FLOW SLIDES AND FLOWS IN GRANULAR SOILS OLDRICH HUNGR

advertisement
FLOW SLIDES AND FLOWS IN GRANULAR SOILS
OLDRICH HUNGR
Department of Earth and Ocean Sciences, University of British Columbia, Canada
ABSTRACT: Three types of flow-like landslides involving granular soils are among the most destructive of all landslide phenomena:
debris avalanches that develop from rapid shallow failures on steep slopes, debris flows that take place in established channels and
gullies and flow slides that are often deep-seated and are caused by spontaneous or earthquake liquefaction of granular soil. This
article reviews the geological, hydrological and morphological context under which such landslides occur. Debris flows and
avalanches require steep slopes and shallow, loose soil veneers. The initial sliding failure may occur in saturated or partly saturated
soil. Rapid loading and incorporation of material is a very important process. Flow slides require deep deposits of loose soil. Full
saturation must be present on the rupture surface, although much of the overlying material can be dry. Loose saturated sands,
lacustrine and eolian silts, residual soil, pyroclastic deposits and artificial fills are most prone to flow sliding. Triggering and motion
mechanisms of flows and flow slides are reviewed.
Keywords: Rapid landslides, debris flows, debris avalanches, flow slides, dynamics
1. INTRODUCTION
Landslide dynamics is a subject that has long been rather
neglected in the geotechnical literature. Morgenstern (1978)
remarked on this deficiency with the following words:
“… There is a class of problems that arises in practice that is not
concerned with the evaluation of whether a slope will fail or not;
but instead obliges us to assume movement and design against
the consequences. Where it is not practical to eliminate
movements, the following questions arise:
1)
2)
3)
How much material will move?
What will be the time history of the movements in
terms of velocities and accelerations?
How are protective structures designed against moving
masses? “
Flow-like landslides are the most destructive among all types of
landslide phenomena. Those that occur under natural conditions
are also the most difficult to prevent. Questions 1 to 3 above
then formulate the main objectives of research directed to such
landslides.
4) "Clay flow slide" is a very rapid to extremely rapid flow of
liquefied sensitive clay, at, or close to its original water content.
The first process, "debris avalanche" initiates on steep slopes and
thus obtains its mobility mainly from a large store of potential
energy. Pore-pressures generated by rapid undrained loading
probably play an important role in mobilizing this landslide type.
The second type, "debris flow", occupies an established channel,
although it may develop from a debris avalanche (or rock slide)
at the point of initiation. It is characterized by entrainment of
saturated debris from the path, surging, internal sorting and
mixing with surface water. The third and fourth types, "flow
slides" involve in situ liquefaction of the source material. These
landslides are therefore limited to materials prone to spontaneous
or earthquake liquefaction, such as quick clays, loose saturated
sands, silts or unsorted debris. Obviously, there is some overlap
between the categories.
For example, a shallow debris
avalanche triggering on a steep slope may contain some material
which has spontaneously liquefied. Nevertheless, the proposed
definitions facilitate the identification of dominant mechanisms
responsible for the triggering and propagation of these extremely
rapid and dangerous landslides.
In an attempt to improve the classification of flow-like
landslides, Hungr et al. (2001) introduced the following
definitions:
2. BASIC INITIATION MECHANISMS
1) "Debris avalanche" is a very rapid to extremely rapid shallow
flow of partially or fully saturated debris on a steep slope,
without an established channel.
To conceptualise the process of slope failure, it is useful to think
in terms of the resultant driving, Fd, and resisting, Fr, forces
acting on the failing mass (Figure 1a). Their possible change
with displacement is shown in Figure 1b.
2) "Debris flow" is a very rapid to extremely rapid flow of
saturated non-plastic debris in a steep channel. ("Mud flow" is
preferred if the Plasticity Index of the matrix (sand and finer) is
greater than 5%).
3) "Sand (silt, debris) flow slide" is a very rapid to extremely
rapid flow of sorted or unsorted granular material on moderate
slopes, involving excess pore-pressure or liquefaction at the
source.
2.1 General
The driving force is a component of gravity and depends on the
mean inclination of the current velocity vector field. Depending
on the curvature of the rupture surface, this force may decrease
gradually with displacement (rotational slide), or remain
approximately constant (translation).
