"FWscHIST" custom compset vs. out-of-the-box "FHIST" compset

[adamrher]

@cecilehannay I was wondering if we could get some amwg diagnostics showing the impact of specified chemistry in L58. I see in your campaign storage directory you have:

f.e21.FHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf
f.e21.FWscHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf

Which I think would address differences due to specified chemistry.

Thanks!
Adam

[cecilehannay]

Hi @adamrher,

The diags are here:
https://webext.cgd.ucar.edu/FHIST/f.e21.FHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf/atm/

Indeed, it was the purpose of that simulation to address the differences between FHIST and FWscHIST in the low top model.

At the time, we decided that the climate of FHIST and FWscHIST were similar enough and the extra cost was small enough (10-20%) to justify to use FWscHIST in both the low top and high models. But we can revisit that.

[swrneale]

There are certainly consistent +1K differences in the tropical tropopause temp.
(https://webext.cgd.ucar.edu/FHIST/f.e21.FHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf/atm/f.e21.FHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf.1980_1989-f.e21.FWscHIST.ne30_L48_BL10_cam6_3_019_plus_CESM2.2.001_zm2.hf.1980_1989/set5_6/set5_ANN_TTRP_TROP_c.png)

I wonder if we could do the WACCM version of the diagnsotics?

[JulioTBacmeister]

What is the impact on the water vapor bias?

[adamrher]

Since set4 doesn't work in the amwg diags ... Ill make some zonal height difference plots of T and Q

[islasimpson]

I can make the tape recorder plot

[swrneale]

Cecile, I'm pretty sure I fixed set4 in the past (my script works). Can you point me to your version of the code and then I can compare?

[cecilehannay]

I use: /glade/p/cesm/postprocessing/atm_diag

[adamrher]

Because both these runs use ZM2 from the "cam6_3_019_plus_CESM2.2.001" runs, I added both of these runs to the bottom row of the Q plots here. In those runs, the UT is a bit drier than our current L58+zm2 control. But FHIST has a bit moister lower stratosphere.

Note I also have 5 years of the new PUMAS simulation in the bottom as well. Ill update this tmrw morning to have more PUMAS years.

[islasimpson]

Here are some tape recorders.

[islasimpson]

Also, someone sent me these papers the other day. They discuss how better resolving tropical waves in the TTL can lead to dehydration of the stratosphere because the waves have warm and cold regions and the cold regions control how much gets into the stratosphere. I haven't had a chance to read them yet, but I wonder if there are some diagnostics in there that we could use to test if that's what's happening. If it is, then maybe we're barking up the wrong tree if we think it's physics that's responsible.

https://acp.copernicus.org/articles/15/3517/2015/
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2014GL062737
https://acp.copernicus.org/articles/21/2469/2021/

[adamrher]

Thanks. We could re-create their figure two (in the second paper) during the wet phase.

maybe we're barking up the wrong tree if we think it's physics that's responsible.

I've been arguing for a while it's the dynamics, not the physics. This was the point of the DQCORE diagnostics. I don't think my understanding is inconsistent with the mechanisms discussed in these papers.

[adamrher]

@dan800 copying your post into this discussion on the impacts of the "sc" compset modifier.

@JulioTBacmeister @adamrher @islasimpson: So I had a look at a FWscHIST.ne30_L83 case to see what's in the name list and chemistry. I think this is at least part of how it's set up in 'sc'

Molecular diffusion of species is off:
do_molec_press = 0.1D0 and the top level is 0.0081 hPa (0.8Pa)

non-LTE cooling calcs are on (I think) and use constituents in the WACCM forcing file as part of the calculation - CO2, O, O2 and O3 if I recall correctly.

The only prognostic constituents are H2O, CH4, N2O, CFC11, CFC12, H2O2, H2SO4, SO2 & DMS. The chemistry (if you can call it that) for CH4, N2O, CFC11 and CFC12 is just a simple loss mechanism - water is produced from the loss of CH4.
DMS is lost with reactions with OH and NO3 (both specfied) to produce SO2 which then reacts with OH to produce H2SO4. Not sure what H2O2 does in the model - probably something to do with the aerosols (its reaction with OH is a source of water but I doubt it's significant).

Looks like there's 20 aerosol species in total.

So not sure what switching off SC for a low top model gets you unless it drops some of these tracers. I can't imagine reading in a heating rate and calling Fomichev cost 10% of the model. But we can certainly see if turning off those two things really changes the 50 to 80 km region much. If it doesn't we can just ditch it for all CAM configurations.

[islasimpson]

Here is a plot of H2O. It's comparing WACCM with with L83, both coupled piControl, FV. It's a bit hard to separate out any influence of the upper boundary condition from the overall dry bias. @dan800 - let me know if there are other diagnostics that would show this better. I don't think there's a rush - we maybe don't care about this too much for the Scripps seasonal prediction project as they are short runs, initialized from obs.

[dan800]

@islasimpson Can you check the value at the top of the model in L83? Is it setting it to the value used in WACCM (i.e., 20 ppb)?

     subroutine set_tgcm_ubc( lchunk, ncol, mmr, mw_dry )
        !--------------------------------------------------------------------
        !       ... Set the upper boundary values h2o, h2, and h
        !--------------------------------------------------------------------
         real(r8), parameter ::  h2o_ubc_vmr = 2.e-8_r8            ! fixed ub h2o concentration (kg/kg)

[islasimpson]

I can look at the top lev in the netcdf files but I guess that's not the top. Here's a snapshot of the top lev. Minimum values here are higher than that (on the order 5e-7). Is this helping or do you mean actually figuring it out from the code?

