CrIS SDR Product Review UMBC Validation Status · B(T) in K Number of Obs AIRS CrIS 160 180 200 220...
Transcript of CrIS SDR Product Review UMBC Validation Status · B(T) in K Number of Obs AIRS CrIS 160 180 200 220...
1
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS SDR Product ReviewUMBC Validation Status
L. Larrabee Strow, H. Motteler, P. Schou, B. Imbiriba, P. Schou,S. Hannon
Physics Department andJoint Center for Earth Systems Technology
University of Maryland Baltimore County (UMBC)
October 23, 2012
2
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Overview
Frequency calibration
Short-wave high-resolution mode results
Radiometric validationInternal consistency (among FOVs)Relative to AIRS, IASI (SNOs)Relative to AIRS, IASI (Double Differences)
Main Results
Frequency calibration working very well
SDRs exhibit boxcar ringing inconsistencies: up to 1K
Hamming apodization reduces these problems significantly
Comparisons to AIRS, IASI within 0.2K or less
Evidence for a ∼0.2K systematic calibration error in far long-wave
High resolution works very well!
3
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS Frequency Calibration: Two Distinct Parts
Pseudo ν Cal: Alignment of detectors (Tobin)
Detector offsets from interferometer axis produces frequency shifts
Corrected with CMO operator, once 3x3 positions known
In-orbit 3x3 positions determined from relative frequency offsets
Original plan: Absolute shifts using B(T) Obs vs Calcs, gives both offsetsand Neon cal.
Relative approach more accurate (Tobin), < 1 ppm LW, MW, SW maybe 2-3ppm?
Absolute Calibration: Upwelling -→ Neon Lamp -→ Met Laser
Neon spectral calibration and alignment to interferometer axis needed.
ITT expected 1/month Neon calibration (so far not needed).
In-orbit Neon calibration done using LW window regions, sharp water lineson flat background. Limits one to very clear ocean scenes.
Unable to achieve similar accuracy with opaque channels in LW or MW.
SW hopeless for Neon calibration with 0.2 cm OPD.
4
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Present Focal Plane Positions and Neon Calibration
The Neon lamp calibration was determined using data fromFeb. 25 LW window region observations.
Errors in treatment of FOV5 in IDPS SDR code forced us to“fudge” the positions of the remaining 8 detectors in the LWand MW arrays.
These adjustment were small, LW=1.4 and MW=2.2 ppm. Notsufficient information to modify SW.
Formal monthly Neon calibrations are presently not beingdone. Only offline at UMBC.
Has production of new CMO operator happened? Maybe soon?
5
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
ν Calibration Overview
CrIS In-orbit Neon Cal unchanged from TVAC!
Examine time series of Neon calibration using (a) Long-waveand (b) Mid-wave bands.
Calibration done daily using clear ocean tropical subsets
650 700 750 800 850 900 950 1000 1050 11000.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
Wavenumber (cm 1)
FOV3
F
OV8
in K
Spectral Cal Difference between 2 Detectors with existing LUTs
7.5 ppm Cal Error
6
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Metrology Laser Wavelength vs Time
Left: Metrology laser λ (and temperature) versus time a/c to Neonlamp, Right: Zoom showing daily variation.
Feb Mar Apr May Jun Jul Aug Sep Oct
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5
3
3.5
Time
∆ ν
(pp
m)
Met Laser νLaser Temp
07/27 07/28 07/29 07/30 07/31 08/01 08/02 08/03 08/04
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Time∆
ν (
ppm
)
Linear drift starting in mid-September when temperature offsetoccurred?
7
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS Frequency Calibration vis Upwelling RadiancesUsing Long-wave Band
Left: All fovs, Right: Mean and Std (over FOVs)
Apr May Jun Jul Aug Sep Oct Nov−2.5
−2
−1.5
−1
−0.5
0
0.5
ν E
rror
(pp
m)
123456789
Apr May Jun Jul Aug Sep Oct Nov0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
ν E
rror
(pp
m)
−1 x Mean band1 ν calStd over Fovs
Upwelling frequency calibration mirrors the Neon calibration of the metrologylaser. If the CMO has not been updated (waits for a 2 ppm change), this meansthe metrology laser has indeed drifted. If the Neon was drifting, we would notsee the same shift in the upwelling spectra (again assuming the CMO operatorremains unchanged). Slight differences among FOVs.
