Computers and Geotechnics 129 (2021) 103876

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Two-phase dynamic MPM formulation for unsaturated soil

Francesca Ceccato a, *, Alba Yerro b, Veronica Girardi a, Paolo Simonini a
a
University of Padua, Department of Civil, Environmental and Architectural Engineering, Via Ognissanti 39, Padua, Italy
b
Virginia Polytechnic Institute and State University, Department of Civil and Environmental Engineering, 750 Drillfield Drive, 200 Patton Hall, Blacksburg, VA, USA

A R T I C L E I N F O A B S T R A C T

Keywords: In this paper, a computational framework based on the Material Point Method is developed to study the coupled
MPM seepage-deformation process in unsaturated soils. Governing equations are derived from the balance equations of
Unsaturated soil solid and liquid phases accounting for partial saturation effects, and the soil is discretized with a single set of
Large deformations
material points which move according to the displacement of the solid phase. The implementation of different
Dynamic coupled analysis
types of boundary conditions, such as transient hydraulic head, seepage face, and infiltration/evaporation, is
Infiltration
Seepage discussed in detail. The method is applied to simulate levee collapse due to rapid drawdown and rainfall. The
methodology proposed in this paper is a promising tool to advance the current practice, which customarily
evaluates slope safety only by means of small deformation analyses based on FEM or LEM, while here a large
deformation approach is presented.

1. Introduction The results of these analyses allow to evaluate FS at different time

instants accounting for stress changes induced by various types of
Many natural hazards involve large deformations of unsaturated boundary conditions and the corresponding most likely failure mecha­
soils, e.g. rainfall-induced landslides, embankment collapses due to nisms, but the collapse evolution cannot be assessed. The knowledge of
wetting, seepage-induced instabilities of dams and levees, etc. The study failure evolution is important for risk management and planning of
of these phenomena requires to account for the complex hydrome­ attenuation measures. Hence there is a need to develop solution schemes
chanical interactions between solid skeleton and pore fluid and to model that are capable of simulating failure initiation as well as post-failure
large deformations in order to predict the post-failure behaviour, which dynamics. To this end, a numerical method suitable for large deforma­
poses significant computational challenges. In common practice, the tion problems is necessary. Several methods have been proposed in the
safety of these slopes is evaluated calculating a factor of safety (FS) with last decades to overcome the difficulties arising with FEM, e.g. ALE, CEL,
Finite Element Methods (FEMs) implementing hydromechanically PFEM, FEMLIP, SPH, MPM etc. A review of these methods applied to
coupled formulations or using Limit Equilibrium Methods (LEMs) landslide mass movements can be found in Soga et al. (2016). Among
(Griffiths and Lane, 1999; Duncan, 1996). Standard Lagrangian FEMs them, the Material Point Method (MPM) has recently increased its
can be used to derive the deformations of the slope under different popularity in the geotechnical community and it has been successfully
boundary conditions, but the solution is limited to small displacements. applied in a number of problems involving multiphase interactions in
FS of a slope can be obtained with the shear strength reduction soils (Yerro et al., 2015; Bandara and Soga, 2015; Ceccato and Simonini,
method in which the strength parameters are progressively reduced 2016; Bandara et al., 2016; Ceccato et al., 2019b; Pinyol et al., 2017;
until the convergence to static equilibrium is impossible to reach Wang et al., 2018; Lei et al., 2020).
(Griffiths and Lane, 1999). With this approach, the failure surface shape MPM was initially formulated for one-phase materials in solid me­
results from the analysis. Differently, LEMs assume the sliding surface chanics by Sulsky et al. (1994) and a number of numerical improve­
shape and compute FS as the ratio between the soil shear resistance and ments have been developed since then. In MPM, the body is discretized
the mobilized shear stress along the surface. The pore pressure and by a cloud of material points (MPs), which move attached to the material
effective stress distribution can be evaluated with preliminary FEM and carry all the updated information such as velocities, strain, stresses,
analyses. Since soil is assumed rigid-plastic, LEMs do not provide any and state variables. Large deformations are simulated by MPs moving
information on the deformations. through a computational grid that covers the full problem domain. This

* Corresponding author.
E-mail address: [email protected] (F. Ceccato).

https://doi.org/10.1016/j.compgeo.2020.103876
Received 3 July 2020; Received in revised form 7 September 2020; Accepted 7 October 2020
Available online 3 November 2020
0266-352X/© 2020 Elsevier Ltd. All rights reserved.
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

grid is used to solve the system of equilibrium equations, but does not means that ponding conditions cannot be simulated. Contrarily, the
deform with the body like in Lagrangian FEM. infiltration BC proposed here evaluates the infiltration discharge at the
A number of mathematical formulations have been proposed in the boundary nodes to switch between natural and essential boundary
literature for unsaturated soils based on different assumptions and using condition. With this approach, ponding conditions can eventually be
different primary unknown variables. Zienkiewicz et al. (1999) present accounted for.
several formulations for saturated and unsaturated soils referring to FEM The paper is organized as follows. Section 2 presents the details of
implementations and discuss their applicability for different kinds of the governing equations, followed by Section 3 where the MPM
problems in geomechanical engineering. These formulations can also be formulation is derived. The implementation is validated in Section 4 by
extended for MPM. comparing MPM results with FEM. Finally, the applicability to real cases
Recent developments of MPM for unsaturated soils include the work is shown in Section 5 by considering two common causes of levee
by Yerro et al. (2015), which proposed a three-phase approach in which collapse: (i) rapid drawdown and (ii) seepage combined with heavy
the governing equations are derived from the momentum balance and rainfall.
the mass balance equations of solid, liquid and gas phase assuming non- The proposed methodology allows to investigate failure and post-
zero gas pressure. In this approach the primary unknowns are the ab­ failure evolution of slopes in a unified numerical framework, thus of­
solute accelerations of the phases (aS − aL − aG formulation). The fering a more comprehensive understanding of the phenomenon
formulation is Lagrangian for the solid phase; MPs move with the ki­ compared to the current practice based on LEM or FEM. The study is also
nematic of the solid skeleton, but carry the information of all phases relevant for authorities and practitioner involved in the field of geo-
(single-point approach). In contrast, Bandara et al. (2016), Wang et al. environmental risk assessment as the methodology can improve the
(2018) and Lei et al. (2020) used a simplified approach, which neglects design of efficient and cost-effective mitigation measures.
the momentum and the mass balance equations of the gas, thus reducing
the computational cost. Bandara et al. (2016) used a formulation 2. Governing equations
derived from Biot’s phenomenological approach in which the solid
skeleton is represented using material coordinates and fluid motion is Unsaturated porous media are mixtures of three different phases:
represented in terms of average relative velocity with respect to the solid solid (S), liquid (L) and gas (G). The solid phase is made by solid grains
skeleton (seepage velocity, w). Relative fluid acceleration and its that constitute the solid skeleton of the media; meanwhile the liquid and
convective terms are assumed negligible. Acceleration of solid skeleton the gas phases fill the voids. A comprehensive hydromecanically
and seepage velocity are the primary unknowns (aS − w formulation). In coupled formulation for unsaturated soils should consider the balance
the formulation proposed in Wang et al. (2018) the governing equations equations of each phase and the interaction between them. However, in
are derived from the dynamic equilibrium of the liquid phase and the many cases the presence of the gas phase can be accounted in a
mixture and the primary unknowns are the absolute accelerations of the simplified way, thus reducing the computational cost.
solid and the liquid (aS − aL formulation). The spatial gradient of liquid In this paper, the following assumptions are adopted:
volumetric fraction is neglected in the liquid mass balance equation and
this is acceptable when considering problems where the degree of 1. Gas pressure is neglected (pG = 0)
saturation varies in a limited range and the slope of the soil–water 2. Gas density is negligible compared to liquid and solid ones (ρG ≈ 0)
retention curve (SWRC) is small (Ceccato et al., 2019a). Lei et al. (2020) 3. Compressibility of solid grains is negligible compared to solid skel­
extended this approach to include soil erosion. Governing equations for eton one (ρS = const)
describing skeleton deformation, liquid flow, liquefaction of solid fine 4. Isothermal conditions
particles, and convection of liquidised fine particles are formulated for
the explicit dynamic MPM framework, and discretized by the General­ The formulation is derived considering dynamic momentum balance
ized Interpolation Material Point (GIMP) method. of the liquid phase, dynamic momentum balance of the mixture, and
This paper presents a two-phase dynamic MPM formulation for non- mass balances and constitutive relationships of both phases involved (i.
erodible unsaturated soils based on the aS − aL formulation. Governing e. solid and liquid). All dynamic terms are taken into account; acceler­
equations are derived from the balance equations of solid and liquid ation of the solid skeleton aS and acceleration of the pore liquid aL are
phases and the soil is discretized with a single set of MPs which move the primary unknowns. Unsaturated conditions are accounted consid­
according to the displacement of the solid phase, i.e. a two-phase single- ering that the liquid phase does not fill entirely the voids. The balance
point approach is used. Unlike previous works, here liquid acceleration equations of gas phase are neglected, thus the presented formulation can
and gradients of liquid volumetric fraction are accounted, which renders be considered two-phase and is an extension of the one presented in
the formulation more suitable for dynamic problems where wave Jassim et al. (2013) and Wang et al. (2018). Moreover, it can be
propagation is important (van Esch et al., 2009; Chmelnizkij et al., considered as a simplified version of the three-phase formulation by
2019) and cases with large changes in suctions and degree of saturation Yerro et al. (2015) as discussed in Ceccato et al. (2019a), with a lower
Ceccato et al. (2019a). computational cost.
The implementation in the MPM framework of foundamental The momentum balance of the liquid phase per unit of liquid volume
boundary conditions, such as transient hydraulic head, seepage face, is given in Eq. (1),
and infiltration/evaporation, is discussed in detail. These conditions are
ρL aL = ∇pL − f dL + ρL g (1)
necessary to simulate typical geomechanical problems with unsaturated
soils, but they have not been presented before. Indeed Yerro et al.
where ρL is the liquid density, f dL is the drag force which accounts for
(2015), Wang et al. (2018), Lei et al. (2020) simulated rainfall infiltra­
tion in a simplified way, considering full soil saturation at the boundary solid–fluid interaction, pL is the liquid pressure and g is the gravity
and did not present other types of hydraulic boundary conditions. vector.
Infiltration boundary condition and seepage face are introduced in The flow is considered laminar and stationary in the slow velocity
Bandara et al. (2016) in the framework of aS − w formulation. In this regime. Hence, the drag force is governed by Darcy’s law (Eq. (2)),
approach, the switch between natural (prescribed pressure) and essen­ n L μL
f Ld = (vL − vS ) (2)
tial (prescribed velocity) boundary condition is based on the soil con­ κL
dition; if the soil is saturated, no inflow is allowed and the pressure of
MPs next to the surface is set to zero, otherwise, if the soil is unsaturated, where μL is the dynamic viscosity of the liquid, κL is the liquid intrinsic
inflow is allowed and the prescribed seepage velocity is imposed. This permeability, nL is the liquid volumetric fraction and vL , vS are the

2
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

absolute liquid and solid velocities respectively. ( ) ( ) ( )

The isotropic intrinsic permeability of the liquid κL can also be ∂ρ ∂SL DS pL
n SL L + ρL = ρL nSL div vS − vL − ρL SL div vS (13)
expressed in terms of Darcy permeability, or hydraulic conductivity, kL ∂pL ∂pL Dt
(Eq. (3)). This simplified form has been adopted by Wang et al. (2018). Yerro
μ et al. (2016), Ceccato et al. (2019a) showed that the use of Eq. (13) gives
κ L = kL L (3)
ρL g reasonably good results for applications where the degree of saturation
varies in a limited range and the derivative of the SWRC is small. In this
The mixture dynamic momentum conservation can be written as Eq. paper, the more general Eq. (12) is used to update liquid pressure.
(4), Hydraulic and mechanical constitutive equations are used to close
nS ρS aS + nL ρL aL = div(σ ) + ρm g (4) the system of governing equations as explained in the following sections.

where ρS is the solid grain density, nS is the volumetric concentration 2.1. Mechanical constitutive equation
ratio of solid, and ρm = nS ρS +nL ρL is the density of the mixture. Note
that nS = 1 − n and nL = SL n, where n is the porosity of the solid skeleton In unsaturated soils, two stress variables can be used to capture the
and SL is the degree of saturation. σ is the total stress tensor, which can soil behaviour, e.g. net stress σ net = σ − pG and suction s = pL − pG . The
be computed with the Bishop’s effective stress equation for unsaturated incremental stress–strain equation becomes Eq. (14),
soils and has the form of Eq. (5), where χ is an effective stress parameter,
DS σ net DS ε S
′D s
here assumed equal to SL , and m is the unit vector, equal to (111000)T in = Dep +h (14)
Dt Dt Dt
3D. In this paper, stresses and pressures are positive for tension, thus
suction is equal to pL . where Dep is the tangent stiffness matrix, h is a constitutive vector. Both
′

are defined by the constitutive model.

