Quantification of prior knowledge in geotechnical site characterization

Geotechnical characterization of a project site often starts with desk-study and site reconnaissance, which provide site information available prior to the project (e.g., existing data in literature, engineering experience, and engineers' expertise). Such information can be used as "prior knowledge" under a Bayesian framework and be quantitatively reflected by a prior distribution in Bayesian methods. However, it is not a trivial task for engineering practitioners to properly quantify prior knowledge as a prior distribution. This paper develops two different methods to quantify prior knowledge during geotechnical characterization of a project site. Where there is no prevailing prior knowledge on the site, a non-informative prior distribution (e.g., uniform prior distribution) is used to reflect quantitatively the engineering common sense and judgment. As prior knowledge improves and becomes much more informative, a subjective probability assessment framework (SPAF) is proposed to estimate the prior distribution from prior knowledge. The proposed SPAF framework assists geotechnical engineers in formulating and expressing their engineering judgments in a quantifiable and transparent manner. These two different methods are illustrated using information from a sand site of the US National Geotechnical Experimentation Sites (NGES) at Texas A&M University (TAMU).