Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac-Larder Lake Fault

Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 © 2011 Canadian Institute of Mining, Metallurgy and Petroleum. All rights reserved. Printed in Canada. 0964-1823/00 $17.00 + .00 Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault O. Rabeau1,†, M. LegauLt2, a. CheiLLetz3, M. JébRak4, J. ROyeR3, and L. Cheng5 (Received April 9, 2008; accepted February 7, 2011) Abstract—By compiling geological, structural, geophysical, and geochemical information into a 3D geological model, we evaluated the orogenic gold potential in the vicinity of a hidden segment of an important Archean fault zone, the Cadillac–Larder Lake fault (CLLF) in the region of Rouyn-Noranda. The segment of CLLF in the present study is partly covered by Proterozoic sedimentary rocks. Because more than 2000 t Au have been extracted along the CLLF to date, our objective is to evaluate the gold potential at depth along a poorly known segment of this fault. A 3D geological model (50 km × 9 km × 1.5 km) including the covered segment was built through the compilation and homogenization of available geological data and the construction of 23 cross sections. The geology under the Proterozoic cover was evaluated using geophysical inversions, drill holes (42 in total), and surrounding geology. All available assays were iltered and upscaled to a 250 m × 250 m × 250 m regular cell grid to determine and quantify spatial relationships between geological features and mineralized occurrences using the weights of evidence method. Structural features, such as E–W-trending faults and fault intersections, and certain lithologies with a high primary porosity such as volcanoclastic rocks of the Blake River Group and Timiskaming sedimentary rocks, proved to be very prospective, yielding favourable factors with a weight of evidence index W+ > 0.24. These salient features were then assigned a combination index for ultimately evaluating the orogenic gold potential under the sedimentary cover. The zones resulting in an optimization of exploration targeting were attributed the highest probability, representing ~1% of the initial volume. © 2011 Canadian Institute of Mining, Metallurgy and Petroleum. All rights reserved. Key Words: Orogenic gold, 3D geomodelling, Spatial analysis, Cadillac-Larder Lake fault, Mineral potential map. Sommaire — La compilation des données géologiques, géophysiques, structurales, et géochimiques et leur intégration au sein d’un modèle géologique 3D a permis d’évaluer le potentiel aurifère dans les environs d’une importante zone de faille archéenne sous couverture protérozoïque. Cette méthodologie a été développée en utilisant le cas de la Faille Cadillac–Larder Lake (FCLL) dans la région de Rouyn-Noranda. Puisque plus de 2000 t d’Or ont été extrait à proximité de la FCLL, l’objectifs des travaux est l’évaluation du potentiel aurifère en profondeur de ce segment. Un modèle géologique 3D (50 km × 9 km × 1.5 km) incluant le segment sous couverture de la FCLL a été construit en utilisant l’ensemble des données géologiques disponibles et 23 coupes géologiques. La géologie du socle archéen sous la couverture protérozoïque a été interprétée avec des inversions géophysiques, des forages (42 au total) et les contacts géologiques exposés. Les teneurs aurifères dans le secteur ont été iltrées et ajusté à une grille régulière de 250 m × 250 m × 250 m ain d’évaluer et de quantiier les relations spatiales existantes entre certaines entités géologiques et les emplacements minéralisés en utilisant la méthode du poids de la preuve. Les failles de direction Est-Ouest, les intersections de failles ainsi que certaines lithologies possédant une porosité primaire importantes comme les certains tuf du Groupe de Blake River et les sédiments du Timiskaming se sont montrés très prospectifs avec des valeurs de W+ > 0.24. Ces entités géologiques ont ensuite été utilisées par le biais d’un indice de combinaison pour évaluer le potentiel en or de type orogénique sous le couvert sédimentaire. Cette méthodologie permet une optimisation des campagnes d’exploration puisque les zones pour lesquelles la probabilité de minéralisations a été évaluée à la plus haute valeur représentent 1% du volume initial. © 2011 Canadian Institute of Mining, Metallurgy and Petroleum. All rights reserved. 1 Unité de recherche et de services en technologie minérale, Université du Québec en Abitibi-Témiscamingue, Rouyn-Noranda, Québec J9X 5E4. Ministère des Ressources naturelles et de la Faune du Québec, Rouyn-Noranda, Québec J9X 6R1. 3 Nancy Université, Centre de Recherche Pétrographique et Géochimique, France. 4 Département de Sciences de la Terre et de l’Atmosphère, Université du Québec à Montréal, Montréal, Québec H2X 3Y7. 5 Département de Sciences Appliquées, Université du Québec en Abitibi-Témiscamingue, Rouyn-Noranda, Québec J9X 5E4. † Corresponding Author: E-mail: olivier.rabeau@uqat.ca 2 116 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Introduction Mineral discoveries along known metallotects in highly explored regions are becoming fewer and costlier because they commonly require deeper and riskier exploration campaigns. However, when used to their full potential, the large amounts of accumulated data in heavily prospected sectors can enhance mineral exploration by improving targeting for deep-seated deposits. This paper presents an example of optimized use of this data with a 3-dimensional (3D) quantitative method to reduce discovery costs for gold exploration at depth in the vicinity of a hidden segment of an important Archean fault zone. Transcrustal-scale fault zones within Archean greenstone belts are commonly of great economic importance because of their spatial association with orogenic gold deposits (Groves, 1993; Kerrich et al., 2000; Goldfarb et al., 2001; Dubé and Gosselin, 2007). The location and geometry of these major fault zones are crucial for gold prospecting even if a minority of gold deposits directly hosted within these structures. The Cadillac-Larder Lake fault (CLLF; Fig. 1) located in the Abitibi Subprovince of the Superior province, is a typical example of a large transcrustal fault zone that hosts many world-class mining camps (>100 Mt) such as Val-d’Or, Malartic, Cadillac, Larder Lake, Kirkland Lake, and Matachewan (Poulsen et al., 2000). However, gold deposits are not distributed regularly along the CLLF. The segment located between the town of Cadillac and the Quebec–Ontario border has a very poor gold endowment when compared to other segments of comparable length. This lack of known gold deposits can be partially attributable to the Proterozoic sedimentary cover, which masks the CLLF for more than 30 km. The thickness of the sedimentary cover is locally greater than 700 m, rendering typical exploration methods inadequate. The aim of this work is to deine a methodology using the Gocad® 3D modeling software (Mallet, 1992b), to combine geological, geophysical, and geochemical data to identify high potential zones