-
Proceedings, Slope Stability 2011: International Symposium on
Rock Slope Stability in Open Pit Mining and Civil Engineering,
Vancouver, Canada (September 18-21, 2011)
Back Analysis of Landslides Triggered by Earthquakes Some
Implications for Future Practice
W. Murphy School of Earth & Environment, University of
Leeds, United Kingdom
R.N. Parker Department of Geography, University of Durham,
United Kingdom
G. Hancox GNS Science, 1 Faiway Drive, Avalon 5010, PO Box
30-368, Lower Hutt 5040, New Zealand
Abstract Earthquake induced landslides are an important
secondary effect of strong earthquakes. While the driving forces
for such instabilities (e.g. topographic amplification) have
received considerable attention, the actual mechanics of how such
slope failures occur has not. Current thinking suggests that the
weakest sector of the slope will fail. Field evidence does not
always support this assertion. Analysis of three landslides
triggered by the 1929 Ms=7.8 Murchison Earthquake indicates that
the geometry of the landslides are inconsistent with the critical
slip surface for ambient stress conditions. The landslides at
Little Wanganui Head and Whitecliffs show critical slip surfaces
that do not match the locations or geometry of the subsequent
failures. Analysis of the Goat Landslide also indicates critical
slide surfaces that are inconsistent with the actual slope failure,
but the existence of a disturbed zone, either from weathering or
unloading, can be used to explain geometric differences.
It is suggested that zones within rock slopes that are
characterised by low shear velocities are a more significant
control on the stability of rock slopes during earthquakes than the
ambient stress conditions. The implications of this on-going
research may impact substantially on the design of mine slopes in
seismically active regions.
1 Introduction The current thinking that underpins the analysis
of slopes during earthquakes is relatively simple. The model
presumes that for almost any earthquake induced landslide (EIL),
the slope behaviour is governed by the weakest slope configuration.
To put this is slope stability terminology, it is the critical slip
surface for ambient effective stress states that will slide during
an earthquake. The model is further developed to outline how, at
some threshold level, the slope may deform when subjected to
seismic shaking. It is this model that underpins the work of
Newmark (1965) and which has effectively been developed by other
workers who have considered how to analyse the stability of slopes
dynamic conditions. The principle outlined here applies to both
soil or rock slopes and has been employed at site specific and
regional scales (see inter alia Sarma, 1981; Ambraseys and Menu,
1988; Ambraseys, & Srbulov, 1995; Jibson, 1993; Miles and Ho,
1998)
The other major issue associated with EIL is that the forcing
function that is driving slope failure i.e. the earthquake ground
motion is in many cases poorly described. For analytical purposes
this can be described as a seismic coefficient a peak horizontal
ground acceleration (PGA) or an Arias Intensity (Arias, 1970). True
accelerograms can also be used employing a variety of methods. From
an analytical perspective these tend to be conservative, however,
due to the fact that many of these estimates have been derived from
free field earthquake records few incorporate topographic
amplification (Paolucci& Rimoldi, 2002; Sepulveda et al. 2005).
The absence of topographic amplification effects may significantly
underestimate ground motions acting on slopes. The work by Ashford
et al. (1997) and Ashford and Sitar (1997) provided a valuable
insight into the processes involved. Further challenges to
effective assessment of slope stability during earthquakes has been
outlined by Murphy et al (2002).
While there are specific problems with a variety of
implementations that have been outlined by Murphy et al (2002), the
basic presumption that the weakest sliding block is the one that
will fail during an earthquake has
-
never been really questioned. In this paper the stability
conditions relating to slopes which subsequently suffered from
earthquake induced landslides have been evaluated. A comparison is
made with the geometry of the critical slip surface derived from
numerical stability analysis for ambient stress states and the
geometry of the landslide mass that actually slipped during the
1929 Murchison (Buller) Earthquake in South Island, New Zealand.
This comparison is considered in terms of the implications for the
analysis of slopes during earthquakes, and whether the current
approach is justified in the majority of cases.