Fd
Fr
(a)
the friction angle, φ. The presence of cohesion, c, permits the
slope to stand steeper than φ. This cohesion may be true
cohesion, due to electrostatic forces or chemical cementing at
particle contacts. It may also be apparent cohesion, arising from
suction forces at particle contacts of an unsaturated soil.
Alternatively, a bulk cohesion may result from vegetation root
reinforcement. At limit equilibrium, the driving stress, τ and
resisting stress, S, must be equal:
τ = γH sin β
Fd
;
S = c + γH cos β tan φ
(1)
Fr
Displacement
(b)
Figure 1. Forces acting on a landslide at the point of failure.
(a) Landslide profile, (b) Variation of resultant forces with displacement.
The resisting force must equal the driving force at the onset of
failure, then reduce as a result of brittleness. The vertical
distance between the resisting and driving force curves in Figure
1b gives the net unbalanced force, ∆F, available to cause
acceleration of the slide. From the work-energy theorem, the
kinetic energy developed by the slide at any value of
displacement will equal the work done by the unbalanced
force,
∫ ∆Fdx , i.e. the shaded area between the two curves.
Thus, the pre-requisite of rapid landslide acceleration is always a
sudden loss of shear strength. In the context of the CoulombTerzaghi strength criterion, this can occur by three processes:
1) loss of cohesion, 2) decrease of friction angle and 3) increase
of pore-pressure or, in the extreme, spontaneous liquefaction.
Once oversteepening, overloading or some other trigger initiates
failure displacement, cohesion will almost instantly disappear.
Then, an unbalanced tangential force equal to S-τ will be acting
on the column. Substituting from (1) and dividing by column
mass, we obtain a formula for the resulting tangential
(downslope) acceleration:
a=g
c
γH
(2)
Due to the need to equate τ and S at failure, the amount of
cohesion is functionally related to the slope angle (as obtained
from (1)). The full line in Figure 3 is derived from (1) and (2),
using a φ of 30º. Evidently, the acceleration increases rapidly
with rising slope angle. Failures initiating by cohesion loss will
be rapid, if starting from steep slopes, even if no pore-pressure
changes occur. The amount of acceleration will increase with
increasing cohesion or decreasing depth and weight of the slide.
Cohesion Loss
Friction Decrease (20%)
Liquefaction
2.2 Strength Loss
0.8
0.7
0.6
ACCELERATION (g)
The dynamic consequences of sudden strength loss can be
quantified using a simple infinite slope stability model (e.g.
Morgenstern and Sangrey, 1978). As shown in Fig. 2, such an
analysis is applicable to a shallow translational slope failure, H
thick, where end effects can be neglected and the equilibrium of
the slide is represented by forces acting on a soil column of unit
normal area.
0.5
0.4
0.3
0.2
0.1
H
β
0
15
20
25
30
35
40
EQUILIBRIUM SLOPE ANGLE
45
Figure 3. Theoretical acceleration resulting from three types of strength
loss, as derived from Equations 1,2 3 and 4.
Figure 2. The infinite slope problem, shown in profile.
Let us first assume that a slope exists in a dry condition,
supported by soil strength containing both a cohesive and
frictional component. If there was no cohesion, the slope would
stand at the angle of repose and the slope angle, β, would equal
A similar derivation can be made to consider the effects of
friction decrease. A decrease will result from particle realignment or asperity breakage in passing from peak (φp) to
residual friction angle. Neglecting cohesion, the equilibrium
slope angle β must equal φp. The acceleration associated with
the decline of friction angle from peak to residual will be:
(3)
The resulting acceleration for a friction decrease by 20% is
shown by the long-dashed line in Figure 3. This process is not
too effective in causing acceleration, if the decline of friction
requires long displacement.
Liquefaction occurs only under a high degree of saturation.
Assuming zero cohesion and a steady seepage flow parallel to
the slope face, the equilibrium slope angle, β, cannot be more
than approximately one-half the peak friction angle. Once
structural collapse of the soil skeleton occurs, the pore-pressures
will approach the full weight of the soil column (geostatic
value). The resisting stress, S will be reduced to zero and the
unbalanced force will equal τ, as given in (1). The acceleration
will then equal:
a = g sin β
(4)
As shown by the short-dashed line in Figure 3, this process is
capable of producing high acceleration even on moderate slopes.
On the other hand, it is unlikely to occur on steep slopes, as such
slopes are unlikely to be sustained by loose, saturated material
required for liquefaction.
