The following NPA article was published in the November 1999
issue of Mapping Awareness. It introduces some of the principles,
issues and recent developments of satellite radar interferometry
specifically applied to ground displacement mapping.
New Developments in Wide-Area Precision
Surveying from Space
Mark Haynes, Chief Engineer, Applications
Development, NPA Group
From mapping the areal coverage and magnitude of ground motions
following cataclysmic earthquakes, to measuring discrete subsidence
displacements over several years, satellite Radar Interferometry
(InSAR) has a wide variety of uses in ground movement surveying
applications.
For several years now, the NPA Group, a satellite mapping
company based in Edenbridge, Kent, has been routinely detecting
and mapping to sub-centimetric levels many types of large-area
ground motions using InSAR techniques. As well as heading
major research projects in this field for the European Space
Agency and the European Commission, NPA is contracted to undertake
this work, not only in the UK but throughout the world, on
behalf of local government agencies, risk management companies,
commercial organisations such as mining companies, and geological
survey institutions.
The purpose of this article is to present some of the capabilities
and recent advances of satellite Radar Interferometry for
displacement mapping, illustrated with examples of our work
in this domain.
The physical concept is analogous to the school physics experiments
of Newton's Rings or Young's Slits. They all involve two coherent
sources of electromagnetic (e.m.) radiation physically spaced
apart a small distance, which, because of the precise, specific
geometry, combine to create bright and dark patterns of interference,
or fringes.
Where spaceborne radar interferometry differs from these
experiments is firstly that it employs microwave radiation
from imaging radar instruments on board orbiting satellites
and not visible light; secondly, the two sources in radar
interferometry consist of two separate imaging radar passes
over a common area of the Earth's surface. The satellite imaging
radar, or SAR, instrument works by beaming microwave radiation
to the Earth's surface and recording the backscattered signal
reflection from points on the ground.
Orbiting radar platforms particularly suited to InSAR include
the European Space Agency's ERS-1 and ERS-2 satellites, which
have created a growing archive of imaging radar data stretching
back to 1991. Other platforms are also available, and continuity
of data into the future is ensured by forthcoming satellite
launches, such as ESA's ENVISAT satellite. Using two radar
scenes of a common area of the Earth's surface (typically
100 km × 100 km) acquired by these satellites on two
different occasions in time, we can generate an interferogram
of the scene. This interferogram is created, through a complex
and computer-intensive process, from the phase (or wave) information
of the radar data and consists of fringes, which cycle from
black, through shades of grey, to white. These cycles of grey
represent phase differences between the corresponding reflected
signals, for each common ground point, acquired at the two
different satellite passes. As in the physics experiments,
these fringes result from the physical separation, and hence
the slightly different imaging angles of the radar instruments.
This separation exists because repeated satellite orbits and
sensor pointing are not perfectly coincident, though the position
and orientation of the sensor are very precisely known and
taken into account in the InSAR process.
Within each fringe, a given shade of grey represents a particular
phase difference, and all points of a particular shade will
exhibit the same phase difference. From the precise knowledge
of the acquisition geometry and radar sensor parameters, ground
points characterised by a constant phase difference within
a fringe represent a constant path length difference (measured
in actual wavelengths from the satellites to those ground
points), which in turn corresponds to points of constant ground
elevation. Hence, an interferogram over a volcano, for instance,
would appear as fringe contours (slightly irregular, due to
the varying topography) encircling the crater peak and increasing
in perimeter down the volcano sides.
A major application of SAR interferometry is thus in the
generation of Digital Elevation Models (DEMs), since the contour
fringes of constant ground elevation can be converted to heights.
An enhancement of the technique, differential SAR interferometry
(difSAR) relies on the fact that the interferometric fringes
comprise not only phase difference fringes due to the ground
topography but also phase differences resulting from displacement
(if any) of the ground. Using existing DEM height data, the
difSAR process removes topographic fringes to yield a differential
interferogram, in which any fringes present correspond solely
to ground movement between the two radar acquisitions.
Each displacement fringe in a differential interferogram
is a result of the satellite-to-ground path length difference,
and is measured down to units of wavelengths of the radar
sensor, which, for the ERS satellites, is 5.6 cm. Again, from
the satellite orbit acquisition geometry and the physical
parameters of the radar sensor, each distinct fringe cycle
in the differential interferogram actually represents 2.8
cm (half a wavelength) of movement in the look-direction,
or line-of-sight of the sensor on the satellite platform.
Moreover, measurements can be derived from incomplete or portions
of fringe cycles, i.e. a half- or a quarter-cycle represent
1.4 cm or 0.7 cm of displacement respectively. Considering
the radar sensors are orbiting several hundreds of kilometres
above the Earth, this precision of measurement is remarkable,
even incredible, but this is nevertheless the case, as has
been shown by the corresponding a priori ground truth (where
available) and numerous ground validations.
The inset of Fig. 1 is a differential interferogram generated
from two radar scenes covering the region and straddling in
time (i.e. one either side of) the magnitude 6.1 earthquake
near the town of Dinar in Turkey, on October 1st
1995. The contours of ground movement have been reproduced
onto optical satellite imagery of the corresponding area:
moving inwards, a summation of the individual 2.8 cm fringes
indicates a maximum displacement at the epicentre of the order
of 56 cm. Furthermore, the varying fringe separations (coded
from green through yellow to red) indicate the increasing
severity of movement over unit distances.
