move some parameters in mynn surface layer to namelist options #137
Conversation
| @@ -3419,6 +3423,10 @@ subroutine control_initialize (Model, nlunit, fn_nml, me, master, & | |||
| real(kind=kind_phys) :: bl_mynn_closure = 2.6 !< <= 2.5 only prognose tke | |||
| !< 2.5 < and < 3.0, prognose tke and q'2 | |||
| !< >= 3.0, prognose tke, q'2, T'2, and T'q' | |||
| logical :: sfclay_compute_diag = .false. | |||
| logical :: sfclay_compute_flux = .false. | |||
There was a problem hiding this comment.
This differs from the original defaults. I'm testing sfclay_compute_flux=.true. right now to see if it fixes the problem.
There was a problem hiding this comment.
I've run a bunch of rrfs and rap regression tests with sfclay_compute_flux=.true. and they're passing. It seems this is the problem.
There was a problem hiding this comment.
So, is the solution to:
- add sfclay_compute_flux=.true. to the HRRR/RRFS namelists
- keep the non-HRRR/RRFS namelist as is (default to .false.). Also Keep RAP namelists as is.
- run new baselines for the HRRR/RRFS namelists
- then complete the regression testing
- merge
?
There was a problem hiding this comment.
For (2), we also need to configure nsst to turn off.
There was a problem hiding this comment.
I don't understand step 2. Are you talking about the workflow?
Edit: I don't understand step 2.
There was a problem hiding this comment.
I think so, assuming it's the workflow that determines which namelists are used for each test/suite. My thinking is that we should probably create a regresion test that actually mimics the way RRFS is run (with sfclay_compute_flux=.true.). This would require a new baseline. Most of the rest of the regression tests can stay with the default setting (sfclay_compute_flux=.false.) but the tests that use the mynn surface layer scheme and also use nsst will require new baselines since the original regression tests (with sfclay_compute_flux hard-coded as .false.) were overwritten with the baselines generated from Mike's PR.
|
In a meeting, we've decided to include this modification to change the results since most configurations using MYNN should not have MYNN compute the fluxes. The RRFS parallels that need to do that will enable the new namelist option (sfclay_compute_flux=.false.) However, this means we need to update the regression tests for RRFS at the ufs-weather-model level. @SamuelTrahanNOAA is working on that now. |
|
These changes are in NOAA-EMC#526 and will be merged back to gsl/develop after the community repositories are updated. |
|
#144 will merge these changes |
|
This PR was already merged. |
Description
Testing
Dependencies
If testing this branch requires non-default branches in other repositories, list them.
Those branches should have matching names (ideally)
Do PRs in upstream repositories need to be merged first?
If so add the "waiting for other repos" label and list the upstream PRs