[MITgcm-devel] seaice

[Martin Losch]

Dimitris and other seaice modelers,

I have modified
dynsolver.F
diffus.F
advect.F
to replace all grid information (dxtice,dytice,etc) with generic  
MITgcm grid parameters (dxF,dyF, etc, defined in GRID.h)
So far, this has NOT changed the results of lab_sea experiment,  
except for the advection part.
I tried to use a flux-form, similar to the second-order central  
difference form in pkg/generic_advection/gad_c2_adv_x.F etc. but  
there are small differences (the model needs a better advection  
scheme that is positive!!!).

Unfortunately on a cubed_sphere grid with seaice, the results do  
change a lot, but is that surprising?

What also changes the results:
in dynsolver.F:
RADIUS = rSphere (default = 6370. _d 3)
TWO*OMEGA*SINEICE(I,J,bi,bj) = _fCoriG(I,J,bi,bj) (=2._d0*omega*sin 
(_yG(i,j,bi,bj)*deg2rad)

I don't see why.

how should I proceed?:
I could send the routines around for testing or looking at and we  
could agree on checking them in or not checking them in.
Or I could forget about the project altogether, because it is tedious  
and unrewarding. Is it worth it? Can you deal with the small  
differences that these changes imply?

Martin


On Feb 13, 2006, at 3:37 PM, Dimitris Menemenlis wrote:

> Martin, I am attaching a draft wish list for MITgcm sea-ice code  
> for you and others to consider and modify.  The list is based on  
> your recent e-mails and work on ice shelves as well as on  
> discussions with Jean-Michel and Patrick during ECCO meeting two  
> weeks ago and with Elizabeth Hunke and Bill Lipscomb at NCAR last  
> week.  I am not sure when these things will happen and in what  
> order.  I am presently trying to hire someone that can help move  
> the MITgcm sea-ice code forward.  Timing also depends on Jinlun's  
> timetable and availability.
>
> 1. Clean-up grid parameters as you suggest.  Make sure that there  
> are no duplicate or repeated computations.  Use model/inc/GRID.h  
> parameters where possible, define the rest in pkg/seaice/ 
> SEAICE_GRID.h and seaice_init.F.
>
> 2. Adapt LSR solver for curvilinear coordinates so that it is  
> "properly" compatible with cubed-sphere and terraced grids.
>
> 3. Port LSR solver to C-grid.  This is something that Jinlun is  
> working on for GISS, and he may be able to spend two months next  
> year to implement on MITgcm.
>
> 4. Better coupling of pkg/seaice thermodynamics with surface level,  
> similar to what Jean-Michel has done for pkg/thsice.
>
> 5. Coupling of pkg/thsice thermodynamics with pkg/seaice dynamics.
>
> 6. Ice shelves and cavern representation.
>
> 7. Import more modern, multi-category sea-ice model.  The plan is  
> to import the Los Alamos CICE sea ice model (http:// 
> climate.lanl.gov/Models/CICE/), ideally with interchangeable  
> dynamics between Jinlun's LSR VP solver and Elizabeth's explicit  
> EVP solver.
>
> Dimitris
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel