Deposit-scale structural architecture of the Sigma-Lamaque gold deposit, Canada—insights from a newly proposed 3D method for assessing structural controls from drill hole data

Deposit-scale structural analysis is a practical method of structurally analyzing 3D drilling data. This interpretation method uses readily available distributed 3D data at the most relevant scale for mining and exploration. This method does not depend on collecting traditional structural data from core, such as orientation of planar and linear data, but relies on the premise that grade distributions sampled from drilling provide insight into the structural control of the deposit, with the interpretation of structural geometry done using a computer 3D viewer. This method is applied to the well-studied Archean Sigma-Lamaque gold deposit, which features late-orogenic vein mineralization, and which is controlled by three orders of structural features. The first-order control is the easterly plunging pipe-like conduit parallel to F2 fold plunge defined by the intersection of bedding (S0) and foliation (S2) (~ 1.5 km radius × 3 km down-plunge). The second-order controls, nested within the first-order, are the younger subvertical Late Diorite plugs that intruded vertically into the host rocks along the established S2 foliation anisotropy (1250:400:150 m dimension for Main Plug). The third-order control, laterally limited by first- and second-order structural controls, is the well-documented gold-bearing vein sets (1–10 m thick and up to 1.5 km in extent) that developed in the Late Diorites and in the older folded volcanogenic stratigraphy and diorites. It is proposed that the first-order control defined by the moderately plunging F2 fold is the main fluid pathway resulting from structural permeability formed during earlier folding. The second-order Late Diorite plugs and other competent lithological units, which intersected and fractured during late-stage D2 compression, served as chemically reactive sites causing gold to precipitate along this first-order pathway. It is predicted that intrusions that do not intersect the first-order structural conduits are less likely to be mineralized. Greenfield orogenic-gold exploration requires that the first-order controls are identified, and the geometry of hydrothermal alteration is best understood in the context of the structural architecture of the host rocks.


Introduction
Drill data are the best three-dimensional (3D) dataset the mineral industry has access to.
Despite the historical limitations of unsophisticated datamost databases contain only a single column of data for a single element (e.g., Au), and many are also under-sampled with many sample gaps-the drilling database is still best for studying mineral deposits. But, scant attention is given to the deposit-scale 3D patterns that can be gleaned from these deposit-wide data, yet the structural patterns gleaned from these data at the deposit-scale can guide further exploration in the surrounding region. Using historical drilling and mined stope data, the Sigma-Lamaque gold deposit is used as a case study for a new approach to examine mineralization patterns. This approach allows sophisticated deposit-scale analysis of the structural patterns within the data to be conducted from single element and lithology table data; that is, data lacking traditional structural measurements (such as orientations of planar and linear data) but commonly available for most mineral deposits. By applying this evaluation method to other ore bodies, where only unsophisticated historical data are available, exploration opportunities can be assessed in 3D, thereby opening up new opportunities for exploration in close proximity to existing deposits.
This study also offers a new structural perspective on the Sigma gold deposit (one half of the Sigma-Lamaque deposit and already extensively studied), which is interpreted as a shear zone-hosted orogenic gold deposit (Robert and Brown 1986a, b;Sibson et al. 1988). This shear zone-hosted interpretation has been used as a template to interpret numerous gold deposits around the world.
The practical importance of deposit-scale structural analysis Like many industries, the mineral industry is currently in the midst of a data-driven revolution. Substantial increases in drilling activities have resulted in the rapid sampling and delineation of mineral deposits in 3D. The ability to obtain previously unobtainable data from drill cores (e.g., high-precision multi element data, hyperspectral data, X-ray tomography and fluorescence, topography scans) has increased exponentially over the last two decades, and this trend is likely to continue (Haest et al. 2012a,b;Schodlok et al. 2016;Arne et al. 2016;Sjöqvist et al. 2017;Bergqvist et al. 2019). Implied in the collection of such multiple datasets, and the ability to computationally process this information, is an assumption that there will be a direct translation into better understanding of ore deposits, increased identification of near-mine exploration targets, and increased confidence in Resource estimations.
Despite this increase in capacity and technical capability, massive write-downs of Resources and Reserves in structurally controlled gold deposits continue. Despite the large amounts of drill sampling examples of these problems include the Phoenix gold mine in Canada in 2016 (Financial Post 2016) and more recently in Guyana (Financial Post 2019). How can these Resource write-downs occur when mining operations claim to efficiently obtain abundant drilling data to minimize the risk associated with Resource estimation? Could a fundamental geological knowledge gap be the cause of these write-downs?
One critical knowledge gap may be the cause: the study of the structural control of mineral deposits at the deposit scale.
"Deposit-scale" refers to observations that can be made from the entire drill sampling, with or without modeled Resource wireframes. Traditionally, and reflecting limited access to the deposit and its the drill hole database, structural controls have been determined on subsets of a deposit then scaled-up. This practice applies to most academic studies and many studies commissioned by companies.
Deposit-scale structural analysis is a simple, rapid, and practical approach to mineral deposit analysis that focuses on the structural analysis of drilling data at the scale of the mineral deposit. These data are usually single element ± logged lithology data. Importantly, they are not reliant on structural data obtained from the core or outcrop. As introduced in this study, this approach contrasts with typical published mineral-deposit analysis, which rarely examine the geometric patterns at the mineral deposit scale.
The traditional exploration workflow of mineral deposit analysis at a regional scale focuses mainly on regional 2D datasets (e.g., aeromagnetic, gravity, soil geochemistry). These are interpreted to establish the tectonic setting and are followed by field work to obtain outcrop, core, and microscale observations (note: "Outcrop" is a general term describing macroscopic scale, which includes stope-scale that applies to underground mines). Geochemical analyses of hand samples provide information about primary and alteration mineral assemblages. All these 2D datasets are commonly analyzed in the context of a published theoretical ore deposit or structural model (Vearncombe and Zelic 2015), and an interpretation is made to explain and predict the origin of the mineral deposit(s).
Discussion of the observed geometric nature of the deposit at the deposit scale in 3D, and its likely structural controls, is conspicuously absent in most mineral deposit research articles and deposit descriptions (e.g., Phillips 2017). This is despite nearly all mining and exploration companies having 3D wireframes of their ore deposits at their disposal. These took time to meticulously digitize in serial sections or which they constructed rapidly using 3D interpolation (Cowan et al. 2002(Cowan et al. , 2003. In the mining industry, these Resource wireframes are treated as merely shapes: practical containers in which various commodities are estimated using geostatistical procedures. However, very few studies ascribe geological meaning to the 3D distribution of mineralization. Even if geological meaning is assigned to grade distribution, the traditional workflow uses indirect observations and interpretations (largely from microto meso-scale, but not directly from observed patterns at the deposit-scale) to infer the deposit-scale patterns and controls (Fig. 1). The 3D wireframes of a Resource are rarely explained in terms of geological patterns or controls that are recorded by geologists at other scales (Fig. 1).
Historically, analytical tools were developed to analyze structural patterns at all scales except the deposit scale, which is "siloed" from being considered as a structural geological problem. Instead, deposit-scale analysis is considered a Resource definition problem and is largely ignored in applied structural analysis (Fig. 1). Traditionally, structural controls at the deposit scale are predicted from outcrop and hand sample scales, with interpretation occurring along the "Inside-Out" direction shown by the red arrow in Fig. 1. This practice of inferring large-scale geological patterns from the small scale, assuming scale-invariant relationships across a range of scales, is known as the Pumpelly's Principle (Pumpelly et al. 1894;Van Hise 1896). In contrast, this study advocates an "Outside-In" analytical approach for interpreting structural features from a larger scale, which is more reliable because predictions can be made, and field tested. Deposit-scale patterns are predicted inward from regional 3D data, and outcropscale patterns are predicted inward from deposit scale (1 to 2 then 2 to 3 along the blue arrow in Fig. 1).
In this study, the main benefits of deposit-scale structural analysis are highlighted by discussing a classic orogenic gold deposit-the Sigma-Lamaque gold deposit in Quebec, Canada, that has been the subject of many decades of highly influential studies, beginning with François Robert's seminal work (1983). This example of deposit-scale structural analysis shows how simple and powerful this "Outside-In" method is to decipher the structural controls of the mineral deposit and answer many issues that were not addressed by previous research on this deposit, all of which focused on up-scaling outcrop observations. The Sigma-Lamaque gold deposit