2 Data sets used and methodology At each site a ground model was
constructed based on field observations and limited a priori
information. Rock mass data were collected from using subjective
assessments and rock mass classification. These data were then
validated using a test site near Karamea (see Figure 1 for
locations) where stability conditions were assessed from field
evidence. Additionally the methodology outlined by Banks (2005) to
evaluate rock mass classification from each site using
geomorphological data was used to provide additional site specific
data. In addition to the Karamea sub division test site a more
complex FEM model of the current geometry of the Little Wanganui
Slump (marked as 1 on Figure 1) to match predicted stability
conditions to observed stability conditions as a further control.
Table 1 shows observed material and rock mass classification data
in comparison with those required to reach limit state. GSI
observations were made on an unweathered section to allow a better
representation of rock mass state at the depth of the landslides
which may have been as much as 300m. Input values were derived from
Geological Strength Index (Hoek and Brown, 1997).
Table 1. Observed and calculated rock mass properties. Intact
strength properties were estimated from field indices (material
could be broken easily by hammer) and the Mohr-Coulomb parameters
are calculated in RocLab. *The calculated GSI is derived from the
method outlined by Banks (2005). The modelled parameters are those
that are required for limiting stability at the Little Wanganui
Slump site at a range of depths from 20 to 100m selected on the
basis of the likely depths of the slip surface. UCS observations
are constrained based on Read & Millar (1990).
Parameter Observed Calculated Modelled
Intact strength (MPa) 1-5 5
GSI 80 82* 85
Cohesion (kPa) 181 432
Friction ( o) 28 32
Anisotropy cohesion (kPa) 0 0
Anisotropy Friction ( o) 15 15
Mass unit weight (kN/m3) 26
-
Figure 1. Landslides triggered by the 1929 MS=7.8 Murchison
Earthquake (from after Hancox et al. 2002).
2.1 Little Wanganui and Whitecliffs landslides The Little
Wanganui Slump (LW) is a large rock slump (Cruden and Varnes, 1996)
located approximately 41 km NNW of the epicentre of the 1929
Earthquake (Table 2). However it is c. 15-20 km to the west of the
likely projection of the White Creek Fault. The freefield
acceleration was therefore likely to have been higher than that
suggested by attenuation models that use the epicentre as a point
source.
The geological model for this site was of an almost uniform
succession of light grey, fine to medium grained, weak, siltstones
and marls that weather to a yellow, fine grained, extremely weak
surface material. Field index tests of these materials at the back
of the Little Wanganui slump indicates Unconfined Compression
Strengths of the order of 0.5-5MPa (crumbles under firm hammer blow
and can be peeled with a pocket knife). There is little evidence on
this landslide of ongoing deep seated landsliding. Using the method
outlined by Banks, (2005) the slope angles of the scarp indicate
RMR values of the order of 65-75. Field observations of GSI (Hoek
and Brown, 1997) tend to suggest value of the order of 75-80. The
major, persistent discontinuity present in the mass was bedding
plane fractures. These dipped at c. 10-15o to the NE. The
significance of these bedding planes in the formation of the
Glasseye Landslide is a dip slope failure at the back of the little
Wanganui Slump and is noted by the clear bedding plane control seen
on aerial photography. That bedding is better exposed in the
Whitecliff landslide to the south.
Goat Slide
Whitecliffs SlideLittle Wanganui Slump
-
Table 2. Landslides used in this study from Hancox et al 2002.
(*classification according to Cruden and Varnes, 1996).
Both landslides outlined in this section are coastal failures
where there is evidence of active coastal erosion and ongoing
instability.
The geological model at this site had to allow four identifiable
conditions (see Figure 2):
1. The presence of single and multiple rock block falls, and
shallow slides on the coastal slopes;
2. The presence of shallow debris slides from weathered material
forming on the main scarp of the landslide and evidence of
instability on occasional exposed blocks showing small
displacements;
3. The presence of water ponding at the foot of the main scarp
indicating a high water table at this point;
4. The presence of a thick and undisturbed vegetation blanket at
the foot of the main scarp and two secondary scarps within the
landslide system indicating the lack of movement at these
points.
