Landslide Monitoring in Three Gorges Area By Joint Use of Phase Based and Amplitude Based Methods

Xuguo Shi , Lu Zhang, Mingsheng Liao, Timo Balz

LIESMARS, Wuhan University

Outlines

• Introduction

• InSAR Results

• Pixel Offset Tracking Results

• Summary

Part 1: Introduction

Test Site: Three Gorges area (Began to function in 2003)

2003 20052004 2006 2007 2008 2009 2010

Dam

66m

135m

156m

185m

175mReservoir impoundment events along

with dam construction/operation

Geohazards in the Three Gorges area

dominant 67% of total

Adapted from Liu et al., 2009

Difficulties for landslide monitoring in TG

Steep terrain

Dense vegetation cover

Complicated atmospheric condition

Part 2: Datasets and InSAR Results

TerraSAR-X datasets

HS: High resolution Spotlight SM: StripMap

Basic parameters

SM data HS data

Orbit direction Descending Descending

Heading 190.7 189.6

Look angle(°) 24 39

Polarization VV HH

Azimuth spacing(m) 1.96 0.87

Range spacing(m) 0.91 0.45

Temporal Coverage Jul 2008-May 2010 Jan 2009-Apr 2010

Az

Rg

N

Fanjiaping Shuping

Qianjiangping

Kaziwan

Xintan

Lianziya

Locations of major landslides

Mean amplitude of 34 TerrraSAR-X StripMap images

What can InSAR do in landslide monitoring ?

What can InSAR observe?

InSAR is useful for monitoring slow moving landslides!

TerraSAR-X Interferogram (SM mode)

Kaziwan

Fanjiaping Shuping

Qianjiangping

Az

Rg

N

Wangjiawan

Xintan

Lianziya

(20091128 -20091220, Bp=13 m, Bt=22 days)

Fanjiaping landslide motion detected by SBAS

The southern part of Fanjiaping landslide is very active.

Deformation rate can reach 5cm/y.

Time series deformation indicated Fanjiaping was moving during the

whole period.

Part 2: Pixel offset tracking results

1000

Fast moving landslides in Three Gorges

Miao et al., Eng. Geology, 2014

Underestimation on these landslides will happen with InSAR analyses.

How to catch the deformation of the fast-moving landslides using high-

resolution SAR images ?

InSAR Point-like targets offset tracking

Shuping Landslide (PS-InSAR)

Active deformation observed in differential interferogram.

Deformation rate estimated by PS-InSAR was unreasonably

less than 1 cm/year.

Underestimation happened with sparse PS points identified.

Shuping landslide is a south-north oriented slope.

Point-like Targets - Corner Reflectors

14 CRs were identified on Shuping landslide.

4 CRs were installed outside Shuping landslide.

Point-like targets offset tracking (PTOT)

– Based on SAR image matching at subpixel-level accuracy

– Make use of pixels with high amplitude values

– Avoid noisy measurements in vegetated areas

– Without phase unwrapping

– Can measure displacements at centimeter-level accuracy in both line-of-sight direction (LOS) and azimuth directions

* LOS ≈ vertical and east-west directions

* Azimuth ≈ north-south direction

Suitable for measuring large displacements.

=>>> Hu, Wang &Liao, IEEE GRSL, 2014 Wang & Jonsson, IEEE JSTARS, In Press

Workflow of PTOT

PT Detection

Time Series

SAR Images

Master Selection

Co-register

Mean Amplitude Sinc Function

PT Candidates

Offset Tracking

Amplitude Correlation and

PT Reselection

Cull PointsCommon

PTs

Orbital Ramp Estimation

Pixel Offset Extraction from

Slave Images

Time Series

Deformation

-

– Mean amplitude image were used to select PT candidates.

– Common PTs were used in the final time series analysis.

Displacement at Shuping measured by HS data (20090221-20100415)

The displacement in azimuth

and range directions can

reach more than 0.8 meters

and 0.6 meters respectively.

Azimuth (m)

Range (m)

InSAR measurement

significantly underestimated

deformations in the LOS

direction.

Displacement at Shuping measured by SM data (20080721-20100501)

The displacements in azimuth

and range direction can reach

more than 1 meter and 0.8

meters respectively.

Observations coincide with the

conclusions in Wang et al. 2008

that the eastern part of Shuping

landslide is more active.

Moving towards north direction

into the Yangtze River.

Azimuth (m)

Range (m)

Time series analysis

Azimuth

Range

Water level

decline VS

displacement ?

Good agreement

achieved on

CR14 between

HS and SM.

BUT ?

Comparison between HS and SM

CR6-Azimuth Range

CR15-Azimuth Range

Projection

sin sin cos sin cos

cos sin

rg rgN E V

az azN E

D D D d

D D d

DN, DE, DV :displacements in the northing, easting and vertical

directions.

α and θ : heading angle and nominal incidence angle at the target

point.

drg and daz : displacements measured in range and azimuth.

δrg and δaz : observation errors to be minimized.

Is it possible to combine HS and SM measurements ?

AX B

sin sin cos sin cos

cos sin 0

sin sin cos sin cos

cos sin 0

HS HS HS HS HS

HS HS

SM SM SM SM SM

SM SM

A

T

N E VX D D D

THS HS SM SMrg az rg azB d d d d

1( )ˆ T TA AX A B

Typical Design matrix

0.115 0.679 0.725

0.986 0.167 0

0.083 0.436 0.896

0.982 0.187 0

A

North East Vertical

HS SM

Rang measurements by HS data is more sensitive to displacement in

easting direction than that by SM data while the latter is more

sensitive to vertical displacement.

Three dimensional displacement (200902-201004)

Horizontal

Vertical

• InSAR and pixel offset method could be jointly used to more accurately map the landslides in TG.

• It is possible to derive 3D displacement from two descending orbit with different look angles.

Summary