The simple models behind Figure 3 indicate that rapid initial
acceleration will most likely occur on steep, often unsaturated
slopes maintained by cohesion, or on flatter slopes made of
loose, saturated and liquefiable (or at least contractive) material.
Shallow rupture and low in situ unit weight enhance initial
acceleration in the first case.
2.3 Undrained Loading
Significant initial acceleration will propel the toe of the landslide
mass forward, overriding the ground surface downslope. The
soil forming this surface will be very rapidly loaded. This will
set the conditions for rapid undrained loading (Hutchinson and
Bhandari, 1971) or associated liquefaction (Sassa, 1985).
The process is illustrated in Figure 4, produced using the
dynamic flow model DAN (Hungr, 1995). In this example, a
rectangular block of soil fails by means of cohesion loss. In the
DAN model, this is simulated by instantaneously turning the
block from a solid to a frictional fluid. Initial acceleration
occurs such as predicted by the full line in Figure 3, propelling
the slide toe forward. The substrate of the ground downhill is a
relatively loose, low permeability, saturated soil, that will
respond to the sudden loading by structural collapse and rise of
pore-pressure.
As a result, the toe of the slide will now be carried on top of a
cushion of liquid mud, or will displace the latter and mix with it
to form a fluid composite mass. This is reflected in the DAN
model by changing the rheological properties at the base of the
slide from frictional to fluid-like. This change is reflected by the
noticeable elongation of the front and acceleration of the leading
edge. As long as saturated material is available on the path, the
same process may continue indefinitely and a mobile, flow-like
movement will result, similar to a flow slide.
60
ELEVATION (m)
a = g cos φ p (tan φ p − tan φ r )
70
50
40
30
20
10
-10
0
10 20 30 40
DISTANCE (m)
50
60
Figure 4. Motion of a block of frictional material entraining an erodible
substrate layer (Produced using model DAN, Hungr, 1995). Profiles
shown at 0 and 4 seconds.
2.4 Liquefaction
The principles of spontaneous or earthquake liquefaction failure
have been discussed by many authors. Perhaps the clearest
explanation of the process uses the "collapse surface" concept,
proposed originally by Sladen et al. (1985). Collapsive materials
exist in situ at void ratios higher than the steady-state void ratio.
Under full saturation, granular collapsive soils are formed by
rapid deposition, which prevents proper packing of the soil
skeleton. Typical examples are sand or silt deposits on some
river deltas, or man-made hydraulic fills (e.g. Casagrande,
1976). In saturated clays, collapsive structure may be due to
electro-chemical changes induced by pore-water leaching (e.g.
Lefebvre, 1995).
The origin and behaviour of collapsive structure in partly
saturated soils is less-well understood. For example, loess soils
form by slow dry aggradation of windborne dust and are held in
a loose state by means of pore-water suction (Dijkstra et al.,
1994). On saturation, the suctions are removed and the soil
collapses (e.g. Lefebvre, 1995). However, this would not,
normally be an instantaneous process, as saturation of a mass of
fine-grained soil must take a long time, measured in hours or
days. Yet, we often observe sudden liquefaction of parts of such
deposits, particularly at basal contacts where perched water
tables can be expected to occur (Dijkstra et al., 1994). How does
loess maintain its unstable skeleton until sufficient quantity is
saturated to the point of collapse? One possibility is that a part
of the cohesion is of a chemical nature and is therefore able to
support the soil structure even after some saturation occurs, until
sudden rupture. This rupture is likely brought about by
overstress, due to stress redistribution by progressive failure, or
to earthquake shaking. Cementing by calcium carbonate, iron or
aluminum oxides is a possibility (Professor E.Derbyshire, pers.
comm.).
A similar process may be effective in producing collapsive
structure in well-graded materials such as natural colluvium, or
artificial fills (e.g. mine waste). Such material is loosely
deposited in a moist condition and held in this state by light
cementing. With time, some fine-grained portions of the
originally dry deposits become saturated, without change in void
ELEVATION (m)
ratio. Collapse occurs when the loose structure of the soil is
suddenly destroyed by shear rupture. It is of interest to observe
that 30 years ago, only uniform loose saturated sands were
considered capable of collapse liquefaction (Terzaghi and Peck,
1967). In fact, many well-graded materials have now clearly
been shown to have collapsive structure, as shown in Figures 5
and 6 (Dawson et al., 1998).
