Seminar: Resolution-Dependent Restoration of Baroclinic Perturbation Growth by Dynamic Kinetic Energy Backscatter: A Breeding Analysis in FESOM2

Abstract

Ocean models at eddy-permitting resolution dissipate too much kinetic energy near the grid scale. As a result, they underestimate baroclinic instability, the process by which mesoscale eddies extract energy from the large-scale flow. Dynamic kinetic energy backscatter schemes return part of this lost energy to the resolved flow and improve simulated eddy energy levels. It has remained open whether they also restore the instability itself, or only add energy.

We address this question with breeding, a technique from ensemble weather forecasting that isolates the fastest-growing perturbations of an evolving flow. We apply it to an idealized baroclinic channel simulated with the unstructured-mesh ocean model FESOM2 at 20 km and 10 km resolution, and compare the results with an eddy-resolving 5 km simulation without backscatter. Over 50 thirty-day breeding cycles, we measure how efficiently perturbations convert available potential energy into kinetic energy.

Without backscatter, both coarse resolutions strongly underestimate perturbation growth. Their baroclinic conversion rate reaches only 2% (20 km) and 13% (10 km) of the reference value. Backscatter strengthens growth at both resolutions, but through different mechanisms. At 20 km, where the deformation radius is not resolved, it raises the peak perturbation kinetic energy by about 40% while the conversion rate stays nearly unchanged: it adds energy without strengthening the instability. At 10 km, it increases the conversion rate by a factor of 2.2–2.8, to about one third of the reference, and perturbations keep growing throughout the cycle as they do at 5 km. Smoothing the backscatter tendency is essential to avoid spurious grid-scale growth, while the number of smoothing passes matters less.

Backscatter can therefore partially restore baroclinic instability, but only once the deformation radius is resolved. Breeding also offers a practical way to evaluate eddy parameterizations by their effect on the dynamics, not only on energy levels.