Using Geological Strength Indices as a starting point for
estimating apparent values of cohesion and friction a series of
back analysis were run to establish a set of strength conditions
that allowed for the modelled conditions shown in Table 1 to
replicate the stability conditions outlined above using the finite
element code Phase2 v7 from Rocscience to carry out the analysis.
The initial estimate of GSI as observed in the field resulted in
stability conditions that met conditions 1 and 2 above, but not
condition 4 as movement was indicated on the main slip surface.
Input parameters were systematically increased within the range of
uncertainty of the field observations until all stability criteria
were satisfied. Estimates of mi were derived from using the ratios
of Unconfined Compressive Strength to Tensile strengths outlined in
Read and Miller (1990) as this has been found to yield a good
approximation (Read pers comm.) and a value of 4 was used. Figure 3
shows the outputs of the finite element model for the current
stability conditions representing limit state conditions.
Uncertainty associated with elastic properties (=0.25, E=10GPa) and
the associated flow rule means that the magnitude of the predicted
displacements must be treated with some caution.
Name Size (x 106m3)
Epicentral distance (km)
Geology Type* Slope angle (Dip/Dip Direction)
Little Wanganui Slump
210 41 km NNW Tertiary calcareous mudstones/siltstones with
limestone interbeds. The materials are generally weak and weather
to a buff yellow, medium grained extremely weak soil,
Rock slump 20 o/315o
Whitecliffs
Slump 120 39 km NNW Rock slump 50 o/315o
Goat
3.8 26 km N Rock slide /
avalanche 30o/135o
-
Figure 2. a) top left - Looking down the LWS; b) top right - the
body of the LWS the break in slope in the
centre of the image is the location of a secondary slip surface;
and c) bottom - The main scarp of the LWS showing water at an
elevation of c. 200m above sea level.
Figure 3. Phase 2 v7.0 Finite Element Model of the Little
Wanganui Slump. Sliding can be seen on the
seaward sections of the slope (c. 150-200m horizontal distance)
and small displacements can be seen on the main scarp. No movement
is indicated at the daylighting slip planes at c. 400m, 820m and
1090m horizontal distance. This corresponds with field
observations.
2b
Second slideblock
2c
2a
main scarp
main slide block
second slideblock
landslide toe
ponding indicatingbacktilted block
third slideblock
scarpscarp scarp
100
200
300
400
00
200 400 600 800 1000 1200 1400 1600 1800
Horizontal Distance (m)
Elev
atio
n (m
)
Little Wanganui Slump - current conditionsMaximum displacement
1.3mat eroded toe on the coast Main scarp (see fig 2c)
Groundwater observedat land surface
Secondary scarp (see fig 2b)
Secondary scarp (see fig 2b)
Toe (fig 2a)
?
??? postulated slip
surfaces
-
While Figure 3 relates to testing the model parameters against
the stability conditions for the current slope surface, to look at
the evolution of the slumps it is necessary to reconstruct the
slope profile prior to failure. In the absence of pre-1929
topographic maps of this area, the pre-earthquake long profile was
constructed by two complimentary methods. The first was by simply
extending the current contours across the landslide system,
connecting the 20m contour intervals at each side of the landslide
system, partly informed by the slope geometry in other sections of
the coastal land system. The second is by rotating the block back
up the slip surface. This process was carried out for both the
Little Wanganui and Whitecliffs landslides. The latter approach was
not possible at the Goat landslide as the whole slide mass had
effectively vacated the scar. In this case the contours were
extended around the slide mass. Sensitivity analysis was used to
investigate the impact of errors in long slope profile associated
with these approaches, and it was found that the variation in
modelled slip surface were significantly less than the errors in
rock mass properties and groundwater. Groundwater was considered to
be below the slip plane in the Goat Landslide.