(5)
′
σ = σ + χ pL m
A great number of constitutive models for unsaturated soils have
The mass conservation of the solid phase is expressed as given in Eq. been presented in the literature (Gens et al., 2006; Vanapalli et al., 1996;
(6), Loret and Khalili, 2000; Mašín and Khalili, 2008; Sheng et al., 2004;
( ) Francois and Laloui, 2008), but in this paper a basic elastoplatic model
∂(nS ρS )
+ div nS ρS vS = 0 (6) with Mohr–Coulomb failure criterion is used. More advanced constitu­
∂t tive models for unsaturated soils can be easily employed in order to
capture more accurately the soil behaviour.
where vS is the velocity of the solid phase.
Similarly, the conservation of liquid mass can be written as Eq. (7),
2.2. Hydraulic constitutive equation
( )
∂(nL ρL )
+ div nL ρL vL = 0 (7)
∂t The relationship between pore liquid pressure and degree of satu­
ration or liquid content is essential to model the behaviour of unsatu­
where vL is the (true) velocity of the liquid phase rated soil. This is given by the SWRC which can assume different
The material derivative with respect to the solid can be expressed as analytical expressions. In the following, two alternative relationships
S
( ) ( ) are used: (i) a linear relation (Eq. (15)) where av is a constant parameter,
D • ∂( • )
= + vS ∇ • (8) Smin is the residual degree of saturation and Smax is the maximum degree
Dt ∂t
of saturation, and (ii) the Van Genuchten relationship (Eq. (16)) where
When considering incompressible solid grains and disregarding the pref , λ, are fitting parameters.
spatial variations in density and porosity, the expressions for the con­ ( )
SL = Smin + Smax − Smin paLv (15)
servation of mass of the solid and the liquid reduce to Eqs. (9) and (10)
respectively. ⎛ ⎞⎡ ⎤− λ
( ) ( )1−1 λ
DS nS ⎜ ⎟⎢ pL ⎥
+ nS div vS = 0 (9) SL = Smin + ⎜ ⎟⎢ ⎥ (16)
Dt ⎝Smax − Smin ⎠⎣1 + pref ⎦
( ) ( ) ( )
DS (ρL nL )
= vS − vL ∇ nL ρL − nL ρL div vL (10) Partial saturation modifies the soil hydraulic conductivity; in general
Dt
partially saturated soils are less permeable than fully saturated soils. The
Including Eq. (9) into Eq. (10), taking into account the definitions of ratio between actual hydraulic conductivity and saturated hydraulic
liquid volumetric concentration ratios in terms of porosity and degree of conductivity k/ksat is a function of the degree of saturation as given by
saturation, nL = SL n, and rearranging terms give Eq. (11) the hydraulic conductivity curve (HCC). A number of relationships has
( ) been proposed in the literature, the functions proposed by Hillel (1971)
DS (ρL SL )
n = div[ρL nSL (vS − vL )] − ρL SL div vS (11) (Eq. (17)) and Mualem (1976) (Eq. (18)) are implemented in the applied
Dt
software, in which r and λ are fitting parameters.
Finally, the material derivative in Eq. (11) is solved assuming liquid k
pressure as state variable, which yields to Eq. (12). = SrL (17)
ksat
( ) ( )
∂ρ ∂SL DS pL
n SL L + ρL = div[ρL nSL (vS − vL )] − ρL SL div vS (12) ⎡ ⎛ ⎞λ ⎤2
∂pL ∂pL Dt
k √̅̅̅̅̅⎢ ⎜ 1
⎟ ⎥
= SL ⎣1 − ⎝1 − SLλ ⎠ ⎦ (18)
The derivative of liquid density with respect of pressure is given by ksat
the state equation of liquid; while the derivative of the degree of satu­
ration is given by the soil–water retention curve (SWRC).
When neglecting the spatial variability of porosity and degree of
saturation, i.e ∇(ρL nSL ) ≈ 0, Eq. (12) can be simplified in Eq. (13)

3
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

3. MPM formulation (Eqs. (9) and (12)) and constitutive equations (Eq. (14)) are posed
locally at the MPs to update secondary variables.
3.1. Discretized equations
3.2. Numerical algorithm
The discretized equations (Eqs. (19), (20)) are obtained by deriving
the weak form of the momentum balance equations (Eqs. (1) and (4)), The formulation presented in previous section is implemented in the
using the Galerkin procedure, and approximating the acceleration and open source code Anura3D (www.anura3d.eu). The numerical algo­
velocity fields by means of finite element shape functions N. rithm is based on the modified lagrangian algorithm originally proposed
( ) by Sulsky et al. (1995) for one-phase material, and successfully extended
M̃ L aL = f ext − f int − QL vL − vS
L L (19) to multiphase materials (see e.g. Jassim et al., 2013; Yerro et al., 2015).
The MPM solution scheme for each time step can be summarised as
MS aS + ML aL = f ext − f int (20) follows.

aL ,aS ,vL , and vS are the nodal acceleration and velocity vectors; MS ,ML , 1. Liquid momentum balance equation (Eq. (19)) is assembled and
and M ̃ L are solid and liquid lumped mass matrices (Eqs. (21)–(23)); f ext ,
L solved for the liquid nodal acceleration aL .
f Lint , f ext , f int are internal and external nodal force vectors of the liquid 2. Mixture momentum balance equation (Eq. (20)) is assembled and
phase and the mixture (Eqs. (24)–(27)), and QL is the drag force matrix solved to obtain the nodal acceleration of the solid aS .
(Eq. (28)). 3. Velocities and momentum of the MPs are updated from nodal ac­
celerations of each phase.
∑
nMP
̃L ≈
M ̃ MP
m (21) 4. Nodal velocities are calculated from nodal momentum and used to
L N
MP=1 compute the strain increment at the MP location and the terms on the
right-end-side of Eq. (12).
∑ 5. Mass balance equation (Eq. (12)) and soil stress–strain equation (Eq.
nMP
ML ≈ mMP
L N (22)
MP=1 (14)) give the increment of pore pressure and effective stress
respectively.
∑
nMP
6. State variables at MPs are updated. Degree of saturation and hy­
MS ≈ mMP
S N (23) draulic conductivity are updated according to SWRC and HCC
MP=1
respectively.
∫ ∑
nNP 7. Displacement and position of each MP are updated according to the
f Lext ≈ NT p
̂ L d∂Ωp + mMP T
L N g (24) updated velocity of the solid phase.
∂Ωp
8. Nodal values are discarded, the MPs carry all the updated informa­
MP=1

∫ ∑
nMP tion, and the computational grid is initialised for the next time step.
f ext ≈ NT τ d∂Ωτ + mMP T
m N g (25)
∂Ωτ MP=1 Steps 1 and 2 are commonly referred to as Lagrangian phase, while
steps 3 to 8 are called convective or Eulerian phase of the method to
∑
nMP
f Lint ≈ BT pMP
L mVMP (26) emphasize the fact that MPM combines the advantages of Lagrangian
MP=1 and Eulerian approaches.

∑
nMP
f int ≈ BT σ MP VMP (27) 3.3. Treatment of boundary conditions
MP=1

The proposed formulation requires the definition of the following

∑ boundary conditions (BC):
nMP
nMP g ρ
QL ≈ NT L MPL NVMP (28)
MP=1 kL
1. prescribed liquid displacement or velocity on ∂ΩvL and prescribed
where τ is the prescribed traction vector in the boundary ∂Ωτ , p̂ L is the pressure on ∂Ωp , where ∂Ω = ∂ΩvL ∪ ∂Ωp and ∂ΩvL ∩ ∂Ωp = 0
prescribed liquid pressure in the boundary ∂Ωp , N is the matrix of nodal 2. prescribed solid displacement or velocity on ∂Ωu and prescribed
shape functions and B is the matrix of the gradients of the nodal shape traction on ∂Ωτ , where ∂Ω = ∂ΩvS ∪ ∂Ωτ and ∂ΩvS ∩ ∂Ωτ = 0.
functions evaluated at local MP positions. The treatment of boundary
conditions for unsaturated soils is described in details in Section 3.3. Essential boundary conditions on ∂ΩvL and ∂ΩvS are imposed on the
The phase mass of the MP is calculated as Eqs. (29)–(32), where VMP nodes of the computational grid. Natural boundary conditions on ∂Ωp
is the volume of the MP. and ∂Ωτ are included in the weak form of the momentum balance
equations.
̃ MP
m L = ρL VMP (29)
In typical problems with partially saturated soil in geomechanics,
prescribed liquid velocity can be applied on infiltration boundaries as
mMP MP
L = nL ρL VMP (30)
described in Section 3.3.2, and prescribed pressures can be applied
either by defining a pressure load ̂
p L or by assigning a total hydraulic
mMP MP
S = nS ρS VMP (31)
head H ̂ as explained in Section 3.3.1. Assuming the validity of Ber­
mMP (32) noulli’s equation and neglecting the kinematic head, the total hydraulic
m = ρm VMP
head can be written as
Eqs. (19) and (20) are integrated in time using the Euler-Cromer
explicit method. This time integration scheme is conditionally stable pL
̂ = hg − ̂
H (33)
and the critical time step size is computed as indicated in Mieremet et al. ρL g
(2015) and later extended in Mieremet (2015).
The momentum balances (Eqs. (19) and (20)) are discretised and where hg stands for the potential head or geometric head and ̂
p L /(ρL g) is
solved at the nodes of the mesh as in Jassim et al. (2013). Mass balances the pressure head. The minus sign in Eq. (33) is introduced because
pressure is assumed negative for compression.

