SAR INTERFEROMETRY AND SPECKLE TRACKING APPROACH FOR GLACIER VELOCITY
ESTIMATION USING ERS-1/2 AND TERRASAR-X SPOTLIGHT HIGH RESOLUTION DATA
Vijay Kumar, G.Venkataraman, Y. S. Rao,
Centre of Studies in Resources Engineering
Indian Institute of Technology, Bombay
Mumbai 400076 India
[email protected]
ABSTRACT
Glacier retreat and advance is general phenomenon in Himalayan
region due to change in climatic conditions. For quantifying the
global warming effects on local scale glacier system monitoring and
estimating its dynamic geophysical parameters viz. movement and
volume change are important. In this study glacier movement
estimation is attempted in north western Himalayan region using
spaceborne InSAR technique, which is based on preserving the
coherence between two acquisitions of the same scene. It is observed
that ERS-1/2 tandem data give high correlation over Gangotri glacier
and Siachen glacier and movement in LOS direction is deciphered.
However, with the use of SAR Speckle tracking method two
dimensional (Azimuth and range directions) velocities of glaciers can
be obtained. In this study attempt has been made to measure 2-D
velocity components of Gangotri glacier. New generation TerraSARX (TS-X) high resolution spotlight mode (HS), single look slant
range complex (SSCs) data of 28th August, 2008, 08th September,
2008, 19th September, 2008 and 30th September, 2008 are exploited
for this study. Interferogram, coherence image and intensity image
are generated. It is observed that Interferometric SSC data of 28th
August, 2008 and 08th September, 2008 give some fringes outside
glacier area and complete decorrelation is shown on the Gangotri
glacier due to high movement of glacier.
Keywords: Coherence loss, glaciers, SAR interferometry, DInSAR
1. INTRODUCTION
Spaceborne Differential Interferometric Synthetic Aperture Radar
(DInSAR) has been successfully used to measure millimeter- to
meter-level deformation on the surface due to earthquake, landslide,
glacier movement and physical process at depth including magma
accumulation and migration, pressurization, and crystallization [1-4]
during the past decade. However application of SAR interferometry
is limited in faster-moving areas [4]. With a 1day repeat period, ERS1/-2 tandem SAR data are much better matched to fast motion
estimations. Goldstein [5] clearly demonstrated the value of InSAR
for mapping glacial ice motion, but also indicated that use of
differential phase could be limited by temporal decorrelation.
However, the relatively large temporal separation of the repeat
coverage for current satellites (RADARSAT; 24 days, ERS-2; 35
days, ENVISAT; 35 days, ALOS; 46 days, TerraSAR-X; 11 days),
coupled with the large speeds that can occur in land locked and outlet
978-1-4244-3395-7/09/$25.00 ©2009 IEEE
V - 332
glaciers and ice streams, complicate the analysis of differential phase
for radial ice motion for coastal or high speed regions. In this study
InSAR as well as speckle tracking approach is used for glacier
movement estimation. ERS-1/2, ENVISAT ASAR and TerraSAR-X
data sets are exploited for this important research.
Gangotri, Siachen, Bara Shigri and Chhota Shigari are major glaciers
in the Himalayan region, which are showing retreat, and their
respective tributary glaciers completely disconnected from main body
of glaciers. ERS-1/2 observations show high correlation on glacier
area and hence movement of Siachen and Gangotri glacier are
measured. Information about dynamism of glaciers can be retrieved
by counting differential interferograms. Displacement of Gangotri
glacier in the radar look direction has been observed as 8.4 cm
represented by 3 fringes. Siachen glacier exhibits a displacement of
22 cm represented by 8 fringes. ERS-1/2 tandem data over all these
glaciers show highest correlation over glacier areas but ENVISAT
ASAR data shows coherence loss over glacier area due to
decorrelation [6]. It is due to change in climate between two
acquisitions at glacier locations. Coherence loss is usual phenomena
in glaciated terrain as repeativity of sensor is high (35 days for
ENVISAT). Among all these, Siachen glacier shows highest
coherence and then Gangotri, respectively. A tandem pair of ERS1&2 acquired on April 1 and 2, 1996 in descending pass over Siachen
shows high coherence than the ascending pair acquired on May 2 and
3, 1996. It is due to change in climate between two acquisitions at
glacier locations. A systematic monitoring of dynamism of
Himalayan glaciers can be done using Permanent Scatterer
Interferometric SAR (PSInSAR) [7] if we place a corner reflector on
the glaciers.
