Ccpp release 3 work #100

grantfirl · 2019-05-14T17:38:44Z

This PR constitutes initial preparation for the third CCPP release. The major work was to:

fix how ozone and h2o production and loss was handled (issues Clean up how ozone and h2o data is handled #70 and More cleanup for ozone, h2o, aerosols, and ICCN #94; notes in comment below)
work with the GSD suite (notes in comment below)
update the code to work with the latest ccpp-framework and ccpp-physics
add a python run script (written by @climbfuji with mods by @grantfirl)
open logfiles for writing in gmtb_scm
update the example Theia batch script to SLURM

Other notes:

moved physics namelists from scm/etc/case_config directory to ccpp/physics_namelists directory since they are more closely associated with the physics SDFs (also updated to latest corresponding values from FV3)
SDFs in ccpp/suites correspond to GFS_v15, GFS_v15plus, CPT_v0, and GSD_v0 (plus *_prescribed_surface versions of each)
the NETCDF environment variable must be set prior to running cmake (changed from using fragile NETCDF installation cmake finder module)
ccpp_init() call in scm/src/gmtb_scm.F90 expects only the suite name as an argument now (due to changes in ccpp-framework)
many changes to scm/src/GFS_typedefs.F90 and gmtb_scm_type_defs.f90 to account for changes to ccpp-physics since the last update and the fractional landmask (note below)

New method for running:
In the run directory, execute:
./run_gmtb_scm.py -c CASE_NAME [-s suite_name] [-n physics_namelist_path] [-g flag to invoke through gdb]

the case_name is the only required argument
if no suite_name is specified, the default value (SCM_GFS_v15) is used
specified suite names must correspond to the names of the suites in the SDFs
a default_namelists.py file exists to map SDFs->input namelists, but a user can specify a non-default namelist
the output directory path is created from the case name and the suite name automatically unless specified in the case_config file
there is no need to specify *_prescribed_surface in the suite name; if sfc_flux_spec is true in the case configuration file, a suite with *_prescribed_surface will automatically be substituted

…n oz_phys or oz_phys_2015; add ozone and h2o data files to repo

… call now expects the name of a suite (rather than a path/filename), modified case_configurations namelists to contains names of suites in the physics_suite variable; modified the string passed into ccpp_init() in gmtb_scm.F90; added code to link SDFs from the directory where they are stored to the run directory

…efs.F90

…uire that NETCDF environment variable is set (and not using custom CMake find); add SIONLIB to CMakeLists; change me and master to 0 (from 1) to work with SIONLIB stuff in Thompson init

…riables to GFS_typedefs for cellular automata, changes to Init_parm, and lheatstrg

…d SCM host metadata to work with latest ccpp-physics

… couple of compilation errors in GFS_typedefs.F90; fix type in gmtb_scm_physical_constants.f90

…ove 'hrrr' from GSD suite; uncomment Tiedtke scheme from prebuild

…s; rename/modify input namelists to reflect values in FV3 and use new suite names; rename and modify case configuration files to use new namelists and suite names; edit 'physics_test' plot configuration file; add MYNN section to GFS_typedefs.F90/control_initialize; add section to gmtb_scm_type_defs.f90/physics_associated for setting fractional landmask fractions

…named case configuration files for arm_sgp_summer_1997_A as a test of prescribed surface suites; added plot_config file to plot

…est code

… experiment_config since it contains both information for setting up a case and information for specifiying physics; associated minor code changes to handle path changes

…appen in a separate PR

grantfirl · 2019-05-20T19:13:00Z

Associated PR:
NCAR/ccpp-physics#247

grantfirl · 2019-05-20T19:20:29Z

Notes on ozone/h2o forcing work:

staged 3 data files related to ozone and h2o production and loss (from FV3) into scm/data/GFS_physics_data
updated GFS_typedefs.F90 to handle ozone and h2o forcing data like FV3 (also links the appropriate ozone file depending on the ozone scheme chosen)
simplified gmtb_scm_setup.F90/GFS_suite_setup for ozone (removed code that was no longer needed)
GFS_phys_time_vary_init and GFS_phys_time_vary_run now work like FV3

