Content uploaded by Hengxing Lan
Author content
All content in this area was uploaded by Hengxing Lan on Mar 17, 2015
Content may be subject to copyright.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMO TE SENSING, VOL. 5, NO. 2, APRI
L 2012 643
Complex Urban Infrastructure Deformation
Monitoring Using High Resolution PSI
Hengxing Lan, Langping Li, Hongjiang Liu, and Zhihua Yang
Abstract—The rising concern on
the urban vulnerability to the
intensive infrastructure development requires enabling technolo-
gies offering a prompt and accurate monitoring of urban infra-
structures deformation. Urb
an infrastructures vary dramatically
both spatially and temporally and their deformation characteris-
tics arecomplex. We evaluated the potentialofhighresolution Per-
sistent Scatterer Inte
rferometry (PSI) technology using coherent
stacks of Spotlight mode TerraSAR-X images in monitoring the
deformations of different types of infrastructures in a new eco-
nomics center in Chin
a. The high density of Persistent Scatterers
(PSs) was identified and therefore facilitates analyzing the defor-
mationcharacterofindividualstructures.AllPSsweercategorized
by identifying t
heir corresponding ground object so as to enable
to characterize deformation pattern of certain type of urban in-
frastructure. The spatial and temporal varying patterns of the de-
formations of
typical building infrastructures and transportation
infrastructures are revealed. They are strongly related to the in-
teractive effects between the typ es, engineering structures, g eome-
tries,engi
neeringgeological s ettings and various loading scenarios.
Besides subsidence of the ground surface, thermal dilation of the
infrastructure itself might be another factor accounting for the ob-
served d
eformation of infrastructure. Although the interpretation
for the observed deformation patterns could be quite site-specific,
high resolution PSI is shown to have the p otential to revealdetailed
defor
mation characteristics of complex urban infrastructures at a
relatively large scale.
Index Terms—Deformation monitoring, persistent scatterer in-
terferometry (PSI), TerraSAR-X, urban Infrastructure.
I. INTRODUCTION
R
APIDLY swelling urban areas present a complex set of
infras
tructure. It isc om posedo f div erse types of elem ents
arranged by humans in complex ways to residential and com-
mercial buildings, transportation systems and utiliti es [ 1] . The
recen
timpressive engineeringconstructionin many urbanareas,
particularly i n China posts major potential vulnerability to city
subsidence [2] as the dewatering of groundwater was controlled
stri
ctly in most of the urban areas to minimize its effect on
ground deformation [3]. Therefore, m onitoring the d eformation
of complex urban infrastructures and their impact on the ground
Manuscript received August 25, 2011; revised October 09, 2011 and De-
cember 12, 2011; accepted December 13, 2011. Date o f publication January 31,
2012; date of current version May 23, 2012. This work was supported in part by
National Science Foundation of China (41072241), National Key Technology
R&D Program (No. 2008BAK50B05).
TheauthorsarewithStateKeyLaboratory of Resources and Environ-
mental Information System (LREIS), Institute of Geographic Sciences and
Natural Resources Research, Chinese Academy of Sciences, Beijing 100101,
China (e-mail: lanhx@lre is.ac.cn; lanhx@igsnrr.ac.c n; lilp@lreis.ac.cn;
hjliu@lreis.ac.cn; yang zh@lreis.ac.cn).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Dig
ital Object Identifi er 10.1109/JSTARS.2011.2181490
deformation is becoming a major
concern for the well-being of
the urban development.
However, there exist big challenges since types of urb an
infrastructure v a ry dramat
ically in both spatial and temporal
pattern. Their deformation characteristics are affected by var-
ious factors such as 3D shape, engineering str ucture, fo undation
type and geological configu
ration. Monitoring such complex
infrastructures in a large regional scale r equires enabling
technologies. The coverage of trad itional ground based survey
techniques, such as leve
ling, total station surveys and GPS,
is limited. Such point-by-point in-situ measurement is hard
to ge nerate high den se survey network required by regional
infrastructur e moni
toring [1], [4].
The advanced Persistent Scatterer Interferometry (PSI) tech-
nique is a multi-temporal radar-b ased remote-sensing techniq ue
to measure a nd monito
r surface deformation. Since the idea of
PSI was firstly introduced as PSInSAR [5], [6], various other
PSI t echniques have been developed, such as Small Baseline
Subset Approach (
SBAS) [7], Coherent Pixels Technique
(CPT) [8], Interfe rometric Point Target Analysis (IPTA ) [ 9],
Stanford Method for Persistent Scatterers (StaMPS) [10],
Spatio-Tempora
l Unwrapping Network algorithm (STUN)
[11], Stable Point Netwo rk (SPN) [12] and Coherent Target
Monitoring (CTM) [13]. PSI technique performs interfero-
metric phase a
nalysis only on tho se permanent scatterers (PSs)
with stable backscattering character to overcome problems
including baseline d ecorrelation, temporal decorrelation and
atmospheri
c di stur bance that tr adit ion al InSAR techni ques
might have difficulties to deal with. PSI shows powerful capa-
bilities of precise deformatio n monitoring motions at a level
of mm/year
by simultaneously processing all the available
SAR images acquired during repeated satellite passes [14]. A
test in Las Vegas shows that ERS and TerraSAR-X d ata allow
detectin
gupto450
and 30,375 respectively
[15] that conventional groun d based approaches can hardly
achieve. P SI has been p roved to be one of the most advan ced
techni
ques for urban deformation mapping [2], [14], [16]–[25].
