Title: Adjoint observation impact for a limited area model

Authors: Clark Amerault (Naval Research Laboratory, Monterey, CA, USA)
James D. Doyle (Naval Research Laboratory, Monterey, CA, USA)

Many global numerical weather prediction systems employ adjoints of a forecast model and data assimilation system to monitor the impact of observations on forecast error. Recently, the adjoint observation impact methodology has been applied to the US Navy's atmospheric limited area forecast model and 3D-Var data assimilation system. This presentation will focus on aspects of the adjoint observation impact system that are unique to a limited area model. Similar to global models, large reductions in forecast error are due to conventional observations from soundings and aircraft. However, the forecast error is much more variable when compared to a global system due to weather features moving into and out of the limited area model's domain. The importance of observations relative to one another is dependent on the physical location of the model's domain. For example, when compared to radiosondes, cloud winds are more important for a domain centered over the eastern north Pacific Ocean than the continental United States. Furthermore, the area over which the forecast error is calculated must be carefully chosen so that the information is not lost to the lateral boundaries during the adjoint forecast model integration. Results from impact calculations of differing resolutions will also be presented.

GMAO Head: Michele Rienecker Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: March 1 2011