TEMPO microphysics for WRF

[AndersJensen-NOAA]

Addition of TEMPO microphysics

TYPE: new feature

KEYWORDS: Microphysics, Operational models

SOURCE: Anders Jensen NOAA/GSL

DESCRIPTION OF CHANGES:
Problem:
Microphysics development of a Thompson-Eidhammer based scheme for operational applications has advanced to a submodule, TEMPO. This new development should be made available to WRF users.

Solution:
Connect TEMPO to WRF. A cleaned and refactored version of TEMPO is being tested in the UFS community's MPAS. Additionally, a PR, MPAS-Dev/MPAS-Model#1393, to add TEMPO to NCAR's MPAS is in progress. This current PR will allow the community to use this same microphysics across the UFS, MPAS, and WRF.

LIST OF MODIFIED FILES:
M .gitmodules
M main/depends.common
M Makefile
M clean
M phys/Makefile
M phys/module_physics_init.F
M phys/module_microphysics_driver.F
M Registry.EM_COMMON
M dyn_em/module_initialize_real.F

TESTS CONDUCTED:

1. compiled and ran em_quarter_ss
2. restart test with em_quarter_ss
3. bit-for-bit with different number of processes using em_quarter_ss.
4. Passed regression tests.

RELEASE NOTE:
TEMPO is the Thompson-Eidhammer Microphysics Parameterization for Operations. TEMPO lives here: https://github.com/NCAR/TEMPO and technical documentation lives here: https://ncar.github.io/TEMPO/.

TEMPO specific release notes are here: https://github.com/NCAR/TEMPO/releases/tag/tempo_v3.0.0
TEMPO version 3.0.0 is the current release version.

TEMPO is being tested in the ufs-community fork of MPAS for next-generation convection-allowing forecast systems (https://github.com/ufs-community/MPAS-Model). Thus, TEMPO is being continually evaluated for improvement.

Additionally, there is a PR in progress to connect TEMPO (v3.0.0) to NCAR's MPAS (MPAS-Dev/MPAS-Model#1393). The version of TEMPO currently in the CCPP (https://github.com/ufs-community/ccpp-physics) will be updated to v3.0.0 soonish. TEMPO is connected to WRF, MPAS, and CCPP as a submodule, and the same v3.0.0 code will be connected to WRF (this PR), MPAS, and CCPP.

In WRF, supported namelist options include tempo_aerosolaware==1 to use the aerosol-aware version, and tempo_hailaware==1 to use the hail-aware version. Both options are on by default.

Key improvements include:

• TEMPO is in a submodule that allows for flexibility when connecting TEMPO to dynamical cores
• Code modularity was been improved and technical documentation has been added
• Bux fix for cloud droplet number concentration during evaporation
• Surface emissions of water-friendly aerosols have been reduced based on a sensitivity simulation that aimed to better constrain water-friendly aerosol concentrations over 24-36 hour forecast periods and reduce the accumulation of aerosols in the atmosphere
• A few modifications were made to graupel number concentration tendencies (when using the hail-aware option) for processes that cause rain to freeze to increase reflectivity in deep convective storms
• Added simple sedimentation tests
• Added unit tests
• Added a program to build lookup tables with MPI capability
• Lookup tables have been unified

[weiwangncar]

@AndersJensen-NOAA Can you compile your code on your system with manage_externals? It is showing some errors doing the regression tests.

[AndersJensen-NOAA]

@AndersJensen-NOAA Can you compile your code on your system with manage_externals? It is showing some errors doing the regression tests.

@weiwangncar I think that I have the code now working with manage_externals pulling a specific tag of TEMPO.

[AndersJensen-NOAA]

@AndersJensen-NOAA Can you compile your code on your system with manage_externals? It is showing some errors doing the regression tests.

@weiwangncar I think that I have the code now working with manage_externals pulling a specific tag of TEMPO.

Actually, maybe not. Thanks for the email though -- I will check things again. Note that this compiles on my machine, but the regression tests are still failing.

