ConvertExodusToSCRIP ignores rank=2 in RLL grids

[czender]

This has the effect of "flattening" all 2-D Exodus files into 1-D (aka unstructured) SCRIP files. I'm unsure whether ConvertSCRIPToExodus flattens 2-D SCRIP files into 1-D Exodus files. Preserving the rank of grids would facilitate using TR and MBTR (which requires Exodus or connectivity in some form) in computing weights for a large number of existing SCRIP grids. Short demonstration of what I mean. The value of grid_rank is 1 (not 2) in the output SCRIP file after the last command. A grid_dims array would also be required in the SCRIP output to fix this issue.

zender@firn:~$ GenerateRLLMesh --lat 180 --lon 360 --file cmip6_180x360.g
Parameters:
  --lon <integer> [360] 
  --lat <integer> [180] 
  --lon_begin <double> [0.000000] 
  --lon_end <double> [360.000000] 
  --lat_begin <double> [-90.000000] 
  --lat_end <double> [90.000000] 
  --global_cap <bool> [false] 
  --flip <bool> [false] 
  --in_file <string> [""] 
  --in_file_lon <string> ["lon"] 
  --in_file_lat <string> ["lat"] 
  --in_global <bool> [false] 
  --verbose <bool> [false] 
  --file <string> ["cmip6_180x360.g"] 
  --out_format <string> ["Netcdf4"] 
=========================================================
..Generating mesh with resolution [360, 180]
..Longitudes in range [0, 360]
..Latitudes in range [-90, 90]
..Writing mesh to file [cmip6_180x360.g] 
Nodes per element
..Block 1 (4 nodes): 64800
..Mesh generator exited successfully
=========================================================
zender@firn:~$ ConvertExodusToSCRIP --in cmip6_180x360.g --out cmip6_180x360.nc
Parameters:
  --in <string> ["cmip6_180x360.g"] 
  --out <string> ["cmip6_180x360.nc"] 
  --out_format <string> ["netcdf4"] 
------------------------------------------------------------

..Loading input mesh
Mesh size: Nodes [64442] Elements [64800]
..Writing mesh
Nodes per element
..Block 1 (3 nodes): 720
..Block 2 (4 nodes): 64080
..Done writing
..Mesh converter exited successfully
=========================================================
zender@firn:~$ ncks -m cmip6_180x360.nc
netcdf cmip6_180x360 {
  dimensions:
    grid_corners = 4 ;
    **grid_rank = 1 ;**
    grid_size = 64800 ;

  variables:
    double grid_area(grid_size) ;
      grid_area:units = "radians^2" ;

    double grid_center_lat(grid_size) ;
      grid_center_lat:_FillValue = 9.96920996838687e+36 ;
      grid_center_lat:units = "degrees" ;

    double grid_center_lon(grid_size) ;
      grid_center_lon:_FillValue = 9.96920996838687e+36 ;
      grid_center_lon:units = "degrees" ;

    double grid_corner_lat(grid_size,grid_corners) ;
      grid_corner_lat:_FillValue = 9.96920996838687e+36 ;
      grid_corner_lat:units = "degrees" ;

    double grid_corner_lon(grid_size,grid_corners) ;
      grid_corner_lon:_FillValue = 9.96920996838687e+36 ;
      grid_corner_lon:units = "degrees" ;

    int grid_dims(grid_rank) ;

    int grid_imask(grid_size) ;
} // group /
zender@firn:~$ 

[czender]

@whannah1 sorry I missed your comment last week. I use RLL to mean Rectangular/Regular Longitude Latitude. The horizonal dimensions are separate coordinate axes so the grid should be Rank 2 (logically rectangular axes) not Rank 1 (unstructured).

[paullric]

This should be fixed on branch updated-api. Need to wait on merge until mbtempest is updated to new API.