[MITgcm-support] SOM
Christopher L. Wolfe
clwolfe at ucsd.edu
Tue Feb 5 14:00:00 EST 2008
Hi Mike,
I've been doing some experiments with both SOM and OS7MP in an eddy
resolving (dx ~ 5 km) configuration. The scheme code for plain SOM is
80 and the code for SOM with a flux limiter (what I believe Alistair
is referring to when he speaks of "limited" cases) is 81. You should
be warned that the SOM limiter is not yet a "proper" limiter, but
simply acts to ensure that the tracer field never becomes negative.
This can do bad things if the tracer field is *supposed* to go
negative. The code for OS7MP is 7.
In my experience, SOM is about 30% more diffusive than OS7MP (as
measured by buoyancy variance destroyed by the advection scheme), but
both of them are significantly better than the next runner up (scheme
33). With a Laplacian diffusivity of 1e-5, OS7MP is roughly as
diffusive as the explicit diffusion, but scheme 33 is about twice as
diffusive.
I don't know about compatibility issues; I haven't had many. The main
caveat is that SOM uses a different set of pickup files from the
other schemes, so an SOM run can't be restarted from a non-SOM run
without additional work.
The SOM scheme requires 10 times the memory of the other schemes to
store tracer information, while the OS7MP scheme requires a 4 point
halo (compared to the 3 point default halo). On non-memory limited
machines and a tile size of 56x56, I find that both schemes are about
as efficient as each other. On memory-limited machines (like SDSC
BlueGene) the SOM scheme simply won't run. As the tile size goes
down, the performance hit due to the extra halo point becomes more
significant.
Hope this helps,
Christopher
-----------------------------------------------------------
Dr. Christopher L. Wolfe 858-534-4560
Physical Oceanography Research Division OAR 357
Scripps Institution of Oceanography, UCSD clwolfe at ucsd.edu
-----------------------------------------------------------
On Feb 5, 2008, at 9:05 AM, Michael A. Spall wrote:
> Hi All,
>
> I have seen some discussions of the SOM advection scheme and I would
> like to try it. I have just downloaded the latest version of MITgcm.
> I can't find the code to turn on the scheme, what do I set
> tempAdvScheme to and are there any compatibility issues
> (staggeredTimeStep, etc)? I am doing high resolution problems
> (dx = O(1 km)) and would like to minimize horizontal and
> spurious diapycnal mixing. Is SOM the best one to use? I looked
> at Alistair's ppt talk where OS7MP is better for "limited" cases,
> but I don't know what "limited" means.
>
> Thanks,
> Mike
>
> --
> ==================================================
>
> Michael A. Spall
> Department of Physical Oceanography
> W.H.O.I. MS #21
> 360 Woods Hole Road
> Woods Hole, MA 02543
>
> mspall at whoi.edu
> (508) 289-3342 (office)
> (508) 457-2181 (fax)
>
> =================================================
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
More information about the MITgcm-support
mailing list