[MITgcm-devel] lab_sea

[Jean-Michel Campin]

Hi Martin, Dimitris, and Patrick,

It seems to converge to the C-grid choice.
But regarding LSR/EVP solver, should we keep (for lab-sea)
the one that you, Dimitris, mostly use (LSR), or switch to EVP ?
Does one solver behave better with the adjoint ?
But if both have some advantages, we could think of 2 experiments
(or 2 tests, with similar set-up) to test them both ?

Jean-Michel

On Fri, Dec 08, 2006 at 05:00:09PM -0800, Dimitris Menemenlis wrote:
> >Assuming the B-grid LSR is a thing of the past,
> 
> Yes B-grid is a thing of the past.
> 
> >we'd change lab_sea to use the new C-grid code plus EVP solver.
> 
> I don't have much experience with EVP solver.  I've used it a bit and it 
> works well, it's often faster than LSR, but I have not looked at solutions 
> in detail.
> 
> >I would then also add an adjoint sensitivity to that,
> >similar to the natl_box_adjoint/code_seaice (with full dynamics).
> >OK? 
> 
> That would be awesome!
> 
> D.
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel