ext_sonde functionality mismatch (GSI vs JEDI)

[delippi]

Description

For RRFSv1, GSI currently uses ext_sonde=.true. which applies synthetic level extensions to raob soundings. JEDI does not implement an equivalent capability (that we know of), so the two DA systems ingest raob profiles differently. This creates inconsistencies when comparing GSI and JEDI diagnostics particularly in observation counts.

For the current validation work, we should set ext_sonde = .false. in GSI so both systems are consistent.

Long-term question(s):

• Should RRFSv1 GSI be using ext_sonde=.true. or should it be disabled?
• Should JEDI implement an ext_sonde-like capability for raob extension, or should both systems use native levels only?

[delippi]

I tested the impact of ext_sonde during a 01z cycle (2024050601). The result is for airTemperature only.

GSI nobs with ext_sonde=.true. + "sonde time-window override": 7253
GSI nobs with ext_sonde=.false. + "sonde time-window override": 6522
JEDI nobs: 4121

For details on the "sonde time-window override", see:
#496

These two options (ext_sonde=.true. + "sonde time-window override") causes about 75% more observations to be used for this particular case. Over 700 of these observations were not measured by an instrument. They are synthetic, interpolated values that GSI treats as if they were real observations.

[ShunLiu-NOAA]

@delippi Thank you. We should conduct the simplified DA retro test with FV3JEDI to determine the difference between ext_sonde=.true. and ext_sonde = false. so that we can decide whether we should move this function from GSI to JEDI.