The Quasi-Biennial Oscillation (QBO) of zonal-mean zonal winds in the tropical stratosphere is a conspicuous feature of the Earth climate. Its influence virtually extends to the whole atmosphere. The QBO is driven by the dissipation of waves generated by deep convection in the tropics. Yet, a significant fraction of these waves have spatial scales that are not explicitly resolved in current climate models, so that the evolution of the QBO in a changing climate still remains unclear. A study by \citeA{Vincent2020} shows that recent long-duration balloon observations, combined with information on the vertical structure of the tropical atmosphere, could contribute to provide significantly improved constraints on wave parameterization in models.

Tropical gravity waves that are generated by convection are generally too small in scale and too high in frequency to be resolved in global climate models, yet their drag forces drive the important global-scale quasi-biennial oscillation (QBO) in the lower stratosphere, and models rely on parameterizations of gravity wave drag to simulate the QBO. We compare detailed properties of tropical parameterized gravity waves in the Whole Atmosphere Community Climate Model Version 6 (WACCM6) with gravity waves observed by long-duration super-pressure balloons, and also compare properties of parameterized convective latent heating with satellite data. Similarities and differences suggest that the WACCM6 parameterizations are excellent tools for representing tropical gravity waves, but the results also suggest detailed changes to the gravity wave parameterization tuning parameter assumptions that would bring the parameterized waves into much better agreement with observations. While WACCM6 currently includes only non-stationary gravity waves from convection, addition of the component that is stationary relative to convective rain cells is likely to improve the simulation of the QBO in the model. The suggested changes have the potential to alleviate common biases in simulated QBO circulations in models.