for mineralization at depth along a lesser known segment of the CLLF. A 3D regional geological model for a partly covered 50 km segment along the CLLF and its adjacent structures was built. This model includes a 3D representation of the Archean–Proterozoic unconformity, and an interpretation of the Archean units and structures underneath the cover. The geophysical data were used to validate and interpret the geological model. The regional model provides a platform to conduct a 3D spatial analysis on a compiled gold assay database to identify the geological features controlling the distribution of gold mineralization. The spatial association of these geological features with gold was then quantiied using the weights of evidence method, and a knowledgebase approach was used to generate exploration targets at depth underneath the sedimentary cover using a combination of factors. Geological Setting The Archean Abitibi Greenstone Belt located in the southern part of the Superior craton is composed of an assemblage of plutonic (50%), volcanic (40%), and sedimentary (10%) rocks. The 50 km × 9 km study area is situated along the limit between the Abitibi and Pontiac subprovinces (Dimroth et al., 1982; Couture et al., 1996; Thurston et al., 2008b), which lie respectively north and south of the CLLF (Fig. 2). The northern part of the Pontiac Subprovince is mainly composed of highly folded and deformed turbiditic sedimentary rocks (<2685 Ma; Davis, 2002), with some horizons of maic to ultramaic volcanic rocks and crosscutting syenite and granite. In the Abitibi Subprovince, volcanic rocks range in composition from komatiite to rhyolite. The majority of the volcanic rocks in the study area are part of the Blake River Group, for which an age of 2701 Ma has been reported near the study area (Lafrance et al., 2005). The McWatters Formation, observed south of the CLLF, is composed mainly of volcanic rocks that are suggested to be equivalent to sequences in the Pontiac Subprovince (Morin et al., 1993). The Piché Formation is composed of ultramaic to maic volcanic rocks wedged within the CLLF. The Cadillac Group (<2687 Ma; Davis, 2002; Lafrance et al., 2005; Mercier-Langevin et al., 2007) is located in the northeastern corner of the study area and consists mainly of a mudstone and wacke assemblage. The Archean sedimentary rocks adjacent to the CLLF are part of the Timiskaming Group, and are composed of unsorted polygenic conglomerates, sandstone, and local interdigitations of alkaline volcanic rocks (2677 to 2670 Ma; Corfu et al. 1991; Davis, 2002; Ayer et al. 2005). Many dikes and sills of gabbroic to dioritic composition crosscut the study area. Their relative ages suggests syn-volcanic to syn-tectonic emplacement (Thurston et al., 2008a). Minor alkaline and calc-alkaline intrusions are observed on both sides of the CLLF. An age of 2685 to 2672 Ma was found for some of these intrusions (Corfu et al. 1989; Wilkinson et al., 1999; Davis, 2002). Proterozoic sedimentary rocks of the Gowganda Formation at the base of the Cobalt Group lie unconformably on the Archean basement in the western and central part of the study area. This formation consists mainly of a basal conglomerate overlain by wackes, mudstones, and sandstones, which cover the CLLF and the Archean basement for over 30 km. A Rb-Sr isochron gave an age of 2288 Ma for the Gowganda Formation (Fairbairn et al., 1969). Two inliers in the Cobalt Group expose the Archean basement in the central part of the study area (Legault and Rabeau, 2006). All Archean rocks are crosscut by N–S- and NE-trending Proterozoic diabase dikes. The tectonic evolution of the study area is associated with numerous events of thrusting and dextral transpression (Daigneault et al., 2002; Daigneault and Mueller, 2004). The irst episode is characterized by the accretion of the Pontiac and Abitibi subprovinces (2690–2680 Ma) followed by dextral strike-slip movement along the CLLF, which led to the formation of the Timiskaming Group (2680–2670 Ma). Renewed thrusting (2670–2661 Ma), exhumation of the Pontiac Subprovince (2661–2642 Ma), and late dextral transpression (<2642 Ma) complete the tectonic events along the CLLF. Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. More than one type of mineralization can coexist in a single deposit, as observed in the Kerr-Chesterville deposit, located within the CLLF west of the study area. This deposit (the second largest along the CLLF) consists of both quartz-carbonate vein mineralization (carbonate ore) and replacement type (low ore) (Kishida and Kerrich, 1987; Smith et al., 1990). Most of the gold mineralization considered in this study belongs to a broad class of orogenic Fig. 1. Simpliied geological map along the Cadillac–Larder Lake fault (CLLF) showing the location of past and present gold, copper, and zinc mines. Regional Gold Metallogeny Although no mines are currently active in the sector, 50 t Au have been extracted along the CLLF in the study area (Couture, 1991). Additionally, a world-class deposit, the Kerr-Chesterville (336 t Au produced from 1930– 1996; Guindon et al., 2007), occurs along the western extension of this segment of the CLLF in Ontario. Four different types of gold mineralization have been recognized in the study area (Legault and Rabeau, 2007). 1. Gold-rich quartz-carbonate ± tourmaline ± sulides constitute the most common type (70%). Typical examples are the Stadacona, Astoria, and Senator-Rouyn mines (Couture, 1996). 2. Replacement type deposits (~20%) are associated with strong, pervasive albite/sericite-carbonate alteration and disseminated sulides. The Francoeur and Wasamac deposits are representative examples (Couture and Pilote, 1993). 3. Mineralization spatially associated with syenitic intrusions (~5%) includes two subtypes: quartz-carbonate veins such as the Granada mine (Couture and Willoughby, 1996), and less common disseminated sulphides enriched in Au and Cu (Couture and Marquis, 1996; Legault and Lalonde, 2009). 4. Polymetallic volcanogenic massive sulphide (VMS) deposits (~5%) are also encountered in the study area (Legault and Rabeau, 2007). Additionally, some occurrences of epithermal-type mineralization have been observed north of the study area near the transcrustal Porcupine–Destor fault (Legault et al., 2005), but have not been described near the CLLF. 117 118 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Fig. 2. Geological map of the study area with location of mines and mineral deposits. Modiied from Legault and Rabeau (2006). Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. gold deposit types. These deposits are characterized by epigenetic, structurally controlled mineralization located in accretionary orogens that formed over a large crustal depth range from deep-seated luids (Groves et al., 1998). The quartz-carbonate vein, replacement, and veins associated with syenite type deposits it within this deinition. Orogenic gold mineralization postdates regional metamorphism, plutonism, and early phases of orogenic deformation (Groves et al., 2000; Goldfarb et al. 2001). This, combined with their location within a craton, imply that the present geometry of the host terranes can be considered to be very similar to the one geometry at the time of ore formation. At a local scale displacement of ore might be observed, but at regional scale the displacement is minor. Geological Modeling Geological modeling enhances the geological understanding of a speciic region by allowing interpretations using all available geoscientiic data in a 3D environment. A thorough comprehension of the geological setting and an abundance of quality 3D data are prerequisites for the construction of such models. As opposed to a 2D environment, the 3D approach offers the beneit of avoiding bias generated by projection of data on the surface, and allows the calculation of true distances between objects taking depth and 3D geometry into account (Fig. 3). The geology 119 and structural context of the study area are well deined and its proximity to the proliic Rouyn-Noranda central camp provides access to a multitude of high-quality data for the sector (Table 1). These data were irst compiled with standardized units and referenced using the UTM NAD83 spatial projection. A 3D geological model (50 km × 9 km × 1.5 km) was built based on 23 vertical cross sections (Fig. 4a) at a mean spacing of 2 km. The sections were interpreted from recent geological mapping in the study area (Legault and Rabeau, 2006, 2007) and compiled structural and drill hole data (Table 1). These sections then served as geometrical constraints for building triangulated surfaces representing the boundaries of geological units (Fig. 4b). The creation of triangulated surfaces respecting the range of constraints of available data was carried out using the Discrete Smooth Interpolation method (Mallet, 1992; Fig. 4c). Finally, a volumetric model was derived from the partition of space by triangulated surfaces (Fig. 4c) representing geological contacts and faults. Modeling the Archean–Proterozoic Unconformity and the Covered Archean Units One of the main challenges in the modeling process was the evaluation of the geometry of the unconformity between the Proterozoic cover and the underlying Archean units from structural data and very sparse data derived Fig. 3. Illustration of the advantages of 3D representation versus a 2D map. A 3D approach allows better visualization of spatial relationships as shown by the connection between the intersection of faults 1 and 2 and the mineralized occurrence. A 3D representation also avoids distortion due to projections to surface and data stacking (as seen with the drill hole representation on both images) and offers the possibility of calculating true distances between geological features. 120 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Table 1. Data Used to Construct the Geological Model Data Type Quantity/Type Drill holes >9000 Mining companies, compilation of historical data Gold assays >209 000 Mining companies, compilation of historical data Structural data >1600 Compilation maps (CG 32D/02, 03, 04; Leduc, 1986; Legault and Rabeau, 2006, 2007) Geological map 1:50 000 map Legault and Rabeau (2006) Provincial survey (south) MegaTEM survey (north) Dion and Lefebvre (1997); Xstrata, unpublished data (2002) 1249 stations Dion (1993) 308 100 points USGS (http://eros.usgs.gov/#/Find_Data/Products_and_Data_Available/gtopo30_info) Magnetic map Gravimetric measurements Digital elevation map Source from drill holes penetrating through the sedimentary cover (Fig. 5). A total of 418 surface bedding measurements was used to estimate the shape of the unconformity at depth. The structural data show that the Proterozoic sedimentary formation has a basinlike shape with very shallow dipping borders (Fig. 5a). The geometry of the unconformity was reined using 81 compiled diamond drill holes, of which 42 intersect the base of the Proterozoic cover (Fig. 5b). The surface representing the unconformity was built to accommodate both sets of data giving priority to drill hole control points. The inal estimation shows that, in the study area, the Cobalt Group has a thickness of up to 790 m in its western portion and becomes shallower toward the east where two inliers are present (Fig. 5c). The 42 drill holes that intersect the unconformity also provide valuable information for the interpretation of geology beneath the sediments (Fig. 6a). In combination with magnetic data, this information was used to evaluate the location of contacts of certain highly magnetic units (e.g., ultramaic rocks in the Pontiac Group, and syenites) in sectors where the sedimentary cover is thin, because of the very low magnetic susceptibility of the Cobalt Group (measured to <0.001 SI). The gravimetric data were mainly used to determine the trace of the CLLF (Fig. 7) which separates two contrasting subprovinces in terms of their nature and density: the relatively dense volcanic rocks of the Blake River Group to the north, compared with the less dense sedimentary rocks of the Pontiac Subprovince to the south. The major part of the covered sector is occupied by metamorphosed sediments and ultramaic lavas of the Pontiac Subprovince. The lithologies composing the Blake River Group only occupy the northern part of the covered Archean basement. The Timiskaming sediments are intersected along the CLLF on the full length of the covered Archean basement. They were interpreted on both sides of the CLLF, but only on its hanging wall in the central part of the covered area. The alkaline volcanic rocks of this group are intersected by drill holes over a large area in the western part of the unexposed Archean rocks. Finally, two syenite intrusions have been intersected underneath the Cobalt Group. The interpretation of the nature and geometry of the hidden Archean units is locally based on sparse and heterogeneously distributed data. To distinguish sectors where the interpretations are based on hard data from others, the degree of conidence linked to interpretations was quantiied (Fig. 6b). The two main constraints for the interpretation of the geometry of the Archean–Proterozoic discontinuity and the underlying 3D geological map are drill hole data and observed contacts near the edge of the Proterozoic sedimentary cover. Because the reliability of the interpretations is directly linked to the distance from these control points, this property was calculated and added into the geological model. A high degree of conidence can be attributed to interpretations in sectors colored in red in Figure 6b, which are located less than 100 m from a control point, whereas sectors colored in white are located more than 1.5 km from a control point and imply that the geological interpretation is highly subjective. Geophysical Inversion The gravity and magnetic data were inverted to recover the density and the magnetic susceptibility distribution in three dimensions in order to assist interpretation of the 3D geometry of certain geological entities. The airborne magnetic data of the northern half of the study area were acquired via MEGATEM technology with a 125 m line spacing (Xstrata, unpublished data, 2002), whereas the data from the southern part are from a wider spaced (200 m) public database of the Ministry of Natural Resources and Fauna of Quebec (Dion and Lefebvre, 1997). Ground gravity data are compiled from two databases from the Ministry of Natural Resources and Fauna of Quebec, and the Geological Survey of Canada: distribution of the stations is heterogeneous and mostly located along roads and lake shores (Dion, 1993). The inversion of gravity data deines a density distribution model that provides the best it to the observed data on surface. The inversion algorithm used (GRAV3D) was Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. 