Geostatistical-learning-based site-optimum 3D integration of borehole logs and geophysical data in urban area in South Korea

Success in civil engineering projects fundamentally depends on thoroughly understanding the site-specific subsurface characteristics. Site investigation, a critical process in the early stages of construction and design, serves as the foundation for ensuring the safety and efficiency of structural development and safeguards against potential disasters. However, owing to financial and time constraints, the number of site investigations is often limited, making spatial uncertainty one of the most significant challenges in geotechnical engineering. Geostatistics-based spatial interpolation techniques are widely used to overcome the limitations of spatial variability and information scarcity in geotechnical engineering. Reliable geospatial analysis is essential for identifying site-specific subsurface stratification information. In this study, site investigation data were collected at a subway construction site at which subsidence occurred during tunnel excavation. Borehole data were optimized using outlier removal to maximize reliability, and geophysical data were digitized to create a 3D integrated database with borehole data. Considering the subsurface characteristics, the optimal stratigraphic boundary elevations were determined using seismic wave velocities values, which clarified the optimized stratigraphic boundaries. Using kriging and simulation-based integrated analysis techniques, the subsurface stratigraphic information was predicted in 3D, and the cross-sectional and longitudinal geotechnical profiles confirmed that the layers with the least deviation effectively reflect the actual strata, which is consistent with the evaluation results, through a learning process that seeks the optimal method and parameters that produce the least prediction residuals. This approach highlights the importance of integrating advanced geostatistical-learning-based integration and geotechnical engineering practices to improve the accuracy and reliability of subsurface evaluations, thereby ensuring safer and more efficient construction.