Running Single Column Model over Seaice

[chris_riedel]

I am trying to run the single column model (SCM) portion of WRF in the Arctic over seaice. I would like to be able to adjust the seaice fraction within the column. I have ran the most recent version of WRF real and gotten the seaice fraction to work and all. Within the module_initialize_scm_xy.F code, I have changed the XLAND and LANDMASK values to represent that the grid is over water. Furthermore, I noticed in the module_initialize_scm_xy.F code there is section to change the xice value to something other than 0. I have ran the code successfully with xice being set to zero in module_initialize_scm_xy.F and namelist flag fractional_seaice set to zero, which means I am only using a flag and not actually seaice fraction. When I set namelist flag fractional_seaice to one (which means I want to use seaice fraction) and leaving xice set to zero within module_initialize_scm_xy.F I get a Segmentation fault error when trying to run.
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0x7FAA988012F7
#1 0x7FAA988018FE
#2 0x38A563250F
#3 0x278D12D in __module_sf_sfclayrev_MOD_psim_unstable
#4 0x2791C35 in __module_sf_sfclayrev_MOD_sfclayrev1d
#5 0x279677D in __module_sf_sfclayrev_MOD_sfclayrev
#6 0x1EB975C in __module_surface_driver_MOD_sfclayrev_seaice_wrapper
#7 0x1F00B16 in __module_surface_driver_MOD_surface_driver
#8 0x17A01F5 in __module_first_rk_step_part1_MOD_first_rk_step_part1
#9 0x1198A99 in solve_em_
#10 0x106FA0D in solve_interface_
#11 0x47137E in __module_integrate_MOD_integrate
#12 0x406C93 in __module_wrf_top_MOD_wrf_run
Segmentation fault

Furthermore, if I change the xice value in module_initialize_scm_xy.F to 1 I get the same Segmentation fault error which I guess is not all that surprising. I have dug around some into the code to try and figure out what the issue is but I have made no progress. First, I was wondering if this has been done before and is a simple fix? If not, I was wondering if you guys had some ideas on how to resolve this issue.

Thanks so much for your help,
Chris Riedel

 

[chris_riedel]

I figured out my issue. It was setting one of global attributes to a different value.

Chris

 

[Ming Chen]

Chris,
Thanks for the update. Can you clarify which global attributes you have specified to make it work? This information might be helpful for others who experience the similar issue.

 

[tshepherd]

It would be useful to obtain a solution to this issue. I have just started to receive this exact error when running on GNU-compiled WRF v3.8.1 on a cloud-based system inside Docker.

At first I thought it was an issue with the met_em input data, as I am detailing in this forum thread: https://forum.mmm.ucar.edu/phpBB3/viewtopic.php?f=31&t=9879

But, I have since discovered that this invalid memory reference issue persists regardless of met_em input data type. I have even tried to run a previously known-to-run-successfully case that I simulated, and even that now fails. Strangely enough, this error does not occur on an Intel HPC-based compile using the same data. This leads me to think it is compiler-related. Compiling with GNU with flags to be less precise has no effect.

The backtrace reveals the cause of the crash is in:

function psim_unstable(zolf)
      integer :: nzol
      real    :: rzol
        nzol = int(-zolf*100.)
        rzol = -zolf*100. - nzol
        if(nzol+1 .le. 1000)then
           psim_unstable = psim_unstab(nzol) + rzol*(psim_unstab(nzol+1)-psim_unstab(nzol))
        else
           psim_unstable = psim_unstable_full(zolf)
        endif
      return
      end function

I discovered a bug fix that I thought related to this issue, but when I implemented this, it does not solve the issue.

https://github.com/wrf-model/WRF/commit/2d702198352f130424b5c494dcfe48ef4301d876#diff-aa8ddc7037b7a4feefbfcff1719741a80abccd700d23ac28986cb22a5013cac8

Running with fractional_seaice = 0, rather than = 1 results in the issue being resolved every time, but this is not a fix, rather it helps to find the root cause. One cannot simply run without sea ice.

So, I am definitely perplexed by this issue. If anyone has suggestions on remedial steps to take, or information on what some of the function calls do, or where/how to include print statements in the function for diagnosis, I would be happy to hear them!

Key domain settings:

 &domains
 time_step                           = 48,
 time_step_fract_num                 = 0,
 time_step_fract_den                 = 1,
 max_dom                             = 4,
 e_vert                              = 57,    57,    57,    57,
 p_top_requested                     = 5000,
 num_metgrid_levels                  = 38,
 num_metgrid_soil_levels             = 4,
 sfcp_to_sfcp                        =.false.,
 dx                                  = 16668, 5556, 1852, 1852,
 dy                                  = 16668, 5556, 1852, 1852,
 grid_id                             = 1,     2,    3,    4,
 parent_id                           = 1,     1,    2,    2,
 parent_grid_ratio                   = 1,     3,    3,    3,
 parent_time_step_ratio              = 1,     3,    3,    3,
 feedback                            = 0,
 eta_levels                          = 1.0000 , 0.9979 , 0.9958 , 0.9937 , 0.9916 ,
                                       0.9895 , 0.9874 , 0.9853 , 0.9832 , 0.9811 ,
                                       0.9790 , 0.97695, 0.9749 , 0.9728 , 0.9707 ,
                                       0.9684 , 0.9661 , 0.9635 , 0.9609 , 0.9579 ,
                                       0.9549 , 0.95145, 0.9480 , 0.9439 , 0.9398 ,
                                       0.93505, 0.9303 , 0.9246 , 0.9189 , 0.91215,
                                       0.9054 , 0.8974 , 0.8894 , 0.8704 , 0.8481 ,
                                       0.8221 , 0.7922 , 0.7583 , 0.7205 , 0.6791 ,
                                       0.6346 , 0.5877 , 0.5393 , 0.4900 , 0.4407 ,
                                       0.3922 , 0.3450 , 0.2996 , 0.2564 , 0.2156 ,
                                       0.1773 , 0.1417 , 0.1086 , 0.0755 , 0.0475 ,
                                       0.0224 , 0.0000,

Key physics settings in my namelist (4 domains):

 &physics
 mp_physics                          = 5,     5,    5,    5,
 ra_lw_physics                       = 1,     1,    1,    1,
 ra_sw_physics                       = 1,     1,    1,    1,
 radt                                = 15,    15,   15,   15,
 sf_sfclay_physics                   = 1,     1,    1,    1,
 sf_surface_physics                  = 2,     2,    2,    2,
 bl_pbl_physics                      = 5,     5,    5,    5,
 bldt                                = 0,     0,    0,    0,
 cu_physics                          = 1,     1,    0,    0,
 cudt                                = 5,
 isfflx                              = 1,
 ifsnow                              = 1,
 icloud                              = 1,
 surface_input_source                = 1,
 num_soil_layers                     = 4,
 num_land_cat                        = 21,
 sf_urban_physics                    = 0,     0,     0,    0,
 bl_mynn_tkebudget                   = 1,     1,     1,    1,
 bl_mynn_tkeadvect                   = .true., .true., .true., .true.,
 rdmaxalb                            = .false.,
 sst_update                          = 1,
 tmn_update                          = 1,
 usemonalb                           = .true.,
 lagday                              = 150,
 sst_skin                            = 0,
 slope_rad                           = 1, 1, 1, 1,
 prec_acc_dt                         = 60., 10., 10., 10.,
 fractional_seaice                   = 1,
 seaice_threshold                    = 0.,
 windfarm_opt                        = 0, 0, 0, 1,
 windfarm_ij                         = 0,
 windfarm_tke_factor                 = 0.25