cris spectral calibration and trending · 1 spectral cal. high spectral res. mode sinc ringing obs...

1

Spectral Cal. High Spectral Res. Mode Sinc Ringing Obs Cold Scene SNOs Background Slides

CrIS Spectral Calibration and Trending

L. Larrabee Strow1, Howard Motteler1, Breno Imbiriba1, ChrisHepplewhite1, and Dave Tobin2

1Department of Physics, JCETUniversity of Maryland Baltimore County (UMBC)

2Space Science Engineering CenterUniversity of Wisconsin Madison

NOAA CrIS SDR ReviewCollege Park, MD

2


Overview

Spectral calibration and apodization corrections

High spectral resolution mode validation

Methodology: Measuring excess “sinc ringing”

Methodology and Update: Analysis of cold-scene SNOs

Terminology

LW = Long-wave band1MW = Mid-wave band2SW = Short-wave band3ν = Frequency

Focal Plane FOV Definitions

7 4 1

8 5 2

9 6 3

3


Spectral Calibration SensitivityRemove self-apodization and shift frequency scale

Apodization Corrections

Adjust asymmetric off-axis FOVILS to on-axis

Shift spectra to fixed ν scale

Performed with CMO operator

Requires: focal plane geometry, νof metrology laser

725 730 735 740 745230

240

250

260

270

280

290

Wavenumber (cm−1

)

B(T

) in

K

Filled Spectrum

CrIS Obs

Adjustments can be very large: 20K+

1 ppm ν calibration error = ±0.06K in B(T)

CrIS ν calibration specification is 10 ppm

NWP applications require uniform calibration among FOVs

4


Relative ν Calibration

ν calibration relative to center FOV(FOV-5)

Provides FOV positions relative tointerferometer optical axis which areused to generate self-apodizationcorrections.

Tables below are ν calibration errors(ppm) relative to FOV-5.

From Dave Tobin/SSEC

LW0.20 0.03 0.030.15 0 0.030.15 0.12 -0.01

MW-0.23 -0.12 -0.310.14 0 0.000.03 0.12 -0.18

SW-0.06 -0.41 -0.68-0.04 0 0.100.27 -0.64 -0.08

Self-apodization parameters well characterized and very stable. FOV frequencymismatches due to incorrect focal plane geometry should be well below the 1ppm level except possibly for short-wave.

5


Absolute ν Calibration (units are ppm)

Based on CCAST radiances

Feb. 2012 CCAST agreement with TVACNeon Cal: 0.6 ppm, so Neon cal notchanged

Absolute ν calibration with IDPS SDRsproblematic: Neon lamp value used inCMO unknown until Nov. 2013

Largest ν error: IDPS requires 2 ppmNeon change to update

Absolute LW ν Calibration(CCAST SDRs in Feb. 2012)

0.1 -0.0 0.00.1 -0.6 -0.5

-0.8 -1.0 -0.4

Jan12 Apr12 Jul12 Oct12 Jan13 Apr13 Jul13 Oct13 Jan14−3.5

−3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

Time

Fre

quency C

alib

ration (

ppm

)

Upwelling (Truth)

Neon

Difference

Neon−Resamp

SW Aug. 28 ν Cal (hi-res mode) LW ν Cal minus Short-wave ν Cal

0.08 -1.37 -1.25-0.72 -0.53 -1.85-0.46 -1.10 -1.16

Mean/Std over FOV: -0.93 ± 0.58 ppm

0.02 1.37 1.250.82 -0.07 1.35

-0.34 0.10 0.76

Mean/Std over FOV: 0.58 ± 0.67 ppm

The LW and SW spectral calibration agree to <1 ppm (using CCAST).

6


Absolute ν Calibration: Figure ZOOM

Jan12 Apr12 Jul12 Oct12 Jan13 Apr13 Jul13 Oct13 Jan14−3.5

−3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

Time

Fre

quency C

alib

ration (

ppm

)

Upwelling (Truth)

Neon

Difference

Neon−Resamp

Data using IDPS long-wave SDRs; Very few updates due to 2 ppm threshold

SDR’s exhibit ∼3 ppm variability

Correct operation of CMO generation started in Nov. 2013

7


IDPS Upgrade

Feb. 2012: IDPS ILS parameters for FOVs (1:4, 6:9) fudgeddue to suspected SDR algorithm code errors for FOV 5. CCASTSDR results did not show these errors.

