Update master branch to use master branches of ccpp-physics and ccpp-framework #185

grantfirl · 2020-07-10T16:27:48Z

This PR allows the SCM to work with the latest master branches of ccpp-physics and ccpp-framework. It also contains changes to work with HWRF physics, which will presumably be going to the master branches of the CCPP soon too.

Most of the changes are to facilitate new/different variables in the physics, although a few of the commits in July (specifically 75f3475) change the way multiple independent columns could be used. Rather than have multiple instances of GFS DDTs holding the different columns (and potentially allowing different physics suites to be run in each of them simultaneously), the horizontal dimension of the individual variables is used instead. E.g. instead of physics%Statein(i)%tgrs(1,k), you now have physics%Statein%tgrs(i,k). This closes the door on the potential to run multiple physics suites simultaneously, but this wasn't being done in practice due to module-level data anyway.

This PR also marks the ability to use the updated tendencies, although the non-physics tendency is not working due to memory management. This can be enabled in a future PR.

Lastly, the GFS_DCNV_generic_pre routines have been rearranged in many SDFs.

…d a namelist that exercises the hurricane version of moninedmf

…ting with Weiguo Wang

…e; add a multi_run configuration file for the using the hurricane PBL namelist for all supported cases

…H LSM

…edefs

…e initialized to something other than 0 in gfdl_sfc_layer)

… HWRF LSM and surface layer for testing

…rk (temporary)

…-framework (temporary)" This reverts commit de084cb.

…th latest dtc/develop branch; update submodule pointers

Update dtc/develop branch to work with ccpp-physics/framework as of 20200608

…se new phys_tend scheme and MYNN-related changes; add new MYNN-related variables, new tendencies, and new aux[2,3]d variables to GFS_typedefs.[meta,F90]; edit gmtb_scm.F90 to call output_init after physics are initialized and to be able to calculate non-physics tendencies; edit do_time_step to calculate non-physics tendencies; edit gmtb_scm_output to write out new tendencies and aux variables

…_analysis.py to read new tendency variables; add new supported_suites.ini plotting configuration file

…k with the latest CCPP framework; update optional arguments to match FV3

…amework

…e the way more than one column is implemented -- use the horizontal_dimension of physics variables rather than having arrays of GFS DDTs with horizontal extent

climbfuji

Thanks for all the cleanup work, glad to see that the new ccpp-framework code is working as expected in SCM as well (when no blocking is used). I left several comments, but happy to approve if you think that it is not worth addressing them now.

climbfuji · 2020-07-10T21:23:10Z

scm/src/GFS_typedefs.F90

@@ -113,6 +116,7 @@ module GFS_typedefs
    real(kind=kind_phys) :: dt_phys              !< physics  time step in seconds
 !--- restart information
    logical :: restart                           !< flag whether this is a coldstart (.false.) or a warmstart/restart (.true.)
+    logical :: cycling                           !< flag whether this is a coldstart (.false.) or a cycled run (.true.)


Note that this will be removed again in fv3atm in the near future, see NOAA-EMC/fv3atm#140.

OK, thanks. No one would (should) ever change that from the default false value when running the SCM anyway. It was only included to minimize differences between the SCM/FV3 GFS_typedefs.

climbfuji · 2020-07-10T21:25:42Z

scm/src/GFS_typedefs.F90

+      write(0,*) "Error, the SCM has not implemented time averaging of diagnostics in GFS_typedefs.F90"
+      stop
+    end if
+


How is the time averaging done in SCM? The tendencies are accumulated in ccpp-physics. Are you subtracting the previous accumulated value after each time step?

diag_phys_zero is called every time step and right before it is written out, these accumulated change variables are divided by the timestep to get the tendencies.

climbfuji · 2020-07-10T21:26:15Z

scm/src/GFS_typedefs.F90

@@ -3290,6 +3537,15 @@ subroutine control_initialize (Model, nlunit, fn_nml, me, master, &
    Model%fhlwr            = fhlwr
    Model%nsswr            = nint(fhswr/Model%dtp)
    Model%nslwr            = nint(fhlwr/Model%dtp)
+    if (restart) then


Have you tried this in SCM? What differences do you see? Would be interesting to check (not for merging this PR, though).

