The GFS grib2 files changed on November 8, 2019.
Before that date SST was filled in for water areas in
TSOIL:0-0.1 m below ground:anl:
TSOIL:0.1-0.4 m below ground:anl:
TSOIL:0.4-1 m below ground:anl:
TSOIL:1-2 m below ground:anl:
which become in metgrid variables
ST000010,ST010040,ST040100,ST100200.
the current metgrid table entry for these fields are
name=ST000010
interp_option=sixteen_pt+four_pt+wt_average_4pt+wt_average_16pt+search
masked=water
interp_mask=LANDSEA(0)
missing_value=-1.E30
fill_missing=285.
flag_in_output=FLAG_ST000010T000010
In the new grib2 files after November 7,2019,
water points are marked as missing with a value of 9.999e20.
The wrf_input file coming out of real.exe had TSLB variable entries in wrfinput_d01 with values of -3 Kelvin at a few places where land mask of GFS grib2 files and land mask from geogrid were different. These strange values for sub soil temperature caused NaN values because they cause logaritm of a negative number.
To try and fix this, I changed the TBL to have these values
name=ST000010
interp_option=sixteen_pt+four_pt+wt_average_4pt+wt_average_16pt+search
masked=water
interp_mask=LANDSEA(0)
missing_value=0.0
fill_missing=0.0
flag_in_output=FLAG_ST000010
Also I did this for the other ST entries.
Then metgrid put out zero for water points which caused real.exe to fill in SST values for water points
and fixed the strange values at those few land points.
I don't know if the my new values are the optimum values but they seemed to cause the desired behavior.
For your info
Allen Lenzen
SSEC University of Wisconsin-Madison, WI
Before that date SST was filled in for water areas in
TSOIL:0-0.1 m below ground:anl:
TSOIL:0.1-0.4 m below ground:anl:
TSOIL:0.4-1 m below ground:anl:
TSOIL:1-2 m below ground:anl:
which become in metgrid variables
ST000010,ST010040,ST040100,ST100200.
the current metgrid table entry for these fields are
name=ST000010
interp_option=sixteen_pt+four_pt+wt_average_4pt+wt_average_16pt+search
masked=water
interp_mask=LANDSEA(0)
missing_value=-1.E30
fill_missing=285.
flag_in_output=FLAG_ST000010T000010
In the new grib2 files after November 7,2019,
water points are marked as missing with a value of 9.999e20.
The wrf_input file coming out of real.exe had TSLB variable entries in wrfinput_d01 with values of -3 Kelvin at a few places where land mask of GFS grib2 files and land mask from geogrid were different. These strange values for sub soil temperature caused NaN values because they cause logaritm of a negative number.
To try and fix this, I changed the TBL to have these values
name=ST000010
interp_option=sixteen_pt+four_pt+wt_average_4pt+wt_average_16pt+search
masked=water
interp_mask=LANDSEA(0)
missing_value=0.0
fill_missing=0.0
flag_in_output=FLAG_ST000010
Also I did this for the other ST entries.
Then metgrid put out zero for water points which caused real.exe to fill in SST values for water points
and fixed the strange values at those few land points.
I don't know if the my new values are the optimum values but they seemed to cause the desired behavior.
For your info
Allen Lenzen
SSEC University of Wisconsin-Madison, WI