For mapping urban infrastructure characteristics, using PSI
technique is particularly of inter ests for discrim inating very
small s
cale features. Generally, the spatial scale for urban
infrastructures (e.g., road and bridge) may have the magnitu de
as low as several meters that can be hardly discriminated by
SAR im
ages w ith spatial resolution more than 10 meters (e.g.,
ERS-1/2, ENVISAT and JERS-1). The high density o f persis-
tent scatterers (PSs) extracted by 1-m resolution TerraSAR-X
imag
e (i.e., tens of PSs for individual building) allows inve s-
tigating the structural deformati on of individual infrastructure
[22], [23], [25]–[27], and permits 3D and 4D reconstructions
of i
ndividual bu ild ing s [26], [28]. In addition, com pared to
1939-1404/$31.00 © 2012 IEEE
644 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, A
PRIL 2012
TABLE I
TerraSAR-X SAR I
MAGES USEDINTHESTUDY AREA
the first generation of satellite SAR sensors (e.g., ERS-1/2
and RADARSAT-1) launched b efo re 2006, the new generation
of TerraSAR-X (X-band) Persistent Scatterer I nterferom etry
performs much better in resolving small scale geometric ambi-
guities due to lay-over, shadow, and multiple scattering effect
[14], [26]. It benefits su bstantially f rom its X-band (
3cm)
acquisition, different concurrent polarizat ions, high geometric
and r adiometric resolution and high revisit frequency. The
dramatic increase of PS density allows us to achieve a dense
samplingofasingleinfrastructure(e.g.,abuilding),which
opens promising application perspectives for complex urban
infrastructure monitoring [22], [27].
The objective of this paper is to examine how high resolu-
tion TerraSAR-X PSI may be of v alue for discriminating a wide
variety of urban infrastructure categories and mapping their de-
formation characteristics. The correspon din g urban infrastruc-
ture type of each PS point is identified fir stly. Then, detailed de-
formation patterns are investigated in respect of different types
of urban i nfrastructures. T h e spatial and tempo ral deformation
characters of building and transporting infrastructures are pre-
liminarily presented and discussed to show the potential of high
resolution TerraSAR-X PSI to monitor the deformation of in-
frastructures in complex urban environ men ts.
II. S
TUDY AREA AND DATA
A. Study Area
The study area Tanggu District is located in the southeast of
Tianjin (Fig. 1). It is the heartland of the “Tianjin Binhai N ew
Area”, the rising new economics center in China. As a center of
North China’s coastal area, Tanggu District has been becoming
one of the most important import/expo rt gates f or Central North
China and Northwest China. I t has been suffering from sub-
sidence since its foundation. Latest survey has shown that the
average annual settlem ent is nearly 20 mm with a maximum
value of 68 mm. Since t he dewatering of underground water
has been strictly controlled with an aim at “zero underground
exploitation” in 2014, the primary urban subsidence is there-
fore attributed to the disturbance of massive engineering in-
Fig.1. Tanggu studyareawiththe coverageof TerraSA R-X imageindicated by
a red frame. Insert map in the top-right indicates the location of Tianjin in Chin a.
Insert photo in the bottom-left shows a future view of the New International
Financial Center in Tanggu. An ongoing construction to build a 300 m high-rise
building is also presented.
frastructures. Now adays, tremendous engineering co nstructions
are taking place in Tanggu with an effort to build an Interna-
tional Financial Center (Fig. 1).
B. Data Sources
TerraSAR-X provides high resolution and short wavelength
SAR imagery at a repeat cycle of eleven days [ 29]. The most
interesting part of using TerraSAR-X images in this research
for PSI is the high resolu tio n Spotlight mod e of the sensor [26],
[27]. Therefore we used a series of TerraSAR-X repeat observa-
tions in very fine-resolution Spotlight mod e (1 m azimuth reso-
lution and 0.4 m range resolution), single polarization (VV) and
an incidence angle range of about 41 degree. Until June, 2008,
11 ascending scenes (
9.65 GHz) have been acquired (Table I).
They are also characterized by perpendicular baseline values
between
338 and 96 m with respect to the central refer e nce
image of March 15, 2008 (Table I). The spatial coverage of Ter-
raSAR-X scenes is shown in Fig. 1. DEM from STRM with res-
olution o f 90 m is used to generate differential interferograms.
In addition, a 0 .5 m h igh resolution optical aerial photo is used
LAN et al.: COMPLEX URBAN INFRASTRUCTURE DEFORMATION MONITORING USING HIGH RESOLUTION PSI 645
Fig. 2. Comparison of 1-m resolution TerraSAR-X image to 0.5-m resolution
optical image.
to assist the classification of PS points. The comparison of Ter-
raSAR-X image with the optical aerial photo indicates the high
capabilities of TerraSAR-X image i n discriminating fine urban
features (Fig. 2).
III. PSI P
ROCESSING
The Interferometric Point Target Analysis (IPTA) technique
[9], [30], o ne of the PSI techniques, is used in processing the
time series of 1-m high resolution TerraSAR-X images. The
IPTA processing on time series o f trad itional low resolution
SAR images [19], [21], [24 ], [31]–[36] or high resolution Ter-
raSAR-X SAR images [2], [37] has been successfully used to
investigate earth surface de formations in a mm/yr scale.
Similar to other PSI techniques, the scattering stability o f
point targets in both spatial and temporal domain per mi ts inter-
pretation of interferometric phase of pairs with long baselines,
and thus makes a more complete use of the achieved data. For
large S L C data stack, a criterion of low temporal var iabi lit y and
high intensity of backscattering is adopted for the selection of
point target candidates, while low spectral diversity criterion is
used for small data stack. The differential interferometric phase
is expressed as the sum of topographic error ,de-
formation
, atomospheric ,and terms:
(1)
The phase model indicates a linear dependence of the to-
pographic error phase on the perpendicular baseline compo-
nent. By intro ducing a constant deformation rate component, a
linear dependence of the differential interfero metric phase on
the linear deformation rate is also fo unded. Consequently, a
two-dimensional linear regression analysis of differential inter-
ferometric phase on perpendicular b aseline and time is possible,
with the slope of th e regression in dicat ing the height correction
and the linear deformation rate. For large interferogram stack,
phase unwrapping is achieved in the mean time during the re-
gression, while spatial phase unwrapping may be required be-
fore regression step for small stack. This 2D regression analysis
in the temporal domain is performed for the entire point list.
The deviation of differential interferometric phase from the re-
gression, i.e., the residual phase, is composed of nonlinear de-
formation phase, atm ospheric phase, and noise phase. Different
components of residual phase can be discriminated according
to their dissimilar spatial and temporal correlations. Parameters
can be improved stepwisely by iteratively using the obtained
parameters including height corrections, refined baselines, and
atmospheric phase to refine the phase m odel in (1).
To make the point target analysis m ore sophisticated, it is
possible to detect the quality of points and reject those poi nts
not suited for analysis according to their phase standard devi-
ation from regression, and to expand the list of points eligib le
for IPTA based on refined phase model. The outcome of pro-
cessing mai nly consists of the deformation history (both linear
and nonlinear), the corrected heights and th e atmospheric phase
for each point. It has been reported that observed deformations
might be in fact domin ated by thermal dilation for short period
monitoring (less than o ne year) in urban area [22]. Extended
phase model w ith thermal dilation co mponent considered was
also introduced in some recent studies [38], [39].
Low temporal variability criterion is not suitable to detect
PS points for small stack of images because statistical char-
acters of backscattering intensity rely on large stack of images
to be fairly revealed. As only 11 TerraSAR-X images is avail-
able in our experiment, the PS candidates were selected based
on the com bin e d use of low spectral diversity and a low tem-
poralvariability criteria to achievea balancebetween the quality
and quantity of candidates. A stack of interferograms with 56
pairs was created using multiple reference images instead of
single reference image. The use of multiple reference images
can significantly increase the number of interferometric pairs
and therefore help to advance the reliability of regression pa-
rameters. Low spectral diversity criterion takes those well-fo-
cused point targets whose backscattering intensity almost keeps
constant when processing different looks with fractional az-
imuth and range bandwidth as PS can did ates. Th e number of
initial PS candidates i s 261,074. After iterative model refine-
ment and quality control, a total of 40,12 0 high quality persis-
tent scattere rs were identified accounting for a density of nearly
1,000 poin ts per
. In a dd ition, a large number of interfero-
metric pairs also allows phase unwrapping to be achieved during
the regression step.
The TerraSAR-X scene acquired in March 15, 2008 (im age
ID 4) was used as SLC reference geometry. Geocoding was per-
formed on the m ulti-look intensity image of the reference scene.
The geometry of d igital elevation model was also transformed
into the SAR geom etr y to assist the coregistration of images.
Offsets between the coregistered images reach below 0.1 SLC
pixel and therefore suggest very fine coregistration. Very short
baselines and temporal intervals of each TerraSAR-X interfer-
ometric pairs (Table I) facilitate the unwrapping and interpre-
tation of interferometric phase. After iterative refinement of the
phasemodel and filtering analysisof the final residual phase, the
linear and nonlinear deformation histo ries of PSs were obtained.
646 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, A
PRIL 2012
Fig. 3. TerraSAR-X PSI derived deformation rates of selected PS points. The
deformation rates are divided into three levels: high, medium and low. The lo-
cations of two villages (Dongg u Petro leum N ew Village and Bohai Petroleum
New Villa ge) and f o ur infrastructur es (Nanjiang Overpass Bridge, Ha ihe Nan-
jiang Bridge, Tanggu Beach Road and Nanjiang Conveyor Belt) are illustrated.
The linear deformation rates in vertical direction of s elected
PSs are presented in Fig. 3. They have been categorized into
three classes according to the specification of China geological
disaster prevention: high (annual rate
50 mm/y r), medium
(annual rate: 20–50 mm/yr), and low (annual rate
20 mm/yr).
Points with high, medium, and low deformati on rates a re ex-
pressed with red, yellow and green respectively. It can be seen
from Fig. 3 that most of the research area present low deform a-
tion r ate, although some PS points with high defor m at ion rate
are f ound in especially Beijiang Port, Nanjiang port and south-
west region of the research area. Detailed discussion about the
deformation patterns o f different infrastructure types will be is-
sued later in this paper.
PSs have been categorized into their corresponding types of
urban ground objects (i.e. building, road, b rid ge and etc.). The
classification of PSs has to be accomplished in the first place
for further deformation analysis in respect of certain urban in-
frastructure type. An urban map (1:2000) in the year 2008 of
the “Tianjin Binhai New Area” provided by the local governing
department is used to recognize the urban structure type that
each PS corresponds to. Overlay analysis was performed be-
tween the geocoded PS points and the urban map to classify
the PSs. The accuracy of PS classification can be improved
by rechecking the locations of PS points on the 0.5 m resolu-
tion aerial photo. T he extracted PSs from TerraSAR-X data are
then subdivided into predefined urban structure categories (e.g.,
ground, building, bridge, road and etc.) and subcategories (e.g.,
high-rise, m ediu m-rise and low-rise building). The distribution
of PSs in a number of representative infrastructure categories in
the research area is shown in Fig. 4.
IV. R
ESULT
A. Building Infrastructure
As the most important urban infrastructure, build ing s fre-
quently suffer from severe damage caused by uneven deforma-
tion. It often results in wall cracking, floor and structure tilting
Fig. 4. Classification of PS points r egarding their corresponding urban ground
objects. Some representative classes of PSs are presented.
and even collapse. Buildings in the Tanggu D istrict are classi-
fied into three sub-categories based on the number of stories and
height, namely low-rise, m ediu m-rise and high -rise building.
The deformations of buildings in D onggu Petroleum New Vil-
lage (Donggu PNV) and Bohai Petroleum New Vi llage (Bohai
PNV) were analyzed. Their locations are indicated in Fig. 3.
Both residential and commercial buildings are located in these
two Villages. Due to different geograp hical location and geolog -
ical conditions, different characteristics of building deformation
present in these two villages. By comparing the tempo-spatial
characteristics of building PSs, it is of interest to reveal the ef-
fect of the building type, co nstruction time, geographical and
geological environment on the b uilding subsidence.
TerraSAR-X PSI monitoring results of building defo rmation
in Bohai Petroleum New Village and Donggu Petroleum New
Village are shown in Fig . 5 an d Table II. It can be seen that
over 90% of buildings in the two Villages are characterized by
slow deformation (annual deform ation rate
20 mm/yr). The
1,712 building PSs located in Bohai PNV suggest an average
annual deform ation rate of 8.9 m m/yr w ith a maxim um of 44.2
mm/yr. Donggu PNV has totally2,205 m onitoring points on the
buildings indicating an average annual deformation rate of 11.6
mm/yrwitha maximumof60.3mm/yr.The proportion ofbuild-
ings characterized by fast subsidence in Dongu PNV is larger
than in Bohai PNV (Table II). For the same type of buildings,
the average deformation rate in Bohai PNV is far less than that
in the Dogng u PNV. For example, the average deformation rate
of medium-rise buildings is 6.0 mm/yr in the Bohai PNV, while
those in D o nggu PNV show much larger average rate of 14.6
mm/yr.
