Land subsidence from repeated SAR observations

Dr. Anuphao Aobpaet (anuphao@eoc.gistda.or.th)

Earth Observation Center Geo-Informatics and Space Technology

Development Agency (Public Organization)

GEO2TECDI-SONG Final Symposium,

27 May 2013, Crowne room 21st floor, Crowne Plaza Lumpini Park Hotel

The Radar concept

At light speed, c Object scatters energy

back to radar!

- Much like sound waves, radar waves carry information that echoes from distant objects.

- The time delay of the echo measures the

distance to the object.

- The changes of the message

in the echo determines the object

characteristics.

Transmitter/

Receiver

Source: Paul A. Rosen, JPL-NASA

!!!olleH !!!olleH

Hallo!!! Halo!!

What Radar can tell us?

Transmitted radar signals have known

characteristics.

–Amplitude

–Polarization

–Phase and Time Reference

–Wavelength, or Frequency

A distant object that scatters the radar signal back

toward the receiver alters the amplitude,

polarization, and phase, differently for different

wavelengths.

Comparison of the received signal characteristics to

the transmitted signal allows us to understand the

properties of the object.

This is the principle of active remote sensing.

Schematic showing the orientation of signal launch toward the earth’s surface.

Radar imaging properties

• Radar images are distorted relative to a

planimetric view. Slopes facing toward or

away from the radar appear foreshortened.

Steep slopes are collapsed into a single

range cell called layover and areas occulted

by other areas are said to be shadowed.

• Radar is primarily sensitive to the

structure of objects being imaged whereas

optical images are primarily sensitive to

chemistry.

• The scale of objects relative to the radar

wavelength determines how smooth an

object appears to the radar and how bright or dark it is in the imagery.

Rough

Surfaces UrbanArea

Smooth

Surface Mountains Forest

Foreshortening

Source: Paul A. Rosen, JPL-NASA

What is Radar Interferometry?

Radar interferometry can be broadly defined by use of phase measurements to precisely

measure the relative distance to an object when imaged by synthetic aperture radar from two or more observations separated either in time or space.

– Interferometric phase is simply another means of measuring (relative) distance.

• Phase measurements in interferometric

systems can be made with degree level

accuracy, and with typical radar wavelengths

in 3-80 cm range this corresponds to relative

range measurements having millimeter accuracy.

InSAR time series analysis The Basic Idea…

Date

Goal: Solve for the deformation history that, in a least-squared sense, fits the set

of observations (i.e., interferograms).

A stack of interferograms provides multiple constraints on a given time interval

The StaMPS-MTI approach

InSAR Differential Phase Equation For pixel n in interferogram i:

Hooper et al. 2004

• Select pixels with stable scattering behavior over time

• Only focus on “good” pixels

Processing (PSI)

master slave slave

ROI_PAC - Raw data

- Focusing

- Cropping

DORIS - SLC data

- Coarse orbits

- Coarse correlation

- Fine correlation

- Polynomial computation

- Master timing error

- Master amplitude

- Reference DEM computation

- Interferogram computation

- Compute reference phase

- Subtract reference phase

- Subtract reference DEM

- Geocoded

StaMPS - Amplitude dispersion threshold (PSI=0.4,SB=0.6)

(1-8)

- Phase noise estimation

- PS weeding

- Phase correction for DEM

- Merging patches

- Unwrapping

- Correcting wrapped phase for

Master atmosphere

DEM error

- Unwrapping

- Correction for DEM and master atmosphere

- Filtering and stratification estimation SB processing

- Bias estimation

- Long wavelength tropospheric signal

- Correcting unwrapped phase

Multi-Temporal

Combined

Multiple masters (SB)

•••

•••

Time series analyses

Patching data

PS candidates

PS

points

Unwrapped Interferograms

Read Precise orbits

Coarse coregistration

Fine coregistration

DEMs

Resample

Raw Raw Raw

SLC SLC

SLC

SDFP

candidat

es

Interferogram (SB) ••• Interferogram (PSI)

Processing flow chart (Modified from Bekaert, 2010).

