Physics update in cumulus convection, PBL & surface layer for UFS_P7 #665

JongilHan66 · 2021-05-24T12:12:54Z

Cumulus convection:
a) More strict convection trigger
b) Reduced entrainment rate below cloud base
c) Enhanced downdraft detrainments starting from 60mb above the ground surface
d) Reduced rain evaporation
e) Modification of cloud depth separating shallow convection from deep convection

PBL and surface layer:
a) Inclusion of wind shear effect reducing characteristic mixing length
b) Reduction of background diffusivity in the inversion layers as a function of surface roughness and green vegetation fraction
c) PBL updraft overshooting limited by bulk Richardson number-based-PBL depth
d) Inclusion of new canopy heat storage parameterization
e) Modification of thermal roughness calculation over land
f) Increase of momentum roughness length over sea
g) Inclusion of sea spray effect parameterization

The verification results for the physics updates above can be found in https://www.emc.ncep.noaa.gov/gmb/jhan/vsdbw/ccaa14
More detailed description for the updates and ccaa14 results are given in the attached ppt file.
Tests_physics_MRF_MJO_JHan_WLi.pptx

Verification results for additional physics updates are in:

Increased sea spray effects: https://www.emc.ncep.noaa.gov/gmb/jhan/vsdbw/ccaa27
Inclusion of new maximum z/L parameterization in the GFS surface layer scheme:
Winter: https://www.emc.ncep.noaa.gov/gmb/jhan/vsdbw/ccaa41
Summer: https://www.emc.ncep.noaa.gov/gmb/jhan/vsdbw/ccaa41x

The final update is based on 'ccaa41'.

Description for the new maximum z/L parameterization can be found in the attached ppt file.
PBL_SL_inv_JHan.pptx

Dependencies

NOAA-EMC/fv3atm#314
ufs-community/ufs-weather-model#592

grantfirl · 2021-06-21T15:39:40Z

@JongilHan66 Please consider merging this PR from me (JongilHan66#1) into your PR branch. These changes WILL NOT affect FV3/UFS in any way; they are only to make sure that the SCM stays compatible with the changes in this PR when specified surface fluxes are used in the SCM. If JongilHan66#1 is not merged into your branch, it will need to go into ccpp-physics in a followup PR, so including it now will probably save some work on @climbfuji and my side. Thanks!

JongilHan66 · 2021-06-21T15:58:04Z

@JongilHan66 Please consider merging this PR from me (JongilHan66#1) into your PR branch. These changes WILL NOT affect FV3/UFS in any way; they are only to make sure that the SCM stays compatible with the changes in this PR when specified surface fluxes are used in the SCM. If JongilHan66#1 is not merged into your branch, it will need to go into ccpp-physics in a followup PR, so including it now will probably save some work on @climbfuji and my side. Thanks!

It is now being regression test for each machine with a new baseline. Could you check the merging issue with Jun Wang (Jun.Wang@noaa.gov) who has a PR schedule for each update?

grantfirl · 2021-06-21T16:05:50Z

@JongilHan66 Please consider merging this PR from me (JongilHan66#1) into your PR branch. These changes WILL NOT affect FV3/UFS in any way; they are only to make sure that the SCM stays compatible with the changes in this PR when specified surface fluxes are used in the SCM. If JongilHan66#1 is not merged into your branch, it will need to go into ccpp-physics in a followup PR, so including it now will probably save some work on @climbfuji and my side. Thanks!

It is now being regression test for each machine with a new baseline. Could you check the merging issue with Jun Wang (Jun.Wang@noaa.gov) who has a PR schedule for each update?

Sure, I can ask Jun. The changed files in ccpp-physics in JongilHan66#1 are not even parsed by the FV3 build system, so I think it is impossible for those changes to affect the UFS regression tests.

climbfuji · 2021-06-21T16:11:08Z

@JongilHan66 Please consider merging this PR from me (JongilHan66#1) into your PR branch. These changes WILL NOT affect FV3/UFS in any way; they are only to make sure that the SCM stays compatible with the changes in this PR when specified surface fluxes are used in the SCM. If JongilHan66#1 is not merged into your branch, it will need to go into ccpp-physics in a followup PR, so including it now will probably save some work on @climbfuji and my side. Thanks!

It is now being regression test for each machine with a new baseline. Could you check the merging issue with Jun Wang (Jun.Wang@noaa.gov) who has a PR schedule for each update?

Sure, I can ask Jun. The changed files in ccpp-physics in JongilHan66#1 are not even parsed by the FV3 build system, so I think it is impossible for those changes to affect the UFS regression tests.

