Numerical integration fails at the time it reaches the radiation timestep

[geraldogomes]

Dear WRF Users,

I'm trying to run a case of extreme rain that occurred in January 2024 with 500 meters of horizontal resolution (320 x 250), with a timestep of 2s and 60 processors. I am using as a boundary condition a 1 km simulation that I previously carried out with WRF-4.3 without errors. The numerical integration fails at the time it reaches the radiation timestep (radt = 9 min), see below.

Timing for main: time 2024-01-13_06:08:54 on domain 1: 1.57103 elapsed seconds
Timing for main: time 2024-01-13_06:08:56 on domain 1: 1.58430 elapsed seconds
Timing for main: time 2024-01-13_06:08:58 on domain 1: 1.55535 elapsed seconds
Timing for main: time 2024-01-13_06:09:00 on domain 1: 1.56854 elapsed seconds

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0x7F1A73018E08
#1 0x7F1A73017F90
#2 0x7F1A71FB04BF
#3 0x1CFEBBC in taugb3.7253 at module_ra_rrtmg_lw.f90:?
#4 0x1D213BD in __rrtmg_lw_taumol_MOD_taumol
#5 0x1D453B8 in __rrtmg_lw_rad_MOD_rrtmg_lw
#6 0x1D566DE in __module_ra_rrtmg_lw_MOD_rrtmg_lwrad
#7 0x1741A2E in __module_radiation_driver_MOD_radiation_driver
#8 0x1809EB9 in __module_first_rk_step_part1_MOD_first_rk_step_part1
#9 0x131593C in solve_em_
#10 0x11B4DE0 in solve_interface_
#11 0x47B88E in __module_integrate_MOD_integrate
#12 0x40A513 in __module_wrf_top_MOD_wrf_run

I tried several radt values (less than 9 min) but was unsuccessful. Any idea what it could be? A few months ago I ran the model with 250 meters for another case and in another city and it did not show this type of error for me.

Thanks

 

[kwerner]

Hi,
When you say that you're using a 1-km WRF simulation as the boundary condition, do you mean that you're using the ndown program to run this? If not, can you let me know the procedure for using the wrf output as an input boundary condition, and where the initial conditions come from? Please also package all of your rsl* files into a single *.tar file and attach that. Thanks!

 

[geraldogomes]

Hi,
When you say that you're using a 1-km WRF simulation as the boundary condition, do you mean that you're using the ndown program to run this? If not, can you let me know the procedure for using the wrf output as an input boundary condition, and where the initial conditions come from? Please also package all of your rsl* files into a single *.tar file and attach that. Thanks!

No, I meant that I took an existing 1 km WRF output simulation, transformed it into grib via Unified Post Processor (UPP) and used it as a boundary condition for another simulation but this time with 500 meters of horizontal resolution. I did exactly as recomended in the UPP FAQs: FAQ's | dtcenter.org. I've already done this for another 1-km simulation but with 250 meters and the model ran successfully.

 

[kwerner]

Thanks for the clarification. That makes sense. Is it possible for you to try using more processors to see if that makes a difference? If not, can you package all of your rsl* files into a single *.tar file and attach that so I can take a look? Thanks!

P.S. - I will be out of the office through next week, but will get back to you as soon as I'm able after my return. I apologize for the inconvenience.