sfcship (q, tsen, p) validation and QC flowcharts

[azadeh-gh]

Review current use of sfcship_p, sfcship_tsen, and sfcship_q and add a test YAMLs to parm/atm/obs/testing in GDASApp to reproduce the GSI (HofX and QC)

[azadeh-gh]

sfcship_tsen Yaml

obs space:
name: sfcship
obsdatain:
engine:
type: H5File
obsfile: !ENV sfcship_obs_${CDATE}.nc4
obsdataout:
engine:
type: H5File
obsfile: !ENV sfcship_diag_${CDATE}.nc4
overwrite: true
simulated variables: [stationPressure, airTemperature, specificHumidity]
geovals:
filename: !ENV sfcship_geoval_${CDATE}.nc4

obs operator:
name: Composite
components:
- name: VertInterp
variables:
- name: airTemperature
- name: specificHumidity
vertical coordinate: air_pressure
observation vertical coordinate: pressure
- name: SfcPCorrected
variables:
- name: stationPressure
da_psfc_scheme: GSI
geovar_sfc_geomz: surface_altitude
geovar_geomz: geopotential_height
linear obs operator:
name: Identity

obs post filters:

1. Observation range sanity check

• filter: Bounds Check
  filter variables:
  • name: airTemperature
    minvalue: 195
    maxvalue: 327
    action:
    name: reject

2. Reject all ObsType 183

• filter: RejectList
  where:
  • variable:
    name: ObsType/airTemperature
    is_in: 183

3. Reject data based on PreQC

• filter: PreQC
  minvalue: 0
  maxvalue: 15
  action:
  name: reject

4. Inflate obs error based on obs type

• filter: Perform Action
  filter variables:
  • name: stationPressure
    where:
  • variable: PreQC/airTemperature
    is_in: 3, 7
    action:
    name: inflate error
    inflation factor: 1.2

5. Calculate obs error inflation factors for duplicated observations at the same location

• filter: Variable Assignment
  assignments:
  • name: ObsErrorFactorDuplicateCheck/airTemperature
    type: float
    function:
    name: ObsFunction/ObsErrorFactorDuplicateCheck
    options:
    use_air_pressure: true
    variable: airTemperature

6. error inflation based on vertical spacing (subroutine errormod in qcmod.f90) for airTemperature and specific humidity

• filter: Perform Action
  filter variables:
  • name: airTemperature
    action:
    name: inflate error
    inflation variable:
    name: ObsFunction/ObsErrorFactorConventional
    options:
    pressure: ObsValue/stationPressure
    test QCflag: PreQC
    test QCthreshold: 4
    inflate variables: [airTemperature]

7. gross check based on QM

• filter: Background Check
  filter variables:

  • name: airTemperature
    absolute threshold: 4.9
    where:
  • variable:
    name: PreQC/airTemperature
    is_in: 3

• filter: Background Check
  filter variables:

  • name: airTemperature
    absolute threshold: 7.0
    where:
  • variable:
    name: PreQC/airTemperature
    is_not_in: 3

8. Inflate obs error based on duplicate check

• filter: Perform Action
  filter variables:
  • name: airTemperature
    action:
    name: inflate error
    inflation variable:
    name: ObsErrorFactorDuplicateCheck/airTemperature

[azadeh-gh]

Results from run_ufo_hofx_test.sh

[azadeh-gh]

QC count check

[azadeh-gh]

[azadeh-gh]

[azadeh-gh]

[azadeh-gh]

[azadeh-gh]

[azadeh-gh]

[azadeh-gh]

To check the HofX values. I selected a location that the JEDI-GSI_HofX was non zero(-0.000335):
hofx=0.008216
obsValue: 0.011418
GSIHofX= 0.0085511
Pressure: 101660
lat=-22.6
lon=11.4
From the geoval file this value is between 101733.3 and 101473.9 air_pressure value
Linear interpolation between two points is as follows:
Given two points (x1, y1) and (x2, y2), the value of y at some point x is given by the equation:
y = y1 + ((y2 - y1) / (x2 - x1)) * (x - x1)

y is specific humidity and x is pressure
y1 = 0.008551092
x1 = 101733.3
y2 = 0.007364262
x2 = 101473.9
x = 101660

y = y1 + ((y2 - y1) / (x2 - x1)) * (x - x1)

y= 0.008215723306861987

So base on this value the JEDI HofX 0.008216 is more accurate

[azadeh-gh]

