WRF 4.6 Stops Progressing to Next Time Step

[sciguymp]

I am running the same domain for many individual random days.

In 90% of the runs, WRF completes successfully in a reasonable amount of time on my cluster (128 cores per run).

But in the remaining 10%, the model will run as normal for a random interval of time (sometimes approaching the end of the run) before stalling. It never proceeds to the next time stamp, but is still running, and no errors are present.

Any tips for how I might debug this behavior?

I suspect that something is going screwy with the complex terrain (Rocky Mountains), but that is just a hunch based on combing prior forum posts. My namelist parameters are reasonable for what has worked in those prior cases that caused WRF to crash.

 

[Ming Chen]

Can you upload your namelist.input of a stalling case for me to take a look? I am suspicious that something went wrong and resulted in the stalling.

 

[sciguymp]

Sure thing, see attached.

Note that we are trying to perform sub-km runs in complex terrain regions trying to quantify boundary layer parameters.

One clue to the stalling is that it happens more often during the summer than the winter.

 

[Ming Chen]

Hi,
Thank you for uploading the namelist.input.
Your case is a triply nested case over 3 domains with resolution of 5km, 1km, and 200m. Note that 5km is a typical grey-zone resolution for cumulus scheme while the 200m falls within the grey-zone of PBL. Specifically, the assumptions used in PBL scheme is no longer valid for 200m resolution. Meanwhile, this resolution is way too coarse for LES simulation. Under this situation, the model may run or it may fail due to unreasonable physics.
This explains why some of your cases were done successfully while other cases failed.