crisnoise and calibration uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as...
TRANSCRIPT
![Page 1: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/1.jpg)
CrIS Noise and Calibration Uncertainty
David Tobin, Joe Taylor, Lori Borg, Michelle Feltz, Dan Deslover, Bob Knuteson, Hank RevercombCIMSS/SSEC, UW-Madison
The 22nd International TOVS Study Conference (ITSC-22) Saint-Sauveur, Canada
![Page 2: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/2.jpg)
CrIS Operational Concept
RDR = Raw Data RecordSDR = Sensor Data RecordEDR = Environmental Data Record
2,200 km Swath
Downlink
EDR Algorithm
Decode Spacecraft
Data
CrIS SDR Algorithm
�50�Cross track
Scans+
Space andICT views
3x3 Array of CrIS FOVs (Each at 14-km
Diameter)
Ground Station
30 Earth FORs
SDRs
RDRs
EDRs
Co-Located ATMS SDRs
Interferograms
Calibrated / Geolocated Spectra
Global Temperature, Moisture, Pressure Profiles
SDRs
RDRs
CrIS on Suomi-NPP,built by ITT Exelis
Key Sensor FeaturesLarge 8 cm Clear ApertureThree Spectral Bands3x3 FOVs at 14 km DiameterPhotovoltaic detectors in all 3 bands4-Stage Passive Detector CoolerPlane-Mirror Interferometer With DAInternal Spectral CalibrationAmbient Internal Calibration TargetModular Construction
Band Wavelength Range Sampling No. Chan. (cm-1) (µm) (cm-1)
SWIR 2155-2550 4.64-3.92 2.5 159 MWIR 1210-1750 8.26-5.71 1.25 433 LWIR 650-1095 15.38-9.14 0.625 713
![Page 3: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/3.jpg)
20406080100120
Rad
20406080100120
Rad
20406080100120
Rad
660 680 700 720 740 760 780 800 820wavenumber
20406080100120
Rad
MonochromaticIASI L1C
(+/-2cm OPD w/ Gaussian apodization)
MonochromaticAIRS
(near Gaussian SRFs)
MonochromaticCrIS unapodized
(+/-0.8 cm OPD, SDRs)
MonochromaticCrIS with Hamming apodization(R’i = 0.23 Ri-1 + 0.54 Ri + 0.23 Ri+1 , BUFR)
Example Longwave Spectra
![Page 4: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/4.jpg)
“Observation Error” contributors:• Instrument Noise• Instrument Calibration uncertainties• Forward Model error (Fast model and underlying LBL)• Representativeness Error• Cloud contamination• Quality Control errors• …
Example hyperspectral IR Clear sky Covariancefrom Bormann et al., “Enhancing the impact of IASI observations through an updated observation-error covariance matrix”
Observation-error correlations for assimilated IASI channels:
Error standard deviations:• Close to instrument noise for upper tropospheric and stratospheric
temperature sounding channels, with weak error correlations; • Larger than the instrument noise for water-vapour channels,
combined with significant interchannel error correlations; and • Larger than the instrument noise for lower temperature sounding,
window and ozone channels, together with weaker, but still significant, interchannel error correlations.
