
Landslides
DOI 10.1007/s10346-020-01451-1
Received: 26 September 2019
Accepted: 28 May 2020
© Springer-Verlag GmbH Germany
part of Springer Nature 2020
Zhichen Song I Xiang Li I José J. Lizárraga I Lianheng Zhao I Giuseppe Buscarnera
Spatially distributed landslide triggering analyses
accounting for coupled infiltration and volume change
Abstract Rainfall infiltration in unsaturated slopes alters the effective
stress through pore water pressure changes, thus causing ground defor-
mation. Although important to assess the timescale over which the
margin of safety of a slope decreases, such coupled processes are rarely
accounted in the context of spatially distributed hazard assessment
procedures. In this paper, a physically based, spatially distributed model
accounting for full hydro-mechanical coupling is discussed. The model
relies on a vectorized finite element (FE) solver to calculate the stability
of deformable unsaturated infinite slopes subjected to transient flow .
First, the FE solver is used to study the response of individual slopes to a
prolonged rainfall for three scenarios (i.e., rigid, swelling, and collapsible
soil). Then, the model is used in the context of spatially distributed
computations to assess spatiotemporal variations of factor of safety over
a large area. For this purpose, a series of shallow landslides occurred in a
mountainous landscape covered by collapsible loess deposits in north-
western China was used as test site. The analyses show that hydro-
mechanical couplings affect the performance of the model in terms of
computed failure time and areal extent of the unstable zones. Specifi-
cally , volume collapse due to suction decrease is found to reduce the
time of failure compared with uncoupled computations obtained for a
rigid soil scenario. The most substantial advantages of using coupled
analyses have been reported with reference to gentle slopes, for which
the higher rate of suction reduction driven by volume change was crucial
to capture landslide source areas that would otherwise be overlooked by
uncoupled analyses. The proposed methodology offers a complete tool
for landslide hazard assessment, in that it incorporates sources of
coupling between hydrology and mechanics that are crucial to replicate
the physics of landslide initiation.
Keywords Hydro-mechanical coupling
.
Infiltration
.
Spatially
distributed analysis
.
Unsaturated soils
.
Collapse
Introduction
Rainfall-induced landslides are among the most widespread and
frequent hazards around the world (Petley 2012). Water infiltration
is indeed a well-known source of soil strength deterioration, either
by increasing pore water pressure (hence reducing the frictional
resistance) or by changing the soil rheology through enhanced
deformability and wetting-induced volume change (Alonso et al.
1990; Rahardjo and Fredlund 1995; Mihalache and Buscarnera
2016). Specifically, as water infiltrates in an unsaturated soil, suc-
tion and degree of saturation vary, eventually giving rise to alter-
ations of the stresses acting on the skeleton and volume changes.
At the same time, changes in the state of saturation controlled by
the volume change response may affect the hydraulic characteris-
tics of the soils, thus influencing the timescale of the infiltration
process and the rate at which deformation and failure may occur
(Wu and Zhang 2009; Garcia et al. 2011; Kim et al. 2016a, b). It is
therefore apparent that, under the most general circumstances,
water infiltration and soil are coupled, in that they affect each
other and determine the hydro-mechanical response of natural
unsaturated soil slopes (Zhang et al. 2005).
Several approaches have been used to evaluate rainfall-induced
landslide hazards at regional scale, such as empirical rainfall threshold
methods (Ti ranti and Rabuffetti 2010;Godtetal.2006; Brunetti et al.
2010;Salciarinietal.2012;DeVitaetal.2013) or statistical and proba-
bilistic methods based on historical records (Ohlmacher and Davis John
2003;Coeetal.2004). Over the last decades, a growing number of
physically based regional models for the assessment of rainfall-induced
landslide susceptibility have also been proposed (Montgomery and
Dietrich 1994;Iverson2000; Salciarini et al. 2008;Baumetal.2010;
Lepore et al. 2012;Parketal.2013;Suetal.2015;Buietal.2017;Zhaoetal.
2019). The recent improvement of the computing performance and the
development of increasingly accessible geographical information system
(GIS) platforms and remote-sensing technology have further contribut-
ed to the widespread use of such physically based models, by making
them increasingly more powerful and reliable for regional-scale land-
slide forecasting. A crucial characteristic of such class of landslide
assessment tools is the simulation of subsurface hydrologic processes
in light of well-defined balance equations and constitutive relations.
Although such models proved useful in several geological settings, they
often recur to simplified descriptions of the soil behavior by neglecting
the deformability prior to failure or hypothesizing frictional slip as the
only mechanism originating instability . However, recent studies on the
mechanics of shallow landslides have shown that volume changes prior
tofrictionalfailureplayacrucialroleforthetriggeringofshallow
landslide in so-called collapsible soil, i.e., deposits which may experience
volume loss upon water infiltration (Buscarnera and Prisco 2012;
Buscarnera and Di Prisco 2013; Lizárraga et al. 2017). Such studies
suggest that neglecting the coupling between water intake and soil
rheology may lead to inaccurate assessments of the rate and magnitude
of the deterioration of the margin of safety during a storm, thus
potentially rendering the analysis unconservative.
The purpose of this paper is to take into account the coupling
between fluid flow and deformation throughout the course of a rain-
storm, thus acknowledging the role of volume changes on the transients
that control the variation of pore pressure within a slope. In standard
uncoupled models, a seepage analysis is used to predict pore water
pressures within a given time, eventually using them as input in stability
calculations (Cai and Ugai 2004;YooandJung2006). In such analyses,
the soil is essentially assumed rigid, in that no soil property enters into
the mass balance equations used to compute pore pressure transients.
Abundant field and laboratory evidence, however, suggests that the
hypothesis of rigid soil during infiltration may be overly restrictive.
For example, T abarsa et al. (2018) conducted a series of collapse poten-
tial tests on loess samples taken from a site susceptible to landslides,
showing a significant risk for wetting-induced collapse. Along the same
lines, Schulz et al. (2018) found a considerable role of swelling in the
dynamics of slow-moving landslides in clay soils subjected to seasonal
rainfall infiltration. In all these cases, it is arguable that the infiltration
processes responsible for the strength deterioration that eventually led
to ground failure took place within deformable soil slopes, thus being
influenced by coupled fluid flow-deformation processes. The impor-
tance of hydro-mechanical couplings has been extensively documented
Landslides
Original Paper