Segmentation fault with real.exe and MOZCART

[varun1490]

I have run wrf using chem_opt=10 and it works fine but when I want to use chem_opt=112 & 131 I am facing the below error for both chemistry options.

Assume Noah LSM input
d01 2018-09-04_00:00:00 Timing for processing 0 s.
STARTING WITH BACKGROUND CHEMISTRY
Subroutine input_chem_profile:

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

Backtrace for this error:
#0 0x155554bb97c2 in ???
#1 0x155554bb8995 in ???
#2 0x155551f26d8f in ???
#3 0x275698d in ???
#4 0x275c605 in ???
#5 0x44eacd in ???
#6 0x44f68b in ???
#7 0x407400 in ???
#8 0x155551f11eaf in ???
#9 0x155551f11f5f in ???
#10 0x407434 in ???
#11 0xffffffffffffffff in ???

I have tried this solution mentioned on this previous post but did not work for me Segmentation fault using real.exe with MOZART chemistry.

 

[mlanger]

Hi all,

It seems that this is also a problem with other Moz[c]art variants (101, 112, 201, 202,...).
I tried WRFv4.5.[1,2], compiled with ifort and gfortran, and I got the same errors.
This problem ist also discussed here: Segmentation fault using real.exe with MOZART chemistry.
The error occured in module_input_chem_data, subr.make_chem_profile (file module_input_chem_data.F, line 845).
For moz[c]art chemistry the values of the index array iref are not assigned.
For some other chemistry variants iref is assigned in subr.setup_gasprofile_maps,
but for variants of moz[c]art there is only the message: "setup_profile_maps: nothing done for..."
I don't think it makes sense to try out the values of iref (and other arrays) assigned for other chemistry variants.
We need a serious solution from the experts at NCAR.

Matthias
.

 

[varun1490]

I have discussed this in wrf-chem group and I have found out that we only need to run real with no chemistry and then mozbc. And wrf.exe and it work fine.

 

[mlanger]

Unfortunately this does not work for me. real.exe with chem_opt=0 will only shift the problems to the wrf.exe

#=running real.exe with chem_opt=202:
forrtl: severe (408): fort: (3): Subscript #4 of the array STOR has value 0 which is less than the lower bound of 1
==> module_input_chem_data: Subroutine make_chem_profile

#=running real.exe with chem_opt=0: real_em: SUCCESS COMPLETE REAL_EM INIT (as expected)
# AND THEN running wrf.exe with chem_opt=202:
forrtl: severe (408): fort: (3): Subscript #1 of the array XL has value 0 which is less than the lower bound of 1
==> module_input_chem_data: SUBROUTINE bdy_chem_value


The indices of some arrays (STOR, XL) are defined in these routines with the values of iref.
...in subr.make_chem_profile: stor(i,k,j,is) with is=iref(l-1)
...in subr.bdy_chem_value: xl(irefcur,kx+1-k) with irefcur=iref(nch-1)
For some chem_opt values the array iref is assigned in SUBROUTINE setup_gasprofile_maps, but not for moz[c]art chemistry.
With moz[c]art, however, the values are never assigned to iref (except for the default initial value, which depends on the compiler option and is 0 in my case).
> .....
> select case(chem_opt)
> .....
> case (MOZCART_KPP)
> call wrf_debug("setup_profile_maps: nothing done for mozcart_kpp")
> case (T1_MOZCART_KPP)
> call wrf_debug("setup_profile_maps: nothing done for t1_mozcart_kpp")
> case (MOZART_MOSAIC_4BIN_KPP)
> call wrf_debug("setup_profile_maps: nothing done for mozart_mosaic_4bin_kpp")
> case (MOZART_MOSAIC_4BIN_AQ_KPP)
> call wrf_debug("setup_profile_maps: nothing done for mozart_mosaic_4bin_aq_kpp")
> ...

I don't think it makes sense to try out the values for other chemistry options (radm, racm, saprc,...) without knowing whether they are also valid for moz[c]art.
I would be happy to get a a serious solution for useful values from iref.