1
Calibration of SNPP-CrIS for the NASA L1bProduct
L. Larrabee Strow, and Howard Motteler
Department of Physics, JCET, University of Maryland Baltimore County (UMBC)
AIRS STM: October 2015
2
Overview
Compare using ∼2200 clear ocean tropical scenes
Calculated radiances use kCARTA (monochromatic) andECMWF
Key issues:
C4 RTA needs instrument responsivity applied
CCAST does not need in-band part of responsivity
Does it matter?
3
FOV7 Improvements (Dave Tobin, UW (SSEC )
FOV$7$$Differences)wrt)FOV9))with))MW7)a2)x)1:)
Differences)wrt)FOV9))with))MW7)a2)x)0.70:)
9$
4
Calibration EquationsCCAST now supporting many SDR approaches.
Calibration Equation Definitions
For this talk the term “UMBC” denotes the CCAST reference calibration equationwhile “NOAA” refers to the SDR Team equation C4.
UMBC:
rES = F · ·f · SA−1 · f · ES − 〈SP〉〈ICT〉 − 〈SP〉SA−1rICT
NOAA-C4:
rES = rICT
F · f · SA−1 · f ·{∆S1∆S2
· |S2|}
F · f · SA−1 · f · |∆S2|
∆S1 = FIR−1(ES − 〈SP〉), ∆S2 = FIR−1(〈ICT〉 − 〈SP〉)
rES is calibrated earth-scene radiance at the user grid
F is resampling from sensor to user grid
rICT is the expected ICT radiance
f , UMBC: is a raised-cosine bandpass filter with wings at or inside the instrumentresponsivity, NOAA: modified ATBD filter
SA, UMBC: Periodic sinc ILS wrapping at the sensor grid, NOAA: Periodic sincwrapping at the undecimated sensor grid.
Non-linearity corrections are included
〈SP〉 and 〈ICT〉 are averages over 9 scans
5
Components Operated on by SA−1
Wavenumber (cm-1)
650 700 750 800 850 900 950 1000 1050 1100 1150
Arb
Un
its
0
0.2
0.4
0.6
0.8
1
1.2
α = ∆ S1 / ∆ S
2
Responsivityα * |S
2|
6
Components Operated on by SA−1
NOAA Wavenumber (cm-1)
600 700 800 900 1000 1100 1200
Arb
Un
its
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
7
Components Operated on by SA−1
CCAST Wavenumber (cm-1)
650 700 750 800 850 900 950 1000 1050 1100
Ra
dia
nce
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
8
Simulations (RTA)
RTA with NOAA approach: apply responsivity tomonochromatic radiance, then divide out responsivity afterthe FFT.
This implies a different forward model, RTA, for eachinstrument.
The instrument response function is no longer a sinc function(some disagree with this statement). An individual channel isnot the spectral space convolution with a sinc...
If the detector filter (part of responsivity) changes (spacecrafttemperature), the SDR will change and the RTA will need tochange. Unlikely, but possible.
The radiance record will be different for the next CrIS sincethe reponsivity will be slightly different.
Next slides: UMBC w/ FILT legend in plots means we arecomparing the NOAA simulated radiance to the UMBC CCASTgenerated observations.
9
NOAA, UMBC Observations are DifferentA consequence of the SDR calibration equation.
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
∆ in K
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4UMBC Obs B(T) -NOAA Obs B(T)
10
NWP Biases: All FOVsThe NOAA and UMBC biases are very similar.
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
NW
P B
ias in K
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
UMBCNOAA
11
NWP Bias Difference between NOAA and UMBCRinging only, quite small.
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Bia
s D
iff in
K
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
UMBC-NOAA
12
Ringing DiagnosticsShowing BiasHamming-BiasSinc
UMBC w/ Filt: Btcalc with Responsivity applied before doing the FFT.
Sin
c E
rro
r in
K
-1
0
1UMBC
Sin
c E
rro
r in
K
-1
0
1NOAA
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
∆ B
(T)
in K
-1
0
1UMBC w/ Filt
Mid-wave worse with UMBC w/ Filt.
13
NOAA and UMBC Differences: Responsivity FilterNOAA C5 numerator embeds responsivity info into the SDR
∆ B
(T)
in K
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
UMBC - NOAA Obs BT
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
∆ B
(T)
in K
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
UMBC w/ Resp. Filter - UMBC no filter
14
NOAA and UMBC Differences: Responsivity FilterNOAA C5 numerator embeds responsivity info into the SDR
15
SDR Variability over FOV ID: Sinc ILSNOAA has smaller variability over FOV, esp. in long-wave
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Bia
s S
td o
ve
r F
OV
(K
)
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
UMBCNOAAUMBC-NOAA
16
SDR Variability over FOV ID: Hamming ILSWith Hamming ILS, larger UMBC is gone.
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Bia
s S
td o
ve
r F
OV
(K
)
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
UMBCNOAAUMBC-NOAA
17
Bias with Scan Direction: Sinc ILSNOAA and UMBC very similar.
Even-O
dd in K
-0.5
0
0.5 FOV5 Scan Direction Bias
UMBCNOAA
Even-O
dd in K
-0.5
0
0.5 Side FOVs Scan Direction Bias
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Even-O
dd in K
-0.5
0
0.5 Corner FOVs Scan Direction Bias
18
Bias with Scan Direction: Hamming ILSVery small, DC offsets may be scene differences?
Even -
Odd in K
-0.5
0
0.5
All FOVs Scan Direction Bias
UMBCNOAA
Wavenumber (cm-1)
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Even -
Odd in K
-0.5
0
0.5
All FOVs Scan Direction Bias: Hamming
19
Conclusions
Two reasons for CCAST approach:CrIS RTA does not need complicated in-band responsivity filter forreference truth (good for external users).Differences in CrIS radiance response among different instrumentsminimized using CCAST approach.
Remaining WorkUMBC performance almost identical to C5 in the low end of thelong-wave band, where the UMBC f cutoff matches the responsivitycutoff, hence the out-of-band is handled properly in simulation.Improve? CCAST by modifying the other f filter cutoffs to moreclosely match the responsivity cutoffs.These out-of-band contributions are quite small, and can possiblybe ignored for even climate applications.Examine sensitivity of these proposed calibration equations tochanges in CrIS bands for J2+. Want to keep instrumentinter-calibration to the climate level if possible.
20
LW Filters
wavenumber
600 700 800 900 1000 1100 1200
norm
aliz
ed w
eig
ht
0
0.2
0.4
0.6
0.8
1
1.2
1.4LW CrIS Filters
responsivitynumeric filterccast bandpasssensor griduser grid
21
MW Filters
wavenumber
1100 1200 1300 1400 1500 1600 1700 1800 1900
norm
aliz
ed w
eig
ht
0
0.2
0.4
0.6
0.8
1
1.2
1.4MW CrIS Filters
responsivitynumeric filterccast bandpasssensor griduser grid
22
SW Filters
wavenumber
2000 2100 2200 2300 2400 2500 2600 2700
norm
aliz
ed w
eig
ht
0
0.2
0.4
0.6
0.8
1
1.2SW CrIS Filters
responsivitynumeric filterccast bandpasssensor griduser grid