Sie sind auf Seite 1von 8

Engineering Geology Landslide Investigations and SAR Interferometry

Janusz Wasowski (1) and Paul Gostelow (2)


(1) CNR/CERIST- National Research Council, via Orabona 4, 70125 Bari - Italy;
wasowski@area.ba.cnr.it
(2) British Geological Survey, NG12 SGG Nottingham, UK; P.Gostelow@btinternet.com

ABSTRACT
We can distinguish three phases of using the
SAR and other EO satellite data in landslide
studies:
1. Surface characterisation, ie initial
topographic and geomorphological
classification, DEMs, photogrammetry (using
Spot and Landsat);
2. Measurement of slow to extremely slow
slope movements by means of SAR
interferometry;
3. Systematic monitoring of temporal-spatial
variations in some slide producing agents (eg,
relative change detection in soil moisture,
surface drainage, vegetation, and land-use).
The most serious limitation for the practical
or operational applications of current space
technologies in landslide investigations
appears to be the coarse resolution of satellite
images. Thus the available EO satellite data
could be best exploited for small scale
landslide investigations. Thanks to a regular
schedule and frequent coverage of large areas
by EO systems, satellite images will be
particulary attractive for regional landslide
hazard zonation and possibly for preliminary
warning. A reliable and timely warning
requires the use of a continuous site-specific
monitoring system. This is usually impractical
in most situations due to high costs of in-situ
operations. The development and routine
introduction of EO methods which employ
periodic change detection of slide producing
agents thus seem to offer a cost effective
alternative for providing initial wide area
assessments of ground deformation
susceptibility. This approach thus simply
targets those areas where there is a potential
hazard and where more detailed geotechnical
investigations may ultimately be required.

Keywords: engineering geology, landslide


investigations, SAR interferometry, user
requirements

Introduction

The scientific literature from this decade has


provided many examples of successful
applications of space-borne synthetic aperture
radar interferometry (InSAR) for detecting ground
surface deformations induced by seismic and
volcanic activity, as well as for the monitoring of
glacier motion (e.g. Massonet and Feigl, 1998).
However, to our knowledge, only in the last few
years have there been isolated cases indicating the
potential usefulness of SAR techniques in the
detection of landslide movements (e.g. Fruneau et
al. 1996; Ferretti et al. 1999 Rott and Siegel1,
1999). Leaving aside the time involved in data
processing and interpretation, it appears that the
encouraging results from these studies were
achieved thanks to particularly favorable
circumstances and environmental conditions
(large, scarcely vegetated and very slowly
deforming or moving slopes, availability of
interferometric pairs with short temporal and
spatial baselines, suitable viewing direction of
SAR sensor and absence of on-site precipitation
during or shortly before the SAR satellite pass).
Like some other natural and man-induced hazards,
slope movements cover an extremely wide range
of spatial and temporal scales. However, the areal
extent of the great majority of damaging or socioeconomically significant landslides is typically
many orders of magnitude smaller with respect to
that of other disastrous phenomena such as
earthquakes, volcanic eruptions, floods. Thus the
single most important factor which limits the
applicability of the currently available SAR and
other Earth Observation (EO) satellite data to
landslide (recognition and interpretation)
investigations is the coarse resolution of spaceborne imagery. This problem, and other important
limitations, (e.g. a short historic record which
reduces usefulness of EO data to landslide

susceptibility and activity evaluations), have been


thoroughly discussed by Soeters and van Westen
(1996).
Most previous uses of remote sensing imagery
have emphasised landslide and surrounding
terrain characterisation eg classification and
inventory mapping, geomorphic mapping ,
empirical factor mapping and hazard zonation
(see Soeters and van Westen, 1996 for a detailed
review). However, more integrated scientific
research, based on a deeper appreciation of
complex user requirements and focused also on
causative processes will be needed to foster the
practical use of satellite remote sensing data for
landslide hazard assessment and warning. We
address this problem from an engineering geology
perspective by discussing and presenting features
related to the physical mechanisms of landslide
initiation, which might be detected using InSAR
or EO satellite data in general.

Basic slope failure mechanisms: Factors of


Safety and the Shear Strength of Slope
Forming Materials

The susceptibility of slopes to failure is controlled


by many variables which change continuously
with time. However, only some of the
destabilizing factors have surface expression and
might be detected from space. This significant
limitation implies that the determination of
several very important aspects of slope stability
such as sub-surface geology, geotechnical
properties, ground water flow patterns, water
pressures, the depths and shape of slip surface will
have to rely on secondary (indirect) qualitative
indicators.
For engineering purposes, ie for remedial
design/construction it is necessary to assess slope
stability quantitatively and this has often proved
to be difficult even with good subsurface
information.
The limit equilbrium of slopes is usually
expressed in terms of an engineering factor of
safety F, where,
F=

Average Shear Strength (s)


Average Shear Stress

When F is equal to 1.0 an average shear stress is


equal to an average shear strength and a slope is at
a point of shear failure resulting in a displacement
and downslope mass-movement.
The average shear strength (s), of saturated slope
forming materials can be expressed in general by,

s = c' + ( u w

) tan

'

where c is the material cohesion, is the total


normal stress on a potential or actual failure plane,
uw is the pore water pressure and tan is the
effective friction angle.
In many parts of the world soils are not saturated,
especially in the summer months and an
alternative equation introduced by Fredlund
(1987) is now widely accepted for these materials,
where the shear strength (s) is given by,

s = c' + ( u a )tan ' + u a u w tan

where ua is the pore air pressure and uw is the pore


water pressure. The friction angle b is equal to
the slope of a plot of matric suction (ua - uw)
versus shear strength when ( - ua) is held
constant.
The average cohesion and friction angle in a slope
depends mainly on geology, landslide history and
to some extent vegetation, while spatial variations
in the normal effective stresses ( - uw) are
controlled by the depth of unsaturated zone,
vegetation, hydraulic conductivity and seasonal
groundwater movements (ie, climate). The
average in situ shear stresses depend on geometry
of the slope (height, slope angle) and unit weight
(bulk density) of slope forming materials.
For a landslide to occur there must be an increase
in i) the average shear stresses, for example
through erosion, earthquakes, man-made changes,
and/or ii) a decrease in average shear strength (s).
From the equations above it can be seen that this
can occur in either the c, , ( - ua), (ua - uw), or
( - uw) terms, although the latter, ie brought
about under saturated conditions by intense
rainfall is often a very common cause of
landsliding. These changes in shear stress and/or
effective strengths which precede landsliding
were referred to as slide producing agents by
Terzaghi (1950), but are also known as causative
and triggering mechanisms (eg, Cruden and
Varnes, 1996 and references therein).
It is well established that some deformation
usually precedes a slope failure, ie when F = 1.0
and Fig. 1 from Terzaghi (1950) illustrates
diagrammatically how in a susceptible slope
which he considered had an initial F 1.5, the
downhill displacement D, time t, and F depend on
the introduction of a final slide producing agent or
trigger.

Figure 1 also suggests that if the basic distribution


of susceptible natural slopes are already known,
for example through a general knowledge of
geology, slope angle and ground water conditions,
then there may be two obvious techniques of

providing warnings of impending slope failure at


F = 1.0: i) by detecting and measuring slope
deformation and ii) by monitoring the slide
producing agents.

Fig. 1. Diagram showing factor of safety changes and slip surface movements preceding, during and after a
landslide (after Terzaghi 1950).
Conventional ground surveying can be used for
both approaches, but this is often costly and
difficult to justify over large areas. The EO
techniques thus provide a possible solution for
firstly contributing to initial characterisation and
secondly for the long-term regular monitoring of
relative changes in factors of safety F, through the
detection of slope deformation and/or the slide
producing agents.
Assuming that the spatial resolution of future EO
systems will make InSAR interferometry more
attractive for landslide investigations, it is of
interest to briefly review how these might relate to
the assessment of slope instability and changes in
F for both first-time and reactivated mass
movements.
Terzaghi (1950) and Skempton and Hutchinson
(1969) divided landslide ground movements into
three phases,
1. Pre-Failure
2. During Failure
3. Post Failure

