[MITgcm-support] documentation for pkg/smooth

[Menemenlis, Dimitris (329D)]

Gael and Matt, thank you both for prompt and helpful replies!

I will start a MITgcm/pkg/smoooth/smooth_description.tex
file and include a link to this thread:
as a temporary placeholder.

Cheers, Dimitris

On Apr 6, 2015, at 5:56 PM, gael forget <gforget at MIT.EDU<mailto:gforget at MIT.EDU>> wrote:

Hi Matt,
It is, mathematically, the full diffusion, because we apply sqrt on the gradients and the sqrt on the ctrls. So for a full optimization cycle it really does essentially become full diffusion on the ctrl fields.
Correct except that Dimitri may only be running the forward model.
Your comment w.r.t. the normalization v. variance preserving is also right.
Gael


On Apr 6, 2015, at 5:36 PM, gael forget <gforget at mit.edu<mailto:gforget at mit.edu>> wrote:

Hi Dimitris,

Dear MITgcmers,
is any documentation for pkg/smooth available?
Unfortunately this is still missing from mitgcm documentation.
I hope to take time to add to MITgcm documentation regarding
this package (as well as ecco,ctrl, and profiles …) this spring.

In particular, I am trying to understand how pkg/smooth parameters relate
the (2 k_1 T)^0.5 length scale discussed in the Weaver and Courtier paper:
http://mitgcm.org/~gforget/pdfs/weavercourtier01.pdf
The parameters as set in data.smooth follow WC01’s notations
(if I recall correctly) specify the spatial correlation modeling
for use with pkg/ctrl & autodiff. Is this what you are trying to do?

Or more precisely, what is the half-power bandwidth
of a filter based on following parameters:
&SMOOTH_NML
smooth2Dnbt(1)=5000
smooth2D_Lx0(1)=100000.
smooth2D_Ly0(1)=100000.
smooth2Dtype(1)=1
smooth2Dsize(1)=0
smooth2Dfilter(1)=0
In the case of the WC01 correlation modeling (if I recall correctly)
it is only the squared root of the diffusion operator that is applied.

If the matter is simply to smooth out a field up to a specified scale,
things are somewhat different. You want to use smooth_hetero2d.F
or smooth_basic2D.F and will find examples in pkg/ecco.

I hope this helps,
Gael

Does above define a lowpass filter for scales greater than 100 km,
200 km, 628 km, or some other value?

Thanks, Dimitris


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20150407/58def890/attachment.htm>