Online Learning in Idealized Ocean Gyres

Abstract Ocean turbulence parameterization has principally been based on process‐based approaches, seeking to embed physical principles so that coarser resolution calculations can capture the net influence of smaller scale unresolved processes. More recently there has been an increasing focus on the application of data‐driven approaches to this problem. Here we consider the application of online learning to data‐driven eddy parameterization, constructing an end‐to‐end automatically differentiable dynamical solver forced by a neural network (NN), and training the NN based on the dynamics of the combined hybrid system. This approach is applied to the classic barotropic Stommel‐Munk gyre problem—a highly idealized configuration which nevertheless includes multiple flow regimes, boundary dynamics, and a separating jet, and therefore presents a challenging test case for the online learning approach. It is found that a NN which is suitably trained can lead to a coarse resolution NN parameterized model which is stable, and has both a reasonable mean state and intrinsic variability. This suggests that online learning is a powerful tool for studying the problem of ocean turbulence parameterization. A test of generalizability with a modified wind forcing shows some positive results. However a test of symmetry preservation demonstrates that the NN parameterized model fails to respect an intrinsic symmetry property of the underlying system.