4
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Fig. 1. Typical boundary conditions for liquid phase (a) and mixture (b).

Fig. 3. Flow chart of the infiltration boundary condition.

e. element without MPs) are detected (boundary elements); then the

nodes belonging to the element side adjacent to an empty element
(boundary side) are marked as boundary nodes. The MPs next to the
boundary side are identified as boundary material points as shown in
Fig. 2. Finally, if the boundary node lies inside the area where a specified
condition has to be applied, e.g. in the infiltration zone or on the po­
tential seepage face, the corresponding boundary condition is applied.
Another difficulty with moving boundaries appears when prescribed
velocity or traction (pressure) do not have a constant direction, but it is
normal to the boundary. This means that if the shape of the contour
changes, the direction of the applied condition has to be updated during
the calculation. The normal direction at the node is determined by
Fig. 2. Definition of boundary nodes and boundary material points. means of the gradient of mass as Eq. (34)
n∑
MP

Sometimes, the boundary condition is part of the problem solution, i. mMP

m B
e. it is not known a priori if the boundary belongs to ∂ΩvL or ∂Ωp . This is n≈ ⃦ MP=1
⃦ n∑
MP
⃦
⃦ (34)
typical of free surface flows across porous media along the so-called ⃦
⃦ mMP ⃦
m B⃦
potential seepage face (Scudeler et al., 2017; Hu et al., 2016). More­
MP=1

over, the size of this boundary condition can evolve along time. This
3.3.1. Hydraulic head BC
boundary condition can be applied as presented in Section 3.3.3. Fig. 1
Examples in which hydraulic head BC can be used are the seasonal
represents schematically how these BCs simulate different hydraulic
fluctuations of groundwater tables (Hong and Wan, 2011), impounde­
loading acting on a levee.
ment or drawdown of a water reservoir (Hendron and Patton, 1985), the
In classical FEM, the application of prescribed boundary conditions is
sudden increase of river levels due to an extreme flood event, the rupture
simple as these can be specified directly on the boundary nodes, which
of buried water pipes altering the surrounding water table distribution
coincide with the boundary of the continuum body and are well defined
in a urban environment, or the effect of introducing a drainage system to
throughout the computation. However, the computational mesh in MPM
stabilize a slope.
does not necessarily align with the boundary of the material making the
application of the prescribed boundary conditions more challenging. In
the numerical framework proposed here, zero-traction and zero- Table 1
pressure boundary conditions are automatically enforced to be satis­ Material parameters for 1D infiltration example and 2D seepage flow.
fied by the solution of equations of motion, but difficulties arise when
Solid density [kg/m3 ] ρS 2700
dealing with non-zero boundary conditions. The nodes belonging to the
Liquid density [kg/m3 ] 1000
boundary are determined with a procedure similar to the surface
ρL
Porosity [–] n 0.4
boundary algorithm presented in Bandara and Soga (2015), here
Liquid bulk modulus [kPa] KL 80000
extended for unstructured mesh. For each time step, firstly the active
Liquid dynamic viscosity [kPa⋅s] μL 1⋅10− 6
elements (i.e. elements containing MPs) adjacent to an empty element (i.

5
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Fig. 4. Geometry of the problem (a). Evolution of liquid pressure (b) and degree of saturation (c) along the column. Comparison between MPM and FEM.

3.3.2. Infiltration BC
This boundary condition is necessary to simulate rainfall or evapo­
ration boundary condition. It consists in applying a prescribed specific
discharge q = w ̂ ⋅n along the boundary, where n is the outward normal
unit vector (Eq. (34)) and the seepage velocity ŵ is defined as Eq. (35)
̂ = nL (vL − vS )
w (35)
The application of this boundary condition is based on a pre­
dictor–corrector scheme: liquid and solid velocity are first predicted
Fig. 5. Geometry and mesh of the MPM model. Positions of sections S1 and S2. assuming zero-pressure boundary conditions at the infiltration bound­
ary and then (eventually) corrected to ensure the prescribed infiltration
The hydraulic head BC is converted to an imposed pressure condition rate. The procedure can be summarized in the following steps, also
pL , using Eq. (33) solved for the liquid pressure. This means that a hy­
̂ shown in Fig. 3:
drostatic approximation is adopted, neglecting local turbulence induced
by the river flow or by the potential collapse of the internal bank. The 1. atL and atS are computed by solving Eqs. (19) and (20) assuming ̂ pL =
element’s nodes affected by this condition are detected with the algo­ 0 at the infiltration boundary
rithm introduced in the previous section (Fig. 2). 2. Nodal velocities are predicted as ̃ vt+Δt = vtL +aL Δt and ̃vt+Δt = vtS +
L S
A series of hydraulic head values in time, like the ones provided by aS Δt
water gauges readings, is assigned with an input file, and at each time 3. Infiltration condition is checked. If the net infiltration discharge qnet
step the current hydraulic head Hcurr is computed by linear interpolation. (Eq. (36)) is positive, ponding conditions occur and if fluid accu­
The resulting nodal vector is part of the vector of external forces in the mulation above the boundary is not allowed (it must remain at zero
liquid momentum equation (f ext L in Eq. (19)) and in the mixture mo­ pressure) no correction is necessary. If the net infiltration discharge
mentum equation (f ext in Eq. (20)) to account also for water weight. is negative, or liquid ponding is allowed above the surface, then
Concerning the boundary nodes included in the hydraulic head BC liquid velocity must be corrected to ensure the correct infiltration
but lying above Hcurr , two approaches are proposed: (i) constant pressure rate.
and (ii) linear pressure distribution (in suction), following the same ( ( t+Δt ) )
hydrostatic gradient given by γW . In the case of a homogeneous slope in qnet = nL ̃vL − ̃ vt+Δt
S − w
̂ ⋅n (36)
stationary flow regimes, it is common to introduce the unsaturated
condition above the water table (Lu et al., 2004; Griffiths and Lu, 2005),
which reflects in the pore pressure distribution. Close to the phreatic 4. If necessary, the liquid nodal velocity is corrected by Eq. (37)
surface, a linear increase of suction can be explained by the capillary rise vt+Δt vt+Δt
=̃ + ΔvL (37)
controlled by the SWRC. The upper portion, closer to the interface with
L L