I think I can speak for the ufs code managers, this PR can be merged as long as you update the submodule pointers in fv3atm and ufs-weather-model accordingly. I recommend taking a local backup of the entire ufs-weather-model directory, then merge in Grant's changes and update the submodule pointers. Then you confirm that only the two files have changed.

And then you check out your ufs-weather-model PR recursively in a different directory and make sure it is identical to where you did the merge and submodule pointer updates. I can do this last step for you if you prefer. But this way you are 100% certain that everything is as expected and that the ufs-weather-model will not be affected.

grantfirl · 2021-06-21T16:18:07Z

@JongilHan66 Please consider merging this PR from me (JongilHan66#1) into your PR branch. These changes WILL NOT affect FV3/UFS in any way; they are only to make sure that the SCM stays compatible with the changes in this PR when specified surface fluxes are used in the SCM. If JongilHan66#1 is not merged into your branch, it will need to go into ccpp-physics in a followup PR, so including it now will probably save some work on @climbfuji and my side. Thanks!

It is now being regression test for each machine with a new baseline. Could you check the merging issue with Jun Wang (Jun.Wang@noaa.gov) who has a PR schedule for each update?

Sure, I can ask Jun. The changed files in ccpp-physics in JongilHan66#1 are not even parsed by the FV3 build system, so I think it is impossible for those changes to affect the UFS regression tests.

I think I can speak for the ufs code managers, this PR can be merged as long as you update the submodule pointers in fv3atm and ufs-weather-model accordingly. I recommend taking a local backup of the entire ufs-weather-model directory, then merge in Grant's changes and update the submodule pointers. Then you confirm that only the two files have changed.

And then you check out your ufs-weather-model PR recursively in a different directory and make sure it is identical to where you did the merge and submodule pointer updates. I can do this last step for you if you prefer. But this way you are 100% certain that everything is as expected and that the ufs-weather-model will not be affected.

Thanks @climbfuji. The alternative (and what typically happens) is to do a follow-up PR with any SCM-only changes to ccpp-physics to be merged along with a future PR. It's not a big deal to do it that way, there is just a lag for SCM users, and there has been a request to use Jongil's changes with the SCM already.

climbfuji · 2021-06-21T16:32:16Z

@JongilHan66 @grantfirl @junwang-noaa here is what we will do. We merge Jongil's PR as is (assuming that all the testing is successful). Afterwards, we merge Grant's changes into ccpp-physics main directly w/o testing in UFS, but the changes must be tested in SCM (since this is the only model reading the two files that changed). The next PR has to update to the head of main anyway and will get Grant's SCM-only changes automatically.

grantfirl · 2021-06-21T16:36:19Z

@JongilHan66 @grantfirl @junwang-noaa here is what we will do. We merge Jongil's PR as is (assuming that all the testing is successful). Afterwards, we merge Grant's changes into ccpp-physics main directly w/o testing in UFS, but the changes must be tested in SCM (since this is the only model reading the two files that changed). The next PR has to update to the head of main anyway and will get Grant's SCM-only changes automatically.

OK, @climbfuji that is fine. I'll remove the PR into Jongil's branch and put it into main, to be updated once #665 is merged.

SMoorthi-emc · 2021-06-21T17:45:53Z

Well, Grant may have to update some more with my updates next.

SMoorthi-emc · 2021-06-21T18:57:29Z

physics/sfc_diff.f

          else
-            ztmax1 = 99.0_kp
+            xkzo = gdx / xkgdx
          endif


why not write this as just "xkzo = min(one, gdx/xkgdx)"?

Sure. You can write that as "xkzo = min(one, gdx/xkgdx)".

SMoorthi-emc · 2021-06-21T19:01:43Z

physics/sfc_ocean.F

 !
 !  --- ...  flag for open water
      do i = 1, im
-
+        flag(i) = (wet(i) .and. flag_iter(i) .and. .not. use_flake(i))


Why do you need to define this extra array "flag"? It is used only once!

I take that back. I didn't see the code below

climbfuji · 2021-06-25T14:54:51Z

Since there were more problems with these code changes on Cheyenne (which uses a newer Intel compiler) than on other systems, I am going to test this code with Intel 2020.2 on Hera today.

climbfuji · 2021-06-25T20:44:48Z

Since there were more problems with these code changes on Cheyenne (which uses a newer Intel compiler) than on other systems, I am going to test this code with Intel 2020.2 on Hera today.

