cu_physics=3

Giamik · Jul 21, 2023

I am running WRF v4.5 in 2 domains(one nested) configuration using GFS data. They cover the Black and Caspian Seas (the first one with a 9km resolution) and Georgia (the nested one with a 3km resolution).
By the user guide recommendation (WRF Physics — WRF Users Guide documentation)
"Domains with grid spacing >=3km to <=10km: This is a “gray zone” where cumulus parameterization may or may not be necessary. If possible, try to avoid domains this size, but if it is unavoidable, it is best to use either the Multi-scale Kain Fritsch or Grell-Freitas scheme, as these take this scale into account."
When I use the Grell-Freitas scheme (cu_physics=3) real.exe runs successfully but wrf.exe stops at the very beginning with the error:
------------------------------------------------------------------------------------------------------------------
d02 2023-07-21_00:00:12 calling inc/PERIOD_BDY_EM_MOIST_OLD_inline.inc
d02 2023-07-21_00:00:12 calling inc/HALO_EM_A_inline.inc
d02 2023-07-21_00:00:12 calling inc/PERIOD_BDY_EM_A_inline.inc
d02 2023-07-21_00:00:12 calling inc/HALO_EM_PHYS_A_inline.inc
d02 2023-07-21_00:00:12 Top of Radiation Driver
d02 2023-07-21_00:00:12 calling inc/HALO_PWP_inline.inc
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
wrf.exe 000000000386115A for__signal_handl Unknown Unknown
------------------------------------------------------------------------------------------------------------------
When I use Multi-scale Kain Fritsch (cu_physics=11) real.exe does not run!
But with cu_physics=1 and cu_physics=5 model runs flawlessly!
Here are my namelist files.

Can anyone help me?

Deleted member 3607 · Jul 21, 2023

Good afternoon @Giamik

I'm not very familiar with the physics parameterizations but I can at least show you what I do with my double nest 9km/3km over Texas.

cu_physics = 11, 0,

The Multi-scale Kain Fritsch scheme is used for the grey zone of 9km. Then 0 for 3km since that is where convection is explicitly resolved.

Giamik · Jul 21, 2023

Many thanks, Whatheway,

Please show me Your namelist.input.

Best regards
George

Deleted member 3607 · Jul 21, 2023

Giamik said:
Many thanks, Whatheway,

Please show me Your namelist.input.

Best regards
George

@Giamik

I can't show all of it but I can show part of it.

Code:

 &physics
 physics_suite                       = 'CONUS'
 mp_physics                          = 28,   28,   
 use_aero_icbc                       = .F.
 aer_opt                             = 0
 ra_lw_physics                       = 4,     4, 
 ra_sw_physics                       = 4,     4,
 levsiz                              = 59
 paerlev                             = 29
 cam_abs_dim1                        = 4
 cam_abs_dim2                        = 45
 radt                                = 10,    10,
 sf_sfclay_physics                   = 4      4,
 sf_surface_physics                  = 4,     4,
 sf_urban_physics                    = 0,     0,
 bl_pbl_physics                      = 4,     4,
 bldt                                = 0,     0,
 cu_physics                          = 11,    0,
 cudt                                = 5,     
 isfflx                              = 1,
 ifsnow                              = 1,
 icloud                              = 1,
 surface_input_source                = 1,
 num_soil_layers                     = 4,
 num_land_cat                        = 21,
 usemonalb                           = .true.
 sst_update                 = 1
 tmn_update                          = 1,
 lagday                              = 150,

/

Giamik · Jul 21, 2023

Many thanks, Whatheway,
Can You show me &dynamics and &noah_mp sections?
for my namelist, it doesn't work!
&physics
mp_physics = 28, 28,
ra_lw_physics = 4, 4,
ra_sw_physics = 4, 4,
radt = 9, 9,
sf_sfclay_physics = 4, 4,
sf_surface_physics = 4, 4,
bl_pbl_physics = 4, 4,
bldt = 0, 0,
cu_physics = 11, 0,
cudt = 0, 0,
num_soil_layers = 4,
prec_acc_dt = 60, 60,
usemonalb = .true.,
isfflx = 1,
ifsnow = 0,
icloud = 1,
sf_urban_physics = 0, 0,
/
Thanks in advance

