Inter-calibration of OperationalIR Sounders using CLARREO

Bob Holz, Dave Tobin, Fred Nagle, Bob Knuteson, Fred Best, Hank Revercomb

Space Science and Engineering Center, University of Wisconsin-Madison

2009 CLARREO WorkshopLangley, VA

May 11-15, 2009

Question:Given a CLARREO mission optimized for producing the climate benchmark products (e.g. consistent with the DS mission consisting of three 90 degree polar orbits, 100 km footprints, 0.1K radiometric accuracy, broad and continuous spectral coverage at high spectral resolution), how well can we meet the secondary objective of CLARREO to serve as an inter-calibration reference for the operational IR sounders ?

Answer: Yes, the CLARREO calibration can be transferred to operational IR sounders with an uncertainty comparable to the CLARREO radiometric accuracy.

Approach:A simulation study using real MODIS data.

Same basic approach as presented at the first CLARREO workshop, the CLARREO science team meeting, and recent AGU and SPIE meetings.

Find Simultaneous Nadir Overpasses (SNOs) of CLARREO and EOS Aqua, and for each SNO use MODIS radiances to estimate the spatial and temporal sampling differences between CLARREO and CrIS/AIRS or IASI.

Opposed to actual inter-comparisons studies involving two sensors, this approach removes the unknown sensor biases and allows spatial and temporal inter-calibration differences to be examined.

CLARREO/Aqua SNOs in 2006Three 90-degree CLARREO orbits with right ascension separated by 120 degrees are “launched” on January 1st, and the SNOs for CLARREO and EOS Aqua are identified for the year of 2006.

Estimating brightness temperature differences due to spatial and temporal

sampling differences

Spatial: Compute mean of 1km MODIS observations within the CLARREO footprints and also within the CrIS/AIRS or IASI footprints. (Also record the standard deviation of the MODIS observations within the CLARREO footprints.)

Temporal: simple Lagrangian approach; compute the mean of MODIS observations within a displaced CLARREO footprint using known CLARREO/Aqua time difference and 13 m/s wind.

Spatial Sampling Differences

CrIS IASI

100 km CLARREO

100km

~55% coverage ~18% coverage

11 m

MO

DIS

BT

(K)

11 m

MO

DIS

BT

(K)

11 m

MO

DIS

BT

(K)

Time Sampling Differences1

1 m

MO

DIS

BT

(K)Displaced

FOV

• We use all sky cases for inter-calibration• The approach presented is consistent with the

GSICS ATBD (in development), and as has been practiced for Geo/Geo and Geo/Leo intercal for many years.

GSICS ATBD Uses All Sky Inter-calibration

Uncertainty in Monthly Means MODIS Band 31@11m; 100km CLARREO FOVs every 14s; CrIS/AIRS

Uncertainty including Space/Time errors onlyUncertainty including Space/Time errors and CLARREO (1K NEDT) and Sounder single channel NoiseMean # points per month

Conclusions: - CLARREO will be capable of inter-calibration uncertainties of less then 0.1 K 3σ- Inter-calibration uncertainties are improved when non-uniform FOV are included

Threshold method and Weighted means MODIS Band 31@11m; 100km CLARREO FOVs every 14s; CrIS/AIRS

168 pts 464

696

866

976

1163

Conclusion: A CLARREO single channel noise of less then 1.0 K will allow for a monthly inter-calibration accuracy of 0.1 3σ

FOV Size Dependence on Monthly Uncertainties

100 km, 50 km, and 25 km diameter footprintsMODIS Band 31@11m; CrIS/AIRS

100km (14s), 1276 pts, STDEV = 0.214 K50 km (8s), 2286 pts, STDEV = 0.460 K25 km (4s), 4474 pts, STDEV = 0.918 K

Solid = threshold method, BTSTD ≤ 10KDashed = uncertainty in weighted mean

Conclusion: The inter-calibration goal of 0.1 3σ can be accomplished for a range of CLARREO FOV sizes including 100 km.

“Nonlinearity Curve” from Annual data MODIS Band 31@11m; 100km CLARREO FOVs every 14s; CrIS/AIRS

10K bin Uncertainties

Blue = space/time only

Cyan = plus CLARREO (1K NEDT) and Sounder Noise

Yellow = plus CLARREO (0.5K NEDT) and Sounder Noise

Red = CLARREO (1K NEDT) and Sounder Noise, averaging 4 adjacent spectral channels

Conclusion: Instrument nonlinearities can be investigated using yearly averages for single channels or monthly using spectral averaging.

Aircraft (S-HIS)

Sounder (AIRS/IASI/CrIS)

CLARREO

Aircraft Inter-Calibration

0.5 3σ 0.1 3σ

Conclusion: Aircraft to Sounder inter-calibration is the limiting link in evaluating CLARREO using Aircraft

This uncertainty limits the evaluation of CLARREO

(K)

21 November 2002, Gulf of Mexico

Aircraft is keyapproach for direct radiancevalidation of EOS & NPOESS

FantasticAgreement,but 3-sigmauncertainty in validation is at least 0.5 K**

**Contributions from Sampling, Representativeness, Noise, Double differences, as well as S-HIS Accuracy

(70% chance error <0.16 K)

Aircraft Inter-Calibration

Non-uniform Spatial Response

Uniform Linear, ±10% Linear, ±25%

Cone ~1/r2 (~4km FWHM)

Uncertainty in Monthly Means Uniform, Linear 20%, Linear 50%, ~1/r2

MODIS Band 31@11m; 100km CLARREO FOVs every 14s; CrIS/AIRS

Tophat, STDEV = 0.214 K20% linear: STDEV = 0.220 K50% linear: STDEV = 0.290 K~1/r2: STDEV = 0.803 K

Solid = threshold method, BTSTD ≤ 10KDashed = uncertainty in weighted mean

Conclusion: Non-uniform spatial response has minimal impact on inter-calibration uncertainty.

Summary, ConclusionsFor a nominal CLARREO DS mission and reasonable SNO thresholds (three 90 deg polar orbits, 100km nadir footprints every 14s, 10 deg max sounder scan angle, 15 min max time difference), the monthly intercal uncertainty budget is dominated by the assumed CLARREO radiometric noise level.

With CLARREO single channel radiometric noise of 1K NEDT, the 1-sigma uncertainty in monthly intercal is less than 0.03K. Lower uncertainties are obtained with smaller (and more numerous) footprints and/or lower radiometric noise.

Non-uniform spatial response has negligible impact on these results.

For annual ensembles, a wide range of intercal BTs and low uncertainties in single channel 10K bins are obtained. When multiple channels are averaged, monthly 10k bins can be used.

Need to remember spectral considerations for intercal objectives.

1 ≤ STDEV ≤ 2 (5074 pts, stdev=0.08)3 ≤ STDEV ≤ 4 (2783 pts, stdev=0.18)5 ≤ STDEV ≤ 6 (1179 pts, stdev=0.27)7 ≤ STDEV ≤ 8 ( 563 pts, stdev=0.34)9 ≤ STDEV ≤ 10 ( 211 pts, stdev=0.47)

Spatial Sampling Differences MODIS Band 31@11m; 100km CLARREO FOVs every 14s; CrIS/AIRS

Conclusion: The spatial sampling differences are Gaussian when separated by STDEV