Slope stability analysis by considering root effect with random field finite element method

Abstract

Soil nailing, geosynthetic reinforcement, retaining wall and cut-off drain are some common engineering measures for slope stabilization. In recent years, however, they have been frequently challenged by their environmentally non-friendly nature. Comparing to these conventional methods, using live vegetation as a protective slope cover appears to be an promising alternatives. Plant roots help reinforce the soil. The term root cohesion has been adopted to characterize the increase in shear strength due to the root-soil interaction. Meanwhile, due to the growth diversity of plants, variability of root cohesion is inevitable. Therefore, a probabilistic approach, which describes the variability of root cohesion, to evaluate the vegetation effect on slope stabilization is required. A probabilistic framework based on Random Finite Element Method (RFEM) is developed and implemented in this study. Comparing to other probabilistic methods (FOSM, FORM, PEM, etc.), RFEM is able to fully consider the spatial variability of geotechnical properties. As an application of the developed framework, the effect of spatial variability of soil shear strength and uncertainty of the soil-water characteristics curve (SWCC) on slope stability of unsaturated soil is analyzed. The additional increase of shear strength due to unsaturation, which is termed as apparent cohesion, can be linked with SWCC through the van Genuchten model. By using Monte Carlo simulation (MCS), the spatial variability of apparent cohesion for any fixed saturation degree can be quantified and described as an isotropic log-normal random field. In order to analyze the effect of correlation length on slope stability, a series of parametric study was conducted for four different saturation degrees. Root cohesion estimation and the uncertainty quantification play an crucial role in the investigation of vegetation effect. On the basis of Wu’s model, a novel probabilistic estimation method of root cohesion was proposed in this thesis. The uncertainty of root cohesion was decomposed into the variability of root architecture and tensile strength. To describe the root architecture, its two descriptors, the root number and root diameter distribution at each depth level, were expressed by enumerated variable and log-normal distribution respectively. The variability of root tensile strength with root diameter was modeled by a power decay relationship. As an application of the proposed approach, a database containing measurements of root system for two native species in Hong Kong, including one shrub (Melastoma sanguineum) and one small tree (Schefflera heptaphylla), was compiled from literature. Based on MCS, the root cohesion at different percentiles was estimated at each depth level. It was found that the root cohesion effect dominates at relatively shallow depths. Moreover, log-normal distribution was found to be proper to characterize the root cohesion at each depth level. Based on the probabilistic results of the estimated root cohesion, the vegetation effect on slope stabilization was investigated. The effect of root cohesion of shrubs was considered as a non-stationary log-normal random field. While the root cohesion influencing region of each tree plant was idealized as semi-circle shape, the effect of vegetation position was analyzed. The lower part of slope area and part of foot area near slope toe are most effective for slope stabilization. Probability of failure decreased linearly with increasing number of plants. The effect of distance between plants was also analyzed. The plants with small gap was shown to provide better stabilization effect than the closely neighboring plants. Three vegetation plans were tested and the vegetation of shrubs gives the highest score comparing to the vegetation of tree (one plant) and mixed vegetation of shrubs and tree (one plant).