Victor99Huang
New member
I've been using WRF v4.4.2 to run a 5-month simulation with 3-hour, 5-km resolution over North China, with Unified Noah LSM, mosaic option on, and single-layer urban canopy model (SLUCM, sf_urban_physics=1). The urban canopy parameter is from GloUCP (Liao et al, 2025), and the urban fraction data is from GAIA global artificial impervious areas dataset. The simulation is driven by ERA5-interim field. The LULC is modified modis with 3 urban categories (1-61, with effective values 1-17 and 51-53).
The problem is that the model crash after running for 20-40 minutes, corresponding to 6-20 hours of integration, the stop due to segmentation fault:
rsl.error.0001:[comput25:25339:0:25339] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0002:[comput25:25340:0:25340] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0009:[comput25:25348:0:25348] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0010:[comput25:25349:0:25349] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
==== backtrace (tid: 15167) ====
0 0x0000000002a3c7fd __module_sf_sfclayrev_MOD_psim_stable() ???:0
1 0x0000000002a4067d __module_sf_sfclayrev_MOD_sfclayrev1d() ???:0
2 0x0000000002a457e5 __module_sf_sfclayrev_MOD_sfclayrev() ???:0
3 0x00000000021e200b __module_surface_driver_MOD_surface_driver() ???:0
4 0x00000000019fc98f __module_first_rk_step_part1_MOD_first_rk_step_part1() ???:0
5 0x000000000147d72e solve_em_() ???:0
6 0x000000000132781b solve_interface_() ???:0
7 0x000000000047d502 __module_integrate_MOD_integrate() ???:0
8 0x000000000047dae2 __module_integrate_MOD_integrate() ???:0
9 0x0000000000405f64 __module_wrf_top_MOD_wrf_run() ???:0
10 0x000000000040591d main() ???:0
11 0x00000000000223d5 __libc_start_main() ???:0
12 0x0000000000405954 _start() ???:0
=================================
I modified the phys/module_sf_sfclayrev.F to find out the specific problem, then found dz8w1d became NaN right before the segmentation fault. The moment that dz8w1d became NaN drifts as the e_vert, time_step, epssm, w_damping changes, I recorded the bad points information and wrote it in the .txt file.
I thought the problem I met may be similar to the following two threads, but the detailed solution was not provided by the thread starters:
DZ8W, ZLVL become NaN value with sf_sfclay_physics=1(MM5), sf_surface_physics=4(NOAH MP)
[SOLVED] Segfault for WRF V4.5 with the new hybrid 100-m MODIS-LCZ dataset using SLUCM urban physics
I will be very grateful if anyone would like to provide his/her insights to solve the problem.
Thank you!
The problem is that the model crash after running for 20-40 minutes, corresponding to 6-20 hours of integration, the stop due to segmentation fault:
rsl.error.0001:[comput25:25339:0:25339] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0002:[comput25:25340:0:25340] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0009:[comput25:25348:0:25348] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
rsl.error.0010:[comput25:25349:0:25349] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe07138d60)
==== backtrace (tid: 15167) ====
0 0x0000000002a3c7fd __module_sf_sfclayrev_MOD_psim_stable() ???:0
1 0x0000000002a4067d __module_sf_sfclayrev_MOD_sfclayrev1d() ???:0
2 0x0000000002a457e5 __module_sf_sfclayrev_MOD_sfclayrev() ???:0
3 0x00000000021e200b __module_surface_driver_MOD_surface_driver() ???:0
4 0x00000000019fc98f __module_first_rk_step_part1_MOD_first_rk_step_part1() ???:0
5 0x000000000147d72e solve_em_() ???:0
6 0x000000000132781b solve_interface_() ???:0
7 0x000000000047d502 __module_integrate_MOD_integrate() ???:0
8 0x000000000047dae2 __module_integrate_MOD_integrate() ???:0
9 0x0000000000405f64 __module_wrf_top_MOD_wrf_run() ???:0
10 0x000000000040591d main() ???:0
11 0x00000000000223d5 __libc_start_main() ???:0
12 0x0000000000405954 _start() ???:0
=================================
I modified the phys/module_sf_sfclayrev.F to find out the specific problem, then found dz8w1d became NaN right before the segmentation fault. The moment that dz8w1d became NaN drifts as the e_vert, time_step, epssm, w_damping changes, I recorded the bad points information and wrote it in the .txt file.
I thought the problem I met may be similar to the following two threads, but the detailed solution was not provided by the thread starters:
DZ8W, ZLVL become NaN value with sf_sfclay_physics=1(MM5), sf_surface_physics=4(NOAH MP)
[SOLVED] Segfault for WRF V4.5 with the new hybrid 100-m MODIS-LCZ dataset using SLUCM urban physics
I will be very grateful if anyone would like to provide his/her insights to solve the problem.
Thank you!