Boundary layer dynamics parametrization by generative machine learning

Even though global simulations of the Earth system on monthly timescales reach now resolutions of 2.5 kilometers, essential turbulent transport processes in the lowest part of the atmosphere still have to be modeled. First, we discuss an ML-based mass-flux parametrization of the subgrid-scale heat flux for a shear-free dry atmospheric boundary layer by a generative adversarial network. Training data and prediction capability of the algorithm are increased by incorporating the physics of the growth of the convective boundary layer following from classical mixed layer similarity theory by Deardorff. Our model is compared successfully to standard mass-flux parametrizations based on an equation model and a plume model. It is additionally found to reproduce the intermittent fluctuations of the vertical velocity and the convective buoyancy flux as well as the horizontal organization of the transiently evolving turbulence correctly. Secondly, we report first steps to model the dispersion of Lagrangian tracer particles in a turbulent boundary layer by a hybrid quantum-classical generative algorithm. In detail, we apply a direct variational generator by using of the efficient sampling and encoding of high-dimensional data in quantum algorithm.