40
30
20
10
0
10
20
30 40 50
DISTANCE (m)
60
70
80
Figure 7. Dynamic analysis of a block of soil subject to sudden and
complete spontaneous liquefaction (Hungr, 1995). Profiles shown in 2
sec. intervals.
3. EXAMPLES
3.1 Debris Avalanches and Flows in Shallow soil Veneer
Figure 5 Grain size distributions of collapsive coal waste material
(Dawson et al., 1998).
Perhaps the most frequent scenario for producing rapid flow-like
landslides is the failure of a thin veneer of relatively loose soil,
covering a steep slope made of stronger material. The most
common stratigraphy involves a colluvial veneer covering
residual soil, bedrock or glacial till. Even more unstable are
veneers of volcanic ejecta covering steep eroded topography
(e.g. Del Prete et al., 1998). Existence of such veneers on
slopes that often approach or exceed 45º indicates the presence
of cohesion, belonging to any or all of the three kinds mentioned
previously, i.e. cement, negative pore-pressure or root systems.
Considerable controversy exists in literature, concerning the
degree of saturation of such slope profiles. In any event,
cohesion is quickly lost upon initial failure displacement and
rapid acceleration ensues. Even if material covering the steep
upper slope is essentially unsaturated, gradual accumulation of
seepage may well saturate the lower parts of the slope, which are
over-ridden and subject to liquefaction by rapid undrained
loading. Thus, debris avalanches starting as localized failures of
perhaps a few cubic metres of soil grow in width as they
propagate downslope, increasing in volume many times beyond
the magnitude of the initial slide.
Figure 6. Results of a large-diameter undrained triaxial test on mine
waste material, exhibiting collapsive behaviour (Dawson et al., 1998).
The effect of spontaneous or earthquake liquefaction is simply a
sudden change of the material from solid to liquid. This change
can be simulated using fluid dynamics models such as DAN,
previously mentioned. Liquefaction may affect the full depth of
the material (usually if fully saturated) as shown in Figure 7.
Alternatively, only a thin basal layer may liquefy, perhaps due to
an accumulation of pore water above an impervious contact.
The analysis must then consider the internal friction and stiffness
of the sliding mass (Hungr, 1995).
One of the best documented examples of such a slope failure is
the largest known natural rapid landslide in Hong Kong, the
Tsing Shan debris flow of 1990 (King, 1996). The initial slide
was a failure of a small colluvial wedge, infilling a structurallycontrolled steep bedrock gully and less than 400 m3 in volume
(Figure 8) In moving down a steep bedrock slope thinly covered
by colluvium, the event grew to its final volume of 20 000 m3
(Figure 9). Similarly, many of the infamous debris flows and
debris avalanches of the Campania region in Italy began by
relatively small sliding failures of a pyroclastic veneer above cut
slopes of artificial trackways, or the crests of natural cliffs
(Professor F. Guadagno, pers. comm.). They widened and grew
many times in magnitude, while traversing the middle and lower
slopes covered with a similar veneer (Del Prete et al., 1998).
This mechanism of enlargement is the most important aspect of
debris avalanches, yet we have little quantitative understanding
of it. Predictions of runout of such events are severely hindered
by our inability to estimate the magnitude (volume) of the event,
which depends on the width of the landslide path and the depth
the yield rate exist. Measurements involving relatively confined,
narrow debris flow paths on Lantau Island in Hong Kong
showed yield rates of the order of only 2-5 m3/m (Lau and
Woods, 1997). However, the yield rates involved in the Tsing
Shan event shown on Figure 9 reached as high as 29 m3/m.
Extensive surveys of debris flow/avalanche channels on Queen
Charlotte Islands in British Columbia, Canada, showed a great
range of yield rates (Figure 10), but no-one has yet succeeded in
correlating these values with suitable predictor variables (Mr. T.
Rollerson, Golder Associates Ltd., Pers. Comm., 1997).
Systematic collection of yield rates and correlation with soil and
slope characteristics is necessary to improve our ability to
predict the failure behaviour of debris flows and avalanches.
200
Figure 8. The source area of the trigger landslide on Tsing Shan, Hong
Kong. Photo taken in 2002.
180
160
FREQUENCY
140
120
100
80
60
40
20
0
-50 -40 -30 -20 -10 0
10 20
YIELD RATE (m3/m)
30
40
50
Figure 10.
Distribution of measured channel yield rates from
approximately 1000 reaches of 300 debris flow/avalanche events on the
Queen Charlotte Islands, British Columbia, Canada (Data courtesy Mr.
