Using Conditioned Random Field to Characterize the Variability of Geologic Profiles

Abstract

Geologic profiles cannot be identified with certainty due to inherent spatial variability and limited site-investigation data. Random field theory is a powerful tool for evaluating the spatial characteristics of soil and rock properties, but may overestimate the site variability if the site investigation data at borehole locations within the site are not considered. This paper aims to propose a method to generate a conditioned geologic random field based on available measurement information. It incorporates both direct and indirect information within the site to constrain the random field. The conditioned random field matches the measured data at the measurement locations, and has much-reduced uncertainty adjacent to the measurement locations due to spatial correlation. The proposed method is applied to evaluate the depth of Grade III rock surface at a construction site in Hong Kong using borehole data. With the proposed method, a more accurate description of the geologic profile can be obtained, which is desired for geotechnical design and construction.