Update GPSRO obs error using the new hybrid 3-cornered hat error model

[XuanliLi-NOAA]

This update includes the following for GPSRO:

1. Replacing the current GPSRO observation error with the new hybrid three-cornered hat error model.
2. Allowing Grace-FO (SAID=803 & 804) into GSI.
3. Turning off the previous MetOp 8-km check.
4. Adding more QC categories.
5. Adding profile quality flag for Grace-FO and Sentinel-6.

[XuanliLi-NOAA]

The impact of the new hybrid obs error on GFS v16.3.16 is summarized here. The metplus plots are available here. The overall impact of the new obs error is significantly positive across most fields at most levels.

MetOp data below 8-km has been testes alongside the new obs error model. The impact of enabling MetOp data under 8-km is minimal due to the small data volume remaining after all QC filters (super refraction, statistic QC, etc).

The impact of the GRACE-FollowOn (GRACE-FO) data is summarized here. The metplus plots are available here. Overall, the impact of the GRACE-FO data is neutral due to its small data volume.

Sentinel-6 data has been added in read_obs.F90 and read_gps.f90.

global_convinfo.txt has been updated to add GRACE-FO and activate Sentinel-6.

QC flag: previously qc=4 contains several QCs. The new QC flags are listed below:
qc = 1: beyond model top or bottom
qc = 2: gpstop (55 km or 45 km for commercial data)
qc = 3: gross check
qc = 4: super refraction
qc = 5: background bending angle > 0.05 rad
qc = 6: statistic (cutoff) QC
qc = 7: MetOp data < 8 km
qc = 8: ddnj < 0

GSI single cycle tests have been conducted and confirmed that all changes are integrated and functioning as expected.

[XuanliLi-NOAA]

Results for the regression tests:
ctest -j 6

Test project /scratch1/NCEPDEV/da/Xuanli.Li/git/gsi_obserr/GSI/build
Start 1: global_4denvar
Start 6: global_enkf
Start 2: rtma
Start 3: rrfs_3denvar_rdasens
Start 4: hafs_4denvar_glbens
Start 5: hafs_3denvar_hybens
1/6 Test #3: rrfs_3denvar_rdasens ............. Passed 797.24 sec
2/6 Test #6: global_enkf ...................... Passed 1952.58 sec
3/6 Test #4: hafs_4denvar_glbens .............. Passed 2128.77 sec
4/6 Test #2: rtma ............................. Passed 2176.11 sec
5/6 Test #5: hafs_3denvar_hybens .............. Passed 2189.85 sec
6/6 Test #1: global_4denvar ...................***Failed 2691.87 sec

83% tests passed, 1 tests failed out of 6

Total Test time (real) = 2691.90 sec

The following tests FAILED:
1 - global_4denvar (Failed)

From the log file: /scratch1/NCEPDEV/da/Xuanli.Li/git/gsi_obserr/GSI/build/Testing/Temporary/LastTest.log:
1.
grep -c cost,grad,step penalty.global_4denvar_loproc_updat-global_4denvar_loproc_contrl.txt
[[ 24 = 0 ]]
echo 'The results (penalty) between the two runs are nonreproducible,'
echo 'thus the regression test has Failed on cost for global_4denvar_loproc_updat and global_4denvar_loproc_contrl analyses.'

ncdump siginc.nc.global_4denvar_loproc_updat
ncdump siginc.nc.global_4denvar_loproc_contrl
diff -s siginc.nc.global_4denvar_loproc_updat.out siginc.nc.global_4denvar_loproc_contrl.out
echo 'The results between the two runs (global_4denvar_loproc_updat and global_4denvar_loproc_contrl) are not reproducible'
echo 'Thus, the case has Failed siganl of the regression tests.'

These are expected because the updated RO obs error results in differences in the penalty term and the increment.