SFARS - Seismic failure and post-failure response of slopes

Abstract:

Earthquakes have long been recognised as a major cause of landslides involving development of massive soil movement and large

post-failure deformations both in natural slopes and earth structures, such as dams or embankments. For example, the 2016

Kumamoto earthquake in Japan triggered 3460 landslides within an area of about 6000 km2 (Xu et al., 2018); the 2016 Kaikoura

earthquake in New Zealand generated more than 10000 landslides, the largest one exceeding a volume of 20 million m3 (Massey et

al., 2018). Extensive lateral spreads were also observed in Italy during the 2012 Emilia earthquake (Fioravante et al., 2013).

Recently, dams that underwent large deformations are the Niwaikumine and Makio dams (Tani, 2000), three dams in India during

the 2001 Bhuj earthquake (Singh et al., 2005) and several dams in China during the 2009 Wenchuan earthquake. Therefore, there is

great impetus to deepen the understanding of the deformation processes and failure mechanisms of the slopes subjected to

earthquakes and develop suitable numerical tools to assess the consequences associated with the occurrence of these phenomena

in the failure and post-failure stages of the landslides.

A considerable amount of work has been conducted to analyse seismic slope stability. Unfortunately, a limited number of detailed

experimental studies focused on the seismic and post-seismic behaviour of slopes are available in the technical literature. Most of

the existing studies are theoretical or deal with the back-analysis of earthquake-induced landslides. They move from simplified

methods to fully coupled continuum analyses using the finite difference (FDM) or the finite element method (FEM). In some studies,

these methods were also used to analyse landslide triggering due to soil liquefaction (Boulanger, 2019). To date, however, it is still

an open challenge to model the entire process of a landslide, including failure initiation, run-out and deposition. Indeed, a strong

limitation of the above-mentioned numerical techniques is their inability to maintain numerical accuracy due to mesh-distortion

effects when the slopes undergo large deformations. Thus, these methods are unsuitable to model the post-failure stage of

landslides, especially when a flow-type failure occurs due to soil liquefaction. The material point method (MPM) is an innovative

numerical method that overcomes this limitation (Sulsky et al., 1994), allowing for the solution of engineering problems involving

large deformations. To date, MPM has been also used to investigate the slope response to earthquakes (Bhandari et al., 2016; Tjung

and Soga, 2021). These studies have demonstrated the potential of MPM in modelling the seismic and post-seismic response of

slopes.

In the current state of practice, simplified methods are often used. Considering that a flow-type failure represents a condition

whereby static shear stresses in a slope of saturated cohesionless soils exceed the strength of the liquefied soil (i.e., the so-called

undrained steady-state or residual strength), a static limit equilibrium analysis is usually performed using the residual strength of the

involved soil. This soil strength is usually evaluated using empirical relationships derived from the back-analysis of case histories

(Idriss and Boulanger, 2008; Kramer and Wang, 2015). Simplified methods are also available for a rough estimate of the lateral

spread displacements (Youd et al., 2002; Faris et al., 2006). For clayey soils susceptible to cyclic softening, a reduced value of the

undrained strength is instead used in the slope stability analyses, as suggested by Boulanger and Idriss (2007). When a flow-type

failure is not expected to occur (static shear stresses not exceeding the residual strength), pseudo-static methods are generally

applied in which the inertial forces acting on the potentially unstable soil mass are evaluated using a seismic coefficient that is

usually provided by the current regulations. When the resulting safety factor drops below unity, the Newmark sliding block model is

often used to predict the slope displacements. According to this model, sliding would occur when the input acceleration exceeds a

critical value and cease when the relative velocity becomes zero. On the other hand, practical methods capable of predicting the

consequences of the deformation processes (from triggering to the run-out of the landslide) occurring in slopes subjected to

earthquakes are still missing in the technical literature.

The present proposal is framed in the above-described context.