A model for geotechnical analysis of flow slides

and debris flows
Xiaobo Wang, Norbert R. Morgenstern, and Dave H. Chan
Abstract: Flow slides and debris flows incorporate a broad range of sediment–fluid mixtures that are intermediate between
dry rock avalanches and hyperconcentrated flows. Following a comprehensive review of some existing analytical ap-
proaches to debris flow runout analysis, a new analytical model based on energy conservation has been formulated. The
new analytical model was developed to deepen the understanding of fundamental aspects in modeling of granular flows
and to improve the geotechnical mobility analysis of flow slides and debris flows. The Lagrangian finite difference method
was used to solve the governing equations. The model and numerical scheme have been tested against analytical solutions
and experiments of granular flows with simplified geometries for sliding mass and basal topography. Results of granular
flow simulations indicate that the model based on energy conservation performs well and is robust. The model can be
used for geotechnical analysis of a wide range of dense granular flows, such as flow slides and debris flows.
Key words: flow slides, debris flows, runout analysis.
Re´sume´ : Les glissements par lique´faction et les e´coulements de de´bris comprennent un large e´ventail de me´langes fluide-
se´diments qui sont a` mi-chemin entre les avalanches de roches se`ches et les e´coulement hyperconcentre´s. Suite a` une re-
vue exhaustive de quelques approches existantes pour l’analyse du parcours des e´coulements de de´bris, un nouveau mode`le
analytique base´ sur la conservation de l’e´nergie a e´te´ formule´. L’objectif de ce nouveau mode`le est de mieux comprendre
les aspects fondamentaux de la mode´lisation des e´coulements granulaires ainsi que d’ame´liorer l’analyse de mobilite´ ge´o-
technique des glissements par lique´faction et des e´coulements de de´bris. La me´thode de diffe´rence finie de Lagrange est
utilise´e pour re´soudre les e´quations. Le mode`le et le sche´ma nume´rique ont e´te´ compare´s a` des solutions analytiques et a`
des expe´rimentations d’e´coulement granulaire ayant des ge´ome´tries simplifie´es de masse glissante et de topographie de
base. Les re´sultats des simulations d’e´coulement granulaire indiquent que le mode`le base´ sur la conservation de l’e´nergie
performe bien et est robuste. Le mode`le peut donc eˆtre utilise´ pour des analyses ge´otechniques d’une grande varie´te´
d’e´coulements granulaires denses, tels que les glissements par lique´faction et les e´coulements de de´bris.
Mots-cle´s : glissement par lique´faction, e´coulement de de´bris, analyse de parcours.
[Traduit par la Re´daction]
Introduction
Flow slides and debris flows are extremely rapid flows of
a highly concentrated mixture of water and predominantly
coarse granular material. This type of granular flow is com-
posed of a poorly sorted, sediment–water mixture that com-
monly contains more than 50% solids by volume. The
constituent sediment usually varies widely in size, from
clay particles to boulders of several metres in diameter. Be-
cause of high flow velocities, large impact forces, long run-
out distance, and poor temporal predictability, flow slides
and debris flows are among the most dangerous and destruc-
tive natural hazards (Johnson and Rodine 1984; Jakob and
Hungr 2005).
Flow slides and debris flows are complex in their physical
behavior and demand subtle theoretical descriptions and nu-
merical models. However, validation of flow slide and de-
bris flow simulation models against field observations is
often difficult or impossible because of the dangers involved
with in situ experimental campaigns, uncontrollable geo-
physical conditions, and unpredictable time and locations of
natural flow events (Hutter 2005). Although laboratory tests
of flow slides and debris flows of reduced size can be per-
formed under well-defined and well-controlled conditions,
these dense granular flows are known to be scale-dependent
and runout distance is greatly influenced by the rate of pore
pressure dissipation (Hutchinson 1986; Iverson 1997). The
significance of scale effects raises questions regarding the
application of experimental observations to predicting flow
slide and debris flow behaviors in practice. Because of these
difficulties in experimental tests and field observations, nu-
merical modeling has become an important and promising
alternative in flow slide and debris flow studies.
Within the framework of dense granular flow simulations,
Received 20 November 2008. Accepted 7 April 2010. Published
on the NRC Research Press Web site at cgj.nrc.ca on
18 November 2010.
X. Wang. Thurber Engineering Limited, Edmonton, AB T6E
6A5, Canada.
N.R. Morgenstern. Department of Civil and Environmental
Engineering, University of Alberta, 3-075 Markin/CNRL Natural
Resources Engineering Facility, Edmonton, AB T6G 2W2,
Canada.
D.H. Chan.
1
Department of Civil and Environmental
Engineering, University of Alberta, 3-038 Markin/CNRL Natural
Resources Engineering Facility, Edmonton, AB T6G 2W2,
Canada.
1
Corresponding author (e-mail: dave.chan@ualberta.ca).
1401
Can. Geotech. J. 47: 1401–1414 (2010) doi:10.1139/T10-039 Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
depth-averaged equations have been predominantly applied
to modeling debris avalanches, debris flows, and other types
of fast-moving landslides. The depth-averaged model is es-
tablished by integrating the mass and momentum conserva-
tion equations in differential form over the flow depth with
the use of Leibnitz’s law. Integration of the differential form
of the conservation laws over the flow depth considerably
reduces the computational time required in rapid landslide
runout analysis. Constitutive properties of the sliding mass
are incorporated into governing equations through the basal
flow resistance in the depth-averaged model. The flow re-
sistance is expressed in terms of averaged velocities when
fluid models such as the Bingham (Johnson 1970) and
Herschel–Bulkley (Huang and Garcı ´a 1998) models are
selected. The Coulomb frictional law has been widely
employed to compute basal shear resistance in the depth-
averaged model for dense granular flow simulations in
geotechnical engineering. The magnitude of resistance is
determined by the product of overburden normal stress and
coefficient of basal friction.
In this paper, a slice-based model based on energy conser-
vation was formulated on a macroscopic scale after a review
of the existing numerical models for granular flow simula-
tions. Compared with the existing depth-averaged slice-
based models, our new model incorporates deformation
work and internal energy dissipation into the governing
equations. The numerical method for practical implementa-
tion of the governing equations is presented following the
model formulation. Model verification has been undertaken
by comparing numerical predictions with analytical solu-
tions and experimental observations of granular flows.
