Ore Geology Reviews 69 (2015) 268–284

Contents lists available at ScienceDirect

Ore Geology Reviews

journal homepage: www.elsevier.com/locate/oregeorev

Regional dome evolution and its control on ore-grade distribution: Insights from 3D implicit modelling of the Navachab gold deposit, Namibia

Stefan A. Vollgger a,⁎, Alexander R. Cruden a, Laurent Ailleres a, E. Jun Cowan b a School of Earth, Atmosphere and Environment, Monash University, Australia b Orefind Pty Ltd, Fremantle, Australia article info abstract

Article history: We introduce a novel approach to analyse and assess the structural framework of ore deposits that fully inte- Received 28 November 2014 grates 3D implicit modelling in data-rich environments with field observations. We apply this approach to the Received in revised form 20 February 2015 early Palaeozoic Navachab gold deposit which is located in the Damara orogenic belt, Namibia. Compared to tra- Accepted 20 February 2015 ditional modelling methods, 3D implicit modelling reduces user-based modelling bias by generating open or Available online 21 February 2015 closed surfaces from geochemical, lithological or structural data without manual digitisation and linkage of sec- fi Keywords: tions or level plans. Instead, a mathematically de ned spatial interpolation is used to generate 3D models that Folding show trends and patterns that are embedded in large drillhole datasets. In our 3D implicit model of the Navachab Ore shoots gold deposit, distinctive high-grade mineralisation trends were identified and directly related to structures ob- Implicit modelling served in the field. The 3D implicit model and field data suggest that auriferous semi-massive sulphide ore shoots 3D deposit modelling formed near the inflection line of the steep limb of a regional scale dome, where shear strain reached peak values Fold inflection lines during fold amplification. This setting generated efficient conduits and traps for hydrothermal fluids and associ- Orogenic gold ated mineralisation that led to the formation of the main ore shoots in the deposit. Both bedding-parallel and highly discordant sets of auriferous quartz-sulphide veins are interpreted to have formed during the later lock- up stage of the regional scale dome. Additionally, pegmatite dykes crosscut and remobilise gold mineralisation at the deposit scale and appear to be related to a younger joint set. We propose that kilometre-scale active folding is an important deformation mechanism that influences the spatial distribution and orientation of mineralisation in ore deposits by forming structures (traps and pathways for fluids) at different preferred sites and orientations. We also propose that areas that experience high shear strain, located along the inflection lines of folds can act as preferred sites for syn-deformational hydrothermal mineralisation and should be targeted for regional scale ex- ploration in fold and thrust belts. Our research also suggests that examination of existing drillhole datasets using 3D implicit modelling is a powerful tool for spatial analysis of mineralisation patterns. When combined with fieldwork, this approach has the potential to improve structural understanding of a variety of ore deposits. © 2015 Elsevier B.V. All rights reserved.

1. Introduction such fluids because of their ability to transport gold and related compo- nents (Phillips and Powell, 2010). Ore deposits form where hydrother- Ore deposits are economic concentrations of chemical elements that mal fluids are either chemically (e.g., fluid mixing, host rock reactions, result from a number of processes that occur within the Earth's crust pH change) or physically trapped (e.g., rapid change of pressure or and deliver elements from a source region to a deposit location. These temperature). The importance and efficiency of these two trapping processes can be chemical and/or physical and they control the libera- mechanisms is still under debate. However, abundant research has tion, transport, focusing and trapping of hydrothermal fluids (Cox and shown that many gold deposits are hosted in deformed rocks, suggest- Etheridge, 1987; Groves et al., 2000; Ridley, 1993; Robb, 2004; ing a strong physical (structural) control on mineralisation. Famous Tomkins, 2010). Most genetic models of gold deposits are ascribed to examples are the lode gold deposits of the Victorian goldfields (Cox et al., 1991, 1995; Leader et al., 2012; Mueller, 2014; Willman, 2006; Willman et al., 2010; Wilson et al., 2013) and the Yilgarn Craton, West- ⁎ Corresponding author at: School of Earth, Atmosphere and Environment, Monash ern Australia (Cox and Ruming, 2004; Groves et al., 1995, 2000; University, Clayton, VIC 3800, Australia. Hodkiewicz et al., 2005; Micklethwaite and Cox, 2006; Ridley and E-mail addresses: [email protected] (S.A. Vollgger), [email protected] (A.R. Cruden), [email protected] (L. Ailleres), jun.cowan@orefind.com Mengler, 2000). Structures associated with these deposits such as (E.J. Cowan). folds, faults, veins and shear zones highlight the importance of a