8
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS Frequency Calibration vis Upwelling RadiancesUsing Mid-wave Band
Left: All fovs, Right: Mean and Std
Apr May Jun Jul Aug Sep Oct Nov−4
−3.5
−3
−2.5
−2
−1.5
−1
−0.5
0
0.5
1
ν E
rror
(pp
m)
123456789
Apr May Jun Jul Aug Sep Oct Nov0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
ν E
rror
(pp
m)
−1 x Mean band2 ν calStd over Fovs
The mid-wave frequency calibration shows a frequency drift very similar tolong-wave. I do not know why the mid-wave frequency calibration varies by up to3.5 ppm among FOVs. We saw this in the Feb. 25 data. D. Tobin’s relativecalibration indicates this is incorrect.
9
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Neon vs Upwelling ν Calibration
Apr May Jun Jul Aug Sep Oct Nov0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
ν E
rror
(pp
m)
−1 × Mean band1 ν calStd over FovsNeon ν cal +1.3 ppm
SDR algorithm waits for a 2 ppm Neon shift to re-compute new CMOPresumably that has not yet happened, so cannot testThus, upwelling calibration roughly follows NeonDifferences may be upwelling algorithm issues?
10
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS High Resolution Mode
UMBC has processed all the CrIS high-resolution mode datafrom Feb. 23, 2012 into calibrated radiances using CCAST
Liens on these radiances:Non-linear correction not applied (no effect on shortwave)Nominal geolocation (good enough for most purposes)Not in SDR formatUses our best-effort CMO apodization removal operators fromthe July time-frame.These data recorded with the old FIR decimation filter
We have computed clear-sky observed radiances for everyobservation
For a single southern ocean granule we have compared thesedata to IASI data in the same region, suitably degraded to CrIS0.8 OPD resolution.
11
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Example High-Resolution Spectra
800 1000 1200 1400 1600 1800 2000 2200 2400
180
200
220
240
260
280
Wavenumber (cm−1)
B(T)
in K
Normal ModeHigh−Res Mode −50K
12
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Example Spectra: Shortwave Only
2150 2200 2250 2300 2350 2400 2450 2500 2550
180
200
220
240
260
280
300
Wavenumber (cm−1)
B(T
) in
K
Normal ModeHigh−Res Mode
Not Noise!
13
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
High-Resolution CMO Operator
UMBC developed a high-resolution CMO operator
Matrix inversion to derive CMO operator condition numbergoes from 1 for center FOV to 106 for corner FOVs.
With careful filtering, high condition number can be handled
Note below: large relative ν offset of off-axis spectra; Left:Normal Mode, Right: High-resolution Mode
2354 2356 2358 2360 2362 2364
−0.2
0
0.2
0.4
0.6
0.8
1
Wavenumber (cm−1)
Impu
lse
Res
pons
e
CornerSideCenter
2354 2356 2358 2360 2362 2364
−0.2
0
0.2
0.4
0.6
0.8
1
14
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Relative Error in CMO Corrections
Plot shows single uniform 3x3 spectraDifference curves (from FOV5) are uniform in frequencyCorrection errors close to radiances values for cold observations.
