Three-Dimensional Geological Modeling with Multisource Data Fusion

Three-dimensional (3D) geological modeling is a modern way to characterize subsurface conditions and support underground digital twins. An essential task is to effectively utilize all available site investigation data and quantify geological uncertainty. This paper presents a generic 3D probabilistic geological modeling framework to fuse multisource data and quantify and reduce geological uncertainty. Data from geophysical tests, boreholes, standard penetration tests (SPTs) and cone penetration tests (CPTs) are integrated utilizing Bayesian sequential updating and density-corrected k-nearest neighbors (kNN) interpolation techniques. Compared with standard kNN, the density correction mitigates bias from clustered data. This framework was applied to two large areas in Hong Kong, and demonstrated more-robust performance and higher computational efficiency than traditional methods. Step-by-step integration of different data sources improves model accuracy and reduces uncertainty, with borehole data contributing the most, followed by CPT and then SPT. In areas with limited borehole data but sufficient geophysical, SPT, or CPT data, the method still can accurately identify geological types. The resulting geological model enables reliable spatial-temporal settlement prediction considering geotechnical and geological uncertainties. The framework enhances the accuracy of 3D geological modeling for large-scale sparse data sites and supports interactive model updates when new data become available.