Tom Wise
Geological Survey of South Australia
Download this article as a PDF (3.8 MB); cite as MESA Journal 90, pages 36–41
Published August 2019
Introduction
Prospectivity modelling can be a powerful predictive tool in mineral exploration targeting. As ‘big data’, artificial intelligence, and ‘intelligence amplification’ trends are becoming more relevant to mineral exploration (Hronsky and Kreuzer 2019), varying forms of data and knowledge integration can be applied to predict where mineral deposits may form. Prospectivity modelling is built on the concept of a mineral system, which is a description of a number of favourable physiochemical processes for the formation of ore deposits (Wyborn, Heinrich and Jaques 1994). Mapping these processes using spatial data as proxies enables spatial integration and development of a prospectivity ranking (e.g. Ford et al. 2019). A variety of methodologies have been developed to combine spatial data types, including artificial neural networks, fuzzy logic (Niiranen, Nykänen and Lahti 2019) and weights of evidence (Rodriguez-Galiano et al. 2015).
The scale at which prospectivity modelling can be employed varies from area selection at craton, terrane or province scale; to a prospective fairway within a belt of known deposits; to the camp or mine scale as a guide for resource estimation. The appropriate scale and methodology used in prospectivity modelling is dependent on the purpose of work. For example, regional-scale studies are useful for government agencies in land use and infrastructure planning, as well as for mineral explorers new to a region (Ford et al. 2019). Smaller scale studies may be used for direct targeting of exploration drilling.
This article presents a weights of evidence mineral prospectivity analysis of the Olympic Domain, within the Olympic Cu–Au Province in the eastern Gawler Craton. This metallogenic province is host to iron oxide – copper–gold (IOCG) deposits including Olympic Dam and Prominent Hill (Fig. 1). As much of the province is buried beneath younger sedimentary cover, mineral prospectivity analysis in this ‘fairway’ is particularly relevant in enabling ranking of camp- to deposit-scale geophysical features. The analysis presented here integrates regional gravity and magnetic potential field data, magnetotelluric data and basement geological interpretations. A knowledge-based weights of evidence approach (Bonham-Carter et al. 1989) was chosen as the IOCG deposits within this region comprise a wide variety of geological hosts, structural settings and alteration/mineralisation styles (Reid 2019). The high degree of variability in deposit controls and manifestation, coupled with variable data density, renders an evidence-based approach using training datasets (e.g. Ford et al. 2019) unsuitable. Therefore, applying a simple mineral systems approach (Wyborn, Heinrich and Jaques 1994; McCuaig and Hronsky 2014) using regional-scale data with unbiased or uniform coverage represents the most appropriate methodology for this fairway.
This study represents an update from previous work of Skirrow, Schofield and Connelly (2011), who assessed areas of interest for U-rich IOCG deposits across east-central South Australia. Skirrow, Schofield and Connelly (2011) used a variety of data types, including mapped and interpreted geology, regional geophysics and point-source information from drillholes. Recent improvements to the interpreted basement geology of the region (Wise, Cowley and Fabris 2015), geophysical anomaly delineation (Katona and Fabris in press) and lithospheric-scale magnetotelluric imaging (Thiel et al. 2016) have warranted an update to prospectivity models across the Olympic Domain.
Methodology and input datasets
Weights of evidence modelling requires key geological ore-forming criteria (mineral system components) to be represented by proxies in spatial data (e.g. Hronsky and Kreuzer 2019; Ford et al. 2019). Figure 2 details the geological criteria and spatial proxies used in this prospectivity model, and associated weightings. Weightings applied are subjective and represent the relative importance of each spatial input at the scale of investigation, as defined by the user. For this study, similar weightings have been applied for most of the geological input criteria (Fig. 2), with slightly higher weightings for basement structures and the presence of mafic rocks, common themes across many deposits. Weighting for cover thickness information (Cowley et al. 2018) was applied on a sliding scale reflecting the continuous nature of this dataset.
Geological polygons were extracted from Wise, Cowley and Fabris (2015) and Cowley (2006), with buffers applied to plutons (5 km) and fault structures (1 km for basement faults, 5 km for crustal-scale structures). Buffers are intended to represent both the spatial inaccuracy in interpretation, as well as a zone of influence of these features, such as advective heat transfer away from a pluton, or lower order fractures adjacent to a fault. Given that the solid geology datasets used to provide the geological inputs to modelling are themselves interpretations of potential field and, where available, drillhole data, the reliability of interpretation is greater in areas with more constraints. In the Olympic Cu–Au Province the best constraints on basement lithology derive from exploration drillholes. Extrapolation from known mineralised areas is a key factor in the prospectivity modelling process, so is therefore highly reliant on the quality of the input geological interpretation.