2150 2200 2250 2300 2350 2400 2450 2500 2550−0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
Wavenumber (cm−1)
Rad
ianc
e
FOV1FOV2FOV3FOV4FOV5FOV6FOV7FOV8FOV9
2150 2200 2250 2300 2350 2400 2450 2500−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Wavenumber (cm−1)
Rad
ianc
e
FOV1FOV2FOV3FOV4FOV5FOV6FOV7FOV8FOV9
15
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CMO Correction Errors Reduced with HammingApodization
Same data as in previous slide, but Hamming apodized
2150 2200 2250 2300 2350 2400 2450 2500 2550−0.5
0
0.5
1
1.5
2
2.5
3
3.5
Wavenumber (cm−1)
Rad
ianc
e
FOV1FOV2FOV3FOV4FOV5FOV6FOV7FOV8FOV9FOV corner − FOV5FOV side − FOV5
2200 2250 2300 2350 2400 2450 2500 2550
−0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Wavenumber (cm−1)
Rad
ianc
e
FOV1FOV2FOV3FOV4FOV5FOV6FOV7FOV8FOV9FOV corner − FOV5FOV side − FOV5
16
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CMO Correction Errors in Brightness Temperature:Hamming
FOVn-FOV5 differences multiplied by 5X and offset by 260K
2150 2200 2250 2300 2350 2400 2450 2500 2550220
230
240
250
260
270
280
290
300
Wavenumber (cm−1)
B(T
) in
K
FOV1
FOV2FOV3
FOV4
FOV5
FOV6
FOV7FOV8
FOV9
+−1K (FOVn−FOV5)
2160 2180 2200 2220 2240
245
250
255
260
265
270
275
280
285
290
295
Wavenumber (cm−1)
B(T
) in
K
FOV1
FOV2
FOV3
FOV4
FOV5FOV6
FOV7
FOV8
FOV9+−1K (FOVn−FOV5)
17
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
High-Resolution Validation: CrIS vs IASI B(T)
2200 2250 2300 2350 2400 2450 2500
230
240
250
260
270
280
Wavenumber (cm−1)
B(T
) in
K
IASICrIS
(a) CrIS/IASI Obs, Hamming Apodized.
2170 2180 2190 2200 2210 2220245
250
255
260
265
270
275
280
285
Wavenumber (cm−1)
B(T
) in
K
IASICrIS
CO Lines
(b) CrIS/IASI Obs, Hamming Apodized:Zoom.
18
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
High-Resolution Validation: CrIS vs IASI BiasesBiases with respect to ECMWF
2200 2250 2300 2350 2400 2450 2500 2550
−1.5
−1
−0.5
0
0.5
1
1.5
Wavenumber (cm−1)
Bia
s in
K
IASICrIS
(c) CrIS/IASI Bias, Hamming Apodized.
2170 2180 2190 2200 2210 2220−1
−0.5
0
0.5
Wavenumber (cm−1)
Bia
s in
K
IASICrIS
Window Channels
Atmospheric CO Lines
(d) CrIS/IASI Obs, Hamming Apodized:Zoom.
Note in (c,d) above, one expects IASI and CrIS window channels to differby 0.1K due to diurnal variation in the SST. Here we use a constantdiurnally averaged SST. Thus, the bias difference between CrIS and IASI isabout 0.1K less than shown here for window channels!
19
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CO Retrievals from High-Resolution Mode SpectraLeft: CrIS, Right: AIRS Color Scale in K
“Retrieval” is just bias between Obs and Calc radiances for a single CO channel.Calc radiances use ECMWF.Remove scenes where window radiance bias > 5K (Clouds).
20
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CO Retrievals from High-Resolution Mode SpectaLeft: CrIS, Right: MOPPIT
21
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CO Retrievals from High-Resolution Mode Specta
22
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Frequency Calibration Using High-ResolutionShort-Wave
High-resolution data used to calibration Neon, 1 day’s worth
High-resolution brings out very stable features with spectralcontrast
23
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
High-Resolution SW Calibration of Neon
2200 2250 2300 2350 2400 2450 2500
220
225
230
235
240
245
250
255
260
265
270
Wavenumber (cm−1)
B(T
) in
K
All ChannelsNeon Cal Channels
Existing Neon calibration limited to non-polar clear ocean scenes (morebelow).
As previously stated, unable to achieve high Neon calibration accuracy withopaque LW, MW channels, which would allow calibration over the entireorbit.
High-resolution in the SW allows us to use the very high-peaking lines in the2350 cm−1 region, indicated by green circles above.