the atmosphere, is governed by climate conditions and suction depends where ΔvL is derived by imposing qnet = 0 (Eq. (38)) and it is given
on the water mass balance at the soil surface (Griffiths and Lu, 2005; by Eq. (39).
Rahardjo et al., 2013). The constant approximation reproduces this last ( ( t+Δt ) )
effect, while the linear approximation simulates better the capillary ef­ vL + ΔvL − ̃
nL ̃ vt+Δt
S − w
̂ ⋅n = 0 (38)
fect. The drawback of the later approach is the overestimation of suction
(( ( t+Δt ) ) )
at higher locations. In the literature, the BC above the river level has ΔvL = vL − ̃
nL ̃ vt+Δt
S − w
̂ ⋅n n (39)
been reproduced in different ways. In some cases, an infiltration/evap­
oration BC is applied (Griffiths and Lu, 2005; Rahardjo et al., 2013) or a
zero nodal flux is imposed (Geoslope, 2007), in other cases the condition 5. The corrected liquid acceleration is computed as atL = (vt+Δt
L − vtL )/Δt
has been approximated with the use of a potential seepage face (Scud­ 6. The MPM solution scheme can proceed with the convective phase as
eler et al., 2017; Callari and Abati, 2009). explained in steps 3–8 of Section 3.2.

6
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Fig. 6. Evolution of liquid pressure in the levee body. Comparison between MPM (left) and FEM (right).

Note that this boundary condition is applied at nodal level, thus must be corrected with Eqs. (37) and (39) in which w
̂ = 0 (switch to
nodal liquid volumetric fraction is necessary. This can be computed by essential boundary condition).
mapping nL = nSL from the MPs to the nodes of the mesh.
4. Validation
3.3.3. Potential seepage face BC
A potential seepage face can be defined as an interface between soil Validating the presented formulation is not straightforward because
and atmosphere where the fluid is free to exit at zero pressure when the no analytical solution is available for coupled soil deformation-fluid
soil is saturated, but it cannot enter when the soil is partially saturated. flow in unsaturated conditions. For this reason, the fluid-flow problem
An example of potential seepage face is the downstream surface of a is validated in this section considering two simple benchmarks: (i) a 1D
river embankment (Fig. 1a), or it can also arise in after a rapid draw­ column infiltration (Section 4.1 and (ii) a 2D seepage flow (Section 4.2),
down of the river level. This boundary condition is necessary where it is by comparing the results obtained by MPM and the well-established
unknown if the boundary is an essential or a natural boundary condition. FEM software SEEP/W by Geostudio. In FEM seepage analyses the
The implementation can be considered as a special case of the deformation of the solid skeleton is neglected, analogously solid velocity
infiltration boundary condition described in Section 3.3.2, where w ̂ = is assumed to be zero in MPM.
0. Liquid and solid nodal velocities are predicted assuming zero-
pressure at the potential seepage face (natural boundary condition). If 4.1. 1D infiltration
vt+Δt
(nL (̃L vt+Δt
− ̃S ))⋅n > 0 it means that fluid is flowing out of the soil at
vt+Δt
zero pressure and no correction is required. If (nL (̃ vt+Δt
− ̃ ))⋅n < 0, A 1 m-high soil column is considered with the material parameters
L S
then fluid is flowing into the soil at zero pressure, thus liquid velocity listed in Table 1. For simplicity, the intrinsic permeability is assumed
constant and equal to κL = 1⋅10− 11 m2 /s, corresponding to an hydraulic

7
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Table 2
Material parameters for levee collapse example.
Solid density [kg/m3 ] ρS 2700
Liquid density [kg/m3 ] ρL 1000
Porosity [–] n 0.4
Liquid bulk modulus [kPa] KL 80000
Liquid dynamic viscosity [kPa⋅s] μL 1⋅10− 6

2 − 11
Intrinsic permeability [m ] κL 4⋅10
Young modulus [kPa] E 50000
Poisson ratio [–] ν 0.30
Friction angle [◦ ] ϕ 27
Cohesion [kPa] c
′
2

saturation SL0 = 0.87.

The liquid infiltrates from the top and flows down though the column
accumulating at the impervious bottom. Here suction starts decreasing
and the soil saturates increasing SL to 1. Fig. 4 plots the evolution of pore
pressure and degree of saturation distribution along the column at
different time instants. MPM results are in very good agreement with
FEM. As expected, it can be noted that at long time (t = 600 s) soil is
saturated at the top boundary, i.e. pL = 0 kPa, thus the infiltration BC
switches from essential, i.e. applied infiltration velocity, to natural, i.e.
zero-pressure, and the pore pressure does not increase further.

4.2. 2D seepage flow

The hydraulic head BC and the seepage BC are applied in a simple

case of a river levee affected by seepage flow from the river side. A 1 m-
high river embankment is considered, with slopes 2/3 on the river side
and 1/2 on the land side, as reported in Fig. 5. The material parameters
Fig. 7. Pore pressure along vertical sections S1 (a) and S2 (b) for MPM (dots) are summarized in Table 1. A linear SWRC (Eq. (15)), with slope coef­
and FEM (dashed line). ficient av = 4.0⋅10− 4 1/kPa is accounted, while the intrinsic perme­
ability is constant and equals to κL = 4⋅10− 11 m2 /s, corresponding to an
conductivity of k = 1.0⋅10− 4 m/s. Van Genuchten SWRC is accounted hydraulic conductivity of k = 3.92⋅10− 4 m/s.
(Eq. (16)), with parameters pref = 3 kPa,λ = 0.7,Smin = 0.125,Smax = 1. The initial phreatic level is at 0.0 m, corresponding to the ground
The column is discretized with 20 rows of 2 square triangular ele­ level at both levee’s sides, visible in Fig. 6a. Above the initial water
ments filled with 3 MPs each (Fig. 4a). The bottom and lateral bound­ table, the pore pressure is linearly approximated. The seepage face and
̂ y = 1.0⋅10− 4 m/s
aries are impervious, while a vertical infiltration rate w the hydraulic head BCs are applied, respectively on the land and on the
river side. The river level increases linearly during 180s, till its
is applied at the top boundary. An initial suction of 2 kPa is assigned at
maximum value of 0.6 m. Then, it remains constant until the end of the
t = 0 s along the column, which corresponds to an initial degree of
simulated time, equal to 200s. The mesh is unstructured, made of 3-
nodes triangular elements of average size 0.05 m. The FEM model
mesh counts 2934 elements while, due to a wider covered domain, the
MPM model counts 5660 elements; a number of 3 MPs is assigned to the
elements belonging to the earth embankment.
Fig. 6 compares FEM and MPM results in terms of pore pressure
distribution in the levee body at four increasing time instants (0s, 60s,
120s, and 200s). Furthermore, the phreatic line is highlighted with blue
dots in MPM (corresponding to MPs with a pressure between − 0.2 and
0.2 kPa) and dashed blue lines in FEM, visualizing the progression of the
saturated front in both models, according to the water level rise.
The progression along time is also shown in Fig. 7, where the pore
water pressure distribution along two vertical sections (S1 and S2) is
plotted at three different instants: 30s, 90s, and 180s. The fitting of the

Fig. 8. (a) Simplified representation of the levee example. (b) Geometry and Fig. 9. Geometry and boundary condition of the model simulating drawdown
discretization of the MPM model. on a levee.