T. Rollerson, Golder Associates Ltd., Vancouver. Note: negative yield
rate represents deposition of material.
3.2 Flow Slides in Saturated Fine Sand
Figure 9. The Tsing Shan debris flow in Hong Kong. The flow volume
increased from 400 m3 to 20,000 m3 through material entrainment.
Photo courtesy of Mr. J.P. King, GEO, Hong Kong.
of material erosion. The product of the latter quantities was
designated by Hungr et al.(1984) as the "Channel Yield Rate",
measured in m3 per meter of path length (a negative value
represents deposition in flatter or less confined reaches of the
path). Unfortunately, only a few reliable field measurements of
Large flow slides occur periodically on the front of the Fraser
River delta near Vancouver, British Columbia, Canada. The
main channel of the Fraser enters the Pacific Ocean at a location
on the front of the delta called Sandheads, 15 km south of the
City of Vancouver. In the past, the river channel meandered
over the surface of the delta, and its mouth migrated laterally
across the delta front (Figure 11). Wherever the channel crossed
the delta crest, a submarine canyon formed. In 1932, a
permanent jetty was constructed, confining the river outlet
permanently to a location just south of the Sandheads lighthouse.
A major canyon developed in front of this location during the 7
decades following the completion of the jetty.
At present, Public Works Canada (PWC) conducts annual
dredging of the main channel immediately to the south of the
jetty, to maintain a minimum depth of 10 m, required for
navigation. Submarine landslides occur periodically at the
western end of the navigation channel, just south of the jetty (Mc
Kenna et al., 1992).
This periodic cycle of landsliding and sedimentation suggests a
possible explanation for the occurrence of the large failures. The
Fraser delivers large quantities of fine sand to the delta front
during the freshet (late June-July). Based on sounding data, as
much as 10 m of vertical aggradation can occur at the slope crest
per year. Considering that much of this occurs during less than
two months, the aggradation rate can be as high as 5 m per
month. Morgenstern (1967) considered the effects of
Figure 11. Location plan of the Fraser delta front. The square frame is 4
km wide . Note submarine canyons and historic channels of the Fraser
River
Consolidation degree (%)
Sandheads
100
80
Cv (cm2/sec)
60
40
0.001 (silt)
20
0.01 (sand)
0.1 (sand)
0
0
1
10
100
1000
Rate of sedimentation (cm/year)
10000
Figure 13. A nomogram derived from the theory of coupled
consolidation and sedimentation, following the approach of Morgenstern
(1967).
underconsolidation on the stability of rapidly aggrading delta
slopes. Following his approach, Figure 13 shows a nomogram
of underconsolidation, derived using a theory of coupled
sedimentation and consolidation. Given the high sedimentation
rate of fine sand at Sandheads, it is quite possible for the degree
of consolidation to remain at only 50% during the freshet period.
This may create excess pore pressures, capable of triggering
local instability. The triggering of instability may be aided by
methane gas exsolution, formed hysteretically during tidal cycles
(Grozic et al., 2000)
Figure 12. Position of the 10 m depth contour preceding (i) and
following (f) each of the documented submarine landslide events (from
McKenna et al., 1992). The dashed line on the left and the full line on
the right side of the picture represent the 1985 landslide.
The most significant of the recorded landslide events occurred
on June 30, 1985. A PWC survey crew was engaged in
sounding the depth of the shipping channel. They found that the
depth of the channel bottom changed from 10 to 30 metres
between morning and afternoon, over an area nearly 300 m long
along the river axis and 250 m wide. Thus, more than 1 million
m3 of sediment vacated the river mouth in course of several
hours and flowed west over the face of the delta and into the
submarine canyon (Figure 12). Neither the jetty nor the
Sandheads Lighthouse were affected by the adjacent landslide.
No unusual waves, or even turbidity were reported in the
shipping channel on that day. The space emptied by the 1985
landslide contained fresh river sediment, which accumulated
during river freshets since the last preceding failure (Figure 12).
Since 1985, the same space infillled again by new sediment.
Thus, the landslides appear to periodically empty the same
approximate location, to be filled anew by river sedimentation.