Background
The Lamaque and Sigma deposits (discovered by surface prospecting in 1924 and 1933, respectively) were the first discoveries in the Val-d'Or region of Quebec, Canada (Fig. 2) and became the region's largest gold producers (Poirier et al. 2015a). When operational (1935-1985, the Lamaque mine produced approximately 24 million tonnes of ore, grading 5.9 g/t gold for a total of 4.6 million ounces of gold, and for a time it was the largest gold producer in Quebec. The Sigma mine (operational 1937-1997) produced approximately 24 million tonnes of ore, grading 5.8 g/t gold for a total of 4.5 million ounces of gold. Minor production continued at both mines intermittently from 1997 to 2012. From 1935 to 2012, the combined production from the Sigma and Lamaque mines was approximately 55.9 million tonnes of ore, grading Fig. 1 Scales of observation and associated tools of investigation of mineral deposits. No specific tools have been developed to study structural patterns at the deposit scale 5.3 g/t gold for 9.5 million ounces of gold, which is more than half the historic production of the Val-d'Or mining camp and makes the two deposits world-class (Poirier et al. 2015a).
In 2003, Integra Gold Corp acquired the Lamaque South property, south of and adjacent to the Sigma-Lamaque mines (Fig. 2). Mineral Resource estimates and preliminary economic assessment studies followed (Poirier et al. 2015a, b), and, late in 2014, Integra Gold acquired the Sigma-Lamaque mines. Eldorado Gold Corporation subsequently acquired the property from Integra Gold in 2017, and commenced commercial production in March, 2019.
The study presented here was conducted in late 2015 as one of the submissions to the public Integra Gold Rush Challenge competition (Integra Gold Corp 2015), which was modeled on the Goldcorp Challenge crowdsource competition held in 2000 (Wilde and Kreuzer 2016). For the competition, Integra Gold released its entire geological and drilling databases, including wireframes of modeled ore bodies and stope shapes; the drilling database was cleaned by Integra Gold to remove errors and inconsistencies. This data contained more than 35,000 surveyed holes, comprising 700,000 assay and 170,000 lithology logs. Of these thousands of holes, only 77 were assessed in this study to have data integrity issues and they were omitted from the analysis.

Geological setting
The Sigma-Lamaque gold deposit (previously referred to as two separate deposits) is located in the eastern segment of the southern volcanic zone of the Archean Abitibi Subprovince, immediately north of the boundary with the Pontiac Subprovince (Fig. 2). The gold deposit is part of the Cadillac tectonic zone, a 250 km long east-west trending high-strain zone in eastern Ontario and western Quebec (Robert 1989). The Cadillac tectonic zone, which is the boundary between the two subprovinces, is interpreted as a Fig. 2 A synthesis of the regional geologic setting of the Sigma-Lamaque gold deposit in the Val-d'Or camp, southern Abitibi Subprovince, Quebec. Geology compiled from Imreh (1982), Robert (1989), Desrochers and Hubert (1996), Tremblay (2001), and Scott et al. (2002). See Fig. 3 for geochronological references and ± age ranges. Red stars show the Sigma (north) and Lamaque (south) mines. The city of Val-d'Or is located at 48.098709°, − 77.796768°and UTM Zone 18U steep northerly dipping high-strain suture between the two subprovinces and is one of the greatest gold belts on Earth with a current Resource of about 130 million ounces of gold (Rafini 2014). The geology of the Val-d'Or district has been described in detail by previous workers and their articles contain a more complete discussion of the regional geology (Robert 1989;Desrochers and Hubert 1996;Scott et al. 2002;Olivo et al. 2006;Bedeaux et al. 2017). This section only summarizes the geology relevant to the discussion of the deposit-scale controls of Sigma-Lamaque gold mineralization.
The Val-d'Or district comprises a sequence of volcanosedimentary and intrusive rocks that were emplaced from 2714 to 2611 Ma (Fig. 2). The Jacola Formation contains pillowed and massive tholeiitic basalts, intercalated with komatiites, and it interfingers with the Val-d'Or Formation to the south, which comprises discontinuous and interstratified massive to pillowed lavas and volcaniclastic andesite, dacite, and rhyolite. The Jacola and Val-d'Or Formations are intruded by felsic to mafic bodies that are grouped into three age groups: 2700 Ma synvolcanic Bourlamaque Batholith; the 2680 Ma syn-to late-D 2 diorite/granodiorite intrusions (Jemielita et al. 1989), and 2675 to 2611 Ma late-to post-tectonic intrusions (East Sullivan Stock, La Corne Batholith, and Bevcon Pluton). Most of the high-grade gold mineralization at Lamaque is hosted in the fractured 2680 Ma "Late Diorites," also referred to as the "Main Plug" because of its prolate intrusion geometry (i.e., pencil-shaped but with an elliptical cross-section [Burrows and Spooner 1989]). The bulk of the remaining lower-grade gold mineralization is hosted within the Jacola and Val-d'Or Formations in various intraformational lithologies (Fig. 2).

Structural architecture at multiple scales
Primary volcanic features in both the Jacola and Val-d'Or Formations indicate a younging direction to the south, with the bedding steeply dipping variously to the southsouth-east and east within the Sigma-Lamaque tenement south of the Bourlamaque batholith (Robert and Brown 1986a). The bedding strike trajectory trace is S-shaped and mimics the S-shaped trace of the contact boundary of the Jacola and Val-d'Or Formations to the west under the city of Val-d'Or (fig. 4 in Desrochers and Hubert 1996) (Fig. 2). All rock units older than the posttectonic intrusions (Fig. 2) record ductile and brittle strain that is the result of D 2 deformation. D 1 deformation is localized closer to the Cadillac tectonic zone (Robert 1989) and west of the city of Val-d'Or and does not affect the rocks in the study area (Desrochers and Hubert 1996). The Jacola and Val-d'Or Formations are pervasively foliated at various intensities by S 2 foliation formed during D 2 , and measurements of S 2 within the tenement and surrounding area consistently strike east-west and, on average, dip steeply to the north, very similar to the steep inferred northerly dip of the Cadillac tectonic zone (Fig.  2). The level of ductile strain, reflected in the S 2 intensity, increases southward towards the Cadillac tectonic zone (Robert 1989). The main schistosity in the study area is labeled as S 2 and follows the convention adopted in the literature published prior to Bedeaux et al. (2017).
To determine the chronological relationship of the lithologies at Sigma-Lamaque, the major lithological units available as logged units were grouped and ordered by relative age. These units were then matched with the geological units in the chronological chart compiled from Morasse et al. (1995) and Scott et al. (2002) and references cited in these papers (Fig. 3).
The drilling data are divided into the Sigma and Lamaque deposits and the volume for geological modeling was defined (Fig. 4). The boundary between the two volumes is not defined by the two separate mining operations: instead, the planar northerly-dipping boundary is within the low-grade zone and trends parallel to the pervasive S 2 foliation recorded in the host rocks of the Jacola and Val-d'Or Formations. An implicit modeling software (Leapfrog Mining version 2.4.5.17) was used to model the geological units, but this was conducted after the structural geological architecture was determined from raw data by applying the "Outside-In" interpretation process summarized in Fig. 1.

