Model output changing only every 3 hours

Hi all,

I ran a 72 h forecast using the UFS MR Weather App on Cheyenne everyday during the month of July 2018. I intend to use these simulations to represent meteorology in the Community Multiscale Air Quality (CMAQ) model simulations. I need hourly output. So, I set nfhout: 1. I see that the hourly output is generated but it is changing only every 3 hours, i.e., the simulated values at 00, 01, 02 hours are same and then changes at 03 hours but stays the same for 04 and 05 hours. Is there some other setting that I need to change? 



Hello Rajesh,

Is this strange behavior happening with all variables, such as temperature and winds?




Rajesh, would you be able to share your directory on cheyenne so I could take a look?   Both your source code dir, and the run directory.  I am not able to reproduce this issue at this time, but I do have an old case I ran last summer that shows the same behavior, but it is incomplete.


Hi Laurie,

The source code is located here: /glade/p/ral/nsap/rkumar/FV3/my_ufs_sandbox/cime/scripts/GFSv16_NOAAfire

The model run is located here: /glade/scratch/rkumar/GFSv16_NOAAfire/20180701

Since I ran a 72 h forecast every day, there is one directory for each day under /glade/scratch/rkumar/GFSv16_NOAAfire/

Please let me know if you have any question. Thanks a lot for helping with this. 




Rajesh - This looks like it is an unusual combination of inconsistent configuration settings, and a bug in the UFS MRW App (and possibly also the ufs-weather-model).  I will do a bit more diagnosis and open some issues for those software components.

In the meantime, there is a workaround, and that is to set a namelist parameter.  If you are using the App, and running case.submit, then add this line to the file, user_nl_ufsatm in the CASE directory:

  fhmaxhf = 240.0

If you are running these cycles with a different script, then add that same line to the file, input.nml, in the &atmos_model_nml record.  

This value needs to be longer than your forecast length, 240 is safe, but 72 would work as well.  The "case.submit" task will generate a new input.nml, so that's why there are two ways to work around this.


I'm glad that worked for your case, I've put a known-issue on the ufs-mrweather-app wiki page, and opened an issue to address this in future releases.