adding inner loop to Thompson microphysics by RuiyuSun · Pull Request #685 · NCAR/ccpp-physics

RuiyuSun · 2021-06-27T20:54:10Z

This is to add an inner loop to the Thomspon microphysics scheme with a namelist option.

climbfuji · 2021-06-28T12:56:56Z

This PR is trying to do the same a the existing subcycling capability. I suggest we test if the bug fix for Thompson subcycling #684 fixes the issue with decreasing hydrometeors and associated low radar reflectivity. If #684 fixes the problem, I suggest to not merge this PR #685 in favor of the more general CCPP approach. If #684 doesn't fix the problem, we should consider removing the subcycling changes from Thompson MP and just use the inner loop approach from @RuiyuSun.

yangfanglin · 2021-06-28T13:06:14Z

Dom,

Won't it be more flexible to use namelist options to turn on and off sub-stepping Thompson MP and control the number of sub-steps ? Using the sub-cycling approach one has to update SDF and recompile the model every time a change is made.

climbfuji · 2021-06-28T13:13:37Z

Dom,

Won't it be more flexible to use namelist options to turn on and off sub-stepping Thompson MP and control the number of sub-steps ? Using the sub-cycling approach one has to update SDF and recompile the model every time a change is made.

Both approaches have their pros and cons. The SDF approach is more general, because the same capability can be used for other schemes in the same way. The namelist option has the advantage that one doesn't have to recompile. I am fine either way, but we should decide on one or the other option in my opinion. Having both approaches coded up seems like a recipe for confusion (although Ruiyu made it work - I will have one request for change later to avoid duplicating the diagnostics code in module_mp_thompson.F90). Greg, what do you think? Any preference from your side? @gthompsnWRF

yangfanglin · 2021-06-28T13:24:09Z

Dom,

We can discuss this at the meeting today and make a decision later based on testing results. Both approaches still have issues.

climbfuji · 2021-06-28T14:12:33Z

Dom,

We can discuss this at the meeting today and make a decision later based on testing results. Both approaches still have issues.

Sounds good to me.

climbfuji · 2021-06-28T14:17:55Z

One of the "issues" I am seeing is the slightly complicated logic in lines 1155-1157 in module_mp_thompson.F90 that "corrects" the user dt_inner value so that you get an integer number of steps. Another approach is to specify the number of steps, but either way the user has to make sure that dt_atmos and nsteps / dt_inner are set consistently.

RuiyuSun · 2021-06-28T14:44:43Z

A potential advantage of this method is to set an upper limit time step (default value) that the microphysics scheme can use. So when the physical time step is too large the inner loop will be automatically activated.

…

On Mon, Jun 28, 2021 at 10:18 AM Dom Heinzeller ***@***.***> wrote: One of the "issues" I am seeing is the slightly complicated logic in lines 1155-1157 in module_mp_thompson.F90 that "corrects" the user dt_inner value so that you get an integer number of steps. Another approach is to specify the number of steps, but either way the user has to make sure that dt_atmos and nsteps / dt_inner are set consistently. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#685 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKFOCLFOV5HLKUX6MEV5JSDTVCAB3ANCNFSM47MYAPVA> .

-- Ruiyu Sun, PhD *IMSG* at NOAA/NWS/NCEP/EMC 5830 University Research Ct., Rm. 2097 College Park, MD 20740 ***@***.*** ***@***.***> 301-683-3787

climbfuji · 2021-06-28T15:29:17Z

Yes, that is a clear advantage. We should have a diagnostic message written out as part of the first time step, so that the user knows what is going on.

…

On Jun 28, 2021, at 8:44 AM, RuiyuSun ***@***.***> wrote: A potential advantage of this method is to set an upper limit time step (default value) that the microphysics scheme can use. So when the physical time step is too large the inner loop will be automatically activated. On Mon, Jun 28, 2021 at 10:18 AM Dom Heinzeller ***@***.***> wrote: > One of the "issues" I am seeing is the slightly complicated logic in lines > 1155-1157 in module_mp_thompson.F90 that "corrects" the user dt_inner > value so that you get an integer number of steps. Another approach is to > specify the number of steps, but either way the user has to make sure that > dt_atmos and nsteps / dt_inner are set consistently. > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#685 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AKFOCLFOV5HLKUX6MEV5JSDTVCAB3ANCNFSM47MYAPVA> > . > -- Ruiyu Sun, PhD *IMSG* at NOAA/NWS/NCEP/EMC 5830 University Research Ct., Rm. 2097 College Park, MD 20740 ***@***.*** ***@***.***> 301-683-3787 — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub <#685 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB5C2RN2U2DWJ3XB6FZFTWLTVCDGJANCNFSM47MYAPVA>.

