Operational AIRS-MODIS Co-location System AIRS Spatial Response - - PowerPoint PPT Presentation

Operational AIRS-MODIS Co-location System

AIRS Spatial Response Function & MODIS IR channel Spectral Response Function

Haibing Sun2, W. Wolf2, C. Barnet1, Lihang Zhou2 and M. Goldberg1

1 NOAA/NESDIS/ORA, 5200 Auth Road, Camp Springs, MD 20746 USA 2 QSS Group Inc, Lanham, MD, USA

AIRS-MODIS Collocation AIRS-MODIS Collocation Processing in ORA Processing in ORA

• Operational Collocation Algorithms: Co-register
• bservations from AIRS and MODIS
• AIRS spatial response function simulation
• Algorithm validation analysis and results
• MODIS IR channel spectral response function .
• Summary

Apply AIRS EFOV Spatial Response Apply AIRS EFOV Spatial Response Function for Collocations Function for Collocations

The AIRS radiance is contributed by all the points within the EFOV of the sensor. In one AIRS EFOV, hundreds of MODIS

• bservations are collocated.

A more realistic AIRS EFOV Spatial Response Function is used to: 1: Select collocated MODIS observations 2: Calculate the weights for each MODIS FOV

• 30.9
• 30.8
• 30.7
• 30.6
• 30.5
• 30.4
• 30.3
• 30.2
• 30.1
• 30

40.1 40.2 40.3 40.4

• 30.9
• 30.8
• 30.7
• 30.6
• 30.5
• 30.4
• 30.3
• 30.2
• 30.1
• 30

40.1 40.2 40.3 40.4

• 30.9
• 30.8
• 30.7
• 30.6
• 30.5
• 30.4
• 30.3
• 30.2
• 30.1
• 30

Collocated MODIS observation 40.1 40.2 40.3 40.4

Simplized AIRS EFOV SRF .Fp=1 Simulated AIRS EFOV SRF Fp=1

AIRS Spatial Response Function AIRS Spatial Response Function Simulation: Physical Model Simulation: Physical Model

• The radiance for each AIRS footprint is a combination of 51 instantaneous

fields of view (IFOV)

• IFOV spatial response function:

– Pre-launch measurements – circle, ellipse, rectangle

• The spatial response function of the AIRS EFOV is the convolution product
• f the spatial response function of each IFOV and the integration time.

Other factors that we need to take into account are: a: Time integration pattern

• b: Scanning pattern / stepwise or continuous
• c: Reflect mirror rotation / whether mirror is used
• d: Satellite movement
• d: Earth rotation
• e: No spherical earth
• AIRS Sampling model: AIRS scan speed:8/3; 119 Sampling; 90Earthview
• 0.022 Integration Period; 51subsampling

AIRS Spatial Response Function AIRS Spatial Response Function Simulation Methodology Simulation Methodology

• Time Integration

–0.022 second sampling period –Simulate sub sampling IFOV during the EFOV integration period –Project all sub sampling IFOV spatial response function on earth –Integrate IFOV spatial response function over EFOV sampling time

• Sub sampling

–51 times –51 sub scan angles –51 sub satellite positions

• Satellite orbit calculation

–along track movement, 51 pointing vectors

• Add Reflection mirror
• Use Earth elevation model (WSR84), non spherical
• AIRS EFOV spatial response function in earth surface coordinates:

–90 spatial response function for 90 scan angles.