The spatial and temporal heterogeneity of building sub-
sidence was observed in both villages (Fig. 5). In Donggu
PNV, buildings located near Haihe River show much larger
deformation rate that th ose in inland regions, whil e buildings in
the so uth west part of Bo hai PNV were sinking faster than those
in northeast region. The time series deformatio n of buildings
basically exhibits linear trend, while visible non-linear defor-
mations were detected during the period from March to May,
2008 in Bohai PNV (Fig. 5).
LAN et al.: COMPLEX URBAN INFRASTRUCTURE DEFORMATION MONITORING USING HIGH RESOLUTION PSI 647
Fig. 5 . TerraSAR-X PSI monitoring results for buildings in Donggu an d Boh ai Petroleu m New Villages. The color code of d eformation rate of Fig. 3 is ado p ted.
Figures in the righ t show the time series deformatio n histori es in different r egions.
TABLE II
P
ARTITION OF BUILDING DEFORMATIONS IN DONGGU AND BOHAI PETROLEUM
NEW VILLAGE
Different types of buildings also show distinctive different
deformation characteristics (Fig. 6). The deformation rate o f
low-rise buil dings is generally more than 10 mm/yr which is rel-
atively faster com pared with medium-rise and high-rise build-
ings. For example, the typical high-rise buildings in Bohai PNV
have an average deformation rate of only 2.6 m m/yr, while the
value in Donggu PNV is 3.8 mm/yr which is slightly higher.
High rise buildings also show slightly nonlinear characters in
spite of small deformation rate (Fig. 6). M eanw hile, PSI mon-
itoring results show that large deformation occurred more fre-
quently in the corner of buildings adjacent to major roads.
The building subsidence processes are rel a ted to multiple
causative factors including engineering geological conditions,
foundation structure and construction time. As expected,
buildings are sinking much faster in the regions along Haihe
River (as shown in Dong gu PNV) an d the southwest side o f
the study region (as shown in Bohai PNV ) where soil layers
with high compressive feature and low-strength are widely
distributed. High-rise buildings usually utilize deep concrete
pile fo undation which minimizes the effect of load ing on the
underneath uncompacted soil layers and therefore are less
prone t o subsidence compared with low-rise buildings with
no concrete pile structure available for their foundations. The
Fig. 6. Time series defor mation characteristics of different types of buildings
in Donggu and Bohai PNV.
interaction bet ween building loading an d the surface geolog-
ical layers takes t im e to reach its equilibrium state. Buildings
usually demonstrate faster subsidence i n the initial stages after
construction. This mig ht explain why the building s in Donggu
PNV were sink ing faster than in Bohai PNV since Donggu
PNV i s much younger th an Bohai PNV. A nother factor that
may account f or the faster sinking rate in Donggu PNV is
that the underlaid sensitive soil layer in Donggu PNV is much
thicker than other regions and therefore exaggerates the effect
of building loading.
Thermal d ilation is another po ssible factor that might con-
tribute the diversity of defo rmation pattern of buildings. Typical
annual range of temperature in study area is about 30
which
would account for an expansion of 3.6 mm for a 10 m long struc-
ture if thermal dilation coefficient is assumed to be
. Our observation time (2007-12-29 2008-06-22) is ap-
proximately a span from peak winter to peak summer. In other
words, for buildings, 10 m d ifference in height cou ld probably
648 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, A
PRIL 2012
Fig.7. TerraSAR-XPSI Monitoringresults forNanjiang OverpassBridge. The
deformation histories of the main part and the northwest approach of the bridge
are shown in the up-right. Insert photo in the up-left also shows heavy duty
vehicle passing through.
result in 2.7 mm d ifference of therm al d e formation durin g our
observation time. Therefore, more vertical dilations of hig h-rise
buildings could be a possible explanation for their less observed
deformations compared with medium-rise and low-rise bu ild-
ings (less than 4 mm averagely, see Fig. 6). Similar correla-
tion of the observed movement with the height of bu ild ings is
also detected [40]. Unfortunately, as thermal effect m odule is
included in the phase mode of current codes, we therefore are
not able to precisely evaluate the effects of t herm al dilation on
the observed deformations.
B. Transportation Infrastructure
High resolution PSI technique is capable of recognizing
meter-scale linear features and allows performing engi-
neering-scale deformation monitoring on narrow u rban lin ear
transportation infrastructures. It helps engineers evaluate the
conditions of transportatio n infrastructure in a fast and reliable
way. The d eform ation processes of selected typical transporta-
tion infrastruc tur e s i ncluding bridge, road and linear conveying
belt w ere analyzed in this section. Their influencing factors and
associated mechanism were also discussed.
Nanjiang Overpass Bridge is one of the two major represen-
tative b rid ges in the study area. It is located in the south of the
study area and is the i mp ortant transportation channel to Tanggu
port. It is a double-way ro ad br idge with total of six lanes and
transportation capacity of 75 tons. Totally 205 PS points were
detected over the surface of the bridge (Fig. 7). The overall av-
erage deformation rate is 15.1 mm/yr. Different parts of the
bridge are characterized by different clusters of deform ation
rates. It can be seen that main part of the bridge is relatively
stable indicated by a large number of green PS points on its sur-
face, while all the approach roads connecting to the main part
of the bridge are characterized by rapid deformation with an av-
erage rate of 31.5 mm/yr. The maximum deformation occurred
in the secti on of northwest approach bridg e with a deformation
rate up to 73.3 mm/yr as shown in Fig. 7 by a red cross. U n -
even subsidence often results in sharp height change on the road
Fig. 8. Th e diversity of deform atio n between the west part and the east part
of Haihe Nanjiang Bridge. A n interpolated map of deformation rate has been
generated based on monitoring results of PS points.
surface. As a result, vehicles are prone to jump when passing
through the intersection between the approach and main part o f
bridge. The sinking time series of both the northwest ap pro ach
and th e main bridge show primary a linear behavior (Fig. 7).