Deleted member 3607 · Jul 21, 2023

Giamik said:
Many thanks, Whatheway,
Can You show me &dynamics and &noah_mp sections?
for my namelist, it doesn't work!
&physics
mp_physics = 28, 28,
ra_lw_physics = 4, 4,
ra_sw_physics = 4, 4,
radt = 9, 9,
sf_sfclay_physics = 4, 4,
sf_surface_physics = 4, 4,
bl_pbl_physics = 4, 4,
bldt = 0, 0,
cu_physics = 11, 0,
cudt = 0, 0,
num_soil_layers = 4,
prec_acc_dt = 60, 60,
usemonalb = .true.,
isfflx = 1,
ifsnow = 0,
icloud = 1,
sf_urban_physics = 0, 0,
/
Thanks in advance

Code:

 &noah_mp
 dveg                               = 4,     
 opt_crs                            = 1, 
 opt_sfc                            = 1
 opt_btr                            = 1,
 opt_run                            = 3,
 opt_frz                            = 1,
 opt_inf                            = 1,
 opt_rad                            = 3,
 opt_alb                            = 2,
 opt_snf                            = 4,
 opt_tbot                           = 1,
 opt_stc                            = 3,
 opt_gla                            = 1,
 opt_rsf                            = 1,
/

 &fdda
 /

 &dynamics
 w_damping                           = 0,
 diff_opt                            = 2,
 mix_full_fields                     = .true.,
 tke_drag_coefficient                = 0.,       
 tke_heat_flux                       = 0.,       
 km_opt                              = 4,         
 diff_6th_opt                        = 0,       
 diff_6th_factor                     = 0.12,     
 base_temp                           = 290.
 damp_opt                            = 3,
 zdamp                               = 5000., 5000., 
 dampcoef                            = 0.2,     
 khdif                               = 0,         
 kvdif                               = 0,           
 epssm                               = 0.5, 0.5,     
 non_hydrostatic                     = .true., .true.,
 moist_adv_opt                       = 1,  1,         
 scalar_adv_opt                      = 1,  1,
 use_theta_m                         = 1,                             
 hybrid_opt                          = 2,                             
 gwd_opt                             = 1,                                         
/

Giamik · Jul 21, 2023

Many thanks, Whatheway,
Doesn't work!

Deleted member 3607 · Jul 22, 2023

Giamik said:
Many thanks, Whatheway,
Doesn't work!

Send me your namelist file here I'll take a look at it @Giamik

Giamik · Jul 24, 2023

Hi Whatheway,
They are attached to the first message of the "cu_physics=3" thread.

Deleted member 3607 · Jul 24, 2023

Giamik said:
Hi Whatheway,
They are attached to the first message of the "cu_physics=3" thread.

Afternoon @Giamik

You are using some physics schemes I am not familiar with and know what all they are compatible with.

Perhaps @Ming Chen or @kwerner might know what is all compatible

Ming Chen · Jul 24, 2023

Your namelist.input looks fine and I didn't see anything wrong. However, there are two issues I am concerned:
(1) you run this case with adaptive time step, which sometimes can cause problem during the integration. Can you turn off this option?
(2) you have 75 vertical levels with p_top at 50hPa, indicating that the vertical resolution is way high. Can you reduce the vertical levels to a smaller value, e.g., 51 or 46, then try again?

Giamik · Jul 25, 2023

Ming Chen said:
Your namelist.input looks fine and I didn't see anything wrong. However, there are two issues I am concerned:
(1) you run this case with adaptive time step, which sometimes can cause problem during the integration. Can you turn off this option?
(2) you have 75 vertical levels with p_top at 50hPa, indicating that the vertical resolution is way high. Can you reduce the vertical levels to a smaller value, e.g., 51 or 46, then try again?

Many thanks, Ming,
When I use Multi-scale Kain Fritsch (cu_physics=11) scheme real.exe. stops at the very beginning, without the creation of any wrfbdy_??? and wrfinput_d0? (boundary and initial) files! Do you think this happens because of the adaptive time step? Or this comment concerns to Grell-Freitas scheme?