Existing mathematical models for granular
flow simulations
Mathematical modeling of granular flows was originally
introduced by Savage and Hutter (1989, 1991). Based on
mass and momentum conservation equations for flow on a
rough inclined plane, and using the depth-averaging process
and making scaling arguments, Savage and Hutter (1989)
derived the one-dimensional, depth-averaged equations for
the shallow free-surface flow of dry granular materials. The
model assumes that a moving granular mass behaves as a
cohesionless Coulomb frictional material and the relation-
ship between shear and normal stresses on internal and
rough bounding surfaces obeys the Coulomb friction law.
Multi-dimensional extensions of the Savage–Hutter model
have been formulated for analyzing dry granular flows.
Iverson (1997) and Iverson and Denlinger (2001) intro-
duced Coulomb mixture theory and derived the governing
equations for a wide spectrum of grain–fluid mixture flows
based on a two-phase analysis. The Coulomb mixture model
assumes that solids and interstitial fluids in debris flows be-
have constitutively as Coulomb frictional materials and
Newtonian viscous fluids, respectively. Advection–diffusion
equations are postulated to describe pore pressure changes
in response to the movement of solids (Iverson and Denlin-
ger 2001). Denlinger and Iverson (2004) developed a mathe-
matical depth-averaged model using a rectangular Cartesian
coordinate system to simulate granular avalanches across ir-
regular three-dimensional terrains. The equations are solved
numerically using a hybrid finite volume and finite element
scheme. Fluxes of mass and momentum across vertical cell
boundaries are computed using the finite volume method
and stresses accompanying the deformation within the gran-
ular avalanche are calculated using the finite element
method. The stresses are first evaluated based on constitutive
relations for an isotropic linear elastic material and then cor-
rected using the Coulomb yield criterion. The model was
tested against laboratory experiments of dry sand avalanches
across irregular basal topography. Good agreement was ob-
tained between theoretical prediction and experimental tests
(Iverson et al. 2004).
Pitman and Le (2005) proposed a two-fluid model for
granular flows of the mixture of solid particles and fluid.
Continuity and momentum equations are formulated explic-
itly for both solid and fluid phases in the two-fluid model.
Interactions between particles and the fluid are taken into
account in the model with a velocity-dependent force. Puda-
saini et al. (2005a, 2005b) extended the Savage–Hutter
model for debris flow simulation by including pore pres-
sures in the model. The extended model has been applied to
analyzing debris flow flume tests, and good agreement is
obtained between theoretical predictions and experimental
observations (Pudasaini et al. 2005a, 2005b). However, the
pore pressures are not predicted; they are merely assumed
by employing an advection–diffusion equation similar to
those proposed by Iverson and Denlinger (2001).
The Savage–Hutter model and its generalized versions
have been tested against laboratory experiments of rapid
granular flows over a wide variety of bed topographies
(Hutter and Koch 1991; Greve and Hutter 1993; Wieland et
al. 1999; Iverson et al. 2004). Theoretical predictions were
found to be in good agreement with experimental measure-
ments. The Savage–Hutter model and its various generalized
versions have been established as the leading models in the
area of dry granular flow analysis (Pudasaini and Hutter
2007). However, their broad applications are limited to mod-
eling of dry granular flows over the simple topography ana-
lyzed in most laboratory experiments.
Slice-based dynamic analysis has been developed and
used in studies of fast-moving gravitational flows, such as
rapid landslides, debris flows, and liquefaction flow slides.
Hungr (1995), Tinti et al. (1997), Miao et al. (2001), and
Kwan and Sun (2006) are examples. In the slice-based
model, the flowing mass is represented by an ensemble of
contiguous slices (two dimensions) or blocks (three dimen-
sions) that are subjected to gravitational forces, basal resist-
ance, and internal forces. While sliding down over a
specified sliding surface, the slices interact with each other,
dissipating energy along their bases and exchanging momen-
tum between adjacent slices. The shape of the slices can be
changed during the flow process, but total volume is con-
served and no mass can penetrate between slices. Newton’s
laws of motion are employed to define the relationship be-
tween slice movement and forces acting on each slice. A
Lagrangian description is generally adopted to easily deter-
mine locations of the slices. A general survey of the
momentum-based models indicates that the methods of cal-
culation of gravitational forces and basal resistance are very
similar. It is the treatment of internal forces that distin-
guishes one method from another (Wang 2008).
1402 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Hungr (1995) presented a dynamic model (DAN) for run-
out analysis of rapid landslides. DAN is based on an explicit
solution of the Saint-Venant equations with the integration
of a variety of constitutive relationships that have been
widely applied in debris flow modeling. The displaced mate-
rials of a landslide are represented by a number of boundary
elements and mass elements. Formulation of the governing
equations is based on the law of conservation of momentum
for the boundary elements. Mass conservation is applied to
the mass elements to calculate changes in flow depth and in-
ternal deformation within the landslide. Flow resistance
along the sliding surface is determined using the constitutive
law specified in the analysis. The resultant longitudinal pres-
sures acting on the boundary elements are determined from
the product of the hydrostatic pressure gradient of the equiv-
alent fluid and a lateral pressure coefficient (Hungr 1995).
The value of the lateral pressure coefficient is dependent on
the coefficients of active and passive earth pressures, the
stiffness coefficient, and the average tangential strain of the
mass elements. Initially, the coefficient of lateral pressure at
rest is used for each element. The magnitude of the lateral
coefficient is then increased or decreased by a value equal
to the product of the incremental strain times a stiffness co-
efficient. The lateral coefficient can vary between the mini-
mum and maximum values, which correspond to the active
and passive states, respectively. Hungr (2003) proposed that
the values of the active and passive earth pressures can be
determined based on internal frictional strength of the slid-
ing material using the equation proposed by Savage and
Hutter (1989).
An improved debris mobility model (DMM) was pre-
sented by Kwan and Sun (2006). DMM is based on the
model for the dynamic analysis of rapid landslides (DAN)
developed by Hungr (1995). The formulation of flow resist-
ance in DMM is based on the entire wetted perimeter of the
flow channel. For simplicity, the cross section of the flow
channel is approximated by a trapezoid in the modified
DMM formulation. Calculation of internal forces in DMM
is based on the hydrostatic gradient and lateral pressure co-
efficient, which is dependent on the local internal deforma-
tion states of the landslide. The method proposed by Hungr
(1995) is used in DMM to calculate the coefficient of lateral
earth pressure. The total flow resistance consists of resistan-
ces developed at the base of each slice and on the two side
slopes of the channel. The basal resistance is calculated us-
ing the same procedure as in the original formulation of
DAN (Hungr 1995). The resistances on side slopes are de-
termined from the sliding mass above the side slopes and
the constitutive law specified in the analysis. Applications
of DMM and DAN in modeling rapid landslides in Hong
Kong show that very similar front velocities are predicted
by both models (Kwan and Sun 2006). However, the back-
calculated strengths from DAN are greater than those deter-
mined from the improved DMM due to incorporation of re-
sistances on the side slopes in the DMM model.
