Agrif sponge

[jchanut]

I guess that the definition of sponge coefficients does not account for possible regional domain open boundary corners.
Moreover, as it is defined, the sponge layer is active all along the regional domain boundaries. This implies that even if the outermost points of the zoom domain are in land, the sponge is still active up to 6 points inside. This means that you may have have unwanted damping in some bays close to the regional domain boundaries.
Jerome

[jchanut]

"6 points inside" is valid for a refinement factor of 3... That's rather "2 coarse grid points".

[smasson]

Can you gave us a little more details about your first sentence? I didn't really understood your point :-)

[jchanut]

That's not really clear, indeed.
The "zlocalviscsponge" array (which defines the sponge viscosities) is reset for each open boundary. One should have maybe something like that:

ztmp = visc_tra * ( ispongearea-jj ) * z1spongearea
zlocalviscsponge(ji,jj)=Max(zlocalviscsponge(ji,jj),ztmp)

to account for bdy corners. Otherwise you may have a discontinuity in the coefficient
Hum. May be more easy in French ;-)

[jchanut]

Here is what I think would be the correct way to do. Seb: I wait for your approval.

[jchanut]

Possible patch for nemo_3_5_alpha attached. Seb: waiting for your approval.

[smasson]

I carefully looked at your code.
I liked it and It looks ok to me expect 3 points:

• when you compute the arrays for the dynamics (spe1ur2 and spe2vr2) you use zlocalviscsponge that was multiplied by visc_tra or in this part of the code it should be multiplied by visc_dyn
• there is a small typo

  IF (( .NOT. spongedoneT ).OR.( .NOT. spongedoneT )) THEN
  

should be

IF (( .NOT. spongedoneT ).OR.( .NOT. spongedoneU )) THEN

• lk_spdone should be named ll_spdone