Regional-scale observations
The regional-scale geological map of Val-d'Or and the surrounding district (Fig. 2) was compiled from previously published research (Imreh 1982;Robert 1989;Desrochers and Hubert 1996;Tremblay 2001;Scott et al. 2002), including the most recent geological survey mapping (Pilote 2015a, b).
The regional geological relationships relevant to the Sigma-Lamaque deposit are: & The Sigma-Lamaque deposit is located on the southern limb of the ESE-plunging overturned La Motte Anticline. & S 2 readings from the Sigma-Lamaque site are very sparse.
However, the literature indicates that S 2 has an approximate dip-azimuth range of 000°to 015°and a dip range of 70°to 90° (Morasse et al. 1995;Scott et al. 2002;Bedeaux et al. 2017), making it axial planar to the La Motte Anticline. & The bedding readings in the Val-d'Or Formation vary, dipping steeply south as well as being overturned and dipping north; this is consistent with an overturned southern limb of the La Motte Anticline (Fig. 2). & All younging directions within the Val-d'Or Formation and the underlying Jacola Formation are consistently south (Fig. 2), and younging reversals are only found in the Heva Formation. This is consistent with the Val-d'Or Formation being located on a limb of a large fold. & The lower contact of the Val-d'Or Formation with the Jacola Formation is documented as being gradational (Scott et al. 2002). However, the change in strike of this contact west of Sigma-Lamaque suggests local parasitic folding of this interface (Fig. 2). & Very few bedding readings are available from the Sigma-Lamaque site. Desrochers and Hubert (1996) suggest the bedding strike changes to mimic the lower contact of the Val-d'Or Formation (Fig. 2); again, this is consistent with  Desrochers and Hubert 1996). & The tightness of the folding is uncertain. The consistently documented (Scott et al. 2002) southerly-facing stratigraphy from the Val-d'Or Formation suggests open folding but Robert and Brown (1986a) and Robert and Poulsen (1997) have indicated isoclinal folding.

Deposit-scale observations
Despite the extensive geological work conducted at the Sigma-Lamaque deposit (Perrault et al. 1984;Robert and Brown 1986a, b;Boullier and Robert 1992;Gaboury et al. 2001;Garofalo et al. 2002;Garofalo 2004;Olivo et al. 2006), many geological peculiarities and features that are recorded at the deposit-scale are not addressed by previous work. The Sigma and Lamaque gold occurrences are only 500 m apart, separated by a northerly-dipping low-grade zone (Fig. 4). Although they are considered a single deposit because of their shared geological history, one of their most striking features is how dissimilar they are in terms of host rock characteristics and deposit geometry. Key observations, which provide important clues to the geologic control of the gold deposit recorded at the depositscale, are listed below.
Observation 1: gold occurrence and lithology are dissimilar at Sigma and Lamaque There are significant differences in the two deposits between gold occurrence and host lithological units (Fig. 5). The highest mean gold grades are associated with the "Late Diorite" phases that occur as plugs in the Lamaque area: "Late Diorite" refers to the phases of diorite logged as Diorite, Early Quartz Diorite, Porphyritic Granodiorite, and Granodiorite (Figs. 5 and 6), which are also referred to as the "Main Plug" (Burrows and Spooner 1989).
Gold data separated into Sigma and Lamaque volumes ( Fig. 5) show that the Late Diorite plug intrusions are highly mineralized at Lamaque, but are largely absent at Sigma. Conversely, the relatively poorly-mineralized lithological units at Lamaque are the highly mineralized lithological units at Sigma (e.g., Early Diorite, Felsic to Intermediate Dykes) (Fig. 5).
The quartz-tourmaline veins, which carry the high grades right across the Sigma-Lamaque deposit, are late-stage brittle fractures that cross-cut all units (Fig. 3). Early Diorite and Late Diorite plugs are associated with the highest grades and are present at both locations. These relationships suggest that the gold mineralization is not entirely influenced by lithology.

Observation 2: Late Diorite plugs only exist at Lamaque
Although some minor dykes are logged as the Late Diorite phase, characteristic plug-like (prolate, or pencil-shaped) geometries are absent at Sigma. The very linear and steeply plunging Late Diorite plugs are only present in the Lamaque area-this puzzling detail requires geological explanation.
Observation 3: the Late Diorite plugs exhibit preferred orientation and geometry The geometry and significance of the Lamaque Late Diorite plugs have not been discussed by previous workers despite their importance as hosts of the Lamaque gold mineralization. The green volume extends to the tenement boundary, which is indicated in Fig. 2 The Late Diorite plugs exhibit a preferred orientation and are very linear, with the Main Plug at Lamaque plunging towards 76°→ 026°. The main mineralized plug is elongate but is triaxial in shape and not perfectly linear with axial lengths a:b:c of 1250:400:150 m (Fig. 6), and the high-grade gold distribution mimics this overall geometry. The a-b plane defines a planar orientation dipping steeply to the north ( Fig. 6d) with an approximately ellipsoidal axial ratio of 8:3:1, which is a prolate geometry according to k-value of 1.3 computed from these axial ratios (Flinn 1962). Although Flinn (1962) developed the k-value measure as a technique to document marker geometries in ductile-strained rocks, this technique is also a convenient way to classify overall shapes of geological units and 3D grade distributions (where k > 1 is prolate [pencil shape]; k < 1 is oblate [pancake shape]; and k = 1 is equally prolate and oblate). The Lamaque intrusions have a peculiar geometry and it would be of interest if this geometry is related to the structural architecture of the Jacola and Val-d'Or Formations that can be deciphered at the deposit-scale.

