CALCON 2003 GIFTS On-orbit Spectral Calibration On-orbit Spectral Calibration of the Geosynchronous Imaging Fourier Transform Spectrometer (GIFTS) USU/SDL.

CALCON 2003GIFTS On-orbit Spectral Calibration

On-orbit Spectral Calibration of the Geosynchronous Imaging Fourier Transform Spectrometer (GIFTS)

USU/SDL

CALCON 2003

David C. Tobin, Henry E. Revercomb, Robert O. Knuteson

University of WisconsinSpace Science and Engineering Center

Slide 2CALCON 2003

GIFTS On-orbit Spectral Calibration

Outline

• Introduction to GIFTS• Optical design and spectral characteristics• Calibration requirements• Spectral calibration using Earth scene spectra

– Example: ground based AERI spectra– Example: aircraft based Scanning-HIS spectra

• Summary and future work

Slide 3CALCON 2003


Introduction to GIFTS

Slide 4CALCON 2003


Combine Advanced Measurement Technologies on a Geosynchronous Satellite to obtain

4-D Observations of the Atmosphere

• Horizontal: Large detector arrays give near instantaneous wide 2-D geographical coverage

• Vertical: Michelson interferometer (FTS) gives high spectral resolution that yields high vertical resolution

• Temporal: Geosynchronous orbit allows high time resolution (i.e., motion observations)

GIFTS Measurement Concept

Slide 5CALCON 2003


Spectral Coverage

CrIS

CO2

CO2

O3

N2O CO

N2O

CO2

H2O

H2O

CH4

CrISGIFTS

GOES Sounder

685 – 1129 1/cm 1650 - 2250 1/cm

Slide 6CALCON 2003


GIFTS Optical Design and Spectral Characteristics

Slide 7CALCON 2003


Expanded View of the Electro-optical Design

LWIR SWIR

laser 1064nm, 9398 1/cm

N 2048 4096

R 8 4

X, max OPD (N/2)R/laser= 0.87 cm

Dv 1/(2X) = 0.57 1/cm

Slide 8CALCON 2003


• With constant x sampling, the wavenumber scale is determined by the effective laser frequency and the resulting interferogram sampling rate:

– Optical Path Difference (OPD) scale:• dx = R/ laser X = (N/2) dx

– Wavenumber scale:• d = 1/(2X) = laser /(NR) » scales with laser

• Spectral calibration primarily deals with knowledge of the effective laser

On-Axis Wavenumber Scale

Slide 9CALCON 2003


• The beams of light reaching each FPA pixel pass through the interferometer at different angles,

• With respect to the on-axis beam, the off-axis beams have slightly shorter OPDs: OPD() = OPD(0) cos().

• For the GIFTS geometry, this causes two primary effects: a different (but correct) wavenumber scale for each pixel of the FPA, and small distortions in the Instrument Line Shape (ILS).

• Integrating over a single FPA pixel and making small angle approximations, the GIFTS interferograms are represented as:

xbeNdxF

bxi

2sinc,2

1222

where is the mean off-axis angle for a given pixel and b is the half-angle subtended by a single pixel (~0.38 mrad).

Off-axis Effects

Slide 10CALCON 2003


OPD()=OPD(0) cos()

after Fourier Transform Spectrometry, James W. Brault

M2

M1

On axis beam

Off axis beam

X/2

FTS axis

x2 x1 x2

x1

X cos

OPD = x1 - x2 = X cos OPD = x1 - x2 = X

X

Slide 11CALCON 2003


FPA Geometry

FPA

interferometertelescope

1 = 14.2 mrad

2 = 97.4 mrad

1 = afocal ratio = 6.855

km4km/pixel

b = single pixel half angle = 2/2/128 = 0.38 mrad

(i,j) = off-axis angle to center of pixel = b[(2i-1)2 + (2j-1)2]1/2

Off-Axis Angle, mrad

Not to scale

(i,j)

b

FPA pixel i,j

FPA pixel index i

FP

A p

ixel

inde

x j

Slide 12CALCON 2003


ILS Variations:

very little ILS change over the detector array

xbeNdxF

bxi

2sinc,2

1222

Slide 13CALCON 2003


• In expression for the measured interferogram, F(x), expand sinc function as a power series of (2xb):

...!5

2

!3

2 244

222

2 xixixi eNdxb

eNdxb

eNdxF

• Compute perturbation terms and subtract from measured interferogram.

» Similar process currently performed for AERI, HIS, S-HIS, NAST-I.

Self-apodization correction process

Slide 14CALCON 2003


GIFTS Off-Axis Interferogram Sampling

Samples are triggered with the (on-axis) laser signal, but each IR beam/pixel experiences a different OPD according to its angle through the interferometer. But all sampled points lie on the same continuous interferogram.