A block-based model was proposed by Tinti et al. (1997)
to simulate movements of flow slides over slopes. Interslice
forces in the block-based model are expressed in terms of an
interaction coefficient, which is dependent on the dynamic
states of the slices and the instantaneous distance between
the slice centers. The value of the interaction coefficient
varies between 0 and 1, which represent two limiting condi-
tions for interslice actions. When the interaction coefficient
is equal to 0, interacting slices have the same post-interaction
velocities irrespective of their pre-interaction states, and
slices move as if they are adhered to each other; this is
the maximum possible interaction. When the interaction
coefficient is equal to 1, the slices retain their pre-interac-
tion velocities as if no interaction had taken place. In gen-
eral, the interaction coefficient can be determined from a
function in terms of interaction intensity, deformability pa-
rameter, and shape parameter (Tinti et al. 1997).
Miao et al. (2001) proposed a sliding block model for the
runout analysis of rapid landslides. The model starts with a
limit equilibrium assessment and incorporates mass dynam-
ics and soil deformation into the calculation of sliding
movement. In the sliding block model, the sliding mass is
divided into a number of slices. Forces acting on a single
slice consist of gravity, basal resistance, and interactions be-
tween slices. The initial interslice forces are determined
from the limit equilibrium analysis using the method of un-
balanced thrust, in which the resultant interslice force is as-
sumed to be parallel to the base of the preceding slice and
acts at the midpoint of the height of the slice. The critical
state corresponding to a factor of safety of 1 is considered
as the initial state. Initial acceleration of the slices is calcu-
lated based on the initial unbalanced forces. After the mass
movement is triggered, the interslice forces are determined
from the deformation energy, which is calculated using the
macroscopic deformation of the slices and bulk deformation
modulus. For slices in a tensile state, the deformation energy
vanishes and the interslice forces are assumed to be zero.
The basal shear resistance is determined based on the over-
burden normal force using the Coulomb frictional law, sim-
ilar to the methods used by other sliding block models.
Application of the sliding block model to two rapid land-
slides in China (Miao et al. 2001) indicated that the sliding
block model can result in considerable fluctuations in the
velocity field within very short periods of time.
Formulation of the slice-based model using
the energy conservation law
The existing slice-based models incorporate the effects of
internal deformation on the flow dynamics using either the
interaction coefficient (Tinti et al. 1997) or deformation en-
ergy (Miao et al. 2001). The interslice forces proposed in
the slice-based model are primarily dependent on the magni-
tude of deformation and dynamic state of the slices. How-
ever, application of the slice-based models of Tinti et al.
(1997) and Miao et al. (2001) to rapid landslide simulation
is either difficult or impractical because of the great uncer-
tainties involved in the calculation of the interslice force and
the determination of model input parameters required. For
instance, the significant fluctuations in the velocity field pre-
dicted by Miao et al.’s (2001) model are caused mainly by
inappropriate computation of the interaction forces in terms
of the deformation energy. A slice-based model incorporat-
ing internal energy dissipation is presented in this paper.
The formulation of the new model is based mainly on the
conservation of energy of slices during the flow process.
Wang et al. 1403
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Mechanical energy equations
The rate of change of the kinetic energy of a material vol-
ume can be written as the sum of three parts (Aris 1962):
(1) The rate at which the body forces do work, i.e., change
in potential energy.
(2) The rate at which the surface stresses do work.
(3) The rate at which the internal stresses do work, i.e.,
change in internal energy due to deformation.
The kinetic energy equation for a unit volume mass is
given by:
½1Š r
d
dt
1
2
u
i
u
i
_ _
¼ rg
i
u
i
þ
@
@x
j
ðu
i
t
ij
Þ À t
ij
e
ij
where r is the density of the fluid, t is time,
u
i
u
i
¼ u
2
1
þ u
2
2
þ u
2
3
is kinetic energy per unit volume, u
i
is
the component of the velocity vector in the x
i
direction, rg
i
is the body force per unit volume in the x
i
direction (where
g
i
is the component of the acceleration vector), t
ij
represents
components of the stress tensor, and e
ij
represents compo-
nents of the strain rate tensor.
An integral form of the mechanical energy equation can
be derived by integrating each term in eq. [1] over a mate-
rial volume or a slice of a landslide body, V. Using Gauss’
theorem, eq. [1] in the Lagrangian system becomes
½2Š
d
dt
_
V
1
2
ru
i
u
i
_ _
dV ¼
_
V
rg
i
u
i
dV þ
_
A
u
i
t
ij
dA
j
À
_
V
t
ij
e
ij
dV
Each term in eq. [2] is a time rate of change: the first
term is the rate of the kinetic energy of a material volume,
the second is the rate of total work done by the body force
acting on a material volume, the third represents the total
work done by the forces acting on the surface of a material
volume, and the fourth represents the rate of energy dissipa-
tion due to deformation of a material volume. The deforma-
tion work includes two parts: the work due to volume
expansion and the work due to irreversible deformation. For
an incompressible fluid, the work due to volume expansion
is zero and the deformation work represents irreversible ki-
netic energy dissipation. Thus, the term
_
V
t
ij
e
ij
dV repre-
sents a rate of loss of kinetic energy and a gain of internal
energy due to deformation of the volume.
The body force can be represented as the gradient of a
scalar potential. The rate of work done by body forces can
be taken to the left-hand side of eq. [2] and can be inter-
preted as a change in the potential energy. For an incom-
pressible fluid and debris, r is constant and the mass within
a material volume can be determined from
½3Š m ¼
_
V
r dV
and eq. [2] can be rewritten as
½4Š
d
dt
1
2
m u
2
i
þ mgz
W
_ _
¼
_
A
u
i
t
ij
dA
j
À
_
V
t
ij
e
ij
dV
where u
i
is the average velocity component of a material vo-
lume, g is gravitational acceleration, z
W
is the elevation of
the center of gravity of the sliding mass, and A
j
is the com-
ponent of the surface area vector. The left-hand side of
eq. [4] thus represents the rate of change of mechanical en-
ergy (the sum of kinetic energy and potential energy).
Derivation of governing equations based on the
conservation of energy
The sliding mass is divided into a series of contiguous sli-
ces, as shown in Fig. 1. Forces acting on a typical slice of
width b and height h are shown in Fig. 2. In the figures, b
is the width of the slice, h is the height of the slice, N is
the normal force, subscript n is the slice number, P is the
interslice force, T is the shear force acting along the base of
the slice, u is the mean velocity of the slice along the base
of the slice, W is the weight of the slice, q is the inclination
of the base of the slice with respect to the horizontal, and
subscripts L and R denote properties on the left and right
sides, respectively, of a slice.
