segmentation fault (signal 11)

[kermarrec]

Dear team,
I am facing an error which I do not manage to fix. I try to compile a very small domain and just 18 hours for testing. I have tried to extend the memory to unliminited, reinstall wrf but it still does not work. I am using GNU Fortran (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0. All tests were OK. I can run ./real.exe but the problem happens at ./wrf.exe I also tried to use GIS4WRF and I had the same problem so I think it comes from my computer.
I attach a few files, I hope this can help me to solve the problem. Many thanks in advance,
best
gael.

 

[kwerner]

Hi,
The cause of your segmentation fault is likely due to the time_step you're using. If you search "cfl" in your rsl* files, you will see lines like this:

rsl.error.0000:d01 2020-08-20_00:00:00         2839  points exceeded cfl=2 in domain d01 at time 2020-08-20_00:00:00 hours
rsl.error.0000:d01 2020-08-20_00:00:00  MAX AT i,j,k:           20           6          16  vert_cfl,w,d(eta)=   1337.87329      -273244.844       4.93803620E-02
rsl.out.0000:d01 2020-08-20_00:00:00         2839  points exceeded cfl=2 in domain d01 at time 2020-08-20_00:00:00 hours
rsl.out.0000:d01 2020-08-20_00:00:00  MAX AT i,j,k:           20           6          16  vert_cfl,w,d(eta)=   1337.87329      -273244.844       4.93803620E-02

This means the model has become unstable. You should be using a time_step that is no larger than 6xDX. Since your dx=1km, then this value should be set to a value no larger than 6 (your currently have it set to 180) in the namelist.

After you fix that, there are a couple other things that are likely going to cause issues.
1) Your domains are too small. The values for e_we and e_sn should never be smaller than 100x100 grid spaces.
2) I see you are using a resolution (dx) of 1km. What is the resolution of the input data? The ratio between resolution should be no more than about 5:1, so if you are wanting to simulate a 1km domain and your input data resolution is too coarse, you will need to include a parent (outer) domain around the finer 1km child domain.

I would recommend taking a look at this page that discusses some of the recommendations for setting up your domain.

 

[kermarrec]

Many thanks for your kind reply. Indeed, I changed the time step and it works. The small domain was just to test if I could compute a first case but I will definitively have a look at the page you mention now that it runs. My intention is to run a LES simulation.

 

[kwerner]

I'm happy to hear you were able to get your test to run. Thanks for the update!