For the first time, a formal comparison is made between gravity wave momentum fluxes in models and those derived from observations. Although gravity waves occur over a wide range of spatial and temporal scales, the focus of this paper is on scales that are being parameterized in present climate models, sub-1000-km scales. Only observational methods that permit derivation of gravity wave momentum fluxes over large geographical areas are discussed, and these are from satellite temperature measurements, constant-density long-duration balloons, and high-vertical-resolution radiosonde data. The models discussed include two high-resolution models in which gravity waves are explicitly modeled, Kanto and the Community Atmosphere Model, version 5 (CAM5), and three climate models containing gravity wave parameterizations, MAECHAM5, Hadley Centre Global Environmental Model 3 (HadGEM3), and the Goddard Institute for Space Studies (GISS) model. Measurements generally show similar flux magnitudes as in models, except that the fluxes derived from satellite measurements fall off more rapidly with height. This is likely due to limitations on the observable range of wavelengths, although other factors may contribute. When one accounts for this more rapid fall off, the geographical distribution of the fluxes from observations and models compare reasonably well, except for certain features that depend on the specification of the nonorographic gravity wave source functions in the climate models. For instance, both the observed fluxes and those in the high-resolution models are very small at summer high latitudes, but this is not the case for some of the climate models. This comparison between gravity wave fluxes from climate models, high-resolution models, and fluxes derived from observations indicates that such efforts offer a promising path toward improving specifications of gravity wave sources in climate models.

Previous versions of GISS climate models have either used formulations of Rayleigh drag to represent unresolved gravity wave interactions with the model-resolved flow or have included a rather complicated treatment of unresolved gravity waves that, while being climate interactive, involved the specification of a relatively large number of parameters that were not well constrained by observations and also was computationally very expensive. Here, the authors introduce a relatively simple and computationally efficient specification of unresolved orographic and nonorographic gravity waves and their interaction with the resolved flow. Comparisons of the GISS model winds and temperatures with no gravity wave parameterization; with only orographic gravity wave parameterization; and with both orographic and nonorographic gravity wave parameterizations are shown to illustrate how the zonal mean winds and temperatures converge toward observations. The authors also show that the specifications of orographic and nonorographic gravity waves must be different in the Northern and Southern Hemispheres. Then results are presented where the nonorographic gravity wave sources are specified to represent sources from convection in the intertropical convergence zone and spontaneous emission from jet imbalances. Finally, a strategy to include these effects in a climate-dependent manner is suggested.

The 1-s-resolution U.S. radiosonde data are analyzed for unstable layers, where the potential temperature decreases with increasing altitude, in the troposphere and lower stratosphere (LS). Care is taken to exclude spurious unstable layers arising from noise in the soundings and also to allow for the destabilizing influence of water vapor in saturated layers. Riverton, Wyoming, and Greensboro, North Carolina, in the extratropics, are analyzed in detail, where it is found that the annual and diurnal variations are largest, and the interannual variations are smallest in the LS. More unstable layer occurrences in the LS at Riverton are found at 0000 UTC, while at Greensboro, more unstable layer occurrences in the LS are at 1200 UTC, consistent with a geographical pattern where greater unstable layer occurrences in the LS are at 0000 UTC in the western United States, while greater unstable layer occurrences are at 1200 UTC in the eastern United States. The picture at Koror, Palau, in the tropics is different in that the diurnal and interannual variations in unstable layer occurrences in the LS are largest, with much smaller annual variations. At Koror, more frequent unstable layer occurrences in the LS occur at 0000 UTC. Also, a “notch” in the frequencies of occurrence of thin unstable layers at about 12 km is observed at Koror, with large frequencies of occurrence of thick layers at that altitude. Histograms are produced for the two midlatitude stations and one tropical station analyzed. The log–log slopes for troposphere histograms are in reasonable agreement with earlier results, but the LS histograms show a steeper log–log slope, consistent with more thin unstable layers and fewer thick unstable layers there. Some radiosonde stations are excluded from this analysis since a marked change in unstable layer occurrences was identified when a change in radiosonde instrumentation occurred.