As indicated in Figs. 1 and 2, the derivation of governing
equations is based on the deformation in the flow direction.
Effects of side-wall deformation perpendicular to the flow
direction are not included in the formulation. With the as-
sumption that the interslice shear forces in the vertical direc-
tion are negligible, the rate of total work done by the surface
forces in eq. [4] is given by
½5Š
_
A
u
i
t
ij
dA
j
¼ P
L
u
L
cosq
L
À P
R
u
R
cosq
R
À T u
where q
L
and q
R
are inclinations between velocities and cor-
responding interslice forces acting on the left and right
sides, respectively, of the boundary.
The rate of change of the total potential energy of a slice
in eq. [4] can be expressed as the sum of the rate of poten-
tial energy changes due to the displacement and deformation
of the slice, as shown in Fig. 2
½6Š
d
dt
ðmgz
w
Þ ¼
d
dt
ðmgz
w1
Þ þ
d
dt
ðmgz
w2
Þ
½7Š m ¼ rbh
where (d/dt)(mgz
w1
) represents the rate of change of poten-
tial energy due to displacement of the center of the gravity,
(d/dt)(mgz
w2
) represents the rate of change of the potential
energy due to deformation of the slice, and r is the mean
density of the sliding mass. Thus
½8Š
d
dt
ðmgz
W1
Þ ¼ Àmg u sinq
½9Š
d
dt
ðmgz
W2
Þ ¼ À
1
2
mgh e
zz
where e
zz
is the mean vertical strain rate of a slice defined in
Fig. 3.
Using eqs. [5] to [9], eq. [4] becomes
½10Š
d
dt
1
2
m u
2
_ _
¼ mg u sinq þ
1
2
mgh e
zz
þ P
L
u
L
cosq
L
À P
R
u
R
cosq
R
À T u À
_
V
t
ij
e
ij
dV
1404 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Summing both sides of eq. [10] for all slices (k = 0 to n
in Fig. 4) gives the rate of change of the kinetic energy of
the overall sliding mass:
½11Š
d
dt

n
k¼0
1
2
m
k
u
2
k
_ _
¼

n
k¼0
m
k
g u
k
sinq
k
þ
1
2
ðh e
zz
Þ
k
_ _
À

n
k¼0
T
k
u
k
À

n
k¼0
_
V
k
t
ij
e
ij
dV
k
It is obvious that the rates of work done by interslice forces
cancel each other in the preceding derivation. Equation [11]
states that the rate of the change of kinetic energy of a sliding
mass is equal to the sum of the rate of potential energy
change, the rate of work done by resistance force along the
base of the sliding mass, and the rate of deformation work.
Numerical scheme for the analytical model
based on energy considerations
Depth-averaged governing equations can be solved using
either a Lagrangian or Eulerian scheme. Two finite difference
Fig. 3. Uniform deformation assumed in the slice-based model. e
xx
, mean horizontal strain rate.
Fig. 2. Forces on a typical slice.
Fig. 1. Representative slice in the slice-based model.
Wang et al. 1405
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
methods — one Lagrangian the other Eulerian — have been
tested against flume experiments on dry granular flows
(Savage and Hutter 1989). Simulations indicated that the Eu-
lerian scheme based on the upwind flux correction method
gave poor predictions of experimental avalanches with gen-
eral initial profiles. The Lagrangian finite difference approach
(Savage and Hutter 1989, 1991) is simple, efficient, and reli-
able for the granular rapid flow problem that involves a free
surface, dry bottom, and moving boundaries. The approach is
based on a Lagrangian, moving mesh, finite difference
scheme in which the flowing material is divided into quadri-
lateral cells in two dimensions or triangular prisms in three
dimensions. Boundary locations are determined for each time
step. The depth of a cell is calculated from the cell volume
and boundary locations. Numerical simulations of the flume
experiments on dry granular flows showed very good agree-
ment between theoretical predictions and experimental data
(Savage and Hutter 1989, 1991; Wieland et al. 1999).
Numerical methods proposed by Hungr (1995), Tinti et al.
(1997), Miao et al. (2001), and Kwan and Sun (2006) for
solving the governing equations of fast-moving landslides
have many similarities. These numerical methods are based
on the Lagrangian finite difference scheme similar to that
formulated by Savage and Hutter (1989).
Following the procedures of Savage and Hutter (1989), the
Lagrangian finite difference scheme is presented here for solv-
ing the equations of the slice-based model with internal energy
dissipation. In the Lagrangian scheme, the sliding mass is dis-
cretized into a number of slices, as shown in Fig. 4. In Fig. 4,
(h
c
)
k
denotes the average height of the center of slice k and
(x
c
)
k
, (x
b
)
k
, and (x
b
)
k+1
denote the locations of the center (sub-
script c) and boundaries (subscript b) of slice k. Mass conser-
vation of slice k during the landslide motion indicates
½12Š
d
dt
ðV
k
Þ ¼ 0
where V
k
is the volume of slice k. The mean height of a
slice at time t can thus be determined from
½13Š ðh
c
Þ
t
k
¼
V
k
ðx
b
Þ
t
kþ1
À ðx
b
Þ
t
k
Solving the governing equations of rapid landslides re-
quires determination of the positions of the boundaries of
each slice at time t. The numerical scheme assumes that all
the variables involved in the calculation at t + Dt are known
from the previous time t, where Dt is the time step interval.
In the framework of the Lagrangian finite difference
scheme, the governing eq. [10] can be written as
½14Š
E
tþDt
k
À E
t
k
Dt
¼
_
W
t
k
½15Š E
k
¼
1
2
m
k
u
2
k
½16Š
_
W
k
¼ m
k
g u
k
sinq
k
þ
1
2
m
k
gh
k
ð e
zz
Þ
k
þ ðP
L
u
L
cosq
L
Þ
k
À ðP
R
u
R
cosq
R
Þ
k
À T
k
u
k
À
_
V
k
t
ij
e
ij
dV
where E
k
is the kinetic energy of slice k, and
_
W
k
is the sum
of the rate of the work done by body force, surface force,
and energy dissipation due to deformation of slice k. E
k
and
_
W
k
are determined from eqs. [15] and [16], respectively.