The deform ation characteristi cs of bridge are directly re-
lated to the interactive effect between vehicle loading and the
engineering structures of bridges, while thermal effect plays
minor role. Field survey has observed a shuttling of heavy duty
vehicles crossing the bridge. The slow sub siding main bridge
is equipped with the concrete pile structure, while natural soil
foundation is used in the approach sections which results in
more severe deform atio n under the s ame veh icle loading.
An obvious uneven deformation pattern was also observed on
Haihe Nanjiang Bridge which locates in adjacent east of Nan-
jiang Overp ass Bridge (Fig. 8). The east part of Haihe Nanjiang
bridge (15 PSs) shows a larger av erage deformation rate of 18.4
mm/yr with a maxim um of 27.0 mm/yr, while an average defor-
mation rate of only 4.7 mm/yr with a maximum of 10.0 mm/yr
was found in the west part of the bridge. A deformatio n map is
generated by interpolation with the 23 PS points on the bridge
(Fig. 8) for better illustration of the uneven deformation phe-
nomenon. Visible no n-linear and differential deform ation char-
acteristics are also detected between both parts of the b ridge
(Fig. 8). This might imply an existence of possib le u neven sub-
sidence hazard, although the difference within the monitoring
time is quite minor (
1 cm). Field surv ey has found a number of
large piles of raw construction materials adjacent to the east part
of b ridge. They provid e extra loadings which impact on the sta-
bility of the east part of the bridge. Considering the main bridge
over Haihe River (about 2 00 m) is sensitive to the t her mal de-
formations, the o bserv ed diversity of deformation between the
west part and the east part of this bridge (about 7 mm, see Fig. 8 )
is also possibly attribut ed to thermal dilation.
LAN et al.: COMPLEX URBAN INFRASTRUCTURE DEFORMATION MONITORING USING HIGH RESOLUTION PSI 649
Fig. 9. TerraSAR-X PSI monitoring results for TangguBeach Road. There sec-
tions are marked on the map to present various deformation characteristics of
different road sec tions.
Tanggu Beach Road is a major transportation infrastructure
connecting the north and the south of the city. It serves as
an important gateway for transporting materials in Tianjin
Binhai New Area. The PSI deformation mo nitoring results
for t he Tanggu Beach Road is shown in Fig. 9, in which the
deformation rates of PS points are also interpolated for better
illustration. An average deformation rate of 23.2 mm/yr is
observed for the 5 km long road. The maximum annual defor-
mation (nearly 60 mm) is located at the southern intersection
between Tanggu Beach Road and Nanjiang Overpass Bridge.
Distinct spatial variation of deformation was observed along
Beach road. It can be divided into three sections (A, B and
C as indicated in Fig. 9). Section A is characterized by slow
deformation with an averag e rate of 13.9 mm/yr. Much higher
deformation rate was observed in Section B with an average
rate of 38.0 mm/yr. A number of abnormal sites can be iden-
tified in Section B indicated by red color in Fig. 9. Similar as
Section A, Section C has relatively low deformation rate with
an average rate o f 1 6.9 mm/yr. The relative larger de formation
in Secti on B more or less resulted from the d istr ibutary of
loading n ear the approach of the Nanjiang Overpass Bridge. As
this road structure principally lies perpendicular to the ground
range direction, thermal dilation, if exists, would have limited
influences on its observed spatial deformation pattern.
Another interesting transportation infrastructure is the Nan -
jiang conveyor belt in south of Nanjiang Po rt. It is one of the
important transportat ion infr astructures for conveying mining
materials from Nanjiang Port. The PSI monitoring on Nanjiang
conveyor belt sugg ests an annual average deform atio n amount
of 16.4 mm with a maximum of 50.1 mm. The conveyor belt is
divided into three sections for further discussion (Fig. 10).
An interesting deformation pattern has been observed from
distribution of PSs and deformation profile along conveyor belt
(Fig. 10). The deformation rate goes up and down regularly
Fig. 10. H igh resolution PSI Monitoring results for Nanjian g Conveyor Belt.
There sections are illustrated in the figure to representvarious deformation char-
acteristics. Temporal patterns of deformations of the three sections are also pre-
sented.
along the conveyor belt. The distribution of engineerin g struc-
ture is found to be responsible for the sensitive fluctuation of
deformation rate. The interval of deformation distribution is
closely corresponding t o the interval of the piers supporting the
conveyor belt. The stronger the piers are, e.g., the middle part
of the conveyor belt ( Sectio n B), the less the deformation takes
place. The regular fluctuation of deformation rate in section A is
also possibly derived from thermal dilation which would m ake
two parts of the tracks move towards each other in the construc-
tional gaps [41] an d in turn induce distinct contrast of observ ed
deformations between adjacent points.
V. C
ONCLUSION AN D DISCUSSION
Rapid urbanization processes involve intensive infrastructure
development which increases the urban vulnerability to subsi-
dence.A promptand accuratemonitoring ofcom plex infrastruc-
ture deformation at a relativ ely large scale is one of the m ost
important concerns in urbanized area. In this study, P ersistent
Scatterer Interferom etry has been carried out using high resolu-
tionTerraSAR-X images tomonitorthe deformation of complex
urban infrastructures in one fast developing area.
The classification of PS points r egarding their correspondin g
urban structure ty pes was firstly accomplished for deform ation
analysis. Detailed deformation patterns of different representa-
tive types of urban infrastructures were successfully revealed.
The def or mat ion s of all in vestigated infr astructures in the study
area unanimously present a remarkable spatial and temporal
variation. The com plex ity of urban infrastructures with regard
to their types, engineering structures, geometries, construction
ages, engineering geological conditions and external loadings
results in complicated spatial-temporal defo rmation patterns of
urban infrastructures.
As expected, newer buildings and those buildings located in
relatively inferior engineering geological conditions sink much
faster. High-rise buildings show lower deformation rates com -
pared with low-rise build ings thanks to t heir stro nger pile f oun-
dation and possible larger thermal expansions.
