[MITgcm-support] Re: nonHydrostatic

[Dimitris Menemenlis]

Geoffroy, the hypotheses that I make in explaining what you see are:

1. Your set-up, in the hydrostatic case, is marginally stable.  If left 
unperturbed it can relax to uniform temperature via diffusion but a small 
perturbation can trigger convection and water motions that eventually make the 
model blow up.

2. Without salt diffusion, there are no perturbations, or more likely the 
perturbations are symmetric in a way such that no instabilities are triggered.

3. Turning on vertical diffusion for salt causes small perturbations at the 
truncation level, which trigger the instabilities in your setup.

Hypothesis 1 and 2, you can test by adding a different kind of perturbation to 
your setup and see what happens.  For example, try adding a small initial 
temperature or salinity perturbation anywhere in your domain and then try the 
case with salinity diffusivity zero and see what happens.  If hypotheses 1 and 2 
hold, then a small perturbation to the initial conditions will trigger the 
instabilities, the same way as a non-zero vertical salt diffusivity term does.

Hypothesis 3, I have already checked in a different domain, see message from 
8/9/2007.  You can also verify that this is the case in your setup by monitoring 
dynstat_salt_max, dynstat_salt_max, dynstat_salt_max, dynstat_salt_sd, and 
dynstat_salt_del2.

I warn you, however, that I am the wrong person to advise you on this particular 
problem and that my explanations can be way off the mark.  D.