Add SatWinds assimilation to JEDI

[HuiLiu-NOAA]

Description:

I will soon start adding major satellite wind (AMVs) types (244-247, 252-260) into JEDI, as planned in the RDAS_JEDI_Observation. The focus is to test and develop required (but may be not available) MPAS-JEDI functions to enable assimilation of those AMVs with MPAS-JEDI. Details will follow soon.

Major AMV types of interest for RRFS:

242 JMA visible 0.8%
243 EUMETSAT visible 0.2%

245 GOES IR 40%
246 GOES WV cloud top 23%
247 GOES WV deep layer 7%

250 JMA WV 2%
252 JMA IR 2.%
253 EUMETSAT IR 3%
254 EUMETSAT WV 2.9%

[keenaneure]

Hi Hui,

I have done some successful MPAS-JEDI tests with some AMVs. If needed or helpful, I am happy to share yamls with you or anything else that may be helpful. This is only with ob type 245 so far. One example of an increment I have shown is in PR #262

[HuiLiu-NOAA]

Thanks @keenaneure.
Yes, it will be helpful to learn what you have done.
Would be good to share your yaml and settings here or email to [email protected].

[HuiLiu-NOAA]

It is noticed that there was no GOES AMVs in the CONUS domain from the rap prepbufr dump, as seen from RDASApp/fix/expr_data/mpas_2024052700/obs/ioda_satwnd_dc.png for 2024052700z. This seems a bit strange.

UPDATE: It turns out that the prepbufr files are not used for AMVs in GSI. Instead, the regular bufr files are used.

[HuiLiu-NOAA]

I processed the RAP bufr data for 2024052700 cycle and extracted AMVs, and see a lot of GOES AMVs in CONUS domain.

It turns out that prepbufr format file is used to produce the ioda file: RDASApp/fix/expr_data/mpas_2024052700/obs/ioda_satwnd.nc, which shows large difference vs the bufr file.

[keenaneure]

Specifically, where did you find these RAP data that are plotted? Also, I am not able to visualize (ncview) or ncdump anything from ioda_satwnd.nc in RDASApp/fix/expr_data/mpas_2024052700/obs/ioda_satwnd.nc. I return an error stating "can't open file" with ncview or "Unknown file format" with ncdump -h

[HuiLiu-NOAA]

@keenaneure,
I used offline_domain_check.py in RDASApp/rrfs-test/IODA/ to get the plot.
also modified the bufr2ioda yaml.

See bufr_satwnd_emc.yaml and emc.sh in
/scratch1/NCEPDEV/fv3-cam/Hui.Liu/RDASApp20250201/RDASApp/expr/mpas_2024052700/AMV_obs/ioda/

The HPSS dir is /NCEPPROD/hpssprod/runhistory/rh2024/202405/20240527/

[HuiLiu-NOAA]

for the ncdump issue, may need to load a newer version RDAS modules.

e.g.,
module purge
module use /scratch1/NCEPDEV/fv3-cam/Hui.Liu/RDASApp2024050201/RDASApp/modulefiles/
module load RDAS/hera.intel.lua
module load EVA/hera.lua
module load ncl

[keenaneure]

@HuiLiu-NOAA concerning the lack of CONUS AMVs in ioda_satwnd.nc. I did some digging and based on 'satelliteIdentifier' in the MetaData, the 3 satellites included are:

56: Meteosat 9
57: Meteosat 10
174: Himawari-9

Per this resource:
https://www.nco.ncep.noaa.gov/sib/jeff/CodeFlag_0_STDv25_LOC7.html

There are some obs here, but none of them appear to be over our desired CONUS domain.

[HuiLiu-NOAA]

Thanks @keenaneure for the info.

It looks like the NCEP prepbufr files exclude the GOES AMVs (as shown in the ioda_satwnds.nc), although the original bufr files do have the GOES AMVs (as shown above).

[HuiLiu-NOAA]

It turns out that the global analysis and RRFS use the satellite winds from the satwnd bufr files, rather than prepbufr files. GOES AMVs are available in CONUS domain as showed above.
I am developing a python tool for the bufr2ioda adding some additional needed variables, which would be ready for test soon.

[keenaneure]

Which .tar file within the HPSS directory did you use? I see quite a few when I check.

[HuiLiu-NOAA]

Which .tar file within the HPSS directory did you use? I see quite a few when I check.

com_obsproc_v1.2_rap.2024052700-05.obsproc_bufr.tar