Thematic set:

COST Action TU1202 Impact of climate change on

engineered slopes for infrastructure Quarterly Journal of Engineering Geology and Hydrogeology
Published Online First https://doi.org/10.1144/qjegh2016-079

Numerical modelling of slopevegetationatmosphere interaction:

an overview
Gaetano Elia1*, Federica Cotecchia2, Giuseppe Pedone2,3, Jean Vaunat4,
Philip J. Vardon5, Carlos Pereira6, Sarah M. Springman7, Mohamed Rouainia1,
John Van Esch8, Eugeniusz Koda9, Josif Josifovski10, Alessandra Nocilla11,
Amin Askarinejad7,12, Ross Stirling1, Peter Helm1, Piernicola Lollino13 &
Piotr Osinski9,14
1
Newcastle University, Newcastle upon Tyne NE1 7RU, UK
2
Technical University of Bari, via Edoardo Orabona 4, 70125 Bari, Italy
3
Present address: University of Southampton, University Road, Southampton, SO17 1BJ, UK
4
Universitat Politecnica de Catalunya - BarcelonaTech, calle Jordi Girona 1-3, 08034 Barcelona, Spain
5
Delft University of Technology, Stevinweg 1, 2628 CN, Delft, The Netherlands
6
LNEC, Av. do Brasil 101, 1700-066 Lisbon, Portugal
7
ETH Zurich, Rmistrasse 101 8092 Zurich, Switzerland
8
DELTARES, Boussinesqweg 1, 2629 HV, Delft, Netherlands
9
Warsaw University of Life Sciences, Nowoursynowska 166, 02-787 Warsaw, Poland
10
Ss. Cyril and Methodius University Blvd. Goce Delcev 9, 1000 Skopje, Republic of Macedonia
11
University of Brescia, Via Branze 43, 25123 Brescia, Italy
12
Present address: Delft University of Technology, Stevinweg 1, 2628 CN, Delft, The Netherlands
13
IRPICNR, Via Amendola 122i, 70126 Bari, Italy
14
Present address: University of Nottingham, Nottingham, NG7 2RD, UK
F.C., 0000-0001-9846-4193; J.V., 0000-0003-3579-9652; C.P., 0000-0001-8805-8685; A.N., 0000-0003-0665-8826;
R.S., 0000-0002-0069-6621
* Correspondence: [email protected]

Abstract: The behaviour of natural and artificial slopes is controlled by their thermo-hydro-mechanical conditions and by
soilvegetationatmosphere interaction. Porewater pressure changes within a slope related to variable meteorological settings
have been shown to be able to induce soil erosion, shrinkageswelling and cracking, thus leading to an overall decrease of the
available soil strength with depth and, ultimately, to a progressive slope collapse. In terms of numerical modelling, the stability
analysis of partially saturated slopes is a complex problem and a wide range of approaches from simple limit equilibrium
solutions to advanced numerical analyses have been proposed in the literature. The more advanced approaches, although more
rigorous, require input data such as the soil water retention curve and the hydraulic conductivity function, which are difficult to
obtain in some cases. The quantification of the effects of future climate scenarios represents an additional challenge in
forecasting slopeatmosphere interaction processes. This paper presents a review of real and ideal case histories regarding the
numerical analysis of natural and artificial slopes subjected to different types of climatic perturbations. The limits and benefits
of the different numerical approaches adopted are discussed and some general modelling recommendations are addressed.
Received 23 August 2016; revised 20 February 2017; accepted 12 April 2017

The equilibrium conditions within slopes, and hence their level of presence of significant suctions in the soil pores (i.e. above the water
stability, depend on several factors (Terzaghi 1950), which include table), which are reduced upon rainfall infiltration. In any case, in
the boundary conditions at the ground surface. These generate general the slope stability varies with time as a result of climate.
exchanges of water between the topsoils and the atmosphere, Several observations are reported in the literature of damage to
through processes such as rainfall infiltration, water evaporation structures and infrastructure interacting with slope movements
from the soil pores and transpiration through vegetation, which, as a connected to the slopeatmosphere interaction, which hence
whole, are referred to as slopeatmosphere interaction. Changes represents a source of risk for society. Such risk may apply to
with time of the soil porewater pressures, and hence of the soil either man-made slopes or natural slopes. For man-made soil
stressstrain conditions, are consequent to such interaction. In embankments and fills, made of partially saturated compacted soils
particular, the porewater pressure changes generate, on one hand, and location of either road or railway infrastructure, movements due
variations of the mobilized shear strengths and, on the other, to variations in the soil degree of saturation as a result of the soil
variations of the available shear strengths, and, as such, may bring atmosphere interaction have been widely observed to jeopardize the
about either the onset or the progression of slope failure, with serviceability and safety of the infrastructure (e.g. Fleming & Taylor
eventual final slope instability. Failure results from the slope 1980; Alexander 1986; Hungr et al. 1999; Guzzetti et al. 2003;
atmosphere interaction especially within slope portions whose Schmertmann 2006; Dijkstra & Dixon 2010; Klose et al. 2014).
stability is marginal and may be lost as a result of even small With regard to natural slopes, the damage to buildings and
increases in porewater pressure. Otherwise, failure may occur within infrastructure interacting with unstable slopes is increasing,
shallow slope portions whose stability is strictly dependent on the mostly as a result of the increase of structures built in areas of

2017 The Author(s). This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/
licenses/by/3.0/). Published by The Geological Society of London. Publishing disclaimer: www.geolsoc.org.uk/pub_ethics
 G. Elia et al.

high landslide hazard connected to climate, and hence of structures The paper reports first an outline of the phenomena to be treated
put at risk by wrong planning schemes (Cascini et al. 2005; in an advanced modelling, with a review of the possible
Cotecchia et al. 2010, 2011). The safety of railways and roads is mathematical formulations of the processes, according to different
often threatened by debris flows involving soil cover on the steep levels of advancement. In this way a methodological framework is
flanks of infrastructure. Such flows show the highest rates and are portrayed as the basis for the classification of modelling strategies of
the most damaging landslides (Cruden & Varnes 1996; Picarelli different levels of accuracy. Several case histories are then
et al. 2004; Jakob & Hungr 2005; Cascini et al. 2010b; Hungr et al. considered as examples of the application of the different modelling
2014), and are triggered by drops in soil suction at shallow depths strategies.
due to rainfall infiltration. Furthermore, despite the fact that the
highest soilatmosphere exchange flows take place at shallow
depths, recent research has demonstrated also that deep slope
General formulation
movements found to jeopardize the safety of roads and railways may The stability of either natural or artificial slopes depends on thermo-
be triggered by slopeatmosphere interaction. The deep movements hydro-mechanical processes taking place in the soil, which are
may relate to the variations in seepage conditions at depth connected to both the climatic and vegetation conditions at the
consequent to the slopeatmosphere interaction, as the whole ground surface (Fig. 1). The climatic factors representing the
seepage domain in the slope reacts to the phenomena taking place at atmospheric conditions are, primarily, rainfall, relative humidity,
the top boundary. Seasonal excursions of the piezometric heads of temperature, net solar radiation and wind speed, which, together
even 2 3 m have been measured from 30 m down to 50 m depth in with the vegetation, determine the top boundary conditions for the
clayey slopes (Cotecchia et al. 2014) and shown to be due to the seepage taking place through the soil pores. When the water table is
seasonal climatic processes. Such seasonal cycles in porewater below the ground surface, the soils above the water table may be in
pressure have been shown to bring about yielding of deep weak unsaturated conditions. Hence, seepage phenomena within the soil
clays and consequent seasonal deep movements (Cotecchia et al. pores in the slope may involve both liquid and gas transport and may
2015). Furthermore, in slopes that are the location of old landslide be also affected by thermodynamic processes taking place within
bodies of marginal stability, seasonal cycling of porewater pressure the pores. The atmospheric conditions vary with time and, as such,
due to climate has been shown to trigger seasonal reactivation of determine a variable boundary condition that causes variations of
sliding (Cotecchia et al. 2008). Also, seasonal reactivations of the porewater pressure distribution across the whole slope and, in
damage to buildings and infrastructure interacting with such slopes turn, variations of the available soil strengths and slope stability (Lu
have been recorded. & Likos 2004; Gens 2010; Fredlund et al. 2012; Lu & Godt 2013).
The work presented in this paper aims at contributing to the The main variables defining the thermo-hydro-mechanical state
dissemination of knowledge about the modelling strategies that of the slope soils, assuming the porewater to be fresh (with either
can be employed to assess the effects of slopeatmosphere zero or constant low solute concentration) or disregarding any
interaction on the stability of engineered slopes and has been chemical process, may be set as (Table 1) the pore liquid pressure
conducted within the COST programme Impact of climate (scalar Pl ), the pore gas pressure (scalar Pg), the temperature (scalar
changes on engineered slopes for infrastructure. The assessment T ) and the solid phase displacement (vector u). The values attained
of the processes activated in the soils by climate is a necessary by these variables are essentially controlled by the balance
background for the evaluation of the instability risk for slope equations (Olivella et al. 1994; Olivella 1995; Gens 2010; see
structure systems and has to be based on interdisciplinary analyses. Table 1): (1) mass balance of liquid, variable Pl (kPa); (2) mass
These entail knowledge in various scientific fields, mainly soil balance of gas, variable Pg (kPa); (3) internal energy balance for the
hydromechanics, hydrology, meteorology, agriculture and thermo- medium, variable T (degrees); (4) momentum balance for the
dynamics. The source of the disciplinary broadness of the subject medium, variable u (m). Because the thermal, hydraulic and
and of the complexity of the modelling is the coupling of several mechanical processes in the soil are coupled, the variables of the
phenomena taking place in the slope as a result of the soil different balance equations are related to each other and the balance
vegetationatmosphere interaction. Over time, the capacity of equations should be solved accounting for such coupling.
models to simulate such phenomena has increased. The paper Nevertheless, different levels of coupling are accounted for in the
presents a review of the possible strategies to model these different modelling strategies, resulting in different levels of
phenomena, from the variation of the piezometric conditions, to accuracy in the assessment, or prediction, of the slope conditions.
the evolution of the slope stability with time, and the correspond- The physical variables involved in the processes are controlled by
ing soil displacements across the slope. physical laws, such as the law expressing the liquid and the gas

Fig. 1. Schematic slope model and

potential slopevegetationatmosphere
interaction phenomena. GWT, ground
water table.
 Slopevegetationatmosphere interaction

Table 1. Balance equations and boundary conditions

Nodal variables
Balance equations (unknowns) Physical laws and constitutive properties Boundary conditions
Mass balance of liquid: liquid water and air dissolved Pl Darcys law Pressures/fluxes/
in water (H balance) meteorological factors
Ficks law
Henrys law
Psychrometric law
Ideal gas law
Retention curve
Conductivity functions
Diffusion/dispersion coefficients
Mass balance of gas: water vapour and air (G balance) Pg
Internal energy balance (T balance) T Fouriers law Temperatures/fluxes
Thermal conductivity
Density variation with T
Momentum balance (M balance) u Soil constitutive model and corresponding Displacements/forces
parameters

transport, the diffusion law, the change of phase law and the from the combination of the climatic variables, vegetation and soil
constitutive law of the soil skeleton. In the theoretical framework of state and behaviour, are negative fluxes. The difference R ET
the hydraulics of equivalent porous media, the laws that may be used RO represents the infiltration water flux at ground surface (Fig. 1).
to simulate some of the basic processes involving the pore fluids are As will be discussed below, whereas the rainfall will be always an
Darcys law, to describe either the liquid or the gas advective flux, input flux of the model at the ground surface, the evapotranspiration
and Ficks law, for the vapour and the air non-advective fluxes. flux may be a result of the model simulation, if thermohydraulic
Furthermore, Henrys law may be used to predict the air dissolution coupling is accounted for, while it has to be introduced as an input
in the liquid phase, the psychrometric law to correlate suctions and boundary condition in the analyses disregarding such coupling. As
relative humidities at a given temperature and predict the transition specified in the section on thermohydraulic modelling, thermo-
of water from the liquid to the gas phase, and Fouriers law to hydraulic analyses require the characterization of the slope material
describe conductive heat fluxes through either the fluid or the solid properties also in terms of thermal conductivity and volumetric heat
skeleton (see Table 1). Moreover, soil constitutive laws are required capacity and account for the direct interference in the thermohy-
in the integration, which are typically: the soil water retention curve draulic conditions of the soil from all the climatic variables, such as
(SWRC) expressed as the variation of the volumetric water content, air temperature, net solar radiation, wind speed, relative humidity
, or the degree of saturation, S, with the soil porosity, n, and the soil and rainfall. However, to date not all the climate effects can be yet
suction, s; the hydraulic conductivity function, k(s, n, T ); the modelled by simulating the specific physical phenomena taking
mechanical constitutive law; the thermal conductivity law, which place at the top boundary of the slope and some empirical laws, or
depends on T, S and n; the air conductivity law; and the diffusion and some simple assumptions, have to be used (e.g. for the effects of
the dispersion coefficients, both dependent on T, S, n and Pg. wind). Common to most models, either solely hydraulic or coupled,
In the framework of continuum thermo-hydro-mechanics and runoff processes are modelled in a simplified way (Blight 1997).
equivalent porous medium, the numerical modelling will require a They are often assumed to occur when the rainfall rate exceeds the
spatial discretization of the slope (Fig. 1). Therefore, the variables, saturated permeability of the soil (Krahn 2003). At this stage, the
or unknowns, of the analysis are obtained as simulation output at a very top porewater pressure is set to zero, hence the top boundary
discrete number of points. The following should be noted: condition turns into a pressure value from being a flux value.
Another established approach assumes that runoff occurs when the
(1) Usually the shallow layers, just below the ground surface, very top porewater pressure becomes zero (Smith et al. 2008). Also
which are location of partial saturation conditions and host in this case, at the zeroing of the top porewater pressures the top
the vegetation roots, require finer discretization due to the boundary condition starts by being provided as a pressure value,
more highly variable gradients with time triggered by the until the very top porewater pressure becomes negative. In both
slopeatmosphere interaction (as shown in Fig. 1). cases, the onset of ponding can be allowed by specifying positive
(2) The variability of the slope soil properties, assessed by values of porewater pressure corresponding to the desired height of
means of site investigations and laboratory studies, has to the pond (Smith et al. 2008). However, a more advanced
be included in the numerical model through appropriate hydrological assessment of the runoff flow may be implemented
discretization strategies. in the modelling at the ground surface. Probabilistic analyses may
(3) Three-dimensional analyses may be required, especially offer a rational way for engineers to implement the variability of the
when the hydraulic, morphological and mechanical meteorological conditions at the top boundary of the slope model
features of the slope are spatially highly variable in the within hazard analyses (e.g. Li & Lumb 1987; Griffiths & Fenton
out-of-plane direction. 1993; Rezaur et al. 2002).

