test NASA AIRS product

[HaixiaLiu-NOAA]

NOAA AIRS product has encountered issues since late 2020 and has been discommissioned since Feb 2021. EMC looks for a new way to get AIRS products, such as the NASA AIRS product. The NASA AIRS bufr contains AMSUA_AQUA data and has 324-channel subset for airs instrument. ObsProc team generated sample bufr dumps for the date 20210520 4 cycles. Work needs to be done to evaluate the NASA AIRS product.

[HaixiaLiu-NOAA]

preliminary evaluation of NASA AIRS product is here using the 4 cycles dump for the date 20210520.

https://docs.google.com/presentation/d/15I9SKHWMeXkaJnrl3Nd-MF_-GNwt6_xB/edit#slide=id.p11

[HaixiaLiu-NOAA]

NASA AIRS bufr dumps are available for the period of 20210728 to 20210810 and I have asked ObsProc team to extend the data covering a longer period.

Cathy pointed out that there is not a suitable time where a real time parallel covers the period of 20210728-20210810, so have to cold start the control run. Cathy helped me generate IC files (chgresed from operational run to low-reso and ensemble members are identical to each other). So a control run has been started on Hera and will be cycled for a full week to build the appropriate ensemble spread.

[HaixiaLiu-NOAA]

Obtained the gdas and gfs bufr dumps from ObsProd team for the period of 20210728-20210829. A low-reso control run has been started on Hera one week ago. The IC files were chgresed from the opr v16 to low-reso @ 2021072000. The ensemble members are identical to each other and chgresed from a deterministic background. This control run will be cycled a week to build ensemble spread.

[HaixiaLiu-NOAA]

Ensemble spread plots from the control run:

The parallel experiment with NASA AIRS data was warm started from the control run @ 2021072812 with zeroed out bias correction for airs_aqua and amsua_aqua data.

[HaixiaLiu-NOAA]

checked the 1st analysis @ 2021072812, gsistat shows expected differences from the control run. The AIRS product was successfully read in and processed. There are no AIRS data assimilated at the 1st analysis cycle due to the zeroed out bias correction coeffs for both airs_aqua and amsua_aqua except for the channel 14 amsua_aqua due to no bias correction applied to this channel. The 2nd analysis cycle was checked as well. Seeing AIRS data being assimilated at the 2nd cycle. Bias correction has been applied at this cycle.

A check on the obs for a LW window channel:

And a check on a SW window channel OmF (data passing QC)

Will continue to monitor the control and experiment runs on Hera.

[HaixiaLiu-NOAA]

preliminary plots from PyGSI using gsistat files, serving as a sanity check.

[HaixiaLiu-NOAA]

The parallel test encountered a segmentation fault at 2021082400 cycle. Investigation revealed that the AIRS BUFR data contains incorrect channel numbers.

++++++  SCBTSEQN  REPLICATION #   239  ++++++

005042 CHNM -517867749.0 NUMERIC CHANNEL NUMBER 025076 LOGRCW -21.15857808 LOG/M LOG-10 OF (TEMPERATURE-RADIANCE CENTRAL WAVENUMB 033032 ACQF MISSING FLAG TABLE CHANNEL QUALITY FLAGS FOR ATOVS 012163 TMBR -14975087.83 KELVIN BRIGHTNESS TEMPERATURE (HIGH ACCURACY) ++++++ SCBTSEQN REPLICATION # 240 ++++++ 005042 CHNM -1638377645.0 NUMERIC CHANNEL NUMBER 025076 LOGRCW 3.61657062 LOG/M LOG-10 OF (TEMPERATURE-RADIANCE CENTRAL WAVENUMB 033032 ACQF 1922994.0 FLAG TABLE(MANY BITS ON) CHANNEL QUALITY FLAGS FOR ATOVS 012163 TMBR 0.36 KELVIN BRIGHTNESS TEMPERATURE (HIGH ACCURACY) ++++++ SCBTSEQN REPLICATION # 241 ++++++ 005042 CHNM -488875011.0 NUMERIC CHANNEL NUMBER 025076 LOGRCW 4.72782366 LOG/M LOG-10 OF (TEMPERATURE-RADIANCE CENTRAL WAVENUMB 033032 ACQF 2006478.0 FLAG TABLE(MANY BITS ON) CHANNEL QUALITY FLAGS FOR ATOVS 012163 TMBR 124.37 KELVIN BRIGHTNESS TEMPERATURE (HIGH ACCURACY)

A similar issue happened to the 2021082406 cycle AIRS dump. Asked ObsProc to regenerate the BUFR dumps for these two cycles. The segfault were gone. The parallel has been finished.

[HaixiaLiu-NOAA]

[HaixiaLiu-NOAA]

The NASA AIRS data generally reduced the wind and humidity biases in the upper troposphere globally shown in the above plots generated from gsistat files.

Here is the VSDB