Application of geostatistical methods for the quantification of multiple-scale uncertainty due to aleatory geologic variability, The

Abstract

Tunneling projects in rock are characterized by a high degree of spatial uncertainty, which is due in part to the natural, random (aleatory) variability the rock possesses. Some degree of variability is intrinsic to all rock, and is present due to the complex nature of its deposition or emplacement and subsequent tectonics. This variability is present at multiple spatial scales, from heterogeneous grains to the project scale, where tectonics cause variability in discontinuity properties. As this variability contributes to overall uncertainty in tunneling projects, it is critical to understand and characterize this variability at multiple relevant scales. This research isolated the component of spatial uncertainty associated with aleatory geologic variability and evaluated statistical and geostatistical methods for quantification and characterization of this variability. Geostatistics has been commonly used in natural resource extraction and other data-sparse environments, and has been used extensively in this research as a means by which to better predict, characterize or quantify spatial uncertainty associated with aleatory geologic variability. As the first contribution of this thesis, 2-D covariance maps were generated for rock core specimen photos and were analyzed to identify the number of specimens required in order to adequately represent rock strength. This contribution identified a method by which to quantify this without testing large numbers of specimens at great cost. Next, sequential indicator cosimulation was used to integrate sparse borehole data with a geologist’s interpretation of subsurface lithology, identifying the value added by having a geologist’s interpretation over borehole data alone in uncertainty quantification. This identifies uncertainty in a geologist’s interpretation for use in tunneling projects, whereas geologist interpretations do not typically reflect spatial uncertainty besides boundary uncertainty (besides qualitative indications of confidence in specific parts of geologic boundaries). Finally, indicator kriging was used to quantify uncertainty in ground conditions both prior to and during excavation of the Caldecott Fourth Bore Tunnel in California, USA, demonstrating an approach by which engineers and geologists could quantify uncertainty to inform high-level decision making. The completion of these works provides valuable insight into aleatory variability at multiple spatial scales and demonstrates novel approaches to integrate different types of geotechnical data, including subjective and interpreted, into geostatistical algorithms to better understand spatial uncertainty in the context of tunneling.