Thanks in advance

Giamik · Jul 25, 2023

Dear Ming,
I removed the adaptive time step from the namelist and used various time steps (30, 20, 10) and vertical levels (60, 50, 46) with Grell-Freitas scheme (cu_physics=3). The real.exe executes flawlessly but wrf.exe stops at the very beginning with this output:
----------------------------------------------------------------------------------------------------------------------------
d02 2023-07-25_00:00:00 calling inc/HALO_EM_MOIST_E_5_inline.inc
d02 2023-07-25_00:00:00 calling inc/HALO_EM_SCALAR_E_5_inline.inc
Timing for main: time 2023-07-25_00:00:03 on domain 2: 1.37573 elapsed seconds
d02 2023-07-25_00:00:03+01/03 calling inc/HALO_EM_MOIST_OLD_E_7_inline.inc
d02 2023-07-25_00:00:03+01/03 calling inc/PERIOD_BDY_EM_MOIST_OLD_inline.inc
d02 2023-07-25_00:00:03+01/03 calling inc/HALO_EM_A_inline.inc
d02 2023-07-25_00:00:03+01/03 calling inc/PERIOD_BDY_EM_A_inline.inc
d02 2023-07-25_00:00:03+01/03 calling inc/HALO_EM_PHYS_A_inline.inc
d02 2023-07-25_00:00:03+01/03 Top of Radiation Driver
d02 2023-07-25_00:00:03+01/03 calling inc/HALO_PWP_inline.inc
-----------------------------------------------------------------------------------------------------------------------------

Ming Chen · Jul 25, 2023

Such kind of error message cannot provide any helpful information for us to figure out what is wrong. However, if the model crashes immediately with segmentation fault, it often indicates that either the input data is wrong, or there is no sufficient memory for running the case. Since you can run the same case with other options, I suppose the input data is correct. Please try with more number of processors (which will give you larger memory) and hope this case would work.

It it still fails, I would suggest that you recompile WRF in debug mode, i.e., ./configure -D, then rerun this case with debug_level = 0.
The rsl files will contain information where the model crashed first. This will us some hints why the model crashed.

Giamik · Jul 26, 2023

Hi Ming,
When I run the model I set "ulimit" values to Unlimited ($ ulimit -s unlimited).
Then I run the model:
$ mpirun -n 4 ./wrf.exe (I tried 10 CPUs but the results are same)
memory usage is not more than 60% of total RAM (16Gb).
but the output is:
starting wrf task 0 of 4
starting wrf task 1 of 4
starting wrf task 2 of 4
starting wrf task 3 of 4

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 0 PID 7025 RUNNING AT server
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 2 PID 7027 RUNNING AT server
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================
I have uploaded namelist and geographical files for my domain, but I have 4 more met netcdf files:

24989890 met_em.d01.2023-07-26_00:00:00.nc
24992081 met_em.d01.2023-07-26_03:00:00.nc
11644334 met_em.d02.2023-07-26_00:00:00.nc
11704647 met_em.d02.2023-07-26_03:00:00.nc
But I couldn't upload them due to their large size. (If You give me the right to upload such large files I'll upload them also!)
This data are correct because if one changes cu_physics=3 on 5 the model completes successfully!

Sarath Guttikunda · Jul 26, 2023

time_step = 10,

Isn't this too small?
For 9km, this can be 40 at least?

Giamik · Jul 26, 2023

Sarath Guttikunda said:
time_step = 10,

Isn't this too small?
For 9km, this can be 40 at least?

Hi Sarath,
but even it can't help! I tried time_step = 30, and 20 also!

Giamik · Aug 4, 2023

Giamik said:
Sarath

Hi Sarath,
Did You manage to run Your namelist for my domain?

Sarath Guttikunda · Aug 5, 2023

Giamik said:
but even it can't help! I tried time_step = 30, and 20 also!

The files you uploaded could not be used.

One because of netcdf mismatch. Two -- don't know why your met_em files have different sizes for different times for the same domain. I have never seen that.

Adjusted my namelist files to match your domain and configuration.
The attached namelists works -- real rsl file is attached
Ran wrf for a few minutes only and it seems to run okay

Hope this helps. All the best.

Ming Chen · Aug 9, 2023

Would you please change the options below from:

cu_rad_feedback = .true., .false.,

cu_diag = 1, 0,

to

cu_rad_feedback = .false., .false.,

cu_diag = 0, 0,

Then try again? Please let me know whether it works.

cu_physics=3

New member

Attachments

Deleted member 3607

Guest

New member

Deleted member 3607

Guest

New member

Deleted member 3607

Guest

New member

Deleted member 3607

Guest

New member

Deleted member 3607

Guest

Moderator

New member

New member

Moderator

New member

Attachments

New member

New member

New member

New member

Attachments

Moderator