In this study glacier movement estimation has been
attempted using satellite radar interferometry. ERS -1/2 tandem data,
Envisat ASAR and TerraSAR-X data pairs are used for this study.
TerraSAR-X is a new German radar satellite carrying a high
frequency X-band (3.1cm) SAR sensor that works in Spotlight-,
Stripmap- and ScanSAR-modes providing high resolution SAR
images for detailed analysis as well as wide swath data whenever a
larger coverage is required. Imaging is possible in single, dual and
quad-polarization [8]. In this paper we present our initial assessment
of applicability of spotlight high resolution SAR SSC interferometric
pairs for monitoring the movement of Himalayan glaciers. Due to the
X-band frequency TerraSAR-X interferometry will be more sensitive
to orbit errors than current SAR sensors that operate in C-band or Lband [9]. A single frequency GPS receiver plus an additional dualfrequency GPS flown as an experimental payload deliver an orbit
accuracy in the order of centimeters. TerraSAR-X will supplement
IGARSS 2009
and enhance the InSAR based observations using other satellite data
sets because of its high phase to deformation sensitivity, high spatial
resolution (1 meter in High Resolution Spot Light Mode) and short
(11 day) repeativity. Short orbital cycle (11 days) of TS-X allows
more frequent interferometric observations than the orbits of, e.g.
ERS, ENVISAT and PALSAR. Since Himalayan climatic conditions
are very dynamic and hence getting interferometryc SAR (InSAR)
fringes on the body of glacier is very difficult because of
decorrelation between two acquisitions of the same scene.
For Gangotri glacier movement study, we have processed ENVISAT
ASAR image mode data using Gamma software. After processing a
pair of Single look complex data of 26 Nov 2003 and 19th Jan 2005
we found interferograms. In similar fashion we generated power
image as well as coherence image. Other seven pairs of ENVISAT
ASAR data are processed for interferograms, coherence image and
power image. Out of seven pairs only one pair combination gives
reasonably good interferogram and coherence. We have processed
ERS-1/2 data for developing Gangotri elevation map and glacier
movement map by DInSAR technique. The relation between
differential phase change Δφ and the surface movement (Δρ ) is given
by [10]
Δφ = φ ' − ( B ' / B )φ = (4π/λ )Δρ
(1)
Where φ' is the phase of interferogram after some changes φ is the
phase of interferogram without changes and BII refers to parallel
component of baseline of a pair. Thus, InSAR (derived DEM) with
an accuracy of 5 metres is very useful for topography mapping,
whereas the differential InSAR with an accuracy of a few millimetres
is a very efficient technique for various applications. Since glacier
and snow covered area interferogram is prone to decorrelation and
hence coherence loss because of movement of the glacier and change
in meteorological conditions, higher temporal base line day
difference. Hence deciphering the movement of glaciers is very
difficult. More recently correlattion based feature tracking techniques
have been used involving coherence tracking and Intensity tracking
approach [11-14] for glacier and ice velocity estimation. This paper
investigates the SAR feature tracking for Gangotri glacier movement
estimation in western Himalayas.
1.1 SAR Features Tracking for Glacier Movement Studies
Two single look complex (SLC) SAR data of Gangotri
glacier region are processed for intensity image using Gamma
software. Gamma software automatically read the parameter file
having information about satellite orbit parameters and coregister two
intensity images precisely. Both coregistered intensity images are
superimposed over each other. One intensity image is georeferenced
with the toposheet and other intensity image is also georeferenced
using same ground control points so that georeferencing error
remained same in both intensity images. Intensity images are
projected with Universe Transform Mercator (UTM) projection
showing positional coordinates in units of meters. Intensity images
are interpreted for objects of high backscattering intensity and the
lowest back scattering intensity, which remain visible on both
images. Water in Ponds shows lowest intensity and rock boulders
show relatively higher intensity than others on the body of the
Gangotri glacier. Central pixel coordinates of glacier boulders and
water bodies on the glacier, it is found different on both images.
