[MITgcm-support] Inconsistency in cubed sphere grid generator

[Nils Hutter]

Hi all,


I am using the northern face of the cubed sphere grid for Arctic 
simulations. While refining the grid, we found that there is a jump in 
the cell area for all cells that lie on a line in grid direction, which 
crosses the center of the each face. This is nicely seen as a darker 
cross in the attached figure. We checked the matlab grid generator for 
cubed sphere 
(http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/high_res_cube/matlab-grid-generator/) 
and found that the following changes in map_xy2xyz.m solves the problem 
(see attached figure of grid cell area for new mesh):


RCS file: 
/u/gcmpack/MITgcm_contrib/high_res_cube/matlab-grid-generator/map_xy2xyz.m,v
retrieving revision 1.1.1.1
diff -r1.1.1.1 map_xy2xyz.m
20,21c20,21
< kx=find(x<0);
< ky=find(y<0);
---
 > kx=find(x<=0);
 > ky=find(y<=0);
23a24
 >
57,58c58,59
< X(find(x==0))=0;
< Y(find(y==0))=0;
---
 > %X(find(x==0))=0;
 > %Y(find(y==0))=0;


For cs-tan-c-510-1020, the grid points generated with these modification 
are on average 20cm displaced compared to original grid. Maximum 
displacement is 79cm.

Does anyone know, why especially the last two lines that set some values 
to zero were necessary initially? These lead to the slightly lower dx at 
the boundary of the quadrant (-> the center of each face) and thereby 
smaller grid cell size.


Thanks,

Nils



-------------- next part --------------
A non-text attachment was scrubbed...
Name: cs-510-ra-new.png
Type: image/png
Size: 85295 bytes
Desc: not available
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20180704/45260032/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cs-510-ra-old.png
Type: image/png
Size: 85793 bytes
Desc: not available
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20180704/45260032/attachment-0003.png>