Reviewing probabilistic soils modelling

The probabilistic approach to soils and foundation engineering requires a suitable set of models. Consistency and comparability require that these models are to a certain extent standardized but capable of incorporating the most important aspects. The concept of ergodicity is introduced as an important tool of probabilistic modelling. For layered soils some standard random field models are proposed. Their properties are discussed. Especially for soils and foundation engineering a Bayesian approach to handle different types of knowledge is mandatory - even in simplified form. An attempt is made to quantify prior information about different soil types and a simple update formula is presented. A methodology to treat classification errors is proposed. An illustrating example is presnted. (C) 2000 Elsevier Science Ltd. All rights reserved.