Open cecilehannay opened 1 year ago
I am still having problems with the TEM diags in cam6_3_102.
FLDLST: VTHzaphys in fincl(1, 5) not found
FLDLST: WTHzaphys in fincl(2, 5) not found
FLDLST: UVzaphys in fincl(3, 5) not found
FLDLST: UWzaphys in fincl(4, 5) not found
FLDLST: Uzm in fincl(5, 5) only available with FV dycore
FLDLST: Wzm in fincl(6, 5) only available with FV dycore
FLDLST: Vzm in fincl(7, 5) only available with FV dycore
FLDLST: THzm in fincl(8, 5) only available with FV dycore
ERROR: FLDLST: 4 errors found, see log
I restarted without the TEM diags for now as I would like to know if the rest is working.
Is it that do_tem_diags=.true. thing needed? I may have got it wrong what exactly the namelist parameter is.
I looked at the output (without TEM diags): /glade/scratch/hannay/archive/f.cam6_3_102.FLTHIST_v0a.ne30.001/atm/hist
I am curious to know who is looking at the fincl2. This is a lot of output. If it is needed, it is fine but I don't remember seeing a lot of plots of these fields.
@fvitt merged the TEM code yesterday and the issue was closed. https://github.com/ESCOMP/CAM/pull/770
adopts the zm naming convention for the zonal mean outputs.
add 'Uzm','Vzm','Wzm','THzm' output fields
So any idea why TEM output is not working?
FLDLST: Uzm in fincl(5, 5) only available with FV dycore
FLDLST: Wzm in fincl(6, 5) only available with FV dycore
FLDLST: Vzm in fincl(7, 5) only available with FV dycore
FLDLST: THzm in fincl(8, 5) only available with FV dycore
BTW, did we revert back to the old names for TEM terms (e.g., 'UWzm') and do away with the zaphys names?
FLDLST: VTHzaphys in fincl(1, 5) not found
FLDLST: WTHzaphys in fincl(2, 5) not found
FLDLST: UVzaphys in fincl(3, 5) not found
FLDLST: UWzaphys in fincl(4, 5) not found
TEM should work with these settings: phys_grid_ctem_nfreq=-6 phys_grid_ctem_zm_nbas=120 phys_grid_ctem_za_nlat=90 fincl3 = 'Uzm','Vzm','Wzm','THzm', 'VTHzm','WTHzm','UVzm','UWzm'
@cecilehannay In SourceMods you have physpkg.F90
which does not include the updates needed for TEM (updates in cam6_3_02). Remove physpkg.F90
from SourceMods and rebuild. Then the TEM should work with cam_dev physics and the *zm
output fields listed above.
@fvitt: I need this code for COSP but I will merge the two physpkg.F90
I will update the fincl5
for the new zm naming convention.
fincl5 = 'Uzm','Vzm','Wzm','THzm', 'VTHzm','WTHzm','UVzm','UWzm'
@cecilehannay Brian Eaton simplified that PR for COSP and now it only requires modifications to cospsimulator_intr.F90. So you can checkout Ben's PR branch and SourceMod in cospsimulator_intr. And then just ditch the physpkg and convect_diagnostics SourceMods that were the "old" fix for COSP.
An alternative is that @brian-eaton start regression testing for the COSP changes this afternoon (after #783 is tagged as cam6_3_103) and he can make a tag later today or tomorrow morning at the latest. Then the COSP code and cam6_3_102 will work properly together.
Let us know if you want to proceed with this, otherwise we'll wait until the CO2 changes are confirmed and put them into together.
FWIW, I think waiting until we have a code base that can do both COSP and TEM at the same would be better than trying to merge things in source mods.
I merged and submitted a 2-year run before seeing your comments (I was in meeting all afternoon). I will let the run completes to see if there is any showstopper or anything missing. Then, we can use a more recent tag for a longer run.
This run is two years:
/glade/scratch/hannay/archive/f.cam6_3_102.FLTHIST_v0a.ne30.001/atm/hist
Please check that it looks ok to and that you have everything you need. Then we can decide what we want to do (i.e. move to a more recent tag, continue the run, add/remove diags, ...).
@cecilehannay - @brian-eaton has just made cam6_3_105 which contains the fix for COSP. He expects to have cam6_3_106 (with the proposed CO2 fix) ready to commit later today. We will wait to commit this until we hear that it should go into CAM
For the purposes of the coupled evaluation the important thing is whether the prognostic CO2 looks OK along with RESTOM FLNT etc... If the first two years look OK we should just continue this run and evaluate its climate based on the longer run. When the TEM and COSP diagnostics are finished we can start a new F-case run, but I don't think they are essential for deciding whether we can go ahead with couped testing. Thoughts?
I would encourage Cecile to use at least cam6_3_105 in a run before we start a coupled run using it. There are several reasons (and it is just me being super cautious).
