Create F1850SC5-CMIP6 compset #2721

tangq · 2019-02-03T02:15:00Z

Create the F1850SC5-CMIP6 compset (letter "S" in the name stands for special), since the following settings will be different from the F1850C5-CMIP6 compset used by the companion coupled compset (piControl, A_WCYCL1850S_CMIP6)

l_ieflx_fix flag default to .false.
Atmosphere and land initial condition data use results at the end (0501-01-01) of the DECK piControl simulation. This is particularly useful to have a better spun-up land state.
SST and sea ice data use the CMIP6 climatology calculated from 1870-1879.

All the new input files were uploaded to the E3SM input data server: https://web.lcrc.anl.gov/public/e3sm/inputdata/

Create the F1850SC5-CMIP6 compset, since the following settings will be different from the F1850C5-CMIP6 compset used by the companion coupled compset (piControl, A_WCYCL1850S_CMIP6) * l_ieflx_fix flag default to .false. * Atmosphere and land initial condition data use results at the end (0501-01-01) of the DECK piControl simulation. * SST and sea ice data use the CMIP6 climatology calculated from 1870-1879. [BFB] - Bit-For-Bit [NML] - Namelist Changing

set check_finidat_year_consistency = .false. [BFB] - Bit-For-Bit [NML] - Namelist Changing

wlin7 · 2019-02-04T05:12:51Z

Hi @tangq , if the intention of this F compset is to use the piControl spun-up cam and clm initial conditions, the changes look all good to me. But do we really want that, esp. for clm initial condition from a coupled run? When land simulation has substantial biases, it would have (probably unintended) consequence on simulations using this F-case.

tangq · 2019-02-04T05:26:57Z

Hi @wlin7 , the land initial condition is usually taken from a long-time land-only simulation. @golaz and I thought that the DECK piControl simulation spun-up the land for another hundreds of years, so it should provide a better land state than its original initial state. Therefore, we would like to replace the land IC for the F-case.

wlin7 · 2019-02-04T12:00:21Z

I see @tangq . My concern is that the land state from coupled simulation may contain larger biases than that from land-only spun-up. Ideally, I think an extended land-only spin up might suit better for general purpose of F-case. That said, this special compset can have its own purpose.

*********1*********2*********3*********4*********5*********6*********7** Turn off IEFLX fixer by setting ieflx_opt = 0. l_ieflx_fix = .false. is not effective, and thus removed. [BFB] - Bit-For-Bit [NML] - Namelist Changing

tangq · 2019-02-04T22:11:08Z

After discussing with @kaizhangpnl , I found that to turn off IEFLX fixer, ieflx_opt should be 0. l_ieflx_fix = .false. is not effective, and thus removed from the use case xml file.

wlin7 · 2019-02-04T22:22:45Z

Do we also want this compset in the master?

tangq · 2019-02-04T22:25:11Z

We eventually want it on master as well. But the land IC file from the DECK piControl run doesn't work with current master due to post-deck changes. We will need new land input files.

wlin7 · 2019-02-04T22:26:33Z

Got it @tangq . Thanks for the reminder of the land IC format change.

PeterCaldwell · 2019-02-04T22:27:04Z

Sorry if I've asked this question before, but why is l_ieflx_fix still in the model at all if it doesn't actually do anything? Keeping things like this is dangerous, as this PR shows... @kaizhangpnl - can you answer this question?

PeterCaldwell · 2019-02-05T01:51:14Z

I think @wlin7 asks a good question about which finidat file to use. It would be good to use as realistic a finidat as possible because even though our current goal is F2000 vs F1850 sensitivity runs, these F compsets will (hopefully) be what the atmos team uses for all their F runs for v2 and beyond. In this context @thorntonpe describes best practices as spinning up the land using prescribed atmosphere data from reanalysis, using this to run an F case with high-frequency output, then using that output to spin up another prescribed-atmosphere run (see https://acme-climate.atlassian.net/wiki/spaces/Docs/pages/872579110/Adding+support+for+new+atmosphere+resolutions for full description). This 3 step process seems more complex than I expect to be able to convince anyone to do, and it is unclear where we'd get reanalysis for 1850. For our F2010 compset, an alternative (which I prefer) is to use the end of the F20TR simulation. Perhaps the strategy for F1850 should be to start with the F2010 finidat and run it until the land equilibrates, then use that?

As an aside, I just looked in namelist_defaults_clm4_5.xml for what I think are the current F1850 and F2000 finidat files. They start with "climi.I", which makes me think they come from an "I" run = prescribed atmosphere rather than following Peter T's suggestion. Note that 1850_CMIP6_control.xml also sets finidat to to a file that starts with I... which is inconsistent with what I thought @golaz told me...

