Predictability and mean state of the MLT: Importance of resolving gravity waves

The circulation in the Mesosphere / Lower Thermosphere (MLT) region is strongly influenced by atmospheric gravity waves that propagate upward from the lower atmosphere. So far, most global models of the MLT have to parameterize gravity waves, given horizontal model resolution on the order of 100 km. It becomes increasingly clear that the simplified approximations of gravity wave parameterizations, including their inability to simulate gravity wave generation within the middle atmosphere, are a cause for biases in the simulation of MLT circulation, holding back scientific progress in understanding, predicting and projecting MLT circulation.

In this study, the extended German Weather and Climate model UA-ICON is used to demonstrate the effects of moving from a coarse model resolution to a gravity-wave permitting resolution on the simulation of the mean state of the MLT and its predictability. An episode of austral winter to spring is simulated with two UA-ICON set-ups, one with about 160 km horizontal grid spacing and 120 vertical levels from the ground to 150 km height (“coarse resolution”), and one with about 20 km horizontal grid spacing and 250 vertical levels (“high resolution”). The high-resolution set-up is able to resolve gravity waves with horizontal wave length up to about 200 km. Resolving gravity waves is essential to simulate the mean state of MLT circulation in austral winter: while in the coarse resolution model, zonal mean winds around 100 km height are easterly, the high-resolution model version simulates westerlies in this region, in agreement with observations. It is shown that wave forcing by resolved waves with horizontal scales below 2000 km, which are only resolved in the high-resolution model version, impose an eastward force on the zonal mean winds, and thus are essential to maintain the westerly winds. Next to the mean state, the two model set-ups are utilized to demonstrate the effects of resolving gravity waves on estimations of the intrinsic predictability of the MLT region: experiments with imposed small perturbations in the initial conditions show that error growth in the MLT region is substantially faster in the high-resolution simulation with resolved gravity waves compared to the coarse resolution simulation. Thus, the intrinsic predictability time-scale, after which the MLT becomes intrinsically unpredictable, is vastly overestimated by a factor of 3-4 in simulations that do not resolve gravity waves. Overall, this work stresses the importance of exploring high-resolution simulations of the MLT in order to make progress on our understanding of MLT dynamics.