-
Notifications
You must be signed in to change notification settings - Fork 24
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix several issues in file regridding; Also prevent double-flipping of vertical indices #270
Conversation
gcpy/file_regrid.py - Renamed "fin" to "filein", to match the command-line argument - Renamed "fout" to "fileout", to match the command-line argument - Added verbose printout of arguments - Remove nargs=1 from weightsdir and verbose, which turns them into a list of 1 element Signed-off-by: Bob Yantosca <[email protected]>
gcpy/file_regrid.py - Removed a double-flipping of vertical levels in LL -> CS/SG regridding gcpy/regrid_restart_file.py - Removed function check_lev_attribute, it is not defined Signed-off-by: Bob Yantosca <[email protected]>
gcpy/file_regrid.py - Removed several debug printout statements at about line 1333 Signed-off-by: Bob Yantosca <[email protected]>
gcpy/util.py - Routine "rename_and_flip_gchp_rst_vars" now uses the inquiry function "is_cubed_sphere_rst_grid" from cstools.py to determine if the dataset is from a GCHP restart file. If not, the original dataset will be returned unchanged. - Updated Pydoc and comments - Updated algorithm in "rename_and_flip_gchp_rst_vars" so that we create a dictionary of variable name replacements so that the replacement can be done as a single operation. This is more efficient. - If we have flipped the dataset, also set lev.positive="up" gcpy/benchmark_funcs.py gcpy/examples/diagnostics/compare_diags.py gcpy/examples/plotting/plot_comparisons.py gcpy/examples/plotting/plot_single_panel.py - Remove external tests for the GCHP restart grid before calling the "rename_and_flip_gchp_rst_vars" function from util.py. The function can now detemrine internally whether or not the passed dataset is from a GCHP restart file. If not the original data will be returned unchnaged. - Updated comments for clarity. Signed-off-by: Bob Yantosca <[email protected]>
gcpy/grid.py - Remove unused import statement - Added "gchp_indices" argument to get_lev_coords and get_ilev_coords functions. If gchp_indices=True, the lev/ilev coordinates will be returned as an array of indices (as is the custom for GCHP netCDF files). - Updated comments and Pydoc Signed-off-by: Bob Yantosca <[email protected]>
gcpy_file_regrid.py - Move flip_lev_coord_if_necessary so that it occurs before variable renaming/regridding. This prevents the flipping algorithm from being spoofed by renamed variables. - Renamed "drop_and_rename_classic_variables" to "drop_classic_vars", which only drops the non-regriddable variables. - Move variable renaming into "rename_restart_variables". Also improve logic so that variables are renamed properly - Updated logic in the flip_lev_coord_if_necessary routine. Also make sure to save coordinates in the proper order. GCHP level coordinates are now saved as indices. - Updated Pydoc and comments. - Trimmed trailing whitespace. gcpy/util.py - Initialize "old_to_new" to a blank dictionary object Signed-off-by: Bob Yantosca <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This all looks good to me except for the stretched grid. I suggest we do not allow stretched grid regridding using online weights until it is resolved. Using offline weights works so we can stick with that.
I am using this to create a GCHP restart from GC-Classic and noticed this output 6 times: Is this needed? Could we suppress it? |
gcpy/file_regrid.py - Because there is an issue in regridding to stretched-grids, we now throw a RuntimeError if sg_params_in or sg_params_out does not match the default (i.e. non-stretched) values. Signed-off-by: Bob Yantosca <[email protected]>
As suggested by @lizziel, we now prevent regridding to/from stretched-grids for the time being, as the stretched-grid plots above indicate some sort of error (which may be either regridding or plotting). Revisit this sometime in the future. The error looks like: Traceback (most recent call last):
File "/n/holyscratch01/jacob_lab/ryantosca/data/Restarts_Out/./rst.py", line 216, in <module>
main()
File "/n/holyscratch01/jacob_lab/ryantosca/data/Restarts_Out/./rst.py", line 155, in main
file_regrid(
File "/n/home09/ryantosca/python/gcpy/gcpy/file_regrid.py", line 111, in file_regrid
raise RuntimeError(msg)
RuntimeError: Regridding to or from cubed-sphere stretched grids is
currently not supported. Please use the offline regridding
method described in the Regridding section of gcpy.readthedocs.io. |
Thanks @lizziel. I think that is coming in from ESMF. I am not sure how to suppress it. It doesn't appear to be generated by xESMF (at least as far as I can tell from looking through the source code). |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Look good!
Update: the "Using dimensions..." warnings is in xESMF (see frontend.py). I have tried to disable the warnings but somehow they still come up. |
Overview
This PR fixes several remaining issues with GCPy file regridding using online weights.
Routine
rename_and_flip_gchp_rst_vars.py
now uses inquiry functionis_cubed_sphere_rst_grid
internally. If the data is on the GCHP checkpoint/restart grid, then data variables will be renamed according to GCHP conventions and vertical levels will be flipped so that it is positive down. Otherwise the original data will be returned unchanged. This was necessary to prevent a bug in which vertical levels were flipped twice incompare_diags.py
and the example plotting scriptsplot_comparisons.py
andplot_single_level.py
.Added the boolean argument
gchp_indices
to routinesget_ilev_coords
andget_lev_coords
. If true, then the coordinates will be returned as indices (e.g. 1..72), which is used for the GCHP level dimension.Fixed several logic issues in
file_regrid.py
:a. Move
flip_lev_coord_if_necessary
so that it occurs before variable renaming/regridding. This prevents the flipping algorithm from being spoofed by renamed variables.b. Renamed
drop_and_rename_classic_variables
todrop_classic_vars
, which only drops the non-regriddable variables.c. Move variable renaming into
rename_restart_variables
. Also improve logic so that variables are renamed properlyd. Updated logic in the
flip_lev_coord_if_necessary
routine. Also make sure to save coordinates in the proper order. GCHP level coordinates are now saved as indices.Fixed issues with command-line arguments:
a. Renamed Python argument
fin
tofilein
, to match the command-line argumentb. Renamed Python argument
fout
tofileout
, to match the command-line argumentc. Added verbose printout of arguments
d. Remove
nargs=1
fromweightsdir
andverbose
, which turns them into a list of 1 element (instead of a scalar, which is what it should be)Output from
file_regrid.py
is shown below.TODO: Double-check the
Regridding.rst
instructions to make sure they are consistent with this PR. Further documentation fixes will be pushed todocs/dev
.Common use cases
GCClassic to GCHP checkpoint
GCClassic to GCClassic
GCHP checkpoint to GCHP checkpoint
GCHP checkpoint to GCHP checkpoint, stretched grid:
NOTE: Not sure if this is a plotting issue or a regridding issue.
GCHP checkpoint to GCClassic
Less common use cases (shown for completeness)
I wanted to show these cases, just to show that the regridding seems to work OK. Some editing of variable names may still be needed. In any case, these are not the usual use case so I think it's OK.
GCHP checkpoint to GCHP diagnostic
GCHP diagnostic to GCHP checkpoint
GCHP diagnostic to GCHP diagnostic
GCClassic to GCHP diagnostic
GCHP diagnostic to GCClassic