[MITgcm-support] Re: pickup problem

[Jean-Michel Campin]

Hi Tom,

I finally managed to run your set-up, and checked (without
subtracting the mean SSH) that 10it+10it=20it,
starting from iter=4320, and find zero diff between all the pickup
files corresponding to iter 4340.

regarding this problem:
> I'm not really sure why subtracting a constant SSH would be
> so catastrophic. Any ideas?
I suspect that it's due to the Non-Linear free surface:
in fact the model needs 2 time levels of the Eta field
to restart; In addition, the model read also (from the pickup)
the d.Eta/dt (dEtaHdt) and I don't know exactly what happens
when those 3 fields don't match:
 d.Eta/dt <> Eta(n) - Eta(n-1) / deltaT
I remember that it was not easy to get a perfect restart,
with no truncation error, and this is the reason why the
code is like this.
In fact, it might work better to remove the same constant
value to etaN and etaH.
Otherwise, apart from the OBCs (that can be corrected for a
mis-balance), the model conserve the water and there is no real
need to change the SSH during the integration.

Hope this help.

Cheers,

Jean-Michel