[MITgcm-support] Problem with CS96 model set-up

[Stephen Hunter [EAR]]

Dear MITgcm Support Team,



I am trying to create a cube-sphere (CS96) MITgcm Ocean model setup which I plan to force with climatology derived from an Atmosphere-only GCM in a way similar to ECCO (but without the adjoint package and CS96 rather than LLC90 grid). I believe (well, hope) that I can create correctly the appropriate bathymetry and forcing binary files.



My problem lies in the initialisation of the model, in particular regarding the grid setup. I have searched the manual and verification experiments for code but I am still unsure how to create the grid_cs96_face00[1-6].bin files. Fortunately, I found a set within http://mit.ecco-group.org/datasets/ecco_for_las/version_4/grids/grids_input/cs_nostretch/cube_96/, which appear to be the correct size (8bytes*(96+1)*(96+1)*16records = 1204352 bytes). I have renamed the files to grid_cs96.face001.bin etc. and set up my input/data to see read them. Additionally, I can read these grid_cs96.face00[1-6].bin binaries using Matlab GCMfaces and can create a mygrid structure (which I have used to create the bathymetry.bin).  I believe these binary grid files are what I need. I have within input/data



# Gridding parameters

&PARM04

horizGridFile='grid_cs96',

usingCurvilinearGrid=.TRUE.,



On running, the model fails with  the following error

forrtl: severe (36): attempt to access non-existent record, unit 9 file [my_full_path]/build/grid_cs96.face001.bin



I have searched MITgcm-support and it seems the model is trying to access records that are not within the grid_cs96.face00[1-6].bin files - which suggests an incorrect SIZE.h file.  I have tried numerous configurations of SIZE.h. A simple case I have for SIZE.h which (I believe) sets up a CS topology with 24 tiles running on 6 processors.



      PARAMETER (

     &           sNx =  48,

     &           sNy =  48,

     &           OLx =   4,

     &           OLy =   4,

     &           nSx =   4,

     &           nSy =   1,

     &           nPx =   6,

     &           nPy =   1,

     &           Nx  = sNx*nSx*nPx,

     &           Ny  = sNy*nSy*nPy,

     &           Nr  =  50 )



Looking at the cs32 verification experiments which I can run, the experiment uses grid_cs32 .face00[1-6].bin files that are 156816bytes in size (e.g. within global_oce_cs32/input_fields/) whereas I would expect them to be 8bytes*(32+1)*(32+1)*16records=139392 bytes. Hence I think that I am misunderstanding something. I believe that my compiler optfile is correct (the verification experiments work).


Kind regards,
Stephen Hunter

School of Earth & Environment
University of Leeds

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20200227/a0432426/attachment.html>