Waiting for cam6_3_106 (which will contain the CO2 fix) is less problematic as the source code changes for it will be identical with what Cecile is using in her SourceMods. I don't see a need to have Cecile use this tag in an F compset run, but it should certainly be used in the coupled run.
@JulioTBacmeister - Are we at the point where we should be bringing the CO2 changes into cam_development, or is it still to early to make that decision?
@cacraigucar: I am not worried about the COSP change are just diags. The CTSM changes could be more important. I don't know what they are.
Do you need it now or can I just continue this run as @JulioTBacmeister suggested? Do I wait for cam6_3_106
and do another shorter run to test it before run coupled.
@cecilehannay - I should be careful in my statements. The answer changes may be due to CTSM or they could also be NUOPC related or due to any of the other externals which CAM uses. It is up to you on whether you want to see the impact of the external changes on a long run or not. I can say definitively, that during the CAM regression testing, F compsets did have answer changes when we updated the externals.
Thanks @cacraigucar
Cecile will use cam6_3_105 for all new simulations.
We need to check the two years Cecile has already run (@dan800 ?) before making the call about putting the CO2 changes into cam_development. If this run looks OK the CO2 changes (prognostic, radiatively active CO2) should go in to cam_development.
@JulioTBacmeister and @cecilehannay - I'm hoping that if we move forward with the CO2, that @cecilehannay will use cam6_3_106 instead of cam6_3_105 for her new simulations
@cacraigucar what will be in cam6_3_106 besides the CO2 fix?
and when will it be ready? We'd like to get the extended F-case started ASAP.
cam6_3_106 only has the CO2 fix in it. And note that this only has an impact if your run starts from an initial file that does not contain CO2. If you are extending runs either using restarts or using an initial file from the previous run that contains CO2 then it will have no impact. The testing on cam6_3_106 is done so I'm just waiting for the go ahead to merge it to cam_development and create the tag.
OK. This seems completely uncontroversial. Go ahead and merge it.
Some results comparing this FLTHIST run to FWscHIST. These are just two year means so we should expect some noise.
Compare to our prior FLTHIST run, which had the hot top (likely because there is too little CO2 up there).
And here are the Q changes:
Compare that to the Q changes in the prior FLTHIST run. It seems more clear now that these water vapor changes in the old run were probably reflecting the large changes in T. And that the actual impact of the UBC in FLTHIST, is not as drastic as those earlier runs indicated.
Sorry, been on a panel today. I had a quick look at CO2 and the CFCs. The latter have only just been added to the h0s and are input to the RT, so worth checking.
CO2 looks good - not seeing anything odd at the UB:
At the end of 2 years the CO2 in the upper part of the model seems to be lagging the troposphere by 4 ppm, which I guess is about what it might have changed in that time. Also, nice to see the air is 'older' at high latitudes.
CFCs spinning up:
There's a loss of CFCs specified in the model domain, unlike with CO2.
At the end of the 2 years:
Obs (SPARC climatologies):
note: Pressure range is 300-10 hPa in these plots.
Looks to be about half of observed in 2005-7 but that's about the trend difference.
I'm looking at the TEM file - it's got a lot of unnecessary output fields related to COSP. Does it need to be in there? cosp_prs_bnds (time, cosp_prs, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_tau_bnds (time, cosp_tau, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_ht_bnds (time, cosp_ht, nbnd) float64 dask.array<chunksize=(1, 40, 2), meta=np.ndarray> cosp_sr_bnds (time, cosp_sr, nbnd) float64 dask.array<chunksize=(1, 15, 2), meta=np.ndarray> cosp_dbze_bnds (time, cosp_dbze, nbnd) float64 dask.array<chunksize=(1, 15, 2), meta=np.ndarray> cosp_htmisr_bnds (time, cosp_htmisr, nbnd) float64 dask.array<chunksize=(1, 16, 2), meta=np.ndarray> cosp_tau_modis_bnds (time, cosp_tau_modis, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_reffice_bnds (time, cosp_reffice, nbnd) float64 dask.array<chunksize=(1, 6, 2), meta=np.ndarray> cosp_reffliq_bnds (time, cosp_reffliq, nbnd) float64 dask.array<chunksize=(1, 6, 2), meta=np.ndarray>
I have the first two year of TEM output calculated. They are on glade here: ~marsh/projects/CAM7-Evaluation/notebooks/TEM/f.cam6_3_102.FLTHIST_v0a.ne30.001.TEM.nc New zonal mean output seems to be working! Here's the mass streamfunction for the last timestep: @islasimpson Please take a look. w* looks a bit funky near the top:
Very cool to see the larger CO2 "emissions" in the NH spring/summer months, compared with the lower emissions in SH spring/summer.
Let's plan to dump out a new inic file for use in the coupled eval, that has the CFC's spun-up. I believe setting inithist_all
forces all constituents into the cam.i file.
