The High-Resolution Doppler Imager on the Upper Atmosphere Research Satellite has observed the O2 atmospheric band nightglow at extremely high spatial resolution of 50 km in the horizontal over much of Earth's surface. These observations of the oxygen nightglow are shown to be a good surrogate for atomic oxygen outflow from the thermosphere and subsequent recombination in the upper mesosphere. The distribution of enhanced airglow and recombination verify the previous observations that bright regions of the nightglow are associated with descending motion, where the tides are transporting atomic oxygen out of the thermosphere, and regions of decreased airglow are associated with ascending tidal motion. The detailed observations of the global airglow at a scale of 100 km present a much more complex picture, however. The tidally enhanced regions are extremely structured, with strong longitudinal and latitudinal variability in brightness and associated atomic oxygen recombination. These structured regions are persistent on a timescale of a day or more, with certain regions of the globe exhibiting enhanced airglow much of the time. These regions are present over the entire night side of the globe but are strongly modulated by the ascending and descending tidal motions. A theoretical study of these enhancements suggests that they are caused by the upward propagation of gravity waves, which break in the high mesosphere and induce strong vertical motion in the atmosphere. It is interesting that the primary cause of the increased transport of atomic oxygen is vertical motion rather than the conventional enhanced turbulence explanation. These gravity waves are associated with wind shear and strong convective activity in the troposphere and often appear in certain preferred locations where convective activity is expected. These observations have important implications for the basic processes by which momentum and energy are deposited in the upper mesosphere. We suggest that the processes must be modeled as stochastic events that dance around the base of the thermosphere, generally in climatologically preferred regions, rather than the relatively uniform distribution of gravity wave forcings that is being used in general circulation models up to this time.