8
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Fig. 10. Evolution of liquid pore pressure at 0, 3, and 60s considering different pore pressure distributions (hydraulic head BC) above the current river level, related
to (a) constant, (b) linear approximations.

MPM values with FEM shows an overall good agreement, both in terms schematically represented in Fig. 8a: (i) a rapid drawdown of the water
of entity and slope of the plotted curves. In the transient phase of the level on the river side (Section 5.1 and (ii) infiltration due to heavy
seepage flow, we observe a minor discrepancy between the trends, more rainfall on the land side (Section 5.2). MPM results are compared with
evident along section S2, yet never greater than 1 kPa. the ones obtained by FEM seepage analysis coupled with LEM analysis to
Given that, these small gaps may be attributed to underpin diversities evaluate FS.
in the formulations. Indeed, MPM solves the dynamic momentum bal­ The considered levee is 3 m-high with a slope inclination of 2/3. A
ance of the liquid, considering the liquid acceleration as primary un­ low-permeability layer is assumed to lie at a depth of 0.5 m, here
known. The pressure increment is a secondary variable updated solving simulated with an impermeable boundary. The levee is assumed to be
the mass balance equation. On the contrary, the FEM approach solves symmetric, thus the same geometry and discretization is used in both the
directly for the nodal pressure (or hydraulic head), without considering drawdown and the rainfall case. The geometry and mesh discretization
the liquid acceleration terms as in the MPM. In addition, solid skeleton is shown in Fig. 8b. For computational efficiency, only half of the levee
and water compressibility may be the source of some minor differences. section is discretized. The model counts a total of 1669 linear triangular
Nonetheless, the seepage flow is in agreement between the two nu­ elements and 897 nodes. The average element size on the slope is 0.2 m.
merical techniques, both in its spatial and temporal connotations. 3 MPs are placed inside each initially active element. For the seepage
analysis, a mesh of the same average size and type is used in the FEM
5. Application examples model, for a total of 1158 elements. LEM analysis applies Bishop
method.
In this section, the formulation presented in Section 3 is applied to A linear SWRC, with slope coefficient av = 4.0⋅10− 4 1/kPa is used,
study the stability of a levee, with the purpose of showing the potenti­ and the hydraulic conductivity is assumed constant. This permeability is
alities of the method in solving real-scale boundary value problems in higher than commonly found in these structures, and allows to reduce
the realm of unsaturated levee failures. Failure of dams and levees can the simulated time and computational cost. An elastic-perfectly plastic
be caused by several mechanisms, e.g. macro-instability, overtopping, Mohr Coulomb constitutive model is adopted to model the soil behavior,
erosion (internal and external) (Simonini et al., 2014; Deniaud et al., with the material parameters listed in Table 2. Cohesion and friction
2013; Peter, 2014). Here we focus on the macro-instabilities due to angle in unsaturated soils are function of suction (Fredlund et al., 1978;
changes in the pore pressure regime induced by two typical phenomena Jommi et al., 2002; Escario et al., 1989), but this effect is assumed

Fig. 11. Onset of failure: a) Solid displacement norm for MPM and critical slip surface for FEM-LEM b) Concurrent liquid pressure in the same moment.

9
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Fig. 14. Geometry and boundary condition of the model simulating rainfall
infiltration on a levee.

Fig. 12. Horizontal displacement contours at 10s, 20s and 60s. Identification of Fig. 15. Initial pore pressure distribution with (a) FEM and (b) MPM. Safety
MP1, MP2, MP3 along the slip surface. factor with LEM.

negligible in this example. Despite the simplifying assumptions on the As introduced in Section 3.3.1, the present study imposes the pore
hydraulic and mechanical soil behaviour, the model can provide a pressure at the boundary above the river level in two different manners,
realistic representation of the levee response; more advanced constitu­ constant or linear. Two simulations starting from an identical state, are
tive models will be used in further development of this research. used to compare the effects of both approximations when the slope is
subjected to a drawdown (Fig. 10). The linear approximation generates
high suction values in the top portion of the river bank, preventing the
5.1. Levee failure for rapid drawdown slope instability (bottom panels in Fig. 10). In the constant approxima­
tion, a pressure of 5 kPa is applied at the points above water level. This
The initial stress distribution is generated with gravity loading lower value of suction leads to the embankment failure during draw­
assuming a river level at 2.0 m. Then, the applied total head is rapidly down (top panels in Fig. 10). Note that, if suction measurements on river
decreased to the low water level of 0.5 m (Fig. 9).

Fig. 13. Liquid pore pressure (a) and deviatoric strain (b) along time for three selected MPs nearby the slip surface.

10
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

increase till maximum values is reached at 60s. In addition, the graph

shows that from the slope crest towards the toe, the MPs deformation
magnitude coherently increases.

5.2. Levee failure for rainfall infiltration

Enduring high water level on the river side and heavy rainfall can
sometimes concur leading to rapid saturation of the levee and potential
instability. The geometry and boundary condition of the numerical
model are shown in Fig. 14. At the right-end, a liquid pressure linearly
increasing with depth (hydrostatic pressure distribution) is applied to
reproduce in a simplified way a water level of 2.0 m on the river side.
The levee surface is a potential seepage face during the initialization
phase, and then an infiltration rate w ̂ y = 1⋅10− 4 m/s is applied to
simulate a heavy rainfall in the following steps. Horizontal fixities are
applied to the solid phase at the lateral boundaries and the bottom is
fully fixed.
To generate the initial stress distribution, liquid pressure, seepage
face and gravity load are applied assuming an initially high material
cohesion (20 kPa). Then, cohesion is reduced to 2 kPa and the slope is
stable with a stress distribution in equilibrium with the applied initial
loads. The initial pore pressure distribution obtained with MPM and
FEM is in very good agreement (Fig. 15). FS evaluated with Bishop

Fig. 16. Deviatoric strain contour at different time instant.

