Harnik, N., and R. S. Lindzen, 2002. Are TOVS temperature retrievals
capable of resolving the vertical structure of stratospheric planetary
waves? Submitted to Annales Geophysicae.


This study examines how the main limitations of satellite retrievals, which are coarse vertical resolution and a limited vertical domain, affect the observations of stratospheric planetary wave structure and evolution. The approach is to calculate radiances measured by a virtual idealized satellite equipped with TOVS and sitting at the top of a linear model with stratospheric planetary waves, and invert them to obtain a retrieved field. The retrieved and model waves are then compared. A simplified version of the operational retrieval method is used, where the inverse solution is combined with additional information (a constraint). Many different constraints are used to understand the effect of the additional information, in particular vertical correlations, on the retrievals.

It is found that while TOVS retrievals are capable of resolving the gross amplitude and phase of the waves, as well as their overall vertical propagation direction, they are only marginally able to resolve upward propagating waves from waves that are not purely upward propagating, and are not able to distinguish between the various processes that hinder vertical propagation, like partial reflection from a turning surface, or absorption at a critical surface. The retrievals are also not sensitive to wave damping. Looking at time evolution, the retrievals can capture growth or decay of the waves, and temporary downward reflection, but the detailed onset of these processes is not resolved. The form of the constraint significantly affects the retrievals above 37km, and when the waves have very sharp features, also below that. This may have implications for observations of wave mean flow interactions at a critical surface. The implications for real TOVS observations, both for NESDIS retrievals and directly assimilated radiances, and the relevance to ATOVS, are discussed.