The boundary conditions for the balance equations in Table 1, at the

Modelling strategies
lateral and base boundaries of the model (see Fig. 1) may be as
follows: (1) the liquid pressures and/or fluxes; (2) the gas pressures In the following, available numerical strategies to model the effects
and/or fluxes; (3) the temperature and/or thermal flux; (4) the solid of the slopeatmosphere interaction on the thermal, hydraulic and
displacements and/or forces. mechanical state of the slope are discussed, which are of different
At the ground surface, the rainfall (R) represents a positive flux, levels of complexity. The discussion will proceed from the simplest
whereas the evapotranspiration (ET) and the runoff (RO), resulting to the most advanced ones. The simplest integrate only part of the
 G. Elia et al.

algorithms simulating the processes taking place in the slope and hydraulic conductivity functions was provided by Leong &
may even disregard their coupling. This is the case for the hydraulic Rahardjo (1997a,b), but several have been the recent developments
models, which simulate the transient seepage in the slope evolving of the algorithms suited to represent such soil hydraulic properties
with the atmospheric conditions by accounting solely for the fluid (Wheeler 1996; Wheeler et al. 2003; Li 2005; Sun et al. 2007;
mass-balance equations and disregarding the effects of the Khalili et al. 2008; Nuth & Laloui 2008a; Pedroso & Williams
variations in temperature within the soil and the deformation of 2010; Cafaro & Cotecchia 2015), which are not strictly empirical,
the soil skeleton. The models that take account also of the thermal but rather also based on a theoretical interpretation of the
processes within the soil and of the thermohydraulic coupling are phenomena. It is worth highlighting that, when the Richards
more accurate than the hydraulic ones and will be presented equation is used, the water retention curve is expressed in terms of
afterwards. Finally, the models that account also for the coupling of the volumetric water content, (s) = Sn. In this case, the soil
the fluid mass balances with the soil skeleton deformations will be deformability inherent in the drying-wetting process and causing
discussed. the variation in porosity, n, that contributes to the variation of upon
dryingwetting is taken into account; hence the integration accounts
for the soil skeleton deformation caused by the variations in suction.
Hydraulic modelling
However, it disregards the strain compatibility across the soil system
The assessment of the slope stability conditions with time requires and the possible consequent redistribution of stresses and variations
the prediction of the variations of the porewater pressures across the in stressstrain conditions (Tsaparas et al. 2002; Calvello et al.
slope in relation to the atmospheric conditions. If the porewater 2007; Cascini et al. 2010a; Tommasi et al. 2013; Cotecchia et al.
pressure distribution is derived based upon the liquid and gas mass- 2014), hence it does not implement the full hydromechanical
balance equations (Table 1), but is not coupled with both the coupling. Below, the Pisciolo case study provides an example of
momentum and the energy balance (and hence does not account for hydraulic modelling of the evolution of the slope conditions and
the skeleton deformations and the effects of the variations in stability with time, implementing the ingredients noted above.
temperature within both the soil skeleton and the pore fluids), the When the SWRC is implemented by inputting the function S(s) and
modelling will be defined as purely hydraulic. In this case, all the the porosity is assumed constant, the seepage problem is integrated
terms referring to the mechanical and thermal behaviour of the soil totally disregarding the soil skeleton deformability (even the
skeleton and the thermal behaviour of either the liquid or the gas deformations upon dryingwetting).
phase are not accounted for. Once the hydraulic regime within the slope is realistically
Hence, in the hydraulic modelling the porewater pressure simulated, the corresponding pore pressure distribution with time
computation results from the analysis of the transport processes can be employed as input in slope stability analyses, performed, for
(Darcys law, Table 1), the diffusion processes (Ficks law, Table 1) instance, using limit equilibrium (Abramson et al. 1996), to derive
and the dissolution of air in water (Henrys law, Table 1). The the variation with time of the slope stability factor.
liquid and the gas densities are assumed to vary exclusively with the The variation in slope stability factor for different climatic
liquid and the gas pressure; that is, isothermal conditions are conditions (e.g. in winter and summer) may be derived even using
considered (Freeze & Cherry 1979; Fredlund & Rahardjo 1993; the results of steady-state seepage analyses. These represent even more
Gens 2010). basic approaches to estimate the differences in the slope seepage for
Further simplification can be introduced in the hydraulic different climatic conditions (e.g. in summer and winter). They
modelling by disregarding the gas balance (e.g. by assuming the simulate a steady-state slope seepage representative of the conditions
gas pressure as constant) and neglecting both the water vapour at the time of year of interest, without exploring the transient seepage
diffusion within the gas phase and the dissolution of air in water. In that takes place throughout the year. If saturated conditions are
this case the model simulates only the liquid mass balance. If assumed to apply above the water table, the model will have to
also the liquid density is assumed to be constant, the liquid implement hydraulic boundary conditions (either as porewater
mass-balance equation turns into a volume-balance equation of pressures or as fluxes) able to provide a piezometric regime consistent
liquid water, which is commonly known as Richards equation with the regime at the time of interest in the analysis. In this case, the
(Richards 1931). model will be realistic if piezometric data from the field monitoring are
Given the uncoupled approach described above, in the hydraulic available and are used to calibrate the model boundary conditions to
modelling the evapotranspiration of water from the soil must be an generate piezometric predictions across the slope close to the
input boundary outflow at the ground surface. Therefore, evapo- monitored data. Hence, the seepage modelling becomes a back-
transpiration has to be estimated using phenomenological inter- calculation of the surveyed slope seepage. Such an approach may be
pretations and semi-empirical laws, which must relate the climatic useful, for example, when comparing the stability of the slope at the
variables to such outflow and have to be calibrated based upon the end of the rainy season and of the dry season, once seepage is back-
outcropping soil conditions and the vegetation typology and state. calculated for both stages (see the Volturino case history presented
Several approaches have been proposed to estimate evapotranspir- below). The approach is even more accurate if steady-state seepage
ation (Penman 1948; Thornthwaite 1948; Turc 1954; Monteith models that implement partially saturated soil conditions above the
1965). The FAO PenmanMonteith method (Allen et al. 1998) is at water table are considered.
present one of the most commonly adopted to provide physically Strength parameters are input for the limit equilibrium analyses.
based estimates of the evapotranspiration rate. It uses empirical When dealing with the stability of slope portions that are mostly
algorithms to represent the different physical processes that partially saturated, the strength of the unsaturated soils can be
combine in giving rise to the outflow, accounting for the represented using the MohrCoulomb failure criterion modified to
outcropping soil, the vegetation typology and the plant conditions account for partial saturation (Fredlund et al. 1978), as was done by
during the year. Ng & Shi (1998), for example. They used the steady-state hydraulic
The hydraulic properties of the soils are the main internal approach to simulate the seepage in the slope and employed the
parameters of the hydraulic model. For the top partially saturated revised version of the failure criterion and Bishops simplified limit
soils, both the water retention curve and the hydraulic conductivity equilibrium method (Bishop 1955) to assess the variation in stability
function must be implemented in the model, once derived from of a typical unsaturated cut slope in Hong Kong during different
laboratory testing. A wide review of water retention curves and representative rainfall events.
 Slopevegetationatmosphere interaction

Thermohydraulic modelling system, the analysis complies with a full hydromechanical coupling.
If the solution of the mass-balance equations is carried out
The modelling is defined as thermohydraulic when the equations separately from that of the momentum balance, by a staggered
describing the mass balance of both the liquid and the gaseous approach, full coupling is not pursued and the accuracy of the results
phase are solved together with the equation describing the energy will depend on the size of the calculation step.
balance (e.g. Krahn 2003). Generally speaking, such a modelling Irrespective of the adopted solution approach, any hydromech-
approach allows the prediction of the liquid pressure, Pl, the gas anical simulation requires the definition of the initial state of the soil
pressure, Pg, and the temperature, T. However, either because of in the slope, resulting from its geo-hydro-mechanical history.
shortage of input data or to reduce the complexity of the numerical Hence, the initialization of the stressstrain conditions across the
formulation, Pg is often assumed to be constant (i.e. equal to the model represents a very challenging task, which can be rarely
atmospheric pressure) and the mass-balance equation of the gaseous accomplished by simply applying an initial gravitational load. In
phase is disregarded. particular, the model prediction will depend on the initial conditions
With T being calculated at each node, the thermohydraulic especially when more advanced soil constitutive laws are adopted.
modelling has the advantage of predicting the water transition from Elasto-plastic laws are the most commonly used to reproduce the
the liquid to the vapour phase, which takes place when the latent soil behaviour and have also been extended to partially saturated
heat of vaporization is supplied to the soil as a result of heat conditions.
exchanges with the atmosphere. This means that the evapotrans- In the Volturino case history presented below (Lollino et al.
piration fluxes can be derived from the numerical simulations, 2010, 2016), hydromechanical finite-element analyses have been
provided that all the climatic variables are defined (i.e. air performed adopting an elasto-plastic MohrCoulomb model for a
temperature, net solar radiation, wind speed, relative humidity and slope initialized with the K0 condition followed by excavation.
rainfall), together with the main characteristics of the vegetation on These analyses are carried out for two permanent hydraulic steady-
the slope, and employed in the calculation of the liquid, gas and heat state conditions, respectively representative of the average condition
fluxes at the top boundary of the model. Boundary conditions need for summer and winter. When the model is set to investigate the
to be defined also on the lateral and the bottom boundaries of the effects of the slopeatmosphere interaction with time, instead, the
mesh, either as liquid, gas and heat fluxes, or directly as imposed numerical simulations have to reproduce the stressstrain evolution
pressures (Pl and Pg) and temperatures (T ). associated with the pore pressure variations. In this case, the
Similarly to the transient seepage computed through the analysis of the deformation processes has to be coupled with the
hydraulic modelling, for the transient thermohydraulic processes analysis of the transient seepage processes induced by the climatic
all the nodal variables need to be initialized. However, in this case, conditions. The simplest strategy to represent these effects
the initial conditions of the slope have to be defined not only in implements at the ground level, as a boundary condition of the
terms of liquid pressures, Pl, but also in terms of temperatures, T model, suction variations close to those really occurring in situ in
(and, eventually, in terms of gas pressures, Pg). To date, the required the topsoils as an effect of the climatic perturbation. Kovacevic et al.
comprehensive characterization of the initial slope conditions (i.e. (2001) and Nyambayo et al. (2004), among others, adopted this
also in terms of temperature and gas pressure) still poses serious modelling approach to investigate the stability with time of typical
difficulties in the use of this modelling approach in practice. vegetated railway embankments in the UK, developing fully
The solution of the energy-balance equation introduces add- coupled hydromechanical finite-element simulations, but assuming
itional complexity, as the thermal conductivities and the volumetric the soil to be saturated above the water table. Those researchers
heat capacities of the different slope materials also become essential employed an elastic strain-softening MohrCoulomb model,
model ingredients, whose definition requires additional tests (e.g. capable of simulating the key features of the mechanical behaviour
Woodside & Messmer 1961; Midttmme & Roaldset 1999). of the materials forming the investigated slopes.
Moreover, because of the relevant role played by the unsaturated More recent studies have reported the results of fully coupled
soil behaviour within the analyses, the variation of both the thermal hydromechanical simulations in which the unsaturated hydromech-
conductivities and the volumetric heat capacities with the soil anical soil behaviour is also reproduced (Fredlund & Rahardjo
volumetric water content should also be defined. 1993; Wong et al. 1998; Smith 2003). The formulation of the basic
Performing non-isothermal analyses also gives the possibility of equations originally developed for saturated conditions is modified
introducing the dependence of all the material properties on T. The to model the elasto-plastic deformations under partial saturation
hydraulic conductivity, for instance, can be defined as a function of conditions. Moreover, the unsaturated formulation should also
both void ratio and temperature (e.g. Lambe & Whitman 1969; accommodate the definition of both the hydraulic conductivity and
Delage et al. 2000). the water retention curve. The latter is here necessarily defined in
Because of the large amount of input data required and the model terms of degree of saturation, as the soil deformations evolve with
complexity, 2D thermohydraulic modelling of the slopeatmos- time and control the soil porosity. This approach has been adopted
phere interaction has rarely been performed so far (e.g. Cotecchia by Rouainia et al. (2009) to develop fully coupled modelling of the
et al., in prep.), whereas 1D modelling is more frequently discussed effects of future climatic scenarios on the stability of cuttings in
in the literature (e.g. Rajeev et al. 2012). London Clay, as discussed below. In this case, the pore pressure
variations at ground level, in the unsaturated soils, have been
calculated through an advanced modelling of infiltration, account-
Hydromechanical modelling
ing for rainfall, runoff and evapotranspiration, all input at the model
The assessment of the slope displacements resulting from the slope top boundary.
atmosphere interaction is crucial for the evaluation of the When the unsaturated soil behaviour is simulated, the constitutive
serviceability of structures interacting with slopes and requires the models cannot be formulated using the effective stress proposed by
performance of mechanically coupled numerical analyses Terzaghi (1936) as the single stress variable. Following Gens et al.
(Zienkiewicz & Taylor 1989; Potts & Zdravkovic 1999, 2001; (2006) and Nuth & Laloui (2008b), there are two approaches to
Zienkiewicz et al. 1999). Such analyses entail the solution of the formulate the stress variables within the unsaturated soil constitutive
momentum-balance equation complying with the variation in time law, both including the suction in the soil stress state. The first
of the pore liquid and gas pressures. If the mass-balance equations approach accounts for two independent stress variables to describe
and the momentum balance are solved at the same time as a single the behaviour of the unsaturated soil, according to three possible
 G. Elia et al.

