UNIVERSITY OF COLORADO AT DENVER

PLACE: Mathematics Conference Room 626 UCD Building, 1250 14th St., Denver

TIME: NOON (Refreshments served at 11:45 am)

DATE: February 21, 2000

Error analysis for photochemical models and its applications.

Boris V Khattatov, Lawrence Lyjak, John Gille
ACD, NCAR, P. O. Box 3000, Boulder, CO 80307

Traditionally, photochemical numerical models are used to simulate concentrations
of atmospheric chemicals but they do not provide estimates of errors of such
simulations, even when initialized by observed concentrations. We we will describe
an application of the Kalman filter technique to a photochemical box model that
provides mathematically rigorous estimates of errors of model simulations.
Additionally, off-diagonal elements of the error covariance matrices, or
correlations between errors of different chemicals are also computed.

Analysis of linearization matrices arising in the Kalman filter method led to an
interesting
finding concerning observability of the stratospheric photochemical
system. These matrices become severely rank deficient after a few hours of integration.
In practice, this finding means that concentrations of only a small subset of active
chemicals in the system are necessary to completely predict time evolution of the
system for the next few days.