PS Methods

Optimized for pixels dominated

by a single scatterer

Small Baseline

MethodsOptimized for pixels with a

Gaussian distribution of scatterers

Temporal Model

Relying on model of deformation in time (Ferreti et al. 2001)

Spatial

Correlation

Relying on correlation in space: StaMPS (Hooper et al. 2004)

Single-look

small baseline method

Multi-look small

baseline approach

Combined

Method(StaMPS-MTI)

Any InSAR method using

multiple images of the same

area acquired at different times

InSAR time series analysis

InSAR time series analysis

InSAR computational aspects

Hardware 1. HP DL380G7 Server

- 2xCPU E5620 2.4GHz.

- 12GB PC3-10600 RDIMM

- 6x300GB SATA HDD

- DVD-ROM, 2 port Gigabit Network

- APC Smart 3000VA Rack UPS

2. HP Pro 3330 Pro

- Core i5-3.1GHz., 8GB RAM, 1TB HDD

- 24-inch LCD Monitor

- Windows 7 Professional

- Emerson PSA1000BX UPS 3. HP Color LaserJet M451 Printer 4. HP 1910-16G Network Switch

5. Rack 27U 6. Samsung LED 55-inch

Software

1. Ubuntu v.12.04

2. Doris v.4.02

3. StaMPS-MTI v. 3.2.1

4. GAMMA-SAR

5. Map-ready (ASF)

6. PolSARpro V.4.2

7. Nest (Next ESA SAR Toolbox) v.4C-1.1 8. ROI_PAC v.3.0.1

Precision of measurement

Precision of the displacement calculations is an important element in validating PS

data. The most important factors impacting on data quality are:

• Spatial density of the PS (the lower the density, the higher the error bar)

• Quality of the radar targets (signal-to-noise ratio levels)

• Climatic conditions at the time of the acquisitions

• Distance between the measurement point (P) and the reference (P0)

The precision values obtained from many

analyses of data from the ERS, Envisat,

and Radarsat-1 satellites. Typical values of

precision (1 sigma) for a point less than 1

km from the reference point (P0),

considering a multi-year dataset of radar

images.

Source: TRE

More details can be found on Ferretti et al., 2007

Satellite Radar Systems available

Schematic showing how a satellite acquires a strip map of the earth’s surface.

Satellite Radar Systems available now and into the future.

Nominal Resolution Cell Sizes of all Commercial Radar

Satellites. Numbers in brackets, for COSMO-SkyMed

data, refer to the use of 2 satellites of the constellation, operating in tandem.

Bangkok and its vicinity area

- Bangkok and its vicinity area maps with the

footprint of Radarsat-1 SAR imagery used in

this project.

- locations of the past maximum subsidence

area (Ramkamhaeng University, Bang kapi)

- locations of the industrial estate growth in the

eastern part of Bangkok and Samut Prakarn province.

Land Subsidence in Bangkok

The side step of the building separated from the ground. Cracks on the sidewall of the same building.

The irregular sidewalk surface. The collapse on sidewalk and Ladder Bridge.

The strong bumpy of the wayside found several places

in the study area

The strong bumpy of the wayside found several

places in the study area.

Dataset from Radarsat-1 satellite

19 Radarsat-1 images in F1N beam mode

with azimuth resolution 8.9 m. and range resolution 6.0 m.

Time period: October 2005 and March

2010 in ascending orbit (Path-Row: 2-60).

Using Product Generation System (PGS) at

GISTDA earth observation center to

process Single Look Complex (SLC)

product in CEOS format.