Observation 4: vein orientations at Lamaque and Sigma display contrasting vein distributions
Sigma and Lamaque sites contain vein-hosted gold mineralization comprising multiple quartz-tourmaline vein sets with variable orientations (Wilson 1948;Robert and Brown 1986a). Using an implicit drawing tool (Cowan et al. 2004), vein orientations for both Sigma and Lamaque were summarized by digitizing the median surfaces of the mined stopes, which correspond with mineralized veins (Fig. 7a), then by plotting all poles of the triangulation faces of the drawn vein median surfaces (Fig. 7b). Light-blue shading in the stereonets of Fig. 7b shows a departure from perfect cylindricity, and veins at both locations are close enough to be considered a result of plane-strain deformation (i.e., 2D strain). Figure 7b plots the theoretical shear planes and assumes a conjugate vein symmetry relative to the extension veins. Gray shading shows the estimated theoretical compression fields assuming perfect symmetry of thrust shear veins relative to the extension vein orientation (poles of veins, which are extension structures, are not expected to plot in the compression field). The compression field at Lamaque has an angle of 70°and although asymmetric in development, is represented by both north-and south-dipping veins in the extensional field (these are interpreted as thrust shear veins). The compression field at Sigma has an angle of 75°, but the symmetrical low-angle south-dipping shear vein is missing in the extensional field (position of green pole in Fig. 7b for Sigma); instead, as documented by Robert and Brown (1986a), it is replaced by steep south-dipping misoriented reverse shear veins (i.e., "misoriented" meaning they are unfavorably oriented for slip and dilation relative to the compression direction determined by the extension veins Sibson [1992]). Bulk strain inferred from the veins developed at Sigma and Lamaque suggests vertical extension and subhorizontal N-S shortening (Fig. 7b). However, no explanation has been documented as to why the sets of vein orientations in the two areas differ markedly in symmetry and orientation. The complete contrast in vein orientations at Sigma and Lamaque, which resulted from the same N-S shortening event, requires a geological explanation because the two areas are only 500 m apart.
Observation 5: the Sigma-Lamaque vein swarms appear to converge at depth The enveloping surface of the Sigma deposit vein swarm, comprising both shear and extension veins, is subvertical and strikes 260°, whereas the enveloping shape of the Lamaque deposit strikes 274°and dips to the north at an angle of 77°-an angular difference of 19°between the two zones. Although the strike of the deposits is similar, because of this obliquity in dip values, both mineralization zones appear to converge at depth in a V-shaped geometry; however, the zone of apparent convergence is below the current drilling depth. No geological explanation has been put forward to explain this convergence.
Observation 6: the Sigma veins shift orientation systematically with position At Sigma, the orientations of the veins shift with geographic position. This is particularly evident in the flat extensional veins when viewed towards the north (Fig. 8). The uppermost eastern veins are flat, whereas the westernmost and deeper veins exhibit a low-angle westerly dip (Fig. 8). Although there is a similar degree of orientation shift in the south-dipping shear veins, there has been no geological explanation for these gradual spatial shifts in vein orientations.
Observation 7: an easterly plunging western grade boundary exists at both Sigma and Lamaque A feature of both Sigma and Lamaque is a 45°easterlyplunging boundary towards the western side of the goldbearing vein swarms (Fig. 9). No geological control has been proposed for this mineralization limit at the Sigma-Lamaque deposit.
In the next section, a structural geological model for the Sigma-Lamaque deposit is presented that explains all these deposit-scale issues.

Synthesis
The annotated local geological map of Sigma-Lamaque shows an obliquity between first-order bedding (S 0 ) and S 2 foliation traces (Fig. 10a). Other geological maps of this area show a similar bedding-to-foliation angular relationship, which  Fig. 8 The stopes of Sigma horizontal veins (green) exhibit a systematic change to higher dips in the westerly direction. The position of this view relative to the entire Sigma gold distribution is shown in Fig. 9 indicates proximity to a fold hinge.
The Undifferentiated Tuff that are logged from the Sigma-Lamaque deposit display identifiable bedding traces when viewed in a 3D visualization software. A view in the down-dip direction of S 2 , assumed to be dipping 77°→ 004°, estimated from Robert and Brown (1986a: fig. 5b), is shown in Fig. 10b. In this near plan view, there is a 24°angular obliquity between the bedding trace in the Undifferentiated Tuff (green) and the S 2 trace. When viewed in vertical section, there is a similar degree of obliquity between the S 0 and S 2 traces viewed in the S 2 plane (Fig. 10c). When viewed looking 44°→ 081°, Easterly-plunging western gold-grade limit at Sigma and Lamaque (dashed pink line). Grades are rendered with Maximum Intensity Projection, which projects the highest grade through the low-grade halo along the line parallel to view (Cowan 2014a). Position of Fig. 8 shown as dashed outline however, the intersection between S 0 and S 2 shows an obliquity of 50° (Fig. 10d). This is the largest possible obliquity angle between S 0 and S 2 and is evident when viewing in the S 2 plane and rotating the view about the pole of S 2 . This view (Fig. 10d) equates with the downplunge direction of the intersection between S 2 and S 0 (i.e., "down-structure" method of MacKin 1950). In this view, the trace of bedded units in the Undifferentiated Tuff (green) and Felsic Volcanic (orange) is evident in the Sigma side of the deposit (Fig. 10d).
All observations of the Sigma-Lamaque deposit listed above can be explained by a folded architecture of the Jacola and Val-d'Or Formations (Table 1). The a-b plane of the Late Diorite plugs (Fig. 6d) defines a planar orientation subparallel to S 2 (77°→ 004°) developed in the Jacola and Val-d'Or Formations, which has reported dips ranging from 70°to 90°, and dip-azimuth ranges from N-S to NNE-SSW (Claoué-Long et al. 1990;Morasse et al. 1995;Scott et al. 2002;Bedeaux et al. 2017).
The S 2 -parallel orientation of the Late Diorites is consistent with syn-or post-fold emplacement into the axial planar region of a local asymmetrical F 2 fold closure during D 2 . That is, the Late Diorites must have intruded along an existing planar heterogeneity in the country rock fabric, implying that the fabric must have existed before the diorite was emplaced. The fold axis is the region of highest structural dilation resulting from ductile and brittle strain. Metamorphic and/or igneous fluid flow and intrusions would focus along the plunge of the fold axis and by inference have the highest probability of localizing mineralization. This interpretation is consistent with the latest research on how granite melts are transported and emplaced in a compressive tectonic environment (Montanari et al. 2010;Ferré et al. 2012).
The angularity between the bedding (S 0 ) and tectonic foliation (S 2 ) defines an easterly plunge in the intersection line of these fabrics, and this expected fabric intersection plunge line that can be inferred from regional mapping (Fig. 2) coincides with the easterly plunge of the western grade boundary that is evident at Sigma-Lamaque (Fig. 9).
A 3D bedding-form surface model was constructed by interpolating bedding orientations (Vollgger et al. 2015)  If fold-controlled structural permeability is the main control of mineralization, the gold distribution will largely be bedding parallel at the limb position and axial planar at the fold closure, which would transect the bedding. This explains the grade-lithology relationship contrast between Sigma and Lamaque, as Sigma is located in the fold limb position and Lamaque is located in the F 2 axial planar position. 2 Late Diorite plugs only exist at Lamaque Lamaque Late Diorite plugs are located in a fold closure, which is commonly a zone of dilation compared to the fold limbs that tends to undergo layer-orthogonal compression. The low-grade zone between Sigma and Lamaque is parallel to S 2 . 3 The Late Diorite plugs exhibit preferred orientation and geometry The Lamaque Late Diorite plugs are located in the axial planar region of a F 2 parasitic fold and emplaced parallel to the S 2 foliation. The F 2 axial plane had the primary influence on its emplacement. This suggests that the Lamaque Late Diorite intrusions were emplaced during the end stages of D 2 deformation when F 2 had already formed. Temporal orientation of σ 2 to the vertical orientation emplaced the Late Diorites as subvertical linear plugs that are contained within the plane of S 2 foliation. 4 Vein orientations at Lamaque and Sigma display contrasting vein distributions Fracture development and orientation depends on the attitude of the host rock anisotropy (Gomez-Rivas and Griera 2009) relative to the compression direction. Steep bedding in Sigma influenced development of misoriented reverse shear veins during high fluid pressure during an inferred to be N-S compression (Sibson et al. 1988). This resulted in vein development at Sigma that is highly asymmetric compared to the conjugate veins developed at Lamaque where bedding is interpreted to be less steep.
5 The Sigma-Lamaque vein swarms appear to converge at depth Sigma veins are almost bedding-parallel, whereas Lamaque is axial-plane parallel; thus obliquity between these two areas is naturally expected in a fold closure setting. 6 The Sigma veins systematically shift orientation with position Local stresses will be guided by local anisotropy (Dubé et al. 1989; Gomez-Rivas and Griera 2009), thus a spatial change in bedding orientation due to folding will change the vein development orientation to reflect this change in the underlying heterogenetic fabric.