strategies (Fredlund & Morgenstern 1977): (1) PgI, (2) PlI Research into the modelling of the effects of cracking has
and (3) (Pg Pl )I, where is the total stress tensor and I is the increased in recent decades. Approaches that do not explicitly
second-order identity tensor. For the second approach, an effective predict the cracking process but implement the variation in soil
stress tensor is provided by Bishops equation (Bishop 1959) properties due to cracking include the estimation of the crack
evolution with both depth and time (Fredlund & Rahardjo 1993),
s0 s  Pg I xsI bimodal SWRCs, hydraulic conductivity and water storage
where is the effective stress parameter. The experimental results functions (Fredlund et al. 2010) for the soil and the cracks, and
have shown that the parameter depends on factors such as the have been applied also to full-scale slopes by Booth (2015).
wetting and drying history, the void ratio and the soil structure Other research has focused on capturing the processes that initiate
(Rojas 2008). Several equations have been developed to define . cracking, but does not attempt to incorporate discrete cracking into
The equation = S is the most often used, although several full-scale slope geometries. Researchers have followed two
alternative formulations are available (Vanapalli et al. 1996; Khalili alternative numerical strategies, modelling cracking in either a
& Khabbaz 1998; Alonso et al. 2010). Houlsby (1997) has shown continuum or a discontinuous medium. Both finite-element and
that both approaches are valid, provided that the appropriate finite-difference methods (FEM and FDM) have been employed
conjugate variable of deformation is used. and represent the simplest way within traditional slope stability
One of the first and still most commonly used constitutive laws modelling to reproduce the main effects of cracks. Meshing in FEM
for unsaturated soils was presented by Alonso et al. (1990). This and FDM typically limits the ability to capture crack localization
law, known as the Barcelona Basic Model (BBM), is formulated and propagation. Therefore, researchers have developed novel
using the first approach for the stress variables. It has been schemes of mesh fragmentation and discretization (Sanchez et al.
employed in the hydromechanical modelling used by Askarinejad 2014; Stirling 2014; Pouya et al. 2015). Alternatively, the extended
(2013), reported in the case histories section below, to evaluate the finite-element method (XFEM), where additional degrees of
stability of the banks of the River Rhine (Switzerland), with freedom are included to represent cracks within finite elements,
reference to both 2D and 3D conditions. can be used for fluid-driven cracking, such as desiccation
Several other constitutive laws, either stemming from the BBM or (Mohammadnejad & Khoei 2013), although at present the
sharing the same choice of stress variables, have been presented in application of this strategy to soil desiccation is limited.
the literature (Wheeler & Sivakumar 1995; Chiu & Ng 2003; Sheng The application of linear elastic fracture mechanics has also been
et al. 2008). However, recent constitutive models have been explored (Lachenbruch 1961) and included in a modelling strategy
formulated in accordance with the second approach (Pereira et al. that uses a stepped approach to predict the moisture profile (with 1D
2005; Russell & Khalili 2006; Khalili et al. 2008; Khalili & FEM), then crack depth, through linear elastic fracture mechanics,
Zargarbashi 2010). and spacing, through linear elastic FEM (Konrad & Ayad 1997).
It is worth observing that advanced constitutive laws should be Alternatively, the discrete element method (DEM) has been used
employed together with advanced formulations of the water to simulate the medium subjected to cracking as an assemblage of
retention curves (e.g. Gallipoli et al. 2003; Tsiampousi et al. single, interacting grains subject to contact laws (Peron et al. 2009;
2013a; Cafaro & Cotecchia 2015), to take into account the intrinsic Amarasiri et al. 2011; Muslelak & Sliwa 2012). This approach
coupling between suction, void ratio, degree of saturation and soil allows the formation of discontinuities as a result of grain bond
mechanical properties. Such an approach has been proposed by breakage and can even capture fluid flow through the cracks.
Tsiampousi et al. (2013b) and employed by Pedone et al. (2016) to However, DEM is limited in its ability to simulate the multi-phase
investigate the stability of a slope interacting with both hydraulic processes within the matrix before cracking; that is, drying is often
and transport infrastructure. inferred through prescription of porewater pressure gradients, rather
than by applying moisture variations, or heat or mass exchanges
(Bui et al. 2015).
Swellingshrinkage cracking
Desiccation cracking is the result of shrinking in clays and is
Modelling the effects of vegetation at the plant scale
commonly considered to initiate when, as a result of the volumetric
contraction of the top strata where the main effects of evapotrans- Vegetation has to be modelled when dealing with slopeatmosphere
piration are observed, the tensile stresses exceed the soil tensile interaction, because it affects the water balance of the slope. As
strength, which is generally very low. The morphology of observed mentioned above, transpiration processes occurring in the
soil cracking is generally found to depend on the vegetative cover topsoil layers result in a negative flux. Furthermore, leaves and
(i.e. root reinforcement, water uptake potential, etc.), the soil steams shield the slope, intercept water and limit the infiltration of
plasticity index, the soil mesofabric and the drying conditions rainfall.
( potential evapotranspiration magnitude and duration). With However, transpiration processes, although increasing the slope
changes in soil moisture, the soil can crack in a brittle mode stability (Glendinning et al. 2009), may induce cyclic movements,
under dry conditions, after a linear elastic stage, or in a ductile way, as they occur on a seasonal basis. This cyclic loading can be
when the moisture content is high and the soil is soft (Hallett & detrimental for serviceability if it occurs in the vicinity of
Newson 2005). infrastructure. Moreover, when the cyclic pore pressure variations
The ability to capture the generation of discontinuities effect of are pronounced and prolonged, they can also promote progressive
desiccation is crucial for an accurate simulation of the slope failure phenomena (Take & Bolton 2011).
response, as cracking reduces the soil shear strength and increases All these aspects can be predicted by advanced coupled
the water infiltration. The development of coupled hydromechanical numerical modelling, if a boundary condition simulating the
models that include cracking in clayey soils represents a challenge presence of vegetation is appropriately implemented. This is the
for numerical modelling, given the difficulties inherent in the case in the numerical simulations presented by OBrien et al.
mathematical simulation of crack generation and flow through the (2004), who modelled the hydromechanical behaviour of railway
cracks. The presence of both discrete deep cracks and extensive embankments covered by either grass or trees. Transpiration
shallow crack patterns is generally the source of a more rapid processes induced by grass were simulated by applying suctions
response of the seepage domain to climatic changes, which may at the top boundary of the model, whereas the presence of trees was
ultimately lead to slope failure. simulated by applying an internal boundary condition acting in the
 Slopevegetationatmosphere interaction

area of the slope where the roots of the trees are concentrated. The studied the stability of the Radiowo landfill located in the
numerical modelling was validated by means of a comparison with northwestern part of Warsaw (Poland). The stabilizing effects of
the available track monitoring data and proved that significant the vegetation have been considered in the factor of safety
deformations can occur as a result of the seasonal pore pressure calculation, making use of the General Greenwood Method
variations associated with transpiration. Similarly, Lees et al. (2013) (Greenwood 2006). The results of the stability analysis indicate
have shown how the mechanical response of an old clay that a 20% increase of the factor of safety can be reached along the
embankment was strongly influenced by the transpiration processes vegetated slopes of the landfill, thus showing how slope
of some trees located in the lower part of the slope. The presence of reinforcement does not always require heavy engineering
the trees was modelled by applying the suctions induced by their methods, but can be achieved with simple, environment-friendly
roots and the results were verified by comparison with inclinometric and cost-effective techniques, such as the implementation and
and piezometric measurements. maintenance of a vegetation cover. Nevertheless, the mechanical
The direct application of suctions as an internal boundary improvement provided by the root systems is strongly controlled by
condition represents a simplified way of simulating the effects of the depth of the root zone. Therefore, the vegetation mechanical
transpiration and may lead to unrealistic porewater pressure effects are less significant in slopes where deep-seated failure
predictions. Otherwise, root water uptake models should be mechanisms are likely to occur.
employed (Nyambayo & Potts 2010), in which the transpiration
fluxes represent the input ingredient, so that pore pressure variations
Case histories
can be estimated and not imposed.
It is worth observing that vegetation can also induce permeability This section presents a collection of real and ideal cases where the
variations of the topsoil layers, because high suction values can stability of natural and artificial slopes subjected to climatic
trigger the formation of cracks (as discussed above), which increase perturbations has been investigated by some of the participants in
the soil permeability (Li et al. 2016). Similarly, if the vegetation has the COST Action, using different modelling strategies, classified as
been removed from the slope, the outcropping soil layers should be outlined so far (Table 1).
considered more permeable, because the holes created by the roots
will act as preferential pathways for the infiltration of rainwater
Pisciolo slope
(Smethurst et al. 2015).
Advanced numerical modelling should also account for the The influence of the slopeatmosphere interaction on slope stability
mechanical characteristics of the vegetation roots. This is because has been investigated implementing the results of hydraulic slope
the tensile strength of roots within the soil mass can improve the modelling in limit equilibrium analyses for the Pisciolo hill-slope
capacity of the soil to resist the mobilized shear stresses. The (Melfi, Italy) (Cotecchia et al. 2014), which is the location of deep
maximum tensile strength, or pull-out resistance, of the roots slow landslide movements typically observed in fissured clay slopes
together with their size and distribution with depth are the in the south of Italy. The slope has been thoroughly studied because
ingredients of the modelling of the reinforcement effects of important infrastructure has been severely damaged owing to its
vegetation for slope stability analyses. In particular, the experimen- interaction with the slope movements. In particular, a pipeline, a
tal data obtained from direct shear tests performed on blocks of soil road and a railway are located at the toe of the slope, where landslide
containing roots have shown that the presence of vegetation activity is highest. The geological setting of the area has been
produces an increase in soil cohesion, leaving the friction angle carefully reconstructed based upon in situ surveys and underground
unchanged (Wu et al. 1988; Faisal & Normaniza 2008). Such a investigations. The slope is mainly formed of clayey turbidites that
mechanical effect is usually introduced in the MohrCoulomb are part of the Paola Doce formation and consist essentially of high-
failure criterion through an apparent cohesion term, cR, which plasticity fissured clays with interbedded coarse and fractured rock
adds to the soil effective cohesion, c (Waldron 1977; Wu et al. inclusions. Paola Doce clays do not outcrop all over the slope, due to
1979). the presence of an anticline with an axis that crosses the hill-slope
Wu et al. (1979) studied the stability of natural slopes on Prince transversally. The central portion of the slope has been uplifted and
of Wales Island (Alaska), before and after the removal of a forest deeper clays, belonging to the Red Flysch formation and
cover. They incorporated the apparent cohesion owing to the roots characterized by even higher plasticity and fissuring degree than
into simple limit equilibrium analyses, assuming infinite slope and the Paola Doce clays, crop out about the fold axis. Also, more recent
steady-state seepage conditions. The results indicated that the arenaceous and sandy materials, which are part of the Numidian
additional strength provided by the tree roots is important for the Flysch, crop out on the fold flanks. This geostructural reconstruction
stability of steep slopes and the loss of root strength following clear- has supported the generation of the geotechnical model of the slope,
cutting can seriously affect slope stability. Chok et al. (2004) and which implements much higher strength properties and higher
Gentile et al. (2010) analysed the mechanical effect of vegetation on permeabilities where the Numidian Flysch and the rock or sand
the stability of ideal slopes, characterized by dimensions and inclusions are located, and a great reduction in strength parameters
material properties typical of highway embankments, using the FE and permeability within the slope locations of the Paola Doce clays
method. The results show that vegetation plays an important role in (cav = 0 kPa, av = 20.7) and the Red Flysch clays.
stabilizing shallow-seated failure of slopes, indicating that the Ten rotationaltranslational clay-slides have been identified to be
increment of the slope factor of safety (FoS) is more significant if active within the Pisciolo hill-slope, based upon stereoscopic
the slope toe elements are treated as vegetated soil. The effect analyses of past and recent aerial photographs, geomorphological
increases with the deepening of the root system, reaching the zones surveys and the interpretation of inclinometer data. The landslides
where the failure mechanism starts. Nevertheless, the factor of involve mainly the Paola Doce soils (clays and interbedded sands
safety of the slope cannot increase indefinitely, as it reaches an and rocks) and, locally, the Red Flysch. The slope has been
asymptotic value with increasing root additional cohesion. Similar subjected to extensive piezometric monitoring, conducted by means
results were obtained by Ji et al. (2012), who performed 2D FE of Casagrande and electrical piezometers. Most piezometric levels
simulations to investigate the effect of root additional cohesion on reach a few metres below the ground level, also for very deep
the stability of two natural slopes in NW China. piezometers (down to 60 m depth). The landslides are a
Another example of numerical assessment of the root mechanical consequence of the relatively high slope mass permeability,
reinforcement is given by the work by Koda & Osinski (2011), who arising from the fissuring of the clays and the occurrence of the
 G. Elia et al.

coarse strata and the fractured rock inclusions within the clays. As difference between the total rainfalls and the evapotranspiration
well as the presence of the higher permeability inclusions, the fluxes, the latter estimated by means of the FAO PenmanMonteith
average permeability of the clay portions (site measurements: ksat > method (considering the presence of winter wheat and using the
109 m s1), which is higher than for unfissured clays by a few single crop coefficient approach). The numerical results have been
orders of magnitude, eases rainfall infiltration, which generates high compared with the piezometric heads measured along three verticals.
piezometric heads that are detrimental to the available shear As shown in Figure 2a for the vertical P7, the piezometric variations
strengths. Weather data have been collected at a weather station near predicted by the numerical model are in good agreement with those
the slope. The 180 day cumulative rainfall evolves according to a measured in situ, especially once the more permeable inclusions are
seasonal trend, with maximum values at the end of winter and implemented in the model. When the pore pressure distributions
minimum values at the end of summer. The piezometric heads and resulting from the seepage analysis for different stages of the year are
the displacement rates of the most active landslides, located in the implemented in the limit equilibrium analysis, variations of the factor
southern portion of the hill-slope, follow a similar fluctuation trend. of safety from 5 to 20% are calculated for the landslide bodies in
In particular, at the toe of the most active rotationaltranslational Figure 2b. Such safety factor variations reflect the changes in actions
multiple landslide, the displacement rates have been measured using that cause the seasonal accelerations of landsliding on the Pisciolo
inclinometers and global positioning system (GPS) sensors and slope, which results in the recurrence of damage to both the pipeline
have been found to fluctuate seasonally. The correspondence and the road present at the toe of the slope.
between the pore pressure excursions, the displacement rate
variations and the 180 day cumulative rainfall variations confirms
Middelburgse kade peat dyke
that the climatic regime represents the main triggering factor of the
current activity of the Pisciolo landslides, and the relatively high The failure of a peat dyke at Wilnis (Netherlands) during summer
mass permeability represents the internal factor that mainly 2003 provides evidence of the effects of severe droughts on the
predisposes the slope to landsliding, along with the relatively low stability of this type of engineered slope. The failure did not cause
available strengths of both the Paola Doce and the Red Flysch clays. casualties, but its economic impact was important, as the resulting
Two-dimensional transient seepage analyses have been performed economic loss was about 20 M. Since then, interest in the
with the finite-element code SEEP/W (Geo-Slope International evaluation of the safety of the 7000 km of peat dykes existing in the
2004), which integrates the Richards equation, according to the Netherlands has increased, as demonstrated by the Knowledge for
method presented above. The analysed section, which crosses Climate programme (2008 2014) of the Dutch Ministry of
longitudinally the most active landslides, is reported in Figure 2, Infrastructure and Environment, which has supported research
together with plots of the model results compared with some into the behaviour of engineered slopes under changing climatic
monitoring data. The section includes the different materials found in conditions. Within this research programme, the stability of an
the slope: the Paola Doce clays, the Red Flysch clays and the more existing dyke near Boskoop (Netherlands), the Middelburgse kade
permeable fractured rock inclusions. The van Genuchten (1980) peat dyke, has been studied by Van Esch (2012). This example is
model has been used to fit the laboratory retention data for the presented here with the aim of outlining the evolution of both
fissured clays and to simulate the permeability variation with suction. seepage and stability applying to peat dykes and to assess the impact
Pore pressures constantly equal to zero have been imposed at the of climate changes on these engineering structures, for more
ground surface at the upper and lower portions of the slope ground efficient and effective advice for their maintenance and remediation.
surface, according to the corresponding presence of a spring and of The dyke is characterized by a height of about 3 m and a
the Ofanto river. Net rainfalls referring to the period September longitudinal length of 40 m, with height-to-width ratio of 1:14.
2006August 2007 have been cyclically applied along the rest of the Figure 3a shows a schematic cross-section of the dyke and the peat
slope ground surface. The net rainfalls have been computed as the layers that cover a sandy subsurface.

