Structural Agent-Based Approach for 3D Geological Surface Modelling

Abstract

Geological surface modelling is the process of creating 3D digital representations of the Earth's subsurface structures and surfaces based on geological, geophysical, and geospatial data. It's commonly used in geology, mining, petroleum exploration, hydrology, and geotechnical engineering. Traditional interpolation methods often rely on assumptions like spatial stationarity and high data availability, which are not always the case in earth sciences applications. These methods fail to capture local variation in geologically complex and data sparse areas, as they tend to be biased towards global means. This research explores how agent-based modelling (ABM) integrated with statistical and domain-specific knowledge can be used to improve geospatial estimation when data is sparse and incomplete, especially in geological surface modelling. In this study, structural agents are introduced as independent entities that interact with each other using rules from the classic BOID flocking algorithm: cohesion, separation and alignment. Each agent is defined by its spatial position (X,Y,Z), a normal (N₁,N₂,N₃) and a velocity (V₁,V₂,V₃) orientations, and updates its behaviour over time through local communication. A two-phase approach was used: Phase I tested if agents could self-organize, and Phase II evaluated how well agents could propagate information from real data points. The long-term goal of the study is to establish viability of a method for reconstructing surfaces of complex geological features such as faults, poly-deformed folds and regional fault networks. In the short term, this work focuses on determining what data configurations and agent behaviours support the simplest reconstruction scenarios, before addressing more complex geological cases.