North pole imprint in sea ice

[simone-silvestri]

While running a coupled ocean-sea ice simulation, I noticed that there is a rather strong imprint of the north pole on the solutions, both on a RotatedLatitudeLongitudeGrid and on a TripolarGrid. To be clear, this is sea-ice thickness.

arctic_ice.mp4

At the beginning, I thought that it might be a problem connected to how we are treating the polar BC on a LatitudeLongitudeGrid, but now I think it is more of an issue of how we perform vector rotation.
An example of the zonal wind from JRA55 before and after interpolating to the TripolarGrid shows a quite clear north pole imprint:

Before:

After:

The fact that the north pole is visible points to an error for me, but I would like to know what people think. Is this even an error, or is it like the atmosphere looks?

[simone-silvestri]

This is the same interpolation performed on a RotatedLatitudeLongitudeGrid centered at the north pole with u velocities on the left side and v velocities on the right side.
In order the rows are

• first row: original JRA55
• second row: interpolated JRA55 without vector rotation (i.e. on an extrinsic frame)
• third row: interpolated JRA55 with vector rotation (i.e. on an intrinsic frame)

The second row shows already the north pole imprinting, hinting at a BC problem in the lat-lon JRA55 data. However, even the vector rotation seems weird. I would expect one big eddy to appear over the north pole, given that the RotatedLatitudeLongitudeGrid is quite regular in terms of grid distortions.

[simone-silvestri]

This is what we get when we compute vorticity for the interpolated and rotated velocities. I think this is clearly wrong, not only at the pole though, all around the pole

Interpolating directly the vorticity computed on the atmosphere grid leads to

[simone-silvestri]

Same thing for divergence, left -> atmospheric data, center -> computed from interpolated velocities, right -> interpolated from atmospheric divergence

[simone-silvestri]

And finally the same plots for kinetic energy, which (at least!) do not change between the one computed from interpolated velocities and the interpolated one

[simone-silvestri]

The rotation matrix seem to be reversed between intrinsic and extrinsic rotation. By inverting the two matrices this is the output (using CliMA/Oceananigans.jl#4560) for the divergence

[navidcy]

out of curiosity, how's the video on the first post after CliMA/Oceananigans.jl#4560?

[simone-silvestri]

eh, that will take a while to produce 😅 . I will run it and post it here when finished