search

NorESMhub / noresm3_dev_simulations

Storing information for generating and diagnosing noresm3 development cases
1 stars 3 forks source link

n1850.ne30_tn14.hybrid_fatessp.20241125 #61

Open mvertens opened 5 days ago

mvertens commented 5 days ago

Purpose: Duplicate of n1850.ne30_tn14.hybrid_fatessp.20241111 but with major update to the CAM code and updated CTSM/FATES code (that handles MEGAN correctly). TODO: Not sure about the updated twostream scheme for FATES. ??? The goal is to run 20 years.

Description: To get the code

git clone https://github.com/NorESMhub/NorESM.git cd NorESM git checkout noresm2_5_alpha08 ./bin/git-fleximod update

Model grid long name a%ne30np4.pg3_l%ne30pg4.pg3_oi%tnx1v4_r%r05_w%null_z%null_g%gris4_m%tnx1v4

Case directory:

Code version on github:

Output:

mvertens commented 5 days ago

@adagj - this is not ready to go yet - but I wanted to start the documentation.

rosiealice commented 5 days ago

Hey @mvertens it doesn't look like I have the access to edit this issue (except for the title, weridly) but anyway, the new FATES param file with MEGAN and Twostream is in:

/cluster/work/users/rosief/git/NorESM_MEGAN_test/components/clm/src/fates/parameter_files/fates_params_megan_twostream.nc

FWIW, twostream radiation is turned on by setting this in the parameter file: fates_rad_model = 2 ;

To recap also, we should use the version WITHOUT twostream for testing as Ryan is workig through some residual issues with mininor errors in the first timestep which stop the ERS being BFB. It is fine to run with it though. These are being fixed. https://github.com/NGEET/fates/pull/1286

gold2718 commented 5 days ago

@rosiealice, You should have plenty of access, does GitHub think you are logged in when you try? I often see a weird state where it knows who I am but still has the "Sign in" button at the top right of the screen.

rosiealice commented 5 days ago

Maybe the thing is that the issues is a 'comment' and @mvertens made that comment and maybe you can't edit someone else's comments? I am logged in... (but that does also happen a lot!!!)

gold2718 commented 5 days ago

maybe you can't edit someone else's comments?

That's probably it. I would guess you need to be an (evil) admin to do that sort of thing. I use the 'Quote reply' option under the three-dot drop-down on a comment to indicate the purpose of my response. If you make a text selection before doing so, it will just quote that part. Otherwise, it quotes the entire comment.

mvertens commented 4 days ago

@adagj - noresm2_5_alpha08 has been made and I think this simulation is ready to be started. I hope those are not famous last words.

mvertens commented 3 days ago

@adagj - please hold off for now. I'm finding some addition problems with the ctsm external. Please wait for the updated alpha08 tag which I hope to have on Monday.

mvertens commented 3 days ago

@adagj - never mind the above message. I think we are okay with ctsm.

mvertens commented 1 day ago

@oyvindseland @rosiealice @kjetilaas - the clm namelists used in this run have some key differences from those used in n1850.ne30_tn14.hybrid_fatessp.20241111.

A few key diffs: 

< is n1850.ne30_tn14.hybrid_fatessp.20241111_v001
> is this simulation:

nuopc.runconfig - the mediator it looks like we are using the new gust parameterization:

<      add_gusts = .false.
>      add_gusts = .true.

drv_flds_in - we have different dust parameterizations:

<   dust_emis_method = 'Leung_2023'
>   dust_emis_method = 'Zender_2003'
>   zender_soil_erod_source = 'atm'

lnd_in:

<  stream_fldfilename_prigentroughness = '/cluster/shared/noresm/inputdata/lnd/clm2/dustemisdata/[Prigent_2005_roughness_0.25x0.25_cdf5_c240127.nc](http://prigent_2005_roughness_0.25x0.25_cdf5_c240127.nc/)'
<  stream_meshfile_prigentroughness = '/cluster/shared/noresm/inputdata/lnd/clm2/dustemisdata/[dust_0.25x0.25_ESMFmesh_cdf5_c240222.nc](http://dust_0.25x0.25_esmfmesh_cdf5_c240222.nc/)'
<  use_prigent_roughness = .true.
>  use_prigent_roughness = .false.

The above is set this way because we did not set dust_emis_method = 'Leung_2023'

>  interception_fraction = 1.0
>  maximum_leaf_wetted_fraction = 0.05
>  upplim_destruct_metamorph = 175.d00

I think what we need to do is:

NOTE: a lot of this confusion is due to the summary here: https://github.com/ESCOMP/CTSM/issues/2713 (which I've read several times and still find confusing).

oyvindseland commented 1 day ago

Re-running the experiment with new settings.