The files from the previous paragraph are:
clmi.I1850CLM45.ne30_oEC.1155ba0.clm2.r.nc
clmi.ICRUCLM45.ne30_g16.1155ba0.clm2.r.nc
I1850CLM45.ne30_oECv3.edison.intel.36b43c9.clm2.r.0001-01-06-00000_c20171023.nc
for 1850,2000, and CMIP6 1850, respectively. I ncdumped them, but they don't have any identifying metadata.

PeterCaldwell · 2019-02-05T01:59:30Z

I looked through the changes and they look good to me. My understanding is that the 1850S_cam5_CMIP6.xml file you added differs from 1850_cam5_CMIP6.xml only in adding ieflx_opt=0 and 1850_SCMIP6_control.xml differs only from 1850_CMIP6_control.xml in using a different finidat. Is this true? If so, would it be worth adding comments to these files explaining this difference and saying they were needed because the build infrastructure doesn't allow one to build new compsets by copying an old one then changing a thing or two?

PeterCaldwell · 2019-02-05T01:59:57Z

Also, did you try a model run with this new compset yet?

rljacob · 2019-02-05T04:13:09Z

I went back and looked at PR #1303 and it looks like l_ieflx_fix was introduced there but not used. @kaizhangpnl , was it supposed to be used at some point? That should have been caught in review.

If its not used in the code, it should be removed from the namelist and all use cases that try to set it.

rljacob · 2019-02-05T04:18:55Z

Thanks for the sleuthing, Rob! Should we create a separate issue to remove l_ieflx_fix and assign it to @kaizhangpnl?

golaz · 2019-02-05T18:04:08Z

We should get rid of l_ieflx_fix everywhere it appears. Since both @tangq and I were confused by it (even though I should have remembered it!), others might as well.
Regarding initial conditions, I'm open to any reasonable (doable) option. The methodology suggested by @thorntonpe is probably the best, but I suspect that nobody would volunteer to apply it. Furthermore it's not obvious how to extend it for 1850. @PeterCaldwell suggestion above seems also reasonable (reuse land initial conditions from the CMIP6 AMIP simulations) and equilibrate it further for 1850. @wlin7, do you think this is a better option?

tangq · 2019-02-05T18:22:07Z

@PeterCaldwell , regarding your test run question, I successfully ran this new compset for 1 day on Edison with the new input files on my local directories. I uploaded those new files to the E3SM input data server: https://web.lcrc.anl.gov/public/e3sm/inputdata/, but I don't know the procedures for testing downloading files from the input data server. Can I do such tests before the PR is reviewed and approved? The only instructions I could find are on this page (https://acme-climate.atlassian.net/wiki/spaces/Docs/pages/707002387/E3SM+Input+Data+Servers).

tangq · 2019-02-05T18:24:06Z

Once we have a plan for the land initial condition, I can work on it and change the setting in this PR to pick it up.

golaz · 2019-02-05T18:59:53Z

@bishtgautam, maybe you want to chime in regarding the best option for land initial condition for the new F1850 compset?

kaizhangpnl · 2019-02-05T19:33:47Z

Sorry if I've asked this question before, but why is l_ieflx_fix still in the model at all if it doesn't actually do anything? Keeping things like this is dangerous, as this PR shows... @kaizhangpnl - can you answer this question?

l_ieflx_fix was removed from the namelist definition file in PR1498 (https://github.com/E3SM-Project/E3SM/pull/1498/files) , but it seems that it was brought back again by 8afcf7d#diff-c93f649e7d1adfaa25adaf15495ea615. It should have been removed from phys_control.F90 to avoid the confusion.

After checking PR1498 and its merge to master again, I just found that the branch "kaizhangpnl/atm/ieflx_cleanup" has the correct code implementation/cleanup. l_ieflx_fix in both namelist definition file and in all the related F90 files was removed. However, the final merge (8abf0b8 ) missed one file (phys_control.F90). This is why when l_ieflx_fix was brought back again in namelist_definition.xml (8afcf7d ) the model still compiled and ran well.

Should I create a separate PR to remove l_ieflx_fix?

tangq · 2019-02-07T19:28:35Z

It seems that we all agree with the method @PeterCaldwell proposed to generate the land IC. I will work on it. How long is a good length for equilibrium? 50 years?

rljacob · 2019-02-07T19:30:18Z

Which namelist has the path to the sst file? I assume its docn so the file should go in inputdata/ocn/docn7/SSTDATA

wlin7 · 2019-02-07T19:54:35Z

@rljacob , it is ice_in and docn.streams.txt.prescribed. The file is used by both docn and cice. Historically, such sst/ice files required by F-cases are put on inputdata/atm/cam/sst. Starting to see some compsets assuming the the path of inputdata/ocn/docn7/SSTDATA.

tangq · 2019-02-07T20:11:47Z

Which namelist has the path to the sst file? I assume its docn so the file should go in inputdata/ocn/docn7/SSTDATA

The sst file path is the file ice_in and docn.streams.txt.prescribed and controlled by the setting in cime/src/components/data_comps/docn/cime_config/config_component.xml. What the latest default path for those files? I will make change to this PR and consistently on the input data server.

rljacob · 2019-02-07T20:59:54Z

Lets start using inputdata/ocn/docn7/SSTDATA. The cam one is from when CAM had a true standalone mode an its own data ocean model. Those don't exist anymore.

