Computing The Reliability Of Soil Slopes Using Conditional Simulation

The geostatistical technique of conditional simulation is applied to measure the uncertainty inherent in the factor-of-safety (FS) predictions for soil slopes in which there is an incomplete knowledge of the spatial distribution of soil properties. A case study that considers a natural clay slope in Canada is presented. The sequential-indicator simulation technique was used as a means of generating a series of spatial maps of the soil properties at the site. These maps approximate random draws from the conditional probability density function that characterizes existing uncertainty in the spatial distribution. FS determinations were made for each simulated map using the simplified-Janbu approach. In using this approach, computations are performed using a new computer program that can operate on irregular spatial profiles of soil properties specified on a discrete grid. A unique capability of this approach is the ability to generate the FS distribution conditional to the type, position and value of existing site-characterization measurements. This capability should reduce uncertainty in FS predictions.