Summary in google slide form for tomorrow https://docs.google.com/presentation/d/16FZoB8HNIUPR5qkWRlmjvbYpu654orV59LvwdRKqe3k/edit?usp=sharing I'm tied up in a NAS panel but hopefully someone can present on my behalf I have added a link in the agenda
Very cool to see the larger CO2 "emissions" in the NH spring/summer months, compared with the lower emissions in SH spring/summer.
Let's plan to dump out a new inic file for use in the coupled eval, that has the CFC's spun-up. I believe setting
inithist_all
forces all constituents into the cam.i file.
I think the emissions actually peak in winter spring (as things die off/decay) and it's not until summer where they are being drawn down by vigorous plant growth. I'm sure you know, but we are specifying a concentration not emission in this run.
inithist_all
is a relic from when Dave Williamson and Jerry Olsen were attempting to make initial files behave like restart files for their CAPT work. It adds extra variables to the initial file which are not relevant for cam_dev physics.
By default all advected constituents are included in the initial file. Also by default initial files are written yearly on Jan 1, so Cecile's current runs should be generating them.
I stand corrected. No, it's not summer, but winter, when NH is emitting a maxima in CO2, and no, we don't need to set inithist to generate a new inic. I'm adding so much value here.
@adamrher OK, I'm no tropospheric expert but look at those winds in the SH compared to ERA-40 in the slide deck - seems way to strong in winter (JJA). I'd worry about sea ice in coupled runs. Is the SAM a diagnostic we are looking at?
@brian-eaton thanks for the background on _all; clarification on what's in the cam.i. files
I'm looking at the TEM file - it's got a lot of unnecessary output fields related to COSP. Does it need to be in there? cosp_prs_bnds (time, cosp_prs, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_tau_bnds (time, cosp_tau, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_ht_bnds (time, cosp_ht, nbnd) float64 dask.array<chunksize=(1, 40, 2), meta=np.ndarray> cosp_sr_bnds (time, cosp_sr, nbnd) float64 dask.array<chunksize=(1, 15, 2), meta=np.ndarray> cosp_dbze_bnds (time, cosp_dbze, nbnd) float64 dask.array<chunksize=(1, 15, 2), meta=np.ndarray> cosp_htmisr_bnds (time, cosp_htmisr, nbnd) float64 dask.array<chunksize=(1, 16, 2), meta=np.ndarray> cosp_tau_modis_bnds (time, cosp_tau_modis, nbnd) float64 dask.array<chunksize=(1, 7, 2), meta=np.ndarray> cosp_reffice_bnds (time, cosp_reffice, nbnd) float64 dask.array<chunksize=(1, 6, 2), meta=np.ndarray> cosp_reffliq_bnds (time, cosp_reffliq, nbnd) float64 dask.array<chunksize=(1, 6, 2), meta=np.ndarray>
@dan800, these are small arrays and are essentially coordinate metadata for cosp output. Note that there are no horizontal grid dimensions (ncol_d) in them. This is a "feature" of our history output -- all the coordinate data is written into every history/initial file without checking whether there is a variable being written which requires that coordinate.
I am looking at aerosol diagnostics. @cecilehannay could you add soa_a2 also to the list in fincl7, I guess it still slipped in the earlier list.
Here are the budgets (SOA burden is missing, because of the missing fincl), compared to the earlier TCMTHIST runs (that were done in 1996-1999). Differences are due to changes in emissions (like BC , POM, sulfate), but dust and sea-salt look good. Dust is higher, but the earlier run was probably low.
Total AODVISdn is 0.105, usually, I think, we tune for something like 0.13, but I think, we have different emissions in 1980, so this is probably OK for now and we have to compare with later dates.
Compared to AOD from MODIS (though for later years), this compares to a 2000-2020 climatology, so we need to wait for later years of the simulation.
@JulioTBacmeister Regarding tuning clouds, one thought is that we need to make sure we are comparing the same years with the observations when we tune the clouds since emissions (especially BC, POM, and SO4) are changing between 1980 and 2000, which could change clouds as well.
Description: We are starting a new run with 102 that will include:
Case directory: Locally (if still available): /glade/p/cesmdata/cseg/runs/cesm2_0/f.cam6_3_102.FLTHIST_v0a.ne30.001
On github: https://github.com/NCAR/amwg_dev/tree/f.cam6_3_102.FLTHIST_v0a.ne30.001
user_nl_cam
Sandbox: Locally (if still available): /glade/work/hannay/cesm_tags/f.cam6_3_102
On github: https://github.com/ESCOMP/CAM/tree/cam6_3_102 hash: fd1c488
Diagnostics: ADF diags (if available) https://webext.cgd.ucar.edu/FLTHIST/f.cam6_3_102.FLTHIST_v0a.ne30.001/atm/
Contacts: @cacraigucar, @cecilehannay, @julio, @fvitt, @adamrher, @brianpm, @dan800, @klindsay28, @tilmes, @islasimpson