Reduce cost of coupled model (focus on waves) #885
Replies: 7 comments 16 replies
-
|
@pjpegion what resolutions are you working on? @aliabdolali Ali is working on creating low resolution ww3 grid, what will reduce the computation and total number tasks for ww3. @aliabdolali do you have any updates on this? Thanks |
Beta Was this translation helpful? Give feedback.
-
|
You can increase the wave model speed substantially by decreasing the spectral resolution. |
Beta Was this translation helpful? Give feedback.
-
|
I'm looking at the section on how to set time-steps, and how do I know what the maximum latitude is? If I assume it is 90-degrees, then the maximum time-step is zero. I tried looking the mod_def.gwes_30m file, but I have no idea what the format of that file is, and cannot find that information in the documentation. |
Beta Was this translation helpful? Give feedback.
-
The max latitude is 80 for gwes_30m and the time steps are 1800. 450. 900. 10. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks! But looking at this run sequence. it looks like the wave model's time-step is 300 seconds, am I missing something? CMEPS warm run sequencerunSeq:: |
Beta Was this translation helpful? Give feedback.
-
|
if the coupling time step is smaller than model time steps, it rules out the model time step. |
Beta Was this translation helpful? Give feedback.
-
|
I'm running C384/0.25 degree, but even with a lower resolution version, is it necessary to have it run at the fast, atmosphere time-step. Wind waves travel ~10m/s so the longer ocean time-step should be fine. |
Beta Was this translation helpful? Give feedback.
-
|
@SMoorthi-emc I believe there's a reason for the atm model to have waves running at the atm time step, correct? |
Beta Was this translation helpful? Give feedback.
-
|
Yes, because when two-way coupled, the wave model gives surface roughness
over open ocean used in physics.
Moorthi
…On Fri, Oct 22, 2021 at 12:09 PM Jessica Meixner ***@***.***> wrote:
@SMoorthi-emc <https://github.com/SMoorthi-emc> I believe there's a
reason for the atm model to have waves running at the atm time step,
correct?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#885 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ALLVRYRDM23A54HMEFVKHZ3UIGEBBANCNFSM5GQVNIKA>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
--
Dr. Shrinivas Moorthi
Research Meteorologist
Modeling and Data Assimilation Branch
Environmental Modeling Center / National Centers for Environmental
Prediction
5830 University Research Court - (W/NP23), College Park MD 20740 USA
Tel: (301)683-3718
e-mail: ***@***.***
Phone: (301) 683-3718 Fax: (301) 683-3718
|
Beta Was this translation helpful? Give feedback.
-
|
@SMoorthi-emc but does the roughness length really need to be updated every physics time-step, and do the wave need to have winds updated so frequently? Running at the ocean time-step is a cost savings of 6x. |
Beta Was this translation helpful? Give feedback.
-
|
I recall different spectral bins and numbers of directions were tested
(reduced) to speed up WAVE for GFS.v16 implementation. I was not they
should not be reduced any further at that time to maintain science
integrity.
Fanglin
…On Fri, Oct 22, 2021 at 11:26 AM Ali.Abdolali ***@***.***> wrote:
You can increase the wave model speed substantially by decreasing the
spectral resolution.
In the ww3_grid.inp, you can change the number of spectral bins and number
of directions. For example, in the existing glo_1deg, you can change it
from 1.07 0.035 50 36 0.5 to 1.07 0.035 25 24 0.5.
Another way would be to make the spatial grid resolution higher. Then you
can justify the time steps and make them coarser (here is the link to HOW
TO SET TIME STEPS
<https://github.com/NOAA-EMC/WW3/wiki/FAQs-page#CFL-condition>).
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#885 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AKY5N2PUUBO2KYY4UT4P7GTUIF7C7ANCNFSM5GQVNIKA>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
--
*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/>*
|
Beta Was this translation helpful? Give feedback.
