[MITgcm-support] advection scheme for T/S

[Jean-Michel Campin]

Hello Martin,

> Samar,
> 
> my experience with DST3 and the flux limited 2nd order scheme (77) is
> similar. Starting from a stable stratification with some noise to start
> an eddy field in zonal channel with zonal wind forcing and meridional
> buoyancy flux gradient, I get a solution, in which the isopycnal slowly
> rise and out-crop according to wind and buoyancy forcing, when I use
> the default 2n order scheme (2). With DST3 with flux limiting (33) and
> 2n order with flux limiting (77), the first thing that happens is that
> the channel gets homogenized vertically and then the expected solution
> starts to develop. But of course, because of the initial mixing, the
> steady state is very different from the case with the default advection
> scheme. I have refrained from using 33 or 77 for temperature and
> salinity since. (o:
> 
> Martin

I though this ploblem has been fixed : a uniform stratification + noise 
that gets rapidly homogenized vertically when using advec.scheme=33 or 77
I look to the MITgcm-support archive, and find this e-mail that I sent 
a year ago, but didn't find anything later (may-be 
the discussion switch to the devel list ?):

> [MITgcm-support] Re: problem higher order advections schemes
> Wed Sep 24 17:21:04 EDT 2003
> Hi Martin,
> 
> Didn't do any other test, but thinking to what causes
> this strange and very different behavior of 
> those advection schemes (77 30 & 33), it could simply due to the 
> fact that they are not using the Adams-Bashforth 
> making the internal wave mode completely explicit (and
> therefore unstable), unless you turn on the stagger timestep option.
> Might be as simple as that.
> 
> So, Alistair and me strongly recommand to use the stagger timestep 
> option with any of those advection scheme.
> 
> See you,
> 
> Jean-Michel

And then I find a directory called "channel", about 1 yers old,
and run again a set-up with advection scheme 77 for T & S, 
T=active, S=passive:
initial fields for S = T_noise = a stratified Temperature field, with 
a front in the Y direction (quasi zonal), with very small noise added.
a) run without staggerTimeStep, T_initial = zonal average of T_noise
b) same as (a) with staggerTimeStep,
c) same as (b) but T_initial = T_noise

results of 1.y integration:
a) produces the expected instability that mixes quiet efficiently
the channel, and strongly affects the stratification.
b) and c) work nicely and preserve the statification. 

Could you clarify the things (what is still a problem and what 
In summary, Can you clarify the problem you find and mention to Samar,
and is it something new or that has already been explained before 
(I don't remember very well).

Thanks,

Jean-Michel