Approximation and inversion of a complex meteorological system via local linear filters

Dynamical models involving systems of numerous differential equations are c ommonly used to describe meteorological behavior. Approximations to such sy stems are often desired, particularly for inversion problems when the model 's adjoint is unavailable. A method for approximating a general, nonlinear system is explored here. The method is particularly useful for inverse prob lems. An application to a complex multivariate dynamic model for southern C alifornia air quality is given, and the method is shown to provide satisfac tory estimates of vehicle emission-rate inputs using ozone concentration es timates for the Los Angeles basin.