RuiyuSun · 2021-06-28T16:02:45Z

That is right. A message at the first step will be helpful. Another potential advantage of the inner loop is that it does not go through the conversions between the moist and dry mixing ratios. On Mon, Jun 28, 2021 at 11:29 AM Dom Heinzeller ***@***.***> wrote:

…

Yes, that is a clear advantage. We should have a diagnostic message written out as part of the first time step, so that the user knows what is going on. > On Jun 28, 2021, at 8:44 AM, RuiyuSun ***@***.***> wrote: > > > A potential advantage of this method is to set an upper limit time step > (default value) that the microphysics scheme can use. So when the physical > time step is too large the inner loop will be automatically activated. > > > On Mon, Jun 28, 2021 at 10:18 AM Dom Heinzeller ***@***.***> > wrote: > > > One of the "issues" I am seeing is the slightly complicated logic in lines > > 1155-1157 in module_mp_thompson.F90 that "corrects" the user dt_inner > > value so that you get an integer number of steps. Another approach is to > > specify the number of steps, but either way the user has to make sure that > > dt_atmos and nsteps / dt_inner are set consistently. > > > > — > > You are receiving this because you were mentioned. > > Reply to this email directly, view it on GitHub > > <#685 (comment) >, > > or unsubscribe > > < https://github.com/notifications/unsubscribe-auth/AKFOCLFOV5HLKUX6MEV5JSDTVCAB3ANCNFSM47MYAPVA > > > . > > > > > -- > > Ruiyu Sun, PhD > > *IMSG* at NOAA/NWS/NCEP/EMC > > 5830 University Research Ct., Rm. 2097 > > College Park, MD 20740 > > ***@***.*** ***@***.***> > > 301-683-3787 > — > You are receiving this because your review was requested. > Reply to this email directly, view it on GitHub < #685 (comment)>, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AB5C2RN2U2DWJ3XB6FZFTWLTVCDGJANCNFSM47MYAPVA >. > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#685 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKFOCLFRX2G667UIGLXYXZTTVCINPANCNFSM47MYAPVA> .

-- Ruiyu Sun, PhD *IMSG* at NOAA/NWS/NCEP/EMC 5830 University Research Ct., Rm. 2097 College Park, MD 20740 ***@***.*** ***@***.***> 301-683-3787

climbfuji · 2021-06-28T16:46:50Z

That is right. A message at the first step will be helpful. Another potential advantage of the inner loop is that it does not go through the conversions between the moist and dry mixing ratios. On Mon, Jun 28, 2021 at 11:29 AM Dom Heinzeller @.***> wrote:
…

Yes, I have been struggling with whether to perform the conversion only at the first (convert to) and last (convert from) time step. This would be ok and essentially the same as your inner loop, but only as long as nobody puts another scheme/computation in the Thompson subcycle group in the suite definition file.

climbfuji · 2021-06-28T21:13:22Z

+
+!> - Call calc_refl10cm()
+
+           diagflag_presenti: IF ( PRESENT (diagflag) ) THEN


The code duplication in lines 1432-1472 can and should be avoided if I see this correctly by using the appropriate combination of conditions.

RuiyuSun · 2021-06-29T09:55:20Z

Thank you for your suggestion. The logic has been modified.

…

On Mon, Jun 28, 2021 at 5:13 PM Dom Heinzeller ***@***.***> wrote: ***@***.**** requested changes on this pull request. ------------------------------ In physics/module_mp_thompson.F90 <#685 (comment)>: > + + SR(i,j) = (pcp_sn(i,j) + pcp_gr(i,j) + pcp_ic(i,j))/(RAINNC(i,j)+1.e-12) + RAINNCV(i,j) = RAINNC(i,j) + IF ( PRESENT (snowncv) ) THEN + SNOWNCV(i,j) = SNOWNC(i,j) + ENDIF + IF ( PRESENT (icencv) ) THEN + ICENCV(i,j) = ICENC(i,j) + ENDIF + IF ( PRESENT (graupelncv) ) THEN + GRAUPELNCV(i,j) = GRAUPELNC(i,j) + ENDIF + +!> - Call calc_refl10cm() + + diagflag_presenti: IF ( PRESENT (diagflag) ) THEN The code duplication in lines 1432-1472 can and should be avoided if I see this correctly by using the appropriate combination of conditions. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#685 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKFOCLCNGZTWAD6VE7PRCE3TVDQYDANCNFSM47MYAPVA> .

-- Ruiyu Sun, PhD *IMSG* at NOAA/NWS/NCEP/EMC 5830 University Research Ct., Rm. 2097 College Park, MD 20740 ***@***.*** ***@***.***> 301-683-3787

climbfuji

Looks good to me except one tiny change, thanks for updating the logic.

climbfuji · 2021-06-30T20:45:10Z

 !> - Call calc_refl10cm()

-           diagflag_present: IF ( PRESENT (diagflag) ) THEN
+           diagflag_presenti: IF ( PRESENT (diagflag) ) THEN


Can you revert this change (additional i) please?

climbfuji · 2021-06-30T20:45:22Z

             enddo
           endif
-           ENDIF diagflag_present
+           ENDIF diagflag_presenti


Same here, can you revert this change (additional i) please?