[weiwangncar]

The regression tests have passed:

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 23           24
Number of Builds       : 60           57
Number of Simulations  : 158           150        0
Number of Comparisons  : 95           86        0

Failed Simulations are: 
None
Which comparisons are not bit-for-bit: 
None

[AndersJensen-NOAA]

@weiwangncar

Output from dmpar compile with intel and em_quarter_ss case with 1 versus 4 processors
Simulation time = 1 hour with output every 30 mins
Last timestep timing information:
4 processors: 0.05787 elapsed seconds
1 processor: 0.14520 elapsed seconds

[AndersJensen-NOAA]

Restart test: em_quarter_ss case, 2-h simulation with restart files written every hour compared to 1-h simulation started from a restart file at 1 hour:

[weiwangncar]

@AndersJensen-NOAA Thanks for doing these tests!

[AndersJensen-NOAA]

Might consider making it option 48 instead of 88, to keep it within the lower numbers.

@dudhia I realized that TEMPO is 88 in the CCPP, so it probably makes sense to keep those the same.

[dudhia]

I didn't think the CCPP had numbers. Is it the UFS option?

[AndersJensen-NOAA]

I didn't think the CCPP had numbers. Is it the UFS option?

Yes, UFS.

[weiwangncar]

@AndersJensen-NOAA Thanks for updating the git repository. I would still like to see a few key improvements outlined in the PR message, possibly under the RELEASE NOTE section. Thanks!

[AndersJensen-NOAA]

@AndersJensen-NOAA Thanks for updating the git repository. I would still like to see a few key improvements outlined in the PR message, possibly under the RELEASE NOTE section. Thanks!

@weiwangncar I have some RELEASE notes -- let me know if you have questions.

[dudhia]

I resolved conflicts related to being in the same places as MYNN SFC additions.

[weiwangncar]

@AndersJensen-NOAA The updated PR looks good. Thanks!

[weiwangncar]

@dudhia Is option number 88 or or should it be changed?

[dudhia]

Anders gave us a reason to keep 88.

[AndersJensen-NOAA]

Anders gave us a reason to keep 88.

@dudhia @weiwangncar
If you are okay with using 88, then I am as well. That way I don't have to change things in the UFS :). The number 88 is still in the spirit of Thompson/Thompson-Eidhammer (8, 28, 38).

[smileMchen]

@AndersJensen-NOAA

Would you please share the three tables: qr_acr_qs_data_tempo_v3, qr_acr_qg_data_tempo_v3, and freeze_water_data_tempo_v3? Thanks.

[weiwangncar]

@AndersJensen-NOAA Perhaps you can provide the instruction on how to generate the tables. Thanks!

[smileMchen]

@weiwangncar @@AndersJensen-NOAA

The instruction on code compiling is given in the code build_tables.F90.

The issue is that, running this program takes too much time. I am not sure whether it is normal.

I am trying to create the tables but the program has been running more than 1.5 hours and has not finished yet.

[weiwangncar]

@smileMchen I don't see a build_tables.F90 file. But I hear that the tables can be built with MPI. Is that what you are compiling and running the code?

[dudhia]

@weiwangncar Ming found that F90 file in the TEMPO/tools directory.

[AndersJensen-NOAA]

@weiwangncar @smileMchen

Info on table building can be found here:
https://ncar.github.io/TEMPO/program/build_tables.html

If you don't want to build tables, you can grab on from the tempo/tables directory:
tempo/tables/ccn_activate.bin

And then the three bigger tables are on a NOAA server, and can be obtained with wget:
wget -q https://gsl.noaa.gov/thredds/fileServer/retro/jensen/qr_acr_qg_data_tempo_v3
wget -q https://gsl.noaa.gov/thredds/fileServer/retro/jensen/qr_acr_qs_data_tempo_v3
wget -q https://gsl.noaa.gov/thredds/fileServer/retro/jensen/freeze_water_data_tempo_v3

@smileMchen I'm curious to know more about your build issue? Table building should not take 1.5 hours. Did you use intel or gnu?

[smileMchen]

@AndersJensen-NOAA

I compiled the code using gfortran v14.3.0.

It takes more than 1.5hour to finish running the program (I run interactively with a single processor). It would be much faster if we run with multi-processors, I guess?

[AndersJensen-NOAA]

@smileMchen: You can either use mpi or change the optimization in Makefile.gfortran. Changing -O0 to -O2 should give a large speed up.

[smileMchen]

@AndersJensen-NOAA

Thank you for the kind information.