Floating point divide-by-zero error in APM module (apm_phys_mod.F:2213) during GEOS-Chem run

[Jiao-sustech]

Your name

Jiao

Your affiliation

Southern University of Science and Technology

What happened? What did you expect to happen?

Hi @yantosca, I encountered a forrtl: error (73): floating divide by zero during a GEOS-Chem Classic run with the APM module. The error occurs in apm_phys_mod.F at line 2213, and causes the simulation to crash with a core dump. Someone meet the similar issue. Here's the relevant traceback:

forrtl: error (73): floating divide by zero
...   
gcclassic          000000000115F0B4  apm_phys_mod_mp_d        2213  apm_phys_mod.F
gcclassic          000000000114CDE5  apm_phys_mod_mp_a         510  apm_phys_mod.F
gcclassic          000000000089EB7D  apm_driv_mod_mp_a        1392  apm_driv_mod.F90
...
Aborted                 (core dumped)

The model was compiled successfully, and the run began normally before crashing at this point. I discussed this with someone else, and they mentioned they had encountered a similar issue with the APM module as well.
Thanks for your support!

What are the steps to reproduce the bug?

cmake ../CodeDir -DRUNDIR=.. -DAPM=y -DCMAKE_BUILD_TYPE=Debug
make -j
make install

Please attach any relevant configuration and log files.

err.log

GC.log

What GEOS-Chem version were you using?

14.6.1

What environment were you running GEOS-Chem on?

Local cluster

What compiler and version were you using?

gcc 4.8.5, ifort 18.0.5

Will you be addressing this bug yourself?

No

In what configuration were you running GEOS-Chem?

GCClassic

What simulation were you running?

Full chemistry

As what resolution were you running GEOS-Chem?

2×2.5

What meterology fields did you use?

MERRA-2

Additional information

No response

[yantosca]

Thanks for writing @Jiao-sustech. It looks like a logarithm is not able to be done at line 2213:

geos-chem/APM/apm_phys_mod.F

Lines 2212 to 2221 in f4c7574

	DO N=1,NTYP
	JK = INT(DLOG10(ZK(N)/1.d-5)*30.) + 1
	JK = MIN(JK,MKACT)
	JK = MAX(JK,1)
	ZDACT(N,1) = DACTTAB(IT,JK,40) !S=0.8%
	ZDACT(N,2) = DACTTAB(IT,JK,20) !S=0.4%
	ZDACT(N,3) = DACTTAB(IT,JK,10) !S=0.2%
	ZDACT(N,4) = DACTTAB(IT,JK,IS) !S= input
	ENDDO

This could be a side-effect of another issue like an out-of-bounds error. I recommend compiling fresh with

$ cmake ../CodeDir -DCMAKE_BUILD_TYPE=Debug -DAPM=y`

Then you can run a short simulation (ideally around the time of the error) to see if that flags anything.

Also the error may be caused by corrupt or incomplete meteorology. Check to see if the met field data files are all of the same size (more or less, they will differ by a little due to the file compression). If you see a file that is much smaller than the others let us know and we can reprocess it.

I noticed the compilers you are using may be problematic. We would recommend using IFORT 19 instead of IFORT 18 (but not 2021, 2022, 2023), or GCC 10 and later.

Tagging @ganluoasrc, on the APM development team.

[ganluoasrc]

This error arises due to unrealistic aerosol mass concentrations simulated in GC14-APM. The variable ZK approaches zero when the mass concentrations of the relevant aerosol species are zero, a scenario that rarely occurs in the real atmosphere. It is recommended to verify the aerosol tracers in the simulation.

[yantosca]

@ganluoasrc: One thing we can do for the future is to put in a trap that prevents doing the LOG if the argument is zero or negative. Maybe just set JK = 1 in that case.

[ganluoasrc]

Yes, I agree.