Note: Since the ozone/h2o production/loss data now works like FV3 (passed via internal module variables), this can only work with multiple independent SCM columns (potentially using different physics) if all independent columns are using the same ozone scheme (and the same ozone production/loss data)

grantfirl · 2019-05-20T19:53:57Z

Notes on working with GSD suite:

On @grantfirl machine, calculating tables for the Thompson MP scheme was taking > 1 day of wall clock time to complete. Since @climbfuji updated the code to be able to read previously calculated data tables for the Thompson MP scheme, reading the tables in (either the binary files or the SION-created file) works much more quickly for running with the Thompson MP scheme in the SCM.
SIONlib support was added through adding a stanza to scm/src/CMakeLists.txt; to use SIONlib with SCM, set the SIONLIB environment variable to the path where it is installed prior to running cmake (also copy the .sl file for the Thompson tables into the bin directory before running); address issue Enable SIONlib support in SCM #44
If not using SIONlib, binary files (generated on Theia with ifort v18) are available for use with the Thompson scheme. The following files should be downloaded into scm/data/GFS_physics_data prior to running cmake:
https://dtcenter.org/GMTB/freezeH2O.dat
https://dtcenter.org/GMTB/qr_acr_qg.dat
https://dtcenter.org/GMTB/qr_acr_qs.dat

grantfirl · 2019-05-20T20:11:53Z

The fractional landmask changes from FV3 were integrated into the SCM code. These changes will be reverted in an upcoming PR. This will affect many SDFs, scm/src/GFS_typedefs.F90, scm/src/gmtb_scm_type_defs.f90 (both CCPP metadata and a section in physics%associate that set the fractional landmask fractions).

UPDATE on 5/29/19: these changes will not be reverted anymore.

…te_type

…LIST

…tb-scm into grant_ccpp_release_3_work_dom_mods

…ts; edited ccpp_prebuild_config.py and CCPP metadata to correctly pass in needed variables to Thompson MP to turn ltaerosol on without crashing

Python model run script to simplify Fortran code

…iguration; the run script must be run with a case name; if a suite is specified, it is used, otherwise, a default SDF is chosen; suites run with a default namelist unless specified; the output_dir is constructed automatically unless specified in the case configuration file

…sed in GFS_typedefs.F90 and module_gfdl_cloud_microphys.F90); fixes writing of 'fort.2' to run dir

…f, det_mf since SCM is outputting thise

…for that suite

grantfirl · 2019-05-29T21:59:50Z

scm/etc/gmtb_scm_ens.py

@@ -11,8 +11,8 @@
 import sys


this file will need to be updated to work with the new run script if we want to release this functionality.

Will this be a follow-up PR?

Yes, although timing (before/after release) is TBD. A followup PR that will be useful for restoring this functionality in a cleaner way (using your run script) is forthcoming.

climbfuji · 2019-05-30T17:10:55Z

ccpp/physics_namelists/input_GSD_v0.nml

@@ -1,15 +1,17 @@
 &gfs_physics_nml
       fhzero         = 6.
       h2o_phys       = .true.
-       ldiag3d        = .false.
-       fhcyc          = 24.
+       ldiag3d        = .true.


I am not exactly sure why this is needed, but either way it's not a big deal (unallocated variables otherwise?)

Ya, the SCM output expects that tendencies are allocated, and they are only allocated in GFS_typedefs if ldiag3d is true.

climbfuji · 2019-05-30T17:14:40Z

ccpp/physics_namelists/input_GSD_v0.nml

-       lradar         = .false.
+       ltaerosol      = .true.
+       lradar         = .true.
+       ttendlim       = -999.


The best value we have at GSD at the moment is between 0.005 and 0.008. The FV3 regression tests are using 0.005 since the last PR was merged (https://github.com/NCAR/NEMSfv3gfs/pull/166). Not sure what we should do, keep it as is because it's apparently working without the limiter turned on (have you tried multiple test cases with the GSD suite)? Physically more correct is -999. because it doesn't modify the answer coming out of the microphysics.