Fig. 2. Pisciolo slope: finite-element mesh

and boundary conditions employed in the
hydraulic analyses (saturated
permeabilities of the lithological units are
reported on the model, together with the
active landslide bodies for which the limit
equilibrium analyses have been
undertaken) and numerical results (dashed
lines refer to the case in which the more
permeable inclusions have not been
implemented in the model; continuous
lines refer to the model implementing also
the more permeable inclusions) compared
with in situ measurements (dots). b.g.l.,
below ground level.
 Slopevegetationatmosphere interaction

Previous research by Van Esch et al. (2007) has proposed a Po river embankments
procedure to assess the consequences of droughts for peat dykes,
using the finite-element software Plaxflow (Brinkgreve et al. 2003). The Po river (northern Italy) is the Italian main water course in terms
This procedure complemented an agro-meteorological model to of length and capacity. Past flooding events have often caused
derive the evapotranspiration flow at the ground surface, based on damage and casualities. An ad hoc public body, the Po River
the PenmanMonteith expression, and a groundwater flow model to Management Authority AIPO, is in charge of the river safety
predict the piezometric distribution based on the Richards equation issues and early warning system, manages emergencies, and
(hydraulic modelling). Flow through cracks in unsaturated subsoil finances and coordinates research activities addressing risk
was simulated via a double-porosity approach and Scotts model mitigation. In the past, flooding prevention required remediation
(Scott et al. 1983) was adopted to simulate the drying and wetting of existing defence embankments, through height increase (even up
behaviour of peat, as laboratory experiments could not be to 10 m), cross-section enlargement, or improvement of the
reproduced well by a single van Genuchten expression. It was mechanical properties of the construction materials. Throughout
found that the approach adopted to assess the slope piezometric and the river network (about 2000 km), embankments are made of
stability conditions was computationally very demanding; this compacted soils taken from pits along the river banks. To investigate
approach is worth using if detailed information on material the hydraulic behaviour of both the fine-grained embankments
behaviour is available. Therefore, using Richards model it was located in the mid-course of the river and the coarse-grained ones
found that is very important to account for anisotropy in the along its upper course, research studies have been carried out on
permeability tensor, because of the horizontal layering of peat. The full-scale physical models at three sites: in Viadana (MN), for the
horizontal component of the permeability tensor in the mostly study of the mid-course embankments behaviour, and in Bormida
saturated zone is about 10 times its vertical component. Hence, a (AL) and Motta dei Conti (VC) for the upper-course embankments.
Dupuit approximated approach (De Marsily 1986) has been used to The main objective of the research study has been to define a design
simulate the flow through the saturated zone of peat dykes. The method that takes account of the seepage in the banks and in the
Dupuit approximation states that groundwater flow in aquifers of foundation soils, under either partially or fully saturated conditions
high permeability, such as sandy deposits, is oriented horizontally (Calabresi et al. 2013; Belardi et al. 2014).
and the flow through the aquitards interbedding low-permeability In the case of Viadana, according to recommendations
clay layers takes place in the vertical direction. To verify the extent formulated by the Po River Management Authority, an embank-
to which the simplified Dupuit approach is able to provide realistic ment prototype was built on the floodplain, beside the existing
predictions in practice, Van Esch (2012) employed the Penman embankment, forming a pond. The pond was filled to reproduce
Monteith expression to generate boundary conditions for a Dupuit historical floods. The engineering properties of the foundation soils
model, and used a finite-volume scheme, implemented using in- have been investigated through in situ tests and complemented by
house software (Van Esch et al. 2013), to simulate transient suction-controlled laboratory tests (Calabresi et al. 2013). The soil
groundwater flow. In the case under investigation, saturated flow profile at Viadana includes different layers, consisting of medium to
through the peat dyke is oriented horizontally and unsaturated flow fine sands, clayey silts, medium to fine silty sands and sandy
takes place in the vertical direction. The proposed model gravels. Porewater pressures have been measured in the embank-
incorporates effective porosity and effective permeability in the ment and in its foundation before, during and after the experimental
unsaturated zone. Effective parameters have been obtained from a simulation of two floods, performed to reproduce those that
calibration procedure and have been computed more easily than occurred in 1976 and in 2000. Atmospheric variables have been
model parameters that support the soil water retention curves. The monitored at the site during the 6 months of experimental activity.
simplified model is computationally highly efficient. Figure 3a The transient seepage processes occurring in unsaturated conditions
shows the computed piezo-lines in the aquifer and the dyke, from within the prototype, given the ponding conditions and the slope
which the water pressure distribution can be derived. The figure also atmosphere interaction, have been investigated by Calabresi et al.
shows the location of four observation wells, where groundwater (2013) through hydraulic analyses conducted with the finite-
heads have been monitored over a period of 2 years. The numerical element code SEEP/W (Geo-Slope International 2004). The mesh
predictions in terms of hydraulic head obtained at different adopted in the analyses is reported in Figure 4a. The water flows at
observation points with the finite-volume Dupuit model have the ground surface due to slopeatmosphere interaction (i.e. rainfall
been compared with those obtained through the more advanced and evapotranspiration) have been computed separately and input
finite-element simulation using Richards model. Both groundwater into the model, according to an isothermal hydraulic modelling
flow models produce results that compare well, as shown in approach (see Table 1), integrating the Richards equation and
Figure 3b, thus demonstrating the good ability of the Dupuit neglecting both hydromechanical and hydrothermal coupling. The
approximation to simulate groundwater pressure fields. Figure 3c transient seepage stages following changes in ponding conditions
compares the calculated heads at different observation points with have been analysed. With this aim the hydraulic boundary
observed heads in the field, showing a good agreement between conditions have been changed over time, according to steps
numerical predictions and measured data. The pore pressure corresponding to the history of the impounding levels of the Po
distribution calculated with the Dupuit model is then used as river. A seepage surface condition has been applied along the
input for a limit equilibrium analysis, which employs Spencers unsaturated portion of the embankment boundary not subjected to
method (Spencer 1967) to evaluate the evolution of the dyke impounding, where pore pressure proved to become either negative
stability with time. The critical slip surface is shown in Figure 3a, or zero. Along this boundary, soilatmosphere interaction has been
and Figure 3d presents the stability response of the dyke over the simulated. Setting the flow rate as the net balance between the
interval under observation. The figure indicates a stable condition outflow owing to evaporation and the inflow owing to rainfall, the
throughout the simulation period as the stability factor is well above prediction of the soil suction has been less satisfactory than that of
1.0. In general, the stability of the dyke is found to decrease when its the subsoil porewater pressures (see Fig. 4b). The transient stages
weight reduces. For peat dykes the weight can strongly decrease have been found to be persistent and a significant delay in reaching
depending on the amount of water that leaves the system by steady-state conditions occurs, due to the low unsaturated soil
evaporation. A tipping point analysis has, therefore, revealed that an permeability, consistent with the soil suctions varying according to
increase of evapotranspiration by a factor of two would have led to the retention curve of the material. The water retention properties of
the failure of the peat dyke (Fig. 3d). the embankment materials are in reality inhomogeneous, given the
 G. Elia et al.

Fig. 3. Middelburgse kade peat dyke: (a) cross-section; (b) comparison between hydraulic head predictions; (c) comparison between predictions and
measurements; (d) stability response. fem, finite-element model; fvm, finite-volume model.

randomly distributed variations in grain size and/or in compaction slopeatmosphere interaction decrease. Hence, it emerges that
degree of the soils. As a consequence, suctions may have been more modelling soilatmosphere interaction is crucial to predict a reliable
scattered than those predicted by the analyses performed assuming evolution and distribution of the suctions across embankments.
homogeneous embankment properties. With regard to the differ- The distribution of the porewater pressures provided by the
ences in the predictions when accounting for the slopeatmosphere modelling has been input into embankment stability analyses, using
interaction effects and when not, the comparison in Figure 4c shows Bishops traditional limit equilibrium method. The safety factors
that these differences are major for tensiometers close to the computed in transient conditions are increasingly higher than those
embankment boundary (see locations T1A, T2A and T6A). Within associated with steady-state conditions reached when impounding is
the embankment (see locations T3A and T5A), the effects of the maintained for a time sufficient to attain a new hydrostatic regime.
 Slopevegetationatmosphere interaction

Fig. 4. Viadana embankment along the Po river: (a) sketch of mesh adopted in the seepage analyses (all dimensions in metres); (b) predicted v. measured
suction evolution within the prototype; (c) suction evolution predicted within the prototype obtained by accounting or not accounting for soilatmosphere
interaction.

There is little penetration, in the short term, of the saturation zone winterearly spring and reduces in summer, as observed for several
during severe flooding, as the vanishing of suction is limited to a thin slopes formed of clay flysch in the Italian Southern Apennines. The
zone near the wet boundary and, in some cases, along the foundation landslide body has been found to involve mainly the Toppo
boundary. Therefore, given the presence of negative pore pressures in Capuana Flysch, which is mostly made of marly clays with an
large portions of the embankment, realistic stability analyses of the average clay fraction of about 50% and plasticity index increasing
embankment during flooding should be carried out by taking into with depth, from about 10% at shallow depth, to 40% at 45 m depth.
account the strength contribution provided by suction. Geomorphological analyses (Fig. 5a) had revealed that the Fontana
Monte landslide was already active in late nineteenth century, so that
the current slope activity can be considered to be largely the effect of
Volturino slope
further straining, relating to climate, along pre-existing shear bands.
The stressstrain conditions in a slope location of a deep slow- In particular, the seasonal accelerations of the landslide suggest that
moving landslide in fissured marly clays, the Fontana Monte slope periodic variations of the shear strength occur in the slope, probably
at Volturino (Southern Italy), have been investigated by Lollino as a result of seasonal porewater pressure variations. Seasonal
et al. (2010, 2016) through hydromechanical finite-element fluctuations of the piezometric levels, resulting from transient
analyses performed with the code PLAXIS 2D (2012) according seepage processes generated by the slopeatmosphere interaction,
to the method presented above. The landslide is observed to undergo have been monitored in the slope to depth, as reported by Lollino
seasonal reactivations, as the rate of movements increases in late et al. (2016). The landslide activity in the upper portion of the slope
G. Elia et al.

Fig. 5. Volturino slope: (a)

geomorphological map with location of
main cross-section indicated; (b)
incremental shear strains for summer
condition; (b) incremental shear strains for
winter condition; (d) incremental vertical
displacement for winter condition.
 Slopevegetationatmosphere interaction