The constitutive law and assumptions regarding the inter-
slice forces and deformation work are required for calculat-
ing the rate of work in eq. [16]. The basal shear resistance
can be determined based on the constitutive laws for materi-
als. For Mohr–Coulomb materials, the basal resistance along
the base of the slice can be expressed as
½17Š T
k
¼ tA
k
¼ ðc þ s tanf
b
ÞA
k
¼ cA
k
þ N
k
tanf
b
where T
k
and N
k
are shear resistance and normal force acting
on the base of slice k, t is shear stress, A
k
is basal area of
slice k, c is cohesion, s is normal stress, and f
b
is basal fric-
tion angle.
Equation [17] is a generalized form of the Mohr–Coulomb
equation. In dense granular flow simulations, either a
purely cohesive or a frictional model is generally used as
a constitutive law in back-analyses of liquefaction flow
slides or debris flows. It is evident that if the friction angle
is equal to 0 (f
b
= 0), eq. [17] reduces to the purely cohe-
sive model; if the cohesive strength is equal to 0 (c = 0),
eq. [17] reduces to the frictional model. Wang (2008)
back-analyzed case histories of liquefaction flow slides us-
ing the cohesive and friction models. The effects of con-
stitutive models on the dynamic analyses were discussed
by comparing model predictions with field observations in
terms of runout distance, deposit distribution, and velocity
profiles.
In granular flow simulations, lateral pressure can be ap-
proximated as a product of hydrostatic pressure and the co-
efficient of lateral pressure, if the frictional model is used as
a constitutive law. By neglecting interslice shear forces in
the vertical direction, the lateral pressure can be written as
½18Š ðP
L
Þ
k
¼
ðK
L
Þ
k
gðh
b
Þ
2
k
2
½19Š ðP
R
Þ
k
¼
ðK
R
Þ
k
gðh
b
Þ
2
kþ1
2
where (K
L
)
k
and (K
R
)
k
are the lateral earth pressure coeffi-
cients on the left and right sides, respectively, of slice k;
g = rg is the unit weight of the sliding mass; and (h
b
)
k
and
(h
b
)
k+1
are flow depths on the left and right sides of slice k
(Fig. 4), respectively. The lateral stress coefficient can be
active, passive or hydrostatic (coefficient of lateral stress
equal to 1) based on the local strain rate (velocity gradient)
of a slice in the longitudinal direction.
Equations proposed by Savage and Hutter (1989) have
been commonly used for calculating the coefficients of lat-
eral earth pressure in the analysis of dense granular flows.
The derivation of the Savage–Hutter equations for the lateral
stresses assumes that Coulomb failure occurs simultaneously
along the bed and within the sliding mass. Therefore, the
definition of the Savage–Hutter coefficients is often as-
sumed to be more representative than those of Rankine lat-
eral earth pressure coefficients used in classical soil
mechanics. Examination of the Savage–Hutter equations in-
dicates that the coefficient of lateral earth pressure would be
1406 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
greater than unity even as a slice is expanding, i.e., in an ac-
tive state, if the value of the basal friction angle is close to
that of the internal friction angle (Fig. 5). This is physically
inadmissible and inconsistent with the assumptions for the
model presented here. In the model developed in this paper,
the active or passive state of stress on the left side of a slice
is dependent on whether the slice is expanding or contract-
ing. It is also assumed that the contribution to the magnitude
of lateral stress from internal deformations is more signifi-
cant than that from the materials of the thin shear zone
along the bed. The values of the lateral stress coefficients
are calculated for a frictional material using the Rankine
equation
½20Š ðK
L
Þ
k
¼
1 À sinf
1 þ sinf
¼ tan
2
45 À
f
2
_ _
@u
@x
_ _
k
> 0
1
@u
@x
_ _
k
¼ 0
1 þ sinf
1 À sinf
¼ tan
2
45 þ
f
2
_ _
@u
@x
_ _
k
< 0
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
Fig. 5. Coefficients of lateral earth pressure based on the Savage–Hutter (Savage and Hutter 1989, 1991) definition.
Fig. 4. Notations used for the slice-based model. h, height of the slice; b, width of the slice; subscript c, center; subscript b, boundary.
Wang et al. 1407
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
where f is the internal friction angle.
For a purely cohesive material, the internal friction angle
is equal to 0 and the total lateral forces calculated using the
Rankine theory are
½21Š ðP
L
Þ
k
¼
1
2
gðh
b
Þ
2
k
À 2cðh
b
Þ
k
@u
@x
_ _
k
> 0
1
2
gðh
b
Þ
2
k
@u
@x
_ _
k
¼ 0
1
2
gðh
b
Þ
2
k
þ 2cðh
b
Þ
k
@u
@x
_ _
k
< 0
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
According to eq. [21], the calculated horizontal force is
negative when a slice is in the active state and (hb)
k
< (4c/g).
Because tensile stresses occur rarely in soils, eq. [21] is
modified as follows:
½22Š ðP
L
Þ
k
¼
1
2
gðh
b
Þ
2
k
@u
@x
_ _
k
! 0
1
2
gðh
b
Þ
2
k
þ 2cðh
b
Þ
k
@u
@x
_ _
k
< 0
_
¸
¸
¸
_
¸
¸
¸
_
Deformation of the slice is simplified as a pure shear de-
formation, as shown in Fig. 3. The deformation work rate is
approximated by
½23Š
_
V
k
t
ij
e
ij
dV ¼ ð t
xx
e
xx
þ t
zz
e
zz
Þ
k
b
k
ðh
c
Þ
k
where
ð e
xx
Þ
k
¼ Àð@=@xÞðu cosqÞ
k
¼ Àðu
R
cosq
R
À u
L
cosq
L
Þ
k
=b
k
is the mean horizontal strain rate and t
xx
and t
zz
are mean
horizontal and vertical stresses, respectively. Following the
conventions of stress and strain representation usually
adopted in geotechnical engineering, negative signs have
been introduced so that compressive stresses and compres-
sive strains are positive quantities.
For an incompressible fluid
½24Š e
xx
þ e
zz
¼ 0
½25Š ð t
zz
Þ
k
¼
gðh
c
Þ
k
2
Fig. 6. Comparison between analytical and numerical solutions of a dam break over a horizontal plane with a 08 basal friction angle.
