Open ADCollard opened 1 year ago
ompsnp_npp test YAML added to parm/atm/obs/testing in GDASApp to reproduce the GSI HofX, QC and final obs error. JEDI h(x) vs. GSI h(x) was identical.
JEDI h(x) - GSI h(x) were almost zero
Difference of JEDI and GSI final obs error was zero
GSI omb vs JEDI omb for both all obs and obs passed QC in JEDI were identical.
omi_aura test YAML added to parm/atm/obs/testing in GDASApp to reproduce the GSI HofX, QC and final obs error.
JEDI h(x) vs. GSI h(x) was identical.
(1) Create input data (obs and geoval) - need to include obs filtered out in the read routine (2) Update QC filters - there are some minor updates ---> Will issue PR in GDASApp to update the filters in the YAML (3) Update variable names so they follow the IODA naming convention
Here are the updated checks
HofX JEDI vs. GSI
JEDI HofX vs. Obs | GSI HofX vs. Obs |
---|---|
HofX Difference vs. Pressure Layers
ObsError JEDI vs. GSI
(no thinning in the input obs file)
JEDI HofX vs. Obs | GSI HofX vs. Obs |
---|---|
(No data thinning for input data)
JEDI HofX vs. Obs | GSI HofX vs. Obs |
---|---|
The last plot does not look right. Some observations have QC difference = 14, meaning GsiEffectiveQC= 0, but JEDI EffectiveQC =14.
This is not because the filters are wrong. This is related to how we prepared the obs file from the GSI diagnostic.
I will check and redo the obs file from GSI and run the GDASApp test again.
See more information in the following T2O issues related to end-to-end tests for ozone OMI Issue #77 OMPSTC Issue #75 OMPSNP Issue #76
GSI HofX vs Obs | |
---|---|
Here is how to interpret the plot: QC diff = 9 --- GsiEffectiveQC = 1, and JEDI EffectiveQC = 10 (Missing Value) --- observations are screened from both GSI and UFO.
QC diff = 13 --- GsiEffectiveQC = 1, and JEDI EffectiveQC = 14 (RejectList)--- observation are screened from both GSI and UFO
QC diff = 0 --- GsiEffectiveQC (GSI) and EffectiveQC (JEDI)) are zeros
The plot tells us that the QC results between GSI and JEDI are identical.
JEDI HofX vs. Obs | GSI HofX vs. Obs |
---|---|
JEDI HofX vs. Obs | GSI HofX vs. Obs |
---|---|
@ADCollard @azadeh-gh Please check the last three new comments. They are updated results for omi, ompstc, and ompsnp.
These results show that the QC flag differences (GsiEffectiveQC vs. EffectiveQC) between GSI and UFO are identical.
Review current use of OMPS and add a test YAML to
parm/atm/obs/testing
inGDASApp
to reproduce the GSI (HofX, QC and final obs error)