is documented by cracks and fissures that affect buildings and roads location of an active landslide since the nineteenth century, which
located along and at the back of the rear scarp of the landslide, at the has been reactivated several times, most recently in 1963, 1999 and
top of the slope, as well as along the lateral borders of the landslide 2002. The sliding mechanism is composed of two sliding masses,
body (see Fig. 5a), as consequence of differential settlements of the upper one of length 350 m, width 110 m, height 60 m and depth
foundations connected to the landslide movements. The analyses up to 24 m, and a lower one of length 200 m, width 90 m, height
have explored the slope straining processes evolving from summer 35 m and depth up to 18 m (Fig. 6a). The separation of the two
to winter in the slope, assuming the slope to be formed of masses is due to the presence of a rock buttress in the middle of the
undisturbed clays that include a deep pre-existing shear band, whose landslide (see Fig. 6a). The total area covered by the landslide is c.
morphology has been derived from investigations of bored corings 37 600 m2 and involves a mass of around 475 200 m3, making it the
and from inclinometric monitoring. The variations in piezometric largest landslide in the Balkan countries and, possibly, in SE
head from summer to winter have been input into the analyses Europe. The geotechnical investigations indicate that the sliding
through the simulation of both the summer and the winter steady- zone is along the contact between low-plasticity clay-like materials
state seepage. The validity of the seepage simulations has been and the bedrock, which is represented by Palaeozoic amphibolitic
checked against piezometric site measurements at several locations schists that are highly foliated, faulted and locally weathered. In the
along the slope. A standard linear elasticperfectly plastic Mohr upper part of the slope, additional fill is present on the top of the clay
Coulomb model with non-associated flow rule has been adopted for layer and the water table occurs at 8 m depth, whereas in the lower
the materials, in accordance with the moderately brittle soil shear part the water table is 2 m below the ground surface. The
behaviour observed in the laboratory. A cohesion of 13 kPa, an hydrogeological conditions indicate that the zones of higher water
effective friction angle equal to 20 and a coefficient of permeability content correspond to the sliding zone. Also, the piezometric levels
equal to 5 109 m s1 have been adopted in the analyses. The at depth are found to be higher than those at shallower depths in
analyses have been carried out assuming the shear band to be of almost all boreholes (Jovanovski et al. 2003). The material
either 37 m maximum depth or of 50 m maximum depth, to evaluate properties have been determined through extensive laboratory and
the influence of the depth of the pre-existing shear band on the field investigations, performed from 1963 to 2002. According to site
current slope stability. Because the shear band is formed of pre-failed observations, the sliding process appears to be caused by heavy
clays, strength parameters intermediate between peak and post-peak rainfall, as the rainfall water infiltration accesses the weak schistose
have been used for the shear band (i.e. c =8 kPa and = 18.7). The rocks. The climate is semi-arid, characterized by gusty rainfall
summer and winter analyses have been carried out under plane- events from October to March, with an average monthly rainfall of
strain drained conditions, with pore pressure distributions resulting around 100 mm. After the landslide occurrence in 2002, hydro-
from steady-state seepage calculations fitting the site measurements, mechanical simulations have been conducted (according to the
as explained above in the section on hydraulic modelling. The method presented above in the section on hydromechanical
maximum localization of shear straining has been found to occur modelling) to evaluate the effects of slopeatmosphere interaction
within the 50 m deep shear band when the piezometric heads on the stability of the slope and estimate the potential further sliding,
increase from the summer values to the winter ones. Figure 5b and using the finite-element code PLAXIS 2D (2012). Based on the
5c report the contours of incremental shear strain in the slope, which available data about the mechanical behaviour of the slope
provide indication of the shear strain concentrations in the deepest materials, the elasto-plastic MohrCoulomb model has been
shear band. The slope is found to be stable in summer and to fail in adopted. The van Genuchten hydraulic model has been employed
winter (i.e. loss of numerical convergence). The corresponding in the analysis and calibrated against measured data, as presented in
contours of vertical incremental displacement are reported in Figure 6b. The rainfall infiltration has been set as 10 mm h1 at
Figure 5d: negative displacements (i.e. settlements) are recorded in ground surface, plus an additional 1 m3 h1 on the right side as
the upper part of the slope and in the centre, whereas maximum inflow boundary. The runoff has been assumed to be not greater
positive displacements (i.e. heave), of about 20 mm just before loss than 0.5 mm h1 owing to the surface characteristics at the site
of convergence, are observed at the base, where the toe of the (which is fairly urbanized). In the model vegetation and temperature
landslide body is located. Maximum incremental settlements, of have not been taken into account and evapotranspiration has not
about 10 mm, are calculated in the scarp area, where buildings are been modelled. Mesh updating has been performed, due to the
periodically damaged. As such, the accumulation of settlement occurrence of large straining. A realistic simulation of the sliding
along the slope crest over few years provides a reason for the mass behaviour has been obtained, as presented in Figure 6c: the
observed pattern of structural damage. sliding of the mass occurs from around 12 h after the start of the
This numerical study proves the tendency of the deep Fontana rainfall, with maximum suction of 169.7 kN m2 (Fig. 6d) and
Monte landslide to reactivate as an effect of seasonal piezometric maximum displacement of 1.17 m (at the rear scarp). The results
fluctuations. Furthermore, it proves that this type of slope is most show a sliding mechanism whereby the upper part of the slope starts
unstable when very deep pre-existing shear bands are present, with to move as a result of the rainfall event, which causes a significant rise
respect to shallower ancient shear bands. However, it is worth noting of the water table and the porewater pressures. Thus, different ground
that a significant improvement in the displacement predictions would water levels develop in the upper and lower part of the slope,
result from the adoption of a strain-softening model, especially for representing a trigger for the sliding of the upper mass and the loading
the undisturbed clays, outside the ancient shear band. Also, the use of at the top of the lower mass. The slope deformations, recorded by
3D numerical modelling, implementing hydromechanical coupling means of inclinometers located along sections of the slope different
within the simulation of the transient seepage, from summer to from the analysed one (about 0.5 m displacement in the middle of the
winter, would provide a great enhancement in the prediction of the landslide), have been qualitatively compared with those simulated
slope response to seasonal rainfall infiltration and evapotranspiration through the hydromechanical analysis (reported in Fig. 6c). The
during the year (e.g. Pedone et al. 2016). comparison appears to indicate that the hydromechanical simulation
of the failure mechanism so far presented is reasonable.
Ramina slope
Newbury slope
The stability of a clay slope in the highly urbanized area of Veles
(Macedonia), the Ramina slope along the left bank of the river An instrumented cutting on the A34 Newbury bypass in southern
Vardar, has been studied by Josifovski et al. (2013). The slope is the England has been studied by Rouainia et al. (2009) to investigate the
 G. Elia et al.

Fig. 6. Ramina slope: (a) geological profile; (b) calibration of the SWRC adopted in the simulations; (c) total displacements (maximum 1.17 m);
(d) suctions (maximum 169.7 kN m2).

effects of present and future climate on infrastructure slope stability. function (as summarized in Fig. 7a). Through this modelling,
The slope is formed of London Clay overlying Lambeth Group changes in the slope porewater pressure have been calculated as
deposits and the Upper Chalk (see Smethurst et al. (2006) for response to the weather conditions with time. The modelled
further details) and takes the form of an 8 m high, 28 m wide cut porewater pressure response has been calibrated to comply with
(Fig. 7a). In the modelling, the cutting surface is vegetated with published monitored porewater pressure data from Smethurst et al.
grass, but in reality, over time larger shrubs have started to grow on (2006) through a parametric study on the effect of the assumed
the site (Smethurst et al. 2012). coefficient of permeability. It has been observed that the saturated
The modelling has been undertaken by inputting currently coefficient of permeability required by the hydromechanical model
monitored meteorological variables into a SHETRAN (Ewen et al. to reproduce correctly the measured pore pressures is three orders of
2000) hydrological model of the site. SHETRAN uses weather data magnitude greater than the coefficient of permeability recorded
to estimate the water flow in the soil and the evapotranspiration from within triaxial cell testing and two orders of magnitude greater than
the slope surface. The hydrological model assumes that surface soil that estimated on site from bail-out tests. Field permeability has
deformation is negligible and does not significantly affect the been therefore selected, as reported in Figure 7a, to allow for the
porewater pressures. In addition, future climate data have been macroscopic effects of fissures, sand lenses and other heteroge-
created using a weather generator based on the UK Climate Impacts neous features, as well as to compensate for scale factors. The
Programme (UKCIP02) projections (Kilsby et al. 2007) and applied hydromechanical coupling has been accomplished using a mean
to the same SHETRAN model to estimate porewater pressure effective stress-dependent stiffness, strain-softening Mohr
response to future climate. Once the daily surface porewater Coulomb model. The predicted mid-slope horizontal displacements
pressures have been calculated with SHETRAN, they have been and the porewater pressures at the ground surface for both the
transferred to the surface of a finite-difference geotechnical model present and future climate conditions are plotted in Figure 7b. A
developed in FLAC (Itasca 2011) to simulate the coupled number of readily apparent differences between the two responses
hydromechanical slope response according to the method presented are evident. The first concerns the magnitude of the suctions
in the modelling section above. In this way, the effects of present calculated for future climate, compared with the minor present-day
and future climate on the mechanical response of the cutting have suctions, where it can be seen that the peak summer suctions
been modelled. The van Genuchten model has been used to derive generated by the future climate are always larger than those for the
the soil water retention curve and the hydraulic conductivity present: the future minimum suction of 575 kPa exceeds the
 Slopevegetationatmosphere interaction

Fig. 7. Newbury slope: (a) slope cross-section and soil water retention and hydraulic conductivity function properties for the modelled London Clay; (b)
history of mid-slope pore surface water pressures and horizontal displacements for the present and future climate models.

maximum suction for the present climate. The maximum future stress fluctuations are reflected in larger shrinkswell cycles and
climate suctions in turn reach c. 1175 kPa in the summer of year 9 permanent seasonal downslope movements. The larger deformations
and approach or exceed 1000 kPa in eight of the modelled years. for the future climate lead to more rapid softening of the material at
This increased variation in suction appears to be due to the the toe of the slope and, ultimately, to the failure of the slope in only
combined effect of the more intense drying and the larger 22 years, whereas for the present-day climate failure occurs after
evapotranspiration that will be connected to the higher temperatures 61 years. It can be concluded that the model has the potential to assist
expected in future. In both cases, it can be seen that the suctions in the assessment of the future behaviour of clay cuttings and
totally dissipate during winter and the corresponding larger effective embankments subjected to modified climatic conditions.
 G. Elia et al.

Ruedlingen test site ratio determined by conventional drained or undrained triaxial tests,
consistent with what was observed by Anderson & Sitar (1995), Zhu
Within the context of the multidisciplinary research programme & Anderson (1998) and Casini et al. (2010). Several direct shear
Triggering of Rapid Mass Movements in Steep Terrain (TRAMM, tests have also been performed on statically compacted unsaturated
http://www.cces.ethz.ch/projects/hazri/tramm), a full-scale field test specimens and the results are indicative of a peak internal friction
has been set up on a steep forested slope located on the east-facing angle of 30 (Colombo 2009; Askarinejad et al. 2014). Wetting and
banks of the river Rhine near the village of Ruedlingen, in the drying branches of the SWRC of natural samples have been
Schaffhausen canton, northern Switzerland. Here, several shallow determined using the axis translation technique (Hilf 1956). The air-
landslides occur due to extreme rainfall. To analyse the effects of entry value of the drying branch is 1.7 kPa (Askarinejad &
hydrological events on the stability of unsaturated silty sand slopes, Springman 2015). In all the numerical analyses presented in this
a 38 steep forested slope has been monitored during the application section, the unimodal van Genuchten formula has been used to fit
of artificial intense rainfall events, after carrying out an extensive the data for both wetting and drying branches.
geo-hydro-mechanical characterization of the site. The selected Two-dimensional uncoupled hydraulic finite-element simula-
experimental site, of 35 m length and 7.5 m width, is a small part of tions have been carried out by Askarinejad et al. (2012) and
a slope on the east-facing bank of the river. Fifteen hours of artificial Askarinejad (2013) to simulate the conditions of the pilot site slope
rainfall, with an average intensity of 11 mm h1, has been applied to before the artificial rainfall event. The mechanical features of the
the slope, and this resulted in the triggering of a landslide of 130 m3 unsaturated soils and the reinforcing effects of vegetation have been
volume. The lithostratigraphic profile of the slope consists mainly of implemented. The simulations have been performed using the
Seawater Molasse from the Tethys Ocean and Freshwater Molasse. SEEP/W and SLOPE/W modules of GeoStudio (Geo-Slope
Several boreholes, as well as natural scarps, revealed the presence of International 2004), according to the method presented above.
horizontal layering of sedimentary rocks (fine-grained sandstones The hydraulic boundary conditions imposed during the test at the
and marlstones; Tacher & Locher 2008). Fissures, with openings of ground surface have been applied to the finite-element (SEEP/W)
more than some centimetres in size, have been mapped in the lower model and the pore pressure distribution with time has been
Freshwater Molasse and found to be parallel to the Rhine calculated. The factor of safety of the slope has been determined
(Brnnimann 2011). The bedrock depth is between 0.5 and 5 m, using SLOPE/W, based on the pore pressure distribution exported
according to dynamic probing (DPL) and electrical resistivity from SEEP/W for each time step. The pore pressure distribution has
tomography (Lehmann et al. 2013). The bedrock on the right side of been calculated according to Richards equation for flow in
the field (looking upslope) is found to be shallower than on the left unsaturated soil. The shear strength of the unsaturated soils has
side. The soil can be classified as medium- to low-plasticity silty been simulated in SLOPE/W using the extended form of the Mohr
sand (ML), according to the Unified Soil Classification System Coulomb criterion after Fredlund et al. (1978). Different bedrock
(USCS). The clay fraction increases with depth, from 4% at shallow topographies (derived from the DPL results) and inclinations of
depths to 10% at about 2 m depth. The silt fraction also increases fissures have been examined. The bedrock has been considered to be
with depth, from 25 to 32%, whereas the sand fraction decreases elastic and impermeable, compared with the soil layer. The location
from 67 to 56% (Casini et al. 2010). The results of conventional and sizes of the fissures in the bedrock have been determined by
drained and undrained triaxial tests on saturated samples revealed a monitoring the sequences of the changes in the degree of saturation
critical-state friction angle of CS = 32.5 (Casini et al. 2010, 2013). in the bedrock. The fissures have been implemented in the geometry
Moreover, a series of unsaturated constant shear drained triaxial of the model and have been filled with the overlying soil, with
tests (as proposed by Brand 1981) have been performed by similar hydromechanical properties. Figure 8a shows the geometry
Askarinejad (2013) to replicate the stress path that a soil element of the model for the right-hand sides of the test site (looking
experiences during the rise of porewater pressure in the slope owing upslope) in the case of the shallower bedrock. Fifteen hours of
to rainfall infiltration. The results suggest that the stress ratio attained rainfall, with constant but different intensities for seven zones, have
at failure ( = 34.2) is generally higher than the critical-state stress been applied all over the slope, in accordance with the real rain event

Fig. 8. Ruedlingen test site: (a) geometry of the model adopted in SEEP/W for the right-hand side of the test site and location of seven rain zones based on
the Landslide Triggering Experiment (LTE); (b) enlarged view of the contour lines of porewater pressure in the upper part of the slope 7 h after the start of
rainfall during the LTE (white arrows show the flow direction).
 Slopevegetationatmosphere interaction

Fig. 9. Ruedlingen test site: (a) geometry and hydraulic boundary conditions for the 2D numerical model with exfiltration (2D_DE); (b) displacement
contours of the 2D_DE model after 8.57 h of rainfall (displacements in metres).

