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

All value in WRF output is NaN after version 4.4

L

Leon Lin

New member
Hello

I noticed that in some cases which may be a extreme precipitation event the value stored in wrfout is mostly NaN.
This issue only happen after version 4.4 (below version 4.3.2 work well).

I operated WRF model compiled by NVHPC on AWS EC2 C7. The attachments are the sample of wrfout and namelist.
 

Attachments

  • wrf.issue.zip
    7.6 MB · Views: 2
NaN in wrfout indicates the case failed due errors in physics/dynamics. If the data and options are the same and the case can run with V4.3.2, then there is no reason it failed with v4.4. I am suspicious this failure might be related to compiling.
 
Ming Chen said:
NaN in wrfout indicates the case failed due errors in physics/dynamics. If the data and options are the same and the case can run with V4.3.2, then there is no reason it failed with v4.4. I am suspicious this failure might be related to compiling.
Click to expand...
Thanks for your reply.
I have had suspected the issue resulted from configuration during compiling but we used the same configuration on Version 4.4 and 4.3.
 
Just for confirmation, can you run the case using WRFv4.4? Please keep everything unchanged, i.e., wrfinput, wrfbdy, namelist.input, etc. If it works for WRFv4.4 but failed with WRFV4.5, then we will know that this is a computer issue.
 
Ming Chen said:
Just for confirmation, can you run the case using WRFv4.4? Please keep everything unchanged, i.e., wrfinput, wrfbdy, namelist.input, etc. If it works for WRFv4.4 but failed with WRFV4.5, then we will know that this is a computer issue.
Click to expand...
All configuration in namelist is same.
It only work before V4.3.2 but failed with V4.4 including V4.5.
 
There is no essetial difference at least between WRFv4.3.2 and WRFV4.4. I see no reason why the case cannot run with WRFV4.4...

Please consult your computer manage and ask whether the libraries/compilers are updated. I am suspicious this is a computer issue.
 
Ming Chen said:
There is no essetial difference at least between WRFv4.3.2 and WRFV4.4. I see no reason why the case cannot run with WRFV4.4...

Please consult your computer manage and ask whether the libraries/compilers are updated. I am suspicious this is a computer issue.
Click to expand...
The NoahMP code has been moved to an external repository and is linked to WRF repository via a submodule from V4.4. So I doubt it's the reason.
 
But when you download the code, you must download NoAHMP and include it as a component of WRF system, right?
 
You must log in or register to reply here.
Top