levee’s bank are collected along time, it’s possible to set the constant
approximated value based on the real initial condition. The constant
approximation is compared with the FEM-LEM results.
In order to emphasise the hydromechanical character of the depicted
process, the onset of failure and the concurrent pore pressure distribu­
tion for MPM and FEM-LEM analyses are presented in Fig. 11. For MPM,
the norm of solid displacement is used to show the development of the
slip surface, while in the LEM the circular surface is automatically
generated, just by imposing ranges of entry and exit, and the relative FS
is obtained. The failure onset is identified in LEM by the FS passage to a
value lower than one, while in MPM it is here identified when the
deviatoric strain assumes a value equal to 0.05 and at the same time
vertical displacements of the cm-order can be detected for the MPs on
the crest. The shape of the slip surface is consistent among the two
models. The occurrence time of the failure onset differs for a few sec­
onds, which can be explained by looking at the pore pressure distribu­
tion above the current river level, as visible in Fig. 11. This difference
can be explained because in the MPM simulation, a constant pressure is
imposed along the boundary portion above the river height, while in
FEM a potential seepage BC is applied instead (Geoslope, 2007).
Only for MPM it is possible to track the river bank collapse pro­
gression along time, likewise the pore pressure dissipation process. The
post-failure behavior is captured in Fig. 12 by plotting the horizontal
solid displacements at three time instants, together with the initial soil
surface elevation reference. The phenomenon is clearly characterized by
large displacements, at the end exceeding 1 m (Fig. 12c); thus the irre­
versible collapse of the internal bank is determined. Three MPs nearby
the slip surface (MP1, MP2, MP3) are tracked during the slope motion in
Fig. 12 and used to analyze the evolution of liquid pressure (Fig. 13a)
and deviatoric strain (Fig. 13b) along time.
In the first 3 s, all the MPs experience a sudden reduction of liquid
pressure as a result of the dynamics induced by the rapid drawdown. It is
followed by a gentle pressure increase till 20s, and a subsequent interval
of pressure oscillations. In general, the pore pressure values reflect the
overall state of slow desaturation of the slope, antithetic to the rapid
kinematic process of slope instability. Concerning this last, the devia­
toric strain curves which represent the progressive increase of MPs un­
Fig. 17. (a-c) Pore pressure contour at different time instant. (d) Pore pressure
dergoing plastification around the slip surface area, follow a continuous along time for four MP.

11
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Different scenarios could be investigated modifying the geometry and

mechanical behavior of the soil. Taking into account these aspects is
important to implement safe and cost effective mitigation measures.

CRediT authorship contribution statement

Francesca Ceccato: Conceptualization, Methodology, Software,

Validation, Writing - original draft, Writing - review & editing. Alba
Yerro: Conceptualization, Methodology, Software, Validation, Writing -
original draft, Writing - review & editing. Veronica Girardi: Concep­
tualization, Methodology, Software, Validation, Writing - original draft,
Writing - review & editing. Paolo Simonini: Supervision, Writing - re­
view & editing, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.

Acknowledgments

Financial support from Fondazione Ing. Aldo Gini and University of

Padua (BIRD181859) is gratefully acknowledged.
Fig. 18. Evolution of displacement with time and final displacement contour.

References
method is higher than 1, and indeed the slope is stable and deformations
are very small. Bandara, S., Soga, K., 2015. Coupling of soil deformation and pore fluid flow using
After stress initialization, rainfall infiltration is applied at the top material point method. Comput. Geotech. 63, 199–214. https://doi.org/10.1016/j.
compgeo.2014.09.009. URL http://linkinghub.elsevier.com/retrieve/pii/S0266352
surface, soil suction decreases and the slope fails. LEM analyses give
X14001839.
FS < 1 for a time t = 15s, which means that the failure surface is fully Bandara, S., Ferrari, A., Laloui, L., 2016. Modelling landslides in unsaturated slopes
developed at this moment. Fig. 16 plots the evolution of shear strain at subjected to rainfall infiltration using material point method. Int. J. Numer. Anal.
Meth. Geomech. 40, 1358–1380.
significant time instants, showing that a circular failure surface develops
Callari, C., Abati, A.D., 2009. Finite element methods for unsaturated porous solids and
rapidly from the bottom of the slope and propagates upward. The shape their application to dam engineering problems. Comput. Struct. 87, 485–501.
of the failure surface is in agreement with LEM; however, in MPM it is a https://doi.org/10.1016/j.compstruc.2008.12.012.
result of the calculation, while in LEM it is an hypothesis of the analysis. Ceccato, F., Simonini, P., 2016. Numerical study of partially drained penetration and
pore pressure dissipation in piezocone test. Acta Geotech. 12, 195–209. https://doi.
A small decrease of suction is sufficient to trigger the failure as shown org/10.1007/s11440-016-0448-6.
in Fig. 17. During soil movement, pressure oscillations are observed. Ceccato, F., Girardi, V., Yerro, A., Simonini, P., 2019a. Evaluation of dynamic explicit
Displacements increase suddenly between 5 and 20s (Fig. 18). At t = 20s MPM formulations for unsaturated soils. In: Oñate, E., Bischoff, M., Owen, D.,
Wriggers, P., Zohdi, T. (Eds.), Particles 2019. Barcellona.
the slope finds a new equilibrium configuration measuring 17 cm of Ceccato, F., Girardi, V., Simonini, P., 2019b. Developing and testing multiphase mpm
displacement of the toe and 25 cm at the crest. At about 100s there is a approaches for the stability of dams and river embankments, in. In: Conference of
further increase of displacement of 2 cm due to enduring rainfall infil­ the Italian Association of Theoretical and Applied Mechanics. Springer,
pp. 2179–2195.
tration. The crest of the levee moves downward approximately 22 cm, Chmelnizkij, A., Ceccato, F., Grabe, J., Simonini, P., 2019. 1D wave propagation in
which means that the levee in its deformed configuration is probably saturated soils: verification of two-phase MPM. In: 2nt International Conference on
still able to fulfill its retaining function. This consideration, impossible the Material Point Method for Modelling Soil-water-structure Interaction.
Deniaud, Y., van Hemert, H., McVicker, J., Bernard, A., Beullac, B., Tourment, R.,
to get with FEM or LEM, can have significant practical consequences for
Igigabel, M., Pohl, R., Sharp, M., Simm, J., et al., 2013. Functions, forms and failure
planning of cost-effective remedial measures. of levees. In: CIRIA.
Duncan, J., 1996. State of the art: limit equilibrium and finite-element analysis of slopes.
J. Geotech. Eng. 577–596.
6. Conclusions
Escario, V., Juca, J., Coppe, M., 1989. Strength and deformation of partly saturated soils.
In: Congrès international de mécanique des sols et des travaux de fondations. vol. 12,
This paper presents a two-phase single point formulation of MPM for pp. 43–46.
unsaturated materials and discusses in detail the application of hy­ Francois, B., Laloui, L., 2008. Acmeg-ts: A constitutive model for unsaturated soils under
non-isothermal conditions. Int. J. Numer. Anal. Meth. Geomech. 32, 1955–1988.
draulic boundary conditions for the simulation of common geotechnical https://doi.org/10.1002/nag.712.
problems such as transient total head, infiltration/evaporation, and Fredlund, D., Morgenstern, N.R., Widger, R., 1978. The shear strength of unsaturated
potential seepage face. The method is validated with a 1D infiltration soils. Can. Gotech. J. 15, 313–321.
Gens, A., Sánchez, M., Sheng, D., 2006. On constitutive modelling of unsaturated soils.
test and a 2D seepage problem and applied to study a levee collapse due Acta Geotech. 1, 137–147. https://doi.org/10.1007/s11440-006-0013-9.
to rapid drawdown and heavy rainfall. Geoslope, 2010. Stability Modeling with SLOPE/ W 2007 Version.
This method allows to investigate the consequences of slope in­ Griffiths, D.V., Lane, P.A., 1999. Slope stability analysis by finite elements. Geotechnique
49, 387–403. https://doi.org/10.1680/geot.1999.49.3.387.
stabilities, thanks to the possibility of simulating large deformations, Griffiths, D.V., Lu, N., 2005. Unsaturated slope stability analysis with steady infiltration
thus overtaking the concept of safety factor customarily estimated by or evaporation using elasto-plastic finite elements. Int. J. Numer. Anal. Meth.
LEMs or FEMs. In contrast to these methods, MPM allows to get a real­ Geomech. 29, 249–267. https://doi.org/10.1002/nag.413.
Hendron, Jr. A., Patton, F.D., 1985. The vaiont slide. A geotechnical analysis based on
istic estimate of soil movements. The levee example shows that,
new geologic observations of the failure surface. Volume 1. Main text, Technical
depending on the boundary conditions, similar values of FS, slightly Report, ARMY ENGINEER WATERWAYS EXPERIMENT STATION VICKSBURG MS
lower than one, can lead to different displacement evolution after failure GEOTECHNICAL LAB.
Hillel, D., 1971. Soil and Water - Physical Principles and Processes. Academic Press,
initiation and thus different consequences in terms of risk assessment.
London (UK).

