A posteriori learning of quasi-geostrophic turbulence parametrization: an experiment on integration steps
This addresses the long-standing problem of subgrid-scale parametrization for ocean, atmosphere, and climate predictions, which is incremental as it builds on existing neural network approaches but focuses on the specific challenge of two-dimensional flows with energy backscatter.
The paper tackled the challenge of modeling subgrid-scale dynamics in quasi-geostrophic turbulence, where direct numerical simulation is infeasible, by learning a model jointly with the dynamical solver using an a posteriori-based loss function, resulting in stable and realistic simulations.
Modeling the subgrid-scale dynamics of reduced models is a long standing open problem that finds application in ocean, atmosphere and climate predictions where direct numerical simulation (DNS) is impossible. While neural networks (NNs) have already been applied to a range of three-dimensional flows with success, two dimensional flows are more challenging because of the backscatter of energy from small to large scales. We show that learning a model jointly with the dynamical solver and a meaningful \textit{a posteriori}-based loss function lead to stable and realistic simulations when applied to quasi-geostrophic turbulence.