121 Fig. 4. a. Location and example of the interpreted cross sections; b. Surface representation of the CLLF being constructed by joining the dip interpretations in each cross section; c. Surface model of the study area (without the Proterozoic sedimentary cover and Archean units underneath). developed at the University of British Columbia (Li and Oldenburg, 1998). The elevated density contrast between the units of the Pontiac and the Abitibi subprovinces allows proper estimation of the dip of the CLLF along the total length of the model. The 3D inversion results conirm that the CLLF is steeply dipping (~75° N) below the Proterozoic cover (Fig. 7b), a result that is consistent with the magnetic data inversion. The inversion of magnetic data allows evaluation of the distribution of the magnetic susceptibility associated with different lithologies. The 3D inversion (University of British Columbia code MAG3D; Li and Oldenburg, 1996) was used to invert the data without considering geological constraints. The non-uniqueness of the results obtained in the magnetic interpretation was addressed by favoring a solution with emphasis on a deeper explanation of the susceptibility distribution. The results of the 3D inversion reined the dip of late NE–SW-trending structures (Fig. 8) such as the Milky Creek, Beauchastel, Horne Creek, and Davidson Creek faults. Spatial Analysis of Mineralized Occurrences Computer-based methodologies such as 3D modeling can be used as predictive tools by uncovering repetitive controls on mineralization. A common method used to delineate prospective areas is to carry out a spatial analysis on a training set of data to identify factors controlling the 122 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Fig. 5. Data used to model the 3D geometry of the Archean–Proterozoic unconformity: a. bedding measurements at the borders of the Cobalt Group (map view); and b. drill holes within the Cobalt sediments. Drill hole intersections of the Archean–Proterozoic unconformity are shown in red, whereas holes that did not intersect the Archean basement are shown in grey (oblique view). c. Model of the thickness of the Proterozoic cover. Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. distribution of known deposits, and then to use this knowledge on lesser-known regions. The 3D geological model was used to test the spatial association of speciic geological features with gold occurrences. These occurrences were determined using 209 000 compiled gold assays. Samples analyzed for precious metals are generally well distributed within the study area, even if certain sectors associated with closed mines or studied deposits are overrepresented (Fig. 9a). In order to properly evaluate the controls on orogenic gold mineralization, the 123 database was iltered to remove assays derived from other mineralization types with distinct characteristics, source, timing, and formation mechanism. These gold occurrences belong to volcanogenic massive sulide (VMS) deposits, porphyry-type deposits associated with syenites, and epithermal deposits. Individual studies of deposits were taken into account to discriminate orogenic gold deposits from other types (Barrett et al., 1992; Couture et al., 1996; Legault and Rabeau, 2007). Furthermore, certain known characteristics of orogenic gold deposits were used to dis- Fig. 6. a. Geological model of the units lying below the Cobalt Group. b. Calculated distance to drill hole data and observed contacts near the edge of the Proterozoic sedimentary cover, the two main constraints for the interpretation the underlying 3D geological map. The distance to these control points can be considered as an indication of the degree of uncertainty in the model. 124 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 card assays linked to other types of gold mineralization. The low salinity of the mineralizing luids involved in the formation of orogenic gold deposits implies low base metal content. In fact, ores typically have negligible values in Cu, Mo or Pb, Sn, and Zn (Groves et al., 2003) as opposed to porphyry or VMS deposits. Orogenic gold deposits also typically have a high Au/Ag ratio. More than 11 200 Cu and 18 000 Ag assay were used to discard 7538 assays considered too rich in base metals or Ag to represent orogenic gold deposits. The discarded assays were mainly located in the vicinity of the Aldermac mine and of the Bay Renaud syenite. It is important to keep in mind that, because copper and silver were not analyzed systematically, it is possible that some assays from different deposit types have been incorrectly treated as orogenic gold deposits. A regular grid covering the entire study area composed of 250 m × 250 m × 250 m cells was used to upscale the assay database in order to diminish the sampling bias on statistical calculations due to data clustering near known deposits (Fig. 9b). The size of cells was determined by iteration in order to deine well-known deposits while avoiding oversampling. The mean, median, maximum, and minimum gold concentrations (ppb) in each cell were calculated. The maximum value of each cell was used to determine two data classes: mineralized and unmineralized occurrences. All cells containing a maximum gold concentration value of more than 950 ppb were considered as mineral- ized occurrences. This threshold was set to discriminate between anomalous gold concentrations and minimum economic values. The unmineralized occurrences were deined by the remaining cells containing gold concentration values below 950 ppb. This process deined 642 mineralized and 1518 unmineralized occurrences. A total of 54 unmineralized occurrences located in Proterozoic lithologies (32 in the sedimentary rocks of the Cobalt Group and 22 in the diabase dikes) were eliminated from the spatial analysis because they postdate the orogenic gold mineralization. Using the 3D geological model and the upscaled and iltered assay database, an analysis was conducted to examine spatial associations of orogenic gold occurrences with pertinent geological features. These spatial associations were quantiied using the weights of evidence method (Bonham-Carter et al., 1988). This method measures the degree of spatial association by a pair of weights (W+ and W−) calculated as shown in Equation 1: W = ln[ P( X |D)/P(X|D )] + W = ln[P( X |D)/P( X|D )] (1) − where D and D represent respectively the presence and absence of an occurrence. P( X |D) corresponds to the probability of having a factor X in association with a mineralized occurrence D, and P( X|D )is the probability of not having factor X in association with an unmineralized occurrence. Fig. 7. a. Map of the residual gravimetric anomaly. b. Section of the unconstrained inversion of the gravimetric data showing the general attitude of the CLLF under the Cobalt Group sediments (the coloring of density values is relative). Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. 125 Fig. 8. Longitudinal section of the unconstrained inversion of the magnetic data, which allows interpretation of the dip of the Milky Creek, Beauchastel, Horne Creek, and Davidson Creek faults at depth. Fig. 9. a. Location of all gold assays and of major deposits: Arntield 1, 2, and 3 (A1, A2, A3); Astoria (AS); Augmitto (Au); Bazooka (BA); Forbex (FO); Fortune Lake (LF); Francoeur 1, 2, and 3 (F1, F2, F3); Granada (GR); McWatters (MC); New Rouyn-Merger (NR); O’Neail-Thompson (OT); Stadacona (ST). b. Upscaled assays with a grid of 250 m × 250 m × 250 m showing mineralized (>950 ppb) and unmineralized occurrences. 126 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 A positive value of W+ combined with a negative value of W− indicates that the tested factor is positive evidence for the presence of orogenic gold. In the case of a negative correlation with gold deposits, W+ would have a negative value and W− would be positive. An absence of spatial association is implied if W+ = W− = 0. This methodology requires a suficient number of reference ore deposits to calibrate the occurrence and nonoccurrence probabilities. To use this methodology, an extensive data training set is required. The weights of evidence method allows evaluation of the degree of spatial association between mineralized occurrences and several geological parameters, such as distance to structural features, host rocks, and distance to intrusive bodies. To quantify the uncertainty attached to each weight, a conidence interval was estimated for each W+ and W− using a similar technique to the one applied for estimating the conidence interval for probabilities. Because weights of evidences W are calculated from the ratio of the conditioned probability, such as p = P(X/D) = P(X∩D)/P(D), it is possible to estimate a conidence interval for the weights using classical rules for propagation of errors with the conidence interval for each probability involved. The conidence interval (± Ɛp) for probability p was estimated using Equation 2 (Kendall, 1997): Ɛp = zɑ/2 [p.(1−p)/n]1/2, (2) where p is the estimated probability, n the sample number, and zɑ/2 is the risk level deined by a normal distribution for a risk of 95% to be in the interval considered (in this case zɑ/2 = 1.96). The conidence interval Ɛr for probability ratio r = p/q was calculated using Equation 3: Ɛr = r [Ɛp/p + Ɛq/q] (3) The lower W− and upper W+ bounds of the conidence interval for the weight of evidence W = ln[p/q] was estimated reporting the upper and lower values of p and q, respectively, in Equation 4: W− = ln [(p−Ɛp) (q+Ɛq)/] W+ = ln [(p+Ɛp) (q−Ɛq)/] (4) Ɛw = [W+ − W−]/2 Lithologies At greenstone belt-scale, orogenic gold mineralization can be hosted by any rock type (Hodgson, 1993); however, some lithologies are more commonly associated with deposits than others (Hodgson and Troop, 1988). This is explained by certain geochemistries that favor precipitation of gold from the hydrothermal luid, or by competency contrasts that inluence the geometry and density of fractures and correlative luid circulation. The 3D regional geological model of the CLLF was used to evaluate and quantify the spatial association of each ma- jor lithology with mineralized occurrences (Table 2). Lithologies considered to be spatially associated considering the uncertainty intervals on the calculated weights are: the ultramaic rocks of the Piché Formation (W+ = 1.45 ± 1.05, W− = −0.02 ± 1.15), the andesitic tuff of the Blake River Group (W+ = 0.84 ± 0.45, W− = −0.06 ± 0.5), and the sedimentary rocks of the Timiskaming Group (W+ = 0.56 ± 0.32, W−= −0.11 ± 0.34). The other volcaniclastic rocks of the Blake River Group have positive W+ and negative W−, but have a low level of conidence according to the calculated uncertainty. The preferential association of gold mineralization with volcaniclastic rocks and weakly metamorphosed sediments illustrates the importance of primary permeability in the formation of orogenic gold deposits. The spatial association of mineralized occurrences within the ultramaic rocks of the Piché Formation is more likely explained by its location within the CLLF, which hosts numerous deposits. In fact, as mentioned above, the Piché Formation is composed of ultramaic volcanic rocks wedged within the CLLF. The spatial association of syenites with gold deposits is well documented (Hodgson and Troop, 1988; Robert, 2001), but the spatial analysis in this study fails to bring out this association. This can be explained by the fact that syenites associated with gold mineralization are mostly of small dimension (less than 3 km2; Legault and Lalonde, 2009), and that these small intrusions tend to be under-evaluated in a regional model, possibly due to the upscaling procedure used to estimate characteristics in the regular cell grid. Furthermore, because these intrusive bodies are also associated with other gold mineralization types, many of the gold assays associated with these intrusions were iltered out. Faults and Fault Intersections Structural features have an important control on orogenic gold deposit distribution. Major fault zones are thought to act as channelways for gold-bearing hydrothermal luids from deeper crustal levels (Kerrich et al., 2000) whereas second- and third-order faults, which host most of the gold endowment, are believed to form dilatational sites favorable for gold precipitation (Eisenlohr et al., 1989). The faults located in the study area can be separated into two distinct groups based on their orientation. The E–W group contains faults having an orientation between N060° and N120°, and represents 66 fault segments. The N–S group contains 94 faults with orientations between N315° and N045°. Each of these groups was tested for their spatial association with orogenic gold occurrences. The distances of the mineralized and unmineralized occurrences from each of the fault groups were calculated in the model. The statistical analysis showed that the mean distance of mineralized occurrences to E–W faults is 284 m, and 75% are within 360 m of one of these faults. A strong spatial association (W+ = 0.44 ± 0.21, W− = −0.71 ± 0.24) can be observed using a maximum 360 m distance between E–W faults and mineralized occurrences (Fig. 10a). The same statistical analysis of the distances between mineralized occurrences and N–S faults indicates a mean distance of 758 m, and much less association with mineralized occur- Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. 