[adamrher]

Here's some more analysis of the differences between FHIST and FWscHIST. Timings (core hrs per sim yr) using 2100 tasks

FWscHIST: 7364.40, 7333.62
FHIST: 6121.43, 5976.72

Both runs have lots of high frequency output ... and all of the fincls have interpolate_output=T. It is surprising to me how expensive these sims are ... I have FHIST L58 runs w/ more modest i/o and the cost from one timing file is 4680.03 on 1800 tasks. That's much closer to the expected incr. in cost going from L32->L58.

Returning to the matter at hand, there's no denying that FWscHIST is a robust 20% increase in cost over FHIST. That's nothing to balk at. 20% is equal to the cost savings of the pg2 grid.

The reason it's more expensive is probably because it has more tracers. FHIST has 35 tracers, FWscHIST has 43. (@PeterHjortLauritzen any idea of whether the +20% is consistent w/ a back-of-the-envelope calc due to 8 more tracers?). The additional tracers are:

12  CH4       CH4                                                                                                                               dry
13  N2O       N2O                                                                                                                               dry
14  CFC11     CFC11                                                                                                                             dry
15  CFC12     CFC12                                                                                                                             dry
40  AOA1      Age-of_air tracer 1                                                                                                               wet
41  AOA2      Age-of_air tracer 2                                                                                                               wet
42  HORZ      horizontal tracer                                                                                                                 wet
43  VERT      vertical tracer                                                                                                                   wet

Well the first four seem kind of important. How were these treated in prior low-top climate sensitivity runs? I can see the age of air important for diagnostic purposes. What are these HORZ and VERT tracers?

If there are other causes of the increase, maybe someone can glean some info from looking at a diff of the atm_in files between these two compsets. Execute this command on cheyenne:

diff /glade/scratch/aherring/cam6_3_043_FWscHIST_ne30pg3_ne30pg3_mg17_mct_600pes_220120_camdev_dynlb0/run/atm_in /glade/scratch/aherring/cam.clubbmf_FHIST_ne30pg3_ne30pg3_mg17_L58camdev_1800pes_220108_3macmic_Nx2yrs/run/atm_in

Lastly, I want to illustrate the main climate difference between these two configuration. There is a modest moistening of the lower stratosphere in FHIST compared to FWscHIST (here). But I think the temperature signal is more interesting, with the Tropical cold point warmer by about 1K in FHIST. Any ideas why?

[dan800]

Ah - a little confusing! Since the grid is settled, maybe switch case names to L58 and L93?

[cecilehannay]

The label for the L58 is 'L58'
L58 Vertical grid: L48_BL10

The label for the L93 is 'L93'
L93 Vertical grid: L83_BL10

You can find the labels at:
https://github.com/NCAR/amwg_dev/labels

As Isla explained these grids were called:
L48_BL10 and L83_L10
in the long names to differentiate the runs that were done with extra levels in the PBL when we were developing the grid.

Note that all the runs are not available on the amwg_dev github. (Older runs before we created the github are not all there).

[cecilehannay]

Ah - a little confusing! Since the grid is settled, maybe switch case names to L58 and L93?

@dan800: We could change the casenames now the grids are settled.

The reason we did that way was that when we were developing grids, we had a flurry of grids. It was confusing and we wanted to differentiate grids with the BL levels and without.

For instance, at some point, there was one L91 without BL levels and L93 with BL levels.
So differentiate them, we called the grids:
L91 without extra BL levels
L83_BL10 with extra BL level

But I think it is fine to change the names now that all the grids have extra BL levels.

[adamrher]

@dan800 I think we should open up a new discussion about what tracers we want in "specified chemistry" runs. I think we can omit 6 for sure (HORZ, VERT, AOA1, AOA2, and the two graupel tracers). Do you want to open it up with your thoughts on what else to consider?

[dan800]

OK - I'd say the discussion is more 'Tracers in CAM7'. The hope is we unify physics and # of tracers for L93 and L58. There's got to be a really good reason to add complexity and overhead should these choices diverge.

[dan800]

Hi Isla, I think you need to run WACCM on the same grid where they overlap to really get to the bottom of this. If the line commented out in Byron’s code means it it using the fixed boundary condition for waccm at 140 km at 80 km (ie practically zero) that certainly will yield low values higher up. Not convinced the differences in the UTLS are not just from differences in tropopause height. To help with the Scripps project just changing the UBC fixed value might be a big improvement, but for our needs we can do better - a seasonally varying value from WACCM. Regards, Dan

…

Sent from my iPhone

On Feb 7, 2022, at 18:39, islasimpson ***@***.***> wrote:  Here is a plot of H2O. It's comparing WACCM with with L83, both coupled piControl, FV. It's a bit hard to separate out any influence of the upper boundary condition from the overall dry bias. @dan800 - let me know if there are other diagnostics that would show this better. I don't think there's a rush - we maybe don't care about this too much for the Scripps seasonal prediction project as they are short runs, initialized from obs. — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you were mentioned.