DIFFER Seminar: (Quantum) Machine Learning for Atmospheric Turbulence and PDEs

Abstract: Earth system models typically have coarse horizontal resolutions of the order of tens of kilometers, which do not resolve many of the processes occurring in the Earth system, such as turbulence in the atmospheric boundary layer. These processes are typically taken into account via parameterizations, which represent the coarse-scale influence of a given subgrid-scale process on the Earth system model’s state variables. Conventional parameterizations cause errors and biases in the models, which could be avoided with improved data-driven methods. In this talk, I present recent advances in applying classical and quantum machine learning techniques to model fluxes due to atmospheric turbulence in the boundary layer, to improve the turbulent closure in large eddy simulations, and to solve partial differential equations (PDEs), with the goal of taking advantage of the current rapid progress of computational power, data availability and quantum computers.