Atmospheric gravity waves are key drivers of the transfer of energy and momentum between the layers of the Earth’s atmosphere. The accurate representation of these waves in General Circulation Models (GCMs) however has proved very challenging. This is because large parts of the gravity wave spectrum are at scales that are near or below the resolution of global GCMs. This is especially relevant for small isolated mountainous islands such as South Georgia (54° S, 36° W) in the Southern Ocean. Observations reveal the island to be an intense source of stratospheric gravity waves, but their momentum fluxes can be under-represented in global models due to its small size. This is a crucial limitation, since the inadequate representation of gravity waves near 60° S during winter has been linked to the long-standing "cold-pole problem", where the southern stratospheric polar vortex breaks up too late in spring by several weeks. Here we address a fundamental question: when a model is allowed to run at very high spatial resolution over South Georgia, how realistic are the simulated gravity waves compared to observations? To answer this question, we present a 3-D comparison between satellite gravity wave observations and a high resolution model over South Georgia. We use a dedicated high-resolution run (1.5 km horizontal grid, 118 vertical levels) of the Met Office Unified Model over South Georgia and coincident 3-D satellite observations from NASA AIRS/Aqua during July 2013 and June–July 2015. First, model winds are validated with coincident radiosonde observations. The AIRS observational filter is then applied to the model output to make the two data sets comparable. A 3-D S-transform method is used to measure gravity-wave amplitudes, wavelengths, directional momentum fluxes and intermittency in the model and observations. Our results show that although the timing of gravity wave activity in the model closely matches observations, area-averaged momentum fluxes are generally up to around 25 % lower than observed. Further, we find that 72 % of the total flux in the model region is located downwind of the island, compared to only 57 % in the AIRS measurements. Directly over the island, the model exhibits higher individual flux measurements but these fluxes are more intermittent than in observations, with 90 % of the total flux carried by just 22 % of wave events, compared to 32 % for AIRS. Observed gravity wave fluxes also appear to dissipate more quickly with increasing height than in the model, suggesting a greater role for wave-mean flow interactions in reality. Finally, spectral analysis of the wave fields suggests that the model over-estimates gravity wave fluxes at short horizontal scales directly over the island, but under-estimates fluxes from larger horizontal scale non-orographic waves in the region, leading to a lower average value overall. Our results indicate that, although increasing model resolution is important, it is also important to ensure that variability in the background wind vector and role of non-orographic waves are accurately simulated in order to achieve realistic gravity wave activity over the Southern Ocean in future GCMs.
</jats:p>
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)