3-D geological model for Hong Kong

Abstract

Hong Kong initiated a number of world-class civil engineering projects such as Lantau Tomorrow Vision and the Northern Metropolis. Different from the past practices, digital technologies will be routinely applied in these new projects, including ground investigations and underground works design. To facilitate foundation and structural design and to minimize the associated risks during the construction, it is crucial to understand the subsurface conditions before conducting any construction projects. A tremendous amount of geological survey and investigation works have been carried out for this purpose in Hong Kong, leading to an extraordinary number of ground investigation reports including borehole data and laboratory test results. How to fully utilize the existing data from ground investigation works to assist the detailed site investigation for new projects is a hot topic in the geotechnical community. Yet, existing commercial software is not sufficiently comprehensive to provide a data management system and 3-D geological modelling and displaying capabilities. In particular, the ArcGIS platform, a common software used in the industry, demonstrates good ability in managing geo-databases but does not automatically generate geological or stratigraphic layers using built-in algorithms that are not intended for categorical interpolation. This study aims to digitalize ground investigation data supported by the Geotechnical Engineering Office, develop an easy-to-use toolbox for 3-D borehole management and visualization, and eventually establish a 3-D geological model for the entire Hong Kong. The methodologies consist of geological data acquisition, data processing, toolbox development and 3-D model construction. After processing approximately 90,000 boreholes with 667,000 records and 46,000 standard penetration tests (SPT-N) with 371,180 records, 3-D virtual boreholes were established and managed using the toolbox developed in ArcGIS Pro. Furthermore, cross sectional diagrams of borehole logs can be created, together with 3-D boreholes to verify the data collected. Most importantly, 3-D models and the associated information entropy can be generated and visualized in the software with the aid of the toolbox. Building 3-D geological models based on machine learning is proved to be an innovative method to provide accurate estimation of soil layering. The 3-D geological models and the constructed fence diagrams not only help engineers and geologists to have a better understanding of the complicated sub-surface profiles in a 3-D way, but also provide estimates of volumes of different soil layers locally. With an optimal anisotropy ratio δ determined in this study, the complexity of Hong Kong geological layers and the uncertainty reflected by information entropy could be shown in the 3-D model, allowing further site investigation works to be carried out to minimize the uncertainty in a region with higher uncertainty. Enhancements on 3-D soil classification were further performed using the piezocone penetration test (CPTu) data. Joining two soil probabilistic models by Bayes’ Theorem demonstrates a higher reliability and less uncertainty in the 3-D soil classification, improving the evaluation of complex soil layering. GIS web applications were created in a management and modelling system, providing an access for others to retrieve table of information, offering a reference for future city planning and underground constructions, and minimizing the risks during the construction stage.

Date of Award	2022
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Li Min ZHANG (Supervisor)