Automated prospectivity analysis for intrusion-related mineral systems in the Charters Towers-Ravenswood region
Arianne Ford
2
Outline
• Aims and scope • Mineral system models • Data • Prospectivity mapping
Aims and Scope
• Evaluate available data • Develop targeting criteria for intrusion-
related mineral systems – Specific focus on geochemistry
• Generate prospectivity maps • Develop a method for ranking existing
targets
Study Area
Mineral Systems
• Deposits are focal points of much larger mass flux and energy systems
• Focus on critical processes that must occur to form a deposit
• Allows identification of mineralization processes at all scales
• Not restricted to particular geologic settings/deposit type
Charters Towers-Ravenswood Mineral Systems
• Siluro-Devonian – Intrusion-hosted orogenic gold system – Mineralization hosted in Silurian granites – Mineralization dates are Devonian – ?Source?
• Permo-Carboniferous – Intrusion-related gold system – Mineralization related to sub-volcanic intrusions
of Kennedy Igneous Association
Data
• Geochemistry • Geology • Geophysics
Geochemistry • Terra Search/GSQ data
– SSS; RC; Soils – Lots of data points
• limited multi-element analysis • Patchy spatial distribution
– Works well for mapping out the Ravenswood Batholith using automated methods
• Carpentaria Gold (Resolute) data – RC and Soils have good sample density over
Ravenswood district – Good multi-element geochemical analysis
Geochemistry maps
Carpentaria Gold Geochem
Geochemistry – data analysis
• Focus on the data provided by Carpentaria Gold – Best distribution of sample points over
Ravenswood district (soil and rock chip) – Comprehensive multi-element analysis
• Data levelled for geology • Determine element associations for each
camp • Produce anomaly maps for district-scale
prospectivity mapping
Geochemistry – PCA results
• Assess element associations – Within camp – Regionally
• Limitations – Needs sufficient sample density and distribution – Interpolating between sparse points – Requires multi-element analysis for each
sample -> hence use of Carpentaria Gold data
Regional PCA (soil)
[As Au Bi Mo Pb Sb] +/- [Ag Cu Te W]
Geochemistry - zonation
• Only works in camps with sufficient data points
• Inconsistent zonation trends between camps – No characteristic zonation, each IRG camp is
different
Geology
• 1:100,000 scale mapping – Surface geology – Solid geology interpretations – Structures
• Camp scale mapping – Variable resolution: ~1:5,000 – Detailed, but patchy and limited spatial extent – Geology and structures
Geology - maps
Geophysics
• GSQ and GA – Magnetics – Gravity – Radiometrics
• Data reprocessed to create derivative products
• New interpretations – Revised structural maps – Intrusion mapping
Geophsyics maps
Intrusion detection • Automated tool in Geosoft for mapping
intrusions from aeromagnetic data
• Limitations – Resolution of data – “Circularity” – Age?
Human expert at best
Automated detection
Ground truth
Images: Geosoft; Holden et al., 2012
Intrusion detection
Statistical Correlations
• Results need to be statistically significant AND geologically meaningful to be useful for exploration
Layer Max. Contrast* % Deposits % Total Area Favourable
CG soil geochem PCA 1.6098 36% 96%
CG RC geochem PCA 1.8018 25% 92%
Dyke density 1.425 17% 6.5%
Permo-Carb intrusions No statistically significant result
* Contrast ≥ 0.5 is meaningful
• Statistically assess relationship between targeting criteria and known IRG occurrences
Prospectivity Analysis
• Integrating what worked for finding major deposits – Permo-Carb sub-volcanic intrusions
• Mapped and interpreted from geophysics – Mapped dykes – Principle component analysis using Carpentaria
Gold geochemistry data
Prospectivity map
Campoven
Three Sisters
Conclusions
• Tried lots of things that didn’t work – Data needs to be fit for purpose
• Resolution of geophysical surveys • Geochemical survey sample density and elements
analysed in the lab – This can be seen as a positive outcome
• We need to know what works, but we also need to know what doesn’t work; this is rarely discussed
Conclusions • Highlights that even with expert data
analysts working on the problem, a purely data-driven or automated approach doesn’t work here – Needs expert opinions to drive the model – We welcome input from industry
• If you want to get involved in a round-table discussion, we are tentatively planning a meeting for October/November in Townsville – Please email Arianne Ford or Vladimir Lisitsin – [email protected] or [email protected]
Acknowledgements
Geological Survey of Queensland
Resolute Mining Ltd.