I think that it should probably be consistent with whatever is used in FV3 going forward, although it might be worth a special note in the namelist saying that it can be turned off with -999. I did run all cases with it off, and it didn't crash, although I didn't look carefully at the output either. I'll change it in a followup PR to be consistent with FV3 and other namelist changes if they're coming down the pike.

scm/src/run_gmtb_scm.py

climbfuji

Great work, only a few minor comments that shouldn't prevent this PR from going in. See also NCAR/ccpp-physics#247 (review).

Ccpp release 3 work

grantfirl added 11 commits May 1, 2019 21:45

simplified ozone, h2o, aerosol, and iccn handling

2d2d27d

add code to link in the correct ozone prod/loss data file depending o…

a72f692

…n oz_phys or oz_phys_2015; add ozone and h2o data files to repo

clean up arguments to physics_create subroutine

e6cd31c

Merge branch 'fix_ozone' into ccpp_release_3_work

b7babad

change hard linking of ozone and h2o files to symlinking in GFS_typed…

277f288

…efs.F90

updating GSD suite SDF and input namelist; updating CMakeLists to req…

e2315c3

…uire that NETCDF environment variable is set (and not using custom CMake find); add SIONLIB to CMakeLists; change me and master to 0 (from 1) to work with SIONLIB stuff in Thompson init

update .gitignore to ignore large Thompson data table files

6ae02ab

updated gmtb_scm to work with commit 2aa966 in ccpp-physics; added va…

714c2ba

…riables to GFS_typedefs for cellular automata, changes to Init_parm, and lheatstrg

updated GFS_typedefs, ccpp_prebuild_config.py, physical constants, an…

ddde256

…d SCM host metadata to work with latest ccpp-physics

update SDFs to use sfc_diff instead of sfc_ex_coef (as in FV3); fix a…

ecb58e8

… couple of compilation errors in GFS_typedefs.F90; fix type in gmtb_scm_physical_constants.f90

grantfirl added work in progress do not merge labels May 14, 2019

grantfirl requested review from climbfuji and llpcarson as code owners May 14, 2019 17:38

grantfirl mentioned this pull request May 14, 2019

simplified ozone, h2o, aerosol, and iccn handling #99

Closed

grantfirl added 6 commits May 14, 2019 12:06

remove CCPP_typedefs (not needed if not using GFDL fast physics); rem…

9ef8a8d

…ove 'hrrr' from GSD suite; uncomment Tiedtke scheme from prebuild

renamed and edited SDFs for prescribed surface suites; updated and re…

a0b9f00

…named case configuration files for arm_sgp_summer_1997_A as a test of prescribed surface suites; added plot_config file to plot

updated the rest of the case configuration files to work with the lat…

fd4c1ca

…est code

moved physics namelists to scm/ccpp directory; renamed case_config to…

7137964

… experiment_config since it contains both information for setting up a case and information for specifiying physics; associated minor code changes to handle path changes

revert change to version 3.0.0 from 2.0.0 since version change will h…

67c090b

…appen in a separate PR

grantfirl mentioned this pull request May 20, 2019

Version update only #101

Merged

grantfirl removed do not merge work in progress labels May 20, 2019

grantfirl mentioned this pull request May 20, 2019

Ccpp release 3 work NCAR/ccpp-physics#247

Merged

climbfuji and others added 10 commits May 22, 2019 16:22

scm/src/gmtb_scm_type_defs.f90: remove physics_suite_dir from scm_sta…

e7f1942

…te_type

Adjust spacing for commented line in scm/src/gmtb_scm_input.f90

1a0931c

scm/src/run_gmtb_scm.py: improve comment for STANDARD_EXPERIMENT_NAME…

8eda7d5

…LIST

Merge branch 'ccpp_release_3_work' of https://github.com/grantfirl/gm…

beff8b5

…tb-scm into grant_ccpp_release_3_work_dom_mods

updated input namelists and SDFs to match latest NEMSfv3gfs equivalen…

a0e3198

…ts; edited ccpp_prebuild_config.py and CCPP metadata to correctly pass in needed variables to Thompson MP to turn ltaerosol on without crashing