Same pixel has different coordinate point; it means that pixel is
displaced from original position. Finally, relative distance is
calculated between two points on the images.
2. TEST SITES
The two major glaciers of India viz. Gangotri and Siachen are located
in the north western part of Himalayan region of India. Gangotri
glacier located in Uttarkhand state of India form the major source of
water for the major part of northern India. According to some earlier
studies [15] this glacier is retreating at the rate of 19 m/yr. This
glacier is about 26 km long with a maximum width of 2.4 km. The
tributary glaciers joining main trunk are as follows. Right of the main
glacier starting from terminus is 1) Raktyaban, 2) Chaturangi, 3)
Swachand, and 4) Maiandi. The left of the glacier is 1) Meru, 2)
Kirti, and 3) Ghanohim. Some of these tributary glaciers are in turn
having their own tributaries. Siachen (N35029’50”, E77000’04’’) is
the biggest glacier located at higher elevation. It has a length of about
72 km and varying width. It is the second largest glacier in the world
outside the polar region. It feeds the Nubra river (also known as
Shaksgam river) that flows parallel to the Karakoram range befor
entering into Tibet. The trunk glacier and its tributaries are in the
form of a vast ice field. Continuous snow fall can be seen during
winter periods. The average winter snow fall is about 10.5m. The
temperature varies from -100C and -500C. The elevation varies from
the snout (3,620m) to source of the glacier (i.e Indira col) (5,753m).
Table-1 List of ERS-1/2 and ENVISAT ASAR data pairs.
Satellite/
Sensor
Date
Pass
Base
line
No. of
Days
ERS-1/2
25-03-1996
Asc
75
01
Gangotri
Des
110
01
Siachen
Asc
114
01
Siachen
Des
52
35
Gangotri
Des
35
210
Gangotri
Des
581
340
Gangotri
Des
12
35
ENVISAT
ASAR
26-03-1996
01-04-1996
02-04-1996
02-05-1996
03-05-1996
09-07-2003
13-08-2003
26-11-2003
23-06-2004
09-07-2003
23-06-2004
21-12-2004
25-01-2005
Site
Siachen
3. DATA PROCESSING
Gamma software developed by Switzerland based company was used
for the processing of ERS-1 and 2 data sets. Initially the raw data
were processed for Single Look Complex (SLC) data product. ERS-1
and 2 tandem SLC data were coregistered and interferograms were
generated. The interferograms were flattened for phase unwrapping.
The unwrapped phase was converted to height and finally geocoded
DEMs were generated. The other products generated through the
software are power and coherence images. For differential
Interferometry, we took unflattened interferograms. SRTM DEM was
V - 333
conditioned by modifying no data value (i.e -32768) by replacing
with zero for the convenience of Gamma software. The output of this
task is simulated SAR backscatter image in DEM geometry and
geocoding look table. After this task, DEM is transformed into SAR
coordinates. Using the transformed DEM, unwrapped topographic
phase is created using Gamma software. This phase is subtracted
from the complex interferogram (not flattened) to get differential
phase. This differential interferogram is filtered and then the phase is
unwrapped. Finally, displacement map is obtained. Coherence and
power images were generated for the data sets given in table 1.
Coherence window size of 5x5 is chosen for coherence calculation.
In case of ENVISAT ASAR, we acquireSd SLC data from ESA. The
data were then processed for interferograms and coherence images
using Gamma Software.
Table 2. List of TerraSAR-X data used for interferogrm generation
Mode
Date
Asc/Des
Inci ang.
Polariz.
Look Dir.
