Regional Liquefaction Mapping Accounting for Multiscale Spatial Variability of Soil Parameters with Geological Constraints

This paper aims at introducing a framework for estimating regional liquefaction hazard considering the spatial variability of soil properties and geological constraints. Firstly, the ground-truth liquefaction potential index (LPI) at in-situ testing locations is calculated. Then, fake LPI values are randomly sampled according to their distributions within different surficial geological units. Both ground-truth and fake LPI values are then treated as known data and fed into a novel multiscale random field model, which can generate LPI values at places where in-situ testing is not performed. The developed framework is able to make predictions of liquefaction potential within a large region. The multiscale random field model is advanced because it can refine the prediction into finer resolutions. Another advantage of the developed framework is that the geological constrains is easy to apply to the random field model so that the predictions could be corrected by geology conditions of the studied area. An example application is performed to estimate the liquefaction potential in an earthquake prone region in the Alameda county of California. It is found that the developed framework is able to capture the fragile portions of the studied region without significant contradictions to the geological constraints.