An easterly plunging western grade boundary exists at both Sigma and Lamaque
This grade plunge is parallel to the fold plunge towards the east. The lower boundary to the grade is inferred to be a bedding layer within the Val-d'Or Formation, with poor permeability, which channeled the mineralized fluids upwards from its deeper source.
interpreted from the magnetic data on the topography surface and the bedding traces visible in the drilling data when viewed in a 3D viewer (Fig. 11). The modeled bedding surfaces conform to the consistent southward younging of the Val-d'Or Formation, and are also consistent with the down-dip curvature of the bedding at depth visible in the logged lithologies at Sigma (Fig. 10d). Likely bedding traces can be identified in the magnetic data perturbations (reduced to pole, first vertical derivative, and horizontal gradient) and these were incorporated into the model (Fig. 12). No S 2 measurements were available, so a constant S 2 orientation of 77°→ 004°was assumed throughout the area of interest, with an F 2 fold plunge of 44°→ 081°. The sections of Figs. 11 and 13b are viewed parallel to this F 2 plunge line. The stereoplot in Fig. 13a summarizes the main structural orientations at Sigma-Lamaque. The Late Diorites are parallel to the S 2 fabric, but their linear axes have a steeper plunge than the plunge of F 2 (Fig. 13a, c). Their emplacement is likely to have been during protracted N-S shortening, but specifically when σ 2 switched with σ 3 into a near vertical position, with the Late Diorites intruded along σ 2 as prolate bodies parallel to the S 2 foliation plane. Interestingly, the plunge of the Late Diorites (Fig. 13a, c) is also subparallel to the steep NE-plunging stretching lineation L 2 developed in the Val-d'Or area, as reported by Bedeaux et al. (2017: figs. 13 and 16). This may indicate that the planar-linear geometry and orientation of the Late Diorites may have been influenced by pre-existing planar-linear schistosity S 2 developed during D 2 .
Fracture development occurred when σ 2 reverted back to a horizontal E-W orientation after the emplacement of the Late Diorites (Fig. 7). Many quartz-tourmaline veins propagated outside the diorites to mineralize the adjacent lithological units in contact with the diorites. Because the brittle fracturing occurred after the major ductile deformation at the end of D 2 (Fig. 2) the competency of the diorite largely governed most of the fracture locations.
Supplementary video file ESM1.mp4 displays the steps involved in identifying and working out the fold plunge from raw lithological data (Fig. 10), and how this plunge correlates to the western limit of the gold mineralization (Fig. 9). show they are well developed parallel to S 2 , but not along bedding at Lamaque. c Felsic to Intermediate Dykes (brown) distribution shows a clear boundary between Sigma and Lamaque. This low-grade boundary is parallel to S 2 , as shown in red trace. This boundary was used to define the Sigma and Lamaque domains (Fig. 4). d Broadly bedding-parallel mineralization pattern in Sigma, and F 2 axial-plane parallel mineralization pattern at Lamaque. Only gold grades of ≥ 30 g/t shown in Maximum Intensity Projection rendering. See Supplementary video file ESM3.mp4 for an explanation veins (Sigma and Lamaque stopes) to the F 2 fold plunge and the modeled folded stratigraphy.