1408 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
For a frictional material, lateral stresses are computed from
½26Š ð t
xx
Þ
k
¼ K
k
ð t
zz
Þ
k
where K
k
is calculated by averaging the coefficients of lat-
eral stress on left and right sides of the slice
½27Š K
k
¼
ðK
L
Þ
k
þ ðK
R
Þ
k
2
For a purely cohesive material, the coefficient of lateral
earth pressure is equal to 1 and the average lateral stresses
inside slice k can be calculated by
½28Š ð t
xx
Þ
k
¼
ð t
zz
Þ
k
@u
@x
_ _
k
! 0
ð t
zz
Þ
k
þ 2c
@u
@x
_ _
k
< 0
_
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
_
Using eqs. [18] to [28], the kinetic energy of slice k at
time t + Dt can be determined from
½29Š E
tþDt
k
¼ E
t
k
þ
_
W
t
k
Dt
The center velocity of a slice at time t + Dt is
½30Š ðu
c
Þ
tþDt
k
¼
ffiffiffiffiffiffiffiffiffiffiffiffiffi
2E
tþDt
k
m
k
¸
The boundary velocity is approximated as
½31Š ðu
b
Þ
tþDt
k
¼
ðu
c
Þ
tþDt
kÀ1
þ ðu
c
Þ
tþDt
k
2
Displacement of the slice boundary is
½32Š ðx
b
Þ
tþDt
k
¼ ðx
b
Þ
t
k
þ
ðu
b
Þ
t
k
cosq
t
k
þ ðu
b
Þ
tþDt
k
cosq
tþDt
k
2
Dt
where ðx
b
Þ
t
k
and ðx
b
Þ
tþDt
k
are x coordinates of boundaries of
slice k at time t and t + Dt, respectively.
The height of slice k at t + Dt is computed by
½33Š
ðh
c
Þ
tþDt
k
¼
V
k
½ðx
b
Þ
tþDt
kþ1
À ðx
b
Þ
tþDt
k
Š
ðh
b
Þ
tþDt
k
¼
ðh
c
Þ
tþDt
kÀ1
þ ðh
c
Þ
tþDt
k
2
½34Š ðh
b
Þ
tþDt
0
¼ 0; ðh
b
Þ
tþDt
n
¼ 0
Fig. 7. Comparison between analytical and numerical solutions of a dam break over a 308 slope with a 08 basal friction angle.
Wang et al. 1409
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
where V
k
is the volume of slice k, ðh
b
Þ
tþDt
k
is the height of
the left boundary of slice k at t + Dt, and ðh
c
Þ
tþDt
k
is the cen-
tral height of slice k at t + Dt.
At time t = 0, the velocities and kinetic energy of the sli-
ces are equal to zero. The initial movement of slice k can be
determined from momentum conservation
½35Š m
k
d u
k
dt
¼ m
k
g sinq
k
À T
k
þ ðP
L
À P
R
Þ
k
cosq
k
The velocities and displacements of slices during the first
time step can be calculated as long as accelerations are ob-
tained from eq. [35]. The motion of each slice can then be
determined using equations of energy conservation with the
Lagrangian difference scheme presented above. The compu-
tation proceeds until the maximum slice velocity is under
the velocity threshold specified. It is worthwhile to note
that the dynamic model based on energy consideration as-
sumes that lateral pressure and basal resistance on individual
slices are known and defined by the Rankine and Mohr–
Coulomb equations. The momentum equilibrium of the
overall sliding mass is not examined during the calculation
and assumptions about the forces do not satisfy statics. As a
consequence, the dynamic analysis using the new analytical
model cannot converge to the static case because kinematics
are not being considered in the formulation and the internal
force distribution implicit as calculated in the static case dif-
fers from that assumed in the model. The threshold velocity
plays a key role in terminating the calculation process.
Based on the classification of flow-like landslides (Hungr et
al. 2001), a value of 0.05 m/s can be used as a velocity
threshold in the analyses of case histories of flow slides and
debris flows, which are among the extremely rapid class.
Model verification and numerical
experiments
Performance of the slice-based model and numerical
scheme presented in previous sections has been tested by
comparing model predictions with analytical solutions of
one-dimensional granular flows and experimental observa-
tions of granular slumping on a horizontal plane.
Comparison between model predictions and analytical
solutions
Mangeney et al. (2000) presented an analytical solution
for a one-dimensional granular avalanche over an inclined
plane. The analytical solution describes the motion of a
flow front of the dam-break granular flow over an infinite,
uniform slope with a Coulomb-type friction acting at the
base of the flow. It should be noted that the analytical solu-
tion of Mangeney et al. (2000) can only be applied to an
idealized flow where (i) lateral earth pressure is assumed to
be hydrostatic, (ii) the basal friction angle is not greater than
the slope angle, and (iii) the flow never stops on the slope.
Fig. 8. Comparison between analytical and numerical solutions of a dam break over a 308 slope with 108 basal friction angle.
1410 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Figures 6 to 9 present a comparison between the analyti-
cal and numerical solutions of dam break scenarios on hori-
zontal and inclined planes. Numerical analyses were carried
out using the model developed in this paper. An internal
friction angle of zero has been used to provide the hydro-
static lateral pressure distribution in all the simulations. Fig-
ure 6 shows the result of the dram break scenarios over a
horizontal plane with zero basal friction. Figures 7 to 9
present comparisons between analytical solutions and nu-
merical predictions for dam break scenarios on a 308 slope
with and without the basal frictional resistance applied. It is
observed that numerical simulations provide a good repre-
sentation of the dam break–induced granular flows over hor-
izontal and inclined planes.
Simulation of granular slumping on a horizontal plane
Granular flows induced by the collapse of initially static
columns of sand over a horizontal surface were investigated
experimentally by Lajeunesse et al. (2005). Effects of the in-
itial column geometry on the flow runout behavior and in-
ternal flow structure were explored in the experiments.
Granular materials used in the experiments were glass beads
1.15 or 3 mm in diameter. Granular materials were initially
contained within a cylindrical or rectangular tank. Axisym-
metric or two-dimensional granular flows were created by
quickly raising the cylinder or removing the gate. Experi-
mental observation demonstrated that the flow dynamics
and final deposit depends on the initial aspect ratio of the
granular column, which is defined as the ratio of the initial
height to horizontal extent of the column.
Numerical simulations of the spreading of granular col-
umns on a horizontal plane have been conducted using the
dynamic model formulated based on energy conservation.
The value of the internal and basal friction angles used in
the analysis is 258, which is the average of values reported
by Lajeunesse et al. (2005). Figure 10 presents the normal-
ized final profiles of numerical simulations and experimental
observations of the spreading of columns with the same ini-
tial aspect ratio, but witha different initial basal length.
Comparison of theoretical and experimental final profiles in
Fig. 10 indicates that the dynamic analysis provides a rea-
sonable prediction of the runout distance for flows induced
by the collapse of granular columns over a horizontal plane.
It should be noted, as shown in Fig. 11, that when the
Savage–Hutter equations for the coefficients of the lateral
stresses are used in the governing equations based on energy
conservation, the fit is not as good either in terms of runout
distance or shape.