12
F. Ceccato et al. Computers and Geotechnics 129 (2021) 103876

Hong, Y.-M., Wan, S., 2011. Forecasting groundwater level fluctuations for rainfall- Rahardjo, H., Satyanaga, A., Leong, E.C., 2013. Effects of flux boundary conditions on
induced landslide. Nat. Hazards 57, 167–184. pore-water pressure distribution in slope. Eng. Geol. 165, 133–142. https://doi.org/
Hu, R., Chen, Y., Zhou, C., Liu, H.-H., 2016. A numerical formulation with unified 10.1016/j.enggeo.2012.03.017.
unilateral boundary condition for unsaturated flow problems in porous media. Acta Scudeler, C., Paniconi, C., Pasetto, D., Putti, M., 2017. Examination of the seepage face
Geotech. https://doi.org/10.1007/s11440-016-0475-3. boundary condition in subsurface and coupled surface/subsurface hydrological
Jassim, I., Stolle, D., Vermeer, P., 2013. Two-phase dynamic analysis by material point models. Water Resour. Res. 53, 1799–1819. https://doi.org/10.1111/j.1752-
method. International Journal for Numerical and Analytical Methods in 1688.1969.tb04897.x.
Geomechanicsnumerical and Analytical Methods in Geomechanics 37, 2502–2522. Sheng, D., Sloan, S.W., Gens, A., 2004. A constitutive model for unsaturated soils:
URL http://onlinelibrary.wiley.com/doi/10.1002/nag.2146/full. doi:10.1002/nag. Thermomechanical and computational aspects. Comput. Mech. 33, 453–465.
Jommi, C., Marchi, C., Romero, E., Vaunat, J., 2002. Modeling the shear strength of https://doi.org/10.1007/s00466-003-0545-x.
unsaturated soils. In: Proceedings of the 3rd International Conference on Simonini, P., Cola, S., Bersan, S., 2014. Caratterizzazione geotecnica, meccanismi di
Unsaturated Soils UNSAT 2002, vol. 1, AA Balkema, pp. 245–251. collasso e monitoraggio degli argini fluviali - relazione generale.
Lei, X., He, S., Chen, X., Wong, H., Wu, L., Liu, E., 2020. A generalized interpolation Soga, K., Alonso, E., Yerro, A., Kumar, K., Bandara, S., 2016. Trends in large-deformation
material point method for modelling coupled seepage-erosion-deformation process analysis of landslide mass movements with particular emphasis on the material point
within unsaturated soils. Adv. Water Resour. 141, 103578. https://doi.org/10.1016/ method. Géotechnique 66, 248–273. https://doi.org/10.1680/jgeot.15.LM.005. URL
j.advwatres.2020.103578. http://www.icevirtuallibrary.com/doi/10.1680/jgeot.15.LM.005.
Loret, B., Khalili, N., 2000. A three-phase model for unsaturated soils. Int. J. Numer. Sulsky, D., Chen, Z., Schreyer, H., 1994. A particle method for hystory-dependent
Anal. Meth. Geomech. 24, 893–927. https://doi.org/10.1002/1096-9853(200009) materials. Comput. Methods Appl. Mech. Eng. 118, 179–196.
24:11<893::AID-NAG105>3.0.CO;2-V. Sulsky, D., Zhou, S.-J., Schreyer, H.L., 1995. Application of a particle-in-cell method to
Lu, N., Likos, W., (Firm), K., 2004. Unsaturated Soil Mechanics, Wiley, URL http solid mechanics. Comput. Phys. Commun. 87, 236–252. https://doi.org/10.1016/
s://books.google.com/books?id=Rv1RAAAAMAAJ. 0010-4655(94)00170-7. URL http://linkinghub.elsevier.com/retrieve/pii/
Mašín, D., Khalili, N., 2008. A hypoplastic model for mechanical response of unsaturated 0010465594001707.
soils. Int. J. Numer. Anal. Meth. Geomech. 32, 1903–1926. https://doi.org/10.1002/ Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., Clifton, A.W., 1996. Model for the
nag.714. prediction of shear strength with respect to soil suction. Can. Geotech. J. 33,
Mieremet, M.M.J., 2015. Numerical stability for velocity-based 2-phase formulation for 379–392. https://doi.org/10.1139/t96-060.
geotechnical dynamic analysis. In: Reports of the Delft Institute of Applied van Esch, J., Stolle, D., Bonnier, P., Lsl, O.C., 2009. Consideration of pore pressures in
Mathematics. Delft University of Technology, Delft, The Netherlands. MPM. In: Computer Methods in Mechanics, pp. 1–16.
Mieremet, M.M.J., Stolle, D.F., Ceccato, F., Vuik, C., 2015. Numerical stability for Wang, B., Vardon, P.J., Hicks, M.A., 2018. Rainfall-induced slope collapse with coupled
modelling of dynamic two-phase interaction. Int. J. Numer. Anal. Meth. Geomech. material point method. Eng. Geol. 239, 1–12. https://doi.org/10.1016/j.
40 https://doi.org/10.1002/nag.2483. enggeo.2018.02.007.
Mualem, Y., 1976. Hysteretical models for prediction of the hydraulic conductivity of Yerro, A., Alonso, E.E., Pinyol, N.M., 2015. The material point method for unsaturated
unsaturated porous media. Water Resour. Res. 12, 1248–1254. https://doi.org/ soils. Geotechnique 65, 201–217. https://doi.org/10.1680/geot.14.P.163. URL htt
10.1029/WR012i006p01248. p://www.icevirtuallibrary.com/doi/10.1680/geot.14.P.163.
Peter, P., 2014. Canal and River Levées, ISSN, Elsevier Science, URL https://books. Yerro, A., Alonso, E., Pinyol, N., 2016. Modelling large deformation problems in
google.com/books?id=aQ9BAQAAQBAJ. unsaturated soils. E-UNSAT 9 (2016), 1–6. https://doi.org/10.1051/e3sconf/
Pinyol, N.M., Alvarado, M., Alonso, E.E., Zabala, F., 2017. Thermal effects in landslide 20160908019.
mobility. Géotechnique 68, 1–18. https://doi.org/10.1680/jgeot.17.P.054. URL htt Zienkiewicz, O.C., Chan, A.H.C., Pastor, M., Schrefler, B.A., Shiomi, T., 1999.
p://www.icevirtuallibrary.com/doi/10.1680/jgeot.17.P.054. Computational Geomechanics. Wiley & Sons, New York.

13