Pirulli, M. & Pastor, M. (2013). Geotechnique 63, No. 10, 887888 [http://dx.doi.org/10.1680/geot.12.D.

007]

DISCUSSION

Numerical study on the entrainment of bed material into rapid landslides

M . P I RU L L I a n d M . PA S TO R ( 2 0 1 2 ) . G e o t e ch n i q u e 6 2 , N o . 1 1 , 9 5 9 9 7 2 ,
h t t p : / / d x . d o i . o rg / 1 0 . 1 6 8 0 / g e o t . 1 0 . P. 0 7 4

R. M. Iverson, U.S. Geological Survey, Vancouver, WA, USA

A paper recently published in this journal (Pirulli &
Pastor, 2012) uses numerical modelling to study the important problem of entrainment of bed material by landslides.
Unfortunately, some of the basic equations employed in the
study are flawed, because they violate the principle of linear
momentum conservation. Similar errors exist in some other
studies of entrainment, and the errors appear to stem from
confusion about the role of bed-sediment inertia in differing
frames of reference.
Consider, for example, the equation used by the authors to
represent evolution of the depth-averaged x component of
linear momentum in a fixed (Eulerian) reference frame
@(hvx ) @(hv2x ) @(hvx vy ) @gz (h2 =2)

@t
@x
@x
@y

Lagrangian and Eulerian formulations of the entrainment

problem can be established. Using the authors notation, a
valid x-component Lagrangian momentum equation can be
expressed as (cf. McDougall & Hungr, 2005)
h

dvx @gz (h2 =2) 1

@z
T x gx h  vx

@x
r
@t
dt

(22)

where d=dt @=@t vx (@=@x) vy (@=@y) is a Lagrangian

total time derivative. Combining equation (22) with the product of vx and the standard, depth-averaged, mass-conservation
equation used by the authors (i.e. vx @h=@t @(vx h)=@x
@(vy h)=@y  @z=@t 0), yields an Eulerian momentum equation
@(hvx ) @(hv2x ) @(hvx vy ) @gz (h2 =2)

@t
@x
@x
@y

(2)

1
@z
T x gx h  vx
r
@t

(23)

1
T x gx h
r

where all notation is the same as that of the authors. (Most

relevant here, vx is the depth-averaged x component of the
landslide velocity, @z=@t is the local rate of bed-sediment
entrainment normal to the xy plane, r is the landslide bulk
density and h is its thickness.) Although one might quibble
with some subtle aspects of equation (2), the equations
fundamental flaw appears in the term vx (@z=@t) on the
right-hand side. The term should not be present, because it
duplicates inertial effects accounted for on the left-hand side
of the equation. Elsewhere the present author has demonstrated this fact as part of an entrainment analysis that
accounts for the effects of variations of landslide velocity
with depth (Iverson, 2012). Here, focus is placed on the role
of vx (@z=@t) in the authors model, in which vx has no
depth dependence.
The authors reason that vx (@z=@t) is necessary in equation (2) to represent the effects of momentum transfer that
occurs as a landslide exerts force to accelerate static bed
material, thereby enabling that material to reach the landslides velocity, vx : This reasoning would be correct for a
Lagrangian model that uses a reference frame moving with
velocity vx to describe the behaviour of an element of
landslide material with fixed mass per unit area rh: If the
landslide element applies force to entrain static bed material
at a mass rate r(@z=@t) per unit area, then the inertia of the
bed material produces an equal and opposite reaction force
on the landslide: rvx (@z=@t) per unit area. This external
force decreases the acceleration of the landslide material
element, but it has no effect on the momentum of the
landslide as a whole, because the rate of landslide velocity
change is precisely offset by the rate of landslide mass gain.
Likewise, in the authors Eulerian model, the inertia of static
bed material should have no effect on landslide momentum.
In an accurate Eulerian formulation, the inertia of static bed
material is accounted for implicitly, without considering any
external forces.
A clear mathematical relationship between depth-averaged

in which vx (@z=@t) has been cancelled from each side.

Mathematically, the effects of vx (@z=@t) on the right-hand
side of equation (22) are absorbed into terms on the lefthand side of equation (23). Physically, the effects of
vx (@z=@t) do not appear in equation (23) because the lefthand side of (23) accounts fully for the momentum of the
landslide and bed material undergoing entrainment. Any
modification of equation (23) to account for entrainment
effects must entail modification of the resisting basal shear
traction, T x  not modification of momentum per unit area
represented by rhvx :
The discussion above can be recapitulated by stating a
unifying principle: the inertia of entrained bed material can
affect landslide velocity, but it cannot affect the net momentum of the two-body system comprising the landslide and its
bed, because static bed material has no momentum to
contribute. This principle applies irrespective of the frame of
reference.
Momentum conservation in a two-body system consisting
of a landslide and its bed also has implications for formulating valid entrainment-rate equations. For example, if the
basal landslide velocity is vx and the bed velocity is zero,
then momentum conservation demands that the entrainment
rate must satisfy @z=@t T jump =rvx , where T jump represents
the difference between the basal shear traction applied by
the moving landslide and the basal shear resistance exerted
by the static bed (Iverson, 2012). Thus, entrainment can
occur only if the bed is weak enough that T jump . 0:
Development of a weakened bed (which might result from
undrained loading, for example) in turn implies that the
resisting traction T x decreases in equations (22) and (23).
Only such a reduction of T x can enhance landslide momentum in conjunction with entrainment (Iverson et al., 2011).
The equation @z=@t T jump =rvx further implies that entrainment rates must decrease as basal landslide velocities increase  assuming that all other factors, such as slope
887

Downloaded by [ Universidad de Castilla-La Mancha] on [15/10/15]. Copyright ICE Publishing, all rights reserved.

DISCUSSION

888

angle, remain constant. This finding contradicts assumptions

made in many entrainment-rate formulas, including that
proposed by McDougall & Hungr (2005) and adopted by the
authors.
Authors reply
The authors agree, there has been a mistake while writing
the equations in the paper. Three alternative formulations
can be considered: one Lagrangian and two Eulerian (primitive variables and conservative form). The erosion term does
not appear in the third. The equations used in the numerical
model actually are
@h @ hvx @ hvy

Et

@t
@y
@x
hv2x

@ hvx @

@t
@x

@ hvx vy

@y

1
T x gx h
r

(24)

@

gz h2
2
@x

(25)

@ hvy @ hvx vy

@t
@x

 
@ hv2y
@y



gz h2
2

@y
@

(26)

1
T y gy h
r
REFERENCES
Iverson, R. M. (2012). Elementary theory of bed-sediment entrainment by debris flows and avalanches. J. Geophys. Res. 117,
F03006, http://dx.doi.org/10.1029/2011JF002189.
Iverson, R. M., Reid, M. E., Logan, M., LaHusen, R. G., Godt, J.
W. & Griswold, J. G. (2011). Positive feedback and momentum
growth during debris-flow entrainment of wet bed sediment.
Nature Geosci. 4, No. 2, 116121, http://dx.doi.org/10.1038/
NGEO1040.
McDougall, S. & Hungr, O. (2005). Dynamic modelling of entrainment in rapid landslides. Can. Geotech. J. 42, No. 5, 1437
1448, http://dx.doi.org/10.1139/T05-064.
Pirulli, M. & Pastor, M. (2012). Numerical study on the entrainment of bed material into rapid landslides. Geotechnique 62,
No. 11, 959972, http://dx.doi.org/10.1680/geot.10.P.074.

Downloaded by [ Universidad de Castilla-La Mancha] on [15/10/15]. Copyright ICE Publishing, all rights reserved.