Comparison with previous work
The detailed descriptions of the vein systems developed at the Sigma mine in the original structural study of the Sigma deposit (Robert and Brown 1986a, b) and subsequent studies (e.g., Sibson et al. 1988;Robert and Poulsen 2001) have been influential in interpreting orogenic vein-hosted gold deposits as well as fluid flow research worldwide (e.g., Kontak et al. 1990;Cox et al. 1995;McCuaig and Kerrich 1998;Bierlein and Maher 2001;Cox and Ruming 2004;Dubé and Gosselin, 2007). At the time of Robert's research in the 1980s, the Lamaque deposit and the Sigma mine were owned by different companies and not all data could be accessed. However, a detailed description of the structural setting and the characteristics of the gold-bearing veins of the Lamaque mine was previously described by Wilson (1948). This current study extends the analysis to the entire Sigma-Lamaque system and the observations and conclusions of Robert and Brown (1986a) are placed in the context of deposit-scale observations made from the entire gold deposit. Observation and analysis of a deposit's entire drilling data is now possible, some three decades since the seminal work of Robert (1983) and Sibson et al. (1988). In the intervening period, significant computer 3D rendering and processing methods have been developed, most importantly Maximum Intensity Projection (developed in 1989 [Wallis et al. 1989; Wikipedia contributors 2019]) and its application to analyze deposit-scale structural patterns from drill hole assay data (Cowan 2014a), and the introduction of implicit geological modeling software to the mining industry (Cowan et al. 2002(Cowan et al. , 2003Cowan 2014b).
Largely based on Sigma's steep reverse and flat extensional vein systems that were described by Robert and Brown (1986a), Sibson et al. (1988) Sibson (1989Sibson ( , 1992 and Cox (1995Cox ( , 2005Cox ( , 2010. The fault-valve model called for high fluid pressure to account for the necessary lowering of the shearing strain along unfavorably oriented steep reverse faults that are at an unusually high angle (> 55°) to the maximum principal stress inferred from the subhorizontal extension veins. The existence of unexpected reverse shear veins, as opposed to thrusts that have a lower angle to the maximum principal stress, led Sibson et al. (1988) to conclude that the reactivation of the faults could only occur when fluid pressure exceeded the lithostatic load. Boullier and Robert (1992) provided microstructural evidence of cyclic growth and deformation in the veins from gold deposits in the Val-d'Or district and argued in support of the fault-valve model. The existence of unfavorablyoriented reverse shear veins is consistent evidence for high fluid pressure (Cox 2010), but the model does not explain why shear veins that are unfavorable in orientation initially formed, in preference to more favorable orientations, given that high fluid pressure would influence the formation of veins of any orientation and kinematics (S.F. Cox, pers. comm., 2019). Under the presence of high fluid pressure, shear veins in favorable orientations would be expected to form, as recorded at Lamaque (Wilson 1948). However, at Sigma, only 500 m away, favorable shear vein orientations that dip south at a lower angle to the extension veins (< 55°) do not exist (Fig. 7b).
One possible explanation for the formation of the unfavorably oriented reverse shear veins at Sigma is that these veins are a small-scale expressions of a reactivated much larger "master shear zone" with prior history. Sibson et al. (1988) inferred that the small-displacement gold-bearing reverse shear veins at the Sigma mine, which are recorded down to 2 km, are part of a large brittleductile planar shear zone. This shear zone was interpreted to have a constant steep southerly dipping orientation that extended more than 10 km down into the base of the seismogenic zone where the Sigma deposit was interpreted to have been located. This interpretation was a b c d Fig. 13 D 2 structural elements of the Sigma-Lamaque deposit. a S 2 dips 77°→ 004°and the F 2 fold axis plunges to 44°→ 081°. Lamaque Late Diorite plugs are parallel to S 2 but plunge more steeply than the F 2 fold axis. Intersections of the mineralized veins are subhorizontal and lie within S 2 . Poles to the modeled bedding form surfaces (Fig. 12b) are shown in blue dots. Poles that plot in the south represent overturned south-facing bedding. b Gold grades (Au ≥ 30 g/t shown as red dots) with traces of 3D bedding (S 0 = colors) and S 2 (gray) form surface model viewed in F 2 down-plunge direction 44°→ 081°. This is the first-order mineralization continuity. c Same data with view parallel to the long axis (a-axis) of the Late Diorite plugs 76°→ 026°. This is the second-order mineralization continuity. d View horizontally along 00°→ 094°in the S 2 plane and along the average intersections of all veins. This is the third-order mineralization continuity reiterated further by Sibson (1989). The steep veins at Sigma were postulated to be parallel to a tabular shear zone that had an extent that was five times the vertical extent of drilling at Sigma mine. In other words, the 10 km vertical extent and the inferred constant steep orientation of the fault zone proposed by Sibson et al. (1988) is effectively the result of an "Inside-Out" interpretation workflow (as illustrated in Fig. 1) where smallscale field data are scaled up to infer a larger-scale structural feature on the assumption of scalability being valid (Pumpelly's Principle). This detail requires examination as the interpreted presence of a large-scale shear zone directly affects gold exploration at Sigma-Lamaque, and is also relevant to near-mine exploration at many other vein-hosted gold deposits. Although the evidence regarding the cyclic nature of vein formation due to repeated fluid pressure fluctuations may be consistent with the fault-valve model (Boullier and Robert 1992), the inference that the Sigma deposit is part of an extensive steep shear zone is inconsistent with the deposit-scale data described in this study. At the heart of this query is whether the formational mechanics, deduced from studying the outcrop-scale veins, can inform the working geologist of the exploration potential for the mineralized veins at the much larger deposit scale. The problems of inferring the existence of a large-scale brittle-ductile reverse shear zone at Sigma are: & The steep shear veins at the Sigma side of the Sigma-Lamaque gold deposit do not continue at the same dip along the total depth sampled and exhibit a lower dip at depths of 1500-2000 m (red traces in Fig. 14). & Steep reverse shear veins do not continue along a steep southerly-dipping shear zone parallel to the shear veins as predicted by Sibson et al. (1988). The enveloping surface of the veins at Sigma was documented by Robert and Brown (1986a: fig. 3) and Sibson et al. (1998: fig. 1) to be vertical or dipping steeply to the north, and this is confirmed in this study (Fig. 14). This pattern contradicts the southerly-dipping reverse-shear model, but was not emphasized in the discussion by Sibson et al. (1988). & The Sigma and Lamaque vein systems appear to converge at depth (Observation 5). This contradicts the tabularplanar shear plane geometry as represented by the Sigma set of veins as postulated by Sibson et al. (1988: fig. 2). Lamaque veins, which are the most productive goldbearing veins in the Sigma-Lamaque system, exhibit an enveloping surface geometry that dips steeply north; therefore, these veins cannot be reconciled with a southerly-dipping reverse shear zone as proposed by Sibson et al. (1988).
& The North and South Shears, which are vertical with a thickness of 2 m on average and laterally extensive to 1 km both along strike and down-dip, as described by Robert and Brown (1986a), were not integrated into the model by Sibson et al. (1988). These shears, noted also by Gaboury et al. (2001), appear to be geologically significant, as they bound the vein development at Sigma. Although the positions of the North and South Shears were not provided by Integra Gold, these shears appear to be at the same orientation as the modeled bedding trace (Fig. 14) and could represent flexural slip planes along lithological or bedding contrasts that were active during the folding of the Val-d'Or Formation. & In the plane-strain dip-slip shear model of Sibson et al. (1988), which was also reiterated by Robert et al. (1995), the gold distribution would be expected to exhibit the same symmetry as the mechanism proposed by Sibson et al. (1988). That is, the three-dimensional gold distribution at Sigma would exhibit the same symmetry as the proposed dip-slip mechanism (Patterson and Weiss 1961), so, in this scenario, the deposit-scale low-grade gold distribution would be expected to pitch down-dip parallel to the thrust movement along the proposed south-dipping reverse shear zone. Apparently supporting this prediction, Gaboury et al. (2001) noted that the mineralized envelope in the longitudinal section plunged parallel to the slip lineations measured within the Sigma reverse shear veins. In detail, however, the plunge illustrated by Gaboury et al. (2001: fig. 6) is at a smaller pitch angle than the large pitch angle expected from the measured slip lineations. The first-order long-range low-grade gold distribution at Sigma (as well as at Lamaque) pitches to the east at 45° (Fig. 9) and not 70°to 90°as expected, parallel to the measured slip lineations. This 3D pattern of gold distribution does not parallel the symmetry axes of the proposed dip-slip model of Sibson et al. (1988). & In contrast to south-over-north shearing in a narrow reverse shear as proposed by Sibson et al. (1988), bulk north-over-south plane-strain non-coaxial shearing better explains the asymmetry in total vein occurrences at Sigma-Lamaque (Fig. 14, inset). The prevalence of south-dipping shear veins at both Sigma and Lamaque (Fig. 7b) implies that the line of extension plunged steeply to the north during veining, as opposed to vertical, which would result in a symmetrical vein distribution with equal occurrences of north and south-dipping shear veins. The line of extension during the formation of the veins is interpreted to be roughly in the plane of the northdipping S 2 cleavage.
Despite the argument for high fluid pressure in the formation of the vein systems at Sigma-Lamaque (Sibson et al. 1988), the knowledge of this mechanism itself cannot help determine the 3D geometry of the enveloping surfaces of these gold-bearing veins, which is the critical information required for within-mine and near-mine exploration in the district. The localization of high gold grades at the stope scale is parallel to the inferred σ 2 direction, which is horizontal and E-W (Fig.  7); therefore, this symmetry is consistent with the vein symmetry at Sigma-Lamaque. However, the gold distribution inferred by Sibson et al. (1988), from the kinematics of Sigma's south-dipping shear veins at stope scale, clearly cannot be scaled up to infer a deposit-scale gold distribution; neither can it predict the position of gold deposition over long distances away from individual vein positions. As mentioned above, not only is gold distribution at Sigma steeply northerlydipping instead of south-dipping, but the expected first-order plunge of the gold distribution is also inconsistent with the proposal of Sibson et al. (1988). This illustrates a fundamental shortcoming of applying the "Inside-Out" interpretation workflow by scaling up data at the outcrop scale. Sibson (1996) concluded that the intermediate stress axis σ 2 is the axis of maximum structural permeability, but noted that the extent of this permeability "will depend on the 3D interconnectedness of the different structural elements." The vein patterns of the Sigma-Lamaque deposit clearly illustrate that the determination of the σ 2 direction from the orientation of stope-sized veins across the deposit (Fig. 7) is only relevant to the scale of the individual veins. That is, even if the measurements of vein orientations are obtained from across the sampled deposit, the inferred mean σ 2 direction cannot be used to extend the likely high-grade continuity hundreds of meters beyond the points of exposure because the extent of the vein swarms depend on the position of the outer enveloping surface of the vein swarm. The shape and continuity of the enveloping surface is the critical information required for ongoing within-mine exploration because it will affect the 3D interconnectedness of fluid flow (Sibson 1996). However, it is impossible to infer the vein swarm enveloping surface geometry from outcrop-scale observations of the veins or from vein orientation measurements. Instead, it requires a deposit-scale structural analytical approach by viewing and making structural sense of the entire mineralization pattern in 3D.