It has been observed that granular slumping involves the
following two processes: (i) collapse and fall of the column
and (ii) spreading of the granular mass on a horizontal plane
until it comes to rest. During these processes, the initial po-
tential energy in the tall column is converted into kinetic en-
ergy and is also dissipated because of internal deformation
Fig. 9. Comparison between analytical and numerical solutions of a dam break over a 308 slope with 208 basal friction angle.
Wang et al. 1411
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
and basal friction. The complexity of the collective dynam-
ics of momentum transfer and energy dissipation involved in
granular slumping highlights the difficulties in modeling this
class of problems within the framework of classical shallow
water equations. To simulate the slumping of the granular
mass, it appears that a model should be capable of account-
ing for vertical momentum transfer associated with the fall
of the column and also capturing the features of the subse-
quent horizontal spreading. It is possible to apply a depth-
averaged model to modeling sideways flow of a granular
mass. Unfortunately, the initial vertical column collapse and
momentum transfer intrinsically violate the shallow water
assumptions and cannot be accounted for by shallow water
approaches. It is evident that the theoretical predictions are
far from describing the whole process of granular slumping
and there are still many issues to be resolved. However, it
has been observed surprisingly that the simulations carried
out by using the new analytical model are able to reproduce
many features of the spreading of a granular mass. The
model based on energy considerations provides new insights
into the approaches used for investigating granular flows.
The model has been used in modeling a number of debris
flow cases (Wang 2008).
Discussion
Flow slides and debris flows generally exhibit pervasive,
fluid-like deformation. Morgenstern (1978) investigated mo-
bile flows in a wide variety of geological and geomorpho-
logical settings and concluded with the observations that
characterization of mobile soil and rock flows and the de-
sign of protective structures should proceed using principles
of fluid mechanics. The fundamental equations of flow dy-
namics are based on three universal laws of conservation:
conservation of mass, conservation of momentum, and con-
servation of energy. During the last three decades, conserva-
tion laws of mass and momentum have been widely applied
in the formulation of numerical models for granular flow
simulations (Savage and Hutter 1989; Hungr 1995; Iverson
1997; Denlinger and Iverson 2001, 2004; Iverson and Den-
linger 2001; Iverson et al. 2004). However, energy conserva-
tion has not been explicitly accounted for in the granular
flow simulations.
Morgenstern (1978) pointed out that large volumes of soil
or rock can become fluidized by virtue of energy transfer
mechanisms following instability. Iverson et al. (1997) pro-
vided an extended discussion on mobilization of debris
Fig. 10. Nondimensional final deposit of granular slumps with the same initial aspect ratio of 3.2 (simulation using Rankine coefficient of
lateral earth pressure).
1412 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
flows from landslides and noted that three processes are in-
volved in the mobilization:
(1) Widespread Coulomb failure within the sediment mass.
(2) Partial or complete soil liquefaction by high pore-fluid
pressure.
(3) Conversion of landslide translational energy to internal
vibrational energy.
These three processes operate simultaneously and synerg-
istically in many circumstances and the change of pore-fluid
pressures and conversion of energy are crucial components
in flow slide and debris flow mobilization and evolution.
Formulation of the governing equation (eq. [10]) is based
on energy considerations and idealization of complex energy
transfer mechanisms involved in mobilization, motion, and
deposition of flow slides and debris flows. Within the
framework of the universal energy conservation law, energy
dissipation due to internal deformation is integrated into
geotechnical analysis of granular flows explicitly. Based on
conservation laws of mass and energy, the analytical model
developed in this paper provides new insights into quantita-
tive analysis of granular flows in geotechnical settings.
Generally speaking, the momentum conservation and en-
ergy conservation principles provide identical results when
applied to certain flow problems. However, it has long been
realized that particle interactions, such as intergrain collision
and sliding, dissipate energy in granular flows (Jaeger and
Nagel 1992; de Gennes 1999). The internal forces in granu-
lar flows are much more complicated than the assumptions
in the existing depth-averaged continuum models. Formula-
tion of governing equations for granular flows within the
framework of energy conservation offers a simpler and
clearer explanation based on granular flow physics and lends
itself to the incorporation of other phenomena, such as those
related to particle breakdown.
Conclusions
A slice-based model is proposed to simulate debris flows,
flow slides, and other types of rapid landslides from a geo-
technical perspective. The model is formulated based on
universal conservation laws of mass and energy. Effects of
deformation work and internal energy dissipation on debris
flow dynamics are accounted for explicitly in the proposed
model.
A Lagrangian finite difference method has been presented
to solve the governing equations for the granular flow simu-
lations. The terms due to deformation work in granular
flows are incorporated into the solution to account for the
effects of internal energy dissipation. Simulations of simple
granular flows have been undertaken to examine the plausi-
bility of the model and applicability of the numerical
Fig. 11. Nondimensional final deposit of granular slumps with the same initial aspect ratio of 3.2 (simulation using Savage–Hutter coeffi-
cient of lateral earth pressure).
Wang et al. 1413
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
scheme. The results of granular flow simulations provide
evidence that the model based on the energy conservation
law is robust.
Acknowledgements
The authors would like to thank Professor P. Steffler at
the University of Alberta for contributions to the formula-
tion of the analytical model and numerical solution pre-
sented in the paper. The authors also benefitted from
discussions with Professor O. Hungr at The University of
British Columbia and Professor S.M. Olson at the University
of Illinois at Urbana–Champaign.
This research has been sponsored by the Natural Sciences
and Engineering Research Council of Canada.
References
Aris, R. 1962. Vectors, tensors, and the basic equations of fluid
mechanics. Dover Publications, Inc., New York.
de Gennes, B.G. 1999. Granular matter: a tentative view. Reviews
of Modern Physics, 71(2): S374–S382. doi:10.1103/
RevModPhys.71.S374.
Denlinger, R.P., and Iverson, R.M. 2001. Flow of variably fluidized
granular masses across three-dimensional terrain 2. Numerical
predictions and experimental tests. Journal of Geophysical Re-
search. Solid Earth, 106(B1): 553–566. doi:10.1029/
2000JB900330.
Denlinger, R.P., and Iverson, R.M., 2004. Granular avalanches
across irregular three-dimensional terrain: 1. Theory and compu-
tation. Journal of Geophysical Research - Earth Surface 109(F1):
F01014.1–F01014.14. doi:10.1029/2003JF000085.
Greve, R., and Hutter, K. 1993. Motion of a granular avalanche in
a convex and concave curved chute: experiments and theoretical
predictions. Philosophical Transactions of the Royal Society A:
Physical Sciences and Engineering, 342(1666): 573–600. doi:10.