PeterCaldwell · 2019-02-07T21:29:47Z

It seems that we all agree with the method @PeterCaldwell proposed to generate the land IC. I will work on it. How long is a good length for equilibrium? 50 years?

I think the strategy should be to run for ~20 yrs, look at timeseries some measure of land spin-up, and continue running until that timeseries stabilizes. @bishtgautam and @thorntonpe - what variable(s) do you think would be appropriate to use to measure land spin up?

tangq · 2019-02-07T23:26:25Z

Lets start using inputdata/ocn/docn7/SSTDATA. The cam one is from when CAM had a true standalone mode an its own data ocean model. Those don't exist anymore.

I modified the path to use inputdata/ocn/docn7/SSTDATA. The SST and sea ice file was also moved to that directory on the input data server.

tangq · 2019-02-08T02:17:34Z

It seems that we all agree with the method @PeterCaldwell proposed to generate the land IC. I will work on it. How long is a good length for equilibrium? 50 years?

I think the strategy should be to run for ~20 yrs, look at timeseries some measure of land spin-up, and continue running until that timeseries stabilizes. @bishtgautam and @thorntonpe - what variable(s) do you think would be appropriate to use to measure land spin up?

I retrieved the land restart file at 2015-01-01 from the DECK A1 simulation and configured a run to use this F1850SC5-CMIP6 compset with that land restart file. I will do one submission. It can finish about 18 years. I will post the link to the results here when it's done.

PeterCaldwell · 2019-02-08T02:20:09Z

It seems that we all agree with the method @PeterCaldwell proposed to generate the land IC. I will work on it. How long is a good length for equilibrium? 50 years?

I think the strategy should be to run for ~20 yrs, look at timeseries some measure of land spin-up, and continue running until that timeseries stabilizes. @bishtgautam and @thorntonpe - what variable(s) do you think would be appropriate to use to measure land spin up?

I retrieved the land restart file at 2015-01-01 from the DECK A1 simulation and configured a run to use this F1850SC5-CMIP6 compset with that land restart file. I will do one submission. It can finish about 18 years. I will post the link to the results here when it's done.

That sounds great. Thanks!

tangq · 2019-02-09T17:15:16Z

The F1850SC5-CMIP6 land IC simulation will finish 18 years today. The results are on Edison:/global/cscratch1/sd/tang30/ACME_simulations/edison.20190125_maint-1.0.F1850SCMIP6_landIC.ne30_oECv3/run. Next, we will need to know the state of land spin-up to decide if it is necessary to extend this run.

Use the same land IC file as the transient compset F20TRC5-CMIP6 [BFB] - Bit-For-Bit [NML] - Namelist Changing

tangq · 2019-02-14T20:24:08Z

@PeterCaldwell and I discussed options about the land IC, we decided to use the same land IC file as the transient compset (F20TRC5-CMIP6) to maintain consistency between F1850 and AMIP simulations. I have made the change on this PR.

wlin7 · 2019-03-07T18:48:57Z

Hi @kaizhangpnl , @PeterCaldwell , @golaz ,
Can you please take a look of this PR? Qi requested you to give a review. I have also reviewed and performed tests with this PR. Thanks.

kaizhangpnl

The ieflx setting looks fine to me.

rljacob · 2019-03-07T20:00:28Z

Does this need to be on master as well as maint-1.0?

tangq · 2019-03-07T23:12:02Z

Eventually we want this on master too. My understanding is that some new files will be needed to make this compset compatible with master, but I don't know how to create those files. So I only added it to maint-1.0.

wlin7 · 2019-03-12T22:19:28Z

@tangq , do you want all your suggested reviewers to approve for this PR? I feel it is safe to merge.

tangq · 2019-03-12T22:30:22Z

@wlin7 , I think so too. It should be save to merge.

PeterCaldwell · 2019-03-13T00:26:06Z

Sorry for being tardy in my review. Feel free to merge without my review. If you want to wait, I'll try to find time tomorrow.

wlin7 · 2019-03-13T01:32:23Z

No problem at all, @PeterCaldwell . It is definitely very helpful if you can find time to take a look. There is no hurry for this PR.

PeterCaldwell

It all looks great to me. Thanks Qi!

PeterCaldwell

It all looks great to me. Thanks Qi!

wlin7 · 2019-03-13T15:50:13Z

Thanks @PeterCaldwell . I am going to merge it today.

