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

Timeseries inconsistencies with restart files

S

samuele

New member
Hello, I am testing restart files.
I ran a long simulation from time A to time B while saving restart files at time C and time D, with A<C<D<B, and saving the timeseries .TS files of the long simulation.
Then I ran simulations from the restart files at time C and at time D saving the .TS files.
However, when I plot the variables (T2m, U10m, P) I find small differences (negligible for the surface pressure)...are these considered normal?
On WRF website it is mentioned that results of restart files should be identical!
It is effectively a continuous run made of two or more shorter runs; hence the results at the end of one or more restart runs should be identical to a single run without any restart.
Click to expand...

These discrepancies I find are either at the beginning of the restart file, in the middle or at the end...and they normally converge with the original simulation.
I attached some plots. Orange and cyan are the same restart file run for 1hour or 12 hours respectively, and they overlap.

Thank you in advance for your help :)
Have a nice day
Sam
 

Attachments

  • CRYOWRF_Det_T.pdf
    29.9 KB · Views: 3
  • CRYOWRF_Det_U.pdf
    67.1 KB · Views: 1
  • WRF_Det_T.pdf
    46.3 KB · Views: 2
  • WRF_Det_U.pdf
    47 KB · Views: 1
Sam,

I expect that T2 and U10 in wrfout files from continuous run and from restart run should be identical. I don't have an immediate answer why they are different in your results. Did you look at these variables in wrfout or in wrfrst?
 
Dear Ming Chen, thank you for your reply!
I am looking at the variables from the timeseries files saved with tslist.
Things also get a bit worse when we try to save different variables (not from standard WRF tslist) as in the attachment.
 

Attachments

  • uStar.png
    uStar.png
    37.9 KB · Views: 3
This is definitely wrong.....

Did you modify codes? If so, can you run the officially released version of WRF and check the results again?

Please keep me update of your result. Thanks.
 
In one case yes we did modify the codes...but I also tested the official release of WRF 4.7.1 (files labeled WRF).
What it is strange to me with the WRF release is that using the same restart file twice, I got different results, one of which completely overlaps (cyan vs green). I did notice that during the tests the cluster I was using experienced filesystem issues in I/O operations...could that affect restart files?
I will try in the near future to run simulations on a local workstation and test the restart files again.

In the case of our modified code, what could cause these jumps?
I'll keep you updated, and thanks!!:)
 
When you retart WRF, did you run with the same number of processors as that in your initial run? Note that different number of processors will lead to slightly different results.

As for the codes you modified, I would suggest that you follow the same approach used in the original codes.
 
In some test cases I started from 2 nodes 24cpus-per-node to 1 node 48 cpus-per-node, because I couldn't load the restart files when trying on two nodes...do you know why?

Thanks for the tips!
Sam
 
Hi Sam,
This is a computer issue, which could be specific for your HPC. Have you talked to your computer manager?
 
Last edited:
Yes, and they said it could be related to the application...which is strange to me since the problem appears only on multiple nodes...
Thank you for your feedback!
 
You must log in or register to reply here.
Top