add nc4-formatted tile file when generating bcs

[weiyuan-jiang]

This PR modifies make_bcs to create an NC4-formatted tile file, in addition to the ASCII-formatted tile files.

The NC4-formatted tile file also includes the supplemental tile attributes (min/max lat/lon and elevation of each tile) needed by GEOSldas, which are currently stored in the (land-only) "catchment.def" file. The supplemental tile attributes are now also computed for landice and lake tiles.

The PR also includes a py script and an F90 utility program to retroactively create NC4-formatted tile files for existing bcs.

Because the PR only touches make_bcs, it should be trivially 0-diff for the standard GCM and GEOSldas nightly tests.

The PR is also 0-diff for bcs generation (except for the version-specific README file generated by make_bcs).

In the future, the ASCII-formatted tile file and catchment.def file should become obsolete.

Contingencies: #909 must be merged first.

@tclune @biljanaorescanin @sdrabenh @lcandre2

[gmao-rreichle]

cc @gmao-rreichle

[weiyuan-jiang]

@gmao-rreichle It is ready for your review

[gmao-rreichle]

@weiyuan-jiang: To compute the supplemental attributes for landice and lake tiles, I suspect we need to modify the (renamed) subroutine supplemental_tile_attributes(). Specifically, I think the indices ip1 and ip2 appear to constrain the calculations to just land tiles. Perhaps it would not be difficult to calculate the supplemental attributes for landice and lake tiles. We do not want them for ocean tiles, and we do not want them written into the legacy "catchment.def" file. The idea is to have good values for min/max lat/lon and elevation of landice and lake tiles in the new, nc4-formatted tile file.

GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/GEOSsurface_GridComp/Utils/Raster/makebcs/rmTinyCatchParaMod.F90

Line 2138 in 19eb1cb

if((catid(i) > ip1).and.(catid(i) <= ip2))then

[weiyuan-jiang]

@weiyuan-jiang: To compute the supplemental attributes for landice and lake tiles, I suspect we need to modify the (renamed) subroutine supplemental_tile_attributes(). Specifically, I think the indices ip1 and ip2 appear to constrain the calculations to just land tiles. Perhaps it would not be difficult to calculate the supplemental attributes for landice and lake tiles. We do not want them for ocean tiles, and we do not want them written into the legacy "catchment.def" file. The idea is to have good values for min/max lat/lon and elevation of landice and lake tiles in the new, nc4-formatted tile file.

GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/GEOSsurface_GridComp/Utils/Raster/makebcs/rmTinyCatchParaMod.F90

Line 2138 in 19eb1cb

if((catid(i) > ip1).and.(catid(i) <= ip2))then

Is the min/max of lat/lon of landice and lake calculated by the same way as land ?

[gmao-rreichle]

Is the min/max of lat/lon of landice and lake calculated by the same way as land ?

I would think so. The information from the raster (*.rst) file, which is the basis for the min/max lat/lon calculations, isn't specific to land and should work for landice and lake in the same way.

[lcandre2]

Is the min/max of lat/lon of landice and lake calculated by the same way as land ?

I would think so. The information from the raster (*.rst) file, which is the basis for the min/max lat/lon calculations, isn't specific to land and should work for landice and lake in the same way.

I agree with Rolf. The lake and landice tile min and max should be calculated the same way as land. Like land, both are irregular.

[weiyuan-jiang]

Is the min/max of lat/lon of landice and lake calculated by the same way as land ?

I would think so. The information from the raster (*.rst) file, which is the basis for the min/max lat/lon calculations, isn't specific to land and should work for landice and lake in the same way.

I agree with Rolf. The lake and landice tile min and max should be calculated the same way as land. Like land, both are irregular.

OK, I will take care of it

[biljanaorescanin]

Testing summary:
1. PR was zero diff for BCS generation functionality.
I've run C180, EASE M09, C12 MOM6 and CF0270x6C-SG001 to confirm zero diff.
Only difference was as expected in README file. (This difference is really a README documentation bug fix. )
2. PR was zero diff for GEOSldas regression testing.
3. PR was zero diff for 1-day AMIP, REPLAY and Inc REPLAY.

@sdrabenh this PR is also ready for you. To be merged you must first merge #909. That is why I am leaving blocking label.
My v13 BCS PR will now come on top of this PR.

[biljanaorescanin]

We fixed all the labels so we are good to be merged.

The merge-base changed after approval.