Segmentation Error After 6 hours of Output

[FSU_Anna]

Hi,

I've been trying to run WRF with 3 nested domains in the Gulf of Mexico for a couple of weeks now. I've tried changing up the physics and dynamics a little, but I keep getting a segmentation error after 6 hours of output. When I look at the output, I'm getting some vertical instability in the northwest corner of my innermost domain (domain 3, 100% in the Gulf over water), but it's far enough from the boundary I don't think it's a boundary problem. I think it's a grid point or subgrid point instability, so I've been trying to play with the damping and diffusion options.

Does anyone have recommendations on how to handle this? Perhaps any physics or dynamics recommendations?

Does it have to do with number processors? I don't think so, but I'm looking at all possibilities at this point.

I've included my namelist.input, rsl.errors, WRF.out, wrf.exe submission script.

 

[FSU_Anna]

I changed p_top to 5000, and still have the same error after 6 hours of output when running all 3 domains.

I just tried running wrf.exe with 2 domains and it gets a segmentation error after 15 hours of output.

Any help or ideas would be greatly appreciated!

 

[kwerner]

Hi,
Regardless of whether the CFL errors are occurring at the boundary, there is still instability that is causing the model to stop. I see you're using less than a 6xDX time step, but you may have to try decreasing it even more. Take a look at this FAQ that also discusses another option (epssm) you can try.

 

[FSU_Anna]

Thank you! There are some new things in the FAQ link that I will try and let you know how it goes.

 

[FSU_Anna]

It's working now. I'm not entirely sure what fixed it. I think it was switching to diff_6th_opt = 2?

 

[Ming Chen]

Redoing time step is an effective way to control CFL violation. diff_6th_opt only works to prohibit noisy structure in the simulation.