Abstract
The NASA/Goddard three-dimensional chemistry and transport model is driven by winds from a stratospheric data assimilation system. Synoptic- and planetary-scale patterns, apparent in satellite observations of trace constituents, are successfully reproduced for seasonal integrations. As model integrations proceed, however, the quality of simulations decreases, and systematic differences between calculation and measurement appear. The differences are explained by examining the zonal-mean residual circulation. The vertical residual velocity w̄* is calculated two ways: (i) from the diabatic heating rates and temperature tendency and (ii) from the Eulerian vertical velocity and the horizontal eddy heat flux convergence. The results from these calculations differ substantially. Periodic insertion of observational data during the assimilation process continually shocks the general circulation model and produces these differences, which leads to an overestimate of the mean vertical heat and constituent transport. Such differences are expected to be general to all data assimilation products. This interpretation is corroborated by two-dimensional (2D) model calculations. When w̄* is calculated from (ii), the 2D ozone evolution is unrealistic and qualitatively similar to the 3D model simulation. The 2D ozone evolution is reasonable when w̄* is calculated from (i).