Nope. Restarts are not implemented in the SCM since they are so short to run. This is another GFS_typedefs variable that is only there to minimize differences between FV3/SCM. There should probably be a list of namelist and other vars to not change from their default in the SCM.

climbfuji · 2020-07-10T21:28:42Z

scm/src/gmtb_scm.F90

@@ -92,66 +90,68 @@ subroutine gmtb_scm_main_sub()
    case default
      cdata_time_index = 2
  end select
+


Nice, this cleans up the code quite a bit.

climbfuji · 2020-07-10T21:29:42Z

scm/src/gmtb_scm_output.F90

@@ -219,6 +221,31 @@ subroutine output_init(scm_state)
  CALL CHECK(NF90_PUT_ATT(NCID=ncid,VARID=dummy_id,NAME="description",&
     VALUES="temperature tendency due to microphysics scheme"))
  CALL CHECK(NF90_PUT_ATT(NCID=ncid,VARID=dummy_id,NAME="units",VALUES="K s-1"))
+  CALL CHECK(NF90_DEF_VAR(NCID=ncid,NAME='dT_dt_ogwd',XTYPE=NF90_FLOAT,DIMIDS= (/ hor_dim_id, vert_dim_id, time_id /), &


It would be nice if we could suppress the output of the 3d tendencies if ldiag3d/qdiag3d are false.

What about if we just used a netCDF missing value if the logicals are false? Right now, it outputs zeros if they are false, but I suppose that could be misleading. I prefer to keep the variables actually in the output file so that scripts don't need to check for variables existing or not. I guess they would need to check for missing values instead though.

climbfuji · 2020-07-10T21:30:05Z

scm/src/gmtb_scm_output.F90

@@ -318,8 +430,23 @@ subroutine output_init(scm_state)
  CALL CHECK(NF90_DEF_VAR(NCID=ncid,NAME='init_hour',XTYPE=NF90_FLOAT,VARID=hour_id))
  CALL CHECK(NF90_PUT_ATT(NCID=ncid,VARID=hour_id,NAME="description",VALUES="model initialization hour"))
  CALL CHECK(NF90_PUT_ATT(NCID=ncid,VARID=hour_id,NAME="units",VALUES=""))
-
-
+


We could do the same for the 3d tendencies, or?

Sure. I could have an array of variable names and descriptions and shorten the code with a loop for sure. At some point, I was hoping to reimplement the output for the SCM though, so I'm not sure that it is worth it right now. I'd like to implement something where the users could specify which variables to write out by their CCPP standard names, implement time-averaging (instead of only instantaneous), etc.

climbfuji · 2020-07-10T21:31:04Z

scm/src/gmtb_scm_output.F90

  do i=1, scm_state%n_cols
-    dummy_1D(i) = physics%Interstitial(i)%dusfc1(1)
+    dummy_2D(i,:) = physics%Diag%dt3dt(i,:,1)/scm_state%dt


I see, this is where the time averaging happens for the 3d tendencies.

climbfuji · 2020-07-10T21:32:21Z

scm/src/gmtb_scm_output.F90

  end do
  CALL CHECK(NF90_INQ_VARID(NCID=ncid,NAME="lw_dn_sfc_clr",VARID=var_id))
  CALL CHECK(NF90_PUT_VAR(NCID=ncid,VARID=var_id,VALUES=dummy_1D,START=(/1,scm_state%itt_out /)))
-
+


You could use the time_avg logical arrays in the same way as it is done in fv3atm - if it is .true. for a specific slice of the array, simply divide by the integration time (same as you do for the 3d tendencies).

OK, I'll look into that for the output rewrite.

grantfirl · 2020-07-10T22:16:15Z

Thanks for all the cleanup work, glad to see that the new ccpp-framework code is working as expected in SCM as well (when no blocking is used). I left several comments, but happy to approve if you think that it is not worth addressing them now.

Thanks for the comments and review. I'd like to go ahead and merge as-is and do a rewrite of the output routines later (which I had already planned on doing at some point anyway) taking your suggestions into account then. Mike Iacono is waiting for these changes and I need a good base to start some other development.

Update master branch to use master branches of ccpp-physics and ccpp-framework

grantfirl added 30 commits December 10, 2019 12:22

add hurricane-specific variables for moninedmf to GFS_typedefs and ad…

5faeca8

…d a namelist that exercises the hurricane version of moninedmf

apply non-default value of moninq_fac (0.7) in namelist for HWRF PBL

8e38fc1

change value of moninq_fac for HWRF_PBL namelist to -1.0 after consul…

75492b9

…ting with Weiguo Wang

Merge branch 'dtc/develop' into biswas-hafs_gjf

03432b7

rename physics namelist that uses hurricane flag in hybrid EDMF schem…

3d2c920

…e; add a multi_run configuration file for the using the hurricane PBL namelist for all supported cases