AIRS EFOV Spatial Response Function AIRS EFOV Spatial Response Function Nadir View Nadir View

• 0.008
• 0.006
• 0.004
• 0.002

0.002 0.004 0.006 0.008 Along the scan fp=45 Lat=10

• 0.008
• 0.006
• 0.004
• 0.002

0.002 0.004 0.006 0.008 Along the orbit

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50

AIRS EFOV Spatial Response Function AIRS EFOV Spatial Response Function Opposite Edge of Scan Opposite Edge of Scan

• 0.008
• 0.006
• 0.004
• 0.002

0.002 0.004 0.006 0.008 Along the scan fp=1 lat=10

• 0.008
• 0.006
• 0.004
• 0.002

0.002 0.004 0.006 0.008 Along the orbit 5E-005 0.0001 0.00015 0.0002 0.00025 0.0003 0.00035 0.0004 0.00045 0.0005 0.00055 0.0006 0.00065 0.0007 0.00075 0.0008 0.00085 0.0009

10 20 30 40 50 60 70 80 90 100 10 20 30 40 50

Algorithm Validation Results and Analysis

• Validation Method
• AIRS SRF application: AIRS pointing correction
• The problem in validation: MODIS Spectral Response Function

MODIS Spectral Response Function Retrieval

• Algorithm Validation with real Data:
• Ocean Case Land Case
• Polar Case Coastline Case Desert Coastline Case

AIRS EFOV Spatial Response Function AIRS EFOV Spatial Response Function Application Application

MODIS Radiance No simulated AIRS EFOV SRF

2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5

Weight in Degrading Simple AIRS SRF

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9

175.9 175.95 176 176.05 176.1 176.15 176.2 176.25 176.3

MODIS radiance Simulated AIRS EFOV SRF

25.84

175.9 175.95 176 176.05 176.1 176.15 176.2 176.25 176.3

Weight in Degrading Simulated AIRS EFOV SRF

25.75 25.8 25.85 25.9

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Operational AIRS-MODIS Colocation processing With Simulated AIRD EFOV SRF AIRS-MODIS Colocation processing with simplized AIRS SRF

Simple SRF Realistic SRF

AIRS EFOV Spatial Response Function AIRS EFOV Spatial Response Function Application: Pointing Bias Correction Application: Pointing Bias Correction

10 20 30 40 50 60 70 80 90 Pointing direction bias from centroid: In scan direction In orbit direction 0.00E+0 2.00E-4 4.00E-4 6.00E-4 8.00E-4 1.00E-3 Bias setting: x: 30 mili degree y: 100mili degree

Low Low Latitude Region (Land) : 4/16/119 Ascending

Latitude Region (Land) : 4/16/119 Ascending

Radiances Simple SRF Realistic SRF

Coastline: 04/17/143 Coastline: 04/17/143

• 20
• 15
• 10
• 5
• 2

2 4 6 8 10 12 14 16 18 20

10 20 30 40 50 60 70 80 90 100 110 120 130 140

• 20
• 15
• 10
• 5

With Simulated AIRS SRF

• 2

2 4 6 8 10 12 14 16 18 20

Radiance Simple SRF Realistic SRF

Desert Coastline: 04/17/111 Desert Coastline: 04/17/111

Realistic SRF Simple SRF Radiance

MODIS Response Function Retrieval

• Part II MODIS IR channel Spectral Response Function

MODIS Relative Spectral Response Function MODIS Relative Spectral Response Function (RSRF) Problem (RSRF) Problem

For the MODIS infrared channel 20-36, there is no way to monitor the instrument spectral sensitivity. Recent researching about the MODIS

• bservation data show that there may be different instrument spectral

response function shifting in these 16 infrared channels. For AIRS-MODIS co-location processing, an observation data based algorithm is developed to retrieve on-orbit ‘Broad band’ MODIS spectral response function with co-located AIRS high spectral resolution observation data. The retrievaled result provide the information of the ‘real’ MODIS RSRF on the observation system level. Those information can be used to improve the integration quality of different instrument and provide inter-instrument calibrilication ability.

Basic Methodology

• Basic equation: Individual MODIS IR channel

W Rad Rad

i i AIRS MODIS

• =
• To resolve the Wi , A equation group is need:

M X A

i j ij

=

• Rad(MODIS): Low spectrum resolution MODIS observation
• Rad(AIRS)i High spectrum resolution AIRS observation
• Wi Weighting defined by spectral response function
• A(I,j): AIRS observation matrix
• X(j) Weighting vector
• M(ij MODIS observation vector
• I > J is required
• Key Assumption: All no-linear behavior is in-depend of wave length.

