[MITgcm-devel] Re: recent changes not tested

[Martin Losch]

Jean-Michel,
thanks for notifiying me. Your modifications did not break my test. 
Maybe it's time to check in a verification experiment for these 
schemes. I have one that tests all four schemes (pp81,my82,ggl90,opps), 
but has little physical meaning: 3x1x20 points with surface cooling and 
wind stress. If you are planing more modifications to the schemes and 
would like to test them, I can send this suite to you directly, or if 
you find them useful, I can check them in. But again: I am far from 
finished with my work on these things (in fact, at the moment it 
rests), and I cannot guarantee that things will change in the future.

Your "other" thing:
On Oct 18, 2004, at 1:48 AM, Jean-Michel Campin wrote:
> An other thing:
> I've seen that (in fact in all those 3 pkg) you set KappaRS
> to be strictly equal to KappaRT, even if diffKrS is not equal to
> diffKrT.
>
> There might be some cases (for example when you use a more diffusive
> advection scheme for S than for T) where you want a different 
> background
> of vertical diffusivity for T & S (especially in the interior),
> and add on top of it, the vertical diffusitity that pp81 computes.
>
> I know it's already like this with ptracers, but ...

I agree, I explicitly exclude double-diffusion effects etc and for 
numerical flexibility, there should be the possibility to have 
different backgound diffusion for salt and temperature. The reason why 
I didn't implement it that way is Alistair's email when I first asked 
about mixing schemes. I cite it here:
> Martin,
>
> Go ahead but...
>
> I had wanted to change the way we do diffusivity (vertical+horizontal) 
> as a
> whole.
>
> We currently have a 1/2/3D diffusivity per variable which is 
> unmanagable and
> impractical with multiple tracers.
>
> I wanted to have a diffusivity per process described by a 
> non-dimensional
> diffusivity (profile or field) and then a dimensional scalar per
> tracer/process. This gives a matrix of possibilities for configuration 
> and
> reduces memory. The diffusivities would be added up at the two places 
> where
> we need them and only then.
>
> It would be most convenient if this switch happened sooner rather than 
> later
> (ie. Before more processes get added). Since I'm out of it (travelling 
> for
> two weeks and then implementing my best ideas in MOM for a few years) 
> are
> you up to the task?
>
> Otherwise, don't let me stop you adding the second most requested 
> feature.
>
> A.

So I only have one diffusivity for both T and S, but I may have 
misunderstood Alistair completely. I'll add a sepapate diffusivity for 
salinity that as soon as I get around to it, but this will take a while 
(I am swamped with other things at the moment). If you are impatient, 
feel free to do it yourself, but who uses this at the moment, other 
than me?

Martin