release/public-v2: cherry-pick gfs.v16 update dz_min #356
Conversation
|
@llpcarson Thanks for this. So, changing |
|
It does not change answers. The original PR notes that this is expected
behavior (not answer changing in the RT cases)
…On Thu, Jan 7, 2021 at 10:32 AM Dom Heinzeller ***@***.***> wrote:
@llpcarson <https://github.com/llpcarson> Thanks for this. So, changing
dzmin does not change the results of the regression tests?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#356 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB2OWIVEPOGSIOTSIAMKRKTSYXV2FANCNFSM4VZIXNZQ>
.
|
Great, thanks. I will run the regression tests on the other platforms (thanks for doing the first three!) before we merge this. |
|
changing dz_min will change forecasts !
…On Thu, Jan 7, 2021 at 12:37 PM Dom Heinzeller ***@***.***> wrote:
It does not change answers. The original PR notes that this is expected
behavior (not answer changing in the RT cases)
… <#m_2235198163498517156_>
Great, thanks. I will run the regression tests on the other platforms
(thanks for doing the first three!) before we merge this.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#356 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AKY5N2LENGDPK6JD3BPTCRTSYXWL7ANCNFSM4VZIXNZQ>
.
--
*Fanglin Yang, Ph.D.*
*Chief, Model Physics Group*
*Modeling and Data Assimilation Branch*
*NOAA/NWS/NCEP Environmental Modeling Center*
*https://www.emc.ncep.noaa.gov/gmb/wx24fy/fyang/
<https://www.emc.ncep.noaa.gov/gmb/wx24fy/fyang/>*
|
Yes, but not the regression tests. I remember this was the same when we ran the full |
|
Yes, the RT test is using the 20161003 case, the condition of differences
of two levels <6 is not met during the integration.
…On Thu, Jan 7, 2021 at 12:40 PM Dom Heinzeller ***@***.***> wrote:
changing dz_min will change forecasts !
… <#m_-8006480597120275994_>
Yes, but not the regression tests. I remember this was the same when we
ran the full rt.conf set of tests for the develop branches. It only
changes the forecast if you actually hit the limit set by dzmin.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#356 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AI7D6TOO6WZX3IKIHDPWMXDSYXWZDANCNFSM4VZIXNZQ>
.
|
|
Changing dz_min from 2 to 6 will also have adverse effects on simulations
with very fine layer spacing so that delz ~ 6. I believe GMAO is testing a
layer setup that does just this.
This was originally proposed as a short-term fix for instabilities because
EMC needed to meet a hard deadline. It is not a long-term solution---the
root cause should be fixed, whether it is in the topography or somewhere in
the physics---and so I do not recommend that it be used except in the
operational GFSv16.
Lucas
…On Thu, Jan 7, 2021 at 12:46 PM Jun Wang ***@***.***> wrote:
Yes, the RT test is using the 20161003 case, the condition of differences
of two levels <6 is not met during the integration.
On Thu, Jan 7, 2021 at 12:40 PM Dom Heinzeller ***@***.***>
wrote:
> changing dz_min will change forecasts !
> … <#m_-8006480597120275994_>
>
> Yes, but not the regression tests. I remember this was the same when we
> ran the full rt.conf set of tests for the develop branches. It only
> changes the forecast if you actually hit the limit set by dzmin.
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub
> <
#356 (comment)
>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/AI7D6TOO6WZX3IKIHDPWMXDSYXWZDANCNFSM4VZIXNZQ
>
> .
>
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#356 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AMUQRVA66IMRYEOMFRGRO43SYXXNZANCNFSM4VZIXNZQ>
.
|
|
The code change in nh_util is incorrect. It changes the algorithm and could result in undesired behavior. |
|
I wholeheartedly agree with @lharris4 that the root cause should be found and addressed. That's certainly preferable without a doubt. Our (limited) tests at 3 km have only shown fairly small/imperceptible differences (using the NAM's vertical level distribution). We have been running with this change in our real-time parallels since mid-December. Perhaps this would be more palatable if it was a namelist parameter and not hardwired in the code? I don't feel particularly strongly either way, but the up-side is that this gives users a more stable model with the public SRW App release. |
|
@llpcarson I ran the regression tests on gaea.intel, jet.intel, hera.gnu; all tests passed against the existing baseline (i.e. b4b identical). Do you want me to send you the logs, i.e. is this PR going forward? |
|
@climbfuji thanks for running the remaining RTs, you can send me the logs or push them to my fork/branch directly. Should this PR be merged into the SRW App release branch? The discussion has been interesting, but I don't see a consensus yet, @lharris4 @JacobCarley-NOAA @junwang-noaa - thoughts? |
|
@llpcarson please copy the logs from Cheyenne into your PR, I don't want to mess with your branch: Thanks! |
Description
The code changes in fv3 dycore were made to resolve the instability issue. #211
Issue(s) addressed
Testing
Regression tests completed successfully for hera.intel, cheyenne.intel and cheyenne.gnu
Dependencies
GFDL_atmos_cubed_sphere: NOAA-GFDL/GFDL_atmos_cubed_sphere#63
fv3atm: NOAA-EMC/fv3atm#223