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

WRF hangs when running an LES at 250 m

G

gpriftis

New member
I have been trying to run a one day of simulations over the Java island. While Java has a complex terrain with mountain reach 3km height, the inner domain is over flat terrain. The wrf.exe hangs at 9 UTC no matter all the different combinations of schemes I have tried.

I am running a 3 domain simulation at 4km, 1km, and 250m. I also have 69 vertical level. I have experimented with different time steps but haven't been able to run the simulation successfully so far.

I have included the namelist and error files in the zip file attached below. Any help or guidance would be appreciated!
 

Attachments

  • Archive.zip
    1,000.2 KB · Views: 2
I don't know what the issue is, but can I ask why are you using a 4 grid ratio? It is usually recommended to use odd-numbered ratios of grid spacing (but not too large), so 3 or 5.
 
@gpriftis,

Can you try running this again, but with the following changes:
1) set debug_level = 0. This rarely produces any useful output and just adds a lot of junk to the rsl files, making this very large and difficult to read through.
2) Can you try running this with 144 total processors (so that the decomposition will be an even 12x12 processors)? It's not mandatory that the decomposition be squared, but since there's a significant difference between your 8 processors in one direction, and 16 in the other, I just want to eliminate the possibility that that may be causing any issues.

After that, assuming it will still fail, please package your new namelist.input file, and all of your rsl* files into a single zipped file and attach that. Thanks!
 
I found several CFL errors (like those shown below) in your rsl* files, meaning the model is becoming unstable.

Code: 
rsl.error.0130:d03 2020-08-14_18:06:16+04/05             1  points exceeded v_cfl = 2 in domain d03 at time 2020-08-14_18:06:16+04/05 hours
rsl.error.0130:d03 2020-08-14_18:06:16+04/05 Max   W:    741    512      2 W: -176.33  w-cfl:    2.03  dETA:    0.00
rsl.error.0130:d03 2020-08-14_18:06:17+07/25             1  points exceeded v_cfl = 2 in domain d03 at time 2020-08-14_18:06:17+07/25 hours
rsl.error.0130:d03 2020-08-14_18:06:17+07/25 Max   W:    741    512      2 W: -223.54  w-cfl:    2.38  dETA:    0.00
rsl.error.0130:d03 2020-08-14_18:06:17+07/25             2  points exceeded v_cfl = 2 in domain d03 at time 2020-08-14_18:06:17+07/25 hours
rsl.error.0130:d03 2020-08-14_18:06:17+07/25 Max   W:    741    513      2 W: -158.23  w-cfl:    2.23  dETA:    0.00

This is a common issue over complex terrain. See Segmentation Faults and CFL Errors, which discusses some methods to overcome the instability. You are already doing some of those, but there are a couple additional things you can try.

I'm also curious about the resolution of your input meteorological data (that you processed during WPS/ungrib/metgrid). Your parent domain uses 4km resolution, and we recommend that the grid size ratio between the parent domain and the input data is no more than about 5:1. So if the input data resolution is coarser than about 20km, that could also cause complex terrain issues.
 
I adopted some of the recommendations regarding the CFL errors and added one more domain to account for the resolution ratio between the parent and input data. I also changed the domain 3 to a flatter area. The simulation run successfully.

Thank you for your suggestions and time.
 
You must log in or register to reply here.
Top