[Mitgcm-support] some progress

[mitgcm-support at dev.mitgcm.org]

Hello,

I've found that exactConserv is really necessary when we 
use free-surface + filter U,V :

with Held & Suarez, 128x64x20 and no Shapiro filter for T,
1) no exactConserv => the model blows up with deltaT=450s, 
and with deltaT=300s, the minimum potential temperature 
becomes occasionally very low (less than 50.K).
2) turning on exactConserv solve this Pb and produce
less noise in W fields.

Still too early to claim that this fixed the problem of
low pot.T near the pole, but it might be the case:
1) in pre38, the U,V filters (Shap & FFT) were not applied after 
SOLVE_FOR_P but before. Then exactConserv is not necessary.
2) when I don't use exactConserv but filtered (shapiro) T,
the mimimum temperature is OK.
This explains why Daniel did not have this problem when
he run  H&S with 20 levels in last may. 

Olivier, are you using the rigid-lid or the free surface ?

See you,

Jean-Michel