[MITgcm-support] Telescoping grid question

[Martin Losch]

Hi Dustin,

this should be possible, but for an general orthogonal curvi-linear grid (usingCurvilinearGrid = .TRUE., in data), the model requires a lot of grid information that you need to compute offline (all 2D fields). Have a look at ini_curvilinear_grid.F to understand, what you actually need (In your case I would define the cpp-flag OLD_GRID_IO and provide the files individually). in model/inc/GRID.h you’ll find a detailed description of what the grid variables actually are (or in the online documentation).

For a cartesian grid (usingCartesianGrid = .TRUE. in data), you can specify 2 vectors for the grid spacing (dx,dy), it’s probably the easiest to use delXfile and delYFile, but then the model assumes that the cells are locally rectangular, which is definitely not the case for your model domain, so I don’t recommend doing that! … but maybe you can use that to have the model compute the grid information that you need for your curv-linear grid? I am not sure how large the errors would be (eg. for the grid cell areas, maybe you can live with that).

Martin

On Jun 4, 2014, at 6:58 PM, Dustin Carroll <dcarroll at uoregon.edu> wrote:

> Hi All,
> 
> I have a new domain set up as a square box. The resolution is fine (10 m) in a square that is centered in the middle of the box and smoothly transitions to coarse (500 m) at the boundaries (N,S,E,W). I wanted to ask if it’s possible to use matrices of dx,dy for this type of grid? In my previous models, I have used a vector for and dx,dy.
> 
> Thanks,
> 
> Dustin Carroll
> University of Oregon
> 
> <Domain.png>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support