Variational Cloud-clearing with CrIS data at NCEP

The variational assimilation of satellite radiance observations is a major contributor to the forecast skill of the Global Forecast System (GFS) at NCEP. However, satellite observations are underused at most meteorologically important areas due to the presence of cloud. For infrared (IR) channels, only channels unaffected by cloud have been assimilated in the Grid-point Statistical Interpolation system (GSI): the NCEP operational data assimilation system. This limits the ability to continue improving the model initial conditions through assimilating radiance data from partly cloudy and cloudy regions.

NCEP has begun to develop a modified “cloud-clearing” algorithm to use the data in partly cloudy areas by removing cloud radiative effects. With the assumption that the surface, atmospheric state and cloud formation characteristics are the same within one field of regard and only cloud fraction varies among adjacent pixels, the clear-column radiances can be reconstructed by minimizing the difference between the radiances with cloud radiative effects removed and simulated clear-sky radiances from radiative transfer model using first guess. The above process differs from other existing cloud-clearing systems in that it is integrated to the GSI system itself. The reconstructed clear-column radiances are estimated in each GSI outer loop and then assimilated together with all other observations in the inner loop. The impact of the cloud-clearing radiances on the global analysis and forecast skills will be evaluated and preliminary results using the CrIS data will be reported at the conference.