Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Segmentation fault or pointer association errors in WRF-Solar EPS

YOcube

New member
when I running short-term simulations with WRF-Solar EPS (v4.5) driven by GFS data, the model crashed with segmentation fault at beginning. After recompiling WRF-Solar using ./configure -D and rerunning the same setup, the error changed to: forrtl: severe (408): fort: (7): Attempt to use pointer I_START when it is not associated with a target. The error seems to occur during the initialization of the stochastic perturbation scheme, but I’m unsure of the exact cause. The error message and associated code are shown in the image.
1746420814614.png1746420863761.png

Following Issue #1545 on GitHub (Segmentation fault using random_perturb in V4.3 · Issue #1545 · wrf-model/WRF), I tried reducing lengthscale_rand_pert in the namelist, but the error remained. I comment out a part of the SUBROUTINE INITIALIZE_STOCH in the source code (as in issue #1545), the model always stops after running for a period of time.

I’ve attached my namelist.input and rsl.error files for reference. Has anyone else experienced similar issues with pointer association errors in WRF-Solar EPS? Any advice on how to resolve this properly would be greatly appreciated.
 

Attachments

  • namelist.input
    6.3 KB · Views: 0
  • rsl.error.0000
    5.1 KB · Views: 1
After further testing, I've observed some new behaviors that might help diagnose the issue:
(1) This error appears to be hauntingly unpredictable, sometimes when I submit the job, the model actually runs successfully
(2) Modifying lengthscale_rand_pert shows little effect, whether configure with or without the -D option.
(3) The model exhibits different behaviors when partially commented code is used: it aborts after several minutes of integration when compiled with the ./configure -D debugging option, while appearing to run normally when compiled with standard ./configure settings.

I'm uncertain whether this approach is appropriate. Since the commented-out code section relates to the spin up of the perturbations in the initialization, which appears to be connected with the rand_perturb parameter in the namelist, would setting rand_perturb to 0 be a valid way to circumvent this error?
 
Last edited:
I will forward your questions to WRF-SOLAR experts, and hopefully they can get back to you soon.

For your question of rand_perturb, if you don't need to run with Stochastic perturbation, please set rand_perturb = 0.
 
Last edited:
I will forward your questions to WRF-SOLAR experts, and hopefully they can can get back to you soon.

For your question of rand_perturb, if you don't need to run with Stochastic perturbation, please set rand_perturb = 0.
Thanks a lot for your help
 
Top