Oceananigans vs MITgcm comparison in the GCM hydrostatic mode

[simone-silvestri]

@jm-c and I have recently set up a latitude-longitude global ocean simulation at a 1-degree resolution to compare results from Oceananigans and MITgcm.

The setup is the following:

• 48 vertical levels taken from the ECCO2 version 4 grid
• Horizontal laplacian viscosity and diffusivity with values $\nu_h = 50000 \ m^2/s$ and $\kappa_h = 300 m^2/s$ , respectively
• Vertical laplacian viscosity $\nu_v = 10^{-4} m^2/s$ increased to $10^{-2} m^2/s$ in the top 50 meters
• Vertical laplacian diffusivity $\kappa = 10^{-5} m^2/s$
• Linear bottom drag with coefficient $\mu = 10^{-3}$
• Centered advection schemes for tracers
• Energy-conserving vector invariant momentum advection
• Linear equation of state
• No GM or Redi diffusivities

we evolve for 3 months starting from the same initial condition with the same forcing. The results in terms of $$T^{\text{Oceananigans}} - T^{\text{MITgcm}}$$
is shown below. The left plot shows the surface, while the right shows a transect of the Pacific ocean.

[glwagner]

The key difference between the two is the treatment of vorticity on horizontal immersed boundaries correct?

[simone-silvestri]

Yes. MITgcm sets vorticity to zero on peripheral_nodes, while we do not. So we end up having a vorticity at corners (vorticity on flat edges will still be zero)