LiDAR-based 3D litho-stratigraphic models calibrated with limited boreholes

An accurate and reliable 3D engineering geological model (expressed herein as a 3D litho-stratigraphic model) is essential for designing slopes, underground structures, and tunnels to reduce the risk of failure. Due to budget constraints, subsurface investigations with traditional methods (e.g., boreholes and geophysical prospecting) are often limited in quantity. Insufficient investigation often leads to inadequate litho-stratigraphic data that can degrade the reliability of the developed 3D litho-stratigraphic models. To this end, LiDAR, which can collect topographic data more efficiently and economically, may complement traditional geological investigations for developing litho-stratigraphic models. However, the task of integrating the subsurface geological data and LiDAR data to create 3D litho-stratigraphic models for engineering applications is challenging. This paper proposes a methodology that combines LiDAR-derived geological data (Type I data herein) and borehole data (Type II data herein) to develop realistic 3D litho-stratigraphic models. To demonstrate the methodology, we took a 1.2 km × 1.2 km area with sedimentary rocks in northern Taiwan as the study area. Under the assumption that the strata are parallel, we first fitted Type I data to a surface regression model (a polynomial mathematical model). Then, we incorporated Type II data into the regressive process to develop the desired 3D model. Comparing the former model (based on Type I data) with the latter model (based on Type I + II data) through a Root-Mean-Squared-Error (RMSE) analysis, it was found that the RMSE of the latter model decreased significantly, indicating a decrease in the error between the developed litho-stratigraphic model and the ‘ground truth.’ To verify this finding, we compared the newly developed 3D model with the ‘known’ geological cross-section of the Qidu Tunnel constructed nearby. The results indicated that the distribution of the primary sandstone strata uncovered from the previous Qidu Tunnel project and those projected by the newly developed 3D model was highly consistent. Thus, this study demonstrates that LiDAR data combined with a few borehole data has the potential to create reliable and accurate 3D geological models, providing an effective solution for obtaining 3D geological information at the scale of engineering applications with limited boreholes. Finally, the proposed modeling approach is further demonstrated through two hypothetical engineering application cases.