DATE Ppb (m) Ptb (days)

Beam: F1N

Incidence Angle: 36.41 -

39.56

Orbit Sense: Ascending

Path/Row: 2/62

23-October-2005 -479 -1008

10-December-2005 -31 -960

10-February-2008 -622 -168

5-March-2008 525 -144

29-March-2008 -768 -120

22-April-2008 774 -96

16-May-2008 -769 -72

3-July-2008 -475 -24

27-July-2008 0 0

31-October-2008 -546 96

18-December-2008 -597 144

11-January-2009 -80 168

4-February-2009 -324 192

2-October-2009 -114 432

19-November-2009 658 480

13-December-2009 -569 504

6-January-2010 -702 528

30-January-2010 125 552

19-March-2010 136 599

A small variation of scatters in an area within distance 100 m

Spatial patterns of Land subsidence

(a)

(b)

(a): Subsidence rate in vertical direction (mm/yr) from InSAR time series analysis.

(b): Standard deviations of mean rates (mm/yr).

Subsidence rate (mm/yr) plotted on the mean radar amplitude

Subsidence rate (mm/yr) plotted on the mean radar amplitude. (a) Bangkok–Nongchok, (b) Samut

Prakarn–Na Klua, (c) Samut Prakarn–Thepharak, (d) Suvarnabhumi Airport,and (e) Bangna–Trad Road.

InSAR vs. leveling subsidence rate validation

Comparison of subsidence rates between

InSAR and leveling with standard deviation

(mm/yr) using B52 as a reference. Negative

values indicate subsidence and ‘-‘ alone

indicates there were no PS within 100 m of

the benchmark.

InSAR against leveling rates ( InSAR and lev

, respectively) were compared using the t-

test (Fisher, 1925),

Using a level of confidence of 95%, 80% of

the benchmarks agree with the estimation

from InSAR.

Aobpaet et al., 2013

InSAR vs. leveling time series validation

Displacement time series obtained from InSAR and leveling for three

benchmarks for which the mean displacement rates are in agreement. Vertical offsets are added for visualization.

Displacement time series obtained from InSAR and leveling for

benchmark B7, for which rates from the two techniques are not in agreement.

Sand layer

Single bounce

Double bounce

Possible ground to wall reflection case

Single and double bounce echoes.

Subsidence rates in the neighborhood of two leveling benchmarks

(a) B27 is located in Thepharak, Samut Prakarn. (b) B7 is located in a rapid growing suburban area, Minburi, Bangkok.

Comparison of interpolated rates

(a)

(b)

leveling InSAR

Land subsidence map of Bangkok from Radarsat-1 data

in the period of 2005-2010 using InSAR time series analysis

Conclusions

Time series InSAR approach that combines both persistent scatterer and small baseline

methods detected approximately 300,000 pixels that can serve as monitoring points.

Deformation rates have been estimated for the period 2005-2010 using 19 SAR images, with a

maximum rate of about 30 mm/yr in south eastern part of Chao Phraya River.

Velocities estimated at a distance of up to 100 m from the benchmarks by InSAR agree with

those estimated from leveling, within the measurement uncertainty, for 80% of the leveling

benchmarks.

The results from InSAR provide a very high observation density with comparable accuracy to

leveling.

InSAR should be considered as a complementary technique to leveling for monitoring land

subsidence in Bangkok.

References

Aobpaet, A., M. C. Cuenca, A. Hoopper & I. Trisirisatayawong, 2013, InSAR time series analysis

for land subsidence in Bangkok, Thailand. International Journal of Remote Sensing, 34:8,

pp.2969-2982, (Accepted 16/09/12). doi10.1080/01431161.2012.756596

Ferretti, A., G. Savio, R. Barzaghi, A. Borghi, S. Musazzi, F. Novali, C. Prati, and F. Rocca, 2007,

Submillimeter accuracy of InSAR time series: Experimental validation, IEEE Trans.

Geoscience Remote Sensing, 45, 1142-1153.

Hooper, A., Zebker, H., Segall, P., and Kampes, B., 2004, A new method for measuring

deformation on volcanoes and other natural terrains using InSAR persistent scatterer. Geophysical Research Letters, 31(L23611).

THANK YOU Geo-Informatics and Space Technology

Development Agency (Public Organization) Department of Survey Engineering,

Chulalongkorn University Royal Thai Survey Department Delft University of Technology GEO2TECDI and GEO2TECDI-SONG projects