0.25 deg CICE grid misaligned: U points at NW corner

[aekiss]

The 0.25° CICE grid is misaligned: u points are at the NW corner of t cells instead of the NE corner. The latitudes are offset correctly but the longitudes are offset incorrectly - see this notebook. Both u latitudes and u longitudes are incorrectly set at the NW corner - see below

• ulon -280... 79.75 offset 1 cell to the west!
• tlon -279.875... 79.875

Thanks to @anton-seaice for spotting this.

Related issues:

• 0.25deg configuration access-om3#101 (comment)
• CICE grid misaligned in some 0.1 deg configurations #190
• Write further tests to check that ice and ocean grids match #129

[aekiss]

@MartinDix does this also affect ACCESS-OM2-025?

[MartinDix]

It looks like both ulat and ulon were set from the NW corner, so ulat is also wrong in the tripolar region.

See https://github.com/MartinDix/notebooks/blob/master/grid025_check.ipynb

[MartinDix]

@MartinDix does this also affect ACCESS-OM2-025?

Testing in u-cz789 (PD control with ocean/ice starting from standard OM2 initial state) with a corrected CICE grid file shows differences from the start.

[anton-seaice]

FYI @aidanheerdegen :- we haven't finalised a fix yet, but well need a fix before the 0.25 deg ACCESS-NRI release

[aekiss]

It looks like both ulat and ulon were set from the NW corner, so ulat is also wrong in the tripolar region.

thanks for spotting that @MartinDix

[aekiss]

The ulon error is -0.25° throughout the non-tripolar region, but increases in magnitude in the tripolar region up to -180° at the two tripolar points - see https://github.com/aekiss/notebooks-MartinDix/blob/master/grid025_check.ipynb

[anton-seaice]

Does OM2 (OASIS?) map variables from MOM to CICE based on the lat/lon information? And if so, does that use ULON or TLON? Or is it just based on the grid cell index?

i.e. does the [0,0] cell from MOM, end up coupled with the [0,0] cell or the [0,-1] cell?

[anton-seaice]

The coupling code in cice5 refers to tlon, e.g: https://github.com/COSIMA/cice5/blob/edcfa6f9c76ed05b63196ce4b5355fa5a8f4fe3a/drivers/auscom/cpl_interface.F90#L363

[anton-seaice]

This issue was fixed here:

COSIMA/esmgrids@1b6837d

With some testing added too. We should be getting accuracy between mom & cice grids of ~1e-13.

This was added to the OM2 configs here through four PRs linked here: ACCESS-NRI/access-om2-configs#17