650 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, A
PRIL 2012
The spatial def or mat ion patterns of linear transportation in -
frastructures were also well revealed. Especially, a rema rkable
regular fluctuating pattern of deformation rate was detected
along a linear conveyor belt. The interactive effect between
external loading and the engineering structure dominates th e
deformation character of transportation infrastructure. In gen-
eral, sections of transportation infrastructures bearing heavier
loadings or supported b y more sensitive foundations are more
prone to subsidence. In addition, lateral thermal dilation is
another possible factor which would domin ate the diversities
of observed deformatio ns between different parts of infrastruc-
tures, especially for short period observations [22], [39].
The observed deformation patterns of various types of infra-
structures from TerraSAR-S PSI are comparable with o ur field
observations. The deformation m echanism of varieties of urban
infrastructures w as in ter pret e d wit h regar d to th eir types, en -
gineering structures, geometries, geological settings, construc-
tion tim e s and exterior loadings. The spatial-temporal var ying
patterns of urban infrastructure deformatio n in a regional scale
have been revealed althoug h som e systematic deviations migh t
exist between the exact value of our m onitoring and in-situ mea-
surements like leveling.
The availability of SAR images is one possible drawback to
our results. Therefore multiple reference scenes in PSI tech-
niqueweread opted perm iting th e potential of PSI analysis using
small data stacks [42]. For example, ground d e formation in t he
urban area of Guangzhou city was successfully detected based
on 10 ASAR images [21]. I n addition, the reliability of PS clas-
sification highly relies on the backscattering characteristics of
SAR signals, accuracy of the geocoding of PS points and the ac-
curacy of the urban map. The discrepancy of coordinate systems
ofdifferentdata sources wouldintroducecoregistratio n problem
and therefore fallacious classification to some PS points.
To rev eal the temporal variation of ground deformation,
the non-linear deformation has been addressed considerably
in the PSI processing. It was obtained by excluding phase
errors (noise), atmospheric phase from the final residual phase
generated by linear regression model. Therefore, the results
are strongly affected by the linear model. The consolidation
of linear regression model was conducted through iteratively
running linear regression analysis with refined p aram eters,
i.e. refined baselines, refined atmosphere and point height
estimates, using spatially filtered reference point. However, it
is alwa ys a challeng e to precisely discriminate the no n- linear
deformation, atmospheric effect and phase noise in the residual
phase, especially for the short temporal image stacks. For
example, the very minor component non-linear deformation
could be easily attributed to the distu rbing com po nents of phase
noises. Hence, the interpretation of both linear and non-linear
component of deformation should be performed with cautions.
A num ber of advanced functions have not been included in
the current PSI code used in this research. For example, the
quantitative analysis of potential influences o f thermal dilation
effects o n our PSI deformation monitoring is missing due to the
unavailability of cert a in module. Employing a phase mode wit h
thermal dilation considered [ 38], [39] could help to g et more
detailed information (both dynamic and static) regarding the
deformation of infrastructures. The key element in PSI tech-
nique is to extract useful information i.e. deformation from the
original observation. Therefore the role of the “disturbing com-
ponents” has to be considered in processing and interpretation.
The robust spatial and temporal filtering technique might help
improve such discriminat ion. The flexibility of selection of fil-
tering window shape and length could also be a benefit.
Despite of various possible disadvantages, the potential of
high TerraSAR-X PSI to monitor the deformation of complex
urban in frastructures is fairly presented in o ur experim ent.
A
CKNOWLEDGMENT
Theauthors wish tothank TangguCenter forWaterResources
Management for the support of field investigation and the re-
lated data, Tianjin Institute of Survey and Mapping for the sup-
port of high resolution aerial images an d Spot Image CN for the
support of TerraSAR-X SAR data.
R
EFERENCES
[1] J. Jensen and D. Cowen, “Remote sen sing of ur ban/suburban infra-
structure and socio-economic attributes,” Photogramm. Eng. Remote
Sens., vol. 65, no. 5, pp. 611–622, May 1999.
[2] T. Strozzi, P. Teatitni, and L. Tosi, “Terrasar-X reveals the impact of
the mobile barrier works on Venice coastland stab ility,” Remote Sens.
Environ., vol. 113, no. 12, pp. 2682–2688, Dec. 2009.
[3] Z. Cuiand Y.Tang,“Landsubsidence andporestructureof soilscaused
by the high-rise building group through centrifuge model test,” Eng.
Geol., vol. 113, no. 1–4, pp. 44–52, May 2010.
[4] A.Ng, H. Chang, L. Ge, C. Rizos, and M.Omura, “Assessment of radar
interferometry performance for ground subsidence monitoring due to
underground mining,”EarthPlanetsSpace,vol.61,no.6,pp. 733–745,
2009.
[5] A. Ferretti, C. Prati, and F. Rocca, “Permanent scatterers in SAR inter-
ferometry,” in Proc.IEEEInt.Geosci.RemoteSens.Symp., Hamburg,
Germany, 1999, pp. 1528–1530.
[6] A. Ferretti, C. Prati, and F. Ro cca, “Perm anent scatt erer s in SAR in-
terferometry,” IEEE Trans. Geosci. R emote Sens., vol. 39, no. 1, pp.
8–20, Jan. 2001.
[7] P. Beradino, G. Fornaro, R. Lanari, and E. Sansosti, “A new algorithm
for surface deformation monitoring based on small baseline differential
SAR interferograms,” IEEE Trans. Geosci. Rem ote Sens., vol. 40, no.
11, pp. 2375–2383, Nov. 2002.
[8] O. Mora, J. J. Mallorqui, and A. Broquetas, “Linear and nonlinear
terrain deformation maps fro m a reduced set of interferometric SAR
images,” IEEE Trans. Geosci. Remote Sens., vol. 41, no. 10, pp.
2243–2253, Oct. 2003.
[9] C. Werner, U. Wegmüller, T. Strozzi, and A. Wiesmann, “Interfero-
metric point target analysis for defor mation mapping,” in Proc. IEEE
Int. Geosci. Remote Sens. Symp., Toulouse, France, July 21–25, 2003,
pp. 4362–4364.
