Modeling of non-stationary random field of undrained shear strength of soil for slope reliability analysis

The spatial variability of soil properties is often assumed to be modeled as stationary or weakly stationary random fields in slope reliability analyses. However, abundant site-specific data have revealed that the mean and standard deviation of soil properties, such as the undrained shear strength of soil, change with depth. Thus, the non-stationary characteristics of soil properties need to be properly accounted for. The aim of this paper is to propose a non-stationary random field (RF) model for the characterization of the spatial variability and the depth-dependent nature of the undrained shear strength of soil. With the proposed model, the uncertainties of the trend and fluctuating components can be modeled individually. As an example, a clay slope under undrained conditions is investigated to illustrate the proposed model. A subset simulation is carried out to evaluate the slope reliability incorporating the non-stationary characteristics of soil properties. The advantages of the proposed model, relative to the existing non-stationary RF models and the commonly-used stationary RF model in the literature, are demonstrated through a series of sensitivity studies.