Nowadays scientists get to work with a wide range of raw and processed data, geological interpretations, and ever-evolving concepts. Increasing knowledge and data demands effort and resources, especially when a high quantity of historical data is available to be brought into a real 3D interpretation environment. New technologies and a novel approach are essential but not simple to implement.
In early exploration phases, the integration of all available data and the construction of a Common Earth Model are tools to better represent the current knowledge and help de-risk exploration targeting. The geological concepts and mineral system are translated to geological objects and grids to be used for constrained inversion and mineral system targeting. However, in the case of a well-explored deposit, a detailed structural interpretation may be important to identify the frequency and trend of high-grade mineralization to better target higher grades.
Using modern exploration techniques such as a high-quality exploration model driven from a detailed structural and geological interpretation, data integration, geophysical inversion, and Artificial Intelligence, improves unbiased mineralization prediction within a data-driven targeting workflow.
”“Professionals, geologists can have bias, and our own team is very close to the project. By working with Mira Geoscience, we were able to get outside experts, across multiple disciplines, looking at our project with no constraints, no preconceived notions about what they might find, and it’s been very instructive.”Hugh Agro, P.EngPresident & CEO of Revival Gold Inc
Regional litho-structural model developed from a combination of geological interpretation, drilling data, surface mapping and geophysical modelling.
Case study: Revival Gold Inc.’s Beartrack – Arnett Gold Projects, Idaho, USA
The Beartrack – Arnett Gold Projects is a great example of how Mira Geoscience works towards producing an unbiased targeting model that clients can trust through detailed structural and geological interpretation, data integration, geophysical inversion, and Artificial Intelligence. The exploration targeting in this case was conducted using a combination of a knowledge-driven weighted average predictor and a Random Forest predictive model. The feature engineering was conducted in close collaboration with Revival Gold’s geology team to better capture their knowledge of the local geology and mineral system. The Random Forest algorithm for mineral prospectivity index (MPI) prediction offers great resilience to noisy data and the resulting predictive model is highly interpretable, providing insight into the key exploration features for a given deposit type.
The conduct the modelling phase of the project, Mira Geoscience incorporated all drilling data, including historical blast hole drilling from the Beartrack mine, geological, geochemical, and geophysical data. The resulting three-dimensional model that was generated served as a basis for constrained inversions and exploration vectoring. Once a 3D model was constructed and geophysical data was inverted using constrained inversion, Mira conducted a literature review and mineral system analysis to identify key exploration vectors for the observed mineralization at Beartrack and Arnett. The mineral system analysis is put forward to help in elaborating a list of key model objects and conduct feature engineering to provide as input to the MPI modelling.
3D targets generated by combining knowledge and a data-driven workflow for Mineral Prospectivity Index (MPI).
The MPI modelling phase was conducted as a two-prong approach with an initial phase of knowledge-driven targeting using a multiple index overlay. This first instance of the MPI modelling represents the current knowledge of the geology team. The second phase of MPI modelling used a data-driven approach using a Random Forest predictive model. To build this model, the current drilling data was used as learning examples and the predictions were made on the remainder of the 3D Voxet. The MPI resulting from the data-driven approach represents the data signature of the different deposits and aims at finding similar signatures elsewhere on the property. The two MPI models were then combined into a final 3D MPI representation to be used by Revival to conduct their exploration targeting.
The three-dimensional MPI model was then successfully used to identify several new or under-appreciated exploration targets as well as to generate new ideas about the controls on mineralization at Beartrack-Arnett.
The key modelling approaches at Beartrack – Arnett rely on the use of all available data in conjunction with the regional geological and structural knowledge from different sources to produce an accurate representation of the underlying geology. This base model was then refined by integrating the information derived from the inversion of the magnetic and IP data. This approach enables the team to identify new possible sources for the hydrothermal fluids by highlighting the existence of several low-magnetic residuals in close proximity to the known mineralization identified in drilling and surface sampling. These findings were crucial in furthering the understanding of the mineralized system and played a key role in developing novel exploration vectors.
Local structural model for Arnett and Beartrack (top) with overlain 2D MPI index in transparency. The longitudinal section shows the gold distribution along the Panther creek fault combined with the new structural interpretation.
Notes on software
GOCAD Mining Suite (GMS) is the only 3D interpretation environment that allows the creation of plausible geologic models that honour all geoscientific data, whether the data are sparse or dense. It performs especially well in the context of a large amount of borehole information to constrain 3D geological, geochemical, and geophysical data modelling. It provides the tools for developing 3D models in a controlled manner while being able to integrate and interrogate all data sets effectively.
Artificial intelligence at Mira Geoscience:
At Mira Geoscience, we are dedicated towards developing efficient customized solutions to geological problems using different machine learning approaches. Since not all problems are equal, solutions must be customized for efficiency. A great emphasis is put on the geoscientific development and understanding of the datasets prior to the application of machine learning techniques. We offer the right combination of geoscientists and data scientist to solve targeted problems and hep move exploration forward.
Steve is an exploration geologist with over thirty years’ experience managing and developing exploration projects. He was most recently responsible for delineating a 30-million-ounce silver resource at Soltoro’s El Rayo project located in Mexico. Soltoro was acquired by Agnico Eagle Mines Limited in early 2015. Previously, Steve served as Exploration Manager for MinCore Inc. and in various positions with Yukon-Nevada Gold Corporation, A.C.A. Howe International Limited, Queenstake Resources Ltd. and Monarch Resources Ltd. In addition to the El Rayo project, he managed exploration on the advanced-stage Magistral gold deposit and the Tameapa copper-molybdenum porphyry deposit and was involved in exploration programs at the Jerritt Canyon mine property. Steve holds a B.Sc. in Geology and completed his M.Sc. in Geology at the University of Idaho. He is a Qualified Person as defined by NI 43-101.
Jean-Philippe pursues new applications of machine learning to overcome geological and geophysical challenges by combining geological knowledge with both supervised learning and deep learning. He brings 15 years of mineral exploration experience including expertise in geostatistics applied to structural, geological, and geochemical modelling and interpretation. Jean-Philippe has developed multiple workflows and novel approaches to reduce interpretational risks of geological data interpretation. He has a wide range of experience in mineral resource estimation for precious metals, base metals, and industrial minerals across diverse geological environments around the world. In 2016, Jean-Philippe has pioneered the application of Machine Learning to the mineral exploration industry in winning the Integra GoldRush challenge by application of machine learning to mineral deposit targeting.
Stanislawa is an experienced geophysicist with over 10 years of experience in 3D modelling, data integration and interpretation, and QA/QC. She is Mira’s leading expert in geologically-constrained geophysical interpretation and inversion of DC/IP data, in addition to her long and varied experience with electromagnetic and potential fields data. She gained substantial field experience in geophysical data acquisition early on in her career. She now specializes in modelling and integrating complex geophysical data sets for a variety of commodities and deposit styles, from grassroots to near-mine exploration projects around the world. Stanislawa’s project scope varies from data processing, plate modelling, and depth to basement interpretation, to 3D geologically-constrained geophysical inversion. Stanislawa is an experienced trainer in 3D geophysical modelling in GOCAD Mining Suite. She has contributed to software development for Mira Geoscience’s ANALYST platform as well as inversion algorithms (VPmg/SimPEG).
Shaun is an exploration geologist with over 10 years of experience in exploration and geological modelling in varied environments, including Cordilleran arcs and intrusions, Archean greenstone belts, bimodal volcanic systems, and mafic-ultramafic intrusives. Shaun’s focus is on geological models, from the mine-scale to regional-scale, the empirical controls on mineralization, and the use of geological models to solve problems like mineral prospectivity and geological uncertainty. He is skilled in the geological interpretation of geophysical data including magnetics, gravity, EM, and IP. He is an expert in the application of Python and the development of custom solutions for geological data processing, analysis, and machine learning for mineral prospectivity and knowledge discovery. Shaun’s approach to modelling is to develop an understanding of each new mineral system in consultation with client experts and through deep analysis of the available geological and geophysical data. He uses his extensive modelling experience, data science skills, and understanding of mineral deposits and geological data to transform data into constraints and design geological models that help clients accomplish their exploration goals.
Thomas is an experienced geophysicist with skills in interpretation for mineral exploration. His systematic approach to geophysical data processing and interpretation has evolved into 3D geologically-constrained inversion modelling and integrated interpretation of potential fields data. Thomas has developed numerous novel approaches to data processing and preparation to help guide sensible constrained inversion and target modelling. His expertise in data pre-processing and scripting are key contributing factors to the success of the team. Thomas is an experienced trainer in 3D geophysical modelling in GOCAD Mining Suite and VP Geophysics Suite.