Update to MYNN-EDMF

[joeolson42]

TYPE: bug fix, enhancement, and new feature

KEYWORDS: mass-flux, diffusivity, subgrid clouds, memory reduction

SOURCE: Joseph Olson (NOAA) with contributions from Jaymes Kenyon (NOAA/CIRES)

DESCRIPTION OF CHANGES:

• A series of small tweaks to help reduce a cold bias in the PBL:
  - slight increase in diffusion in convective conditions
  - relaxed criteria for mass-flux activation/strength
  - added capability to cycle TKE for continuity in hourly updating HRRR
  - added effects of compensational environmental subsidence in mass-flux scheme,
  which resulted in tweaks to detrainment rates.
• Bug fix for diagnostic-decay of SGS clouds - noticed by Greg Thompson. This has
  a very small, but primarily positive, impact on SW-down biases.
• Tweak to calculation of KPBL - urged by Laura Fowler - to make more intuitive.
• Tweak to temperature range of blending for saturation check (water to ice). This
  slightly reduces excessive SGS clouds in polar region. No impact warm clouds.
• Added namelist option bl_mynn_output (0 or 1) to suppress or activate the
  allocation and output of 10 3D variables. Most people will want this
  set to 0 (default) to save memory and disk space.
• Removed TKE_PBL for memory purposes - only outputting QKE
• Added the forcing of bl_mynn_output=0 when bl_mynn_edmf=0 in check_a_mundo

LIST OF MODIFIED FILES:
M Registry/Registry.EM_COMMON
M share/module_check_a_mundo.F
M dyn_em/module_first_rk_step_part1.F
M phys/module_bl_mynn.F
M phys/module_pbl_driver.F
M phys/module_physics_init.F

TESTS CONDUCTED:
Single case study looked good - similar to GSD's v3.9+ code. Here are some figure's that demonstrate some of the improvement in the upcoming version of the HRRR-Alaska. The focus on HRRR-Alaska is because the data assimilation changes are relatively minor for the new HRRR-Alaska product, making it easier to attribute improvements to the model physics, in this case, primarily the MYNN-EDMF and MYNN surface layer scheme.

New HRRR (Blue), old HRRR (red) and difference (orange):

And here's a figure to demonstrate improvement to 2-m temperature over the CONUS, but here, data assimilation improvements are also contributing to some of the improvement.

New HRRR (Blue), old HRRR (red) and difference (orange):

RELEASE NOTE: This version of the MYNN-EDMF code is very similar to what's in the HRRRv4, which has received unanimous thumbs up from the stakeholders during the NCEP review. This new version will become operation in June 2020.

[davegill]

Remove this change

[joeolson42]

OK. I'll take it out.

[davegill]

@joeolson42 @weiwangncar @dudhia
Joe,
I am interested in what the auto-testing results said, specifically for HWRF.

[davegill]

@joeolson42 @weiwangncar @dudhia
Joe,
In check_a_mundo, if the MYNN EDMF option is not activated, then set both bl_mynn_edmf and bl_mynn_output to zero.

[joeolson42]

Dave,

I backed out the bug fix in module_big_step_utilities_em.F and added the forcing of bl_mynn_output=0 when bl_mynn_edmf=0. However, I can't push these changes back to github yet. I'm getting "Failed connect to github.com" errors. I'll keep trying...

As for the question about auto-testing, the purpose of almost all of these changes are to improve the HRRR (CONUS & Alaska) and the global applications. The changes to the HRRR received unanimous thumbs up from the stakeholders during the NCEP review and this new version will become operation in June 2020.

The HWRF tests has been mixed, but Evan Kalina recently found a critical bug where the MYNN-EDMF was not getting any surface stability information (rmol = 1/L) passed into it, so all results have been tainted. rmol is typically calculated in the surface layer scheme, but the current version of the GFDL surface layer scheme (HWRF's preferred surface layer scheme) doesn't output this field. Once corrected, improvements are practically guaranteed. However, I think the HWRF folks are using a slightly older version of the MYNN-EDMF in their repository, so until their code is refreshed, apples & oranges will be mixed.

