Quantifying spatial uncertainty in rock through geostatistical integration of borehole data and a geologist's cross-section

In order to increase understanding of spatial uncertainty in subsurface conditions for problems in geological and geotechnical engineering, this study develops a geostatistical approach for incorporating a geologist's interpreted cross-section to quantify spatial uncertainty of lithology. In this work, the Sequential Indicator Cosimulation (COSISM) algorithm and the information theory concept of entropy are utilized to generate uncertainty models. A sensitivity analysis is used to study how both borehole data and the cross-section are incorporated into the algorithm, which is quantified using entropy as a metric for uncertainty in geologic unit boundary locations. A verification methodology is then developed to identify the most appropriate combination of input parameters for combining these two data types; recommendations from the verification methodology are made by comparing the calculated entropy to accuracy at the global scale on a point-by-point basis. This methodology is then used to investigate how calibrated input parameters vary due to differences in the input cross-section based on different levels of understanding of the subsurface geology by iteratively providing a geologist with additional borehole data after a cross-section is developed. Results show that two of the parameters, variogram range and the number of conditioning data, can be used to most directly reflect geologist understanding in the simulations. Finally, this study provides guidelines for incorporating a geologist's understanding of a geological environment into this modeling framework, as results show cross-sections that are locally accurate are better represented by different parameters than cross-sections that are not locally accurate but are globally accurate. These guidelines are intended to aid geologists and engineers in understanding how to incorporate a geologist's interpretation of the subsurface lithology or categorical geotechnical parameters at a particular worksite into a structured geostatistical framework.