A typical longitudinal profile along the centerline of the river
channel at the crest of the delta slope is shown in Figure 14a. A
limit equilibrium analysis showed that small-scale slumping
failure can easily occur at the crest, prompted by
underconsolidation and possibly also by spontaneous undrained
collapse of the loose sediment. Once an initial failure occurs,
the liquefied sand will flow down the slope as shown in Figure
14b. The wave of sand will over-ride loose material on lower
slopes. An approximate undrained limit equilibrium analysis of
the front of the propagating sand wave proved that entrainment
of a large quantity of sand from the path of the sand wave is
likely (cf. Sassa. 1985).
In this manner, a relatively shallow flow slide can sweep over
the slope, removing a slice of material as shown in Figure 14c.
The crest of the slope is left unsupported and may fail
retrogressively by the same mechanism. In this way, a more-less
continuous flow of liquefied sand may be set up on the slope,
moving material steadily towards the submarine canyon and
enlarging the landslide crater in the upstream direction. Such a
process would explain how an enormous quantity of material can
be removed in a few hours, without causing a noticeable
disturbance of the water surface. The material entrainment
process may continue on lower slopes of the delta, deepening the
submarine canyon and possibly diluting itself to form a turbidity
current on lower slopes of the delta.
Retrogression, 4m
aa
Liquefaction
Figure 13. Schematic view of a stage in the formation of a retrogressive
flow slide.
3.3 Flow Slides in Overconsolidated Silty Clays: Macroscopic
Brittleness
b
Slides involving overconsolidated, non-sensitive clay tend to be
slow and ductile (e.g. Skempton and Hutchinson, 1969). Many
such slides are found along river valleys of the Canadian
Prairies, exploiting weak surfaces in heavily overconsolidated
glacio-lacustrine clays, or the underlying argillaceous bedrock
(Thomson and Morgenstern, 1977).
An unusual case is the Attachie Slide, which occurred in May,
1973 on the Peace River upstream of Fort St.John, north-eastern
British Columbia. The south side of the 200 m deep valley had
been unstable for at least 30 years, before the first airphotos of
the area were taken. Sliding was taking place in a thick and
complex sequence of glacio-lacustrine clays, clayey silts and
silts, covered by about 30 m of glacial till. The glacio-lacustrine
sequence was underlain by basal gravels and Cretaceous
bedrock, neither of which participated in the instability.
Displacements totaling several tens of metres had occurred at
several levels in the slope by 1973, creating several large scarps
(Figure 15a). The instability occurred in two stages, upper and
lower, separated by an intact intermediate scarp which can be
seen on both in Figures 15a and b and on the profile, Figure 16.
Precedent from similar locations in prairie valleys would
indicate that a slow, ductile failure was in progress. However, in
May, 1973, following a relatively wet year, the lower stage of
the slope became suddenly mobile. A total of 12.4 million m3 of
material moved on both levels. About half of this volume (6.4
million m3) was sufficiently mobile to descend a 60 m scarp at
the toe of the slope and move rapidly across the floodplain of the
Peace River (Figure 15b). This mobile portion corresponded
largely with the lower stage of the instability, between the toe
and the intermediate scarp. The upper stage also displaced, but
with much less mobility. The flow slide had sufficient
momentum to raise a wave and impact the opposite shore.
Details of the case history can be found in Evans et al. (1996),
Fletcher (2000) and Fletcher et al. (2002).
The glacio-lacustrine soil, forming the most mobile part of the
displacement material, consisted of approximately 31% of lowplastic silt, 48% plastic clayey silt and 21% sand (Fletcher et al.,
2002). All these materials are complexly interbedded and very
Figure 15. Vertical airphotos of the Attachie Slide. a) in 1970
(BC5529,75) b) in 1973 following the rapid flow slide of May, 1973
(BC7279,70). The area covered by the photographs is approximately
1,800 m wide.
Upper stage
650
Lower stage
?
600
Pre-failure surface
550
Post-failure surface
Unit 1
Unit 2
Unit 3
Unit 4
Unit 5
Unit 6
Cobble/gravel
Pre-glacial laminated clay and silt
Sand
Till
Colluvium
Post-glacial laminated clay and silt
500
450
400
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
meters
Figure 16. A central cross-section of the Attachie Slide (Fletcher, 2000)
stiff or dense, having been overridden by glacial ice on at least
one occasion. There is some evidence that the low plasticity silt
units are cemented by calcium carbonate or gypsum. The flat
basal part of the rupture surface formed in thinly laminated
highly plastic illite clay with a clay fraction of 60-70%. The
residual friction angles ranged from 17° to 25°, on account of
varying silt content. Three consolidated undrained triaxial tests
on this material indicated dilative behaviour with values of the
pore-pressure parameter, “A” at failure of 0.12 to 0.55,
indicating low sensitivity. Liquidity indices generally below 0.5
also pointed to low sensitivity.