Merge pull request #1 from climbfuji/grant_ccpp_release_3_work_dom_mods

7d3eb89

Python model run script to simplify Fortran code

replace experiment_config with case_config in several files

c4065bf

open a logfile for each column for writing in the output directory (u…

6ea8677

…sed in GFS_typedefs.F90 and module_gfdl_cloud_microphys.F90); fixes writing of 'fort.2' to run dir

update example batch script to run with slurm for Theia

8adb1a7

grantfirl force-pushed the ccpp_release_3_work branch from 9018bde to 8adb1a7 Compare May 29, 2019 20:18

grantfirl added 3 commits May 29, 2019 14:24

get rid of unused variales in gmtb_scm_input.f90

fdbbe27

remove comments for allocation/initialization of dq3dt, upd_mf, dwn_m…

c366f4a

…f, det_mf since SCM is outputting thise

set ldiag3d to T for GSD_v0 suite to avoid write error in SCM output …

2011af0

…for that suite

grantfirl commented May 29, 2019

View reviewed changes

remove repeated value from GFS_v15 namelists

fd2b233

climbfuji reviewed May 30, 2019

View reviewed changes

scm/src/run_gmtb_scm.py Show resolved Hide resolved

climbfuji reviewed May 30, 2019

View reviewed changes

scm/src/run_gmtb_scm.py Show resolved Hide resolved

climbfuji approved these changes May 30, 2019

View reviewed changes

update comments in run_gmtb_scm.py

0636f95

grantfirl merged commit 95140c7 into NCAR:master May 30, 2019

This was referenced May 31, 2019

Clean up how ozone and h2o data is handled #70

Closed

More cleanup for ozone, h2o, aerosols, and ICCN #94

Closed

Enable SIONlib support in SCM #44

Closed

grantfirl deleted the ccpp_release_3_work branch June 11, 2019 21:37

grantfirl restored the ccpp_release_3_work branch June 11, 2019 21:37

grantfirl deleted the ccpp_release_3_work branch June 11, 2019 21:37

dustinswales pushed a commit to dustinswales/ccpp-scm that referenced this pull request May 16, 2022

Merge pull request NCAR#100 from grantfirl/ccpp_release_3_work

28b9d5d

Ccpp release 3 work

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Ccpp release 3 work #100

Ccpp release 3 work #100

grantfirl commented May 14, 2019 •

edited

Loading

grantfirl commented May 20, 2019

grantfirl commented May 20, 2019 •

edited

Loading

grantfirl commented May 20, 2019 •

edited

Loading

grantfirl commented May 20, 2019 •

edited

Loading

grantfirl May 29, 2019

climbfuji May 30, 2019

grantfirl May 30, 2019

climbfuji May 30, 2019

grantfirl May 30, 2019

climbfuji May 30, 2019

grantfirl May 30, 2019

climbfuji left a comment

Ccpp release 3 work #100

Ccpp release 3 work #100

Conversation

grantfirl commented May 14, 2019 • edited Loading

grantfirl commented May 20, 2019

grantfirl commented May 20, 2019 • edited Loading

grantfirl commented May 20, 2019 • edited Loading

grantfirl commented May 20, 2019 • edited Loading

grantfirl May 29, 2019

Choose a reason for hiding this comment

climbfuji May 30, 2019

Choose a reason for hiding this comment

grantfirl May 30, 2019

Choose a reason for hiding this comment

climbfuji May 30, 2019

Choose a reason for hiding this comment

grantfirl May 30, 2019

Choose a reason for hiding this comment

climbfuji May 30, 2019

Choose a reason for hiding this comment

grantfirl May 30, 2019

Choose a reason for hiding this comment

climbfuji left a comment

Choose a reason for hiding this comment

grantfirl commented May 14, 2019 •

edited

Loading

grantfirl commented May 20, 2019 •

edited

Loading

grantfirl commented May 20, 2019 •

edited

Loading

grantfirl commented May 20, 2019 •

edited

Loading