Weird grid signature in coupled runs

[simone-silvestri]

In #1630 we are coupling a relatively coarse ocean (50km res) with a very coarse atmosphere (400km res).
After running for 1 day of simulation time, we are seeing grid imprinting in the ocean that supposedly are caused by the same grid artifacts appearing in the atmosphere's wind fields.

Here is a plot of the zonal velocity at the first near-surface grid cell

The grid signature is quite visible.

This causes the same imprint on the ocean surface:

Any idea of the cause?
Maybe it is a problem that disappears with increasing resolution?

@szy21 @dennisYatunin

[juliasloan25]

This is how we get u_int and v_int in the coupler (here), which is what's plotted above. CA.projected_vector_data should convert them both to physical units, so after that we don't expect to see grid imprinting right?

function Interfacer.get_field(sim::ClimaAtmosSimulation, ::Val{:u_int})
    # NOTE: This calculation is copied from ClimaAtmos (and is allocating! Fix me if you can!)
    int_local_geometry_values =
        CC.Fields.level(CC.Fields.local_geometry_field(sim.integrator.u.c), 1)
    int_u_values = CC.Spaces.level(sim.integrator.p.precomputed.ᶜu, 1)
    return CA.projected_vector_data.(CA.CT1, int_u_values, int_local_geometry_values)
end
function Interfacer.get_field(sim::ClimaAtmosSimulation, ::Val{:v_int})
    # NOTE: This calculation is copied from ClimaAtmos (and is allocating! Fix me if you can!)
    int_local_geometry_values =
        CC.Fields.level(CC.Fields.local_geometry_field(sim.integrator.u.c), 1)
    int_u_values = CC.Spaces.level(sim.integrator.p.precomputed.ᶜu, 1)
    return CA.projected_vector_data.(CA.CT2, int_u_values, int_local_geometry_values)
end

[szy21]

This looks like cubed sphere imprinting, which is always there I think, but I don't remember seeing anything this significant. Does it get better if you run it for longer or use a higher resolution?

[juliasloan25]

This looks like cubed sphere imprinting, which is always there I think, but I don't remember seeing anything this significant. Does it get better if you run it for longer or use a higher resolution?

Is the imprinting always there? I thought after projecting to physical space there shouldn't be imprinting.

We'll try running it for longer

[szy21]

This looks like cubed sphere imprinting, which is always there I think, but I don't remember seeing anything this significant. Does it get better if you run it for longer or use a higher resolution?

Is the imprinting always there? I thought after projecting to physical space there shouldn't be imprinting.

We'll try running it for longer

The imprinting is a common problem for cubed-sphere dycore, but it should be much weaker. I just looked at some (higher resolution) amip results and didn't see a pattern like this, so maybe there is something wrong?