AMIP-type FHIST runs with L58 SE grid, cam_dev physics, updated CLUBB

[bstephens82]

I ran two 10-year simulations, one with clubb_l_predict_upwp_vpwp true and one false. The runscript I used for the clubb_l_predict_upwp_vpwp=.true. test is here, although note that I did not set up the "resubmit" option properly (noob mistake) so I carried out my restarts manually. The clubb_l_predict_upwp_vpwp=.false. runscript is identical apart from that flag setting.

Note that the version of CLUBB in the code base I used here is from November, not the more recent version that I think Kate is putting into the model, but I think the differences are likely to be minor.

The output from these runs is currently at /glade/scratch/stepheba/archive/pmfT.colin.10yr.b/ (prognostic=true) and /glade/scratch/stepheba/archive/pmfF.colin.10yr.b/ (prognostic=false). I've also run diagnostics on these results and that output can be found at /glade/scratch/stepheba/cesm-postprocess/.

I'm not very familiar with all of the avaiable diagnostics yet, but here are some diagnostics comparing the two tests for sea-level pressure and surface stress.

[adamrher]

interesting reductions in surface stress. Can you post the diagnostics on the web? To publish to http://webext.cgd.ucar.edu/ , go to /project/diagnostics/external/FHIST/ on any CGD machine, and follow examples to create directory to dump diagnostics into. If you don't have write permissions, email [email protected]

[bstephens82]

I've gotten access to the directories Adam mentioned and posted the diagnostics online here: https://webext.cgd.ucar.edu/FHIST/pmfT.colin.10yr.b/.

[tto061]

[Pumas-ice-substep_tht_pages.pdf](https://github.com/NCAR/amwg_dev/files/8398565/Pumas-ice-substep_tht_pages.pdf) Hi Adam the weakning winds with prognostic momentum fluxes (and the concomitant, necessary re-ordering) is a result have also got consistently in all my simulations (with FV L32). Will include some graphs here later. Cheers Thomas Thomas Toniazzo Meteorologiska institutionen (MISU) Arrhenius Laboratory, Svante Arrhenius Väg 16C Stockholms universitet SE-106 91 Stockholm Sverige tel. +46 (0)8 16 2416

…

On 2022-03-23 18:30, Adam Herrington wrote: interesting reductions in surface stress. Can you post the diagnostics on the web? To publish to http://webext.cgd.ucar.edu/ <http://webext.cgd.ucar.edu/> , go to /project/diagnostics/external/FHIST/ on any CGD machine, and follow examples to create directory to dump diagnostics into. If you don't have write permissions, email ***@***.*** ***@***.***> — Reply to this email directly, view it on GitHub <#93 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADZGLJF6AEP4MIXWL7COGSTVBNIKZANCNFSM5RORLNKA>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[adamrher]

@bstephens82 @tto061 would you mind explaining to me what is the deal with this namelist clubb_l_diag_Lscale_from_tau? My understanding is that this namelist controls whether to use all those tau variables. Are either of you using these?

I gleaned from @zarzycki that these tau variables are an "alternative" way of specifying the inverse time-scales used in various terms in the prognostic moments. However, I never knew the inverse time-scales to be so complicated. Are these tau variables then an enhancement to the default inverse time-scales, rather than just an alternative? How are the inverse time-scales specified in the default?

[bstephens82]

@adamrher I think the l_diag_Lscale_from_tau code was added shortly before my time working on CLUBB, and I think I've always used it, however I haven't worked on that code much myself. From looking back at the history a bit, it looks like calculating Lscale directly (and then calculating the taus based on Lscale) was expensive and gave weak winds in hurricanes (this was back in 2018 so a lot may have changed this then). So the decision was made to calculate the taus first (based on BV frequency and various tunable parameters) and then calculate Lscale based on that. I believe all the various tau variables are used in either case. If you can see this, the section of code that calculates Lscale first and then the taus is here, while the code that calculates the taus first and then Lscale is just below that section.

[zarzycki]

@adamrher The turbulent length scale can be currently diagnosed in CLUBB in two different ways. The default way (i.e., l_diag_Lscale_from_tau=.false.) is following Section 3b of Golaz et al., 2002.

The newer way (l_diag_Lscale_from_tau=.true.) we have been applying in the CPT version(s) of CLUBB has been calculating Lscale as:

See Section 2b of Nardi et al., 2022.

[adamrher]

I have to say, looking over the way Golaz et al. 2002 calculates L is a bit nauseating. It's like an entraining plume model using a fixed entrainment rate! The clubb_l_diag_Lscale_from_tau method seems like a more general approach. I take comfort in that.

[adamrher]

@bstephens82 @JulioTBacmeister I think it would be useful to do a run with prognostic momentum off, but with the clubb_l_diag_Lscale_from_tau still set to true. This is a thing, right? We are changing two big things at once in the run in this thread?

[bstephens82]

@adamrher the only change between the two runs linked in this thread is the value of clubb_l_predict_upwp_vpwp. They are both run with l_diag_Lscale_from_tau=.true.

[bstephens82]

@adamrher I ran a 10-year FHIST SE experiment with clubb_l_diag_Lscale_from_tau=.true. and clubb_l_predict_upwp_vpwp=.false. starting from a clean 053 tag, and the results compared to obs are here.

I haven't looked closely at all the output yet, but since I've been looking at SWCF in my last few runs, I noticed this one has as hugely amplified SWCF:

[adamrher]

Hi All,

I made a branch of Kate's branch with a first crack at a namelist_defaults. @bstephens82 could you look through this and see what I'm missing:

https://github.com/adamrher/CAM/blob/5759a2c7c5b0f5308d420c8dafdb3af37458122d/bld/namelist_files/namelist_defaults_cam.xml#L1908-L1942

I don't see the e3sm flag yet in Kate's branch. So that has to be added in. I think everything conditioned on clubb_l_diag_Lscale_from_tau is set to Colin's preferred values.

[bstephens82]

@adamrher in personal communication with @zarzycki he has indicated he's using the following values for a few of the C_invrs_tau parameters:

 C_invrs_tau_N2_wp2           =   0.100000000000000
 C_invrs_tau_N2_xp2           =   5.000000000000000E-002
 C_invrs_tau_N2_wpxp          =   0.000000000000000E+000
 C_invrs_tau_N2_clear_wp3     =    1.00000000000000

which differ from the values in your namelist.

Other things:
Colin seems to be using C4=5.2 (same as the diag_Lscale=F case according to the CAM namelist), so you can probably just drop that line.
Also your clubb_c11 for diag_Lscale=T is formatted incorrectly with two decimal points.

Those are the only things I've noticed for now.

[adamrher]

thanks, done. I kept a duplicate C4=5.2 for Lscale=T, even though it's redundant (I think we should have all dp1 and pr1 terms duplicated for Lscale=T, even if they're the same as =F).

Im looking into the e3sm namelist now.

[adamrher]

@bstephens82 do you still have any of the 2 runs at the top of this thread? I'd like to cross reference the atm_in file with the defaults I'm working on (atm_in is usually in scratch/username/case/run/atm_in/)

[bstephens82]

I don't have the output anymore for the first two runs mentioned in the thread, but the atm_in (prognostic momentum flux T and F) files from the [case]/CaseDocs directories are here:
atm_in.pmfF.txt
atm_in.pmfT.txt