Direct measurements of the F region neutral wind field in the northern (winter) hemisphere during late November and December 1981 were obtained by using the Dynamics Explorer (DE 2) satellite and the network of ground-based Fabry-Perot interferometers sited in the North American and Scandinavian sectors. All available data for the study period have been collated and averaged according to universal time (UT) to provide a global-scale experimental determination of the mean of average circulation in the northern hemisphere F region. The data are organized into UT ''bins'' and are presented as a sequence of 12 ''climate maps'' depicting the neutral wind system at 2-hour UT intervals during a ''mean'' day in December 1981. The data set provides the most comprehensive experimental measure of the global-scale high-latitude thermospheric circulation yet reported.

The averaged measurements from the different instruments exhibit a satisfactory degree of internal consistency when viewed in terms of the global-scale neutral flow. A detailed comparison with the prediction of the NCAR thermospheric general circulation model is discussed. The model predictions are in good agreement with the UT-dependent mean circulation, reproducing the basic control of the high-latitude neutral wind system by ion drag forces modulated in UT as a result of the offset between the geomagnetic and geographic poles. Differences in flow directions and velocities between model predictions and experimental results are ascribed to natural variability in the thermosphere as well as to the use in the model of a less than fully realistic, symmetric ion convection geometry and density distribution. The two basic conclusions of this study are that (1) instrumentation deployed around the globe and on satellites can provide powerful, composite data sets that can be used to monitor global-scale mean thermospheric ''climatology'' and that (2) the mean thermospheric (F region) neutral circulation for the given geophysical situation (solar maximum, near solstice) can be modeled reasonably well using a three-dimensional, time-dependent model with appropriate parameterizations for the energy and momentum inputs.