How Subjective Are Geotechnical Reliability Estimates?

In geotechnical reliability analysis and reliability-based design, the distributions of soil parameters are typically determined from multiple information sources, in particular field data but also prior investigations, literature sources and expert knowledge. These different sources of information can be combined in a consistent manner through application of Bayesian analysis, whereby a prior distribution is updated with data to a posterior distribution. The so determined probability distributions and associated predictions will depend on the subjective use of information that manifests in the choice of the prior distributions. This paper shows that the effect of the subjective choices is not necessarily as drastic as discussed in the literature, if consistent modeling choices are made and if the spatial variability is properly accounted for. We consider a shallow foundation resting on spatially variable silty soil and study the influence of the assumptions on the prior marginal distribution of the friction angle on the predicted reliability of bearing capacity.