Prediction of interface of geological formations using generalized additive model

ABSTRACT

Geological information such as geological interfaces is important for the design of underground excavation and supporting measures. This in turn requires a method to predict accurately the locations of geological interfaces for the gap areas between boreholes. This study presents a generalized additive model (GAM) to predict the location of the geological interfaces. The performance of the GAM method is evaluated using both simulated data and borehole data for the determination of rockhead in two different geological formations in Singapore. The results show that the GAM method can provide a reasonable confidence interval (CI) of the mean trend and the prediction interval (PI) in the sense that the 95% CI covers about 95% of the actual mean curve while the 95% PI covers around 95% of testing data. Furthermore, the geological complexity can be well reflected as the prediction uncertainty in the geologically complex area is larger than that in the geologically regular area. More importantly, the users can impose prior information or personal judgment regarding the shape of the geological profile on the model. This is an important feature to enable further improvement in the accuracy of the prediction.

Highlights

• A generalized additive model is used to predict the location of the geological interfaces

• The performance of the GAM method is evaluated using both simulated data and actual borehole data from Singapore

• The results show that the GAM method can provide a reasonable confidence interval of the mean trend and the prediction interval of a prediction

• Geological complexity can be well reflected in the sense that the prediction uncertainty in the geologically complex area is relatively large

• Expert judgment or knowledge on the geological profile can be applied to the model, which improves the prediction accuracy

KEYWORDS:

• Spatial prediction
• geological interface
• generalized additive model
• cubic spline

Acknowledgment

The second author would like to thank the financial support from the National Natural Science Foundation of China (No. 52109144) and the Open Innovation Fund of Changjiang Institute of Survey, Planning, Design and Research (No. CX2020K07).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by National Natural Science Foundation of China: [Grant Number 52109144]; Open Innovation Fund of Changjiang Institute of Survey, Planning, Design and Research: [Grant Number CX2020K07].