wrf.exe crashes when setting one more eta-level near surface in high-resolution run

[userwrfch]

Hi,

I am using WRF 4.1.3 to downscale hourly ERA5 data with a grid ratio of 1:5 (25 km, 5 km, 1 km, 200 m, 40 m) over an area in the Rocky Mountains, using SRTM1 topography in d03, d04 and d05.

I managed to get wrf.exe to work with:
TEST 1:
1) Shifting the domains (such that the boundaries are not over mountainous terrain)
2) Using 1-2-1 smooth option in geogrid.tbl
3) Setting w_damping = 1 and damp_opt = 3 (needed for my usecase)
4) Running on 7 nodes with 20 processors each
5) Setting epssm to 0.1, 0.1, 0.1, 0.7, 1.0
6) Setting e_vert to 90 (while keeping the model top at 5000)
7) Setting the time step to 45 s

I would like to increase the number of vertical levels near the surface. At the moment, I have levels at 0 m, 48.21 m, 109.11 m, … I would like to have one more level, e.g., at 25 m.

I set the eta-levels manually, but I can’t get the model to work. I used the same specifications as above, except:
TEST 2:
- Setting epssm very high to 0.1, 0.5, 0.5, 1.5, 2.0
- Reducing the time step to 40 s (a shorter time step will require too much time for me to run the experiments)
- Setting e-vert to 91 and setting the eta-values manually, while using the same eta-levels from above, and including only one level (at 0.997)

The error is:
[n040:19446] *** Process received signal ***
[n040:19446] Signal: Segmentation fault (11)
[n040:19446] Signal code: Address not mapped (1)
[n040:19446] Failing at address: 0xfffffff605ec2448

I tried many different test runs, e.g.:
- Same as TEST 1, but specifying the eta-levels manually to the same ones as TEST 1 --> worked
- Increasing the number of vertical levels to e_vert = 200 --> gives almost the exact same vertical levels near the surface
- Setting the p_top_requested to 10000 --> gives again almost the same vertical levels near the surface

Can you please help me to get one more vertical level near the surface (e.g. at 25 m)? Am I missing anything?

Thanks!

 

[kwerner]

Hi,
If you didn't already do this, I would recommend first following the procedures mentioned here. After you run real.exe without specifying your eta_levels, yourself, you can use the command:

ncdump -v ZNW wrfinput_d01

to see the model-suggested values for eta-levels. From there, you can play around with the values to see what works. Unfortunately, there is no specific formula for determining the right values. It's just a matter of playing around with it a bit. If you're still having problems, I would recommend trying some debugging to determine exactly where in the code the problem is occurring and see if you're able to correct it with that information.

 

[userwrfch]

Thank you very much for the reply! I appreciate it a lot.

Yes, I have already inserted the model-suggested values for eta-levels and tried to modify them, but couldn't make it work so far. I will further try different options for the eta-levels in combination with epssm and reducing the time step, and use debugging.