ABSTRACT
The values of soil properties used in geotechnical engineering and geomechanics involve a significant level of uncertainty arising from several sources such as: inherent random heterogeneity (also referred to as spatial variability), measurement errors, statistical errors (due to small sample sizes), and uncertainty in transforming the index soil properties obtained from soil tests into desired geomechanical properties (e.g. Phoon and Kulhawy, 1999). In this chapter, only the first source of uncertainty will be examined (inherent random heterogeneity), as it can have a dramatic effect on soil response under loading (in particular soil failure), as will be demonstrated in the following. As opposed to lithological heterogeneity (which can be described deterministically), the aforementioned random soil heterogeneity refers to the natural spatial variability of soil properties within geologically distinct layers. Spatially varying soil properties are treated probabilistically and are usually modeled as random fields (also referred to as stochastic fields). In the last decade or so, the probabilistic characteristics of the spatial variability of soil properties have been quantified using results of various sets of in-situ tests.
