The solution of large scale inverse problems in
atmospheric chemical data assimilation

Adrian Sandu

Computer Science Department
Virginia Polytechnic Institute and State University

The task of providing an optimal analysis of the state of
the atmosphere  requires the development of novel
computational tools  that facilitate an efficient
integration of observational  data into models. We discuss
several new tools developed for the assimilation of chemical
data into atmospheric chemical transport models. The
distinguishing feature of these models is the presence of
stiff chemical interactions.

This talk introduces the four dimensional variational
(4D-Var) method for data assimilation. We present
(surprising) properties of discrete adjoints of several well
known time stepping methods and of numerical schemes for
solving the advection equation. The discrete adjoints  are
used to implement data assimilation systems based on state
of the art chemical transport models.

Results are shown from the application of these
computational tools to assimilate real data sets in regional
and global atmospheric simulations.