Dealing with Nonlattice Data in Three-Dimensional Probabilistic Site Characterization
Jianye Ching, Zhiyong Yang, Kok‐Kwang Phoon
Abstract
In site investigation, it is common to conduct some soundings to explore greater depths that are not explored by remaining soundings. This produces the scenario of nonlattice data, meaning that not all soundings measure identical depths. Recently in 2020, the first and third authors of the current paper developed a probabilistic site characterization method based on sparse Bayesian learning (SBL). This SBL method assumes lattice data (all soundings measure identical depths) to take advantage of the Kronecker-product derivations. These Kronecker-product derivations significantly improve computation efficiency, so the resulting SBL method can be scaled up to address full-scale three-dimensional problems. However, this SBL method is not applicable to nonlattice data, which are common in practice. The purpose of the current paper is to modify the SBL method developed in 2020 to accommodate nonlattice data, while retaining the crucial computational advantage of the Kronecker-product derivations. One real-world case study of underground stratification is used to demonstrate the usefulness of the modified method.