Coping With Model Errors in Data Assimilation

Istvan Szunyogh*
Texas A&M University

Data assimilation schemes must be robust to errors in the model of the dynamics, the model that defines the observation function, and the statistical model that implements the spatiotemporal interpolation of the observed information. The schemes can be made robust by a combination of a variety of techniques, which include correcting the background for model errors, modifying the observation function to account for model errors, altering the observation error statistics to make the schemes resilient to errors in the observation function and statistical outlier observation errors, and altering the background error statistics to make the schemes robust to statistical outlier background errors. I will show examples from my research with different collaborators for the application of these techniques. These examples will include, but may not be limited to, recent results on the assimilation of tropical cyclone (TC) observations and coupled global-limbed-area data assimilation.

*Preference: Oral