The two-dimensional transport of conserved tracers by planetary waves results from two processes: dispersion and organized nondispersive advection. For nonconserved quantities an additional transport mechanism is the chemical eddy effect. This paper is concerned with the latter, which depends on spatial variation in the production and loss of a trace species and can occur in the absence of dispersion. A simple and commonly used form of representing the eddy flux is as a transport matrix multiplied by the mean gradient. Such a matrix is often called a diffusion matrix, although with off-diagonal components it can represent transport across, as well as parallel to, the mean gradient. The two-dimensional transport matrix in this form can, with the assumptions of linearity, stationarity, and a simple form of production and loss, be derived from the eddy tracer continuity equation. We have computed the chemical eddy contribution to the components of the matrix from LIMS satellite observations, using specified photochemical damping time scales. The dominant (Kyy) component of the transport matrices for several winter months are given for

The ozone, nitric acid, and quasi-geostrophic potential vorticity (PV). The parameterized transports for these three species are compared with the ''exact'' transports, computed directly from the eddy LIMS data. These results indicate that the chemical eddy effect can account for most of the observed ozone transport in early winter, decreasing to less than half in late winter. The agreement between the parameterized and observed nitric acid and PV fluxes is not as good. In the lower stratosphere the observed eddy fluxes (the ''exact'' fluxes) for ozone and nitric acid exceed the parameterized chemical eddy fluxes by a substantial margin. This may be because dispersion, which is not included in the parameterization, is the primary mechanism for eddy transport in that part of the atmosphere. Other reasons that might explain the discrepancies between the parameterized chemical eddy fluxes and the observed fluxes are the inability of a simple flux-gradient relation to characterize the transport, the approximations made in the calculation of the transport matrix, the inaccuracies in the specified damping times, or the problems with the observed (''exact'') eddy fluxes, which can be sensitive to errors in the wave data.