TEMPEST regridding tool errors with new oEC60to30 grids

[jonbob]

The TEMPEST regridding tools will not successfully make mapping files using the new oEC60to30v2 or oEC60to30v3 grids. With the v3 grid, the error message in the TEMPEST log says:

`Validate mesh A
Error detected (orientation):
Face 354, Edge 0, Orientation 5.27839e-06
(x,y,z):
n0: 9.93977e-01 1.09429e-01 -5.87795e-03
n1: 9.99983e-01 6.21380e-05 -5.89998e-03
n2: 9.99983e-01 1.10999e-04 -5.85172e-03
(lambda, phi):
n0: 1.09650e-01 -5.87798e-03
n1: 6.21391e-05 -5.90001e-03
n2: 1.11001e-04 -5.85176e-03
X-Product:
5.27655e-06 2.89801e-07 -3.23976e-07

..EXCEPTION (GridElements.cpp, Line 864) Mesh validation failed: Clockwise element detected
`
Note that this exact version of TEMPEST did successfully work with the original oEC60to30 grid.

B-case runs on anvil using oEC60to30v2 have so far all failed, for various reasons ACME #1184 -- though it''s not clear if the failures are machine-related or specific to this grid. So there are also runs using the original oEC60to30 grid underway on anvil, as well as an oEC60to30v2 test on grizzly.

[jonbob]

I tracked down the failure to a lack of checking in the TEMPEST remapping tool. It assumes the SCRIP file specifies the "grid_corners" in degrees, but does not check the attributes for units in those netcdf files. So the internal machinations TEMPEST uses to modify the corresponding coordinates (like making sure they're between -pi/2 and +pi/2) don't work properly, and the grid validation step fails. If I remove the the hard-wired conversion, the regrid creation works fine. And I checked -- the original oEC60to30 SCRIP file was in degrees, while v2 and v3 has all coordinates in radians.

I'll report this to the TEMPEST developers and close this issue.

[jonbob]

OK, this issue is being tracked in tempest #27. Closing it as an MPAS issue.

[mark-petersen]

Great detective work, @jonbob! It's great to know it is just a validation step. The radians/degrees conversion has caught me before in other codes. Thanks!