QUANTIFYING EXPLORATION SUFFICIENCY WHILE ACCOMMODATING JUDGMENT

Abstract

A computationally efficient method to direct and evaluate the sufficiency of site characterization data is demonstrated on two synthetic subsurface geometries. This method, entitled reliability-based exploration (RBE), combines three-dimensional subsurface analysis and uncertainty in the geologic data to produce a quantifiable measure of data sufficiency, which is directly related to the probability of project success. The RBE approach is illustrated with a three-dimensional finite element soil deformation model, where interfaces separating geologic units are the subsurface data of interest. The RBE method begins with a prior model of the subsurface, which can include geologic judgement. As new information is collected at a site, the prior subsurface model is updated with site data. During the RBE analysis, both a probability of project success and the location of the next boring are computed. The robustness of this computation is illustrated with a practical example of the automatic exploration of two different pinching geometries. The subsurface geometry is imposed on these two sites to demonstrate the importance of geology on the computed results. A comparison of the modeled subsurface and the "true" (or defined) subsurface at each site shows a close match directly below the proposed structure, where calculated settlement is most sensitive to uncertainty in geology. The study demonstrated that to achieve the same exploration sufficiency more data are required at the site with abruptly changing subsurface geometry than at the site with gradually changing geometry.