AMSU-A Obs Validation

[ADCollard]

This issue will be used to track progress on the development and validation of radiances in UFO for GDAS in four stages:

1. Validation with GeoVaLs from GSI:

• [x] HofX
  • [x] GSI NC diagnostic converter needs to update to output GSI HofX directly instead of output HofX based on O-(O-HofX). This change is necessary to avoid producing Zero HofX at locations where the observations are missing
  • [x] Regenerate GeoVaLs and Obs
  • [x] Validation results: 100% replication of GSI for all platforms. New input files are available on ORION:

          Date: 2021080100
          Resolution: C768
          GeoDir=/work2/noaa/da/eliu/ufoeval/GSIobserver/2021080100/ioda/geovals/
          ObsDir=/work2/noaa/da/eliu/ufoeval/GSIobserver/2021080100/ioda/obs/
          BCDir=/work2/noaa/da/cmartin/UFO_eval/data/gsi_geovals_l127/nofgat_aug2021/20221027/bc/
          FixDir=/work2/noaa/da/cmartin/GDASApp/fix/

• [x] Update/improve QC filters (GSI QC steps in GSI setuprad.f90 and qcmod.f90 (qc_amsua subroutine)

• [x] The @ObsError is set to missing in the input obs file
• [x] Code change is required for ObsErrorFactorSituDependMW to avoid getting the original obs error from the input obs file
• [x] YAML update
  • Improve the use of generic filters and obs functions in the YAML
  • Add filters to output diagnostic variables in YAML
  • Remove the need to repeatedly re-compute the same obs functions used in various filters
    • [x] QC count and obs error checks using-GSI produced GeoVaLs
• [x] Code change is required GSI NC diagnostic converter to output the correct varinv (output varinv instead of varinv0)
• [x] Validation Results: Obs Error and HofX
  • [x] amsua_n19 --- 100% replication of GSI
  • [x] amsua_n18 --- 100% replication of GSI
  • [x] amsua_n15 --- 100% replication of GSI
  • [x] amsua_metop-a --- 100% replication of GSI
  • [x] amsua_metop-b --- 100% replication of GSI
  • [x] amsua_metop-c --- 100 % replication of GSI
  • [x] amsua_aqua --- data not available in the operational data stream

2. UFO development:

• [ ] Separate HydrometeorCheckAMSUA into three separate checks for clarity --- work in progress The HydrometeorCheckAMSUA will be separated into three single checks: RainCheck, CloudCheck, and SurfaceEmissivityCheck, and these checks will be filtered over land and ocean separately.
• [ ] Modify existing Gaussian thinning filter to reproduce GSI result --- work in progress
• [ ] Check with JEDI team about keeping the thinned data in the obs space
• [ ] Add thinning filter and a few more QC steps (from GSI read_bufrtovs.f90 routine) in YAML
• [ ] Validation with GSI result

3. IODA-converter development:

• [x] Prepare AMSU-A YAML for bufr converter
• [ ] Work with Cory to add the capability of converting AMSU-A bufr to IODA in GDASApp workflow

4. HofX computation in FV3-JEDI with GDASApp workflow:

• [ ] Test workflow (plumbing)
• [ ] Check statistics
• [ ] Summarize and explain differences between GSI and UFO

[ADCollard]

This also references #38

[ADCollard]

It was noted that there are observations where GSIHofX is zero. See plot below. This is only occuring where the observation value is set to missing, so appears to be a quirk in the GSI. I am still trying to track down exactly where the logic is that is doing this, but this should not adversely affect JEDI validation.

[emilyhcliu]

Here are a few more examples of zero HofX from GSI caused by missing observations: This is fixed by output GSI HofX instead of getting HofX from Obs-(Obs-HofX)

AMSU-A N19 channel-12	channel-13

AMSU-A MetOp-A channel-7	channel-8

[emilyhcliu]

Example HofX (No QC) Validation: GSI vs. JEDI UFO

AMSU-A MetOp-A Channel-1	Channel-2	Channel-3

Channel-4	Channel-5	Channel-6

Channel-7	Channel-8	Channel-9

Channel-10	Channel-11	Channel-12

Channel-13	Channel-14	Channel-15

[emilyhcliu]

Initial (Original) / Final Observation Error Comparison: GSI vs. JEDI UFO

Result of QC filtering check (using AMSU-A N19 as an example) The following figures show the GSI final observation error on the left and UFO on the right The data count passed QC is listed above the color bar

Original Observation Error Channel-2 GSI	UFO

Observation Error after a series of data filtering and error inflation processes: Channel-2 GSI	UFO

Channel-1 GSI	UFO

Channel-3 GSI	UFO

Channel-4 GSI	UFO

Channel-5 GSI	UFO

Channel-6 GSI	UFO

Channel-7 GSI	UFO

Channel-8 GSI	UFO

Channel-9 GSI	UFO

Channel-10 GSI	UFO

Channel-11 GSI	UFO

Channel-12 GSI	UFO

Channel-13 GSI	UFO

Channel-14 GSI	UFO

Channel-15 GSI	UFO

[emilyhcliu]

HofX Comparison between GSI vs. JEDI UFO: No QC (Black) and QC (Red)

AMSU-N19 Channel 1	Channel 2	Channel 3

Channel 4	Channel 5	Channel 6

Channel 7	Channel 8	Channel 9

Channel 10	Channel 11	Channel 12

Channel 13	Channel 14	Channel 15