Stratospheric gravity waves in high-resolution global models

As we enter the age of exascale computing, more and more global scale simulations with horizontal grid spacings in the range of 1-10 km become available. Yet, not the full spectrum of gravity waves (GWs) is resolved and new challenges emerge, some of which are directly linked to the representation of convection, which is only partially resolved, but the most important source of GWs in the tropics.

Unlike most climate models that use parameterizations for GWs, the DYAMOND simulations reproduce detailed, satellite-observed features of the global GW momentum flux (GWMF) distribution including the zonal mean. This can be attributed to realistic GWs from convection, orography and storm tracks. Yet, the GWMF magnitudes differ substantially among simulations. Differences in the strength of convection may help explain differences in the GWMF between simulations of the same model in the summer low latitudes where convection is the primary source. For ICON, simulations with explicit convection show 30-50% larger zonal-mean momentum fluxes in the summer hemisphere subtropics than simulations with parameterized convection. Explicit convection is associated with stronger updrafts and GW sources.

Since any kind of observations can only see a fraction of the GW spectrum, we also analyzed the spectra of the horizontal motions associated with inertia GWs and Rossby waves, respectively. A fundamental characteristic of the atmosphere is the distribution of wave energy across different horizontal scales. Observations and numerical modelling have supported the idea of a canonical energy spectrum. Horizontal kinetic energy scales with the horizontal wavenumber k as k**-3 at large scales with a transition towards k**-5/3 at mesoscales.

The simulations produce the expected canonical shape of the spectra, which is encouraging given that some models are stripped down to a minimum set of physical parameterizations. Yet, total energy levels, spectral slopes at sub-synoptic scales, and spectral crossing scales differ significantly. The contribution of inertia GWs to the total wave energy differs by factors of up to two between the simulations. The crossing scales between the inertia GW and Rossby wave spectra also differ by a factor of about two between the simulations and depend mostly on the ratio of integrated wave energies, rather than on spectral slope or details of the spectral shape. The spectra exhibit little variability in time and can serve as an almost instantaneous diagnostic.