Yong Han found three errors in computation of FOV5 CMOoperator, fixed in ADL.

Fudge removed in new ENGR packets and tested via 1-day ofADL runs done at UW. Now using original CCAST derived ILSparameters.

UW relative ν calibration and UMBC absolute ν calibrationfrom new ADL SDR radiances proved that IDPS FOV-5 errorshad been corrected.

IDPS may start using corrected SDR code and new ENGRpacket in Feb. 2014.

8


Short-Wave ν Calibration

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

−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

Shortwave (SW) at high spectral resolution provides very goodν calibration

Spectral region used by IASI for all ν calibration

SW channels emit in upper-strat/mesosphere. This will allownearly continuous spectral calibration during each orbit,unlike LW.

9


High Spectral Resolution Validation

Validate Aug. 2013 high-spectral resolution data

CCAST used to generate high-spectral resolution SDRs

Lien: No non-linear corrections for mid-wave (coming soon)

Validation approach 1: SNOs versus IASI converted to CrIS ILS

Validation approach 2: Analyze biases versus computedradiances using ECMWF and SARTA Radiative TransferAlgorithm (RTA)

10


High Spectral Resolution SNOs with IASI

500 1000 1500 2000 2500 3000

210

220

230

240

250

260

Wavenumber (cm−1

)

B(T

) in

K

CrIS

IASI

CrIS−normal

2300 2310 2320 2330 2340 2350 2360 2370

215

220

225

230

235

240

245

250

Wavenumber (cm−1

)

B(T

) in

K

CrIS

IASI

CrIS−normal

SNO agreement is very goodBlack circles on right graph show CrIS at normal mode resolutionLine structure in 2330 cm−1 region provides very good spectralcalibration throughout most of an orbit

11


High Spectral Resolution SNOs with IASI: Details

B(T) Diff B(T) Diff vs Radiance

1200 1400 1600 1800 2000 2200 2400 2600−1

−0.5

0

0.5

1

1.5

2

Wavenumber (cm−1

)

CrI

S −

IA

SI_

mod in K

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7−1

−0.5

0

0.5

1

1.5

2

2.5

3

3.5

Channel Radiance

BT

CrI

S−

IAS

I in

K

Note: No non-linear corrections in mid-wave bandShort-wave agreement very good (this is sinc ILS!)Larger BT differences in cold channels, noise related.

12


High Spectral Resolution NWP Biases

B(T) Bias % Radiance BiasSTD over FOVs STD versus ECMWF

500 1000 1500 2000 2500 3000−3

−2

−1

0

1

2

3

4

5

6

Wavenumber (cm−1

)

Bia

s/S

td in

K

No Non−linear Corrections in MW!

Bias vs ECMWF

Std Bias over FOV

2150 2200 2250 2300 2350 2400 2450 2500 2550

−2

−1

0

1

2

3

4

5

Wavenumber (cm−1

)B

ias/S

td in

% o

f 2

87

K R

ad

ian

ce

Bias vs ECMWF

Std Bias

Relatively good agreement with ECMWF (clear ocean tropical)Std of bias over FOV implies accurate SW ILS correctionsRinging in short-wave accentuated by cold scene temperaturesShort-wave % radiance bias shows ringing is constant over bandIncorrect CO in RTA gives high bias errors

13


Method for Observing Excess Sinc Ringing(See Dan Mooney’s presentation for more information.)

The CrIS ILS specification is sinc ILS.Most RTAs simulate apodized radiances.Ringing artifacts far smaller than RTA + NWP simulations.Enhance visibility of non-sinc ringing with differences (D1,2,3)

D1 ≡ biasILS = BTobs,ILS − BTcalc,ILS

where ILS = Hamming or Sinc. BTcalc,Hamming is RTA simulation fromECMWF. The Hamming forward operator reduces sinc ringing by an orderof magnitude. Create BTcalc,Sinc with the inverse Hamming operator.

D2(sweep_dir) ≡ (biassinc − biasHamming)sweep_dir

highlights non-sinc ringing. sweep_dir selected using odd or even FORs.Differential ringing bias with sweep direction is then

D3 ≡ D2(odd_FOR)−D2(even_FOR)

This approach removes model, RTA, and smooth radiance calibrationerrors, leaving non-sinc ringing in D2 and D3.

14


Examples of Sinc Ringing MetricErrors shown as percent of a 287K blackbody radiance

Left hand side: This shows the excess (non-sinc) ringing in theCrIS spectra for odd FORs. It is quite large for the short-wave.

