Error in while running wrf.exe for ERA5 data

[june0497]

I wanted to run WRF ARW model at 1km using ERA5 reanalysis, for testing I just ran for 15 days but
However, when running wrf.exe, I encounter the following error:
forrtl: severe (174): SIGSEGV, segmentation fault occurred. I have tried multiple ways to fix it but I am not sure why is that happening.

I ran WPS, and generated the met_em files successfully, my domain configuration
for my WPS namelist is following:
&share
wrf_core = 'ARW',
max_dom = 1,
start_date = '2016-03-01_00:00:00',
end_date = '2016-03-15_00:00:00',
interval_seconds = 21600,
io_form_geogrid = 2,
/

&geogrid
parent_id = 1,
parent_grid_ratio = 1,
i_parent_start = 1,
j_parent_start = 1,
e_we = 1080,
e_sn = 1111,
geog_data_res = 'default',
dx = 1000,
dy = 1000,
map_proj = 'mercator',
ref_lat = 15.00,
ref_lon = 79.00,
truelat1 = 15.0,
truelat2 = 0.0,
stand_lon = 79.0,


I am attaching my namelist.input too which successfully generated bdy and input files while running ./real.exe

also for the running in hpc I used the below configuration.
Also used 10 nodes too.
SBATCH -N 10
#SBATCH --ntasks-per-node=48

mpirun -np 480 ./wrf.exe

I have run the model successfully for the hurricane matthew example but for this single domain run, it is not running.

Would really appreciate your help on this.

 

[kwerner]

Hi,
Can you package all of your rsl.error* files into a single *.tar or zipped file and attach that here? Can you also let me know the resolution of the input ERA5 data? Thanks!

 

[june0497]

Hi,
Thank you so much for the reply. The resolution of ERA5 data is the native 0.25∘ deg . I am attaching the zipped rsl.error* files and also the script that I used for the input ERA5
Would really appreciate help on this.

 

[kwerner]

Thanks for sending those. I think there may be a couple of issues here:

1) The input data are 0.25 deg, or ~27 km, meaning the grid ratio between the input and your domain is about 27:1, which is too much of a difference. It's recommended that grids should be no more than about 5:1, including between the input grid and your domain 01. This doesn't have to be exact, but since you're interested in a 1km resolution grid, I would recommend using nesting and adding a parent domain between the input and the 1km grid. For e.g., you could run a simulation with dx = 5000, 1000, and then parent_grid_ratio = 1, 5.

2) I do see CFL errors in your rsl files. If you search 'cfl' you will find prints like this:

rsl.error.0006:d01 2016-03-01_00:04:42            1  points exceeded v_cfl = 2 in domain d01 at time 2016-03-01_00:04:42 hours
rsl.error.0006:d01 2016-03-01_00:04:42 Max   W:    355     31      3 W:   -8.80  w-cfl:    2.02  dETA:    0.01
rsl.error.0006:d01 2016-03-01_00:04:42            2  points exceeded v_cfl = 2 in domain d01 at time 2016-03-01_00:04:42 hours
rsl.error.0006:d01 2016-03-01_00:04:42 Max   W:    355     32      4 W:  -11.02  w-cfl:    2.40  dETA:    0.01
rsl.error.0006:d01 2016-03-01_00:04:42            2  points exceeded v_cfl = 2 in domain d01 at time 2016-03-01_00:04:42 hours
rsl.error.0006:d01 2016-03-01_00:04:42 Max   W:    355     31      3 W:   -5.82  w-cfl:    2.54  dETA:    0.01

This indicates the model has become unstable. I notice that you have p_top_requested = 10000. Is there a reason why you changed it to 10,000, instead of the default of 5000? As long as the input data supports it, setting that to 5000 may correct that error. Otherwise, see Segmentation Faults and CFL Errors for additional information and solutions.