Spatial distribution estimation by considering the cross-correlation between components with indirect data using Gaussian process regression

Generally, soil properties are measured only at limited locations. To rationally estimate the spatial distribution of soil properties, it is preferable to effectively use all available measurement data, including indirect data. We propose a Gaussian process regression with multiple random fields tha...

Full description

Saved in:
Bibliographic Details
Main Authors: Yuto Tsuda, Ikumasa Yoshida, Shinichi Nishimura
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:Soils and Foundations
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S0038080625000587
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Generally, soil properties are measured only at limited locations. To rationally estimate the spatial distribution of soil properties, it is preferable to effectively use all available measurement data, including indirect data. We propose a Gaussian process regression with multiple random fields that considers the cross-correlation between one of the random fields of direct data and indirect data, and show the application to simulated data and actual measured data. In the application, the direct data are of CPT tip resistance (qc), which was obtained within a narrow area, and the indirect data are of shear wave velocity (Vs) obtained by surface wave exploration, which were obtained over a wide area. We estimate the spatial distribution of qc from the limited qc and wide area Vs data. The estimation accuracy of the proposed method is evaluated by cross-validation, and its effectiveness is discussed.
ISSN:2524-1788