Nested hierarchical structural architecture
The 3D pattern of gold mineralization, and likely accompanying alteration envelope, at Sigma-Lamaque is a "nested" system with three orders of scale, each controlled by specific structural geological features. The gold continuity for each order can be truncated by the large-scale (lower-order) feature, and each order has its own structural symmetry that influences the grade continuity at that scale. The three orders of scale at Sigma-Lamaque are: 1. The first-order control is the zone of maximum dilation due to F 2 folding. For Sigma-Lamaque, it is the broad enveloping surface of the folded region in 3D (Fig. 12) and the mineralized volumes are bounded by local S 2 foliation and bedding (S 0 ). These mineralized zones plunge 44°→ 081°parallel to the intersection of S 0 and S 2 foliation (F 2 fold plunge) with the long-axis of the lowgrade envelope parallel to this plunge line. This first-order control, where grade distribution is constrained in place by both S 0 and S 2 , is obvious in the down-plunge direction (Fig. 13b). However, the grade "continuities" are not continuous parallel to this plunge direction at every location, and highly mineralized volumes can transition to unmineralized zones along this plunge line. This is consistent with patterns recorded in other fold-hosted deposits (Newhouse 1942). It is largely determined by the structural permeability imposed by the foliation, bedding, lithology, and bedding-parallel faults that propagate from bedding surfaces. These are part of the first-order structural control, but are also affected by the second-and third-order controls that are nested within the first-order control. 2. The second-order control is provided by the geometry and position of the steeply plunging, structurally late, Late Diorite plugs, at Lamaque. The line of continuity of the second-order features, which are steeply plunging at Sigma-Lamaque, can be predicted, but the second-order continuity can be truncated or affected by the first-order controls of bedding (S 0 ) and laterally restricted by the S 2 foliation. 3. The third-order control is the geometry of the auriferous quartz-tourmaline veins that develop within the host rocks. These veins are not strictly confined to within the diorites, but there is a positive correlation (Fig. 5). Without close drill sampling, it is not possible to predict the exact positions of these veins. The vein orientation patterns of Sigma-Lamaque (Fig. 4) may be used to explore for local extensions to short distances, but vein swarms are predicted to be truncated by the second-and first-order controls (described above) that define the enveloping surface of the veins. The extents of the veins at Lamaque are laterally limited by the second-order control of the Late Diorite plugs, whereas the veins at Sigma are laterally limited by the first-order control of steep bedding (S 0 ) and S 2 foliation.
Supplementary video file ESM3.mp4 shows how the high gold grade values in veins (third-order control) are laterally terminated by a combination of S 0 and S 2 foliation (first-order control). This animation also illustrates the bedding and Late Diorite plugs that parallel S 2 foliation.

Implications for orogenic gold mineralization and exploration
The greatest long-range exploration potential lies along the first-order trend, parallel to the F 2 fold plunge at Sigma-Lamaque. These first-order fold-controlled conduits are equivalent to the theoretical fluid pathways described in the mineral systems literature ( fig. 3 in Knox-Robinson and Wyborn 1997). At Sigma-Lamaque, the plunging conduit radius is~1.5 km (Figs. 11d and 13b) and the drill-sampled mineralized downplunge conduit length is~3 km (Fig. 9). This conduit is interpreted to tap the lower crustal fluid source due to the plunging nature of the fold axis, and sporadic gold mineralization is predicted to exist in the down-plunge direction. Similar foldcontrolled permeability pathways are possible at surface in the district surrounding the Sigma-Lamaque tenement within the folded Jacola and Val-d'Or Formations. Basic structural geological mapping using the traditional "Inside-Out" approach ( Fig. 1), looking for angularity between bedding and S 2 foliation (Fig. 10a), is one way to locate similar fold-controlled pipelike conduits where drilling data are unavailable. Where drilling data are available, the "Outside-In" workflow described in this study can be used to determine the first-order controls and thus potentially locate near-mine extensions to gold mineralization. However, locating a fold closure at surface may not necessarily lead to a discovery of gold mineralization there due to the possible presence of second-order controls down-plunge. Identifying a geochemical pathfinder and alteration anomaly expected at an outflow zone at surface is required before drilling to depth for possible mineralized pockets down-plunge (Beus and Grigorian 1977;Reis et al. 2001;Ord et al. 2012).
The first-order F 2 fold plunge trend, interpreted as the main pipe-like fluid conduit, is expected to be intersected internally by second-order controls; these are represented by the Late Diorite plugs at Sigma-Lamaque. These diorite plugs and other competent lithological units can be viewed as mineralization "traps", as discussed in mineral systems literature (Knox-Robinson and Wyborn 1997;McCuaig et al. 2010), which borrows terminology from petroleum systems literature that do not entirely apply to mineralization systems. More precisely, these fracture sites, formed in competent lithologies during lateorogenic compression (third-order control), feature the appropriate host composition to chemically react and precipitate gold from the fluid flowing along the first-order permeability pathway (Ord et al. 2012). Similar mineralization pockets hosted by diorite are predicted in the downplunge direction of F 2 . Late Diorite plugs are also expected in the district where there is well developed S 2 fabric, but the potential for gold mineralization may only occur if these bodies intersect with a first-order fluid conduit (Fig. 15). Supplementary video file ESM4.mp4 explains the proposed Sigma-Lamaque F 2 -paralell flow conduit model.
Despite the late-stage introduction of gold at Sigma-Lamaque, the distribution of the gold-as highlighted by the hierarchical arrangement of structural controls-is a function of the deformation history and resulting structural architecture. The structural architecture of the unmineralized rocks is critical to understanding the fluid pathway geometries-the structural patterns within these units offer the best clues to the geometric controls and the as yet undiscovered extensions of gold deposits. In orogenic gold systems in general, most effort should be placed on basic structural mapping that identifies the firstorder controls in potential host rocks away from known mineralization, rather than be distracted with details of third-order features such as gold-bearing veins. It is not possible to work out the large-scale architecture controlling the fluid pathways that existed in orogenic gold systems from vein orientations alone (e.g., Fig. 7) because the more important vein-swarmenveloping geometries are laterally limited by large-scale firstand second-order structural features.
Similarly, fine-resolution multi-parameter datasets obtained from drill cores cannot be used effectively in isolation without considering the large-scale structural architecture of the deposit. Current and future machine learning (ML) systems that maximize the knowledge gained from drill core data (Bérubé et al. 2018;Liu et al. 2019;Schnitzler et al. 2019) hold great promise for the minerals industry. Even though the benefits of ML are great, it is paramount that geologists know how to train these artificial computer algorithms to recognize large-scale hierarchical structural controls, otherwise ML systems may never yield sensible exploration targets.