Figure 4. The Whitecliffs Slump looking south along the coast
(4a) with the rotated block to the right hand
side of the photograph. Erosion of the main scarp can be seen as
shallow earthflows on 4b. The bedding can be seen tilting backward
into the main scarp at the top 4b.
2.2 The Goat Landslide The Goat Landslide (see Figure 1) is a
predominantly planar failure surface and should in theory at least
be the easiest landslide to model. Figures 5a-f show the landslide.
Available evidence suggests that the rock mass is somewhat blockier
than that seen at LW and the WC landslides. However, bedding plane
fracture traces seen in the rock face suggest bedding that dips
into the slope. Initial assessment indicated that dips would be of
the order of 25o and greater, however it is recognized in the
absence of good 3-D control that such estimates of orientation
could be considerably steeper. The bedrock geology was a marly
limestone similar to that seen elsewhere in the area. Sensitivity
analyses were carried out to look at the impact of using different
shear strength parameters for this landslide and these did not
improve the fit between the actual and predicted landslide
geometry.
There was however clear visible differences between the rock
mass observed in the flanks of the landslide and the basal shear
plane evident by the vacated scar and there is some indication that
bedding at the top of the slope is much more steeply inclined than
that seen towards the base of hill. There are two additional
observations that are worth making about the rock mass at this
site. Firstly, it is evident that the top 20m or so of the rock
mass is disturbed. Although disturbance here is natural in origin,
a D value of c. 0.5 (Hoek et al 2002) could easily be ascribed to
the rock mass (Figures 5c, 5e) based on visual inspection and a
comparison D values for engineered rock masses. This degree of
disturbance can be seen in the flanks of the vacated landslide,
while the slip surface exposed in the scar shows significantly less
disturbance and the rock mass is more compact (Figure 5d). The
implications of this observation will be discussed later.
4b4a
Backtilted block
Main backtilted block
Main scarp
Partial obscuredsecondary slipplane
Regionalbedding dip
~15o
~30o
Backtilted bedding dip
Main scarp
-
3 Back analysis of earthquake induced landslides
3.1 Analyses of the landslides The aim of the analysis of each
of these landslides was to identify whether two common methods of
slope stability analysis yielded a reasonable prediction of the
sections of the slopes that failed during this earthquake.
5a 5b
5c 5d
Slip surface
Slip surface
~25o
~25o
Regionalbedding dip
Dis
turb
ed ro
ck
mas
s D
~ 0.
5
Undisturbed rockmass below slip plane
10m
10m10m
5e 5f
Figure 5. The Goat Landslide c. 26km from the epicentre of the
1929 Murchison Earthquake. The overview of the landslide can be
seen in 5a and the debris that remains on the largely vacated slide
scar can be seen in 5b-5f. The left flank of the landslide showing
the disturbed rock mass can be seen in 5e while the less disturbed
rock mass forming the slip plane is evident in 5f.
-
These methods are Janbus (1954) Method and Bishops (1955)
Method. The analyses were carried out using the Slide program from
Rocscience. Field observations, limited strength data and the
technical literature were used to obtain input parameters for the
analysis. Two broad groups of analyses were used: the first used a
Mohr-Coulomb strength model with c and derived from rock mass
classification; the second used an anisotropic strength model with
a c and being the same as that used in the Mohr-Coulomb model, but
with a strength anisotropy dipping parallel with the bedding plane
fractures. These were assigned strength values of c=0 and =15o.
This value was selected on the basis of angles of repose bedding
plane controlled landslides in the area.
3.2 Little Wanganui and Whitecliffs Slumps The results of these
analyses in the absence of seismic accelerations can be
conveniently described together. Figures 6a and 6b shows the
results of the stability analyses for the Little Wanganui and
Whitecliffs Slumps. Both of these analyses use a reconstructed
slope angle to consider the most critical slip surface for sliding
analyses were run using both non-circular and circular slip surface
options to investigate the goodness of fit. In both cases an
anisotropic strength model was used allowing for a set of
weaknesses orientated dipping into the slope. This anisotropy is
consistent with field evidence and explains the stability
conditions seen at the Glasseye Landslide some 500m to the East of
the Little Wanganui Slump.
In the case of the Little Wanganui Slump, the critical slip
surface yields a value of 1.04 using Bishops method of slices. This
slip surface would produce a main scarp at about 550m along the
cross section. This marginal Factor of Safety (F) is significant,
since it indicates that even relatively low ground accelerations
would result in failure of this lower block and subsequent
unloading of the upper sections of the slope. However, whilst the
geomorphology supports progressive failure of this slope during the
earthquake, the location of the scarps is not consistent with the
critical slide surface indicated by the analysis. In fact the slip
surfaces closest to those observed from field data all have
significantly higher values for F ranging from 1.67 to 1.93.
Modification of the slope long profile and re-analysing the
stability conditions allowing for the removal or partial removal of
the block bounded by the critical slip surface shown in green in
6a, yields values of F ranging from 1.1-1.4 depending on the degree
of removal of the sliding block. The most seaward section of this
slump has not completely vacated the scar and so modelling such
conditions is not supported by field evidence. A non-circular slip
model yields a slightly better estimate of current slip surface
geometries than a circular one for the Little Wanganui Slump. This
is not the case with the Whitecliffs landslide where a circular
slip model yields better results.
The conditions at Whitecliffs are somewhat simpler due to having
only one or possibly two slip surfaces. A consideration of the
observations highlighted in 4a demonstrates that there is a large
backtilted block but there may be a small secondary slide,
indicated by a variation in bedding plane dips (Figure 4b) as a
result of unloading by the main slump. The analysis shown in Figure
6b again shows a critical slip surface at the over-steepened sea
cliff. The factor of safety (1.22) indicates stability for ambient
stress conditions. Analysis of the closest approximation to the
actual slip surface assuming one large and a smaller secondary
block indicates values of F of the order of 2.39 and 2.49. The
larger slip surface encompassing the whole slide mass does not fit
well with the available field evidence.
3.3 The Goat Landslide The Goat Landslide provides a somewhat
different state of conditions. The slopes are considerably higher
and there is no marine erosion to complicate stability conditions.
Additionally, given the nature of the slope it is likely that
groundwater is likely to be well below the landslide shear plane as
seen in Figure 5. Observations made in the exposed rock faces
indicate bedding dipping into the slope, there does appear to be
differences in the discontinuity pattern at the bottom and the top
of the slope. Therefore these were modelled as an anisotropic mass,
with different anisotropy orientations (Figure 6c) therefore were
treated as two different materials. The overview of the landslide
seen in Figure 5a indicates a slip surface c. 20m deep. The
variations in fracture pattern seen in Figure 5e and 5f are
evident. Additionally there is a significant degree of rock mass
disturbance
-
evident in 5e. In this example two rock masses units were
considered reflecting different bedding fracture orientations
within the material.
Analysis indicates that the critical slip surface (F=1.25) is
broadly consistent in shape (i.e. significantly more planar) with
the landslide scar, but is far too deep (c. 65m). Shallower slip
surfaces show high Factors of Safety, with slip surfaces of
consistent depth to that observed to have been triggered by the
1929 earthquake to show F>2.
Figure 6. Stability analyses of the Little Wanganui Slump (6a);
The Whitecliffs Slump (6b) and the Goat
Landslide (6c). Stability analyses show high factors of safety
with the exception of a potential slide block on the Little
Wanganui Slump. The somewhat blocky nature of the slip surfaces
highlighted in black in 6a relate to mudstone forming the
anisotropy within the rock mass.
4 Discussion What is apparent from the analyses of these three
landslide systems is that standard limit-equilibrium methods do not
replicate the seismic slope stability conditions well. There are a
number of possible reasons for this that is worth
consideration.
1.98
1.73
1.25
550
620
690
760
830
900
970
10 80 150 220 290 360 430 500 570 640 710 780
Ele
vatio
n (m
)
Horizontal distance (m)
1.00
Fact
or o
f Saf
ety 1.20
1.40
1.60
1.80
2.00+
1.71
Ele
vatio
n (m
)
Horizontal distance (m)
1.22
2.392.402.49
80
160
240
320
400
70 210 350 490 630 770 910
Fact
or o
f Saf
ety
1.00
2.00
3.00
4.005.00
6.00+
closest slip circle
approximation
Ele
vatio
n (m
)
Horizontal distance (m)
1.674
1.929
1.76
1.93
1.04
150
0
300
450
Fact
or o
f Saf
ety
1.00
1.20
1.40
1.60
1.80
2.0+
0 150 300 450 600 750 900 1050
1.67
6a
6c
6b
-
Firstly, there is the very reasonable consideration that the
assumptions that have had to be employed in these analyses are
wrong. In order to consider the impacts of these, sensitivity
analyses was carried out on the output of each model. The main and
most obvious potential error is that groundwater has been
incorrectly located. Groundwater level was selected to mirror the
land surface that was reconstructed from topographic data and
knowledge of the geometry of sliding blocks (in two of the three
cases). In none of these landslides was there evidence of surface
seepage emerging from the slopes in the near vicinity, and this
does tend to suggest that water levels are normally low.
Sensitivity analysis indicates that an error in the groundwater
level would result in a change in Factor of Safety for the critical
and actual slip surfaces, but did not significantly change the
geometry or location of the critical slip surface. There is no
evidence to suggest that there was an ongoing stability problem
prior to the 1929 Murchison Earthquake in these slopes. It is
however recognised that the area was sparsely populated.
Secondly, there is the likelihood that errors in rock mass
classification and parameter selection has resulted in significant
miscalculation. This is in some respects more likely. Material
strength was estimated from field index testing and technical
literature. As such they are prone to wide uncertainties.
Additionally, the rock masses evident in Figures 2, 4 and 5 cannot
be said to be homogenous and isotropic. In fact a significant
anisotropy was introduced onto the rock mass in the form of weak
layers within the stratigraphy that was evident at the Little
Wanganui and Whitecliffs Slumps. In an attempt to overcome this
problem, the selected parameters to be used in the analysis of the
failures were used to evaluate the current stability conditions at
Little Wanganui Head. In order to replicate the stability
conditions seen i.e. only localised instability on the steeper
slopes, it was actually necessary to increase strength parameters,
so the initial rock mass classification and parameter selection was
possibly tending towards the conservative. This probably stems from
the weathered nature of the materials, and the rock mass quality is
likely to be better at depth. Sensitivity analysis demonstrated
that large errors were required to significantly change the
critical slip surface geometry (large in this case being 500-600kPa
in cohesion and 4-5o in terms of friction angle). However while
these changed slip surface geometry they did not get the location
of the slip surface correct. In the Goat Landslide, the changes in
strength actually produced a slip surface less like that observed
in the field.
The third possible explanation is that even with the ground
perfectly characterised, the methods outline would not provide an
adequate description of stability conditions during a strong
earthquake. The rationale behind the Newmark sliding block model is
that the weakest component of the slope is the one that will fail
and that the earthquake load merely contributes to the driving
force. This is a simplification of earthquake ground motions since
during an earthquake effective normal and shearing stresses will
all change markedly during shaking. However, the analysis of the
Goat landslide throws up an interesting observation: that is, there
is a marked velocity contrast between the disturbed rock mass
evident in Figure 5e and the undisturbed rock mass in Figure 5f.
Calculations of shear wave velocity (Vs) based on rock mass
classification indicate a significant difference in Vs. For the
undisturbed rocks below the slip plane, shear waves will be
transmitted at approximately 2400ms-1. However in the disturbed
rock mass evident in the flanks of the landslide wave propagation
speeds were likely to have been of the order of 1500-1600ms-1. In
practical terms the energy moving into the low velocity layer will
be unable to exit at the same speed so the particle amplitude will
increase. That is, there will be higher ground accelerations in the
low velocity layer.
The practical implications of this are potentially significant.
The first is that the method outlined by Newmark (1965) while being
valid for homogenous and isotropic materials does not translate
well to natural heterogeneous and anisotropic slopes. It is
therefore, potentially inappropriate to use such methods without
careful consideration of the ground conditions. If either a Newmark
Sliding block or a pseudostatic method of analysis is being applied
to a slope there are some reasonable questions that may guide
further investigation:
1. Does a low velocity layer exist in the stratigraphy? Such low
velocity layers may include, but not be limited to weathered zones;
disturbance from engineering, mining or quarrying; unloading or
stratigraphic variations.
2. If such a zone is a discrete layer in the geological
succession, is it orientated in a manner which will impact on the
stability of the slope? Should it be treated in the same way that
discontinuities are treated?
-
3. If such a low velocity layer outcrops at the surface of the
slope, will the weakest materials merely be shaken off the top, and
the depth of the landslide be limited by the availability of low
velocity rock mass that can be displaced, rather than the shear
strength of the material within the slope? I.e. does the landslide
deepen progressively during the earthquake, not form as a discrete
slip plane that the overburden slides on?
Figure 7 shows two rockfalls triggered by the 1999 Chi Chi
Earthquake in Central Taiwan. Both failures occurred in a weathered
rock mass where loose material, that is material with low shear
wave velocities, was merely shaken off the rock mass. While neither
of these landslides constitutes a shear failure, there is no
rational reason why poorly consolidated materials at lower slope
angles should not show deformation when subjected to forced
vibration.
Figure 7. Rockfalls triggered by the M=7.9 Chi Chi earthquake in
1999. Are the low velocity layers
associated with weathering and unloading responsible for the
large number of shallow earthquake induced landslides?
The observations presented here must be regarded as preliminary.
There are large uncertainties in almost every stage of the process
and more rigour is required. However, the investigation does tend
to suggest that the presumption that the critical slip circle i.e.
the weakest component of a given slope is the block that will slide
does not appear to be supported by field evidence. This is the
subject of research in progress.
5 Summary and implications This work currently outlines the some
observations that have arisen from research in progress. The
analysis of three landslides, albeit it with observations from the
authors of many others, does not prove a hypothesis and these
comments have to be regarded as preliminary. Having made that
statement, it is felt that these are valid observations and not
merely coincidental. These observations can be summarised as
follows:
-
1. Slope stability analysis does not accurately predict the
geometry of landslides triggered by earthquakes. Where the geometry
of such landslides are known back analysis can be used to derive
some information about forcing functions active during the
earthquake or the potential seismic origin of a landslide
2. Is ambient stress state a valid predictor of seismic slope
stability? Should there be some other parameter be used to guide
further analysis of the stability of slopes during earthquakes
(such as shear wave velocity)?
3. Is the simplest way to mitigate earthquake-induced landslides
in rock masses to try to maintain an almost uniform velocity
profile throughout the slope? Does the uniform transmission of
seismic waves through the ground only present a problem when there
are differences in the velocity profile? In the absence of a low
velocity layer such as a systematic discontinuity zone; a weak
layer within the stratigraphy or even a landslide shear zone to
trap/scatter seismic waves (see Rial, 1996) wave propagation may
occur through the rock mass without significant difficulties.
6 Acknowledgements GNS Science (New Zealand) is gratefully
acknowledged for logistical and financial support during fieldwork.
Rob Parker is funded by the Willis Re Research Network and their
support is gratefully acknowledged.
7 References Arias, A. (1970). A measure of earthquake
intensity. Seismic Design for Nuclear Power Plants. Massachussetts
Institute of
Technology Press, Cambridge, MA, pp. 438483. Ambraseys, N.N.,
Menu, J.M. (1988). Earthquake-induced ground displacements.
