PolSIR BT

[chengdang]

@BenjaminTJohnson
Here are some hofx created using the PolSIR coefficients:

litt

[chengdang]

The above BT were computed with an altered ctest: ufo_test_tier1_test_ufo_opr_atms_crtm.
In which the atms spectral and tau coefficients are replaced with the polsir coefficients (little endian).

Here is the yaml file:

time window:
  begin: 2018-04-14T21:00:00Z
  end: 2018-04-15T03:00:00Z

observations:
- obs operator:
    name: CRTM
    Absorbers: [H2O,O3,CO2]
    Clouds: [Water, Ice]
    Cloud_Fraction: 1.0
    SurfaceWindGeoVars: uv
    linear obs operator:
      Absorbers: [H2O,O3,CO2]
      Clouds: [Water, Ice]
    obs options:
      Sensor_ID: atms_npp
      EndianType: little_endian
      CoefficientPath: Data/
      NC_CoefficientPath: Data/
      CloudCoeff_File: CloudCoeff.nc4
      CloudCoeff_Format: netCDF
      Cloud_Model: CRTM
      AerosolCoeff_File: AerosolCoeff.nc4
      AerosolCoeff_Format: netCDF
      Aerosol_Model: CRTM
  obs space:
    name: Radiance
    obsdatain:
      engine:
        type: H5File
        obsfile: Data/ufo/testinput_tier_1/atms_npp_obs_2018041500_m_qc.nc4
    obsdataout:
      engine:
        type: H5File
        obsfile: Data/polsir_obs_2018041500_m_qc_polsir_crtm_out.nc4
    simulated variables: [brightnessTemperature]
    channels: 1-6
  geovals:
    filename: Data/ufo/testinput_tier_1/atms_npp_geoval_2018041500_m_qc.nc4
  vector ref: GsiHofX
  tolerance: 1.e-7
  linear obs operator test:
    iterations TL: 5
    tolerance TL: 2.0e-5
    tolerance AD: 1.0e-11

[BenjaminTJohnson]

Thanks!

the colorbars don't seem to be showing any numbers for me -- it might be that I'm using dark mode, and the numbers are black on a transparent background (yes, that's it -- my problem, not yours).

Is it relatively easy to create a histogram of TBs per channel?

[chengdang]

Yes it is. Let me update soon.

[chengdang]

Here are the histogram plot:

[BenjaminTJohnson]

Hmm. Visually channels 3 and 4 look identical, and channels 5 and 6 look identical, and they should be different even though the frequency is the same, the polarization should be different -- so over ocean points should be different.

[chengdang]

Yes, you are right, hofx of these paired channels are identical. Hmmm

[BenjaminTJohnson]

I can modify the polarization, would you be able to easily re-run the experiment using a new coefficient?

[chengdang]

Yes. I can do that easily. Let me know when it's ready.

[BenjaminTJohnson]

@chengdang
https://bin.ssec.wisc.edu/pub/s4/CRTM/file/polsir_p01_9_10.SpcCoeff.nc

[chengdang]

@BenjaminTJohnson could you convert this into LE binary file? The UFO pipeline is not ready for netCDF versions of spc and tau coefficient yet (no TauCoeff_Format and SpcCoeff_Format yet).

[BenjaminTJohnson]

@chengdang https://bin.ssec.wisc.edu/pub/s4/CRTM/file/polsir_p01_9_10.SpcCoeff.bin you can use the existing polsir taucoeff file.

[chengdang]

@BenjaminTJohnson Thank you! There are some minor difference between these paired channels, but still almost identical:

[BenjaminTJohnson]

Thanks -- it's entirely possible that the surface is obscured by water vapor, and we won't see polarization effects from the surface. The other big thing that's missing is O3 contribution.
https://amt.copernicus.org/articles/13/53/2020/ (fig 1)

We could determine whether surface is obscured by plotting weighting functions / jacobians wrt the surface.