-joe

[davegill]

@joeolson42 @dudhia @weiwangncar

1. Joe has added the requested check_a_mundo test to set the new NML option to zero if the MYNN PBL schemes are not selected.
2. There was a small "what does this mean" flag that Joe removed.
3. The outstanding piece is the auto-testing for HWRF. With the mods to the Registry.EM_COMMON, does the code build and run with HWRF (with the new options de-selected). Joe can forward me the email from Jenkins and I can verify the status.

[davegill]

@joeolson42 @weiwangncar @dudhia
Folks,
Joe sent along the Jenkins test - the expected HWRF success is there (as well as all other passive passes).
From an infrastructure perspective, I am OK with this PR.

[davegill]

Infrastructure: OK

[weiwangncar]

I'm ok with this PR too.

[weiwangncar]

@joeolson42 Is it possible to provide a pair of plots to show some results before and after the change?

[joeolson42]

I can gather some results but it's almost impossible to isolate the improvements to only these MYNN-EDMF changes, since the development spanned at least a year and was done in parallel with changes to other model components and data assimilation. Maybe it's best to isolate changes to HRRR-Alaska, where the DA system was only slightly tweaked. [image: Screen Shot 2020-01-27 at 1.48.51 PM.png] The HRRR-CONUS results convolve physics improvements & data assimilation: [image: Screen Shot 2020-01-27 at 2.46.28 PM.png]

…

On Mon, Jan 27, 2020 at 1:18 PM weiwangncar ***@***.***> wrote: @joeolson42 <https://github.com/joeolson42> Is it possible to provide a pair of plots to show some results before and after the change? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1063?email_source=notifications&email_token=ADLRR3XW7L6CLIQBHKJPZCLQ746QFA5CNFSM4KLUBF32YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKA5CEY#issuecomment-578933011>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADLRR3WRJB5EKEPUWAAQOFLQ746QFANCNFSM4KLUBF3Q> .

-- Joseph Olson Meteorologist Model Development Branch Global Systems Division NOAA-Earth System Research Lab Boulder, Colorado

[joeolson42]

Here are some figure's that demonstrate some of the improvement in the upcoming version of the HRRR-Alaska. The focus on HRRR-Alaska is because the data assimilation changes are relatively minor for the new HRRR-Alaska product, making it easier to attribute improvements to the model physics, in this case, primarily the MYNN-EDMF and MYNN surface layer scheme. Hew HRRR (Blue), old HRRR (red) and difference (orange):

And here's a figure to demonstrate improvement to 2-m temperature over the CONUS, but here, data assimilation improvements are also contributing to some of the improvement. Hew HRRR (Blue), old HRRR (red) and difference (orange):

[weiwangncar]

@joeolson42 Thanks for providing the plots. Would you expect some improvement from this update even if someone is doing cold start from GFS, for example?

[joeolson42]

Yes, in general, but it may depend on the forecast variable and the magnitude of the improvements will vary seasonally. This version of the MYNN-EDMF does work better for global simulations cold started from the GFS - tested with the rest of the HRRR suite (and with the updated surface layer scheme coming next). The updates that are specific for cycling have a small impact beyond 1hr forecasts, they definitely won't impact cold starts.

…

On Mon, Jan 27, 2020 at 5:40 PM weiwangncar ***@***.***> wrote: @joeolson42 <https://github.com/joeolson42> Thanks for providing the plots. Would you expect some improvement from this update even if someone is doing cold start from GFS, for example? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1063?email_source=notifications&email_token=ADLRR3VEH4HVAQ5WDG5L7V3Q755IZA5CNFSM4KLUBF32YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKBTTRY#issuecomment-579025351>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADLRR3RCZSC5GE5PZP5BPLLQ755IZANCNFSM4KLUBF3Q> .

-- Joseph Olson Meteorologist Model Development Branch Global Systems Division NOAA-Earth System Research Lab Boulder, Colorado

[weiwangncar]

@joeolson42 Thanks for the reply. I'm good with this PR.