I ran rt.conf on Hera with Intel 2020.2. The following two tests failed, because they timed out towards the end of the run:

control_wrtGauss_netcdf_parallel 013 failed
control_wrtGauss_netcdf_parallel 013 failed in run_test
regional_quilt_netcdf_parallel 036 failed
regional_quilt_netcdf_parallel 036 failed in run_test

All other tests completed successfully within the walltime. @junwang-noaa @DeniseWorthen @DusanJovic-NOAA

junwang-noaa · 2021-06-26T02:55:33Z

Thanks Dom for doing these tests. So there is no stability issue with Intel 2020.2 (tests time out rather than failed) while there is with INtel 2019. Currently the two tests take 170s and 290s in the develop branch on hera, it looks to me there is some issue with parallel netcdf library. Can you run Intel 2020.2 with the develop branch to see if the same slowness happens for the two tests?.

…

On Fri, Jun 25, 2021 at 4:45 PM Dom Heinzeller ***@***.***> wrote: Since there were more problems with these code changes on Cheyenne (which uses a newer Intel compiler) than on other systems, I am going to test this code with Intel 2020.2 on Hera today. I ran rt.conf on Hera with Intel 2020.2. The following two tests failed, because they timed out towards the end of the run: control_wrtGauss_netcdf_parallel 013 failed control_wrtGauss_netcdf_parallel 013 failed in run_test regional_quilt_netcdf_parallel 036 failed regional_quilt_netcdf_parallel 036 failed in run_test All other tests completed successfully within the walltime. @junwang-noaa <https://github.com/junwang-noaa> @DeniseWorthen <https://github.com/DeniseWorthen> @DusanJovic-NOAA <https://github.com/DusanJovic-NOAA> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#665 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7D6TKSCSSZSF3BO56HAITTUTTEVANCNFSM45ND3WEQ> .

climbfuji · 2021-06-26T03:01:31Z

Maybe not Intel 19(.1.1), maybe with SGI MPT or something else on Cheyenne. Will try develop on hera with 2020.2.

…

On Jun 25, 2021, at 8:55 PM, Jun Wang ***@***.***> wrote: Thanks Dom for doing these tests. So there is no stability issue with Intel 2020.2 (tests time out rather than failed) while there is with INtel 2019. Currently the two tests take 170s and 290s in the develop branch on hera, it looks to me there is some issue with parallel netcdf library. Can you run Intel 2020.2 with the develop branch to see if the same slowness happens for the two tests?. On Fri, Jun 25, 2021 at 4:45 PM Dom Heinzeller ***@***.***> wrote: > Since there were more problems with these code changes on Cheyenne (which > uses a newer Intel compiler) than on other systems, I am going to test this > code with Intel 2020.2 on Hera today. > > I ran rt.conf on Hera with Intel 2020.2. The following two tests failed, > because they timed out towards the end of the run: > > control_wrtGauss_netcdf_parallel 013 failed > control_wrtGauss_netcdf_parallel 013 failed in run_test > regional_quilt_netcdf_parallel 036 failed > regional_quilt_netcdf_parallel 036 failed in run_test > > All other tests completed successfully within the walltime. @junwang-noaa > <https://github.com/junwang-noaa> @DeniseWorthen > <https://github.com/DeniseWorthen> @DusanJovic-NOAA > <https://github.com/DusanJovic-NOAA> > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#665 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AI7D6TKSCSSZSF3BO56HAITTUTTEVANCNFSM45ND3WEQ> > . > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#665 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB5C2RJZMNQ24WNNLSKS7ODTUU6S7ANCNFSM45ND3WEQ>.

climbfuji · 2021-06-26T03:54:07Z

Same for develop. Could also be a mistake on my end when installing hpc-stack for Intel 2020.2. Definitely something to keep in mind.

…_hanp7

climbfuji · 2021-07-15T16:02:39Z

physics/satmedmfvdifq.F

@@ -917,13 +950,23 @@ subroutine satmedmfvdifq_run(im,km,ntrac,ntcw,ntiw,ntke,          &
              if(n == 1) then
                dz = zl(i,1)
                tem1 = tsea(i)*(1.+fv*max(q1(i,1,1),qmin))
+!               tem1 = 0.5 * (tem1 + thvx(i,n))
+!!              tem1 = 0.5 * (tem1 + thlvx(i,n))


At some point (not this PR) we should remove all these comments.

climbfuji · 2021-07-15T21:14:31Z

Regression testing completed, will merge this PR now.