127 Table 2. Spatial Associations with Mineralization Lithologies Mineralized Unmineralized W+ ± Ɛw W− ± Ɛw Contrast Pontiac Sedimentary rocks 4 69 −2.02 ± 2.46 0.04 ± 0.61 −2.06 ± 3.07 Basalt 0 3 n.d. n.d. n.d. Ultramaic volcanic rocks 0 2 n.d. n.d. n.d. Rhyolite 75 227 −0.28 ± 0.36 0.04 ± 0.34 −0.32 ± 0.7 Basalt 222 603 −0.17 ± 0.25 0.10 ± 0.24 −0.27 ± 0.49 *Andesitic tuff 62 61 0.84 ± 0.45 −0.06 ± 0.5 0.90 ± 0.95 Basaltic tuff 16 19 0.65 ± 0.82 0.01 ± 0.96 0.66 ± 1.78 Rhyolitic tuff 15 27 0.24 ± 0.77 −0.01 ± 0.81 0.23 ± 1.58 Andesite 0 2 n.d. n.d. n.d. 139 181 0.56 ± 0.32 −0.11 ± 0.34 0.67 ± 0.66 0 25 n.d. n.d. n.d. 15 8 1.45 ± 1.05 −0.02 ± 1.15 1.47 ± 2.2 Syenite 4 37 −1.40 ± 3.17 0.02 ± 0.82 −1.42 ± 3.99 Gabbro 90 199 0.03 ± 0.35 0.00 ± 0.35 0.03 ± 0.7 Felsic intrusions 0 1 n.d. n.d. n.d. *Distance to E–W faults 477 686 0.44 ± 0.21 −0.71 ± 0.24 1.15 ± 0.45 *Distance to fault intersections 316 423 0.51 ± 0.24 −0.33 ± 0.26 0.83 ± 0.5 369 857 0.04 ± 0.22 0.05 ± 0.21 −0.01 ± 0.43 Blake River Group Timiskaming Group *Conglomerate, wacke Trachytic tuff Piché Group *Ultramaic volcanic rocks Intrusive Rocks Structural Features Competency Contrasts Distance to plutons Note: * = favourable features; n.d. = no data. rences. Localized zones where the fracture density is important become sectors of enhanced permeability leading to the circulation of large amounts of mineralizing luids and gold precipitation (Tripp and Vearncombe, 2004). Fault intersections at a regional scale were therefore considered as areas of high fracture density. Fault intersections can be represented in 3D space as lines. The distance between the generated lines and mineralized occurrences was computed to deine the spatial association of fault intersections and gold deposits. A histogram showing the cumulative frequency of mineralized occurrences as a function of the distance to fault intersections shows a break at the median value of 519 m (Fig. 10b). This range was used in the calculation of weights, and conirms a good spatial relationship with weights of evidence (W+ = 0.51 ± 0.24, W− = −0.33 ± 0.26). deformation and creates dilatational sites (Groves et al., 2000). Hence, intrusive bodies can be favourable hosts for gold deposits. The spatial distribution of mineralized occurrences around each intrusive unit was tested and quantiied. The lithologies considered for the calculation were: gabbros, felsic intrusions, and syenites. The distance to each mineralized occurrence was calculated for the closest intrusive body. Although 75% of all the occurrences are within 668 m of an intrusive body (Fig. 11), the calculated weights (W+ = −0.04 ± 0.22, W− = 0.05 ± 0.21) show that there is no spatial association between mineralized occurrences and intrusions, meaning that mineralized and unmineralized occurrences are located at the same distance from plutonic bodies. Distance to Intrusive Rocks It has been noted that the competency contrast of plutonic bodies with their host rock favors fracturing during Hidden orogenic gold deposits can be considered to have common characteristics with known deposits in the same area. The geological model built under the Proterozoic P3D Mineral Potential Mapping Under the Proterozoic Cover 128 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Fig. 10. Calculated distance of mineralized occurrences to: a. E–W faults; and b. fault intersections. Histograms show the cumulative frequency of mineralized occurrences versus the distance to both fault groups, and calculated weights with conidence intervals. cover was used to target areas using the aforementioned inluential geological features and their quantiied spatial associations (Fig. 12). Geological features presenting a positive spatial association (Table 2) were used to deine favourable zones. A combination index having a value between 0 and 1 was assigned to each inluential feature in such a way that the cumulative value of the index is equal to 1 (Table 3). This index allows the integration of the speciic knowledge of the user while considering the degree of spatial association measured by the weights of evidence. The index values presented were established by considering that structural factors are the most signiicant Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. 129 Fig. 11. Calculated distance to intrusions with accompanying histogram showing the cumulative frequency of mineralized occurrences versus the distance to the nearest intrusion, with calculated weights and conidence intervals. because of their inluence on the distribution of orogenic gold deposits, and present the highest spatial association. Most of the closed mines or well-deined deposits are located on a fault zone: Francoeur, Wasamac, Arntield, Astoria, Augmitto, McWatters. A combination index of 0.5 was assigned for cells located at less than 360 m from an E–W fault, and a combination index of 0.25 was assigned to cells located less than 520 m from a fault intersection. In the geological model, mineralized occurrences are hosted in a unique lithology; consequently, cells located in a favorable lithology were assigned an index value of 0.25. The Piché Formation, located directly within the CLLF, was treated distinctly. An index value of 0 was attributed to cells within the Piché Formation because they already attract a 0.5 index assigned to E–W faults due to their spatial association with the CLLF. The degree of conidence linked to geological interpretations below the Cobalt Group sediments (Fig. 6b) was also included in the generation of the 3D gold potential map to account for sectors where interpretations were made with less data. A value determined by the level of uncertainty in interpretations was subtracted from the cumulative index factor. The distance to the control points was considered to be representative of the degree of uncertainty. The cumulative index of cells located less than 250 m away from a control point were left unchanged, whereas cells located 250 m to 500 m, 500 m to 750 m, and more than 750 m from a control point were given values of 0.15, 0.2, and 0.3, respectively, subtracted from their combination index. The end result of the combination of all the cumulative indexes (Fig. 12a) helps to discriminate zones of high potential for orogenic gold mineralization. The 3D potential map focuses exploration by targeting high potential zones with consideration to uncertainty and, therefore, associated risk. The majority of the high-potential zones are located along the CLLF because this fault zone has a favorable orientation, and it also spatially controls one of the favorable lithologies, the Timiskaming Group. The inal map (Fig. 12a) shows that less than 6.25% of the total volume was assigned a cumulative index of 0.75, and less than 1% was assigned a value of 1. It is important to also evaluate the depth of these exploration targets to determine their potential economical value. Figure 12b presents the highest potential localities with a colour code for depth. More than 75% of the targets are less than 650 m deep. Cross-Validation of the 3D Potential Map To test the capacity of the presented methodology to predict mineralized volumes using the 14 major deposits of the area, the cumulative index was calculated for every cell within the model. The cells were then sorted according Table 3. Combination Index for Each Geological Feature Used for Mineral Potential Mapping Combination Index Structural Features < 500 m to fault intersection < 355 m to E–W Faults Lithological Features Temiskaming sediments Andesitic tuff Uncertainty–Distance to Control Points 250 m < Control point > 500 m 500 m < Control point > 750 m Control point > 750 m 0.25 0.5 0.25 0.25 −0.15 −0.2 −0.3 130 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Fig. 12. a. Mineral potential map for orogenic gold mineralization under the Proterozoic sedimentary cover. Index values vary from −0.3 to 1. b. Location (in 3D) of zones with highest index value painted accordingly to their depth with respect to the unconformity contact and the CLLF. to their index value and the volume occupied by the cells of each index value was calculated. Figure 13 presents the proportion of the 14 major deposits in the study area against the proportion of the total volume of Archean basement sorted in decreasing order of the combination index. The location of more than 90% of the mines can be predicted by the selection of 19.2% of the total surface area that has been covered. Conclusions and Limitations of the Model The integration of detailed geological observations with structural measurements, drill hole data, and geophysical interpretation has led to the construction of a realistic geological model. This model, coupled with compiled gold assays, was used to perform a spatial analysis to identify the controlling features for gold mineralization, and to create a 3D mineral potential map for a covered segment of a major Archean fault zone. The methodology suggested here uses a training set to evaluate features that inluence the distribution of gold mineralization, but still relies on geological knowledge to assign an importance to each of these factors. This methodology, based on Bayesian statistics and employing 3D modeling, is an eficient way to reduce risk in regional exploration campaigns at depth, and is particularly useful in sectors in which traditional exploration methods are inappropriate but where data are abundant. The methodology presented here has limitations in certain aspects. The main limitation is caused by uncertainties linked to interpretations at depth. Even though the methodology takes into account distances to control points under the Proterozoic cover, many interpretations on cross sections might inluence the spatial relationships between geological features and mineralized occurrences, and were made with unequally distributed data. Another limitation is linked to the assay database. Because copper and silver were not analyzed systematically, assays from different deposit types might have been incorrectly treated as orogenic deposits. Finally, the end result might vary by changing the importance given to each feature with the combination index, because the geologist’s models or assumptions directly inluence this result. Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault • O. Rabeau et aL. Fig. 13. Graph showing the proportion (%) of the 14 major gold deposits in the study area against the total volume of the total study area sorted decreasingly according to the combination factor value. Acknowledgments This work was supported by the University of Quebec in Abitibi-Témiscamingue, the DIVEX consortium, and the Ministère des Ressources naturelles et de la Faune du Québec. The manuscript beneited from careful reviews by E. De Kemp, J. Harris, S. Lin, and J. Richards. The authors would like to thank J. Goutier for his comments on part of the work presented. The assistance of G. Morin in the preparation and 3D integration of numerical data and of J. Jobidon for graphical design are also acknowledged. The following mining companies are also acknowledged for sharing data, information, or expertise: Xstrata Copper, Richmont Mines, Alexis Mining Corporation, Iamgold, Jean Decarreaux, Yorbeau Ressources, and West Cadillac Mining Corporation. Finally, J. Gagné, M. Canada, and P. Mercier are thanked for their meticulous help compiling the data necessary for this study. References Barrett, T.J., Cattalani, S., Chartrand, F., and Jones, P., 1992, Massive sulide deposits of the Noranda area, Quebec. II. The Aldermac mine: Canadian Journal of Earth Sciences, v. 28, p. 1301–1327. Bonham-Carter, G.F., Agterberg, F.P., and Wright, D.F, 1988, Integration of geological datasets for gold exploration in Nova Scotia: Photogrammetric Engineering and Remote Sensing, v. 54, p. 1585–1592. Corfu, F., 1993, The evolution of the Southern Abitibi greenstone belt in light of precise U-Pb geochronology: Economic Geology, v. 88, p. 1323–1340. Couture, J.-F., 1991, Carte géologique des gîtes métallifères des districts de Rouyn Noranda et de Val-d’Or: Ministère de l’Énergie et des Ressources du Québec, DV 90–11. Couture, J.-F, 1996, Gisements metalliques du district de Rouyn-Noranda, in Couture, J.-F. and Goutier, J., eds., Métalogénie et évolution de la région de Rouyn-Noranda: Ministère des Ressources naturelle du Québec, MB 96-06, p. 11–18. 131 Couture, J.-F., and Pilote, P., 1993, The geology and alteration patterns of disseminated shear zone-hosted mesothermal gold deposits: The Francoeur 3 deposit, Rouyn-Noranda, Quebec: Economic Geology, v. 88, p. 1664–1684. Couture, J.-F., and Willoughby, N.O., 1996, Géologie du gisement Granada in Couture, J.-F., and Goutier, J., eds., Métallogénie et évolution de la région de RouynNoranda: Ministère des Ressources naturelles du Québec, MB 96-06, p. 77–80. Couture, J.-F., Goutier, J., and Péloquin, A.S, 1996, Géologie de la région de Rouyn-Noranda, Québec, in Couture, J.