Reducing Ionospheric DecorrelationEffects in InSAR Data Using Accurate

Coregistration

Albert C. ChenHoward A. Zebker

Stanford UniversityElectrical Engineering Department

IGARSS, 2010, Honolulu, HIFR3.L09.2

Overview

1. Background: Ionospheric effects in InSAR

2. Empirical Observations: Azimuth offsets

3. Results: improved coherence and phasecorrection

4. Conclusions

Ionosphere in ALOS Imaging

Earth surface

Altitude (km)

~ 2 Lightning

~ 10 Commercial aircraft, airborne SAR

100 E-layer (due to X-ray and UV ionization)

300 F-layer (due to UV ionization)

700 ALOS satellite orbit altitude

Ionosphere

384,000 moon

plasma, magnetic fieldaffect wave propagation

95% of atmospheric watervapor is below 5 km

Theory: Ionospheric Phase

2122,1,23.404TECTEC

f

SARiono

SARiono

InSARiono

3

2

2

22 cos1

fff

ff

n Bpp Dispersive

Birefringent

e

epp m

enf0

2

21

2

e

BB m

eBf

21

2

Theory: Ionospheric Azimuth Offset

Albert C. Chen 5

point target

synthetic aperture length

…TEC vs. azimuth

23.400

2

21 RxfTECTEC

x

Background: Ionospheric effects inInSAR data

K.E. Mattar and A.L. Gray“Reducing ionospheric electron density error in satellite

radar interferometry applications”Can. J. Remote Sensing

Vol. 28, No. 4, pp. 593-600, Aug. 2002.

ERS 1/ERS 2 Tandem InterferogramMarch 18-19, 1996Northern Canada

C-band interferogram

Background: Ionospheric effects inInSAR data

U. Wegmuller, C. Werner, T. Strozzi, A. Wiesmann“Ionospheric electron concentration effects on SAR

and InSAR”IGARSS 2006

JERS Repeat-pass InterferogramMarch 23 – May 6, 1994

Svalbard, Norway?

L-band interferogram

Central Greenland Dataset

Instrument ALOS-PALSAR (ascending orbit)

Acquisition Dates 09 Mar 2007 – 24 Apr 2007

Temporal baseline 46 days

Perpendicular baseline 155 m

Radar wavelength 23.61 cm

Scene center 76.887° N, -34.772° E

Source: Google Earth

Scene Location Radar Brightness Image

Iceland Dataset

Source: Google Earth

Instrument ALOS-PALSAR (ascending orbit)

Acquisition Dates 02 Sept 2007 – 18 Oct 2007

Temporal baseline 46 days

Perpendicular baseline 511 m

Radar wavelength 23.61 cm

Scene center 64.657° N, -18.573° E

Scene Location

Radar Brightness Image

Observation: Azimuth StreaksGreenland

Interferogram Coherence

Observation: Azimuth StreaksIceland

Interferogram Coherence

Empirical Analysis: Azimuth OffsetsGreenland

Interferogram Coherence Azimuth offset

Empirical Analysis: Azimuth OffsetsIceland

Interferogram Coherence Azimuth offset

-4 0 40 0.5 1-π 0 π

Empirical Analysis:Coherence vs. Misregistration

coregnoisethermaltemporalspatial

C

else,0

1/0),/sinc( coreg

Best fit model:C = 0.693δ = 1.65

Theoretical δ (pixels/resel): 1.4

“background coherence”

Results: Improved Interferogramlow-order

polynomial offsetinterferogram

accuratecoregistrationinterferogram

accuratecoregistration

coherence

low-orderpolynomial offset

coherence

Greenland

Results: Improved Coherence

Results: Residual Decorrelation

“background coherence”

azimuthal profile

Results: Ionospheric Phase ScreenGreenland

Near RangeIonospheric Phase

Results: Ionospheric Phase ScreenGreenland

Original Interferogram Corrected Interferogram

Iceland: Smoothed offsetsEstimated Offsets Outliers Removed Smoothed Offset

Iceland

-4 0 4

Results: Improved Interferogram

Results: Improved CoherenceIceland

Conclusions

1. Spatially variable TEC causes spatially variableazimuth offsets

2. Accurate coregistration method reduces ionosphericdecorrelation

3. Ionospheric phase can be estimated to first-order, andremoved from interferograms

4. Accurate InSAR measurements of ice velocity, crustaldeformation, etc. requires compensation ofionospheric effects.

5. Ionospheric correction methods needed and feasiblefor L-band InSAR (ALOS, JERS, upcoming DESDynI,etc.)

Acknowledgments

This work is supported in part by the Reed-HodgsonStanford Graduate Fellowship.

We would like to thank the following for their helpful suggestions:

Mark SimonsSigrid Close

Piyush ShankerJingyi Chen

Tom Rune Lauknes

Method: Offset Measurement

By parallelizing our code using OpenMP, wecan calculate a “dense” offset field without

sacrificing speed

Method: Offset Smoothing

Iceland Offset Estimate / Coherence Scatter Plot

Accept estimates where:

• Coherence > 0.68

• Azimuth offset < 4 pixels

• Range offset < 2 pixels

Use Delaunay Triangulation tofill in gaps.

Method: Resampling

Sinc Resampler

),(),(

11

11

RxxRxR

),( 111 Rxs

),( 112 Rxs

BilinearInterpolation

*21ss