Abstract We analyze the Lorenz Energy Cycle of the ocean simulated by FESOM on several unstructured meshes, with nominal horizontal resolution ranging from 1/4 to 1/12°. Some meshes are more strongly refined in regions where the eddy kinetic energy (EKE) is large, whereas others follow a standard design, with resolution gently increasing toward high latitudes. We show that, except for the coarsest (1/4°) mesh, the generation of kinetic energy (KE), the generation of available potential energy (APE), and the conversion between mean KE and APE are insensitive to resolution and mesh design. EKE and the conversions between mean and eddy APE, and between eddy APE and KE, show a weak but systematic increase with resolution. The EKE‐based mesh design yields specific advantages at the highest resolutions. However, the choice of absolute versus relative winds has a stronger effect than the differences noted above. It causes several times larger changes in EKE generation and alters the EKE level as much as mesh resolution does. We also describe the implementation of the KE compartment of the Lorenz Energy Cycle diagnostics in FESOM. It is based on online estimates of the power associated with each tendency term in the momentum equation. This procedure eliminates interpolation and reveals the sources of numerical dissipation.
Koldunov et al. (Wed,) studied this question.