JongilHan66 added 30 commits May 21, 2021 15:19

updated hfx2 & qfx2

3307baa

updated evap & hfx

89cd97f

updated hflx & evap

ccac831

updated GFS debug

51f2842

updated GFS PBL generic

662bb1f

updated GFS PBL generic meta

0dcde4a

updated GFS surface generic

989701b

updated GFS surface generic meta

b542460

updated mfpblq

b8a3a7e

updated evap & hflx

4d301b1

updated hflx & qflx

3bdbf1c

updated heat & evap

a034b14

updated heat & evap

b567312

updated deep convection

2683863

updated samfdeepcnv.meta

f44d15a

updated shallow convection

64b3d2d

updated samfshalcnv.meta

b5fca27

updated heat & evap

3e155b4

updated TKE-EDMF

ed61f02

updated satmedmfvdifq.meta

fd20310

updated TKE-EDMF

c4f5394

updated TKE-EDMF

3175790

updated GFS surface layer scheme

a3c4d52

updated sfc_diff.meta

f16f1af

add sea spray effect parameterization

2a518b2

updated sfc_nst.meta

20189cb

add sea spray effect parameterization

3d1e7f3

updated sfc_ocean.meta

f4bdbf5

updated heat & evap

f207919

updated heat & evap

9b88927

grantfirl mentioned this pull request Jun 21, 2021

SCM-only followup to #665 #678

Closed

SMoorthi-emc reviewed Jun 21, 2021

View reviewed changes

SMoorthi-emc approved these changes Jun 21, 2021

View reviewed changes

JongilHan66 added 5 commits July 1, 2021 16:53

fix the conflict

2a52172

fix the conflict

41cbdc2

update czilc

0808096

Merge branch 'main' of https://github.com/NCAR/ccpp-physics into ccpp…

ce60c85

…_hanp7

fix conflict

ace4b58

climbfuji approved these changes Jul 15, 2021

View reviewed changes

climbfuji merged commit 1792e87 into NCAR:main Jul 15, 2021

This was referenced Jul 22, 2021

Physics update in cumulus convection, PBL & surface layer for UFS_P7 #657

Closed

Canopy Heat Storage Parameterization changes for PBL schemes #443

Closed

Similar or duplicated definitions of the same variables in GFS_typedefs and ccpp/physics #640

Closed

rmontuoro mentioned this pull request Jul 26, 2021

Deep convective transport of aerosols in SAMF generates unrealistically large mixing ratios #709

Closed

grantfirl mentioned this pull request Sep 20, 2021

SCM-only updates for latest main #731

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Physics update in cumulus convection, PBL & surface layer for UFS_P7 #665

Physics update in cumulus convection, PBL & surface layer for UFS_P7 #665

JongilHan66 commented May 24, 2021 •

edited by climbfuji

Loading

grantfirl commented Jun 21, 2021

JongilHan66 commented Jun 21, 2021

grantfirl commented Jun 21, 2021

climbfuji commented Jun 21, 2021

grantfirl commented Jun 21, 2021

climbfuji commented Jun 21, 2021

grantfirl commented Jun 21, 2021

SMoorthi-emc commented Jun 21, 2021

SMoorthi-emc Jun 21, 2021

JongilHan66 Jun 21, 2021

SMoorthi-emc Jun 21, 2021

SMoorthi-emc Jun 21, 2021

climbfuji commented Jun 25, 2021

climbfuji commented Jun 25, 2021

junwang-noaa commented Jun 26, 2021 via email

climbfuji commented Jun 26, 2021 via email

climbfuji commented Jun 26, 2021

climbfuji Jul 15, 2021

climbfuji commented Jul 15, 2021

Physics update in cumulus convection, PBL & surface layer for UFS_P7 #665

Physics update in cumulus convection, PBL & surface layer for UFS_P7 #665

Conversation

JongilHan66 commented May 24, 2021 • edited by climbfuji Loading

grantfirl commented Jun 21, 2021

JongilHan66 commented Jun 21, 2021

grantfirl commented Jun 21, 2021

climbfuji commented Jun 21, 2021

grantfirl commented Jun 21, 2021

climbfuji commented Jun 21, 2021

grantfirl commented Jun 21, 2021

SMoorthi-emc commented Jun 21, 2021

SMoorthi-emc Jun 21, 2021

Choose a reason for hiding this comment

JongilHan66 Jun 21, 2021

Choose a reason for hiding this comment

SMoorthi-emc Jun 21, 2021

Choose a reason for hiding this comment

SMoorthi-emc Jun 21, 2021

Choose a reason for hiding this comment

climbfuji commented Jun 25, 2021

climbfuji commented Jun 25, 2021

junwang-noaa commented Jun 26, 2021 via email

climbfuji commented Jun 26, 2021 via email

climbfuji commented Jun 26, 2021

climbfuji Jul 15, 2021

Choose a reason for hiding this comment

climbfuji commented Jul 15, 2021

JongilHan66 commented May 24, 2021 •

edited by climbfuji

Loading