add new PBL diagnostic for satmedmfvdifq

4e390e7

add variables, SDFs, namelist, and data files for HWRF version of NOA…

2c0ff62

…H LSM

update GFS_typedefs.F90 to be in sync with FV3's version

fda4986

change name of physical constant and rearrange a few lines in GFS_typ…

95bdd90

…edefs

edit GFS_typedefs.F90/meta for HWRF RRTMG

fe3010a

add HWRF RRTMG to ccpp_prebuild_config

680691c

Merge branch 'hwrf_rrtmg' into noah_hafs2

ca2df0a

change extension on module_sf_noahlsm.F[90] in ccpp_prebuild_config

8102c0d

add snotime as a sfcprop variable

2b3a978

update GFS_typedefs for sflx_glacial and add HWRF SDF/namelist

facdcee

move sthick

472f8a2

change away from hafs name

95a6986

add variables to GFS_typedefs for HWRF SAS

b8ed24f

Merge branch 'dtc/hwrf-physics' into noah_hafs2

fb8ee19

initial changes for GFDL surface layer scheme

5dc86bf

add roughness length to non-prescribed surface flux cases (needs to b…

18c433d

…e initialized to something other than 0 in gfdl_sfc_layer)

add namelist options for GFDL surface layer; add suites that combines…

e614bc0

… HWRF LSM and surface layer for testing

Merge branch 'gfdl_sfc_layer' into hwrf_sfc

afc899e

Merge branch 'master' into hwrf_sfc

ac57ca3

update GFS_typedefs to work with latest dtc/hwrf-physics

e5bc573

revert changes to CMakeLists.txt to use older version of ccpp-framewo…

de084cb

…rk (temporary)

Revert "revert changes to CMakeLists.txt to use older version of ccpp…

43c4570

…-framework (temporary)" This reverts commit de084cb.

update GFS_typedefs.[F90/meta] and ccpp_prebuild_config.py to work wi…

227ac70

…th latest dtc/develop branch; update submodule pointers

remove hurricane specific files

2d6bf80

remove whitespace

bacbc93

grantfirl added 11 commits June 8, 2020 20:52

Merge pull request NCAR#183 from grantfirl/update_dtc_develop_20200608

cb1dd17

Update dtc/develop branch to work with ccpp-physics/framework as of 20200608

add phys_tend scheme to end of SDFs

fcf0d95

update optional arguments in ccpp_prebuild_config.py; update gmtb_scm…

7a72116

…_analysis.py to read new tendency variables; add new supported_suites.ini plotting configuration file

update ccpp_prebuild_config.py to remove pset to match FV3 and to wor…

8559fb1

…k with the latest CCPP framework; update optional arguments to match FV3

Merge branch 'gsd/develop' into update_master_20200706

49f37ad

update GFS_typedefs.[F90/meta] to match FV3 changes to master

27798da

attempt to change gmtb_scm_type_defs.meta to match changes to CCPP fr…

fcfce46

…amework

update submodule pointers

848f578

remove possibility to run SCM columns with independent physics; chang…

75f3475

…e the way more than one column is implemented -- use the horizontal_dimension of physics variables rather than having arrays of GFS DDTs with horizontal extent

update ccpp-physics submodule pointer

dba6881

grantfirl requested review from climbfuji and llpcarson as code owners July 10, 2020 16:27

grantfirl added the work in progress label Jul 10, 2020

change order of GFS_DCNV_generic_pre in many SDFs

30fe019

grantfirl removed the work in progress label Jul 10, 2020

climbfuji reviewed Jul 10, 2020

View reviewed changes

climbfuji approved these changes Jul 12, 2020

View reviewed changes

grantfirl merged commit ac0765b into NCAR:master Jul 13, 2020

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

Merge pull request NCAR#185 from grantfirl/update_master_20200706

ae5ba05

Update master branch to use master branches of ccpp-physics and ccpp-framework

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update master branch to use master branches of ccpp-physics and ccpp-framework #185

Update master branch to use master branches of ccpp-physics and ccpp-framework #185

grantfirl commented Jul 10, 2020 •

edited

Loading

climbfuji left a comment

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020 •

edited

Loading

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020

climbfuji Jul 10, 2020

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020

climbfuji Jul 10, 2020

climbfuji Jul 10, 2020

grantfirl Jul 10, 2020

grantfirl commented Jul 10, 2020

Update master branch to use master branches of ccpp-physics and ccpp-framework #185

Update master branch to use master branches of ccpp-physics and ccpp-framework #185

Conversation

grantfirl commented Jul 10, 2020 • edited Loading

climbfuji left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

grantfirl Jul 10, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

grantfirl commented Jul 10, 2020

grantfirl commented Jul 10, 2020 •

edited

Loading

grantfirl Jul 10, 2020 •

edited

Loading