RuiyuSun · 2021-07-01T11:07:13Z

'i' is removed. Thanks.

…

On Wed, Jun 30, 2021 at 4:46 PM Dom Heinzeller ***@***.***> wrote: ***@***.**** requested changes on this pull request. Looks good to me except one tiny change, thanks for updating the logic. ------------------------------ In physics/module_mp_thompson.F90 <#685 (comment)>: > !> - Call calc_refl10cm() - diagflag_present: IF ( PRESENT (diagflag) ) THEN + diagflag_presenti: IF ( PRESENT (diagflag) ) THEN Can you revert this change (additional i) please? ------------------------------ In physics/module_mp_thompson.F90 <#685 (comment)>: > @@ -1426,7 +1459,7 @@ SUBROUTINE mp_gt_driver(qv, qc, qr, qi, qs, qg, ni, nr, nc, & refl_10cm(i,k,j) = MAX(-35., dBZ(k)) enddo endif - ENDIF diagflag_present + ENDIF diagflag_presenti Same here, can you revert this change (additional i) please? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#685 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKFOCLFN6XR7APW5F25X74LTVN7BLANCNFSM47MYAPVA> .

-- Ruiyu Sun, PhD *IMSG* at NOAA/NWS/NCEP/EMC 5830 University Research Ct., Rm. 2097 College Park, MD 20740 ***@***.*** ***@***.***> 301-683-3787

climbfuji · 2021-07-01T12:43:41Z

@RuiyuSun since we discussed at the last Thompson MP meeting that the way forward is to temporarily accept both approaches (inner loop and subcycling), would you mind if I sent you a PR to include my bug fixes in your PR?

Merge "Bugfix in Thompson MP, pass correct timestep to core routine" into "Adding inner loop to Thompson microphysics"

adding inner loop to Thompson microphysics

dd04dc3

RuiyuSun requested review from JulieSchramm, climbfuji, grantfirl and llpcarson as code owners June 27, 2021 20:54

climbfuji requested changes Jun 28, 2021

View reviewed changes

change the logic

7178cb2

climbfuji requested changes Jun 30, 2021

View reviewed changes

climbfuji requested a review from gthompsnWRF June 30, 2021 20:46

remove extra i

f7450c2

climbfuji and others added 3 commits July 1, 2021 11:16

Bugfix in Thompson MP, pass correct timestep to core routine

581f5a6

Remove test code from physics/mp_thompson.F90

b498507

Merge pull request #1 from climbfuji/innerloop_ruiyu_dom_mods

9e1a47a

Merge "Bugfix in Thompson MP, pass correct timestep to core routine" into "Adding inner loop to Thompson microphysics"

climbfuji mentioned this pull request Jul 1, 2021

Bugfix in Thompson MP, pass correct timestep to core routine #684

Merged

climbfuji mentioned this pull request Jul 22, 2021

Wrapper PR for: Thompson inner loop, Thompson subcycling bugfix, fix snet bug for Noah LSM #702

Merged

climbfuji merged commit 518968f into NCAR:main Jul 23, 2021

RuiyuSun mentioned this pull request Sep 2, 2021

adding semi-Lagrangian sedimentation of rain into Thompson MP #724

Closed


		!> - Call calc_refl10cm()

		diagflag_presenti: IF ( PRESENT (diagflag) ) THEN

Conversation

RuiyuSun commented Jun 27, 2021

Uh oh!

climbfuji commented Jun 28, 2021

Uh oh!

yangfanglin commented Jun 28, 2021

Uh oh!

climbfuji commented Jun 28, 2021

Uh oh!

yangfanglin commented Jun 28, 2021

Uh oh!

climbfuji commented Jun 28, 2021

Uh oh!

climbfuji commented Jun 28, 2021

Uh oh!

RuiyuSun commented Jun 28, 2021 via email

Uh oh!

climbfuji commented Jun 28, 2021 via email

Uh oh!

RuiyuSun commented Jun 28, 2021 via email

Uh oh!

climbfuji commented Jun 28, 2021

Uh oh!

climbfuji Jun 28, 2021

Choose a reason for hiding this comment

Uh oh!

RuiyuSun commented Jun 29, 2021 via email

Uh oh!

climbfuji left a comment

Choose a reason for hiding this comment

Uh oh!

climbfuji Jun 30, 2021

Choose a reason for hiding this comment

Uh oh!

climbfuji Jun 30, 2021

Choose a reason for hiding this comment

Uh oh!

RuiyuSun commented Jul 1, 2021 via email

Uh oh!

climbfuji commented Jul 1, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants