An Integrated Geodata Science Workflow for Resource Estimation: A Case Study from the Merensky Reef, Bushveld Complex

This study presents a novel integrated geodata science workflow for mineral resource estimation, addressing key challenges in modern exploration and mining operations. Unlike traditional geostatistical methods, which are often manual, fragmented, and limited in scalability, the proposed workflow leverages machine learning, spatial modeling, and combinatorics-based visualization to create an automated and objective resource estimation approach.

The study focuses on the Merensky Reef in the Bushveld Complex, South Africa, a region known for its rich platinum deposits and high nugget effect. The authors demonstrate the effectiveness of their approach using actual mining data, ensuring practical relevance. The workflow incorporates several advanced data-processing techniques, including:

• Geodomaining: A semi-automated, machine learning-based method to delineate spatial domains more effectively than conventional approaches.
• Domain Boundary Delineation: A geostatistical approach to improve accuracy in defining mineralized zones.
• Combinatorics-Based Visualization: A technique that enhances data interpretation and decision-making.
• High Automation Potential: The workflow supports scalable, practitioner-agnostic resource estimation, reducing reliance on manual interventions.

By integrating these elements, the workflow enables faster, more accurate, and replicable resource estimation—a necessity as geodata sets become larger and more complex in modern mining operations. The study highlights the importance of geodata science, which explicitly accounts for the spatial nature of data, distinguishing it from generic data science approaches.

Ultimately, this research provides a scalable and efficient framework for the mining industry, allowing for better remote monitoring, short-term operational feedback, and improved decision-making in mineral resource estimation.