Data_override:: Issue with reading the grid file in data_override when using 0:180:-180:0 grids

[uramirez8707]

Describe the bug
Grid files that have longitude to ranges from 0 to 180 then -180 to 0 are not being read correctly by data_override when using "OCN" and "ICE" grids.

The code gets here:

FMS/data_override/include/get_grid_version.inc

Lines 218 to 224 in 460c592

	if(trim(mod_name) == 'ocn' .OR. trim(mod_name) == 'ice') then
	do j = jsc, jec
	do i = isc, iec
	lon(i,j) = (tmpx(i*2-1,j*2-1)+tmpx(i*2+1,j*2-1)+tmpx(i*2+1,j*2+1)+tmpx(i*2-1,j*2+1))*0.25_lkind
	lat(i,j) = (tmpy(i*2-1,j*2-1)+tmpy(i*2+1,j*2-1)+tmpy(i*2+1,j*2+1)+tmpy(i*2-1,j*2+1))*0.25_lkind
	end do
	end do

At the dateline, it will use values around -180 and 180 and average those which will cause bad longitudes.
You can see the issue in this file:
ncview /gpfs/f5/gfdl_f/scratch/Uriel.Ramirez/REGIONAL_MODEL/FMS/build/test_fms/data_override/LineOfWeirdness_Test /grid_read_in_ocn.nc

MOM6 seems to use the centroids when it reads its grid:
https://github.com/NOAA-GFDL/MOM6/blob/8ecb2603119b8c9d4dee9ee6b174e1864415d1cc/src/initialization/MOM_grid_initialize.F90#L220-L222

To Reproduce
This Test reproduces the error:
https://github.com/uramirez8707/FMS/blob/LineOfWeirdnessTest/test_fms/data_override/test_data_override_LOW.F90

The data files to run it can be found here:
/gpfs/f5/scratch/Uriel.Ramirez/gfdl_f/REGIONAL_MODEL/FMS/build/test_fms/data_override/LineOfWeirdness_Test

Expected behavior
The grid file should be read correctly by data override. We should either:

1. Read the centroids from the file directly
2. Add logic to the averaging to deal with this corner case
3. Model components can pass in their own lat/lon to use by data_override

System Environment
This happens in any system

Additional context
This was seen by @theresa-morrison and @kshedstrom

[uramirez8707]

@marshallward
Can you confirm that MOM6/SIS2 is simply reading in the centroids from the grid file, instead of calculating it using the four corners as data override is doing?

FMS/data_override/include/get_grid_version.inc

Lines 218 to 224 in 460c592

	if(trim(mod_name) == 'ocn' .OR. trim(mod_name) == 'ice') then
	do j = jsc, jec
	do i = isc, iec
	lon(i,j) = (tmpx(i*2-1,j*2-1)+tmpx(i*2+1,j*2-1)+tmpx(i*2+1,j*2+1)+tmpx(i*2-1,j*2+1))*0.25_lkind
	lat(i,j) = (tmpy(i*2-1,j*2-1)+tmpy(i*2+1,j*2-1)+tmpy(i*2+1,j*2+1)+tmpy(i*2-1,j*2+1))*0.25_lkind
	end do
	end do

It seems that way from looking at the MOM6 code, but we will like to confirm and be consistent.
https://github.com/NOAA-GFDL/MOM6/blob/8ecb2603119b8c9d4dee9ee6b174e1864415d1cc/src/initialization/MOM_grid_initialize.F90#L220-L222

[marshallward]

Sorry for the late reply, I am getting over a bout of COVID.

I agree that geoLonT is on the center points, although I'm not sure how tmpZ is constructed.

[theresa-cordero]

Thanks! I am looking forward to using this soon!