![Page 5: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/5.jpg)
Outline
• CrIS instrument noise• Gaussian distribution• Scene independence of NEDN• FOV variability• Spectral correlation• Self-apodization correction and Hamming apodization
effects on NEDN level and spectral correlation• CrIS Calibration uncertainties
• Contributors• Warm and cold scene examples
• Next steps
![Page 6: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/6.jpg)
Calibrated ICT (onboard blackbody) spectra ensembles
~15,000 ICT view radiance spectra
NEDN = Stdev(RICT)
RICT at 825 cm-1
Gaussian
[ mW/(m2 sr. cm-1) ]
![Page 7: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/7.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 200K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 8: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/8.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 233K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 9: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/9.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 260K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 10: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/10.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 287K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 11: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/11.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 299K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 12: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/12.jpg)
NEDN vs NEDT
NOAA20 measured NEDN for 310K scene:1.0
0.1
0.01
0.001600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
wavenumber
NED
N (m
W/m
2sr
. cm
-1)
![Page 13: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/13.jpg)
NEDN converted to NEDT at various scene temperatures
(NEDN [mW/(m2 sr. cm-1)] = 0.1 LW; 0.04 MW; 0.006 SW)
NEDN converted to NEDT at 280K and at scene temperature of a typical clear sky spectrum
700 cm-1
1550 cm-1
2500 cm-1
![Page 14: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/14.jpg)
FOV variability of NEDNSNPP CrIS
• 30% variations in LW NEDN among FOVs
• In the MW, FOV7 is the large outlier, with NEDN ~3 times higher than other FOVs. (This detector also has the largest nonlinearity of the SNPP MW FOVs)
• MW and SW bands show self-apodization noise amplification, with values up to 70% (FOV3) greater than on-axis FOV5 at end of SW band
![Page 15: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/15.jpg)
FOV variability of NEDNNOAA20 CrIS
• ~75% variation among FOVs in the LW band, with FOVs 7 and 4 as notable outliers
• In the MW, FOV9 has ~2x higher noise than other FOVs. (It also has nigh nonlinearity (the other NOAA20 MW FOVs are ~linear) and is from the same detector lot as SNPP FOV7.) Aside from FOV9, MW variations are ~15%.
• SW variations are similar to SNPP.
![Page 16: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/16.jpg)
Journal of Geophysical Research: Atmospheres, Volume: 118, Issue: 23, Pages: 13,108-13,120, First published: 25 November 2013, DOI: (10.1002/2013JD020457)
Correlated (red) and random noise (green) contribution to the total NEdN (blue) estimated from the ECT spectra acquired during dynamic interaction test for center FOV5 in MWIR spectral band. (a) Baseline NEdN is compared with (b) NEdN estimated for an external vibration of 5⋅10−3 g0 injected along the Y axis at 158 Hz. Black line is a spec NEdN value.
SNPP spectrally correlated noise, Pre-launch testingMidwave band FOV5 example from Zavyalov et al., Noise performance of the CrIS instrument
Baseline environment With external vibration
![Page 17: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/17.jpg)
Random/correlated noise contribution to the total NEdN in SWIR spectral band estimated for all nine FOVs from the ICT data acquired on 10 January 2013, Orbit 6245. Note that the blue line (total noise) overlays the green line (random noise).
SNPP spectrally correlated noiseShortwave band example from Zavyalov et al., Noise performance of the CrIS instrument
![Page 18: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/18.jpg)
NEDN versus scene radiance
scene radiance (r.u.)
NED
N2
(r.u.
2 ) wavenum
ber
1100
1000
900
800
700
Longwave
scene radiance (r.u.)
NED
N2
(r.u.
2 )
wavenum
ber
2600
2500
2400
2300
2200
Shortwave
1600
1500
1400
1300
1200
scene radiance (r.u.)
NED
N2
(r.u.
2 ) wavenum
ber
Midwave
NEDN increases with sqrt{scene radiance}, consistent with photon noise. The total noise at scene temperature T is parameterized as
NEDN(T) = [N(T) gphoton + NEDNthermal2]1/2
where NEDNthermal2 (the y-intercepts) and gphoton
(the slopes) are determined for each channel.
Spectrally Correlated Noise
The PCA estimate is of the spectrally uncorrelated noise; the spectrally correlated noise is computed as [total_noise2 - pca_noise2]1/2 and compared to pre-flight determinations performed by JPL/BAE:
wavenumber
NEN
corr
elat
ed/ N
ENun
corr
elat
ed
Ratio of correlated noise to uncorrelated noise
PCA estimatePre-flight (JPL/BAE)
• Very good agreement between two very different and independent analyses.• The correlated noise is a large fraction of the total noise for several arrays.
