UDY
LE
oherence
Tom this
;ouple is
It is also
alysis of
is not in
seful for
perty of
sls.
rties or
nces on
2). The
atterns
ialue of
>lements
ric data
bility of
le, some
en rock
ome are
even at
: lements
of time.
| terrain
1 ERS-1
ondition
for ERS
1 couple
nts into
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002
different landuse-landcover classes by the previous workers
(Stussi et al, 1996). On the other hand, in a 35-day ERS
couple, stability of the terrain is much less because of its larger
temporal baseline and therefore it is not encouraged much for
terrain analysis. In this work, an evaluative study to compare
the utility of ERS tandem couple and 35-day couple using
temporal stability as a criterion for terrain analysis has been
attempted. Coherence images of the couples of longer temporal
baselines have also been studied in this work in an attempt to
characterize temporal stability of different terrain elements.
2. THEORETICAL BACKGROUND
The degree of interferometric correlation between the images is
known as coherence. The value of coherence depends on the
speckle patterns in two SAR images of the interferometric
couple. Speckle effect in SAR image results from the difference
in vectorial sum of backscattered energy from the elementary
targets or scatterers in the adjacent pixels. Number of scatterers
in a pixel depends on radar wavelength being used but their
differential temporal response depends on the nature of the
terrain as a function of stability of the scatterers with time.
When all the scatterers in a particular pixel remains more or
less stable for a given time interval, the vectorial sum of
backscattered energy will remain unchanged for that pixel in
the two images. Consequently, the value of coherence will be
very high (- 1). Temporal stability of the scatterers within a
resolution cell varies from one landcover to the other. For
example, water changes its physical shape and therefore the
stability of the scatterers within a resolution cell of water body
gets disturbed in milliseconds. The value of coherence is
therefore always very low for water bodies independent of the
temporal baseline of the repeat-pass couple. The similar
situation occurs in case of agricultural and heavily vegetated
areas for temporal baselines of several weeks. On the other
hand, scatterers in sparsely vegetated areas remain stable for
longer duration. Scatterers within a resolution cell of dry soil
and sand deposits remain generally stable with time unless
disturbed by natural factors e.g., storm, rain or human
interference which causes change in their positions or dielectric
properties. In bare rock outcrops and urban areas, scatterers
within a resolution cell may be considered to be more or less
stable irrespective of the time interval. So, the value of
coherence for such landuse-landcover classes is generally
always high independent of the temporal baseline of the couple.
The value of coherence also depends on geometric criteria
of the interferometric couple. This includes spatial baseline
between the two images and the areas of layover in the
individual images. With increasing effect of both the
parameters, the value of coherence reduces.
The value of coherence between two complex SAR images
Z, and Z; is calculated as described below:
*
EZ, Z5
Ez |g Iza] |
The expectation value operator 'E' is in practice
approximated with a sampled average over n samples as
described below:
y-
577
Xn ZZ
Ms
Asa zz
However, topography of the terrain contributes to the value
of coherence as a function of phase difference of the
interferometric image. The topography-compensated coherence
value is therefore truer representative of temporal changes in
the speckle patterns of the individual pixels. The topography-
compensated coherence may be estimated using topographic
phase contribution in the interferometric couple ($) as
described below:
Zn ZZ 558
n - ;
VEn|Z1| En|Z2|
3. MATERIALS AND TOOLS
In this study, ERS tandem couple and 35-day couple of a test
site of Mauritania were used to compare and evaluate their
utility for terrain analysis. Tandem couple was acquired in
ascending mode in April 1996. This has a perpendicular spatial
baseline of 111 m. The 35-day couple was acquired in
descending mode during June-July 1999. This has a
perpendicular spatial baseline of 829 m.
In addition, interferometric couples from parts of Iraq and Iran
were available in this study, which helped to gain better
understanding of temporal stability criterion of the terrain
elements. The interferometric data couples are as described
below:
(i) ERS tandem couple from a test site in Iraq, acquired during
May 1996, and a 1121-day couple of the same area acquired in
May 1996 and May 1999 were used in this study. The couples
have perpendicular spatial baselines 105 m. and 62 m.
respectively.
(i1) ERS 1121-day couple from a test site in Iran acquired in
May 1996 and June 1999 with perpendicular spatial baseline of
43 m.
Data processing for generating coherence image from the
interferometric couples was carried out using DIAPASON
interferometric tool, developed by CNES, France. Due to
unavailability of DTM, topographic contribution in coherence
values could not be compensated.
4. RESULTS AND DISCUSSION
It is observed that the overall coherence in land area is
strikingly higher in tandem couple (Figure 1a) than that in 35-
day couple (Figure 1b) of Mauritania test site. On the other
hand, coherence is very low in either of tandem couple or 35-
day couple due to the intrinsic unstable nature of water body at
millisecond interval. In the coherence image of the tandem
couple, various terrain elements such as active coast, low-lying
backshore region, arid open fields with sand deposits and
sparsely vegetated areas can be delineated. This is not possible
from the 35-day couple due to its low overall coherence.
However, in the 35-day couple, beach area, being the most
stable zone in the coast can be mapped nicely, which is not
possible from the tandem couple. Similarly, the urban area
(Nouakchott city area) can be demarcated in the 35-day couple