in the Landslide Triggering Experiment (LTE, March 2009). The location of the fissures in the bedrock and the hydraulic boundary
seepage analyses show that the water applied in rain zone 7 is conditions are shown in Figure 9a. The maximum dimension of the
conveyed through a fissure in the upper part of the slope to the triangular elements used in the FE mesh is 0.3 m. The bedrock is
horizontal permeable layer and then it exfiltrates from the bedrock assumed to be impermeable, with draining fissures that are located
back into the soil layer, applying an uplifting hydraulic pressure beneath clusters 1 and 2. The lower boundaries of the model have
(Fig. 8b). However, the majority of the infiltrated water drains been set as seepage face lines, where the water can drain out from
through the fissures into the bedrock in the lower part of the slope. the model and the pressure head is kept to zero. Exfiltration has been
The evolution of the factor of safety as a function of time during the modelled at a simulation time of 5 h, by applying 9 kPa constant-
LTE indicates that the FoS decreases with the rain and attains a head boundary conditions at the desired locations in the upper slope
minimum value 2.6 h after the rain stops. The results of the limit (cluster 3). The contours of total displacement before the failure for
equilibrium analyses suggest that the reduction in factor of safety the 2D_DE model are shown in Figure 9b. The approximate shape
occurs more rapidly in the case of the shallower bedrock as a result of the failure surface of the Ruedlingen slope from the landslide
of the exfiltration process described in Figure 8b. triggering experiment is also depicted in this figure. Changes in
In addition, a series of 2D and 3D coupled hydromechanical porewater pressure at depth of 150 cm in cluster 3 are shown in
finite-element analyses have been performed by Askarinejad (2013) Figure 10. The differences between the models with and without
to take the coupled response of the unsaturated soil into account, exfiltration are significant. The results of the 2D_DE model have
according to the method presented above. Only the results of the 2D been found to be in fairly good agreement with the field
coupled analyses are commented on here. The simulations have measurements in terms of the evolution of the porewater pressures
been performed using the finite-element software CODE_BRIGHT (Fig. 10) and location of failure (Fig. 9b). However, the time of the
(Olivella et al. 1996). The constitutive modelling of the coupled failure did not match the in situ observations (i.e. failure occurred
hydromechanical response of the unsaturated soils has been c. 13 h after the start of the rainfall). Nonetheless, the numerical
achieved using the Barcelona Basic Model (BBM) (Alonso et al. analysis confirmed the importance of the hydraulic interaction of the
1990). The time dependence of the stresses and strains of the bedrock with the overlying soil layers, in terms of drainage and
material, owing to viscosity, has been also taken into account for the exfiltration, and showed that these interactions can play a major role
plastic region, based on the formulation suggested by Perzyna in the pattern of pore pressure distributions and in stabilizing or
(1966). Accordingly, the model used for these series of simulations destabilizing the slope.
is an elasticviscoplastic model. The constitutive model parameters
have been defined by comparing the yield function of the BBM with
Idealized heterogeneous slope
the modified yield function proposed by Desai & Zhang (1987) for
unsaturated soils, as specified by Askarinejad (2013). It should be A theoretical slope, characterized by the typical geometry of a
noted that the time dependence of the downslope movements is cutting or embankment for linear infrastructure, has been modelled
more related to the cyclic changes in suction and volumetric water by Arnold & Hicks (2011) to analyse the influence of soil
content induced by the infiltrated rainfall than the viscosity of the heterogeneity on the slope stability during a rainfall event. The so-
material, which has been introduced in these simulations for called random finite-element method (RFEM) has been employed
numerical convenience. The equation proposed by Averjanov (Griffiths & Fenton 1993), where a series of different statistically
(1950) has been used to describe the unsaturated hydraulic possible set-ups (realizations) of material properties have been
conductivity function in CODE_BRIGHT, whereas the saturated simulated. Using a Monte Carlo framework, the data have been
hydraulic conductivity has been determined to be 5 105 m s1 compiled into a statistical output of the factor of safety, leading to a
(Askarinejad et al. 2010). Two analyses have been carried out to description of the reliability of the slope. The model uses Bishops
investigate the effect of exfiltration from the bedrock on the effective stress and involves a saturated or unsaturated seepage
response of the overlying soil layer during the LTE. No exfiltration analysis performed in a stepwise staggered manner (as described
has been simulated in the first model (2D_D), whereas water above); that is, a hydraulic analysis is undertaken at a single time
exfiltrating from the bedrock at two locations has been considered in step, with the results imported into a stability analysis. The strength
the second model (2D_DE). The geometry of the 2D_DE model, the reduction method (Griffiths & Lane 1999) has been used to
 G. Elia et al.

Fig. 10. Ruedlingen test site: porewater

pressure at a depth of 150 cm in in the
upper part of the slope, based on the field
measurements and numerical simulations
with (2D_DE) and without (2D_D)
exfiltration.

determine the FoS at each time step. Partial saturation is included wide eight-noded quadrilateral finite elements. Five soil properties
with both the SWRC and the unsaturated hydraulic conductivity have been selected as spatially varying in the slope: the effective
following the van Genuchten formulation. A linear elasticperfectly cohesion and friction angle, the porosity, the saturated hydraulic
plastic soil model has been adopted in the simulations. Hydraulic conductivity and the inverse of the air-entry pressure. Their mean
conductivity and shear strength parameters can be correlated or values have been set equal to 10 kPa, 25, 0.4, 0.0036 m h1 and
independent and it is possible to incorporate anisotropy. The model 1.0 m1, respectively, with coefficients of variation equal to 0.2,
has been designed to investigate the effect of soil heterogeneity on 0.3, 0.15, 1.75 and 0.9. The five properties have been assumed to
stability, and therefore embodies a number of more simplified be log-normally distributed. Based on test results found in the
features for other aspects; for example, no hardening or softening is literature, a typical correlation matrix has been set up to define the
incorporated in the constitutive model, fluid flow is single phase covariance structure for this boundary value problem. In
with air pressures assumed to be atmospheric, and vegetation and particular, positive correlations have been assumed between
other surface effects are neglected, as are any thermal influences. porosity, saturated hydraulic conductivity and the inverse of the
The model has been implemented in an in-house finite-element air-entry pressure, whereas negative correlations have been
code, coupled to a Random Field (RF) generator (Hicks & Samy assumed between friction angle, saturated hydraulic conductivity
2002). The RF generator produces the different set-ups of material and the inverse of the air-entry pressure. In the case presented by
properties by use of the Local Average Subdivision (LAS) method Arnold & Hicks, the effective cohesion and friction angle have
(Fenton & Vanmarcke 1990), which utilizes as input the mean, been considered constant, with only the SWRC parameters
standard deviation and scale of fluctuation (a measure of spatial varying in space. A homogeneous case, for which the mean
correlation between values at different locations). This method values of the above-mentioned soil properties have been adopted,
generates fields with the same scale of fluctuation in both vertical and a series of different heterogeneous slopes have been
and horizontal directions, which can be post-processed to produce investigated. In the heterogeneous case, a vertical scale of
an anisotropic field by squashing or stretching the isotropic field. fluctuation of 2 m and a horizontal scale of fluctuation of 6 m
Rainfall events in this model are simulated by fixing a zero pore have been assumed. The stability of the models has been
pressure value on the nodes at the embankment surface. This investigated applying a continuous flux of 18.0 mm h1 for 48 h
approach is valid where the rainfall event is sufficient to fully at the ground surface, representative of a heavy rainfall event,
saturate the upper layers of the slope. Any additional rain is assumed followed by 1.0 mm h1 for about 550 h.
to run off, and is not included in the simulation. Care must be taken For a given factor of safety, the reliability of the slope is equal to
not to simulate excessive inflow, and it is possible to include mixed R = 1 Pf = 1 Nf/N, where Pf is the probability of failure, N is the
boundary conditions to ensure this automatically. total number of realizations and Nf is the number of realizations in
In the study by Arnold & Hicks (2011), a 45 slope of 5 m which the slope fails. A sufficient number of realizations must be
height resting on a 5 m basement layer has been analysed in two considered to produce a converged result, which increases with the
dimensions, assuming plane strain conditions and using 0.25 m higher reliability levels. In this case, 300 realizations have been

Fig. 11. Ideal slope case study: Bishops

stress for (a) homogeneous slope and (b)
heterogeneous slope.
 Slopevegetationatmosphere interaction

found to provide a satisfactory response. The FoS calculated prior to permeability is the single most influential parameter in the
rainfall, without taking into account suction, is 1.39, which infiltration and evapotranspiration processes in the slope. This
increases to 1.75 taking into account suction stresses. The FoS is finding highlights the importance of suitable methods to measure
observed to fall during and after rainfall as the water front propagates this property in situ. In conjunction with the hydraulic modelling,
through the soil. In the homogeneous case the FoS falls from 1.75 to thermal coupling or mechanical coupling can be included in more
1.58 after c. 500 h. Figure 11 presents a comparison of the Bishops sophisticated simulations of slopeatmosphere interaction pro-
stress contours for the homogeneous and a single realization blems. When considering the hydromechanical coupling, slope
heterogeneous slope, with the corresponding potential failure progressive failure can be predicted. For strain-softening soils,
mechanisms shown by the deformed mesh. The results are presented appropriate constitutive models implementing this aspect of the
after 48 h, immediately after the rainfall event. A reduction in shear behaviour have to be used. Moreover, the soil permeability
Bishops stress can be observed close to the surface in both cases can significantly affect the rate of progressive failure, as higher
owing to increased levels of saturation (and reduction in suction). In permeabilities allow for a greater seasonal porewater pressure
the heterogeneous case, the FoS is reduced from 1.90 to 1.42 over change at depth, thus leading to larger seasonal displacements and
the same time period. The failure mode is also seen to be different, increased strain-softening. In addition, the effects of desiccation
with a shallower failure surface noted for the heterogeneous case. It cracking and corresponding increase of permeability are not fully
is also observed that the mean FoS of the heterogeneous slope is considered in most of the numerical approaches adopted in practice.
lower than that of the homogeneous case; that is, a reliability of 50% Therefore, the development and practical application of advanced
is c. 90% of the homogeneous FoS and a 99.75% reliability gives an numerical tools for the prediction of swellingshrinkage cracking at
FoS that is c. 70% of the homogeneous case, thus indicating how the slope surface represents a future research challenge.
failure can be driven by the spatial variability of the effective shear Finally, the effects of temperature are usually not fully
strength parameters within the slope. incorporated in the modelling. The temperature is not only a
fundamental ingredient to analyse climatic processes at the interface
between the slope and the atmosphere, but it can also affect
Conclusions and research perspectives
the mechanical behaviour of soils. Therefore, thermo-hydro-
The stability analyses performed taking into account the interaction mechanical coupling represents the most advanced numerical
phenomena between the soil and the environment through a approach to solve the full set of balance equations. Coupled
multidisciplinary approach have shown that changes in climate can thermo-hydro-mechanical modelling has been successfully used for
affect the behaviour and stability of artificial and natural slopes, and the analysis of deep geological storage and disposal of high-level
engineers are required to assess their impact to minimize risk and nuclear waste (Gens 2010), but its application to slope stability
financial costs. A range of approaches, from simple limit problems is much less common (Nishimura et al. 2009) and does
equilibrium solutions to advanced finite-element and finite- not appear to be well established yet. Nevertheless, its use should be
difference analyses, have been proposed in the literature to assess promoted further and it is hoped that, with the increase of
slopeatmosphere interaction and the related slope stability. In computational power and resources, coupled thermo-hydro-mech-
particular, the research related to thermo-hydro-mechanical coup- anical simulations of slopeatmosphere interaction will be more
ling processes in soils has, nowadays, provided advanced modelling widely adopted in practice.
tools to take account of climate effects in slope stability and
serviceability analyses. However, these tools are seldom used in
practice, given their mathematical complexity and the in-depth
Acknowledgements This paper is an output of Working Group 1 of EU
COST Action TU1202 Impacts of climate change on engineered slopes for
knowledge of soil constitutive behaviour required for their infrastructure. TU1202 comprises four working groups: WG1, Slope numerical
application. The paper contributes to the dissemination of these modelling; WG2, Field experimentation and monitoring; WG3, Soil/vegetation/
climate interactions; WG4, Slope risk assessment. Outputs from each working
modelling approaches by providing a schematic description of all
group have been submitted to the Quarterly Journal of Engineering Geology and
the alternative procedures, accounting step by step for all the Hydrogeology and are intended to be read as a thematic set.
ingredients useful to represent the different slope conditions and
processes. A review of the available methods to model the slope
behaviour is presented, which allows, through the comparison of the Funding The authors gratefully acknowledge the funding for COST Action
TU1202 through the EU Horizon 2020 program, without which these outputs
methods and of their results, identification of their benefits and would not have been possible.
limitations and the different input requirements. In this perspective,
the paper outlines the work by some of the participants in the COST Scientific editing by Nick Koor; David Taborda
Action TU1202 concerning the effects of climate and vegetation on
slope deformation and stability performance. In all cases the review References
emphasizes the importance, for any reliable modelling of the slope Abramson, L.W., Lee, T.S., Sharma, S. & Boyce, G.M. 1996. Slope Stability and
behaviour, of simulating the fundamental soil hydromechanical Stabilization Methods. Wiley, New York.
properties, as derived from in situ monitoring and/or laboratory Alexander, D. 1986. Landslide damage to buildings. Environmental Geology and
Water Sciences, 8, 147151.
tests. Allen, R.G., Pereira, L.S., Raes, D. & Smith, M. 1998. Crop evapo-transpiration
The scenarios of slope conditions that have been analysed are (guidelines for computing crop water requirements). FAO Irrigation and
large in number and the classical soil mechanics formulation has Drainage Paper, 56.
been progressively generalized to account for a broader range of Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially
saturated soils. Gotechnique, 40, 405430.
phenomena and soil behaviour, including thermohydraulic, hydro- Alonso, E.E., Pereira, J.M., Vaunat, J. & Olivella, S. 2010. A microstructurally
mechanical and thermo-hydro-mechanical coupling. A common based effective stress for unsaturated soils. Gotechnique, 60, 913925.
key aspect of all the numerical approaches described in the paper is Amarasiri, A.L., Kodikara, J.K. & Costa, S. 2011. Numerical modelling of
desiccation cracking. International Journal of Numerical and Analytical
the implementation of the climate effects, from the simplest way of Methods in Geomechanics, 35, 8296.
simulating saturated conditions with suctions calibrated to represent Anderson, S.A. & Sitar, N. 1995. Analysis of rainfall-induced debris flow.
real or monitored porewater pressures, to the simulation of the Journal of Geotechnical Engineering, ASCE, 121, 524552.
hydraulic behaviour of partially saturated soils (through SWRC and Arnold, P. & Hicks, M.A. 2011. A stochastic approach to rainfall-induced slope
failure. In: Vogt, N., Schuppener, B., Straub, D. & Bru, G. (eds) Third
k(s)) and the implementation of rainfall and evapotranspiration. International Symposium on Geotechnical Safety and Risk (ISGSR). Munich,
From these hydraulic analyses it is usually found that the soil Germany, Bundesanstalt fr Wasserbau, Munich, 107115.
 G. Elia et al.