I think I disagree with some of the conclusions/ premises in the ESCOMP. The model expert should not be the model with the most complex use of the parameter, but the model (type) that was used in developing the parameterisation. For the discussion above this means fires, Megan and dust emissions. Dry deposition of gases is dependent on parameters both in the atmosphere and on the surface. Since the processes is governed by parameters in several models then I think the complexity of the user model should decide the place of calculation.

mvertens commented 1 day ago

@oyvindseland - thanks for your input. If I'm understanding your correctly, I agree. I think we can do things differently in noresm and do not have to replicate what is done in ESCOMP.

kjetilaas commented 23 hours ago

For the discussion above this means fires, Megan and dust emissions. Dry deposition of gases is dependent on parameters both in the atmosphere and on the surface. Since the processes is governed by parameters in several models then I think the complexity of the user model should decide the place of calculation.

Not sure I completely understood this. Do you mean that CTSM should set fire, megan and dust emissions? And CAM set dry dep? @oyvindseland. I don't have a clear opinions on this, just want to make sure I understand.

oyvindseland commented 23 hours ago

Kjetil, yes, that are my opinion on this. At the same time my preferences are not very strong so even good technical arguments may be enough for me to change my opinion.

oyvindseland commented 23 hours ago

After discussing with @mvertens we decided to run a number of atmosphere/land tests before re-running the coupled with new settings. Will kep the issue for now, but since the model crashed I will not upload the data to nird

rosiealice commented 22 hours ago

Which simulation was it that crashed? The one described in the initial issue? Any clues as to why?

Noting that dry dep and fire emissions are tbd still... I am going to work on that asap.

oyvindseland commented 22 hours ago

Yes, it was this simulation. Modified the name slightly since I created it yesterday. It is unclear why it crashed after two years, but already after month 1 there seems to a drying of Asia, wettening of Sahara After a couple of year cloud cover is 99 % almost everywhere and there is hardly any cloud water in Asia and North America and quite a lot more of sea-ice. The crash was in CLUBB

890: Error calling advance_xp2_xpyp 890: Fatal error in CLUBB: at timestep 34486 890: LAT Range: 46.7599572659205 -- 48.5538909438917
890: LON: Range: -143.877729653838 -- 0.000000000000000E+000 890: ERROR: clubb_tend_cam: Fatal error in CLUBB library 890: Image PC Routine Line Source

mvertens commented 22 hours ago

@rosiealice - we are systematically trying to see what could be causing this. As a result, @oyvindseland is now carrying out 3 shorter simulations

  1. NF1850mam4 (this has clm60%SP)
  2. NF1850 (this also has clm60%SP)
  3. 1850_CAM70%LT%NORESM%CAMoslo_CLM60%FATESSP_CICE%PRES_DOCN%DOM_MOSART_DGLC%NOEVOLVE_SWAV_SESP
rosiealice commented 21 hours ago

Thanks @mvertens let me know if I can help at all.

adagj commented 21 hours ago

@oyvindseland @mvertens @rosiealice

Can you please add important results you want to share with the community in these slides: https://docs.google.com/presentation/d/1ZQnsRMfczkoCL3BaiUoxG4QBqZjwW1momM8RG2Kpds0/edit?usp=sharing

oyvindseland commented 5 hours ago

Can do. The main problem is how to treat model crashes due to instabilities. Just send the run catalogue to nird and run diagnostics on whatever data? Can the diagnostics work on a run catalogue or do I need to build a xxx/hist manually.

adagj commented 5 hours ago

Thanks Øyvind. I think you need to archive. Maybe you can use the archiving script from the case?

ons. 27. nov. 2024, 08:46 skrev oyvindseland @.***>:

Can do. The main problem is how to treat model crashes due to instabilities. Just send the run catalogue to nird and run diagnostics on whatever data? Can the diagnostics work on a run catalogue or do I need to build a xxx/hist manually.

— Reply to this email directly, view it on GitHub https://github.com/NorESMhub/noresm3_dev_simulations/issues/61#issuecomment-2503135719, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADZUHH5FT3DAOHGGCS7SUXL2CV2FFAVCNFSM6AAAAABSHGQQ2KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKMBTGEZTKNZRHE . You are receiving this because you were mentioned.Message ID: @.***>

oyvindseland commented 4 hours ago

It is very slow / hanging when run interactively. Anyone remember how to submit it as a batch job without editing case.submit

adagj commented 4 hours ago

It is documented here: https://noresm-docs.readthedocs.io/en/noresm2/output/archive_output.html#short-term-archiving

oyvindseland commented 3 hours ago

Stored at nird now.