AIRS Inter-channel Sampling

700 705 710 715 720 725 730 735 740

MODIS channel 35 Spectrum

40 60 80 100 120

AIRS Radiance

0.00 0.20 0.40 0.60 0.80 1.00

MODIS IR Channel(35) RSF

MODIS Channel 35 Spectrum Response Function SPEC35

Deficient Matrix inverse Problem

• The X can be solved by multiplying both sides by the

pseudo-inverse of A. But A is invariably rank deficient.

• Physically, X is continues with un-limited dimensionality.

The A and M are tend to be low dimensionality.

• With rank deficient A, M, Retrieval of X based on matrix

inversion are very sensitive to noise

Retrieval with Simulated Data

10 20 30 40 50 60 70 80 90 100 110 120

MODIS CHANNEL 36( AIRS channel)

0.000 0.200 0.400 0.600 0.800 1.000

Simulated Data with zero noise to valid algorithms

Retrievaled with U=0.01 AVE:1 SVD:111 Pre-launch Measurement SRF

Noise Free Data

Retrieval with Simulated Data

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95100 105 110 115

Channel 36 (AIRS ch Point index)

0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018

Retrieval with Simulted AIRS -MODIS Data with Addtional Noise: 0.02*MODIS(ch)*[-1,1] U = 0.00 SVD:111

• 0.80
• 0.40

0.00 0.40 0.80 1.20

Retrieval with Real Observated AIRS -MODIS Data U = 0.00 SVD:111

Data with Noise NOISE = [-0.5,0.5]

Improved Methodology: Retrieval with deficient Matrix

Data set Optimization:

Co-located MODIS–AIRS Date/ Uniform scene. Apply the AIRS spatial response function Differential equation: Remove system bias. Average Equation to reduce the Observation white noise

Deficient Matrix Equation Optimization:

Regulate Matrix A with Truncated Singular decomposition: Noise Reduction: Truncate small singular values

Constraint Retrieval algorithm : Constraint

Retrieval/Selection. Retrieval average

Simultaneous optimization Retrieval with Physical constraining

Constraints: 1. RSF positivity 2. RSF smoothness 3. RSF Bounded prediction

• 4. Modality
• 5. RSF rank constraint
• 6. RSF border limitation
• X j

T X X

j

j

• +
• |

|

1

• =

=

• |

|

1

X A M

j l j j ij

i

min 1

|| ||

2

=> +

• =

=

• =

X D X X A M

T i

j l j j ij f

• =
• x

f

Smoothness Weighting factor:

• 1

. . . . 1 2 1 1

D

Retrieval Algorithms Numerical Test

Purpose: Validate the Retrieval algorithms Validation Method: Retrieval with simulated AIRS /MODIS Data set to test whether This algorithms can obtain useful retrieval result when observation noise exist. Data set: AIRS Data: AIRS Observation data MODIS Data: AIRS Convolution product With pre-launch measurement MODIS RSRF Noise Model: Free noise/System Bias/White Noise / Observation related noise F noise Model NOISE = -1.242+ [-1,1]+ 0.01*Rad(MODIS)+ 0.01*Rad(MODIS)*[-1,1]

10 20 30 40 50 60 70 80 90 100 110 120 0.000 0.004 0.008 0.012 0.016 0.020

Simulated Data with zero noise to valid algorithms U=0.0 AVE:1 SVD:113 U=0.01 AVE:1 SVD:113 U=0.1 AVE:1 SVD:113 U=0.5 AVE:1 SVD:113

Constrain retrieval with Free noise Data

The smooth constrain will limit the high Frequency Component in retrieval result. ‘Over’ constrain will lead to the ‘under estimate of the high frequency component and thus introduce retrieval Bias between retrieval result and the ‘Truth’.

