Estimation of engineering bedrock layer utilizing ground surface elevation in Gaussian process regression

Abstract

An understanding of the ground structure has a significant impact on the performance of any structure built. Many cases of substantial inclination and settlement of buildings due to insufficient understanding have been reported. In the early stages of design, it is often difficult to estimate the depth to the bedrock layer over a wide area from a very limited number of borings. It is known that the elevation of the engineering bedrock layer and ground surface elevation are highly correlated. Ground surface elevation are useful as information on the elevation of the engineering bedrock layer because such data are available in the early stages. This paper proposes a methodology for estimating the spatial distribution of the engineering bedrock layer by utilizing the ground surface elevation in Gaussian process regression. We focus on the tendency of the engineering bedrock layer to be close in elevation to the ground surface when the slope angle is large. If the slope angle is below a certain value, the ground surface elevation is considered to have no significant information for engineering bedrock estimation and is excluded from observation. The estimation accuracy is evaluated by cross-validation, and the effectiveness of the proposed method is discussed.