LAUSR

Abstract The damage caused by soil liquefaction during earthquakes is substantially affected by the total volume and spatial distribution of the liquefied soils. Therefore, a proper characterization of the spatial distribution of liquefaction potential of subsurface soils plays a critical role in earthquake hazard assessment and mitigation. In engineering practice, in situ tests, such as the cone penetration test (CPT), are widely used for evaluating soil liquefaction potential through, e.g., factor of safety (FS) against liquefaction, and the results of liquefaction assessment in a cross-section are needed for subsequent earthquake hazard analysis and mitigation. Traditional simplified methods only provide the liquefaction potential of soils at individual locations, and the results of liquefaction triggering evaluation at individual locations are often used to represent the whole site. This might lead to an unconservative estimation of earthquake-induced deformations and damage at a site where spatial variation of soil properties is significant. In addition, although CPT provides almost continuous sounding along vertical direction, measurements along horizontal directions are usually limited. This leads to the difficulty in proper application of geostatistical methods in liquefaction potential analysis. To address these challenges, a non-parametric and data-driven method is developed in this study to interpret a cross-section of FS with high spatial resolution directly from sparse CPT soundings. Based on the interpreted cross-section of FS, the predicted total volume and spatial distribution of liquefied soils are estimated. Both simulated and real CPT data are used to demonstrate and validate the proposed method, and the results show that the proposed method performs reasonably well.