-F., and Goutier, J., eds., Métallogénie et évolution de la région de Rouyn-Norand: Ministère des Ressources naturelles du Québec, MB 96-06, p. 1–9. Daigneault, R., and Mueller, W.U., 2004, Abitibi greenstone belt plate tectonics: The diachronous history of arc development, accretion and collision, in Eriksson, P., Altermann, W., Nelson, D., Mueller, W.U., Catuneaunu, O., and Strand, K., eds., The Precambrian Earth: Tempos and events: Development in Precambrian Geology, v. 12., p. 88–103. Daigneault, R., Mueller, W.U., and Chown, E.H., 2002, Oblique Archean subduction: Accretion and exhumation of an oceanic arc during dextral transpression, Southern Volcanic zone, Abitibi Subprovince, Canada: Precambrian Research, v. 115, p. 261–290. Dimroth, E., Imreh, L., Rocheleau, M., and Goulet, N., 1982, Evolution of the south-central part of the Archean Abitibi Belt, Quebec. Part I: Stratigraphy and paleogeographic model: Canadian Journal of Earth Sciences, v. 19, p.1729–1758. Dion, D.J., 1993, Données gravimétrique dans la région de l’Abitibi (Chibougamau, Rouyn-Noranda, Val-d’Or) et de Manicouagan: Ministère des Ressources naturelles et de la Faune du Québec, MB 93-61X. Dion, D.J., and Lefebvre, D., 1997, Données numériques (proiles) des levés géophysiques aéroportés du Québec (32D): Ministère des Ressources naturelles et de la Faune du Québec, DP-96-02. Dubé, B., and Gosselin, P., 2007, Greenstone-hosted quartzcarbonate vein deposits, in Goodfellow, W.D., ed., Mineral deposits of Canada: A synthesis of major deposit-types, district metallogeny, the evolution of geological provinces, and exploration methods: Geological Association of Canada, Mineral Deposit Division, Special Publication 5, p. 46–73. Eisenlohr, B.N., Groves, D.I., and Partington, G.A., 1989, Crustal-scale shear zones and their signiicance to Archaean gold mineralization in Western Australia: Mineralium Deposita, v. 24, p. 1–8. Fairbairn, H.W., Hurley, P.M., Card, K.D., and Knight, C.J., 1969, Correlation of radiometric ages of Nipissing diabase and Huronian metasediments with Proterozoic orogenic events in Ontario: Canadian Journal of Earth Sciences, v. 6, p. 489–497. Goldfarb, R.J., Groves, D.I., and Gardoll, S., 2001, Orogenic gold and geological time: A global synthesis: Ore Geology Reviews, v. 18, p. 1–75. 132 Exploration and Mining Geology, Vol. 19, Nos. 3–4, p. 115–132, 2010 Groves, D.I., 1993, The crustal continuum model for lateArchean lode-gold deposits of the Yilgarn Block, Western Australia: Mineralium Deposita, v. 28, p. 366–374. Groves, D.I., Goldfarb, R.J., Gebre-Mariam, M., Hageman, S.G., and Robert, F., 1998, Orogenic gold deposits: A proposed classiication in the context of the crustal distribution and relationship to other deposit types: Ore Geology Reviews, v. 13, p .7–27. Groves, D.I., Goldfarb, R.J., Knox-Robinson, C.M., Ojala, J., Gardoll, S., Yun, G.Y., and Holyland, P., 2000, Late-kinematic timing of orogenic gold deposits and signiicance for computer based exploration techniques with emphasis on the Yilgarn Block, Western Australia: Ore Geology Reviews, v. 17, p. 1–38. Groves, D.I., Goldfarb, R.J., Robert, F., and Hart, C.R.J., 2003, Gold deposits in metamorphic belts: Overview of current understanding, outstanding problems, future research, and exploration signiicance: Economic Geology, v. 98, p. 1–29. Guindon, D.L., Grabowski, G.P.B., Meyer, G., and Picotte, M.C.M., 2007, Report of activities 2006, resident geologist program, Kirkland Lake regional resident geologist report: Kirkland Lake district: Ontario Geological Survey, Open Report 6204, 42 p. Hodgson, C.J., 1993, Mesothermal lode-gold deposits, in Kirkham, R.V., Sinclair, W.D., Thorpe, R.I., and Duke, J.M., eds., Mineral deposit modeling: Geological Association of Canada, Special Paper 40, p. 635–678. Hodgson, C.J., and Troop, D.G., 1988, A new computer aided methodology for area selection in gold exploration: A case study from the Abitibi greenstone belt: Economic Geology, v .83, p. 952–977. Kendall, M., Stuart, A., and Ord, J.K., 1997, Kendall’s advanced theory of statistics: Volume 2A, Classical inference and the linear Model, 6th ed.: New York, John Wiley & Sons Inc., 300 p. Kerrich, R., Goldfarb, R.J., Groves, D.I., and Garwin, S., 2000, The geodynamics of world-class gold deposits: Characteristics, space-time distribution, and origins, in Hageman, S.G., and Brown, P.E., eds., Gold in 2000: Reviews in Economic Geology, v. 13, p. 500–551. Kishida, A., and Kerrich, R., 1987, Hydrothermal alteration zoning and gold concentration at the Kerr-Addison Archean lode gold deposit, Kirkland Lake, Ontario: Economic Geology, v. 82, p. 649–690. Lafrance, B., Davis, D.W., Goutier, J., Moorhead, J., Pilote, P., Mercier-Langevin, P., Dubé, B., Galley, A.G., and Mueller, W.U., 2005, Nouvelles datations isotopiques dans la portion québécoise du Groupe de Blake River et des unités adjacentes: Ministères des Ressources naturelles et de la Faune, Québec, 15 p. Leduc, M., 1986, Géologie de la région du lac Dasserat, Abitibi (Groupe de Blake River): Ministère de l’Énergie et des Ressources, Québec; MB 86-14, 180 p. Legault, M., and Lalonde, A., 2009, Discrimination des syenites associés aux gisements aurifères de la Sousprovince de l’Abitibi, Québec, Canada: Ministère des Ressources naturelles, RP-2009-04, 10 p. Legault, M., and Rabeau, O., 2006, Étude métallogénique et modélisation 3D dans la region de la Faille Cadillac dans le secteur de Rouyn-Noranda (phase 1): Ministère des Ressources naturelle et de la Faune du Québec, 11 p. Legault, M., and Rabeau, O., 2007, Étude métallogénique et modélisation 3D dans la region de la Faille Cadillac dans le secteur de Rouyn-Noranda (phase 2): Ministère des Ressources Naturelles et de la Faune du Québec, 11 p. Mallet, J.L., 1992a, Discrete modeling for natural objects: Mathematical Geology, v. 29, p. 199–219. Mallet, J.L., 1992b, GoCAD: A computer aided design program for geological applications: Journal of Mathematical and Physical Sciences, v. 354, p.123–141. McNicoll, V., Dubé, B., Goutier, J., Mercier-Langevin, P., Dion, C., Monecke, T., Ross, P.S., Thurston, P., Percival, J., Legault, M., Pilote, P., Débard, J., Leclerc, F., Gibson, H., and Ayer, J., 2008, New U-Pb geochronology from the TGI-3 Abitibi/plan cuivre: Implications for geological interpretations and base metal exploration [abs.]: GACMAC, Quebec 2008, Abstracts, 110 p. Mercier-Langevin, P., Dubé, B., Hannington, M.D., Davis, D.W., Lafrance, B., and Gosselin, G., 2007, The LaRonde Penna Au-rich volcanogenic massive sulide deposit, Abitibi greenstone belt, Quebec, Part I: Geology and geochronology: Economic Geology, v. 102, p. 585– 609. Morin, D., Jébrak, M., Bardoux, M., and Goulet, N., 1993, Pontiac metavolcanic rocks within the Cadillac tectonic zone, McWatters, Abitibi belt, Quebec: Canadian Journal of Earth Sciences, v. 30, p. 1521–1531. Poulsen, K.H., Robert, F., and Dubé, B., 2000, Geological classiication of Canadian gold deposits: Geological Survey of Canada, 106 p. Robert, F., 2001, Syenite-associated disseminated gold deposits in the Abitibi greenstone belt, Canada: Mineralium Deposita, v. 36, p. 503–516. Smith, J.P., Spooner, E.T.C., Broughton, D.W., and Ploeger, F.R., 1990, The Kerr Addison-Chesterville Archean gold quartz vein system, Virginiatown: Time sequence and associated maic “albite” dike swarm: Ontario Geological Survey, Miscellaneous Paper 150, p. 175–199. Thurston, P.C., Ayer, J.A., Goutier, J., and Hamilton, M.A., 2008a, Depositional gaps in Abitibi greenstone belt stratigraphy: A key to exploration for syngenetic mineralization: Economic Geology, v. 103, p. 1097– 1134. Thurston, P.C., Goutier, J., McNichol, V., and Legault, M., 2008b, Maic dikes cutting the Blake River Group, Abitibi Subprovince, [abs.]: GAC MAC 2008, abstracts, p. Tripp G.I., and Vearncombe, J.R., 2004, Fault/fracture density and mineralization: A contouring method for targeting in gold exploration: Journal of Structural Geology, v. 26, p. 1087–1108.