[10] A. Hooper, H. Zebker, P. Segall, and B. Kam pe s, “A new m ethod for
measuring deform ation on volcanoes and o t her nat ural terrains using
InSAR persistent scatterers,” Geophys. Res. Lett., vol. 31, no. L23611,
2004, doi 10.1029/2004GL021737.
[11] B. M. Kampes, “Displacement parameter estimation using p er manent
scatterer interferometry,” Ph.D. dissertation, Delft Univ. Technol.,
Delft, The Netherlands, 2005.
[12] J.Duro,J.Closa,E.Biescas,M.Crosetto,andA.Arnaud,“Highresolu-
tion differential interferometry using time series o f ERS and ENVISAT
SAR data,” in Proc. 6th. Geomatic Week Conf., Barcelona, Spain, Feb.
2005.
[13] M. Van der Kooij, W.Hughes, S. Sato,and V. Poncos,“Coherent target
monitoring at hig h spatial density: Examples of validation results,”
Proc. FRINGE Workshop, 2005.
[14] C. Prati, A. Ferretti, and D. Perissin, “Recent advances on surface
ground deformation m ea surement by means of repeated space-borne
SAR observations,” J. Geodynam., vol. 49, no. 3–4, pp. 161–170, Apr.
2010.
[15] N.Adam,M.Eineder,N.Yague-Martinez, and R. Bamler, “High res-
olution interferometric stacking with TerraSAR-X,” in Proc. IEEE Int.
Geosci. R emote Sens. Symp., Boston, MA, 2008, pp. II-117–II-120.
LAN et al.: COMPLEX URBAN INFRASTRUCTURE DEFORMATION MONITORING USING HIGH RESOLUTION PSI 651
[16] C. Colesanti, A. Ferretti, F. Novali, C. Prati, and F. R oc ca, “SAR mon-
itoring of progressive and seasonal ground deformation using the per-
manent scatterers technique,” IEEE Trans. Geosci. Remote Sens., vol.
41, no. 7, pp. 1685–1701, Jul. 2003.
[17] A. Ferretti, F. Novali, R . Bürgmann, G. Hilley, an d C. Prati, “InSAR
permanent scatterer analysis reve a ls ups and downs in San Francisco
Bay area,” EOS Trans. AGU, vol. 85, no. 34, Aug. 2004, doi 10.1029/
2004EO340002.
[18] R. Lanari, O. Mora, M. Manunta, J. J. Mallorquí, P. Berardino, and E.
Sansosti, “Asmall-baselineappr oachfor investigating deformationson
full-resolution differential SAR interferograms,” IEEE Trans. Geosci.
Remote Se ns., vol. 42, no. 7, pp. 1377–1386, Jul. 2004.
[19] S. Stramondo, F. Bozzano, F. Marra, U. Wegmuller, F. R. Cinti, M.
Moro, and M. Saroli, “Subsidence induced by urbanisation in the city
of Rome detected by advanced InSAR technique and geotechn ical in-
vestigations,” Rem ote Sens. Environ., vol. 112, no. 6, pp. 3160–3172,
Jun. 2008.
[20] N.Adam,A.Parizzi,M.Eineder,andM.Crosetto,“Practicalpersistent
scatterer processing validation in the course of the Terrafirma project,”
J. Appl. Geophys., vol. 69, no. 1, pp. 59–65, Sep. 2009.
[21] Q.Zhao,H.Lin,L.Jiang,F.Chen,andS.Cheng,“Astudyofground
deformationintheGuangzhouurbanarea with persistent scatt erer in-
terferometry,” Sensors, vol. 9, no. 1, pp. 503–518, Jan. 2009.
[22] M. Crosetto, O. Monserrat, R. Iglesias, and B. Crippa, “Persistent
scatterer interferometry: Potential, limits and initial C- and X-band
comparison,” Photogramm. Eng. Remote Sens., vol. 76, no. 9, pp.
1061–1069, Sep. 2010.
[23] G.Herrera,R.Tomás,D.Monells,G.Centolanza,J.J.Mallorquí,F.
Vicente, V. D. Navarro, J. M. Lopez-Sanchez, M. Sanabria, M. Cano,
and J. Mulas, “Analysis of subsidence using TerraSAR-X dat a: Murcia
case s tudy,” Eng. Geol., vol. 116, no. 3–4, pp. 284–295, Nov. 2010.
[24] L. Jiang, H. Lin, and S. Cheng, “Monitoring and assessing reclamation
settlement in coastal areas with advanced InSAR techniques: Macao
city (China) case study,” Int. J. Remote Sens., vol. 32, no. 13, pp.
3565–3588, Jul. 2011.
[25] G. Liu, H. Jia, R. Zhang, H. Zhang,H . Jia, B. Yu, and M. Sang, “Explo-
ration of subsid en ce estimation by persistent scatterer InSAR on time
series o f high resolution TerraSAR-X images,” IEEE J. Sel. Topics
Appl. Earth Observ. Remote Sens., vol. 4, no. 1, pp. 159–170, Mar.
2011.
[26] N. Adam, X. Zhu, and R. Bamler, “Coherent stacking with Ter-
raSAR-X imagery in urban areas,” in Joint Urban Remote Sensing
Event, Shanghai, China, May 2009.
[27] M. Eineder, N. Adam, R. Bamler, N. Yague-Martinez, and H . Breit,
“Spaceborne spotlight SAR interferometry with TerraSAR-X,” IEEE
Trans. Geosci. Remote Sens., vol. 47, no. 5, pp. 1524–1535,May 2009.
[28] X. Zhu and R. Bamler, “Very high resolution spaceborne SAR tomog-
raphy in urban environment,” IEEE Trans. Geosci. Remote Sens., vol.
48, no. 2, pp. 4296–4308, Dec. 2010.
[29] R. Werninghaus and S. Buckreuss, “The TerraSAR -X mission and
system design,” IEEE Trans. G eosci. Remote Sens.,vol.48,no.2,pp.
606–614, Feb. 2010.