24
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
LW Opaque Channel Calibration of Neon
660 680 700 720 740 760
220
225
230
235
240
245
250
255
260
265
270
Wavenumber (cm−1)
B(T
) in
K
All ChannelsNeon Cal Channels
(e) LW opaque Neon cal channels.
0.01 0.02 0.03 0.04 0.05
50
100
150
200
250
300
350
Kernel Function
Pre
ssur
e in
mba
r
LWSW
(f) LW vs SW cal channel kernels.
Opaque LW channels include emission from 200 mbar. Leads to inaccurateNWP calculations (clouds, polar) with poor performance in the polar night.
Moreover, LW opaque channel frequency calibration not accurate;presumably due to NWP radiance calculation errors.
With high-resolution, can use extremely high-peaking CO2 channel (5-10mbar, 30+ km) with very good performance.
25
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Final Results: Neon Frequency Calibration
−80 −60 −40 −20 0 20 40 60 80−20
−15
−10
−5
0
5
10
15
20
25
Latitude
Neo
n O
ffset
in p
pm
LW High−AltitudeSW High−AltitudeLW Operational
Each circle represents a 360 second
period that allowed an accurate
calibration.
Present operational Neon: black circles, LW window, ± 40 degrees latitude.
LW opaque channels provide more observations (blue circles). But, apparent5 ppm offset, and very poor performance in the polar night.
SW opaque using CrIS high-spectral resolution mode gives extremely goodperformance, low noise, well into the polar night portion of the orbit.
SW high-resolution agrees very well with LW window region Neoncalibration, maybe 1 ppm.
IASI (METOP-A/B) uses these channels for all metrology laser calibration forall three bands.
26
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Radiance Intercomparisons
Following slides are a small sample of radiometricintercomparisons between CrIS and AIRS/IASI.
27
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI Double-Difference
800 1000 1200 1400 1600 1800 2000 2200 2400 2600−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Wavenumber (cm−1)
CrI
S −
IAS
I in
K
SST Diurnal Correction
CrIS Bias vs IASI Bias (relative to ECMWF), tropics, ocean only
Very good agreement. But SST in calcs off by 0.1K! (maybe)
28
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI SNOs + DDs: SNOs for May 2012 (LW)SNOs from JPL Sounder PEATE: 10 min, 8 km windows, S. Hemis: -73 deg S.
650 700 750 800 850 900 950 1000 1050
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Wavenumber (cm−1)
CrI
S−
IAS
I in
K
IDPS Box:SNOIDPS Hamming:SNOStd Err:SNOIDPS−CCAST HammingCrIS−IASI:DD
CrIS-IASI boxcar apodization has large ringing. Uncertain to cause, used all 4 IASIFOVs, all 9 CrIS FOVs for now.
Significant (for climate) offset in the longwave!
Red curve is CrIS from CCAST (UW/UMBC Matlab SDR testbed algorithm). CCASTmuch closer to IASI, but more work needed.
CrIS-IASI DD is bias double-difference from ECMWF
29
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI SNOs + DDs: SNOs for May 2012 (LW)SNOs from JPL Sounder PEATE: 10 min, 8 km windows, S. Hemis: -73 deg S.
650 700 750 800 850 900 950 1000 1050
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Wavenumber (cm−1)
CrI
S−
IAS
I in
K
IDPS Box:SNOIDPS Hamming:SNOStd Err:SNOIDPS−CCAST HammingCrIS−IASI:DD
CrIS-IASI boxcar apodization has large ringing. Uncertain to cause, used all 4 IASIFOVs, all 9 CrIS FOVs for now.
Significant (for climate) offset in the longwave!
Red curve is CrIS from CCAST (UW/UMBC Matlab SDR testbed algorithm). CCASTmuch closer to IASI, but more work needed.
CrIS-IASI DD is bias double-difference from ECMWF
30
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI SNOs + DDs: SNOs for May 2012 (LW)SNOs from JPL Sounder PEATE: 10 min, 8 km windows, S. Hemis: -73 deg S.
