Create very low resolution grids

[GeorgeGayno-NOAA]

The UFS group want to run the model at very low resolution - C18 or C12.

Check if UFS_UTILS can create grids this coarse.

Add relevant unit tests.

Related issues: ufs-community/ufs-weather-model#2508

[GeorgeGayno-NOAA]

Updated the cpld_gridgen program (91c08f4) to output C12, C18 and C24. The program ran to completion and a spot check of the files did not indicate any problems.

The files were placed on Hera: /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/UFS_UTILS/fix/orog.lowres

@DeniseWorthen - can you take a quick look?

[DeniseWorthen]

@GeorgeGayno-NOAA It seems to have generated the expected values.

I didn't check every combo, but I did some spot-checking. This for example is a "normal" c96.mx100 plot of land-frac

and this is C12.mx025

The land masses seem consistent, so I agree the generated files appear OK.

[GeorgeGayno-NOAA]

The following grids were created and placed on Hera in /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/UFS_UTILS/fix/orog.lowres

• C12.mx900
• C18.mx500
• C24.mx500

[GeorgeGayno-NOAA]

When creating the unit test for "remove_isolated_pts", I discovered a possible problem at the i-direction edges of the grid. To determine if a point is 'isolated', the surrounding 8 points are checked. So, if point (2,2) is checked for a 3x3 grid, the 8 surrounding points are:

 check point            2           2
 check points to the w            1           2
 check points to the e             3           2
 check points to the nw          1           1
 check points to the n             2           1
 check points to the ne           3           1
 check points to the sw           1           3
 check points to the s             2           3
 check points to the se           3           3

That is correct.

When point (1,2) is checked, some of the eight points are located at i=3.

 check point            1           2
 check points to the w            3           2
 check points to the e            2           2
 check points to the nw          3           1
 check points to the n            1           1
 check points to the ne           2           1
 check points to the sw           3           3
 check points to the s             1           3
 check points to the se           2           3

That is only correct for a global grid. I suspect when the orog program was converted to run on tiles (instead of the global GFS grid), this logic was not adjusted. Without modifying the orog program to process all tiles are once, there is no way to properly check if all edge points are 'isolated'. For now, I will simply not process the 'i' edges. Note, the 'j' edges were not being processed, which makes sense for an algorithm used for the old global GFS grid.

This test was done using 25c6b54.

[GeorgeGayno-NOAA]

When creating the unit test for "remove_isolated_pts", I discovered a possible problem at the i-direction edges of the grid. To determine if a point is 'isolated', the surrounding 8 points are checked. So, if point (2,2) is checked for a 3x3 grid, the 8 surrounding points are:

 check point            2           2
 check points to the w            1           2
 check points to the e             3           2
 check points to the nw          1           1
 check points to the n             2           1
 check points to the ne           3           1
 check points to the sw           1           3
 check points to the s             2           3
 check points to the se           3           3

That is correct.

When point (1,2) is checked, some of the eight points are located at i=3.

 check point            1           2
 check points to the w            3           2
 check points to the e            2           2
 check points to the nw          3           1
 check points to the n            1           1
 check points to the ne           2           1
 check points to the sw           3           3
 check points to the s             1           3
 check points to the se           2           3

That is only correct for a global grid. I suspect when the orog program was converted to run on tiles (instead of the global GFS grid), this logic was not adjusted. Without modifying the orog program to process all tiles are once, there is no way to properly check if all edge points are 'isolated'. For now, I will simply not process the 'i' edges. Note, the 'j' edges were not being processed, which makes sense for an algorithm used for the old global GFS grid.

This test was done using 25c6b54.

This change to "remove_isolated_pts" was made at a7545ea. All grid_gen regression tests passed.

[GeorgeGayno-NOAA]

While writing the unit test for 'remove_isolated_pts', I could not figure out how to trip the 'ELSE' branch of this 'IF' block.

UFS_UTILS/sorc/orog_mask_tools.fd/orog.fd/orog_utils.F90

Line 602 in 21c51d7

XN=(I-1)+1

My guess is that this logic was used for the old GFS gaussian reduced grid - where the number of points in the 'i' direction decreased toward the poles. I will remove this logic.

[GeorgeGayno-NOAA]

While writing the unit test for 'remove_isolated_pts', I could not figure out how to trip the 'ELSE' branch of this 'IF' block.

UFS_UTILS/sorc/orog_mask_tools.fd/orog.fd/orog_utils.F90

Line 602 in 21c51d7

XN=(I-1)+1

My guess is that this logic was used for the old GFS gaussian reduced grid - where the number of points in the 'i' direction decreased toward the poles. I will remove this logic.

The routine was updated at 380a57c. All grid_gen regression tests passed.

[GeorgeGayno-NOAA]

Mid-point check: all grid_gen regression tests passed using 858629b on Hercules.