Checking the values for sfcship_tsen, showed that the pressure value is always higher than air_pressure value in the geoval file, for each location. So the values of JEDI and GSI HofX are as same as the first air_temperature value in the geoval file.
Example:
hofx=298.6753
obsValue:300.15
GSIHofX= 298.6753
Pressure: 101360
lat=-10.3
lon=142.2
air_pressure (geoval)=101258
air_temperature (geoval)=298.6753

[ADCollard]

@azadeh-gh Thanks for doing this! So JEDI seems to be closer.

Just for completeness, I did the same calculation for Ln(p). The result is very close:

 y = y1 + ((y2 - y1) / (np.log(x2) - np.log(x1))) * (np.log(x) - np.log(x1))

In [14]: y
Out[14]: 0.008216030328678606

The GSI value just seems to be y1. I guess that is what we were finding on Thursday?

[azadeh-gh]

Correct, on Thursday we found that if we use surface_pressure for sfcship_q (that is the first value of air_pressure_level for each location in geoval file) the JEDI HofX will be equal to GSI HofX. But using air_pressure the HofX values will be different.
Now we know that using air_pressure from geoval the JEDI values are linear interpolation and GSI are y1.
We did not see this difference in sfcship_tsen because both GSI and JEDI HofX are equal to y1 (the first air_temperature value for each location in geoval file).

[azadeh-gh]

SFCSHIP

[azadeh-gh]

[azadeh-gh]

[CoryMartin-NOAA]

100,000,000 Kelvin is quite the error difference! Did this ship make it to the center of the sun?

[ADCollard]

@azadeh-gh What are we looking at here? Have you modified the YAMLs to agree with the GSI? Did you produce a new flowchart?

[azadeh-gh]

Yes, above plots are related to the modified YAMLs, base on Emily's QCflowcharts for sfcship_ps

[ADCollard]

@azadeh-gh So the usual background check filter does not work in this case?

[azadeh-gh]

[CoryMartin-NOAA]

How are the obs error differences that large for temperature?

[ADCollard]

@azadeh-gh I don't see where you have done the error inflation

[azadeh-gh]

@CoryMartin-NOAA Because the YAML needs error inflation filter for airTemperature. The values of Effective Error are still 2.5 (GsiInitialObsError)

I added obserrorFactorConventional that is errormod subroutine in qcmod.f90
- filter: BlackList
filter variables:
- name: airTemperature
action:
name: inflate error
inflation variable:
name: ObsFunction/ObsErrorFactorConventional
options:
test QCflag: PreQC
inflate variables: [airTemperature]

But get below error:
Test "ufo/ObsFilters/testFilters" failed with unhandled exception: Reason: An exception occurred inside ioda while opening a variable.
name: ObsValue/pressure
source_column: 0
source_filename: /scratch1/NCEPDEV/da/Azadeh.Gholoubi/GDASApp/ioda/src/engines/ioda/src/ioda/Has_Variables.cpp
source_function: ioda::Variable ioda::detail::Has_Variables_Base::open(const std::__cxx11::basic_string<char, std::char_traits, std::allocator> &) const
source_line: 108

MetaData/pressure and ObsValue/stationPressure exist in the observation file, not ObsValue/pressure

[azadeh-gh]

obs space:
name: sfcship
obsdatain:
engine:
type: H5File
obsfile: !ENV sfcship_obs_${CDATE}.nc4
obsdataout:
engine:
type: H5File
obsfile: !ENV sfcship_diag_${CDATE}.nc4
overwrite: true
simulated variables: [stationPressure, airTemperature, specificHumidity]
geovals:
filename: !ENV sfcship_geoval_${CDATE}.nc4

obs operator:
name: Composite
components:
- name: VertInterp
variables:
- name: airTemperature
- name: specificHumidity
vertical coordinate: air_pressure
observation vertical coordinate: pressure
- name: SfcPCorrected
variables:
- name: stationPressure
da_psfc_scheme: GSI
geovar_sfc_geomz: surface_altitude
geovar_geomz: geopotential_height
linear obs operator:
name: Identity

obs post filters:

Observation range sanity check

• filter: Bounds Check
  filter variables:
  • name: stationPressure
    minvalue: 37499.0
    maxvalue: 106999.0
    action:
    name: reject

Reject all ObsType 183

• filter: RejectList
  where:
  • variable:
    name: ObsType/stationPressure
    is_in: 183

Reject surface pressure below 500 hPa

• filter: Bounds Check
  filter variables:
  • name: stationPressure
    minvalue: 50000.00
    action:
    name: reject

Reject data based on PreQC

• filter: PreQC
  minvalue: 0
  maxvalue: 15
  action:
  name: reject

Inflate obs error based on obs type

• filter: Perform Action
  filter variables:
  • name: stationPressure
    where:
  • variable: PreQC/stationPressure
    is_in: 3, 7
    action:
    name: inflate error
    inflation factor: 1.2

Calculate obs error inflation factors for duplicated observations at the same location

• filter: Variable Assignment
  assignments:
  • name: DerivedMetaData/ObsErrorFactorDuplicateCheck
    type: float
    function:
    name: ObsFunction/ObsErrorFactorDuplicateCheck
    options:
    use_air_pressure: false
    variable: stationPressure

Reduce effective observation error based on obs type and subtype

In this case: reduce effective obs error for buoy

• filter: Perform Action
  filter variables:
  • name: stationPressure
    where:
  • variable:
    name: ObsType/stationPressure
    is_in: 180
  • variable:
    name: ObsSubType/stationPressure
    is_in: 0
    action:
    name: inflate error
    inflation factor: 0.7

Reduce original observation error based on obs type and subtype

In this case: reduce original obs error for buoy

• filter: Variable Assignment
  where:
  • variable:
    name: ObsType/stationPressure
    is_in: 180
  • variable:
    name: ObsSubType/stationPressure
    is_in: 0
    assignments:
  • name: ObsError/stationPressure
    type: float
    function:
    name: ObsFunction/Arithmetic
    options:
    variables:
    - name: ObsError/stationPressure
    coefs: [0.7]

Inflate surface pressure observation based on discrepancies between

model and observations due to terrian

• filter: Perform Action
  filter variables:
  • name: stationPressure
    action:
    name: inflate error
    inflation variable:
    name: ObsFunction/ObsErrorFactorSfcPressure
    options:
    geovar_sfc_geomz: surface_altitude

Gross check based on residuals

• filter: Background Check
  filter variables:

  • name: stationPressure
    absolute threshold: 3500
    where:
  • variable:
    name: PreQC/stationPressure
    is_in: 3

• filter: Background Check
  filter variables:

  • name: stationPressure
    absolute threshold: 5000
    where:
  • variable:
    name: PreQC/stationPressure
    is_not_in: 3

Inflate obs error based on duplicate check

• filter: Perform Action
  filter variables:
  • name: stationPressure
    action:
    name: inflate error
    inflation variable:
    name: DerivedMetaData/ObsErrorFactorDuplicateCheck

[azadeh-gh]

@ADCollard above is my sfcship.yaml , it doesn't have any error inflation for airTemperature.

[emilyhcliu]

Here is the updated QC flowchart for PSOB SFCSHIP

[azadeh-gh]

[azadeh-gh]

QC flowcharts

[emilyhcliu]

@CoryMartin-NOAA @ADCollard @azadeh-gh

• I have the station pressure QC work done for obs types 180, 181, 183, 187, and 120. (Combination of SFC, SFCSHIP and ADPUPA (POB at the surface only from ADPUPA).
• The results can 100% replicate GSI results in terms of data count and final observation errors
• I will document the results here (still work in progress)

[azadeh-gh]

Commenting out the duplicate check in GSI changed the sfcship_tsen JEDI-GSI obs error from A to B

A: With duplicate check in GSI

B: Without duplicate check in GSI

Also evaluated the updated sfcship.yaml from GMAO.