[MITgcm-support] Re: problem higher order advections schemes

[Jean-Michel Campin]

Hi Martin,

After further investigation, I have the impression that
this set-up is very sensitive to any local perturbation in the
vertical diffusion.
To summarize: with advection scheme 2 or 4, every thing 
remains stable (after the initial adjustment (~ 1 month) CFL_v,w < 2.e-4) 
and the initial stratification is preserved.
with advection scheme 30, 33, 77 Super-Bee (and multiDim as default),
I get strong mixing due to high vertical velocity (max_CFL_v > 0.02 
& max_CFL_w > 0.1 ).
This is very sensitive because when I use the advection scheme 77 
but changed to Lax-Wendroff, the max Temp start to decrease
almost immediately (with strong W) if I use multiDim but much later 
(after > 2 month) if I use multiDim=.FALSE.

Could this high sensitivity be related to
Coriolis on C grid with not enough horizontal viscosity ?
(note that all remain perfectly zonally symetric, included W).

It is clear that it's related to dynamical feedback, so I decided to
use the stagger time-stepping option and, guess what, it works nicely 
and the initial stratification  is preserved
with 77 Lax Wendroff & SuperBee, 30 and also 33 !

Hope this will help.

Cheers,

Jean-Michel

PS: 2 little things:
a) the minimum temperature is slightly below the initial minimum
and is still decreasing after 1 year, but tends to be lower when I use KPP
compared to the simple case with convective adjustment (ivdc_kappa=100).
b) I tried z* and get perfect global conservation of total heat content.