Evaluate/ Validate Pressure and Temperature QC Procedures in Sonde YAML File and GSI

[PraveenKumar-NOAA]

A Sonde YAML file exists in GDASApp copied from UFO/ewok/jedi-gdas on June 28, 2022.

Before diving into replicating the PrepBUFR preprocessing for Sonde pressure and temperature observations, we should first check that the QC procedures implemented in UFO match those being performed in GSI on PrepBUFR observations.

• Run GDASApp for the first week of August 2021 for Sonde observations
• Compare EffectiveQC against GsiEffectiveQC for pressure and temperature
• If there are differences, look at setupq.f90 or another related file in the GSI source code
• Make any modifications to YAML as necessary

This issue will be used to track progress, including figures, statistics, and paths to files.

[PraveenKumar-NOAA]

Using the sondes_ps.yaml, that Cory wrote, I ran GDASApp for the following observations: surface pressure, air temperature, specific humidity, and wind. Surface pressure plots are attached below that need to be discussed first:

This plot looks pretty good.

[PraveenKumar-NOAA]

This plot needs further discussion, as points are mostly scattered.

[PraveenKumar-NOAA]

Both plots look very similar.

[PraveenKumar-NOAA]

Histogram plots: one is centered at zero, but other one not.

[CoryMartin-NOAA]

I believe these differences can likely be resolved by https://github.com/JCSDA-internal/ufo/pull/2207

[PraveenKumar-NOAA]

Air_temperature plots:

Histograms:

[CoryMartin-NOAA]

@PraveenKumar-NOAA something looks wrong here... is this a plotting issue or a data issue?

[PraveenKumar-NOAA]

@CoryMartin-NOAA most probably a data issue, I will look into the details.

[CoryMartin-NOAA]

@PraveenKumar-NOAA it looks like this is 'pressure' and not 'temperature', note that GSI is ~200-300 and JEDI is ~30000-100000 (K vs Pa)

[PraveenKumar-NOAA]

I believe these differences can likely be resolved by JCSDA-internal/ufo#2207

Created new surface pressure plots after the GSI Ps fix, which look similar to those created without this fix.

[PraveenKumar-NOAA]

@CoryMartin-NOAA @ADCollard @emilyhcliu there was an error in yaml for the air_temperarure. I fixed and regenerated air_temperature plots for H(x), error, histogram, etc. The plots look better compared to the old ones.

[CoryMartin-NOAA]

Thanks @PraveenKumar-NOAA these look great. Can you do virtual_temperature next? The way the sondes input files are currently setup, air_temperature obs are only in the ~stratosphere (where there are no humidity obs) and virtual_temperature obs are in the troposphere.

[PraveenKumar-NOAA]

Thanks @PraveenKumar-NOAA these look great. Can you do virtual_temperature next? The way the sondes input files are currently setup, air_temperature obs are only in the ~stratosphere (where there are no humidity obs) and virtual_temperature obs are in the troposphere.

@CoryMartin-NOAA sure, thanks for the output.

[PraveenKumar-NOAA]

@CoryMartin-NOAA I ran GDASApp and generated virtual-temperature plots for the sonde data. I nned to discuss these plots with you.

Histogram plots:

[CoryMartin-NOAA]

@PraveenKumar-NOAA Yikes, something looks bad here...

[PraveenKumar-NOAA]

Regenerated surface_pressure plots using the updated GDASApp and performed analysis on the output diag file:

Comparison of GSI and UFO H(x):

• The differences between GSI and UFO H(x) is very small.
• The regression is on the 1:1 line, where slope is 1 and the intercept is 0.93 for the observations that passed the QCs.

Number of observations passing/failing quality control

• The totals counts, 471, are the same between GsiEffectiveQC=0 and EffectiveQC=0.
• The totals are not the same between GsiEffectiveQC=1 and EffectiveQC!=0. The number of counts of GsiEffectiveQC=1 are 7 and for EffectiveQC!=0 are 77621. I don't know the reason of this discrepancy, possibly multiple values at the same location.

Observation error assignment

• Please look at the attached histogram plots. The differences between GsiFinalObsError and EffectiveError is small, but need to generate a plot for this.

[PraveenKumar-NOAA]

Updated results for the air_temperature obs. I only used PreQC for the air_temperature in my YAML and the results obtained are discussed below:

Comparison of GSI and UFO H(x):

• The differences between GSI and UFO H(x) are very small.
• The regression is on the 1:1 line, where the slope is 1 and the intercept is 0.0 for the observations that passed the QCs.

Number of observations passing/failing quality control

The total counts which passed QC in GSI are 15103 (GsiEffectiveQC=0), while for JEDI are 15108 (EffectiveQC=0). The totals are not the same between GsiEffectiveQC=1 and EffectiveQC!=0. The number of counts for GsiEffectiveQC=1 is 113 and for EffectiveQC!=0 is 62984.

Observation error assignment

The differences between the GsiFinalObsError and EffectiveError is also small, please look at the histogram plots below.

[PraveenKumar-NOAA]

The number of observations that passed QCs is listed in the following document, https://docs.google.com/document/d/1_w6P-cJkxyYzTvLNMm6kCBWj2ZATk0PMsXOdKW29z8s/edit for three different YAML:

• YAML with no QC,
• YAML with QCs copied from an existing GDASapp sonde YAML, copied from UFO/ewok/jedi-gdas
• New (gsi) YAML