TSVD &Retrieval Bias

10 20 30 40 50 60 70 80 90 100 110 120

MODIS CHANNEL 36 (AIRS channel)

0.000 0.004 0.008 0.012 0.016 0.020

Simulated Data with zero noise to valid algorithms

U=0.01 AVE:1 SVD:8 U=0.01 AVE:1 SVD:24 U=0.01 AVE:1 SVD:12 U=0.01 AVE:1 SVD:111 Pre-launch Measurement

U=0.01 TSVD Truncated small Singular factor to reduce the

Equation Noise sensitivity, It will also lead to the retrieval Result to lost the high frequency component and introduce Retrieval Bias.

TSVD &Retrieval Bias

10 20 30 40 50 60 70 80 90 100 110 120 0.000 0.004 0.008 0.012 0.016 0.020

Simulated Data with zero noise to valid algorithms Pre-launch Measurement U=0.05 AVE:1 SVD:6 U=0.05 AVE:1 SVD:111 U=0.05 AVE:1 SVD:56 U=0.05 AVE:1 SVD:28 u=0.05 AVD:1 SVD:14

U-0.05

TSVD & Constraint&Retrieval Bias

10 20 30 40 50 60 70 80 90 100 110 120

Channel 35(AIRS channel)

0.000 0.004 0.008 0.012 0.016 0.020

SRF

Simulated Data with zero noise to valid algorithms U=0.01 AVE:1 SVD:111 U=0.001 AVE:1 SVD:8 U=0.01 AVE:1 SVD:8 U=0.05 AVE:1 SVD:8 SVD:8 U=0.001 0.02 0.05 SVD:8 will limit the prescsion

Algorithms Validation Simulated MODIS observation with noise

685 690 695 700 705 710 715 720

Channel 36 Frequency (CM-1)

• 0.03
• 0.02
• 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06

MODIS Spectral Response function(CH36)

Simulated Data with additional noise: 0.02* MODIS* Normal[-1,1]

U=0.01 U=0.04 U= 0.08 U=0.1 U=0.2 Pre-launch measurement

RSF smoothness constraint

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Channel 36 Frequency : Ave:20 SVD:10

• 0.07
• 0.06
• 0.05
• 0.04
• 0.03
• 0.02
• 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

64 Point DCT of MODIS Spectral Response Function

PS: Different between simultaneous optimization and low pass fileter

Algorithms Validation with Simulated Data

10 20 30 40 50 60 70 80 90 100 110

Channel 36

0.000 0.004 0.008 0.012 0.016 0.020

MODIS Spectral Response Function (CH36)

Retrieval with noise free simulated data: U=0.01 SVD:8 Retrieval with simulated data with noise : U=0.01 SVD:8

Algorithms Validation with Simulated Data

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100105110115

Channel 36

• 0.00

0.00 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.02 0.02

MODIS Spectral Response Function (CH36)

Simulated Data Testing Retrievaled RSF with Noise free simulated MODIS Data with U=0.05 SVD:111 Modis = AIRS x RSF RSF: Pre-launch Measurement of Spectral Response Function Retrived Spectral Function with Simulated Data Addition MODIS Noise is Added to Modis Observation Single Resolution with Positive/Border/Peak Constrain F:C16.REAL.UTEST.SA20U.GRF

700 705 710 715 720 725 730 735 740

Channel 35 MODIS Frequency

0.0 0.2 0.4 0.6 0.8 1.0

MODIS RSF (Channel 35)

u=0.005 It: 18 Pre-launched

Noise= -1.242+ System Bias [-1,1] + Random Noise 0.01*Rad(MODIS)

Algorithms Validation with Simulated Data

Retrieval Result Analysis

Constraint Retrieval with Observation Data

685 690 695 700 705 710 715 720

Channel 36 Frequency (CM -1) A10SAU01

• 0.02
• 0.01

0.00 0.01 0.02 0.03 0.04

MODIS Spectral Response Function (CH36)

