P3.15
The Assimilation of Satellite Observations with the NRL Atmospheric Variational Data Assimilation System (NAVDAS)
Nancy L. Baker, NRL, Monterey, CA; and R. Daley, S. D. Swadley, J. Clark, E. H. Barker, J. S. Goerss, and K. Sashegyi
This poster discusses the use of satellite observations in the Navy's new three-dimensional variational system NAVAS (NRL Atmospheric Variational Analysis System). NAVDAS is designed to replace the operational multivariate optimum interpolation analysis system, and to satisfy both the global and regional atmospheric data assimilation requirements of Fleet Numerical Meteorology and Oceanography Center.
NAVAS is capable of assimilating several different types of satellite observations, including TOVS and ATOVS radiances and retrievals, SSM/I wind speed and total precipitable water, cloud drift and water vapor winds, and scatterometer winds. This poster will focus on the unique aspects of satellite observation assimilation in NAVDAS.
For polar-orbiting sounders such as ATOVS, NAVDAS is designed to assimilate radiances, but can also assimilate retrievals provided by either outside agencies such as NESDIS, or in-house one-dimensional variational algorithms. The retrievals and radiances are projected onto a subset of the gravest vertical modes of the eigenvectors of the background error correlation matrix. This eigenvector projection provides an efficient data compression technique. The radiance assimilation uses a linearized (instead of nonlinear) forward operator, which introduces several important implications that will be discussed in this poster.
SSM/I wind speeds may be assimilated using either a nonlinear or linearized forward operator. These methods are compared to the ad hoc, traditional method of assigning background wind directions from the background field.
Finally, the NAVDAS adjoint has also been developed. The adjoint of NAVDAS is used to compute the sensitivity of a specified forecast aspect (such as forecast error) to the observations and background. The TOVS radiance sensitivity highlights the importance of properly assigning the observation errors, and the effects of data density on the analysis.
Poster Session 3, Operational Applications
Tuesday, 16 October 2001, 9:15 AM-11:00 AM
Previous paper Next paper