HS
HS
HS
HS
28-08-2008
08-09-2008
19-09-2008
30-09-2008
Desc
Desc
Desc
Desc
38.44
38.44
38.44
38.44
HH
HH
HH
HH
Right
Right
Right
Right
4. RESULTS AND DISCUSSION
Siachen glacier Interferograms are formed from ERS-1/2 SAR data
of April 1 and 2, 1996 in descending pass and after subtracting it
from simulated interferogram differential interferogram obtained. In
differential interferogram each color cycle (magenta-cyan-yellowmagenta) represents one fringe which is equal to half wavelength (2.8
cm) displacement in the radar direction. Differential interferogram
shows good development of fringes over Siachen glacier particularly
at higher elevation as well as lower elevation, i.e. ablation area of
Siachen. In this scene fringes are also seen above the Siachen glacier,
where there are some small glaciers. In this differential interferogram
maximum number of 8 fringes equivalent to 22.4 cmd-1 displacement
towards the radar direction is observed. While for Gangotri glacier
almost 3 fringes (8.4 cm) were observed using ERS-1&2 SAR
tandem data (March 25&26, 1996) is shown in Fig. 1 (right).
Gangotri glacier interferogram, coherence map, uwrapped
interferogram and displacement map superimposed on intensity
image are shown in Fig. 3 (a) (b) (c) and (d) respectively.
For tandem data with one day interval, the coherence is preserved in
most of the glacier areas due to uniform motion. Coherence loss is
observed over Siachen glacier using ENVISAT ASAR data with 35
day interval of Dec. 21, 2004 and Jan. 25, 2005 is shown in Fig. 2. In
this image, coherence loss is observed mainly in the glacier part and
good coherence is seen in the surrounding areas. It may be due to
large motion of the glacier within 35 day interval. Using 70 day
interval data acquired on May 10 and July 19, 2005, we observed
further loss of coherence. It may be due to several changes
(precipitation, erosion, vegetation growth, soil moisture conditions,
etc.) during the 70 day interval. Complete coherence loss is observed
for the data with 350 day interval in Siachen glacier region.
Fig. 2. Coherence images of Siachen glacier (a) for ERS-1&2 April 1
and 2, 1996 pair and (b) May 2 and 3, 1996 pair. Loss of coherence in
May is higher
Figure 3. An (a) An .interferogram from March 25&26, 1996 ERS
data(b) Coherence image (c)Unwrapped interferogram (d)
Displacement map of Gangotri glacier.
Fig. 1. (a) ERS-1 SAR differential interferogram (left) acquired from
data April 1and April 2, 1996 for Siachen and the differential
interferogram (right image) for Gangotri glacier. Whole Scene covers
around 100 km x 100 km.
On the basis of intensity tracking of glacier objects in multitemporal
ASAR images it is observed that high intensity centres and low
intensity centres moved by 10.79 cmd-1 along its flow line. Using
four TS-X data sets three interferometric pairs are formed with 11
days temporal resolution. Only one pair out of these three of 28th
August 2008 and 8th September 2008 show interferograms which can
V - 334
be visually interpreted. It is observed that this pair give interferogram
over snow free area and glacier area also remained completely
decorrelated. Decorrelation on glaciers is observed because of its
rapid movement.Confidence of this observation can be assessed by
high resolution TS-X data based displacement study. Glacier objects
identified on the basis of local intensity differences and it is displaced
by 3 m in 32 days. Hence magnitude of displacement using TS-X
data is 9.37cmd-1 which is less by 1.42cmd-1 of ASAR based
observations.
References
[1] Massonnet et al., 1993 D. Massonnet, M. Rossi, C. Carmona, F. Adragna,
G. Peltzer, K. Feigl and T. Rabaute,The displacement field of the Landers
earthquake mapped by radar interferometry, Nature 364 (1993), pp. 138–
142.
[2] Massonnet et al., 1994 D. Massonnet, K. Feigl, M. Rossi and F. Adragna,
Radar interferometry mapping of deformation in the year after the
Landers earthquake, Nature 369 (1994), pp. 227–230.
[3] Massonnet, D., P. Briole, and A. Arnaud, Deflation of Mount Etna
monitored by spaceborne radar interferometry, Nature, 375, 567– 570,
1995.