In remote, inaccessible areas, this rapid damage mapping
system can enable disaster relief agencies to assess the areal
extents of destruction on the ground and to prioritise efforts
and resources accordingly. Scientifically, this work has provided
additional information to earthquake scientists to improve
the formulation of seismic parameters and modelling of specific
earthquake events, which are crucial to understanding their
re-occurrence and behaviour.
Fig. 3 is an extract from a large differential interferogram
of an area in the north-east of England. The period bracketed
by the two SAR images is 35 days and numerous, mainly circular
phase anomalies were revealed, as highlighted with red rings.
Line-of-sight displacements for these subsidence features
range from 28 mm to 84 mm. Varying in extent from a few thousand
square metres to almost one square kilometre, many of these
features are associated with coal mining, known to be active
over this particular period. It is of great significance that
many of these are in close proximity to - or even underlie
- major roads, railway lines and urban communities. (NPA is
grateful to Mr Dick Stow of Doncaster College who provided
us with the schedule of mining activity further north in Selby).
The differential
interferogram in Fig. 4 has been colour-coded for clarity and overlaid
on an optical satellite scene to reveal the extent of urban subsidence
in Las Vegas, Nevada, in the USA. The subsidence phenomenon in this dry
region is a result of groundwater abstraction. The magnitude of this
subsidence is of the order of 9 cm over 3 years, concentrated on the
outskirts of Las Vegas. Detecting and mapping the extents and scale of
subsidence activity in urban areas is of vital use to city planners'
and civil engineers' development projects, as well as to the water authorities
in order to assist with them with mitigation plans.
Issues of Coherence
The
quality of an interferometric result is a function of several factors,
one major influence being coherence. Coherence is the degree of
similarity of backscattering response (or reflection characteristics,
as measured by the SAR sensor) between corresponding ground cells in
both SAR images of an InSAR pair. That is to say regions exhibiting a
high level of coherence are composed of ground cells whose backscattering
responses have changed very little over the intervening period between
the two SAR image acquisitions. Such stable ground surfaces include dry,
bare rock regions, and built-up urban areas.
Conversely,
low coherence is a feature of surfaces whose backscattering responses
fluctuate to a high degree and provide little similarity, or correlation,
across an InSAR pair of radar images. Notable examples include dense
vegetation and water bodies, which merely yield speckled, incoherent
noise in the interferogram. In general, where an interferogram has been
generated to encompass ground motions, displacement fringes will be most
prominent across areas corresponding to moderate to high coherence, whereas
fringes will tend to break up or even become lost in noise in regions
of poor coherence.
Whilst
regions of high coherence can remain so for many years, the problem of
low coherence regions is exacerbated as the time separation is increased
between the two scenes of an InSAR pair - a phenomenon, known as temporal
decorrelation. This may be a feature of differential interferograms
where, for instance, very slow rates of ground motion, e.g. a few mm
per year, dictate that an InSAR pair spanning several years is required
for displacement fringes of the order of 1cm to be observed in the result.
Methods now exist to overcome to some extent this shortcoming, and NPA
has been successful with these, some of which are discussed on the next
page.
In
areas of low coherence, conventional difSAR can be supplemented by measurements
made on individual artificial or natural reflecting features on the ground
such as buildings or custom-built corner reflectors. InSAR techniques,
applied to this network of distributed, stable point features can allow
the enhancement of difSAR measurements to the millimetric scale.
The
image of West Houston in Fig. 5 shows red contours of subsidence over
a 2-year period extracted from the differential interferogram (inset).
The signal is lost over rural, agricultural regions and so subsidence
contours (dashed) have been extrapolated from known groundtruth. However,
within these regions of poor coherence and signal loss, the yellow points
represent stable ground features, whose response does not degrade with
time, and which may be used as permanent point scatterers. These can
yield measurements in rural, low coherence areas and replace the need
for costly, in-situ monitoring devices.
In
parallel to this natural reflector work, NPA is heading a major, two-year
research program for the European Commission, in collaboration with
several European partners. This program is studying the application
of difSAR methods on discrete, ground-fixed corner reflector devices,
which appear highly visible to the radar sensors, as illustrated in
Fig. 6. A corner reflector consists of a 2 or 3 metre-sided metallic,
trihedral structure whose open base faces the satellite radar beam
and, because of its particular geometrical design, reflects the incoming
beam straight back to the sensor with relatively little power loss
compared to the ground surface in its vicinity. Using difSAR, millimetric motions
of a network of these devices can be detected, corresponding to the
motion of the Earth's surface, such as creep between tectonic plates,
over which the corner reflectors are deployed. In conjunction with
leading world authorities in geodesy and geophysics, this pioneering
work is being undertaken to benefit the Greek government's Earthquake
Planning and Protection Organisation, in order to monitor pre-seismic
motion that occurs on the millimetric scale and is associated with
earthquakes.
As
well as surveying the ground for displacements, the use of corner reflectors
together with conventional differential SAR interferometry, is equally
and very importantly applicable to monitoring the discrete movement
of engineering structures, such as dams, tunnels, quarries, towers,
pipelines and oil and gas drilling rigs. NPA is confident that its
demonstrated difSAR expertise, products and research endeavours will
provide an additional, invaluable displacement-surveying tool to surveyors,
civil engineers and in many other disciplines.
NPA
would like to thank the European Space Agency and the British National
Space Centre for partly enabling some of this work, and is grateful to
its many collaborators and associates, including the UK Defence Evaluation
and Research Agency, Phoenix Systems, WS Atkins Consultants Ltd and the
University of Oxford's Department of Earth Sciences.
