mpirun noticed that process rank 3 with PID 0 on node exited on signal 6 (Aborted).

[hamzanisar470]

Hi, I am using WRF4.3.3 along with WUDAPAT LCZ.
When i use sf_surface_physics= 0 the model runs smoothly. But when I use sf_surface_physics= 3 to activate BEM/BEP the model gives me this error: mpirun noticed that process rank 3 with PID 0 on node exited on signal 6 (Aborted).

 

[hamzanisar470]

I have seen that if I decrease the grid size the models runs perfectly. Can someone explaine why this happens

 

[kwerner]

Can you attach the error/out file(s) for when it's failing? If you have multiple rsl.* files, package them together into a single *.tar file and attach that. Thanks!

 

[hamzanisar470]

Kindly find the attached

 

[hamzanisar470]

What I have noticed is if i run using urban physics = 1 then there is no error but gives error on urban_physics=3

 

[kwerner]

Thanks for sending those and for the new information. I have a few suggestions:

1) The size of your d03 is too small. Domains should never be smaller than 100x100 grid points, so you should make that domain larger.
2) It's very likely you need to use more processors to run this. You are only using 6, and for your domain sizes, you could use up to about 100 (or more, depending on how large you make d03).
3) I notice dx for d01 is 3.5km. If your input data are more coarse than about 20km, you need to have a parent domain surrounding that domain. The recommendation is that the resolution ratio between domain (including between d01 and the input data) is no more that about 5:1.
4) Finally, there have been several bug fixes related to the wudapt option since WRF V4.3.3. After you make adjustments based on the above recommendations, if it still fails, I'd like you to try running with the latest version of the code (we just released V4.6.0).

 

[Kongs]

Thanks for sending those and for the new information. I have a few suggestions:

1) The size of your d03 is too small. Domains should never be smaller than 100x100 grid points, so you should make that domain larger.
2) It's very likely you need to use more processors to run this. You are only using 6, and for your domain sizes, you could use up to about 100 (or more, depending on how large you make d03).
3) I notice dx for d01 is 3.5km. If your input data are more coarse than about 20km, you need to have a parent domain surrounding that domain. The recommendation is that the resolution ratio between domain (including between d01 and the input data) is no more that about 5:1.
4) Finally, there have been several bug fixes related to the wudapt option since WRF V4.3.3. After you make adjustments based on the above recommendations, if it still fails, I'd like you to try running with the latest version of the code (we just released V4.6.0).

Thank you very much for your reply. I would like to ask about the number of grid points mentioned by you, which should not be less than 100*100. For example, if my resolution of an area is 30km, is it appropriate to set it to 90*90? I look forward to your reply!

 

[William.Hatheway]

Thank you very much for your reply. I would like to ask about the number of grid points mentioned by you, which should not be less than 100*100. For example, if my resolution of an area is 30km, is it appropriate to set it to 90*90? I look forward to your reply!

Still should be 100x100 based on the best practices recommended by NCAR as I understand. I have run less than 100x100 for the outermost domain but and some inner when nested but there are usually issues with it not producing accurate data. I can't explain why it does it but it has to do something with the way wrf calculates grids.

 

[Kongs]

Still should be 100x100 based on the best practices recommended by NCAR as I understand. I have run less than 100x100 for the outermost domain but and some inner when nested but there are usually issues with it not producing accurate data. I can't explain why it does it but it has to do something with the way wrf calculates grids.

Thank you very much for your letter. Based on your suggestion, I will set the number of grids to at least 100 x 100, which will be very helpful to me.

 

[William.Hatheway]

Thank you very much for your letter. Based on your suggestion, I will set the number of grids to at least 100 x 100, which will be very helpful to me.

Why domains should be at least 100x100 grid spaces

Each WRF domain should be at least 100x100 grid points (e_we and e_sn). When running a simulation, the input data for the lateral boundaries, every so many intervals (e.g., every 3 hours), are much more coarse than the simulation you’re trying to run. Synoptic-scale systems typically move...

forum.mmm.ucar.edu

this will also explain some more details.