Factors contributing to metal endowment in the western Wabigoon and southern Abitibi subprovinces: a machine learning approach to Precambrian greenstone belts
Date
2023-09-20
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Mineral exploration workflows are including more quantitative techniques and probabilistic
targeting to capture subtle or convoluted relationships to gain insight about geological processes.
However, related methodological improvements often focus on efficiency and sensitivity, leaving
geologically representative feature engineering underdeveloped. Improving modeling capabilities
alone is insufficient, and both geological plausibility and representation of complex processes or
features are critical to generate robust predictive models. Geological features involved in Magmatic
Ni-Cu-PGE, Volcanogenic Massive Sulfide (VMS) Cu-Zn-Pb-Ag(-Au), and Orogenic Au mineral system
prospectivity from two Archean greenstone belts from the southern Superior Province near Timmins
and Dryden, Ontario are examined and compared using a variety of statistical techniques.
Specifically, this PhD research explores 1) current knowledge and characteristic geological features
for both greenstone belts, 2) methods to enhance geological knowledge using whole rock
geochemistry, 3) methods to reduce bias and improve repeatability when mapping structural
complexity, 4) how data science and geological understanding of mineral systems can be integrated
for enhanced feature engineering 5) which factors control, or are most strongly associated with
mineralization, and 6) why the greenstone belts near Timmins and Dryden, Ontario have contrasting
orogenic Au endowment. Data-related outcomes of this research include multi-disciplinary
geoscientific databases (e.g., structural, field observations, geochemistry), new bedrock geology
maps for each area, and reprocessed aeromagnetic grids. Methodological outcomes of this research
include new geochemical classification diagrams for ultramafic to felsic (including tonalite-
trondhjemite -granodiorite and lamprophyre) Archean igneous rocks, Igneous Rock Favorability
indices, automatic mapping of structural complexity from bedding measurements and aeromagnetic
lineaments, mapping pre-deformation fluid path distances, mapping rheological and chemical
contrast, semi-discrete interpolation of characteristic element ratios, as well as mapping mobile
element gain/loss. Geological knowledge outcomes include the importance ranking of factors
controlling magmatic, volcanogenic, and orogenic prospectivity from random forests as well as
geological insight about contrasting orogenic Au endowment in the Timmins and Dryden areas.
Overall, this research demonstrates the importance of integrative studies that leverage multi-
disciplinary data, methods, and knowledge to improve existing geological understanding, maximize
data utility, and generate robust exploration targets. These improvements may enhance exploration
under difficult conditions, such as in data sparse environments, regions affected by clustered/partial
data representation, or targets under cover.
Description
Keywords
Mineral exploration, machine learning, geochemical classification, structural complexity, magmatic Ni-Cu-PGE, volcanogenic massive sulfide Cu-Zn-Pb-Ag(-Au), orogenic Au;, Southern Abitibi, Western Wabigoon