[30] U. Wegmuller, C. Werner, T. Strozzi, and A. Wiesmann, “Multitem-
poral interferometric p oint target analysis,” in A nalysis of Multi-Tem-
poral R emote S ensing Image s , ser. Remote Sensing, P. Smits and L.
Bruzzone, Eds. Hoboken, NJ: World Sci., 2004, vol. 3, pp. 136–144.
[31] P. Teatini, L. Tosi, T. Strozzi, L. Carbognin, U. Wegmuller, and F.
Rizzetto, “Mapping regional land displacem ents in the Venice Coast-
land by an integrated monitorin g system,” Remote Sens. Environ., vol.
98, no. 4, pp. 403–413, Oct. 2005.
[32] P. Teatini, T. Stroz zi, L. Tosi, U. Wegmuller, C. Werner, and L.
Carbognin, “Assessing short- and long-time d isplacements in the
Venice coastland by synthetic aperture radar interferometric point
target analysis,” J. Geophys. Res., vol. 112, no. F01012, 2007, doi
10.1029/2006JF000656.
[33] Y. Guéguen, B. Deffontaines, B. Fruneau, M. Al Heib, M. de Michele,
D. Raucoules, Y. Guise, and J. P lanchenault, “Mon ito ring residual
mining subsidence of Nord/Pas-de-Calais coal basin from differential
and Persistent Scatterer Interferometry (Northern France),” J. Appl.
Geophys., vol. 69, no. 1, pp. 24–34, Sep. 2009.
[34] I. Parcharidis, M. Foumelis, P. Kourkouli, and U. Wegmuller, “Persis-
tent Scatterers InSAR to detect groun d deformation ove r Rio-Antirio
area (Western Greece) for the period 1992–2000,” J. Appl. Geophys.
,
v
ol. 68, no. 4, pp. 348–355, Aug. 2009.
[35] L. Jiang and H. Lin, “Integrated analysis of SAR interferometric and
geological data for inve s tigating long-term reclamation settlement of
Chek Lap Kok Airport, Hong Kong,” Eng. Geol., vol. 110, no. 3–4,
pp. 77–92, Feb. 2010.
[36] Y. Zhang, J. Zhang, H. Wu, Z. Lu, and G. Sun, “Monitoring of urban
subsidence with SAR interferometric point target analysis: A case
study in Suzhou, China,” Int. J. Ap pl. Earth Observ. Geoinform., vol.
13, no. 5, pp. 812–818, Oct. 2011.
[37] U. Wegmüller, D. Walter, V. Spreckels, and C. Werner, “Nonuniform
ground mo tion monitorin g with TerraSAR-X p e rsisten t scatterer in-
terferometry,” IEEE Trans. Geosci. R emote Sens., vol. 48, no. 2, pp.
895–904, Feb. 2010.
[38] M. Cuevas, O. Monserrat, and M. Crosetto, “A new product from per-
sistent scatterer interferometry: The thermal dilation maps,” in Joint
Urban Remote Sensing Event, Munich, Germany, Apr. 11–13, 2011,
pp. 285–288.
[39] O. Monserrat, M. Cuevas, M. Crosetto, and B. Crippa, “The thermal
expansion component of persistent scatterer interferometry observa-
tions,” IEE E Trans. Geosci. Remote Sens., vol. 8, no. 5, pp. 864–868,
Sep. 2011.
[40] J. Duro, O. Mora, M. Agudo, and A. Arnaud, “First results of stable
point network software using TerraSAR-X data,” in Proc. EUSAR,
Aachen, Germany, Jun. 7–10, 2010.
[41] S. G ernhardt, N. Adam, M . Eineder, and R. Bam ler, “Potential o f very
high reso lution SAR for persistent s catterer interferometry in urban
areas,” Annals of GIS, vol. 16, no. 2, pp. 103–111, June 2010.
[42] U. Wegmuller and A. Wiesmann, “Potential of interferometry point
target analy sis u sin g s mall data stacks,” in 3rd Int. Workshop on ERS
SAR Interferometry, FRI N GE’ 03 , Frascati, Italy, Dec. 2003.
Hengxing Lan received the B.S. and M.Sc. degrees
in geology from Shandong University of Science
and Technology in 1995 and 1998, respectively. He
received the Ph.D. degree in geological engineering
from the Institute of Geology and Geophysics,
Chinese Academy of Sciences, in 2001.
He worked in the department of civil and envi-
ronmental engineering in Hong Kong University
in 2003 as a Research Associate. From 2004 to
2009, he was a Research Engineer in the University
of Alberta in Canada. He is currently a Professor
with the State Key Lab of Resources and Environmental Information System
(LREIS), In stitu te of Geographic Sciences and Na tur al Resources Research,
Chinese Academy of Sciences. His current research in terests a re d evelopment
and application of R e mote Sensing and GIS technology to geological hazards
problems.
Langping Li received the B.S. and M.Sc. degrees in physical geography from
Nanjing University in 2007 and 2 01 0, respectively. He is cu rrently a Ph.D. can -
didate majoring in GIS and remote sensing in the State Key Lab of Resources
and Environmental Informa tion System (LREIS), Institute of Geographic Sci-
ences and Natural Resources Research, Chinese Academy of Sciences.
HongjiangLiu received the B.S. degree in physical geography from Southwest
Normal Un iversity in 1993 . He receiv ed the M.Sc. degrees and Ph.D. degree in
physical geography from the Institute of Mountain Hazards and Environment,
Chinese Academy of Sciences, in 1993 and 2008, respectively.
He is currently a Post-Doctoral Fello w in GIS and Remote Sensing in the
State Key Lab of Resources and Environmental Information System (LREIS),
Institute of Geographic Sciences and Natu r al Resources Research, Ch in ese
Academy of Sciences.
Zhihua Yang recei ved the B.S . degree from Shandong Architect University in
2007 and the M.Sc degree in geographic information systems from the China
Geoscience University in 2010. He iscurrently a Ph.D. candidate in GISan d Re-
moteSensingintheStateKeyLabofResources and EnvironmentalInformation
System (LREIS), Institute of Geograph ic Sc iences an d Natural Reso u rces Re-
search, Chinese Academy of Sciences.