Examples for EOS-Aqua Atmospheric InfraRed Sounder
![Page 19: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/19.jpg)
Suomi-NPP CrIS Observed and Calculated Instrument Lineshapes FOVs 5, 4, and 1
FOV5
FOV1 FOV4
![Page 20: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/20.jpg)
!Les = Lict ⋅F ⋅ fATBD ⋅SAs
−1 ⋅ fATBD ⋅ΔS1ΔS2
ΔS2⎡
⎣⎢
⎤
⎦⎥
F ⋅ fATBD ⋅SAs−1 ⋅ fATBD ⋅ ΔS2
• Complex calibration method (Revercomb, 1988) used for radiometric calibration• Onboard neon source for spectral calibration• Instrument self-apodization (SA) correction via inverse self apodization operator (Genest
and Tremblay, 1999; Desbiens et al., 2006)
Ø SA-1 is a de-apodization process, amplifying and correlating signal and noise• Han et al., “Effect of self-apodization correction on Cross-track Infrared Sounder radiance
noise”
DS1=SES - SDS DS2=SICT - SDS
CrIS Calibration Equation/Algorithm
![Page 21: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/21.jpg)
NEDN amplifications due to SA-1
Han et al., “Effect of self-apodization correction on Cross-track Infrared Sounder radiance noise”
FSR Unapodized
NSR Unapodized
![Page 22: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/22.jpg)
CrIS Noise Covariance example
Center FOVSide FOVCorner FOV
Unapodized
HammingApodized
wavenumber
![Page 23: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/23.jpg)
FOV1 noise covariance, Hamming apodized, log scaleCrIS Noise Covariance example
![Page 24: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/24.jpg)
Effects of SA-1 and Hamming Apodization on NEDN
Center FOVSide FOVCorner FOV
![Page 25: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/25.jpg)
Spectral correlation due to SA-1 and Hamming apodization
1 channel away
2 channels away
3 channels away
1 channel away
2 channels away3 channels away
wavenumber wavenumber
![Page 26: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/26.jpg)
CrIS Simplified On-Orbit Radiometric Calibration Equation:
LS = Re {(C’ES – C’
DS) /(C’ICT-C
’DS)} RICT
for observed complex spectra, C, of the Earth scene (ES), Internal Calibration
Target (ICT), and Deep Space (DS) views.
with:
1. ICT Predicted Radiance: RICT = eICT B(TICT) + (1-eICT) B(TICT, Refl)
2. Quadratic Nonlinearity Correction: C’ = C � (1 + 2 a2 VDC)
3. Polarization Error (aka Correction):
for polarization coefficients prpt, scene selection mirror polarization angle δ, sensor
polarizer angle α, and emission from the scene mirror BSSM. (H==ICT, C==DS).
![Page 27: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/27.jpg)
Example Radiometric Uncertainty estimatesFor a warm clear sky scene (~worst case)
TICTeICTTrefl,measTrefl,modelpr pta2Total
![Page 28: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/28.jpg)
TICTeICTTrefl,measTrefl,modelpr pta2Total
Example Radiometric Uncertainty estimatesFor a cold cloud scene
![Page 29: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/29.jpg)
Calibration Uncertainty Covariance examples
Radiance spectrum
Sqrt(Diag(Cov))
Radiance spectrum
Sqrt(Diag(Cov))
Warm clear sky
scene
Cold cloudy scene
![Page 30: CrISNoise and Calibration Uncertaintycimss.ssec.wisc.edu/itwg/itsc/itsc22/presentations... · as [total_noise 2-pca_noise]1/2and compared to preflight determinations performed by](https://reader035.vdocument.in/reader035/viewer/2022070719/5edf89dead6a402d666ae0b4/html5/thumbnails/30.jpg)
• CrIS noise characteristics• Random from footprint to footprint• NEDN is independent of scene; convert to NEDT at scene T if needed• NEDN level is FOV dependent with a few significant outliers• Self-apodization corrections increase NEDN and introduces noise correlation
between channels, with dependencies on FOV position and channel frequency.• Hamming apodization reduces NEDN and further alters the spectral correlation
among neighboring channels• Results are consistent with Han et al., and covariance matrices of various
flavors are available for testing.• CrIS calibration uncertainties
• Generally small and stable, but scene dependent (not a stagnant “bias” in radiance or Tb) and highly spectrally correlated, and spatially correlated to the extent that adjacent scenes are spatially correlated
• Next step: Estimate covariance from “everything else”. i.e. sO-B
2 = sNoise2 + sCal
2 + [ sRT2 + sAtmState
2 + … ]
Summary and Next Steps