Askarinejad, A. 2013. Failure mechanisms in unsaturated silty sand slopes Cotecchia, F., Pedone, G., Bottiglieri, O., Santaloia, F. & Vitone, C. 2014. Slope
triggered by rainfall. PhD thesis, ETH Zurich. atmosphere interaction in a tectonized clayey slope: a case study. Italian
Askarinejad, A. & Springman, S.M. 2015. Centrifuge modelling of the effects of Geotechnical Journal, 1, 3461.
vegetation on the responses of a silty sand slope subjected to rainfall. 14th Cotecchia, F., Santaloia, F., Pedone, G., Vitone, C., Lollino, P., Mitaritonna, G.
International Conference of the International Association for Computer 2015. Diagnosis of slow landslides affecting tectonised clayey slopes
Methods and Advances in Geomechanics (IACMAG), Kyoto, Japan, (Southern Appennines). In: Lollino, G. (ed) Engineering Geology for
13391344. Society and Territory, Springer International Publishing, Switzerland,
Askarinejad, A., Casini, F., Kienzler, P. & Springman, S.M. 2010. Comparison of Vol. 2, 10291032.
the in situ and laboratory water retention curves for a silty sand. 5th Cruden, D.M. & Varnes, D.J. 1996. Landslide Types and Processes. In: Turner,
International Conference of Unsaturated Soils (UNSAT), Barcelona, Spain, A.K. & Schuster, R.L. (eds) Landslide-Investigation and Mitigation.
423428. Transportation Research Board, Special Report, 247, 3675.
Askarinejad, A., Casini, F., Bischof, P., Beck, A. & Springman, S.M. 2012. Delage, P., Sultan, N. & Cui, Y.J. 2000. On the thermal consolidation of Boom
Rainfall induced instabilities: a field experiment on a silty sand slope in clay. Canadian Geotechnical Journal, 37, 343354.
northern Switzerland. Rivista Italiana di Geotecnica, 3, 5071. De Marsily, G. 1986. Quantitative Hydrogeology: Groundwater Hydrology for
Askarinejad, A., Beck, A. & Springman, S.M. 2014. Scaling law of static Engineers. Academic Press, New York.
liquefaction mechanism in geo-centrifuge and corresponding hydro-mechan- Desai, C.S. & Zhang, D. 1987. Viscoplastic model for geologic materials with
ical characterisation of an unsaturated silty sand having a viscous pore fluid. generalized flow rule. International Journal for Numerical and Analytical
Canadian Geotechnical Journal, 52, 708720. Methods in Geomechanics, 11, 603620.
Averjanov, S.F. 1950. About permeability of subsurface soils in case of Dijkstra, T. & Dixon, N. 2010. Climate change and slope stability in the UK:
incomplete saturation. English Collection, 7, 1921. challenges and approaches. Quarterly Journal of Engineering Geology and
Belardi, A., Colleselli, F. et al.et al. 2014. Studio del comportamento idraulico Hydrogeology, 43, 371385, https://doi.org/10.1144/1470-9236/09-036
degli argini del alto Bacino del Po. In: Convegno XXV CNGAGI, Baveno, Ewen, J., Parkin, G. & OConnell, P.E. 2000. SHETRAN: Distributed River
Italy, 771779. Basin Flow and Transport Modelling System. ASCE Journal of Hydrologic
Bishop, A.W. 1955. The use of the slip circle in the stability analysis of slopes. Engineering, 5, 250258.
Gotechnique, 5, 717. Faisal, H.A. & Normaniza, O. 2008. Shear strength of soil containing vegetation
Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad, 39, 859863. roots. Soils and Foundations, 48, 587596.
Blight, G.E. 1997. Interactions between the atmosphere and the Earth. Fenton, G.A. & Vanmarcke, E.H. 1990. Simulation of random fields via local
Gotechnique, 47, 715767. average subdivision. Journal of Engineering Mechanics, 116, 17331749.
Booth, A. 2015. Impacts of desiccation cracking and climate change on highway Fleming, R.W. & Taylor, F.A. 1980. Estimating the Costs of Landslide Damage
cutting hydrology. PhD thesis, Loughborough University. in the United States. US Geological Survey Circular, 832.
Brand, E.W. 1981. Some thoughts on rainfall induced slope failures. In: Fredlund, D.G. & Morgenstern, N.R. 1977. Stress state variables for unsaturated
Proceedings of the 10th International Conference on Soil Mechanics and soils. Journal of the Geotechnical Engineering Division, ASCE, 103,
Foundation Engineering, 373376. 447466.
Brinkgreve, R.B.J., Al-Khoury, R. & Van Esch, J.M. 2003. Plaxflow manual Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils.
(Version 1). Balkema, Rotterdam. Wiley, New York.
Brnnimann, C. 2011. Effects of Groundwater on Landslide Triggering, Nr. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of
5236. PhD thesis, EPFL, Lausanne. unsaturated soils. Canadian Geotechnical Journal, 15, 313321.
Bui, H.H., Nguyen, G.D., Kodikara, J. & Sanchez, M. 2015. Soil cracking Fredlund, D.G., Houston, S.L., Nguyen, Q. & Fredlund, M.D. 2010. Moisture
modelling using the mesh-free SPH method. In: 12th ANZ Conference on movement through cracked clay soil profiles. Geotechnical and Geological
Geomechanics, Wellington, New Zealand. Engineering, 28, 865888.
Cafaro, F. & Cotecchia, F. 2015. Influence of the mechanical properties of Fredlund, D.G., Rahardjo, H. & Fredlund, M.D. 2012. Unsaturated Soil
consolidated clays on their water retention curve. Italian Geotechnical Mechanics in Engineering Practice. Wiley, New York.
Journal, 2, 1329. Freeze, A. & Cherry, J. 1979. Groundwater. PrenticeHall, Englewood Cliffs,
Calabresi, G., Colleselli, F. et al. 2013. A research study of the hydraulic NJ.
behaviour of the Po river embankments. Canadian Geotechnical Journal, 50, Gallipoli, D., Wheeler, S. & Karstunen, M. 2003. Modelling the variation of
947960. degree of saturation in a deformable unsaturated soil. Gotechnique, 53,
Calvello, M., Cascini, L. & Sorbino, G. 2007. A numerical procedure for 105112.
predicting rainfall-induced movements of active landslides along pre-existing Gens, A. 2010. Soilenvironment interactions in geotechnical engineering.
slip surfaces. International Journal for Numerical and Analytical Methods in Gotechnique, 60, 374.
Geomechanics, 32, 327351. Gens, A., Snchez, M. & Sheng, D. 2006. On constitutive modelling of
Cascini, L., Bonnard, C., Corominas, J., Jibson, R. & Montero-Olarte, J. 2005. unsaturated soils. Acta Geotechnica, 1, 137147.
Landslide hazard and risk zoning for urban planning and developmentState Gentile, F., Elia, G. & Elia, R. 2010. Analysis of the stability of slopes reinforced
of the art report. In: Hungr, O., Fell, R., Couture, R. & Eberthardt, E. (eds) by roots. In: Brebbia, C.A. & Carpi, A. (eds) Design and Nature V:
Proceedings of the International Conference on Landslide Risk Management. Comparing Design in Nature with Science and Engineering. WIT
Taylor & Francis, London, 199235. Transactions on Ecology and the Environment, 138, 189200.
Cascini, L., Calvello, M. & Grimaldi, G.M. 2010a. Groundwater modeling for Geo-Slope International 2004. GeoStudio, Users guide. Geo-Slope
the analysis of active slow-moving landslides. Journal of Geotechnical and International, Calgary, AB.
Geoenvironmental Engineering, ASCE, 136, 12201230. Glendinning, S., Loveridge, F., Star-Keede, R.E., Bransby, M.F. & Hughes, P.N.
Cascini, L., Cuomo, S., Pastor, M. & Sorbino, G. 2010b. Modeling of rainfall- 2009. The role of vegetation in the sustainability of infrastructure slopes.
induced shallow landslides of the flow-type. Journal of Geotechnical and Proceedings of the Institution of Civil Engineers: Engineering Sustainability,
Geoenvironmental Engineering, ASCE, 136, 8598. 162, 101110.
Casini, F., Jommi, C. & Springman, S.M. 2010. A laboratory investigation on an Greenwood, J.R. 2006. SLIP4EXa program for routine slope stability analysis
undisturbed silty sand from a slope prone to landsliding. Granular Matter, 12, to include the effects of vegetation, reinforcement and hydrological changes.
303316. Geotechnical and Geological Engineering, 24, 449465.
Casini, F., Serri, V. & Springman, S.M. 2013. Hydromechanical behaviour of a Griffiths, D. & Fenton, G.A. 1993. Seepage beneath water retaining structures
silty sand from a steep slope triggered by artificial rainfall: from unsaturated to founded on spatially random soil. Gotechnique, 43, 577587.
saturated conditions. Canadian Geotechnical Journal, 50, 2840. Griffiths, D. & Lane, P. 1999. Slope stability analysis by finite elements.
Chiu, C.F. & Ng, C.W.W. 2003. A state-dependent elasto-plastic model for Gotechnique, 49, 387403.
saturated and unsaturated soils. Gotechnique, 53, 809829. Guzzetti, F., Reichenbach, P., Cardinali, M., Ardizzone, F. & Galli, M. 2003. The
Chok, Y.H., Kaggwa, W.S., Jaksa, M.B. & Griffiths, D.V. 2004. Modelling the impact of landslides in the Umbria region, central Italy. Natural Hazards and
effect of vegetation on stability of slopes. In: Proceedings of the 9th Australia Earth System Sciences, 3, 469486.
New Zealand Conference on Geomechanics, Auckland, 391397. Hallett, P.D. & Newson, T.A. 2005. Describing soil crack formation using
Colombo, L. 2009. Large shear box for analysing strength mobilisation in elasticplastic fracture mechanics. European Journal of Soil Science, 56,
unsaturated conditions. Masters thesis, Politecnico di Milano. 3138.
Cotecchia, F., Santaloia, F., Bottiglieri, O. & Monterisi, L. 2008. Landslides in Hicks, M.A. & Samy, K. 2002. Influence of heterogeneity on undrained clay
stiff clay slopes along the Adriatic coast (Central Italy). In: Proceedings of the slope stability. Quarterly Journal of Engineering Geology and Hydrogeology,
X International Symposium on Landslide, XIAN 2008, 2, 15251531. 35, 4149, https://doi.org/10.1144/1470-9236/09-036
Cotecchia, F., Santaloia, F., Lollino, P., Vitone, C. & Mitaritonna, G. 2010. Hilf, J.W. 1956. An investigation of pore-water pressure in compacted cohesive
Deterministic landslide hazard assessment at regional scale. In: GeoFlorida soils. US Department of the Interior, Bureau of Reclamation, Design and
2010: Advances in Analysis, Modeling and Design, West Palm Beach, Construction Division, Denver, CO.
Florida. Geotechnical Special Publication, ASCE, 31303139. Houlsby, G.T. 1997. The work input to an unsaturated granular material.
Cotecchia, F., Santaloia, F. & Lollino, P. 2011. Slow landsliding as effect of Gotechnique, 47, 193196.
hydro-mechanical coupled processes: examples of resulting damages and Hungr, O., Evans, S.G. & Hazzard, J. 1999. Magnitude and frequency of rock
mitigation strategies. In: Proceedings of the 2nd International Workshop on falls and rock slides along the main transportation corridors of southwestern
Landslides, 2829 September, Naples, Italy. British Columbia. Canadian Geotechnical Journal, 36, 224238.
 Slopevegetationatmosphere interaction

