Problem creating CLM3.0 land surface dataset using default mksrf.nc raw datasets

[madeleine_patterson77@gmail_com]

Hi,

I have run CAM 3.1 with a land surface input dataset made with the default mksrf_XXXXX.060926.nc raw datasets after specifying the namelist variable FSURDAT = ' ' (i.e. empty string).

The only thing I changed was the number of soil colours in mksoilcol.F90; I changed soil_color_i from 8 to 20 (I had a look at the default mksrf_soilcol.060929.nc, and there were values up to 20!) (prior to this change CAM was crashing before it created a new surface dataset with the raw datasets.)

I also set the initialisation namelist variable to an empty string, as when I tried to set it to the default clm3.0 initialisation file, I get an error message telling me "can not make surface data when finidat is already specified set finidat to empty string in namelist".

CAM runs ok for a little while, but then crashes, however it does produce a land surface dataset that looks fine, (no Nans or anything unusual).

I have then run the model with FSURDAT = 'path/new_landsurface_filename.nc'.

CAM starts off fine, the land model is successfully initialised, but then a series of segmentation faults occur and the model crashes, e.g:

print_memusage: size, rss, share, text, datastack=* 133957** 32094*** 2483**** 808****** 0
print_memusage iam** 6 stepon after dynpkg. -1 in the next line means unavailable
print_memusage: size, rss, share, text, datastack=* 134005** 33722*** 2721**** 808****** 0
[cxxx:18844] *** Process received signal ***
[cxxx:18844] Signal: Segmentation fault (11)
[cxxx:18844] Signal code: Address not mapped (1)
[cxxx:18844] Failing at address: 0xfffffffc0eb8ca20
[cxxx:18846] *** Process received signal ***
[cxxx:18845] *** Process received signal ***
[cxxx:18846] Signal: Segmentation fault (11)
[cxxx:18846] *** End of error message ***
[cxxx:18844] *** End of error message ***

print_memusage iam** 1 stepon after dynpkg. -1 in the next line means unavailable
print_memusage: size, rss, share, text, datastack=* 134042** 34602*** 3154**** 808****** 0
print_memusage iam** 0 stepon after dynpkg. -1 in the next line means unavailable
print_memusage: size, rss, share, text, datastack=* 139792** 39115*** 3171**** 808****** 0
[cxxx:18843] *** Process received signal ***
[cxxx:18843] Signal: Segmentation fault (11)
[cxxx:18843] Signal code: Address not mapped (1)
[cxxx:18843] Failing at address: 0xfffffffc0ec4a620

etc...

According to the error log file there are a couple of messages repeated after a while, i.e:

**** (1)*****a few of these:

ERROR in srfsub********************** NaN


******(2)****and lots of these:

*QNEG4 WARNING from TPHYSAC , lchnk =* 33; Max possible LH flx exceeded at*** 2 points.* Worst excess =********* NaN at i =** 49
*QNEG4 WARNING from TPHYSAC , lchnk =* 33; Max possible LH flx exceeded at*** 2 points.* Worst excess =********* NaN at i =** 50
*QNEG4 WARNING from TPHYSAC , lchnk =* 33; Max possible LH flx exceeded at*** 1 points.* Worst excess =********* NaN at i =** 51

I think the "ERROR in srfsub' comes from the ice model ice_dh.F (not sure if this is the cause of the seg fault).

I had a look at the default land surface dataset that comes with CAM3.1 (clms_2x2.5_c031104.nc), and the global attributes say it is composed of a slightly different set of mksrf.nc files (that date from 11/11/2003), e.g. mksrf_soilcol_clm2.nc. Does anyone know if this soil color dataset has 8 soil colors, and whether this is perhaps this might be the cause of my problems?

Does anyone have any advice for me as to why the seg faults are occurring? Or why I got an initial error (and crash) when using mksrf_soilcol.060929.nc with 8 soil colors? Has anyone else had this problem with the soil color mksrf.nc dataset? Can someone suggest the correct clm namelist settings to create the land surface surface datafile from the mksrf.nc files?

Thanks in advance for any advice anyone has!

 

[madeleine_patterson77@gmail_com]

Hi,

I just thought I'd post an update to this issue: the solution to get CAM3.1 to make an input land surface dataset using the mksrf_x.f90 utilities and the rawdata netcdf files is to use the rawdata files (e.g. mksrf_pft.nc) that are packaged with CLM 2.1 and not CAM 3.0! If you specify a blank string for fsurdat in the CLM namelist when running cam, and then specify the rawdata files that cam packaged with the CLM 3.0, CAM keeps crashing. Yet I specified the CLM 2.1 rawdata files and CAM worked perfectly, creating the land surface data input file for CLM without a problem.

Could it be that the wrong rawdata files were packaged with CLM 3.0? (and yes, my colleagues and I checked, we are running CLM 3.0 within CAM3.1). All very strange...

Cheers,
M

 

[slevis]

Dear Maddy,

Glad to hear that you have a workaround for your problem. Thank you for using this forum to share problems and solutions with the clm community.

For your and other users' information, clm support staff did not respond to your first message because we do not support the clm3 (released in 2004) anymore. Since April 2010 we support the clm4, the land component of the ccsm4 and cesm1.

Regards,
Sam Levis