Open CoryMartin-NOAA opened 2 years ago
The screenshots are for NOAA-15 but should apply to all observation types.
These are the differences in GSI H(x) after changing
prsi
to be computed from delp
rather than ps * bk + ak
. There is a pretty consistent bias that is shown when the colorbar is +-0.1K. It is Old - New, and the H(x) difference is positive. This suggests that there is an impact in the pressure computation difference. Also note, the relatively huge differences in S. America due to topography.
@ADCollard and I have been trying to figure out this problem for a while:
cmp
was executed on the amsua_n19.TauCoeff.bin
file that comes with JEDI and the version that existed already on Orion for CRTM 2.3.0:
amsua_n19.TauCoeff.bin /work/noaa/global/glopara/crtm/crtm_v2.3.0/TauCoeff/ODPS/Little_Endian/amsua_n19.TauCoeff.bin differ: byte 17, line 1
Copying the TauCoeff.bin file to /work2/noaa/da/cmartin/GDASApp/fix/crtm/2.3.0_jedi
and rerunning produces this:
So it seems the CRTM coefficient files provided by JEDI cannot be trusted and we need to use our own CRTM 2.3 little endian coefficient files.
Compare H(x) generated from GSI with H(x) generated by FV3-JEDI in UFO for both unbias corrected and bias corrected radiances for each AMSU-A platform and channel.