Hungr, O., Leroueil, S. & Picarelli, L. 2014. The Varnes classification of Monteith, J.L. 1965. Evaporation and environment. Symposia of the Society for
landslide types, an update. Landslides, 11, 167194. Experimental Biology, 19, 205224.
Itasca 2011. FlacFast Lagrangian Analysis of ContinuaVersion 7.0 Users Muslelak, G. & Sliwa, T. 2012. Fracturing of clay during drying: Modelling and
Guide. Itasca, Minneapolis, MN. numerical simulation. Transport in Porous Media, 95, 465481.
Jakob, M. & Hungr, O. 2005. Debris-flow Hazards and Related Phenomena. Ng, C.W.W. & Shi, Q. 1998. A numerical investigation of the stability of
Springer, Berlin. unsaturated soil slopes subjected to transient seepage. Computers and
Ji, J., Kokutse, N., Genet, M., Fourcaud, T. & Zhang, Z. 2012. Effect of spatial Geotechnics, 22, 128.
variation of tree root characteristics on slope stability. A case study on Black Nishimura, S., Martin, C.J., Jardine, R.J. & Fenton, C.H. 2009. A new approach
Locust (Robinia pseudoacacia) and Arborvitae (Platycladus orientalis) stands for assessing geothermal response to climate change in permafrost regions.
on the Loess Plateau, China. Catena, 92, 139154. Gotechnique, 59, 213227.
Josifovski, J., Gjorgjevski, S. & Susinov, B. 2013. Ramina Landslide from a Nuth, M. & Laloui, L. 2008a. Advances in modelling hysteretic water retention
Natural Hazard to Remediation. Arbanas, S.M. & Arbanas, Z., (eds) In: 1st curve in deformable soils. Computers and Geotechnics, 35, 835844.
Regional Symposium on Landslides in the AdriaticBalkan Region with 3rd Nuth, M. & Laloui, L. 2008b. Effective stress concept in unsaturated soils:
Workshop of the JapaneseCroatian Project on Risk Identification and Land- Clarification and validation of unified framework. International Journal for
Use Planning for Disaster Mitigation of Landslides and Floods in Croatia, Numerical and Analytical Methods in Geomechanics, 32, 771801.
Zagreb, Croatia, Emergency Management Office, Zagreb. Nyambayo, V.P. & Potts, D.M. 2010. Numerical simulation of evapotranspiration
Jovanovski, M., Markovski, V. & Spirovski, D. 2003. Geotechnical report from using a root water uptake model. Computers and Geotechnics, 37, 175186.
geotechnical investigations on the zone of landslide RaminaVeles after Nyambayo, V.P., Potts, D.M. & Addenbrooke, T.I. 2004. The influence
reactivation in 2002. DGR Geotehnika Pelagonija, Skopje. of permeability on the stability of embankments experiencing seasonal
Khalili, N. & Khabbaz, M.H. 1998. A unique relationship for for the cyclic pore water pressure changes. In: Proceedings, Skempton Conference,
determination of the shear strength of unsaturated soils. Gotechnique, 48, Advances in Geotechnical Engineering, London, UK, 2, 899910.
681687. OBrien, A.S., Ellis, E.A. & Russell, D. 2004. Old railway embankment clay fill
Khalili, N. & Zargarbashi, S. 2010. Influence of hydraulic hysteresis on effective laboratory experiments, numerical modelling and field behaviour. In:
stress in unsaturated soils. Gotechnique, 60, 729734. Proceedings, Skempton Conference, Advances in Geotechnical Engineering.
Khalili, N., Habte, M.A. & Zargarbashi, S. 2008. A fully coupled flow London, UK, 2, 911921.
deformation model for cyclic analysis of unsaturated soils including hydraulic Olivella, S. 1995. Nonisothermal multiphase flow of brine and gas through saline
and mechanical hysteresis. Computers and Geotechnics, 35, 872889. media. PhD thesis, Universitat Politecnica de Catalunya (UPC), Barcelona.
Kilsby, C.G., Jones, P.D. et al. 2007. A daily weather generator for use in climate Olivella, S., Carrera, J., Gens, A. & Alonso, E.E. 1994. Nonisothermal
change studies. Environmental Modelling and Software, 22, 17051719. multiphase flow of brine and gas through saline media. Transport in Porous
Klose, M., Damm, B. & Terhorst, B. 2014. Landslide cost modeling for Media, 15, 271293.
transportation infrastructures: a methodological approach. Landslides, https:// Olivella, S., Gens, A., Carrera, J. & Alonso, E.E. 1996. Numerical formulation
doi.org/10.1007/s10346-014-0481-1 for simulator (CODE_BRIGHT) for coupled analysis in saline media.
Koda, E. & Osinski, P. 2011. Slope erosion control with the use of fly-ash and Engineering Computations, 13, 87112.
sewage sludge. Annals of Warsaw University of Life Sciences SGGW, Land Pedone, G., Tsiampousi, A., Cotecchia, F. & Zdravkovic, L. 2016. Effects of soil
Reclamation, 43, 101111. vegetationatmosphere interaction on the stability of a clay slope: a case study.
Konrad, J.-M. & Ayad, R. 1997. An idealized framework for the analysis of Delage, P., Cui, Y.-J., Ghabezloo, S., Pereira, J.-M. & Tang, A.-M. (eds) In:
cohesive soils undergoing desiccation. Canadian Geotechnical Journal, 34, Proceedings of the 3rd European Conference on Unsaturated Soils (E-UNSAT
477488. 2016), Paris, https://doi.org/10.1051/e3sconf/20160915002
Kovacevic, K., Potts, D.M. & Vaughan, P.R. 2001. Progressive failure in clay Pedroso, D.M. & Williams, D.J. 2010. A novel approach for modelling soil-water
embankment due to seasonal climate changes. In: Proceedings, 5th characteristic curves with hysteresis. Computers and Geotechnics, 37, 374380.
International Conference on Soil Mechanics and Geotechnical Engineering, Penman, H.L. 1948. Natural evaporation from open water, bare soil and grass.
Istanbul, 3, 21272130. Proceedings of the Royal Society of London, Series A, 193, 120145.
Krahn, J. 2003. The 2001 R.M. Hardy Lecture: the limits of limit equilibrium Pereira, J.M., Wong, H., Dubujet, P. & Dangla, P. 2005. Adaptation of existing
analyses. Canadian Geotechnical Journal, 40, 643660. behaviour models to unsaturated states: Application to CJS model.
Lachenbruch, A.H. 1961. Depth and spacing of tension cracks. Journal of International Journal for Numerical and Analytical Methods in
Geophysical Research, 66, 42734292. Geomechanics, 29, 11271155.
Lambe, T.W. & Whitman, R.V. 1969. Soil Mechanics. Wiley, New York. Peron, H., Delenne, J.Y., Laloui, L. & El Youssoufi, M.S. 2009. Discrete element
Lees, A.S., MacDonald, G.J., Sheerman-Chase, A. & Schmidt, F. 2013. Seasonal modelling of drying shrinkage and cracking of soils. Computers and
slope movements in an old clay fill embankment dam. Canadian Geotechnical Geotechnics, 36, 6169.
Journal, 50, 503520. Perzyna, P. 1966. Fundamental problems in viscoplasticity. Advances in Applied
Lehmann, P., Gambazzi, F. et al. 2013. Evolution of wetting patterns preceding Mechanics, 9, 244368.
hydro-mechanically induced landslide inferred from electrical resistivity Picarelli, L., Urciuoli, G., Ramondini, M. & Comegna, L. 2004. Main features of
tomography. Water Resources Research, https://doi.org/10.1002/ mudslides in tectonised highly fissured clay shales. Landslides, 2, 1530.
2013WR014560 PLAXIS 2D 2012. Reference Manual. Plaxis, Delft.
Leong, E.C. & Rahardjo, H. 1997a. Review of soil-water characteristic Potts, D.M. & Zdravkovic, L. 1999. Finite Element Analysis in Geotechnical
curve equations. Journal of Geotechnical Engineering, ASCE, 123, Engineering: Theory. Thomas Telford, London.
11061117. Potts, D.M. & Zdravkovic, L. 2001. Finite Element Analysis in Geotechnical
Leong, E.C. & Rahardjo, H. 1997b. Permeability functions for unsaturated soils. Engineering: Application. Thomas Telford, London.
Journal of Geotechnical Engineering, ASCE, 123, 11181126. Pouya, A., Vo, T.-D., Hemmati, S. & Nguyen, V.L. 2015. Numerical modelling
Li, J.H., Li, L., Chen, R. & Li, D.Q. 2016. Cracking and vertical preferential flow of hydric cracking of an unsaturated soil. In: SEC 2015 International
through landfill clay liners. Engineering Geology, 206, 3341. Symposium, Marne-la-Valle, France, 169178.
Li, K.S. & Lumb, P. 1987. Probabilistic design of slopes. Canadian Geotechnical Rajeev, P., Chan, D. & Kodikara, J. 2012. Groundatmosphere interaction
Journal, 24, 520535. modelling for long-term prediction of soil moisture and temperature.
Li, X.S. 2005. Modelling of hysteresis response for arbitrary wetting/drying Canadian Geotechnical Journal, 49, 10591073.
paths. Computers and Geotechnics, 32, 133137. Rezaur, R.B., Rahardjo, H. & Leong, E.C. 2002. Spatial and temporal variability
Lollino, P., Elia, G., Cotecchia, F. & Mitaritonna, G. 2010. Analysis of landslide of pore-water pressures in residual soil slopes in a tropical climate. Earth
reactivation mechanisms in Daunia clay slopes by means of limit equilibrium Surface Processes and Landforms, 27, 317338.
and FEM methods. In: GeoFlorida 2010: Advances in Analysis, Modeling and Richards, L.A. 1931. Capillary conduction of liquids through porous media.
Design. Geotechnical Special Publication, ASCE, 31553164. Physics, 1, 318333.
Lollino, P., Cotecchia, F., Elia, G., Mitaritonna, G. & Santaloia, F. 2016. Rojas, E. 2008. Equivalent stress equation for unsaturated soils. I: Equivalent
Interpretation of landslide mechanisms based on numerical modelling: two stress. International Journal of Geomechanics, 8, 285290.
case-histories. European Journal of Environmental and Civil Engineering, 20, Rouainia, M., Davies, O., OBrien, A. & Glendinning, S. 2009. Numerical
10321053. modelling of climate effects on slope stability. Engineering Sustainability,
Lu, N. & Godt, J.W. 2013. Hillslope Hydrology and Stability. Cambridge 162, 8189.
University Press, Cambridge. Russell, A.R. & Khalili, N. 2006. A unified bounding surface plasticity model for
Lu, N. & Likos, W.J. 2004. Unsaturated Soil Mechanics. Wiley, New York. unsaturated soils. International Journal for Numerical and Analytical
Midttmme, K. & Roaldset, E. 1999. Thermal conductivity of sedimentary rocks: Methods in Geomechanics, 30, 181212.
uncertainties in measurement and modelling. In: Aplin, A.C., Fleet, A.J. & Sanchez, M., Manzoli, O. & Guimaraes, L. 2014. Modelling 3-D desiccation soil
Macquaker, J.H.S. (eds) Muds and Mudstones: Physical and Fluid Flow crack networks using a mesh fragmentation technique. Computers and
Properties. Geological Society, London, Special Publications, 158, 4560, Geotechnics, 62, 2739.
https://doi.org/10.1144/GSL.SP.1999.158.01.04 Schmertmann, J.H. 2006. Estimating slope stability reduction due to rain
Mohammadnejad, T. & Khoei, A.R. 2013. Hydro-mechanical modeling of infiltration mounding. Journal of Geotechnical and Geoenvironmental
cohesive crack propagation in multiphase porous media using the extended Engineering, ASCE, 132, 12191228.
finite element method. International Journal for Numerical and Analytical Scott, P.S., Farquhar, G.J. & Kouwen, N. 1983. Hysteretic effects on net
Methods in Geomechanics, 37, 12471279. infiltration. Advances in Infiltration, 11, 163170.
 G. Elia et al.

Sheng, D., Fredlund, D.G. & Gens, A. 2008. A new modelling approach for Turc, L. 1954. Le bilan deau des sols. Relation entre la prcipitation,
unsaturated soils using independent stress variables. Canadian Geotechnical lvaporation et lcoulement. Annales Agronomiques, 5, 491569.
Journal, 45, 511534. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for
Smethurst, J.A., Clarke, D. & Powrie, W. 2006. Seasonal changes in pore water the prediction of shear strength with respect to soil suction. Canadian
pressure in a grass-covered cut slope in London Clay. Gotechnique, 56, Geotechnical Journal, 33, 379392.
527537. Van Esch, J.M. 2012. Modeling groundwater flow through embankments for
Smethurst, J.A., Clarke, D. & Powrie, W. 2012. Factors controlling the seasonal climate change impact assessment. In: Proceedings, 19th International
variation in soil water content and pore water pressures within a lightly Conference on Computational Methods in Water Resources (CMWR 2012)
vegetated clay slope. Gotechnique, 62, 429446. University of Illinois at UrbanaChampaign, USA.
Smethurst, J.A., Briggs, K.M., Powrie, W., Ridley, A. & Butcher, D.J.E. 2015. Van Esch, J.M., Van, M.A., Hendriks, R.F.A., Van den Akker, J.J.H., Forster, U. &
Mechanical and hydrological impacts of tree removal on a clay fill railway Van Etten, R.J.G. 2007. Geo-hydrologic design procedure for peat dykes under
embankment. Gotechnique, 65, 869882. drying and wetting conditions. In: Proceedings, 10th International Symposium
Smith, P.G.C. 2003. Numerical analysis of infiltration into partially saturated on Numerical Methods in Geomechanics (NUMOG X), Rhodes, Greece.
soil slopes. PhD thesis, Imperial College London. Van Esch, J.M., Sellmeijer, J.B. & Stolle, D. 2013. Modeling transient
Smith, P.G.C., Potts, D.M. & Addenbrooke, T.I. 2008. A precipitation boundary groundwater flow and piping under dikes and dams. In: Pande, G.N. &
condition for finite element analysis. In: Proceedings of the 1st European Pietruszczak, S. (eds) Proceedings, Third International Symposium on
Conference on Unsaturated Soils, 773778. Computational Geomechanics (ComGeo III). Taylor & Francis, London.
Spencer, E. 1967. A method of analysis of the stability of embankments, van Genuchten, M.T. 1980. A closed form equation for predicting the hydraulic
assuming parallel interslice forces. Gotechnique, 17, 1126. conductivity of unsaturated soils. Soil Science Society of America Journal, 44,
Stirling, R. 2014. Multiphase modelling of desiccation cracking in compacted 892898.
soils. PhD thesis, Newcastle University. Waldron, L.J. 1977. The shear resistance of root-permeated homogeneous and
Sun, D.A., Sheng, D., Cui, H.B. & Sloan, S.W. 2007. A density-dependent stratified soil. Soil Science Society of America Journal, 41, 843849.
elastoplastic hydro-mechanical model for unsaturated compacted soils. Wheeler, S.J. 1996. Inclusion of specific water volume within an elasto-plastic
International Journal for Numerical and Analytical Methods in model for unsaturated soil. Canadian Geotechnical Journal, 33, 4257.
Geomechanics, 31, 12571279. Wheeler, S.J. & Sivakumar, V. 1995. An elasto-plastic critical state framework
Tacher, L. & Locher, D. 2008. Internal TRAMM report: Geological for unsaturated soil. Gotechnique, 45, 553.
Characterization of the Buchberg field site (Ruedlingen). EPFL, Lausanne. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic
Take, W.A. & Bolton, M.D. 2011. Seasonal ratcheting and softening in clay hysteresis and stressstrain behaviour in unsaturated soils. Gotechnique, 53,
slopes, leading to first-time failure. Gotechnique, 61, 757769. 4154.
Terzaghi, K. 1936. The shear resistance of saturated soils. In: Proceedings, 1st Wong, T.T., Fredlund, D.G. & Krahn, J. 1998. A numerical study of coupled
International Conference on Soil Mechanics and Foundation Engineering, consolidation in unsaturated soils. Canadian Geotechnical Journal, 35, 926937.
Cambridge, MA, 1, 5456. Woodside, W. & Messmer, J.H. 1961. Thermal conductivity of porous media
Terzaghi, K. 1950. Mechanisms of landslides. Geological Society of America, I. Unconsolidated sands. Journal of Applied Physics, 32, 16881699.
Berkley Volume, 83123. Wu, T.H., McKinnell, W.P. & Swanston, D.N. 1979. Strength of tree root and
Thornthwaite, C.W. 1948. An approach toward a rational classification of climate. landslides on Prince of Wales Island, Alaska. Canadian Geotechnical
Geographical Review, 38, 5594. Journal, 16, 1933.
Tommasi, P., Boldini, D., Caldarini, G. & Coli, N. 2013. Influence of infiltration Wu, T.H., Beal, P.E. & Lan, C. 1988. In situ shear test of soilroot system.
on the periodic re-activation of slow movements in an overconsolidated clay Journal of Geotechnical Engineering, 114, 13511375.
slope. Canadian Geotechnical Journal, 50, 5467. Zhu, J.H. & Anderson, S.A. 1998. Determination of shear strength of Hawaiian
Tsaparas, I., Rahardjo, H., Toll, D.G. & Leong, E.C. 2002. Controlling parameters residual soil subjected to rainfall-induced landslides. Gotechnique, 48,
for rainfall induced landslides. Computers and Geotechnics, 29, 127. 7382.
Tsiampousi, A., Zdravkovic, L. & Potts, D.M. 2013a. A three-dimensional Zienkiewicz, O.C. & Taylor, R.L. 1989. The Finite Element Method. McGraw
hysteretic soil-water retention curve. Gotechnique, 63, 155164. Hill, London.
Tsiampousi, A., Zdravkovic, L. & Potts, D.M. 2013b. A new Hvorslev surface for Zienkiewicz, O.C., Chan, A.H.C., Pastor, M., Schrefler, B.A. & Shiomi, T. 1999.
critical state type unsaturated and saturated constitutive models. Computers Computational Geomechanics (with special reference to earthquake engin-
and Geotechnics, 48, 156166. eering). Wiley, Chichester.