Probabilistic Landslide Simulation: From Initiation to Post-Failure Large Deformation of Soils


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
  • Yu WANG (Supervisor)
  • Dianqing Li (External person) (Supervisor)
Award date4 Sep 2020


Landslides are major geo-hazards in the world, endangering life and property in many hilly and rainy regions. Landslides are challenging to predict due to uncertainties, such as uncertainties in triggering mechanism and spatial variability of soil properties. These uncertainties can be quantitatively and properly accounted for using probabilistic approaches and quantitative risk assessment of landslides, which have been rapidly developed in recent years. Although a landslide is often a dynamic process with multiple stages, from a stable slope, triggering by factors (e.g., rainfall, earthquake, human activity), initiation of a landslide, post-failure large deformation of soils, and final deposition of soils, most existing probabilistic methods are often limited to the landslide initiation stage due to the restriction of traditional slope stability analysis methods (e.g., limit equilibrium methods, LEM, and finite element method, FEM) in simulating large deformation of soils. Material point method (MPM), which is a numerical tool like FEM but capable of simulating large deformation of soils, is adopted in this thesis to simulate the entire process of landslides, particularly large deformation of soils.

This thesis aims to develop probabilistic simulation approaches for landslide from initiation to post-failure large deformation of soils. Before a landslide initiates, there are numerous potential slip surfaces to be considered in probabilistic slope stability analysis. Rainfall, which triggers most landslides in Hong Kong, should also be considered. Once a landslide is initiating, it is critical to model and characterize its evolutionary process with large deformation of soils. This thesis aims to address these challenges. An adaptive Monte Carlo simulation (MCS) method was proposed for probabilistic slope stability analysis. In addition, both single-phase and two-phase hydro-mechanically coupled MPM were developed for landslide simulation. A real rainfall-induced landslide case history (i.e., the Fei Tsui Road landslide occurred on 13 June 1995 in Hong Kong) was simulated using the two-phase MPM to investigate its triggering and failure mechanisms. A random limit equilibrium and material point methods (RLE-MPM) framework was proposed to investigate the evolution of landslide failure modes in spatially variable soils. A random finite element and material point method (RFE-MPM) with hydro-mechanical coupling was proposed for probabilistic simulation of the entire process of rainfall-induced landslides. 

Results showed that the proposed adaptive MCS method is able to consider a large or even unlimited number of circular and/or non-circular potential slip surfaces in slope system reliability analysis in an efficient manner. It is also suitable for the reliability analysis of large series systems (component number >1000). A complete slope system considering all potential slip surface can be represented by a slope subsystem comprised of a small number of slip surfaces. The proposed RLE-MPM framework is able to identify different landslide failure modes and investigate its evolution, and it can be used to estimate both landslide probability and consequences (e.g., runout distance, sliding volume, sliding depth), providing valuable information for risk assessment of landslides. The proposed RFE-MPM framework is able to model the entire process of rainfall-induced landslides, from triggering by rainfall to the post-failure large deformation of soils. It was found that the spatial variability of soil strength and seepage properties not only greatly affects the slope failure probability, but also determines landslide failure mechanisms and consequences. It is, therefore, of great importance to characterize the site-specific spatial variability of soil properties during landslide risk assessment and mitigation.