Simulation of Vertically-Propagating Waves in Comprehensive General Circulation Models: Opportunities for Comparison With Observations

A number of important developments in observations of the middle atmosphere have been discussed at this meeting. Among the most exciting are the display of preliminary data from the Doppler radiometer instruments on UARS, the development of active optical methods for determining Doppler winds throughout the entire height range of the middle atmosphere, and the deployment of some nearly identical MF radars in an array extending over several hundred km. While such developments are extremely promising, the discussion at this meeting has highlighted the limitations of practical observations in producing a characterization of the full three-dimensional, time-dependent flow field in the middle atmosphere. Each method has significant limitations in terms of geographical and/or temporal coverage as well as vertical and/ or horizontal resolution. In addition, there was much concern expressed about the magnitude of the discrepancies that have been observed when the various radar and optical methods are intercompared as directly as possible. These problems all suggest a role for three-dimensional numerical models in assessing and interpreting the available observations. One obvious use of the models is to quantify the influence of the spatial and temporal sampling inherent in the different observation methods. Another possible application is in four-dimensional data assimilation, i.e. effectively inserting observations as they become available into a GCM integration (Rood et al., 1989; Swinbank and O'Neill, 1994). A prerequisite for such applications is a numerical model that produces a reasonably realistic simulation of the middle atmospheric wind field, even at relatively short space and time scales. In this brief paper the simulation obtained by the GFDL "SKYHI" troposphere-stratosphere-mesosphere model will be reviewed with an emphasis on the vertically-propagating gravity waves that account for the variance over much of the atmospheric spectrum. The focus will be on examining the model results presented in formats similar to those typically produced from available observations.