OPD (cm)

Inte

rfer

ogra

m (

coun

ts)

CALCON 2003GIFTS On-orbit Spectral Calibration

GIFTS Interferogram Data Cube Simulations

interferogram point #1 (ZPD, OPD=0)

interferogram point #780 (CO2 resonance)

Pixel #

Pixel #

Pixel #

Pixel #

In the spectral domain, produces apparent shifts of spectral features when plotted versus the on-axis wavenumber scale.

Slide 16CALCON 2003


Puts spectra for all pixels onto a common wavenumber scale• Start with decimated interferograms (2048 pts LW, 4096 pts

SMW)• Zero-pad interferograms by factor of 16.• FFT to get oversampled spectra.• Interpolate to standard wavenumber scale.

» Can be performed before or after radiometric calibration.

Off-axis wavenumber scale normalization

Slide 17CALCON 2003


Calibration Requirements

Slide 18CALCON 2003


Calibration Allocation TreeOverall Calibration Spec

Absolute: 1K (3 )

Spatial

Blackbody Emissivity & Temperature

LinearityInstrument

Temperature Stability

Uncalibrated Mirror

Reflectance

Residual Radiance

Noise

Pointing Mirror Polarization & Scattering

InstrumentLine ShapeUncertainty

Geometric Stability

ROIC Cross-talk

OpticalCross-talk

LaserWavenumber

Aliasing

SpectralAbsolute: 510-6 (3 )Stability: 110-6 (3 ) over 1 hr

RadiometricAbsolute: 0.95K (3 )Reproducibility: 0.2K (3 ) over 24 hrs

Basic philosophy is to constrain the spectral calibration such that spectral errors do not contribute significantly to the total calibration budget

Slide 19CALCON 2003


Simulated Earth Scene Brightness Temperature Errors due to wavenumber scale uncertainties

Tb

Diff

(K

)

wavenumber

Tb

(K)

wavenumber

SMW bandLW band

= ×

110-6

210-6

510-6

(1 sigma

perturbations)

• The absolute knowledge of the spectral calibration shall be known to better than 510-6 (3 sigma).• The stability of the spectral calibration shall be known to better than 110-6 (3 sigma) over 1 hour (Threshold), over 30 days (Objective).

Slide 20CALCON 2003


Spectral Calibration using Earth Scene Spectra

Slide 21CALCON 2003


• The short and long term geometric and laser frequency stability, and the resulting spectral calibration, of GIFTS are expected to be very good (better than requirement) by design. Argues for philosophy that spectral calibration is established on the ground with validation/monitoring on-orbit.

• Spectral calibrations with Earth scene data will be used to monitor and remove any long term drifts as required.

• Spectral positions of selected spectral features are known with high accuracy and are used to determine the spectral calibration.

• For a given clear sky Earth spectrum, the effective laser wavenumber and resulting wavenumber scale of the observed spectrum is varied to produce best agreement with a calculated spectrum.

Slide 22CALCON 2003


Earth view spectral calibration example: ground based AERI spectra

Slide 23CALCON 2003


AERI: Atmospheric Emitted Radiance Interferometer

• UW/BOMEM FTS spectro-radiometer providing 1 cm-1 resolution IR spectra.• 632 nm (~15800 cm-1) HeNe laser. =0, b=23mrad.

• The following presents analysis of AERI spectral calibration using 241 clear sky zenith spectra collected over a 3 year period at the SGP ARM site.

sample longwave zenith viewing radiance spectrum

Slide 24CALCON 2003


730-740 cm-1 region of Longwave spectrum

• Fundamental 2 CO2 spectral line parameters known very well from laboratory work

• Observed spectrum largely sensitive to lower troposphere atmospheric temperature

Slide 25CALCON 2003


For each of the 241 cases, radiance spectra are computed using collocated radiosonde profiles, and the spectral calibration is performed. The optimal effective laser wavenumber is found when the RMS residual (observed minus calculated radiance) over the 730-740 1/cm region is minimized. Uncertainty in the final laser wavenumber is reported as 1 std. dev. over the ensemble.

sampling of the AERI spectra

wavenumber (1/cm)

effective laser wavenumber offset (1/cm)

RM

S r

esid

ual

RMS residual as a function of laser wavenumber perturbation for each case

Slide 26CALCON 2003


AERI laser wavenumber analysis

case number laser wavenumber offset (1/cm)

lase

r w

aven

umbe

r of

fset

(1/

cm)

effective laser wavenumber offset versus case number Histograms, before and after Feb 2000

• Earth view spectral calibration determines effective laser wavenumber to ~0.025 cm-1 / ~1.5 ppm (1 sigma) uncertainty.

• ~0.65 cm-1 change in effective laser wavenumber with laser replacement in February 2000.

Slide 27CALCON 2003


Earth view spectral calibration example: aircraft based Scanning-HIS spectra

Slide 28CALCON 2003


Scanning High-resolution Interferometer Sounder

sample nadir viewing brightness temperature spectrum

• UW/BOMEM FTS spectro-radiometer providing 1 cm-1 resolution IR spectra.• 632 nm (~15800 cm-1) HeNe laser. =0, b=20mrad.

