expected input for 'config_sfclayer_scheme = sf_noah' option

D

dkmiller

New member
I have attempted an MPAS (v 8.3.1) simulation in which I've selected the config_physics_suite = 'convection_permitting' option in namelist.atmosphere, which defaults to the 'sf_noah' [Noah (WRF 4.5)] option under the physics config_sfclayer_scheme. The model crashes after reading in the init files, but before the first time step of the model (see attached file). Guidance [https://www2.mmm.ucar.edu/projects/mpas/tutorial/Boulder2026/lectures/Debugging_MPAS_202608.pdf] suggests that input from either the LSM and/or the MP scheme was missing. I built the MP tables for the Thompson MP scheme, so that's not what is missing. My concern is that my choice of "config_noahmp_static = false" in the 'namelist.init_atmosphere' file may have bypassed an important initialization. I assumed since sf_noah (and NOT sf_noahmp) is the default convection permitting option that this choice of 'false' was correct. What static tables (if any) are expected as input for the sf_noah [Noah (WRF 4.5)] option? I'm not sure where I need to focus to correct this input problem.

I've attached "ncdump -h ..." output of the init file ('dynamic') and the sfc_update files to show what was included as part of the initialization. Would it make a difference if the sfc_update file was created starting with the date 22 Sep 2024, while the actual MPAS init time is 24 Sep 2024? I assumed MPAS would automatically parse the sfc_update file until it located the analysis time that corresponded to 24 Sep 2024 (init time of the simulation).

Thank you for any guidance/ suggestions you can provide!
 

Attachments

@ dkmiller

Your log file indicates this is a regional MPAS run, please let me know if I am wrong. It doesn't contain any error message. Did you find any message in your log.err file that may be helpful for me to figure out what is wrong?

For your question of sst_update file that has different initial time than that in the initial condition, I have never done such kind of cases and I am not sure of the answer. But based on the codes, I expect that the model should check the time and skip sst data prior to the initial time. Would you please create a new sst_update file that has same initial time as that of your case, then try again? Please let me know whether the model performs as expected.

Also, can you upload your namelist.init_atmopshere, streams.init_atmosphere (for both static and initial conditiona generation) and namelist.atmosphere and streams.atmopshere for me to take a look?

if you run grid_rotate to relocate high-resolution mesh, please send me the namelist and region definition files, In addiiton, what is your input data used to create initial and lateral conditions?
 
I tried starting a simulation with a sfc_update file that matched the desired init date/time and the simulation still failed. There are no log.err files created when the simulation fails.

It is possible we're seeing the effects of data files created on derecho using the latest version of MPAS compared to running MPAS (v8.3.1) on a different computer. I will try running the simulation for a brief time on derecho to see if it fails in a similar fashion.
 
I tried running the simulation on derecho and got the following message...

Error termination. Backtrace:
At line 1562 of file mpas_timekeeping.F
Fortran runtime error: Bad integer for item 1 in list input

and again no log.err files were produced.
 

Attachments

@dkmiller

Thank you for testing the SST update issue. It is not surprising that the case still failed.

Your namelist.atmosphere looks fine except that the number of processors you used to run this case is way too small. I am suspicious this is the reason why the case crashed immediately but didn't produce any log.err files.

Can you rerun this case using more processors, e.g., in your namelist.atmosphere, please set

Code: 
&io
    config_pio_num_iotasks = 0
    config_pio_stride = 1

In your job script, please set:

Code: 
#PBS -l select=10:ncpus=128:mpiprocs=128

Note that you need to create x20.835586.graph.info.part.1280 for such a setting.

Please try and let me know how it works. If you run this case in derecho, tell me your case location and I may also be able to repeat your case.
 
You must log in or register to reply here.
Back
Top