Site-scale liquefaction potential analysis using a sectional random field model

Abstract

This paper presents a new random field modeling to assess the site-scale spatial variability of soil properties applied in liquefaction mapping. In this procedure, the initial random fields are primarily generated, based on in-situ measured data. These initial fields are considered as the virtual known values to construct interior random fields, section-by-section and around a central axis. Assembling these sections results in a three-dimensional random field model. Through this approach, more accurate vertical semivariogram models can be available, in addition to a higher number of measured data involved in the generation of random fields. Based on a verification procedure, computational cost reduction and a more appropriate prediction of the layering characteristics are the two most important beneficial points of changing the modeling strategy from planar to sectional. The proposed approach is implemented in a case study in Oceano, California and the spatial distribution of liquefaction probability is estimated. It is concluded that the sectional random field can efficiently locate the liquefiable zone through the soil volume, based on liquefaction evidence. In addition to performing the liquefaction severity assessment, it is also illustrated that the hazard assessments based on liquefaction severity alone cannot provide the comprehensive prediction of liquefaction potential and may lead to unconservative engineering judgments.