Principles of Probabilistic Characterization of Soil Properties

Abstract

Summary by Lulu Zhang: Relevant soil parameters governing most geotechnical performances are spatial average soil properties over some domain of the soil layer. In this paper, a simplified estimation method of spatial average soil properties is proposed to incorporate the three major uncertainties, i.e., inherent spatial variability, systematic uncertainty, and statistical uncertainty due to insufficient samples. The uncertainty due to inherent spatial variability can be evaluated based on the random field theory. Multiple sources of systematic discrepancies can be integrated by combining the bias factors and the systematic uncertainties from each source. When evaluating the statistical uncertainty, the effect of indirect tests coupling with empirical equations and statistical dependence due to spatial correlation structure should be considered. For estimation from censored samples, the mean and the standard deviation of soil property can be obtained using an empirical approach or a Bayesian approach. It should be noted that if several independent sources of information are available to estimate the overall statistics of a soil property, inherent spatial variability is generally not reducible with additional information, whereas the modeling error could be reduced. It is expected that this method would yield a more rational and realistic description of the in situ soil properties for further engineering analysis and design.