-
Right, it is ok for regtesting, otherwise, they should be higher in operational setups. |
Beta Was this translation helpful? Give feedback.
-
|
Phil,
I don't think you will save much moving wave from inner to outer loop -
maybe 1 node for C384.
If you do move, the coupling is not as good where wave action may be
important; however considering all the approximations involved it isn't
that important.
Moorthi
…On Fri, Oct 22, 2021 at 12:21 PM Phil Pegion ***@***.***> wrote:
@SMoorthi-emc <https://github.com/SMoorthi-emc> but does the roughness
length really need to be updated every physics time-step, and do the wave
need to have winds updated so frequently? Running at the ocean time-step is
a cost savings of 6x.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#885 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ALLVRYRJMARRDPVFMY4Z2ZTUIGFPRANCNFSM5GQVNIKA>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
--
Dr. Shrinivas Moorthi
Research Meteorologist
Modeling and Data Assimilation Branch
Environmental Modeling Center / National Centers for Environmental
Prediction
5830 University Research Court - (W/NP23), College Park MD 20740 USA
Tel: (301)683-3718
e-mail: ***@***.***
Phone: (301) 683-3718 Fax: (301) 683-3718
|
Beta Was this translation helpful? Give feedback.
-
|
My understanding is that the code will still work as atm will use the z0 in the fv3 export state until the next time wave updates it, the same for ww3. Maybe we can try to see how much cost this can save. |
Beta Was this translation helpful? Give feedback.
-
|
@junwang-noaa I'm willing to test, but I may need help with the nems.configure |
Beta Was this translation helpful? Give feedback.
-
|
@pjpegion Maybe something like this: runSeq:: |
Beta Was this translation helpful? Give feedback.
-
|
@JessicaMeixner-NOAA Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
@JessicaMeixner-NOAA that did not work, it seems that the atmosphere model is not happy with having the import state be from an older time-step since it is crashing in the 2nd atmospheric timestep: ERROR PET000 fv3_cap.F90:1275 Invalid argument - NUOPC INCOMPATIBILITY DETECTED: Import Field wave_z0_roughness_length not at current time I guess since the wave model is not going through the mediator, is cannot handle the different coupling intervals. |
Beta Was this translation helpful? Give feedback.
-
|
I am curious, so ww3 can run with large time step (1800s vs 300s) with the
same high resolution grids?
…On Fri, Oct 22, 2021 at 12:37 PM Phil Pegion ***@***.***> wrote:
@JessicaMeixner-NOAA <https://github.com/JessicaMeixner-NOAA> Thanks!
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#885 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AI7D6TON5OZQXYYWEBFJ6R3UIGHOBANCNFSM5GQVNIKA>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
|
Beta Was this translation helpful? Give feedback.
-
|
looks like the CFL criteria, is about 1.5 hours if the wave grid is also 1/4 degree. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks, Phil. So the winds must be quite small in the wave model. |
Beta Was this translation helpful? Give feedback.
-
|
I think the winds are just fluxes, i think it is because the ocean swell travels ~10 m/s |
Beta Was this translation helpful? Give feedback.
-
|
I see, with that speed 1800s time steps should be OK. Now I am wondering why we use 300s time step :). |
Beta Was this translation helpful? Give feedback.
-
I am trying to figure out different ways to speed up the coupled model, one thing I am working on is moving from a 75 layer ocean model to a 41 layer model, which reduces the cost of the ocean by about 45%. But only about 7% for the full coupled model. One thing that I noticed is that the wave model now costs more than the ocean model! Looking at the job sequence, I see that the wave model is running at the short-time step with the atmosphere and ice-models.
Would it be possible for me to move the wave model to run at the ocean time-step to reduce its cost? I don't know enough about how the nems.configure works to do this myself and could use help suggestions on how to configure this, and I assume something would need to change in the wave model namelist too.
Thoughts?
Beta Was this translation helpful? Give feedback.
All reactions