• The following presents analysis of SHIS spectral calibration using 82 clear sky nadir spectra collected @ 8 to 12 km altitude over a 3 week period at the SGP ARM site.

Slide 29CALCON 2003


Example Spectral Calibration: S-HIS

Slide 30CALCON 2003


SHIS spectral calibration, 730-740 1/cm, Nominal algorithm.

laser wavenumber = 0.00125 0.018 cm-1, ppm = 0.1 1.1

Slide 31CALCON 2003


Scanning-HIS LW/MW and MW/SW Band Overlap

LW HgCdTe band MW HgCdTe band SW InSb band

LW/MW overlap MW/SW overlap

Slide 32CALCON 2003


Band 1 - 2 (cm-1) note laser (cm-1) laser (ppm)

1 730 - 740 CO2 0.001 0.018 0.1 1.1

1 746 - 760 CO2 0.021 0.021 1.3 1.3

1 1170 - 1205 H2O, high noise 0.052 0.121 3.3 7.7

2 1170 - 1205 H2O 0.172 0.025 10.9 1.6

2 1560 - 1630 H2O 0.169 0.051 10.7 3.2

2 1785 - 1805 H2O, high noise 0.161 0.186 10.2 11.8

3 1785 - 1805 H2O, high noise 0.065 0.172 4.1 10.9

3 2050 - 2100 CO2, CO, H2O 0.036 0.032 2.3 2.0

3 2190 - 2200 N2O -0.010 0.042 -0.6 2.7

Results for different SHIS Bands and spectral regions

starting laser = 15799.706 cm-1

• Accuracy degrades with increasing measurement noise and decreasing spectral

contrast in a given spectral range. Spectral contrast varies with atmospheric conditions.

• Consistent results for various spectral ranges within bands confirms scales with laser

• laser differs by band even though SHIS detectors share same field stop and aft optics

Slide 33CALCON 2003


Summary and Future Work

Slide 34CALCON 2003


• GIFTS ILS effects due to finite-field-of-view are small and correctable

• Off-axis pixel spectral scale varies predictably from pixel-to-pixel. Renormalization via interpolation in ground processing provides all spectra on a common wavenumber scale.

• Good laser and geometric performance of GIFTS design argues for spectral calibration established on ground and validated/monitored on-orbit using Earth views.

• Analysis of AERI and Scanning-HIS datasets suggests that long term variations in GIFTS spectral calibration can be determined/monitored successfully on-orbit using Earth view spectra. Ensemble 1 standard deviation of the Earth view spectral calibration results are ~1.5ppm (AERI) and ~1.1ppm (S-HIS) using 730-740 cm-1.

Summary

Slide 35CALCON 2003


• Comprehensive error analysis of AERI and Scanning-HIS spectral calibration results.

• Ground-up estimate of uncertainty in calculated spectral scale at GIFTS spectral resolution for various spectral ranges and atmospheric conditions.

• Extend Scanning-HIS analysis to include larger range of observed spectra/profiles.

• Develop practical plan for performing GIFTS on-orbit spectral calibration/monitoring.

On-going/Future Work

Slide 36CALCON 2003


The End. Thank you

Slide 37CALCON 2003


Backup Material

Slide 38CALCON 2003


• Two 128 x 128 pixel IR detector arrays with 4 km footprint size

• One 512 x 512 pixel visible detector array with 1 km footprint size

• Views 512 km x 512 km region with all three arrays in ~10 seconds

• Each 10 second observation period provides 16,384 spectra and retrievals

GIFTS: 2-D Detector Array Technology

~4.26x10-4 cm

~8.51x10-4 cm

~1.06x10-4 cm

~2.66x10-5 cm

SMW:(~1175-2350 cm-1)

LW:(~587-1175 cm-1)

A/DAnalog

Samplingt Normalization,

Flat-fieldingCo-add

Numerical Filter /

Decimation

laser = 1E4/1.064m = ~9398 1/cm

SMW: N LW: 4N

N4N

N4N

NN

N/4 = 4096N/8 = 2048

Interferogram Sampling

Slide 40CALCON 2003


Self-Apodization due to finite detector size

Limited to less than 1%

OPD (cm)

apod

izat

ion

1750 cm-1, i=64,j=64

1750 cm-1, i=1,j=64

750 cm-1, i=64,j=64

750 cm-1, i=1,j=64

750 cm-1, i=1,j=1

Slide 41CALCON 2003


Effect of Uncorrected ILS variations

Rarely larger than 0.05 K without correction

Slide 42CALCON 2003


laser wavenumber = 0.00489 0.021 cm-1, ppm = 0.3 1.3

SHIS spectral calibration, 730-740 1/cm,Nominal algorithm but using mean calculation as reference for all cases.

CALCON 2003 GIFTS On-orbit Spectral Calibration On-orbit Spectral Calibration of the Geosynchronous Imaging Fourier Transform Spectrometer (GIFTS) USU/SDL.

Documents