Effect of spatial variability on reliability evaluation of an earth slope

Various uncertainties exist in slope engineering, such as inherent spatial variability of soil properties and uncertainties induced by simplifications and approximations adopted in geotechnical models (e.g., geometry and location of slip surfaces in limit equilibrium methods). This paper illustrates these effects through a cohesive soil slope example using a practical approach to probabilistic slope stability analysis that implements an advanced Monte Carlo Simulation (MCS) method called Subset Simulation in a spreadsheet environment for improving the efficiency and resolution of MCS at relatively small probability levels. It is found that, when spatial variability of soil properties is ignored by assuming perfect correlation, the variance of Factor of Safety (FS) is over-estimated, which may result in either over-(conservative) or under-(unconservative) estimation of the probability of failure [Pf = P(FS < 1)]. When spatial variability of soil property is considered, the critical slip surface varies spatially and such uncertainty should be properly accounted for. Otherwise, the probability of failure can be significantly under-estimated and unconservative.