Pre-Failure Movements

The length of time required to progress along the


pre-failure part of the landslide ground movement
curve (Fig. 1) is clearly critical for warning and
safety purposes. In brittle geological materials
where there is a rapid loss of strength with strain,
the displacements can accelerate quickly at the
onset of failure and hence Terzaghi (1950)
suggested the distances O-D1 and O-t1 (Fig. 1)
would be small and the onset of failure very rapid.
However, in a more plastic material with a flattopped stress-strain curve this distance would tend
to be larger and with a correspondingly slower
onset to failure they might be more easily detected
by remote sensing techniques.
The term creep is used to describe time-dependent
pre-failure shear strains and/or volumetric strains
that develop through the viscous resistance of a
soils structure. However, despite a number of field
and laboratory investigations, the mechanics,
magnitude and significance of creep strains in
relation to mass movement and other processes on

natural slopes are still poorly understood.


Nevertheless all forms of ground deformation
which occur on slopes are of interest to the
engineer whether they ultimately lead to first-time
shear failure or not, but in most cases their
presence would suggest that a slope was close to
limiting equilibrium.
Gravity or Continuous Creep
Terzaghi (1950) was the first to make the useful
distinction between the shallow creep
deformations caused by seasonal processes on
older, degraded slopes and deformations related to
deeper seated gravity or continuous creep which
take place following unloading or stress relief
through downcutting and erosion in younger
geomorphological terrains.
Theoretical gravitational stress analyses of rapidly
formed, ie young cutting slopes in
overconsolidated soils demonstrated that creep
deformation might lead towards overall horizontal
surface movements and vertical settlements
behind slope crests following a stress relaxation
under natural conditions. Bjerrum (1967) and
Bishop (1967) suggested that the associated nonuniformity of stress and strain within a slope
deforming in this way might also lead to the
formation of progressive shear surfaces, generally
circular in shape and that the time scale of prefailure deformation along these would be
controlled by processes such as swelling,
weathering, slow releases of strain energy or
rheological strength reductions.
Potts et al (1997) used finite element analyses to
model strains in excavated cutting slopes
consisting of plastic clays which took into
account strain softening and coupled
consolidation/swelling. They demonstrated that
patterns of surface movements near slope crests
were similar to those predicted with earlier elastic
analyses and it was also suggested that under
these time-dependent controlling conditions they
could act as precursors to delayed progressive
failures and landsliding in high, steep slopes cut in
strain softening brittle soils.
How important such time-dependent strains
actually are, or have been in initiating longer-term
delayed landslide disasters in natural slopes is not
really known. The results of some studies,
however, have suggested that in certain geological
environments the slow development of deepseated strains arising from a strength reduction
due to swelling might also control gravitational
creep strains in natural slopes up to several
thousand years old (eg, Ballantyne et al. 1998).

The typical surface signs that creep and


progressive deformations were, or had been taking
place might be tension cracks at slope crests and
toe features near the base of slopes. With regard
to the detection of pre-cursory movements to
first-time landsliding using SAR interferometry it
thus seems that emphasis should perhaps be given
to processing images from steep natural slopes
which average 20 +, and/or have an identifiable
free-face or cliffed crest. Detailed analysis should
thus cover:
a) The upper parts of slopes (their crests)
b) The flatter areas behind slope crests
c) The base of slopes, especially where natural
erosion is taking place
Weathering and shallow seasonal creep
Weathering is another process which has been
put forward as a possible mechanism which
controls shallow deformation and thus pre-failure
times in soils on older, more degraded slopes
which may already have a surface layer of
colluvium. For geotechnical engineering
purposes, Chandler (1972) defined weathering as
the physical and chemical weakening processes
which lead to an increase in water content at
shallow depths. The former included seasonal
water content changes, frost heave and mass
movement and the latter oxidation/reduction and
leaching. His study of the Lias Clay in the UK,
together with several other geotechnical
investigatons (eg, Vaughan and Kwan 1984) have
shown that these processes reduce the strength
and stiffness of soils and that the accompanying
structural breakdown and particle disintegration
might lead to creep, surface settlements, swelling
and possible shear failure.
Tavenas and Leroueil (1981) discussed generally
how ground water fluctuations might also become
an important control of shallow, seasonal creep
strain rates. An example of this described by
Mitchell and Eden (1972) included downslope
inclinometer measurements to depths of 10m in
14m-29m high intact valley slopes of between
25.5 and 27.5 (ie, unfailed) cut in Pleistocene
marine clays in Canada. These averaged
25mm/year over a 4 year period, with the most
rapid movements always occurring during the
spring when groundwater levels were highest.
Carson and Kirkby (1972) considered that
solifluction was a rapid form of seasonal creep
produced by annual freeze-thaw cycles and/or
flow movements resulting from melting ice
lenses. Typical rates of movement of this process
on a variety of soil slopes (6 - 18) from
periglacial climates in NE Greenland, Colorado,

and Norway ranged from 0mm - 26mm/year


(Harris 1987).
Shallow surface slope creep deformations appear
to vary widely in magnitude on both
comparatively young and older slopes, but seem
to be closely linked to climate and hence would be
most likely to be detected most easily in wet
seasons on remotely sensed images. However, the
evidence presented here suggests there is a need
to reliably detect movements of perhaps only 1 or
2 cm per year before the technique could be used
to predict the onset of shallow instability with any
confidence.

Movements During a First-Time Slide


Figure 1 shows that the movements during a slide
occur comparatively quickly in comparison to
pre- and post failure deformation phases. The
actual rate of displacement will largely depend on
the shape of the stress curve of the materials
involved. If the shear resistance reduces rapidly
after the peak strength is reached, as in a brittle
soil, the slide will accelerate, reaching speeds of
up to 1.5 m/sec in some cases (Skempton and
Hutchinson , 1969). Such high rates are usually
the cause of most landslide disasters which result
in loss of life and/or property.
Conversely in a more plastic material with a flattopped stress-strain curve deformation will tend to
be much slower and more predictable. Terzaghi
(1950) quoted typical values of between
0.0008m/sec to 0.05 m/sec. However, in both
types of slide the deformation takes place over a
comparatively short timescale in a landslides
history and hence it seems unlikely that they will
be easily detected and monitored by periodic airor space-borne remote sensing (Figure 1).

Post-Failure Movements

After failure the strength of the slip surface will


be at residual with a very low or zero cohesion, c
and the factor of safety in a stationary slide will
be just above 1.0. In many naturally degraded
slopes, especially in clays, such surfaces are
commonly present and can become reactivated
through a variety of causes. Skempton and
Hutchinson (1969) noted that a common feature
of these subsequent movements is their low speed,
regardless of whether they are brought about by
seasonal water pressure changes or by an
alteration in loading on a slipped mass. They
quoted rates typically in the range of zero to up to
6m/year and they suggested these could continue
over long time periods, perhaps lasting tens of

years. This phase in a landslides movement


history (Fig. 1) is thus perhaps the easiest to
detect through periodic remote imagery,
especially in large complex slides where such
deformations may also extend over large areas
beneath slipped masses.
However, the majority of reactivated landslides
in areas of socio-economic interest are of limited
areal extent (generally within a few hundred
meters in length and 100 m in width). Figure 2
shows an example of a typical, medium-large size
mass movement recently reactivated in the
southern Apennines of Italy; the detailed ground
topographic monitoring record reveals the spatial
and temporal complexity of the reactivation
process. It is apparent that in this case, due to the
composite nature of the landslide and the
significant variations in the magnitude of
displacements, it might be very difficult to apply
profitably the present-day coarse resolution EO
satellite systems.

Slide Producing Agents (Causative and


Triggering Mechanisms)

Slope movements are complex phenomena, whose


time-space distribution results from an interaction
of several highly variable factors: geological,
geomorphological, physical, and human (eg,
Terzaghi, 1950; Cruden and Varnes 1996). A
combination of these slide producing agents
perhaps acting over a timespan ranging from days
to decades of years, may thus contribute to the
onset of O-t1 conditions (Fig. 1) following a final
trigger. Although not shown on the Terzaghis
original figure, the same final agent could also
lower the factor of safety of a pre-existing shear
surface and control post-failure displacement rates
also.
With regard to the routine wide-area use of
remote sensing imagery in landslide hazard
assessment, it seems that at present the detection
of these ground surface changes which reflect
different slide producing agents may be more
easily achieved than measurements of ground
deformation in both the pre O-t1 and O-t1 phases.
Where such changes have been mapped over a
period of time and examined in relation to
existing conditions, such as geology and slope
angle, perhaps by using Geographic Information
Systems (GIS), they can be used to warn of a) the
approach of possible O-t1 conditions for first time
failure and b) reactivation where the changes have
occurred in previously slipped ground

Fig. 2. Ground topographic monitoring record of spring 1995 reactivation of a landslide in the southern
Apennines of Italy (from Wasowski and Mazzeo, 1998). Horizontal and vertical movements refer to the
periods 25.03.95-18.05.95 (left) and 18.05.95-23.06.95 (right); numbers in brackets with negative and
positive signs indicate respectively subsidence and rise in meters.

Thus, through the use of suitable algorithms and


image classifiers the satellite data can be exploited
for a systematic recognition and quantification of
the changing surface features representing
potentially destabilising factors (eg land-use and
man-made modifications of slopes, topography,
soil moisture and vegetation). Taking the latter as
just one example, it is well known that rooting
depths, and plant-water requirements can
influence bulk soil cohesion and shallow effective
stresses (eg, Gray, 1995. These in turn influence
overall shear strengths and may significantly
affect values of F with regard to subsequent final
rainfall triggering mechanisms. Remotely sensed
images showing vegetation changes, for example
tree clearances, harvesting, or the substitution of
deep for shallow rooting crops can contribute
significantly in the identification of slopes which
may become susceptible to first time instability
from such intense rainfalls (Mc Kean et al, 1991).
Further work on adapting such imagery and
combining it within a GIS with the spatial and
temporal variations in the other slide producing
agents for routine use in landslide hazard studies
is perhaps needed to generate indexes which can
be use as warnings of potential instability from a
final trigger.

Thus the available EO satellite data (and those to


be acquired in near future) could be best exploited
for small scale landslide investigations. Thanks to
a regular schedule and frequent coverage of large
areas by EO systems, satellite images will be
particularly attractive for regional landslide
hazard zonation and possibly for preliminary
warning. This type of zonation (relative hazard)
does not require that both spatial and temporal
occurrences of mass movement are known.
Indeed, the accurate temporal predictions of slope
failures are often not feasible, even where data
from field investigations are available. A reliable
and timely warning therefore requires the use of a
continuous site-specific monitoring system.
However, in most situations this is usually
impractical due to high costs. The development
and routine introduction of EO methods which
employ periodic change detection of slide
producing agents thus seem to offer a cost
effective alternative for providing initial wide area
assessments of ground deformation susceptibility.
This approach thus simply targets those areas
where there is a potential hazard and where more
detailed geotechnical investigations may
ultimately be required.

Conclusions

Ballantyne CK, Stone JO, Fifield LK (1998)


Cosmogenic CL-36 dating of postglacial
landsliding at the Storr, Isle of Skye, The
Holocene, 8, 347-351
Bishop AW (1967) Progressive failure with
special reference to the mechanism causing it.
Proc. Geotech. Conf. Oslo 2, 142-150
Bjerrum L (1967) Progressive failure in slopes of
overconsolidated plastic clay and clay shales.
Proc. Jnl. Soil. Mech. Div. ASCE, 93, 1-49
Carson MA, Kirkby MJ (1972) Hillslope form
and process, Cambridge University Press, 475pp
Chandler RJ (1972) Lias Clay: Weathering
processes and their effect on shear strength.
Geotechnique, 22, 403-431
Cruden DM, Varnes DJ (1996). Landslide types
and processes. In Landslides. Investigation and
mitigation, ed AK Turner, RL Schuster.
Transportation Research Board Spec. Rep. 247.
Nat. Academy Press, 673pp
Fruneau B, Delacourt C, Achache J (1996)
Observation and modelling of the Saint-Etiennede-Tinee landslide using SAR interferometry.
Proc. Fringe 96

We can distinguish three phases of using the SAR


and other EO satellite data in landslide studies:
1. Surface characterisation, ie initial topographic
and geomorphological classification, DEMs,
photogrammetry (using Spot and Landsat);
2. Measurement of slow to extremely slow slope
movements by means of SAR interferometry;
3. Systematic monitoring of temporal-spatial
variations in some slide producing agents (eg,
relative change detection in soil moisture, surface
drainage, vegetation, and land-use).
The most serious limitation for the practical or
operational applications of current space
technologies in landslide investigations appears to
be the coarse resolution of satellite images. The
planned EO systems with finer resolution (5-10
meter or better) and stereo capabilities (cf.
Singhroy and Landslide Hazard Team/CEOS
Disaster Management Project, 1999) should make
the operational applications more feasible.
However, even these new systems will probably
not be sufficient for some site-specific
engineering geology investigations of landslides
that typically require sub-meter pixel resolution
(or 1:1000 or larger mapping scale approach).

References

Ferretti A, Prati C, Rocca F (1999) Permanent


scatteres in SAR interferometry. Proc.
IGARSS99 Conference, Hamburg
Fredlund DG (1987) Slope stability analysis
incorporating the effects of soil suction. In Slope
Stability, ed MG Anderson, KS, Richards. J Wiley
and sons, 648pp
Gray DH (1995) Influence of vegetation on the
stability of slopes. In Vegetation and Slopes, ed
DH Barker, Thomas Telford Press, 296pp
Harris C (1987) Mechanics of Mass Movement in
Periglacial Environments. In Slope Stability, ed
MG Anderson, KS Richards. J Wiley and sons,
648pp
Massonet D and Feigl, KL (1998) Radar
interferometry and its application to changes in
the Earths surface. Rev. Geophys., 36, 441-500
McKean J, Buechel S, Gaydos L (1991) Remote
sensing and landslide hazard assessment,
Photogrammetric Engineering and Remote
Sensing, 57, 1185-1193
Mitchell RJ, Eden WJ (1972) Measured
movements of clay slopes in the Ottawa area.
Can. J. Earth Sci., 9, 1001-1013
Potts DM, Kovacevic N, Vaughan PR (1997)
Delayed collapse of cut slopes in stiff clay.
Goetechnique, 47, 953-982
Rott H, Siegel A (1999) Analysis of mass
movements in Alpine terrain by means of SAR
interferometry. Proc. IGARSS99 Conference,
Hamburg
Singhroy V and Landslide Hazard Team (1999)
Earth Observation for Landslide Assessment report for CEOS Disaster Management Project,
http://disaster.ceos.org/landslide.htm
Skempton AW, Hutchinson JN (1969) Stability of
natural slopes and embankment foundations, State
of the Art Report, 7th Int. Conf. Soil Mech. and
Found. Eng. 291- 340
Soeters R, van Westen CJ (1996) Slope instability
recognition, analysis, and zonation. In Landslides.
Investigation and mitigation, ed A. K. Turner, R.
L. Schuster. Transportation Research Board Spec.
Rep. 247. Nat. Academy Press, 673pp
Tavenas F, Leroueil S (1981) Creep and failure of
slopes in clays, Can. Geotech. J, 18, 106-120
Terzaghi K, (1950) Mechanisms of landslides, in
Applications of geology to engineering practice,
Geol. Soc. Amer. Spec. Pub., Berkey Volume, 83123
Vaughan, PR and Kwan, CW (1984) Weathering,
structure and in-situ stress in residual soils.
Goetechnique, 34, 43-59
Wasowski J. & Mazzeo D. 1998. Some results of
topographic monitoring of the Acquara-

Vadoncello landslide, Italy. Proc. 8th Int.


Congress IAEG, Vancouver, Balkema, 1705-1712

Das könnte Ihnen auch gefallen