It is not difficult to explain the occurrence of instability at this
site, given the existence of pre-sheared layers of plastic clay and
the presence of confined ground water aquifers. However, it is
much more difficult to explain the occurrence of extremely
rapid, catastrophic failure as observed in May, 1973.
In order to quantify the strength loss evidenced by the 1973
displacement, a dynamic analysis of the lower stage source
volume was carried out with the program “DAN” (Hungr, 1995),
using frictional rheology. Used in a back-calculation, the model
can give the strength loss required to obtain the geometry change
observed during the failure. In this case, it was first assumed
that undrained pore-pressure increase was the sole source of
brittleness. The analysis showed that an average pore pressure
ratio increase from 0.26 to 0.50 was required to deform the
sliding body as observed. The corresponding change in the
Factor of Safety is approximately from 1.0 to 0.5 (Fletcher,
2000, Fletcher et al., 2002).
It is difficult to explain such a dramatic, instant strength loss,
given the character of soil materials at this site. None of the soil
descriptions give an indication of the possibility of contractive
behaviour, which could translate into high undrained brittleness.
Further, the large ductile pre-1973 displacements of the slope are
hard to reconcile with the behaviour of a sensitive clay.
Alternative explanations of the catastrophic failure are therefore
needed. After a detailed examination of several alternatives,
Fletcher (2000) concluded that two failure mechanisms are
possible:
The first hypothesis considered the internal strength of the
sliding body. The profile of the lower stage of the Attachie slide
is compound. The rupture surface consists of a long, flat-lying
segment which follows the pre-sheared clay layer and a steep
back scarp cutting across layers.
In order to fail, such a
geometry requires strong internal distortion on secondary shears
complementary to the back scarp (Hutchinson, 1988). The high
internal strength of the unfailed very stiff, possibly cemented
soil, adds considerably to the overall sliding resistance, but
reduces sharply, once the brittle material fails. The quantitative
effect can be modeled using a two-block stability analysis
(Fletcher, 2000).
The second hypothesis concentrates on the occurrence of
multiple tension cracks in the disturbed slope, as a result of pre1973 movements. The cracks divide the silty soil into a network
of stiff blocks separated by discontinuities. Many of the cracks
fill over time by loose silty debris, which may become saturated
by surface water. Once such mixture of intact blocks and loose
matrix is forced to move, localized liquefaction may occur,
driving the toe portion of the slope forward. This effect can be
referred to as “macroscopic brittleness”. Analyses by Fletcher
(2000) show that most likely both of the above mechanisms
participated in the spectacular failure of May, 1973. Neither
mechanism is normally considered in hazard analyses for slopes
in over-consolidated clays and silts.
4. CONCLUSIONS
Although analytical tools are available, quantitative analysis and
prediction of the behaviour of flow-like landslides is still not
straightforward.
Better understanding is required of the
mechanisms forming and maintaining loose soil structure,
providing cohesion and causing changes in the degree of
saturation. The range of materials and conditions known to be
susceptible to liquefaction must be greatly expanded. Practical
methods of identification of the potential for liquefaction must
be developed.
Empirical means of quantifying material
entrainment must be established, in order to facilitate the
prediction of magnitude and runout of flowing landslides.
REFERENCES
Casagrande, A., 1976. Liquefaction and cyclic deformation of
sands: a critical review. Harvard Soil Mechanics Series, No. 88,
Cambridge, Massachussetts, 26 p.
Dawson, R.F., Morgenstern, N.R. and Stokes,A.W., 1998.
Liquefaction flowslides in Rocky Mountain coal mine waste
dumps. Canadian Geotechnical Journal, 35:328-343.
Del Prete M., Guadagno F.M., Hawkins B. 1998. Preliminary
report on the landslides of 5 May 1998, Campania, southern
Italy. Bull. Eng. Geol. Env.,57, pp. 113-129.
Dikstra, T.A., Rogers, C.D.F., Smalley, I.J., Derbyshire, E., Li,
Y.J. and Meng, X.M., 1994. The loess of north-central China:
geotechnical properties and their relation to slope stability.
Engineering Geology, 36:153-171.