Create F1850SC5-CMIP6 compset Create the F1850SC5-CMIP6 compset (letter "S" in the name stands for special), since the following settings will be different from the F1850C5-CMIP6 compset used by the companion coupled compset (piControl, A_WCYCL1850S_CMIP6) l_ieflx_fix flag default to .false. Atmosphere and land initial condition data use results at the end (0501-01-01) of the DECK piControl simulation. This is particularly useful to have a better spun-up land state. SST and sea ice data use the CMIP6 climatology calculated from 1870-1879. All the new input files were uploaded to the E3SM input data server. [BFB]

wlin7 · 2019-03-13T20:32:49Z

Merged to maint-1.0.

…rf_fixes Automatically Merged using E3SM Pull Request AutoTester PR Title: Change perf suite. Add test mod PR Author: jgfouca PR LABELS: AT: AUTOMERGE, AT: Integrate Without Testing

tangq added 2 commits February 2, 2019 12:53

update default clm setting

7f2ea4f

set check_finidat_year_consistency = .false. [BFB] - Bit-For-Bit [NML] - Namelist Changing

tangq added compset BFB PR leaves answers BFB maint-1.0 PRs for or issues about maint-1.0 labels Feb 3, 2019

tangq requested review from wlin7, PeterCaldwell and golaz February 3, 2019 02:15

rljacob assigned wlin7 Feb 3, 2019

tangq requested a review from kaizhangpnl February 4, 2019 20:17

tangq added 2 commits February 4, 2019 14:03

turn off IEFLX fixer by setting ieflx_opt=0

5b60204

Turn off IEFLX fixer

594f14d

*********1*********2*********3*********4*********5*********6*********7** Turn off IEFLX fixer by setting ieflx_opt = 0. l_ieflx_fix = .false. is not effective, and thus removed. [BFB] - Bit-For-Bit [NML] - Namelist Changing

Modify sst and sea ice file path

e97599b

Update the land IC setting

b7b056a

Use the same land IC file as the transient compset F20TRC5-CMIP6 [BFB] - Bit-For-Bit [NML] - Namelist Changing

wlin7 approved these changes Mar 7, 2019

View reviewed changes

kaizhangpnl approved these changes Mar 7, 2019

View reviewed changes

PeterCaldwell approved these changes Mar 13, 2019

View reviewed changes

wlin7 merged commit b7b056a into maint-1.0 Mar 13, 2019

tangq deleted the tangq/atm/F1850-LR-compset branch May 28, 2019 00:26

Create F1850SC5-CMIP6 compset #2721

Create F1850SC5-CMIP6 compset #2721

Conversation

tangq commented Feb 3, 2019

wlin7 commented Feb 4, 2019

tangq commented Feb 4, 2019

wlin7 commented Feb 4, 2019

tangq commented Feb 4, 2019

wlin7 commented Feb 4, 2019

tangq commented Feb 4, 2019

wlin7 commented Feb 4, 2019

PeterCaldwell commented Feb 4, 2019 • edited Loading

PeterCaldwell commented Feb 5, 2019

PeterCaldwell commented Feb 5, 2019

PeterCaldwell commented Feb 5, 2019

rljacob commented Feb 5, 2019 • edited Loading

rljacob commented Feb 5, 2019 • edited by PeterCaldwell Loading

golaz commented Feb 5, 2019

tangq commented Feb 5, 2019

tangq commented Feb 5, 2019

golaz commented Feb 5, 2019

kaizhangpnl commented Feb 5, 2019 • edited Loading

tangq commented Feb 7, 2019

rljacob commented Feb 7, 2019

wlin7 commented Feb 7, 2019

tangq commented Feb 7, 2019

rljacob commented Feb 7, 2019

PeterCaldwell commented Feb 7, 2019

tangq commented Feb 7, 2019

tangq commented Feb 8, 2019

PeterCaldwell commented Feb 8, 2019

tangq commented Feb 9, 2019

tangq commented Feb 14, 2019

wlin7 commented Mar 7, 2019

kaizhangpnl left a comment

Choose a reason for hiding this comment

rljacob commented Mar 7, 2019

tangq commented Mar 7, 2019

wlin7 commented Mar 12, 2019

tangq commented Mar 12, 2019

PeterCaldwell commented Mar 13, 2019

wlin7 commented Mar 13, 2019

PeterCaldwell left a comment

Choose a reason for hiding this comment

PeterCaldwell left a comment

Choose a reason for hiding this comment

wlin7 commented Mar 13, 2019

wlin7 commented Mar 13, 2019

PeterCaldwell commented Feb 4, 2019 •

edited

Loading

rljacob commented Feb 5, 2019 •

edited

Loading

rljacob commented Feb 5, 2019 •

edited by PeterCaldwell

Loading

kaizhangpnl commented Feb 5, 2019 •

edited

Loading