eCLM crashes with soil balance-check error while running EUR0275 setup — same setup runs fine under CLM5

[bnaz2004]

Hi all
I am running eCLM for a EU0275 storyline scenario run, using the exact same setup (namelists, forcing, grid) as Chris's existing CLM5 runs for the same storylines, which have already completed successfully. I switched to eCLM specifically to output some additional variables not available through CLM5's history stream.

The run progressed fine through most of the simulation but crashed in the final year with a soil balance-check error:

WARNING: BalanceCheck: soil balance error (W/m2)
nstep = 38803
errsoi_col = -1.000095792232786E-05
clm model is stopping - error is greater than 1e-5 (mm)
calling getglobalwrite with decomp_index= 245475 and clmlevel= column
local column index = 245475
global column index = 3169111
global landunit index = 1705284
global gridcell index = 1025730
gridcell longitude = 44.0748252868652
gridcell latitude = 56.2096099853516
column type = 223
landunit type = 2
ENDRUN:
ERROR: ERROR in BalanceCheckMod.F90 at line 711

Has anyone else seen this type of soil balance-check error when running with eCLM that does not occur under the same setup in CLM5?

[kvrigor]

as Chris's existing CLM5 runs for the same storylines, which have already completed successfully.

@bnaz2004 does this mean you have a working EU0275 setup in CLM5? Can you share the SLURM log files for both of your eCLM and CLM5 EU0275 runs?

[bnaz2004]

Hi Paul,
The path for the eCLM runs is: /p/scratch/cjibg31/naz1/SCENIC/eclm_run_storyline/2K/logs/15412866.eclm.out.
Regarding CLM5, Christian can provide the log files, as he performed the CLM5 runs, which completed successfully.

[AGonzalezNicolas]

Hi Bibi. I have no access to cjibg31. Can you copy files somewhere into cslts or give me access to your project?

[bnaz2004]

Hi Anne, I copied log files here: /p/scratch/cslts/naz1/eclm_0275_storylines/logs/

[AGonzalezNicolas]

The reason eCLM stops is within this check here :

if (abs(errsoi_col(indexc)) > 1.e-4_r8 .and. (DAnstep > skip_steps) ) then
             write(iulog,*)'clm model is stopping'
             call endrun(decomp_index=indexc, clmlevel=namec, msg=errmsg(sourcefile, __LINE__))
          end if

In the log.out file (Line 2014-2017), there is an error greater than 1.e-4_r8:

 WARNING: BalanceCheck: soil balance error (W/m2)
 nstep         =        38803
 errsoi_col    =  -1.000095792232786E-005

Do you know the threshold error for CLM5?

[kvrigor]

@AGonzalezNicolas eCLM is largely based from CLM5 so they have the same error threshold.

The more relevant question is, why would the soil error exceed 1e-4 in eCLM, but not in CLM5? From Bibi's error log, we see that abs(errsoi_col) = 1.000095792232786E-05 which is almost 1e-4, but not quite. This strongly suggests that the issue is machine-specific.

@bnaz2004 could you also share how you built eCLM? Was it built with GCC or Intel, and which Stage was used? You can simply point us to your eCLM build folder. Please also share your working EUR0275 CLM5 case dir so we could check how it was built.

[AGonzalezNicolas]

@bnaz2004, I don't have access to your folder in project1, I need you to copy:
ECLM_LOADENVS=/p/home/jusers/naz1/jureca/load-eclm-variables.sh to scratch
Please, also copy the log file from the build of eCLM ECLM_EXE=/p/project1/cjibg31/naz1/eCLM/install/.... into scratch.
@Toeroeoe could you indicate the paths to your working directory with the same case using CLM5?

[bnaz2004]

I coiped this file load-eclm-variables.sh here : /p/scratch/cjibg31/naz1/SCENIC/eclm_run_storyline/2K/
I am not sure exactly which log file you need from the eCLM build, but you should have access to /p/project1/cjibg31/naz1/eCLM/install/...

[AGonzalezNicolas]

[AGonzalezNicolas]

I need the log-file created during the build. For instance, in my case, this file is created within the bld folder and then copied into the bin folder. Usually follows a similar name to e.g., JURECADC_eCLM_2026-07-14_09-54.log and in the very first line of this file there is a long cmake command and in the last lines there is

[100%] Completed 'eCLM'
[100%] Built target eCLM
-- Install configuration: "RELEASE"

[Toeroeoe]

This is the run directory for the case in CLM5: /p/scratch/cjibg31/jibg3105/CESMDataRoot/CaseOutputs/CLM5-EUR0275-BGC_SCENIC_2K_prod/run
I copied the log files here, too: /p/scratch/cslts/shared_data/CLM5_storyline_logs/
And the namelist files are also there.

[AGonzalezNicolas]

Chris, could you add the "submit SBATCH" file and the environment?
Also, what are the names of the log.out and log.error files of the submitted job?

[Toeroeoe]

Not sure where to find the submit sbatch? In CLM5 the submission is handled by cime. I copied the clm50.jsc_generic.loadenvs to the folder.

[bnaz2004]

I need the log-file created during the build. For instance, in my case, this file is created within the bld folder and then copied into the bin folder. Usually follows a similar name to e.g., JURECADC_eCLM_2026-07-14_09-54.log and in the very first line of this file there is a long cmake command and in the last lines there is

[100%] Completed 'eCLM'
[100%] Built target eCLM
-- Install configuration: "RELEASE"

I dont have any log files created. In the bin folder, i only have the eclm.exe file generated. may be i am using the older version of eCLM. Should i try running with the new build of eCLM?