650 700 750 800 850 900 950 1000 1050
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Wavenumber (cm−1)
CrI
S−
IAS
I in
K
IDPS Box:SNOIDPS Hamming:SNOStd Err:SNOIDPS−CCAST HammingCrIS−IASI:DD
CrIS-IASI boxcar apodization has large ringing. Uncertain to cause, used all 4 IASIFOVs, all 9 CrIS FOVs for now.
Significant (for climate) offset in the longwave!
Red curve is CrIS from CCAST (UW/UMBC Matlab SDR testbed algorithm). CCASTmuch closer to IASI, but more work needed.
CrIS-IASI DD is bias double-difference from ECMWF
31
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI SNOs + DDs: SNOs for May 2012 (LW)SNOs from JPL Sounder PEATE: 10 min, 8 km windows, S. Hemis: -73 deg S.
650 700 750 800 850 900 950 1000 1050
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Wavenumber (cm−1)
CrI
S−
IAS
I in
K
IDPS Box:SNOIDPS Hamming:SNOStd Err:SNOIDPS−CCAST HammingCrIS−IASI:DD
CrIS-IASI boxcar apodization has large ringing. Uncertain to cause, used all 4 IASIFOVs, all 9 CrIS FOVs for now.
Significant (for climate) offset in the longwave!
Red curve is CrIS from CCAST (UW/UMBC Matlab SDR testbed algorithm). CCASTmuch closer to IASI, but more work needed.
CrIS-IASI DD is bias double-difference from ECMWF
32
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS and IASI SNOs: Data for May 2012 (MW)
1300 1400 1500 1600 1700
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
NH (day)
Wavenumber (cm−1)
CrI
S−
IAS
I in
K
IDPS BoxcarIDPS HammingStd. Err
CrIS-IASI boxcar apodization again has ringing.
Very good agreement. Can we determine interconsistency below 0.05K?
33
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CrIS-AIRS SNOs LocationsWith 10-min,8 km window obtain full latitude range!
−100 −50 0 50 1000
1000
2000
3000
4000
5000
6000
7000
Latitude
SN
Os
per
Mon
th
Unlike IASI-AIRS or IASI-CrIS, wide latitude range of SNO’s.
This allows very detailed inter-comparisons as a function of scene type. Here we
examine SNO differences with scene temperature for one channel.
34
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
2552 cm−1 SNOs for AIRS, CrIS
160 180 200 220 240 260 280 300 3200
2000
4000
6000
8000
10000
12000
14000
16000
B(T) in K
Num
ber
of O
bs
AIRSCrIS
160 180 200 220 240 260 280 300 320180
200
220
240
260
280
300
320
AIRS B(T) in K
CrI
S B
(T)
in K
2552 cm−1
Good number of SNOs over a large range of B(T)’s
CrIS hits a B(T) floor around 200K.
35
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
SNO AIRS-CrIS: LongwaveEarly Global SNO Comparisons Using AIRS-to-CrIS Conversion
700 750 800 850 900 950 1000 1050−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
AIR
S−
CrI
S in
K
1200 1250 1300 1350 1400 1450 1500 1550 1600−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Wavenumber (cm−1)
AIR
S−
CrI
S in
K
36
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
CCAST vs IDPS: Avg Radiometric Differences
Longwave
650 700 750 800 850 900 950 1000 1050
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
Wavenumber (cm−1)
∆ B
(T)
in K
Midwave: 0.006K
1300 1400 1500 1600 1700
−1
−0.5
0
0.5
1
Wavenumber (cm−1)
∆ B
(T)
in K
Shortwave: 0.03K
2150 2200 2250 2300 2350 2400 2450 2500 2550
−1
−0.5
0
0.5
1
Wavenumber (cm−1)
∆ B
(T)
in K
Longwave: FOVs 1-3,4-6
Midwave: FOVs 1,6,9
Averaged over all FOVs forshortwave (no non-linear)
37
Intro ν Cal High-Res Mode CO High-Res ν Cal Rad. Cal.
Possible Errors for High Temperature ScenesReal part of 860 cm−1 vs Imaginary Part
Color scale is number of observations