wrfout timing issues with adaptive time stepping in WRF 4.2.1

[mastdk]

I'm bringing up and old topic which never seemed to have been resolved.

In 2010 Michael had the same issue with WRF 3.2
https://mailman.ucar.edu/pipermail/wrf-users/2010/001770.html

Problem:
We run various WRF models and use adaptive time stepping to get outputs as a stream instead of waiting for all outputs to be completed. What we expect is the output times to be equidistant eg. 18:00:00, 18:15:00, 18:30:31, 18:45:00. Instead its starts to get off by a few seconds up to 90s based on length of run.

I have tried various combinations, based on forum research, but none have succeeded so far. I have attached our input files.

use_adaptive_time_step              = .true.,
step_to_output_time                 = .true.,

Output from: ncdump -v Times

Times =
  "2021-09-20_18:00:00",
  "2021-09-20_18:15:31",
  "2021-09-20_18:30:31",
  "2021-09-20_18:45:31" ;

Any help will be greatly appreciated!

 

[kwerner]

Hi,
Can you send your rsl.out.0000 file so that I can take a look? Thanks!

 

[mastdk]

Hi kwerner,

Yes thank you for taking a look

Hi,
Can you send your rsl.out.0000 file so that I can take a look? Thanks!

I have attached rsl.out.

View attachment rsl.out.0000.txt

 

[kwerner]

Thanks for sending that! I notice your namelist.input file is configured differently - probably for another script to read? Is it possible for you to run a test with a standard namelist.input file (perhaps modifying the default namelist for one specific run) to see if that makes any difference? The code specifically looks for values for "history_interval" and I want to make sure your unique namelist is not causing the problem. Thanks!

 

[mastdk]

Hi @kwerner

Thanks for taking a look at this. I have attached the generated namelist.input that WRF is using. We use a template that automatically changes certain parameters.

Thanks for sending that! I notice your namelist.input file is configured differently - probably for another script to read? Is it possible for you to run a test with a standard namelist.input file (perhaps modifying the default namelist for one specific run) to see if that makes any difference? The code specifically looks for values for "history_interval" and I want to make sure your unique namelist is not causing the problem. Thanks!

View attachment namelist.input

 

[kwerner]

Hi,
Is it possible for you to run a test with a non-automatically-generated namelist? I'd like to see if you still have the same problem when not using that type of namelist. Thanks.