known to be biased, both absolutely and relative to any given forecast model. There are two main sources of bias relative to the forecast model: scan bias and air-mass bias. Scan bias derives from the instrument itself as it views the atmosphere at different zenith angles. Air-mass bias derives from the fast (i.e. approximate) radiative transfer code of the forecast model, and depends on the temperature and moisture characteristics of the underlying atmosphere. Our goal is to find and implement the best possible bias correction scheme for satellite radiance assimilation.

The previous bias correction scheme used at NRL for the NOGAPS global forecast model was based on Eyre, 1992. Recently NRL has implemented a scheme based on Harris & Kelly, 1999. Both schemes have been independently reproduced in Matlab, a development environment that allows easy customization, enhancement, and generalization of these methods. New types of data visualization have highlighted the need for more sophisticated quality control, and demonstrated that commonly used statistical measures such as the mean and variance are, in some cases, inappropriate. The spatial and temporal characteristics of the residual bias will aid us in the development of improved bias correction methods.

Data assimilation and resulting weather forecasts can be improved by superior bias correction of satellite radiances. To achieve this, we have built a general, modular bias correction system in Matlab. The flexibility and generality of the new software will enhance the rapid assimilation of new data types from modern instruments such as hyperspectral sounders.