The gravity anomalies have been adapted from Katona and Fabris (in press) and Katona, Wise and Reid (2018), and represent the spatial extent of residual anomalies extracted from regional data. Katona and Fabris (in press) discuss the application of statistical clustering as a method of identifying significant anomalies within the eastern Gawler Craton. Using a cluster and outlier analysis, Katona and Fabris (in press) concluded that high-amplitude gravity anomalies within a cluster of lower magnitude anomalies (their High-Low cluster type) represent the key outlier anomalies within the Olympic Domain, and correspond to many IOCG deposits. These High-Low cluster anomalies have been utilised here, and weighted higher than other clustered or not statistically clustered anomalies.
Magnetotelluric data has been used to provide a measure of the conductivity of the lower crust, a proxy for fossil fluid enrichment and pathways toward the upper crust. AusLAMP magnetotelluric data from Thiel et al. (2016), accessed through the South Australian Geophysical Reference Model (van der Wielen et al. 2016), has been used in this study. A 35 km depth slice from this model was used to represent relative fertility and permissive lower crustal architecture. Heinson et al. (2018) discuss utility of magnetotelluric data in a mineral systems framework with reference to a 2D section across Olympic Dam, and the concepts behind prospectivity modelling are derived from these interpretations.
Several input criteria fulfill multiple parts of the mineral system classification. For example, Hiltaba Suite granitic plutons act both as a thermal driver (energy source) and as a carrier of metals/fluids. This dependency naturally leads to a summative weighting placing greater importance on criteria that represent multiple parts of the mineral system (Hronsky and Kreuzer 2019).
Results and model reliability
The results of a summative combination of each weighted spatial layer produces a map detailing the relative IOCG prospectivity across the Olympic Domain based on the input criteria (Fig. 3). Warm colours represent greater prospectivity, whereas cooler colours (blue) represent less prospective regions. The main deposits through the central Olympic Domain are highlighted by two main WNW-trending tracts, primarily reflecting where Gawler Range Volcanics and Wallaroo Group metasedimentary units sit either side of a block of the Paleoproterozoic Donington Suite between the Carrapateena deposit and Murdie/Torrens prospects (Fig. 3; Wise, Cowley and Fabris 2015).
As displayed in Figure 3a, the geological and preservation input criteria (see Fig. 2), when combined, highlights all major deposits, and represents a narrowed search space compared to targeting on gravity anomalies alone (Fig. 3b). Combining geological inputs with potential expressions of IOCG mineralisation (gravity anomalies) provides the best rationale for target selection, as the spatial association of favourable geological criteria and geophysical expressions (Fig. 3c) represents a further refined search space.
Aside from the central Olympic Domain, the lower Yorke Peninsula and regions east of Prominent Hill and east of Middleback Range at the top of the Spencer Gulf score highly and are without known deposits. In particular, the area east of Prominent Hill is relatively underexplored.
As a measure of the reliability of the final model (Fig. 3c), prospectivity values for the map cell containing ~20 IOCG deposits (recorded in MINDEP, South Australia’s mineral deposits database) have been extracted and plotted against a histogram of total map cell values (Fig. 4). Key deposits have prospectivity map cell values of >4. As each deposit plots at the ‘more prospective’ end of the total map cell population, this is a measure of the success and reliability of the prospectivity model.
The map score for each deposit represents how well that deposit is represented by the input criteria. For example, the gravity anomaly associated with the Prominent Hill mine has been spatially clustered along with anomalies in the neighbouring Mount Woods Domain, rather than the higher weighted anomalies in the rest of the Olympic Domain with which it has more in common (e.g. low metamorphic grade; Reid and Fabris 2015).
Carrapateena likely scores lower than expected due to a lack of representation of Hiltaba Suite present beneath the basement surface. Given significant volumes of Hiltaba Suite magmatism interpreted across much of the central Olympic Domain from potential field imagery (e.g. Gow, Wall and Valenta 1993) and seismic sections (Drummond et al. 2010; Wise et al. 2015) are not recorded in basement surface geological interpretations, the footprint of the Hiltaba Suite is under-represented in these models. Addressing this lack of a spatial proxy for Hiltaba Suite plutonism at depth is an avenue for future work, as is the spatial incorporation of the role that volumetrically minor mafic magmatism (e.g. Wade et al. 2019) has in the generation of IOCG deposits.
Conclusion
Prospectivity modelling using the weights of evidence methodology has been applied to the Olympic Domain in the eastern Gawler Craton. The resulting models highlight the major IOCG deposits as well as underexplored regions and validate the weights of evidence approach at this domain scale. Incorporation of geological proxies for mineral system components and integration with geophysical anomalism provides a method for narrowing the search space for IOCG deposits beyond targeting gravity anomalies alone, and eliminates many false positives. This ‘intelligence amplification’ process provides a pragmatic framework for scale-reduction from a domain of proven prospectivity to the camp scale.
The resultant prospectivity map can be accessed via SARIG as a downloadable PDF and will also be available as a map layer (go to Explore map by theme, Projects and Products).
References
Bonham-Carter GF, Agterberg FP and Wright DF 1989. Weights of evidence modelling: a new approach to mapping mineral potential. In FP Agterberg and GF Bonham-Carter eds, Statistical applications in the earth sciences, Geological Survey Canada Paper 89-9:171–183.
Cowley WM comp. 2006. Solid geology of South Australia, Mineral Exploration Data Package 15. V 1.1. Primary Industries and Resources South Australia, Adelaide.
Drummond B, Lyons P, Goleby B and Jones L 2006. Constraining models of the tectonic setting of the giant Olympic Dam iron oxide–copper–gold deposit, South Australia, using deep seismic reflection data. Tectonophysics 420:91–103.
Ferris GM, Schwarz MP and Heithersay P 2002. The geological framework, distribution and controls of Fe-oxide Cu-Au mineralisation in the Gawler Craton, South Australia, Part I – Geological and tectonic framework. In TM Porter ed, Hydrothermal iron oxide copper-gold and related deposits: a global perspective, vol. 2. PGC Publishing, Adelaide, pp. 9–31.
Ford A, Peters KJ, Partington GA, Blevin PL, Downes PM, Fitzherbert JA and Greenfield JE 2019. Translating expressions of intrusion-related mineral systems into mappable spatial proxies for mineral potential mapping: case studies from the Southern New England Orogen, Australia. Ore Geology Reviews 111:102943.
Gow PA, Wall VJ and Valenta RK 1993. The regional geophysical response of the Stuart Shelf, South Australia. Exploration Geophysics 24(3-4):513–519.
Heinson G, Didana Y, Soeffky P, Thiel S and Wise T 2018. The crustal geophysical signature of a world-class mineral system. Scientific Reports 8(1).
Hronsky JMA and Kreuzer OP 2019. Applying spatial prospectivity mapping to exploration targeting: fundamental practical issues and suggested solutions for the future. Ore Geology Reviews 107:647–653.
Katona LF and Fabris A in press. Defining geophysical signatures of IOCG deposits in the Olympic Cu-Au Province South Australia, using geophysics, GIS and spatial statistics. In L Corriveau, AH Mumin and EG Potter eds, Iron oxide copper-gold (Ag-Bi-Co-U-REE) and affiliated deposits, Short Course Notes. Geological Association of Canada.
McCuaig TC and Hronsky JMA 2014. The mineral system concept: the key to exploration targeting, Special Publication 18. Society of Economic Geologists, pp. 153–175.
Niiranen T, Nykänen V and Lahti I 2019. Scalability of the mineral prospectivity modelling – an orogenic gold case study from northern Finland. Ore Geology Reviews 109:11–25.
Reid AJ and Fabris A 2015. Influence of pre-existing low metamorphic grade sedimentary successions on the distribution of iron oxide-copper-gold mineralization in the Olympic Cu-Au Province, Gawler Craton. Economic Geology 110:2147–2157.
Rodriguez-Galiano V, Sanchez-Castillo M, Chica-Olmo M and Chica-Rivas M 2015. Machine learning predictive models for mineral prospectivity: an evaluation of neural networks, random forest, regression trees and support vector machines. Ore Geology Reviews 71:804–818.
Skirrow RG, Bastrakov E, Davidson G, Raymond OL and Heithersay P 2002. The geological framework, distribution and controls of Fe-oxide and related alteration, and Cu-Au mineralisation in the Gawler Craton, South Australia, Part II: Alteration and mineralisation. In TM Porter ed., Hydrothermal iron oxide copper-gold and related deposits: a global perspective, Vol. 3. PGC Publishing, Adelaide, pp. 33–47.
Skirrow R, Schofield A and Connelly DP 2011. Uranium-rich iron oxide-copper-gold. In DL Huston and SE van der Wielen eds, An assessment of the uranium and geothermal prospectivity of east-central South Australia, Record 2011/34. Geoscience Australia, Canberra, pp. 37–68.
Wade CE, Payne JL, Barovich KM and Reid AJ 2019. Heterogeneity of the sub-continental lithospheric mantle and ‘non-juvenile’ mantle additions to a Proterozoic silicic large igneous province. Lithos 340–341:87–107.
Wyborn LAI, Heinrich CA and Jaques AL 1994. Australian Proterozoic mineral systems: essential ingredients and mappable criteria. Australasian Institute of Mining and Metallurgy Annual Conference, Proceedings. Australasian Institute of Mining and Metallurgy, Melbourne, pp. 109–115.
Download the map
Prospectivity modelling of the Olympic Cu-Au Province (7.7 MB)