error for branch run

[xgao304]

Dear Scientist,


I am able to successfully run the CLM5 BGC crop model at half degree from 1951 to 2014 over the USA, driven by GSWP3 forcing. Here is the command for creating case:


create_newcase --case /net/fs05/d1/xgao/cesm2.1.3/cases/BgcCrop2000_USA --compset 2000_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_SROF_SGLC_SWAV --res hcru_hcru --user-mods-dir /net/fs05/d1/xgao/cesm2.1.3/USAdata --machine svante --compiler intel --run-unsupported


USAdata contains the surface and domain files over USA, and usr_nl_clm.


I would like to output additional variables based on pft, so I did one year branch run (2014) from the case "BgcCrop2000_USA" for testing. The case was built successfully, but stopped after generating one-month data.


1. create the case:


create_newcase --case /net/fs05/d1/xgao/cesm2.1.3/cases/BgcCropUSA_Pft --compset 2000_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_SROF_SGLC_SWAV --res hcru_hcru --user-mods-dir /net/fs05/d1/xgao/cesm2.1.3/cases/sim_setup/USAdata4pft --machine svante --compiler intel --run-unsupported


USAdata4pft contains the same surface and domain files used for the case of "BgcCrop2000_USA", but user_nl_clm contains additional variables to output.
hist_empty_htapes = .true.
hist_fincl1 = 'GRAINC_TO_FOOD','NFERTILIZATION','NPP','PCT_CFT','PCT_LANDUNIT','PCT_NAT_PFT'
hist_fincl2 = 'GRAINC_TO_FOOD','NFERTILIZATION','NPP','PCT_CFT','PCT_LANDUNIT','PCT_NAT_PFT'
hist_dov2xy = .true.,.false.
hist_type1d_pertape = '','PFTS'
hist_nhtfrq = 0,0

2.
xmlchange RUN_TYPE=branch
xmlchange RUN_REFCASE=BgcCrop2000_USA
xmlchange RUN_REFDATE=2014-01-01
xmlchange STOP_OPTION=nmonths
xmlchange STOP_N=12
xmlchange REST_OPTION=nmonths
xmlchange REST_N=12
xmlchange DATM_CLMNCEP_YR_START=2014
xmlchange DATM_CLMNCEP_YR_END=2014
xmlchange DATM_CLMNCEP_YR_ALIGN=2014


3. case.setup


4. copy the 2014 initials files (BgcCrop2000_USA.clm2.r.2014-01-01-00000.nc, BgcCrop2000_USA.clm2.rh0.2014-01-01-00000.nc, BgcCrop2000_USA.cpl.r.2014-01-01-00000.nc) and rpointer.* (point to 2014) to the "run" directory

5. case.build
6 case.submit

The attached is the cesm.log file. The error seems related to box rearrange. But I have no idea how to resolve the issue.

Your help is appreciated.

 

[erik]

The traceback does point to the history fields. So I suggest to debug, that you simplify the history fields until you find which of the history field options leads to the problem. So start with just outputting the first history file stream "fincl1". And if that doesn't work, just output a reduced set of the fields. And then search for which field isn't working.

Looking at your fincl2 list you are outputting at the PFT level, but some of the fields are at a higher level. So you need to look at what subgrid level each of the fields is at, and put them on different file streams (different fincl levels).

 

[xgao304]

Hi Erik,

Thanks for the suggestion. The case I use to branch from "BgcCrop2000_USA" has all the outputs on the grid level for the primary history files - so I think I can assume "fincl1" should not have any issue.

The thing that confuses me is why the program is able to generate one month of data, then stop. If there is an issue on the history fields, I would assume no data would be generated.

Thanks.

 

[erik]

That is actually what I would expect, that you don't find out until it actually writes data, or closes data files. Unfortunately the earliest you will often find out is on that first write. The only thing that it does catch is if you ask for an invalid history field. That will be caught at initialization.

And yes your reasoning makes sense. So you probably just need to work with the fincl2 list of fields, and can quickly figure out which ones are problematic.

 

[zhangguangzhi]

The traceback does point to the history fields. So I suggest to debug, that you simplify the history fields until you find which of the history field options leads to the problem. So start with just outputting the first history file stream "fincl1". And if that doesn't work, just output a reduced set of the fields. And then search for which field isn't working.

Looking at your fincl2 list you are outputting at the PFT level, but some of the fields are at a higher level. So you need to look at what subgrid level each of the fields is at, and put them on different file streams (different fincl levels).

Dear erik
It may be troublesome to try according to the method you said. Is there any official document that specifies the subgrid level of different history fields?

 

[slevis]

I agree with @erik
First go back to the version that worked.
Then start adding one new variable at-a-time to see when the simulation starts failing.