Earthquake Engineering and Structural
Dynamics 16(7): 9851006. Ambraseys, N.N., Srbulov, M. (1995).
Earthquake induced displacements of slopes. Soil Dyn Earthq Eng 14:
5971. Ashford S. A., Sitar N. (1997). Analysis of topographic
amplification of inclined shear waves in a steep coastal bluff.
Bull
Seis Soc Amer 87(3): 692-700. Ashford S.A., Sitar, N., Lysmer
J., Deng N. (1997). Topographic effects on the seismic response of
steep slopes. Bull Seis
Soc Amer 87(3): 701-709. Banks, D. (2005). Rock mass ratings
(RMRs) predicted from slope angles of natural rock outcrops.
International Journal of
Rock Mechanics & Mining Sciences 42: 440449. Bishop, A.W.
(1955). The use of slip circles in the stability analysis of earth
slopes. Geotechnique 5: 7-17. Cruden D.M., Varnes D.J. (1996).
Landslide types and processes. In Turner A.K.; Shuster R.L. (eds)
Landslides:
Investigation and Mitigation. Transp Res Board, Spec Rep 247, pp
3675. Hancox, G.T., Perrin, N.D., Dellow, G.D. (2002). Recent
studies of historical earthquake-induced landsliding, ground
damage, and MM intensity in New Zealand. Bulletin of the New
Zealand Society for Earthquake Engineering 35(2): 5995.
Hoek, E., Brown, E.T. (1997). Estimation of rock mass strength.
Intl J Rock Mech Min Sci Geomech Abs 34: 1165-1186. Hoek, E.,
Carranza-Torres, C., Corkum, B. (2002). Hoek-Brown failure
criterion 2002 Edition. In Proc. 5th North
American Rock Mechanics Symposium, Toronto, July 2002, 267-273.
Jibson, R.W. (1993). Predicting earthquake-induced landslide
displacement using Newmarks sliding block analysis.
Transportation Research Board, National Research Council,
Washington D.C., TR record 1411, 917. Janbu, N. (1954). Application
of composite slide circles for stability analysis. Proceedings of
the European Conference on
the Stability of Earth Slopes. Stockholm, 3: pp. 43-49. Miles,
S.B., Ho, C.L. (1998). Rigorous landslides hazard zonation using
Newmarks method and stochastic ground motion
simulation. Soil Dynamics and Earthquake Engineering 18: 305323.
Murphy, W., Petley, D.N, Bommer, J.J., Mankelow, J.M. (2002).
Seismological uncertainty in the assessment of slope
stability during earthquakes. Q Jl Engng Geol. 35: 71-78.
Newmark, N.M. (1965). Effects of earthquakes on dams and
embankments. Geotechnique 15(2): 139159. Paolucci, R., Rimoldi, A.
(2002). Seismic Amplification for 3-D Steep Topographic
Irregularities. In 12th European
Conference on Earthquake Engineering, Paper Reference 087.
Elsevier Science.
-
Read, S.A.L., Millar, P.J. (1990). Strength characteristics of
New Zealand Tertiary age sedimentary soft rocks and their stability
in cut slopes. Road Engineering in Soft Rock Materials Road
Research Unit Bulletin 84(4): 37-50. ISSN 0549-0030.
Rial, J.A. (1996). The anomalous seismic response of the ground
at the Tarzana Hill site during the Northridge 1994 Southern
California Earthquake: A resonant, sliding block? Bulletin of the
Seismological Society of America 86(6): 714-1723.
Sarma, S.K. (1981). Seismic displacement analysis of earth dams.
Journal of Geotechnical Engineering Division, ASCE 107(12):
17351739.
Sepulveda, S.A., Murphy, W., Petley, D.N. (2005). Topographic
controls on coseismic rock slides during the 1999 Chi-Chi
Earthquake, Taiwan. Quarterly Journal of Engineering Geology and
Hydrogeology 38: 189196.