Their amplitude grows as they propagate vertically and they impact the middle atmosphere circulation when they break and/or dissipate, even if they have relatively small amplitudes at the source level. The spatial scales of GWs are generally too small to be resolved by general circulation models (GCMs), and the generation, propagation and dissipation of these waves need to be parameterized for GCMs to produce a reasonable circulation. Such parameterizations were first introduced in the 1980s in models with a barely resolved stratosphere, and only high-amplitude orographically forced GWs were needed to be taken into account (Palmer et al., 1986). Nowadays many models resolve the middle atmosphere, requiring the parameterization of the effects of smaller-amplitude, non-orographic gravity waves (Manzini et al., 1997).One way to parameterize non-orographic GWs consists in using the observational evidence of "universal" spectra built over a large number of realizations of GW fluctuations of vertical wind and temperature in the middle atmosphere, which shape is derived from radiosondes and satellite data (e.g., Cot, 2001;Zhao et al., 2017). These spectra are numerically robust (Brethouwer et al., 2007;Lindborg, 2006), and various theories have been developed to explain them. Some involve wave breaking (e.g., Dewan & Good, 1986), other include nonlinear