There's a bug in CICE or MOM at the tripole seam that crosses the north pole along the 100W/80E meridian north of 65N. Apologies for the long post, I wanted to get it clear in my head.
From the grid you can see that on the Atlantic side of the seam positive u velocity runs from Canada to Siberia; on the Pacific side it is the reverse orientation.
v velocity also changes sign at the seam. On both sides of the seam positive v velocity has a northward component. There is also a zonal component of v velocity that changes sign at the seam (positive v is partly eastward in western Siberia/Canada and partly westward in eastern Siberia/Canada).
Here are some views off Siberia. Date is in the heading.
Starting with u velocity (in pairs, first ice, then ocean).
Most of the time the u velocity has a sign change at the seam as it should (positive u is down on the left of 80W and up on the right in this view):
But sometimes the sign of u does not change across the seam i.e. the physical tangential velocity abruptly changes direction across the seam.
It starts small
and grows
This strong shear at the seam leads to flow instability within a few days
The same sequence of days in v velocity shows no such problems (positive v is rightward on the left of 80W and leftward on the right of 80W in this view)
The wind stress also shows no problems at the seam.
Movies of the ice shear show this problem comes and goes in various locations along this seam.
Clearly there's something wrong with at least one term in the u tendency at the seam in MOM or CICE. The fact that the problem is not apparent most of the time suggests the usually-dominant terms are correct, and the problem appears only on occasions when the buggy term is unusually large.
I expect it's a term with a stencil that crosses the seam but doesn't take into account the sign change. The fact that it is asymmetrical (the positive u on the west spreads eastwards) and that the v velocity is eastward suggests the problem is with the y advection of x momentum in the u tendency in MOM or CICE, ie vdu/dx (remembering that x,u are along-seam and y,v across-seam).
It is unclear whether the problem arises first in the ocean or ice (it seems pretty simultaneous in the figures) but I'm trying runs with zero ocean-ice drag to try to identify the culprit. The fact that v is more strongly eastward in the ice than the ocean when the problem starts on 0017-08-03 makes me suspect the bug is in CICE rather than MOM.
There's a bug in CICE or MOM at the tripole seam that crosses the north pole along the 100W/80E meridian north of 65N. Apologies for the long post, I wanted to get it clear in my head.
From the grid you can see that on the Atlantic side of the seam positive u velocity runs from Canada to Siberia; on the Pacific side it is the reverse orientation.
v velocity also changes sign at the seam. On both sides of the seam positive v velocity has a northward component. There is also a zonal component of v velocity that changes sign at the seam (positive v is partly eastward in western Siberia/Canada and partly westward in eastern Siberia/Canada).
Here are some views off Siberia. Date is in the heading.
Starting with u velocity (in pairs, first ice, then ocean).
Most of the time the u velocity has a sign change at the seam as it should (positive u is down on the left of 80W and up on the right in this view):
But sometimes the sign of u does not change across the seam i.e. the physical tangential velocity abruptly changes direction across the seam.
It starts small
and grows
This strong shear at the seam leads to flow instability within a few days
The same sequence of days in v velocity shows no such problems (positive v is rightward on the left of 80W and leftward on the right of 80W in this view)
The wind stress also shows no problems at the seam.
Movies of the ice shear show this problem comes and goes in various locations along this seam.
Clearly there's something wrong with at least one term in the u tendency at the seam in MOM or CICE. The fact that the problem is not apparent most of the time suggests the usually-dominant terms are correct, and the problem appears only on occasions when the buggy term is unusually large.
I expect it's a term with a stencil that crosses the seam but doesn't take into account the sign change. The fact that it is asymmetrical (the positive u on the west spreads eastwards) and that the v velocity is eastward suggests the problem is with the y advection of x momentum in the u tendency in MOM or CICE, ie vdu/dx (remembering that x,u are along-seam and y,v across-seam).
It is unclear whether the problem arises first in the ocean or ice (it seems pretty simultaneous in the figures) but I'm trying runs with zero ocean-ice drag to try to identify the culprit. The fact that v is more strongly eastward in the ice than the ocean when the problem starts on 0017-08-03 makes me suspect the bug is in CICE rather than MOM.