Constaint retrieval with Obsercation Data: Ave:10 SVD:10

U=0.01 U=0.02 U=0.04 U= 0.06 U=0.08 U=0.1 U=0.2 U=0.75 U=0.5 U=0.3 Pre-launch Measurement

MODIS channel 35 RSRF Retrieval

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100105110

Channel 36 Frequency

0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018

MODIS Spectral Reponse Function (CH36)

U=1 AVE:1 SVD:10 U=10 AVE:1 SVD:10 U=3.5 AVE:1 SVD:10 U=4.0 AVE:1 SVD:10 U=0.5 AVE:1 SVD:10 Single solution: choose with Positive/Border/Peak constraint F:C16.REAL.UTEST.RA01UA.GRF

MODIS channel 35 RSRF Retrieval

700 705 710 715 720 725 730 735 740

Channel 35 MODIS Frequency

0.0 0.2 0.4 0.6 0.8 1.0

MODIS RSF (Channel 35)

u=0.1 u=0.02 Pre-launched u=0.01

Precise result and constrain degree

10 20 30 40 50 60 70 80 90 100 110 120

Channel 36 Modis( AIRS channel)

0.000 0.200 0.400 0.600 0.800 1.000

Retrieval with constraint Comparing with simulated noise free data and real data U=0.05 AVE:1 SVD:6 Simulated Noise free Data U=0.05 AVE:1 SVD:6 Real data U=0.05 AVE:1 SVD:6 1: Frequency shift: about 6 channel number 2: Reshape in two side Strong constrain result

1:Some information we can get with this retrieval Shifting and reshape. 2: Strong constraint will lead to detail structure of the RSF be to smoothed.When the noise is compressed, the high frequency component useful information is also depressed. 3: The precision and the compression of the noise,that is a comprise.

Biased Constraint retrieval

10 20 30 40 50 60 70 80 90 100 110 120

Channel 36 Modis( AIRS channel)

0.000 0.200 0.400 0.600 0.800 1.000

Retrieval with constraint Comparing with simulated noise free data and real data U=0.05 AVE:1 SVD:6 U=0.05 AVE:1 SVD:6 Real data U=0.05 AVE:1 SVD:8 SVD: 6==> 8 With Less constraint, More 'high frequenvy' component can be kept.

Noise sensitivity & Constraint bias

The bias introduced by constraint condition depend on: The SRF pattern. If In SRF, the high frequency component is small, the bias introduced by constrain is small. The retrieval can be used as SRF. Constraint strength. Weak constraint mean strong noise sensitivity. Solution: 1: Limit the noise in data will reduce the requirement of the constraint. 2: High frequency component compensation.

Constrainting bias correction

10 20 30 40 50 60 70 80 90 100 110 120

Channel 36(AIRS channel)

0.000 0.004 0.008 0.012 0.016 0.020

Spectral Response function

Pre-launch Measurement Retrievaled and add constraint correction Retrievaled with constrain condition

Simplified correction model basing on the SRF Data and assumption that slight shape change. If the shape change is great, the correct will introduced bias

Evaluation for channel 36

44 46 48 50 52 54 56 58

MODIS CHANNEL 36 Radiance bias between Observation and convolution RSF

0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40

Bias (Radiance)

BIAS = Observation - AIRS X RSF RSF: Pre-launched Measurement Retrievaled RSF

Summary

• AIRS spatial response function can be used to improve the

collocation accuracy.

• AIRS pointing bias and MODIS spectral change make the

collocation problem more complex.

• Constraint retrieval algorithms can be applied to obtain MODIS RSF

information with AIRS high spectral resolution information. This application reveal that the inter-instrument Calibralition is possible and useful.

• The multi-sensor remote sensing data integration and inter-

calibration will beneficial to retrieval and calibration both.