[4] Massonnet, D., and K. L. Feigl (1998), Radar interferometry and its
application to changes in the Earth's surface, Rev. Geophys. 36, 441-500
[5] R. Goldstein, R. Engelhard, B. Kamb, and R. Frolich, “Satellite
radar iterferometry for monitoring ice sheet motion: Application to
an ntarctic ice stream,” Science, vol. 262, pp. 1525–1530,
Dec.1993.
Figure 4. lowchart for measurement of glacier displacement using
SAR Intensity tracking
6. CONCLUSION
ERS 1&2 tandem data of Gangotri and Siachen glacier were
processed for deciphering their movement. Good coherence
was observed over the glacier region using ERS-1/2 tandem
data and differential interferograms are generated. About 8.4
cmd-1 movements has been noticed in case of Gangotri glacier
and about 22.4 cmd-1 movement has been noticed in case of
Siachen glacier in LOS direction. Endeavor is made to time
series study using ENVISAT ASAR data which shows high
decorrelation in the snow bound Himalayan region. On the
basis of this study it can be concluded that InSAR based
movement determination using TS-X HS data over Himalayan
region is not feasible and hence SAR intensity tracking
approach is adopted. Gangotri glacier surface velocity is
observed as 9.37 cmd-1 using TerraSAR-X intensity tracking
method.
[6] Vijay Kumar Y.S. Rao, Singh Gulab, G. Venkataraman, Snehmani.2008.
Spaceborne InSAR technique for study of Himalayan glaciers using
ENVISAT ASAR and ERS data. Proc. IGARSS 2008, Bostan July 6-11,
2008, 1085-1088.
[7] Ferretti A., C. Prati, and F. Rocca, “Nonlinear Susidence RateEstimation
using Permanent Scatterer in differential SAR Interferometry,” IEEE
Transactions on Geoscience and Remote Sensing, vol. 38, no. 5, pp.
2202 -2212, 2000
[8] Roth, A.2004. TerraSAR-X Science Plan. Munich, Germany: DLR
Oberpfaffenhofen, 2004. (Online). Available: http://www.caf.dlr.de/tsx/
documentation/TSX-Science-Plan.pdf.
[9] Eineder,M., Runge,H. , Boerner,E., Bamler,R., Adam,N. , Schättler,B. ,
Breit, H. , Suchandt,S.2003. SAR INTERFEROMETRY WITH
TERRASAR-X, Proc. of FRINGE 2003 Workshop, Frascati, Italy.
[10] A.K. Gabriel, R.M. Goldstein, and H.A. Zebker, “Mapping small
elevation changes over large areas: differential radar interferometry,” J.
Geophys. Res., Vol. 94, No. B7, 9183-9191(1989).
[11] R. Michel and E. Rignot, “Flow of Glaciar Moreno, Argentina, from
repeat- pass Shuttle Imaging Radar images: Comparison of the phase
correlation method with radar interferometry,” J. Glaciol., vol. 45, no. 149,
pp. 93–100, 1999
[12] L. Gray, N. Short, K. Mattar, and K. Jezek, “Velocities and flux of the
Filchner Ice Shelf and its tributaries determined from speckle tracking
interferometry,” Can. J. Remote Sens., vol. 27, no. 3, pp. 193–206, 2001.
[13] E. Rignot, S. Gogineni,W. Krabill, and S. Ekholm, “North and Northeast
Greenland ice discharge from satellite radar interferometry,” Science,
vol. 276, pp. 934–937, May 1997.
[14] I.Joughin , “Ice-sheet velocity mapping: a combined interferometric and
speckle-tracking approach” ,Annals of Glaciology, Volume 34, Number 1,
1 January 2002, pp. 195-201(7)
[15]. A.K. Naithani, H.C. Nainwal, K.K. Sati, and C. Prasad,
“Geomorphological evidence of retreat of Gangotri glacier and its
characteristics,” Current Science, Vol. 80, pp.87-94, 2001.
Acknowledgement
We highly acknowledge European Space Agency (ESA) for providing
ENVISAT ASAR data and ERS data and German Aerospace Centre (DLR)
for providing TerraSAR-X data for this important study under the C, category
and AO scheme respectively.
V - 335