Mismatched dimensions between write_grid and oasis_put
Up to Bugs and debugs
Posted by Anonymous at February 27 2019
Hi,
I'm implementing OASIS-MCT within CICE and am stuck on a couple points. Any clarification/help is greatly appreciated.
1) In oasis_write_grid (Sect. 2.2.4), "ny_global [INTEGER; IN] : second dimension of the global grid (=0 if the grid is expressed as a 1D vector)." In Sect. 5.1, "Unstructured grids or other grids expressed with 1D vectors are supported by setting nx to the total number of grid points and ny to 1." The former generates empty data in grids.nc and masks.nc files. The latter is correct, yes?
2) Does using an Orange vs. Points partition have any sensible impact (e.g., on memory, runtime,...)? What about a logically rectangular vs. unstructured grid after the mapping files are made?
3) I'm testing with the global gx3 mesh with 100x116=11600 points in the horizontal, treating it as an unstructured grid with a Points partition. In the written grids.nc and masks.nc, {lat,lon,msk} are properly dimensioned 1 x 11600. When I go through the same steps to oasis_put to OUTPUT a sea ice variable, the output file variable is dimensioned time x 1 x 11455. The only matching information is that, in debug.root.001, ... (oasis_coupler_print) partid, gsize 1 11455 (oasis_coupler_print) partid, lsize 1 725 (oasis_coupler_print) partid, nx,ny 0 0 (oasis_coupler_print) rpartid,gsize 1 11455 (oasis_coupler_print) rpartid,nx,ny 0 0 ...
Any ideas on how this is possible?
Thanks! Nick
Posted by Anonymous at March 1 2019
Hi Nick,
1) yes you are right, I will modify the User Guide
2) We have never tested this is detail. Thanks for your feedback if you do!
3) We have never observed this and we don’t know why this happens. Please note that the point partition is relatively new in OASIS3-MCT. Do you have the same problem if you describe your grid with another type of partition?
Best regards,
Laure & Sophie