Metamorphic vs magmatic-hydrothermal fluids debate
In the 1980s, a debate raged between researchers who saw Archean gold as originating from either metamorphic fluids (Phillips and Groves 1983;Groves et al.1987), or from synintrusion fluids (Burrows et al. 1986, Burrows and. A groundbreaking study by Jemielita (1989) unequivocally established a~90 million year time gap between the consolidation of Lamaque diorites (2685 ± 3 Ma, 2682 ± 2 Ma) and the quartz tourmaline veins that cross-cut and host the high-grade gold (2593 ± 5 Ma) (Fig. 3). Until that study, the relative timing of the intrusion and the gold-bearing veins was unclear. The veins could have been interpreted as roughly coeval with the consolidation of the Late Diorites, but the difference in nearly 100 million years was unexpected and significant, dispelling the igneous origin for gold (at least) at Lamaque. More studies followed that showed an unequivocal pattern of late-stage gold mineralization in the Abitibi Subprovince (Jemielita et al. 1990;Wong et al. 1991), and it is now generally recognized that gold mineralization globally occurred relatively late in the tectonic history of Archean greenstone belts (Groves et al. 1998(Groves et al. , 2018Dubé and Gosselin 2007;Ord et al. 2012).
Notwithstanding the apparent triumph of the metamorphic fluid theorists in the gold fluid origin debate 30 years' ago, this study raises an important structural geological relationship that was unrecognized until now-the plumbing system of the late-stage gold-bearing metamorphic fluids and igneous intrusions may have been shared at Sigma-Lamaque during D 2 deformation, as manifested by the emplacement of the Late Diorite plugs parallel to S 2 . This new interpretation legitimizes Spooner's intuition and research direction in the 1980s that was based on a critical geological observation-that gold occurrence and dioritic intrusions were spatially associated with each other in the Abitibi Subprovince (Burrows and Spooner 1989;Spooner 1993). The reason for this common occurrence appears to be the shared structural geological plumbing system between the metamorphic fluids and magmas and this has implications for other Archean orogenic terrains where there is co-spatial, but not necessarily co-temporal, occurrence of intrusions and gold (e.g., Allibone et al. 2004;Tourigny et al. 2018). The fluid plumbing system at Sigma was previously interpreted to be a tabular crustal-scale shear zone (Sibson et al. 1988), and this view has had significant impact on the interpretation of orogenic gold deposits (Dubé and Gosselin 2007). However, the evidence from this study, in terms of considering both Sigma and Lamaque, favors a linear foldcontrolled structural permeability pathway ( fig. 9c in Sibson 1996). Folds with moderate to steep plunges have the potential to tap the deep source of these fluids and behave like linear pipe-like conduits. Injected parallel to S 2 foliation, the pathway followed by the Late Diorite plugs at depth is unknown. They may continue down-dip parallel to the S 2 foliation and coalesce into a single dyke as depicted in Fig. 15, or the plugs may follow parallel to the easterly F 2 plunge direction.
The abundance of gold deposits clustered along this boundary leaves little doubt that the Cadillac tectonic zone is a crustal-scale suture, which acted as a primary conduit for fluids (Rafini 2014). Therefore, researchers have focused primarily on the architecture and structural history of the Cadillac tectonic zone to explain the gold distribution along this suture (Robert et al. 1995;Bedeaux et al. 2017). But the Sigma-Lamaque deposit is 5 km from this boundary, and it is debatable whether the local brittleductile strains recorded there are from the distributed strain that resulted from the movement along the Cadillac tectonic zone. Furthermore, there are many more gold deposits in the region that are well away from the Cadillac tectonic zone (Fig. 2). The first-order 3D structural architecture of the host rocks examined at gold deposits distal from the Cadillac tectonic zone is likely to reveal important structural features that focused the fluid flow at each deposit location. It is postulated that the late orogenic gold-bearing fluids, which may or may not have been sourced from Cadillac tectonic zone, would have taken advantage of any structural permeability that had previously existed in the host rocks at shallower crustal levels. These structural pathways are not necessarily associated with strains attributable to movements along the Cadillac tectonic zone.

Tectonic history during D 2 deformation
The structural controls defined in this study, together with previously published geochronological data, give a unique insight into the tectonic history of this part of the Abitibi Subprovince. The structural data presented in this study illustrate the remarkably simple relationships between the D 2 deformation architecture and the form of the Late Diorite intrusions at Lamaque, as well as the symmetry of the gold-bearing veins. Despite the implied temporary rotation of σ 2 to vertical during the Late Diorite plug emplacement, the σ 1 axis recorded by the veins at Sigma-Lamaque is virtually coaxial to the pole of S 2 (Figs. 7b and  14). This is significant because, if the tectonic history has a component of non-coaxial transcurrent strain in which stress and cumulative strain axes are not parallel to each other, there should be an appreciable angular difference between the pole of the cumulative strain fabric S 2 (or the minimum cumulative shortening axis λ 3 ) and the maximum stress axis σ 1 axis, determined from veins, which is an instantaneous record of strain (Fig. 7b). Due to the isotope ages established previously (Jemielita et al. 1989;Wong et al. 1991), the parallelism of σ 1 and λ 3 implies that there was little or no transcurrent deformation since the development of the S 2 fabric and the mineralized veins, which have a gap of~90 Ma between them. Therefore, the interpretation that D 2 is transpressional (Robert 1989) may only relate to the earliest folding stages of D 2 , but not the end stages after S 2 had developed and the mineralized quartz-tourmaline veins were formed at Sigma-Lamaque. The strain history suggests that the goldbearing veins at Sigma-Lamaque formed during N-S shortening, with the only non-coaxiality seen in the vertical plane, which implies north-over-south plane-strain bulk deformation with subvertical extension (Fig. 14). This is consistent with the end stages of a long protracted history of D 2 that formed the La Motte Anticline (Fig. 2), with gold mineralization formed at the end stages with its distribution controlled by the structural architecture on the steep southern limb of this anticline.

Conclusions
Deposit-scale structural analysis is a powerful "Outside-In" method to decipher the geometric controls of mineral deposits using the most readily available 3D data-the drill-hole data.
It is powerful because: 1. It does not rely on collecting modern sophisticated empirical data, but works on the most unsophisticated single column assay data that is abundant and readily available from historical deposits. 2. By applying basic principles of structural analysis, it rapidly establishes the geometric controls at the depositscale, which is the most relevant scale for mineral exploration and Resource definition. 3. It identifies the various structural controls at different scales, which cannot be readily identified using the traditional "Inside-Out" method of applied structural analysis. 4. Structural features that are likely to be seen in outcrop can be predicted using the "Outside-In" method and the interpretation can be readily field tested.
The interpretation of the Sigma-Lamaque gold deposit, made entirely from a historical drilling database, is a model that can be readily field tested. As with any model that explains natural patterns, it will likely be revised or even eliminated in the future if better geological interpretations can explain the deposit-scale patterns. However, what is most useful about deposit-scale structural analysis are the lessons learned by repeating the interpretation process many times by analyzing data from mature mines and advanced exploration projects. With experience, geologists can notice repeated structural patterns in assay data, and these geometric patterns can help them predict how sparsely drilled deposits are likely to behave in space, thus minimizing exploration uncertainty in new exploration projects. When grade distributions are understood as structural geological patterns, rather than merely treated as geometrical envelopes for Resource estimation, the lessons learned from this practice will greatly minimize the risk of Resource write-downs. Resource envelopes will be understood in the context of structural architecture, and continuities of grade can be predicted by integrating structural geological understanding. Structural interpretation and determining the control of orogenic gold deposits using assay data is not only possible but might become a key strategy to move away from the over-reliance of theoretical structural models that have had little impact on gold exploration, as discussed by Vearncombe and Zelic (2015). continuing discussions which are always very educational. This study, including the development costs of proprietary software Orefind used to analyze the data, and the payment of open access publication fee, was self-funded. Monash University is acknowledged for providing access to their online library.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.