Improved Correlations of Geophysical Models with Sparse Borehole Data Using Geostatistical Algorithms

Geotechnical site investigations that incorporate geophysical survey data require an understanding of the relationship between the geological materials encountered and their measured physical properties. Traditionally, this is done in an ad hoc fashion whereby the geophysical models (typically 2D cross-sections) are visually compared to lithology logs from nearby boreholes and/or test pits. We developed a Python-based algorithm that automatically samples the geophysical models at borehole locations such that physical properties can be determined for the different lithology units and contacts encountered. By sampling at multiple locations, site-wide physical property distributions can be determined for any lithology unit or contact of interest. This geostatistical approach provides a means to better classify and interpret the geophysical model data. Probability density functions (PDFs) produced from these sample distributions can be used to convert the geophysically derived physical property models into equivalent models showing the likelihood of encountering a given lithology unit or contact. We tested the algorithm on three data sets from different locations where seismic refraction tomography (SRT) or electrical resistivity tomography (ERT) profiles had been acquired in conjunction with borehole and test pit programs.