Hi everyone,
I have set up the WRF model 4.5.1 for a long-term historical and projection run, spanning 20 years, with forcing data derived from NoeESM2 GCM. The model is configured to run with SLUCM coupled to the Boulac PBL schemes. The outer domain covers Europe at a 12 km grid resolution, while the inner domain operates at a convection-permitting scale over the Alps region.
Typically, the model runs smoothly during the colder part of the year, but it crashes during the summer months, specifically in July or August. Unfortunately, I do not receive any clear error messages that would help diagnose the issue.
We have managed to overcome the crashing points so far using several approaches:
The challenge is that the model never crashes at the same point or line, making it difficult to pinpoint the exact cause of the failure. I suspect that the issue might be related to the complex terrain in the Alps, since the same configuration runs without issues for a convection-permitting domain over northern Europe.
I have tested various timesteps values (e.g., 60s, 30s) and increased epssm values (up to 0.5). Additionally, I attempted to change the PBL scheme, as the WRF debug run suggests that the crash might originate in the Boulac PBL scheme (see an example of the rsl.error.0000 file here).
I am also attaching the namelist.input that I am using.
What could be the cause of these crashes, and what additional tests or adjustments should I consider?
I have set up the WRF model 4.5.1 for a long-term historical and projection run, spanning 20 years, with forcing data derived from NoeESM2 GCM. The model is configured to run with SLUCM coupled to the Boulac PBL schemes. The outer domain covers Europe at a 12 km grid resolution, while the inner domain operates at a convection-permitting scale over the Alps region.
Typically, the model runs smoothly during the colder part of the year, but it crashes during the summer months, specifically in July or August. Unfortunately, I do not receive any clear error messages that would help diagnose the issue.
We have managed to overcome the crashing points so far using several approaches:
- Performing a simple restart from wrfrst file without changing anything, or just changing the restart point
- The run past the crashing point using WRF compiled with -O2 optimizations instead of -O3 (note that we are using WRF v4.5.1 compiled with Intel)
- The run past the crashing point using WRF compiled in debug mode (compiled usising configure -d; version compiled with configure -D could not run)
The challenge is that the model never crashes at the same point or line, making it difficult to pinpoint the exact cause of the failure. I suspect that the issue might be related to the complex terrain in the Alps, since the same configuration runs without issues for a convection-permitting domain over northern Europe.
I have tested various timesteps values (e.g., 60s, 30s) and increased epssm values (up to 0.5). Additionally, I attempted to change the PBL scheme, as the WRF debug run suggests that the crash might originate in the Boulac PBL scheme (see an example of the rsl.error.0000 file here).
I am also attaching the namelist.input that I am using.
What could be the cause of these crashes, and what additional tests or adjustments should I consider?
Last edited: