WRF (v4.5.1) "FATAL ERROR singular matrix in gaussj"

G

gibbsc

New member
Hi all,

I'm writing to see if anyone has any idea what is going on with this error I am having running wrf.exe.

Currently I am running a 1 day test for my setup using BEP+BEM. I am able to run WPS and real.exe with no issue but wrf.exe gives me this error:

"FATAL ERROR... singular matrix in gaussj"

I have attached my namelist.input and rsl.error.0000 files as well as a sample met_em file.
 

Attachments

We did not see this error before. Could you please provide more info (e.g., your namelist.input, study domain, urban parameter and map data source in WPS used, etc.)?
 
For some reason when I try to upload files I get "The upload failed because the temporary directory was missing. The site administrator will need to resolve this before any files can be uploaded". So for now I will have to provide some more detail here.

My namelist.input and domain should be accessible through my initial attached files. I used the default land surface data, nlcd_2011, and used w2w to modify my geo_em files for LCZ use.

I've since done some further looking into this and it looks like I have NaN showing up in the gaussj subroutine.

For example from rsl.out.0000:
"
>>> gaussj: entering outer loop
NaN detected at a( 13 , 14 )
NaN detected at a( 14 , 13 )"
...
"Row icol before failure:
1.7520E-04 2.0295E-04 2.0821E-04 9.3975E-05 0.0000E+00 0.0000E+00 3.0917E-03 4.4776E-03 6.4560E-03 4.4734E-03 0.0000E+00 0.0000E+00 NaN 0.0000E+00 0.0000E+00
b(icol)= NaN
FATAL: pivot a(icol,icol) == 0 at i= 14 , icol= 6
a(icol,icol)= 0.0000000E+00"

I'm not quite sure how to fix this though. Any help would be much appreciated!
 
This is likely caused by physics incompatibility issue that led to NaN values. Have you tried to change PBL or microphysics schemes or model timestep to see if the issue is still there?
 
I've tried various combinations of a time step of 27, 15, and 9, and I've tried microphysics options 3 and 4, all of which with no change. I haven't tried changing the PBL scheme so perhaps I'll try that next.
 
An update - I've since found out there was some issue with the version of module_sf_bep_bem.F I was using. I have since successfully ran WRF with a corrected version.
Thank you for the help!
 
You must log in or register to reply here.
Back
Top