Right hand side: Here we form the triple difference, bysubtracting the odd FOR (biassinc − biasHamming) from the samequantity but for even FORs.

D2(odd_FOR) D3

1000 1500 2000 2500−2

−1.5

−1

−0.5

0

0.5

1

1.5

2FOV5 Sinc − Hamming Odd Bias in Percent Radiance Units at 287K

Wavenumber (cm−1

)

%R

ad

ian

ce

800 1000 1200 1400 1600 1800 2000 2200 2400

−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

0.2

0.3

0.4

0.5

FOV5 Sinc−Hamming, Odd−Even Bias in Percent Radiance Units at 287K

Wavenumber (cm−1

)

%R

ad

ian

ce

15


SNOs: Improved MethodologyCorrects error in last SDR Review

Last Review

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

CrIS B(T)’s are warmer when AIRS SNOB(T)’s are cold.

Methodology Error

Scene subset selection based only on onesensor is biased and non gaussian. Subseton union of scenes with AIRS and CrISB(T)’s within some B(T) bin. Color scale:log(counts).

16


Proper SNO Scene Selection and Binning

205 210 215 220 225205

210

215

220

225

AIRS BT (K)

CrI

S B

T (

K)

Sample population is union of scenesin bin for both instruments.

180 200 220 240 260 280 300−0.4

−0.3

−0.2

−0.1

0

0.1

0.2

0.3

0.4

BT_CrIS or BT_AIRS (K)

AIR

S −

CrI

S in

KGlobal SNO versus scene temperaturefor ν=900 cm−1.

See Dave Tobin’s presentation for detailed SNO results.

17


Shortwave Cold Scene SNOsRe-visit data shown in previous review

No 10K issues as before. SNOdifferences far below noise. Bothstandard SNO and quantile resultsshown.

Same data as on left, but plottedversus scene radiances instead of B(T).Note noise level. This is quantileresult.

Short-wave cold scenes: Too sensitive to noise, avoid. However,inter-comparisons with AIRS are very good, but not perfect. No coldscene issues relative to AIRS.

18


Conclusions

CrIS IDSP SDR spectral calibration good to ∼3 ppm.

CrIS instrument and up-welling ν calibration capable of∼1ppm absolute and <0.5 ppm relative ν calibration.

Achieving instrument capabilities require removal of IDPS 2ppm threshold for computation of new CMO. Can be achievedby interpolation of CMO operators (very smooth) as done byCCAST.

Metrology laser drifts track instrument temperaturevariations. (Not discussed.)

High-spectral resolution mode working very well, short-waveagreement with IASI to ∼0.1K, accurate ILS corrections.

High-spectral resolution mode spectral calibration veryaccurate, can be used over much of each orbit.

Improved methodology for SNOs shows that CrIS and AIRSagree very well for cold scenes (in all bands).

19


Quantile Analysis of SNO Data

10−3

10−2

10−1

100

200

210

220

230

240

250

260

270

280

290

300

Cumulative Probability

Me

an

B(T

) in

K

Why Quantile Approach?

Tropical SNOs have large differences(small cold clouds)

Limited # of cold SNOs, poorstatistics

Quantile analysis quantifies statisticaldistribution of each instrument SNOradiances independently

Allows larger data sets?

Select a cumulative probability vector p = 0.0001:δP:1.0Sort radiances by value, find mean radiance in each δP bin.Distribution difference is difference in radiance/B(T) associatedwith same cumulative probability bin.

Does not require subtraction of SNO pairs, which may have very largeB(T) differences (especially in the tropics).

20


CrIS Radiometric StabilityRelative to SST, CO2

Apr12 Jul12 Oct12 Jan13 Apr13 Jul13 Oct13 Jan14−0.6

−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

0.2

0.3

Bia

s in K

Tropical ocean clear

1-Year differences far below0.1K. Red curve is smoothedtime series.

Apr12 Jul12 Oct12 Jan13 Apr13 Jul13 Oct13 Jan14−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15

0.2

B(T

) C

hange in C

O2 L

ine D

epth

2.5 ppm CO2

CO2 from ECMWF bias (791.5cm−1) - 0.27*bias(790 cm−1).

Second term removes any SST,H2O variability.

Oct 2012 through Oct 2013shows 2.5 ppm growth rate(0.06K).

cris spectral calibration and trending · 1 spectral cal. high spectral res. mode sinc ringing obs...

Documents