Evans, S., Hu, X.Q., Enegren, E.G. 1996. The 1973 Attachie
Slide, Peace River Valley, near Fort St. John, B.C., Canada: a
landslide with a high-velocity flow slide component in
Pleistocene sediments.
In Proceedings, 7th International
Symposium on Landslides, Trondheim, Norway, 715-720
Fletcher, L., 2000. Failure behaviour of two large landslides in
silt and clay.
M.A.Sc. Thesis, Geological Engineering,
University of British Columbia, 170p.
Fletcher, L., Hungr, O., and Evans, S.G., 2002. Contrasting
failure behaviour of two large landslides in clay and silt.
Canadian Geotechical Jounal, 39:46-62.
Grozic, J.P.L., Robertson, P.K. and Morgenstern, N.R., 2000.
Cyclic liquefaction of loose gassy sand. Canadian Geotechnical
Journal, 37:843-856.
Hungr, O., 1995, A model for the runout analysis of rapid flow
slides, debris flows and avalanches. Canadian Geotechnical
Journal: v. 32, p. 610-623.
Hungr, O., 2000. Analysis of debris flow surges using the theory
of uniformly progressive flow. Earth Surface Processes and
Landforms 25:1-13.
Hungr, O., Morgan, G.C. and Kellerhals, R., l984. ‘Quantitative
analysis of debris torrent hazards for design of remedial
measures’, Canadian Geotechnical Journal, 2l, 663-667.
Hungr, O., Evans, S.G., Bovis, M., and Hutchinson, J.N., 2001,
Review of the classification of landslides of the flow type.
Environmental and Engineering Geoscience, v. VII, p. 221-238.
Hutchinson, J.N. and Bhandari, R.K., 1971. Undrained loading,
a fundamental mechanism of mudflow and other mass
movements. Géotechnique, 21:353Hutchinson, J.N., 1988. General report: Morphological and
geotechnical parameters of landslides in relation to geology and
hydrogeology.
Procs., 5th International Symposium on
Landslides, Lausanne, C. Bonnard Ed., 1 3-35.
King, J., 1996. Tsing Shan Debris Flow. Hong Kong
Government, Geotechnical Engineering Office, Special Project
Report SPR 6/96, 133p.
Lau, K.C. and Woods, N.W., 1997. Review of methods for
predicting travel distance of debris from landslides on natural
terrain. GEO Technical Note, TN 7/97, 48 p.
Lefebvre, G., 1995.
Collapse mechanisms and design
considerations for some partly saturated and saturated soils. In
Genesis and properties of collapsible soils, E.Derbyshire et al.,
Eds., Kluver Academic Publishers, pp.361-374.
Morgenstern, N.R. and Sangrey, D.A., 1978. Methods of
stability analysis. in Schuster, R.L. and Krizek, R.E.J., Eds.,
Landslides, Analysis and Control. Transportation Research
Board Special Report 176, Washington, D.C., p. 155-171.
McKenna, G.T., Luternauer, J.L. and Kostaschuk, R.A., 1992.
Large-scale mass-wasting events on the Fraser River delta front
near Sandheads, British Columbia. Canadian Geotechnical
Journal, 29: 151-156.
Morgenstern, N.R., 1967. Submarine slumping and the initiation
of turbidity currents. In Marine Geotechnique, A.F. Richards,
Editor, University of Illinois Press: 189-220.
Morgenstern, N.R., 1978.
Mobile soil and rock flows.
Geotechnical Engineering, 9:123-141.
Sassa, K., 1985, The mechanism of debris flows. Proceedings,
XI Internat. Conference on Soil Mechanics and Foundation
Engineering, San Francisco, v. 1, p. 1173-1176.
Skempton, A.W. and Hutchinson, J.N., 1969. Stability of
natural slopes and embankment foundations. Procs., 7th.
International Conference of Soil Mechanics and Foundation
Engineering, Mexico, State of the Art volume, 291-340.
Sladen, J.A., Hollander,R.D. and Krahn, J., 1985. The
liquefaction of sands, a collapse surface approach. Canadian
Geotechnical Journal, 22:564-578.
Terzaghi, K. and Peck, R.B., 1967. Soil mechanics in
engineering practice. 2nd. Edition, John Wiley and Sons, New
York, 750 p.
Thomson, S. and Morgenstern, N.R., 1977. Factors affecting
distribution of landslides along rivers in southern Alberta.
Canadian Geotechnical Journal, 14:508-523.
Download