Innovative Subsurface Stratigraphy Interpretation by Integrating Electrical Resistivity Tomography and Borehole Data

Subsurface stratigraphy interpretation often relies on sparse borehole data. Due to the inherent spatial variability and limited borehole data, uncertainty inevitably exists in the interpreted subsurface stratigraphy. Geophysical approaches such as electrical resistivity tomography (ERT) provide continuous subsurface data in the concerned cross-section. Though the reliability of the geophysical data derived is much lower than that of the solid borehole data, the continuous subsurface data obtained from geophysical investigations might be taken as a complement to the sparse borehole data in the subsurface stratigraphy interpretation. This paper proposes an innovative subsurface stratigraphy interpretation approach, which takes advantage of the high reliability of sparse borehole data and the abundance of ERT data. This approach is partially developed based on the random field approach recently advanced by the authors, and the relationship between ERT data and borehole stratigraphies is mapped with the algorithm of the support vector machine (SVM). The uncertainty of the interpreted subsurface stratigraphy arising from the random selection of the training and testing datasets (for training the mapping relationship between ERT data and stratigraphies) is analyzed by a ‘bootstrapping’ method. Furthermore, field tests of site investigation at four benchmark stratigraphic cross-sections are conducted, and high-quality ERT and borehole data are collected. Based on the high-quality site investigation data obtained, illustrative applications of the proposed approach are presented.