WRF-Chem hangs without any error

[Merve Eke]

Hi everyone,

I'm running a WRF-Chem simulation with 4 nested domains (d01 to d04). The simulation runs fine for domains d01 and d02 (currently seems fine for domain 3 as well) but it freezes when entering d04. There is no error message — the model just stops at the following point in the log:

d04 2021-05-08_00:00:02 kpp_mechanism_driver: calling mozcart_interface

My settings:
- time_step = 60
- chemdt = 60, 0, 0, 0
- parent_time_step_ratio = 3, 4, 5
- dx for d04 = 200 m

The model seems to hang during the MOZCART chemistry step. There’s no segmentation fault or crash, just a silent stop.

Has anyone experienced similar behavior or have suggestions to fix this? I am sending the namelist and rsl.error attached.

Any help would be greatly appreciated!

Best,  
Merve

[Gabriele Pfister]

Hi Merve

my guess is that you are running out of memory. Have you tried running one-way nesting using ndown. 

I also noticed that for d01 you are setting chemdt=60. chemdt is in minutes, so you want to set it to 1 or at the most 2 minutes. You might also want to change to phot_opt=4, we no longer recommend using option 3 (see https://www2.acom.ucar.edu/wrf-chem).

Also note that it is recommended to use an odd domain ratio for WRF (https://forum.mmm.ucar.edu/threads/domain-design-the-ratio-among-domain-resolutions.329/) 

Gabi

--

================================
Gabriele Pfister

Atmospheric Chemistry Observations & Modeling (ACOM) Lab

NSF National Center for Atmospheric Research (NCAR)

Email: pfi...@ucar.edu

Work Phone: +1 303 497 2915

Web: https://staff.ucar.edu/users/pfister

[Merve Eke]

Hi Gabi,

Thank you for your return! I have tried running it with different memory allocations per CPU—4, 8, and 12 GB. I also set feedback = 0 for one-way nesting, but it still hangs at the exact same point.

Regarding memory and cores, my latest trial setup is shown below. Previously, I tried running ndown with a different chemical mechanism (RADM2), but I encountered issues with the boundary conditions for the 4th domain. The upper-layer concentration profile was significantly different from that of the 3rd domain. That’s why I’m now trying to run all domains together.

#SBATCH -N4 --ntasks-per-node=12                 # Number of nodes and cores per node required

#SBATCH --mem-per-cpu=12G  

srun -n 48 ./wrf.exe

Best,

Merve

31 Mart 2025 Pazartesi tarihinde saat 11:24:37 UTC-4 itibarıyla pfi...@ucar.edu şunları yazdı:

[Gabriele Pfister]

Hi Merve

 can you  run d01 all by itself and does the output look ok? I am not sure what you mean with issues in the boundary conditions for the 4th domain? Was this when using ndown? 

Gabi

[Merve Eke]

Hi Gabi,

Yes, I can run d01 all by itself, also I can run all domains when I set chem_opt into 0. 

I had the boundary condition issue while running ndown before (while using RADM2). Therefore, I am trying to run all 4-domains together this time.

Best,

Merve

Gabriele Pfister <pfi...@ucar.edu>, 7 Nis 2025 Pzt, 11:12 tarihinde şunu yazdı:

[Gabriele Pfister]

Hi Merve

the only I can think of is a potential memory issue. I do not understand why the upper-layer concentration profile in your wrfbdy for the innermost domain would be very different when you are using ndown. Is it only the top-most layer? 

Gabi