1098/rsta.1993.0033.
Huang, X., and Garcı ´a, M.H. 1998. A Herschel–Bulkley model for
mud flow down a slope. Journal of Fluid Mechanics, 374: 305–
333. doi:10.1017/S0022112098002845.
Hungr, O. 1995. A model for the runout analysis of rapid flow
slides, debris flows, and avalanches. Canadian Geotechnical
Journal, 32(4): 610–623. doi:10.1139/t95-063.
Hungr, O. 2003. User’s manual of DAN-W: Dynamic analysis of
landslides. Release 4 [computer program]. Hungr Geotechnical
Research Inc., Vancouver, B.C.
Hungr, O., Evans, S.G., Bovis, M., and Hutchinson, J.N. 2001. Re-
view of the classification of landslides of the flow type. Envir-
onmental & Engineering Geoscience, VII: 221–238.
Hutchinson, J.N. 1986. A sliding–consolidation model for flow
slides. Canadian Geotechnical Journal, 23(2): 115–126. doi:10.
1139/t86-021.
Hutter, K. 2005. Geophysical granular and particle-laden flows: re-
view of the field. Philosophical Transactions of the Royal So-
ciety A: Mathematical, Physical and Engineering Sciences,
363(1832): 1497–1505. doi:10.1098/rsta.2005.1591.
Hutter, K., and Koch, T. 1991. Motion of a granular avalanche in
an exponentially curved chute: experiments and theoretical pre-
dictions. Philosophical Transactions of the Royal Society A:
Physical Sciences and Engineering, 334(1633): 93–138.
Iverson, R.M. 1997. The physics of debris flows. Reviews of Geo-
physics, 35(3): 245–296. doi:10.1029/97RG00426.
Iverson, R.M., and Denlinger, R.P. 2001. Flow of variably fluidized
granular masses across three-dimensional terrain 1. Coulomb
mixture theory. Journal of Geophysical Research. Solid Earth,
106(B1): 537–552. doi:10.1029/2000JB900329.
Iverson, R.M., Reid, M.E., and LaHusen, R.G. 1997. Debris-flow
mobilization from landslides. Annual Review of Earth and Plane-
tary Sciences, 25(1): 85–138. doi:10.1146/annurev.earth.25.1.85.
Iverson, R.M., Logan, M., and Denlinger, R.P. 2004. Granular ava-
lanches across irregular three-dimensional terrain: 2. Experimen-
tal tests. Journal of Geophysical Research. Earth Surface
109(F1): F01015.1–F01015.16. doi:10.1029/2003JF000084.
Jaeger, H.M., and Nagel, S.R. 1992. Physics of the granular state.
Science, 255(5051): 1523–1531. doi:10.1126/science.255.5051.
1523. PMID:17820163.
Jakob, M., and Hungr, O. 2005. Debris-flow hazards and related
phenomena. Praxis Publishing Ltd., Chichester, UK.
Johnson, A.M. 1970. Physical processes in geology. Freeman-
Cooper, San Francisco, Calif.
Johnson, A.M., and Rodine, J.R. 1984. Debris flow. In Slope in-
stability. John Wiley & Sons, Chichester, UK. pp. 257–361.
Kwan, J.S.H., and Sun, H.W. 2006. An improved landslide mobi-
lity model. Canadian Geotechnical Journal, 43(5): 531–539.
doi:10.1139/T06-010.
Lajeunesse, E., Monnier, J.B., and Homsy, G.M. 2005. Granular
slumping on a horizontal surface. Physics of Fluids, 17(10):
103302. doi:10.1063/1.2087687.
Mangeney, A., Heinrich, P., and Roche, R. 2000. Analytical solu-
tion for testing debris avalanche numerical models. Pure and
Applied Geophysics, 157(6–8): 1081–1096. doi:10.1007/
s000240050018.
Miao, T., Liu, Z., Niu, Y., and Ma, C. 2001. A sliding block model
for the runout prediction of high-speed landslides. Canadian
Geotechnical Journal, 38(2): 217–226. doi:10.1139/cgj-38-2-217.
Morgenstern, N.R. 1978. Mobile soil and rock flows. Geotechnical
Engineering, 9: 123–141.
Pitman, E.B., and Le, L.2005. A two-fluid model for avalanche and
debris flows. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences, 363(1832):
1573–1601.
Pudasaini, S.P., and Hutter, K. 2007. Avalanche dynamics: Dy-
namics of rapid flows of dense granular avalanches. Springer,
Berlin, Germany.
Pudasaini, S.P., Wang, Y., and Hutter, K. 2005a. Rapid motions of
free-surface avalanches down curved and twisted channels and
their numerical simulation. Philosophical Transactions of the
Royal Society A: Mathematical, Physical and Engineering
Sciences, 363(1832): 1551–1571. PMID:16011933.
Pudasaini, S.P., Wang, Y., and Hutter, K. 2005b. Modelling debris
flows down general channels. Natural Hazards and Earth System
Sciences, 5(6): 799–819. doi:10.5194/nhess-5-799-2005.
Savage, S.B., and Hutter, K. 1989. The motion of a finite mass of
granular material down a rough incline. Journal of Fluid Me-
chanics, 199: 177–215. doi:10.1017/S0022112089000340.
Savage, S.B., and Hutter, K. 1991. The dynamics of avalanches of
granular materials from initiation to runout. Part I: Analysis.
Acta Mechanica, 86(1–4): 201–223. doi:10.1007/BF01175958.
Tinti, S., Bortolucci, E., and Vannini, C. 1997. A block-based the-
oretical model suited to gravitational sliding. Natural Hazards,
16(1): 1–28. doi:10.1023/A:1007934804464.
Wang, X. 2008. Geotechnical analysis of flow slides, debris flows,
and related phenomena. Ph.D. thesis, University of Alberta, Ed-
monton, Alta.
Wieland, M., Gray, J.M.N.T., and Hutter, K. 1999. Channelized
free-surface flow of cohesionless granular avalanches in a chute
with shallow lateral curvature. Journal of Fluid Mechanics, 392:
73–100. doi:10.1017/S0022112099005467.
1414 Can. Geotech. J. Vol. 47, 2010
Published by NRC Research Press
C
a
n
.

G
e
o
t
e
c
h
.

J
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
n
r
c
r
e
s
e
a
r
c
h
p
r
e
s
s
.
c
o
m

b
y

U
N
I
V

O
F

S
O
U
T
H
A
M
P
T
O
N

H
I
G
H
F
I
E
L
D

o
n

1
0
/
2
4
/
1
1
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.