UFO Evaluation for Conventional Station Pressure --- Re-evaluation

[emilyhcliu]

This issue is to update the station pressure with the following modifications:
(1) Add error assignment from extracting errors from GSI error table (need update from UFO PR #3085)
(2) Update options for error inflation from pressure check (due to updates from GMAO)

To closely replicate the error assignment from GSI error table. The ObsFunctionModelStepwiseLinear needs to add the capability of rounding the output of this function to the nearest integer. A related UFO PR #3085 has submitted in UFO repository.

Notes for test data set:

• The error inflation for duplicate observation is turned on for the UFO evaluation.
• The input obs file is a combined data types from 180 183 (SFCSHIP), 181, 187 (SFC), and 120 (Radiosonde Ps) for stationPressure.
• The GSI input obs error read from the prepbufr file is stored in the obs file for reference and validation. This can be used to validate the obs error values extracted from the errtable using ObsErrorModelStepwiseLinear function.

Observer Flow Chart for Station Pressure

Details are documented here.

[emilyhcliu]

Validation of Input observation error:

Initial (Original) Obs Error Assignment between GSI and JEDI:

The initial observation error values are extracted from GSI error table. The error values are a function of pressure. In UFO, an ObsErrorModelStepwiseLinear function is used to extract the obs errors from the error table. However, to following GSI implementation, the function needs to add the capability of rounding the interpolated error values extracted from the stepwise linear function to the nearest integer. Please see UFO PR #3085 for details.

The result replicates GSI obs errors 100% except one point. This point is related to ObsType 181. The GSI value is 115, and the UFO value is 114. The difference is 1 Pa. The difference is caused by the precision difference between GSI (double) and UFO (single).

[emilyhcliu]

Validation with QC

HofX Comparison

QC Flag Comparison (100% replication)

Final Obs Error Difference vs Pressure (~100% replication; one point difference due to precision difference between GSI and JEDI)

There is one obs that does not replicate GSI result. This the same point where the initial obs error differs by 1 Pa (114 vs 115).

The observation count passed QC comparison between GSI and JEDI is identical!