http://dx.doi.org/10.1016/j.oregeorev.2015.02.020 0169-1368/© 2015 Elsevier B.V. All rights reserved. S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 269 changing history of brittle and ductile deformation for economic Rescources, 2012; Lombard et al., 2013) as part of B2Gold's Otjikoto pro- mineralisation. These changes are important because they influence ject (formerly operated by AuryxGold) and Damara Gold Corp.'s Gold fluid flow and control the formation of dilational sites that can act as Kop target (formerly Helio Resources). These have both been described traps for hydrothermal fluids (Leader et al., 2011). It follows that the as having a similar style of mineralisation to that at the Navachab gold shape, size, orientation and spatial distribution of ore bodies contain im- deposit. portant information about the deformation history of an ore deposit and The Navachab gold deposit is located in the S part of the Central Zone the associated interactions between host rock types, fluid flow and of the Damara belt, also known as the Southern Central Zone. The Cen- structures (Blenkinsop and Kadzviti, 2006). Work by Blenkinsop tral Zone is one of several tectonostratigraphic domains within the Pan- (2004) on two lode gold deposits from the Zimbabwe craton, studies African Damara orogen (Miller, 1983)(Fig. 1). The majority of mineral by Bauer et al. (2014) on VMS deposits in the Swedish Skellefte district, deposits are located in the Central Zone and are associated with rocks 3D modelling and analysis of nickel sulphide shoots at the Kambalda that have been subjected to medium- and high-grade metamorphism dome (Western Australia) by Stone et al. (2005) and investigations on (Pirajno and Jacob, 1991). the vein-hosted gold ore body at Sunrise Dam (Western Australia) by The Damara belt extends to the NE and links with the Lufillian Hill et al. (2013) demonstrate that shapes and orientations of ore bodies orogen in Zambia–Angola–Congo (Pirajno, 2008). The term Pan- or mineralised zones contain important information about the evolu- African orogeny is commonly used to describe magmatic, tectonic and tion of ore deposits. The analysis of structures that transport and trap metamorphic activity during the Neoproterozoic (~870 Ma) to early hydrothermal fluids is therefore of major scientific and economic Palaeozoic (~550 Ma) orogenic episode (Kröner and Stern, 2004). significance. By understanding the genetic relationship between The tectonic evolution of the Damara belt started with the initiation (mineralised) structures and associated physical processes, we can of an intracontinental, asymmetric rift between the Congo craton and potentially predict areas with a greater potential for mineralisation in the Kalahari craton during the break-up of Rodinia around ~780– similar tectonic environments. In this paper, we present a structural 750 Ma (Barnes and Sawyer, 1980; Miller and Frimmel, 2009; Miller interpretation workflow that uses 3D implicit modelling to visualise et al., 2009). The sedimentary fill of the newly opened graben in the S and analyse geochemical and geological trends from an extensive (Nosib Group) exceeds 6 km in thickness and consists of feldspathic drillhole dataset from the Navachab gold deposit, Namibia (provided sandstones, playa lake and evaporitic sabkha deposits (Miller et al., by AngloGold Ashanti Ltd). Such drillhole data are normally used to 2009). Rifting led to siliciclastic and carbonate sedimentation before compute block models for the economic assessment of a project. For and during the subsequent formation of an oceanic basin (Khomas each block of these models, important key values and properties are ei- sea). The final width of the Khomas sea is estimated to about 1200– ther assigned or calculated, such as net present value, metal content, 1500 km (Miller et al., 2009). The Matchless Amphibolite is the meta- density or rock type. Block models are powerful and well-established morphosed remnant of oceanic crust (sea-floor basalt) which formed tools for resource estimation, grade control and scheduling, where the during that time, and is host to several VMS and (epigenetic) copper de- main purpose is to evaluate, maximise and optimise the economics of posits (Breitkopf and Maiden, 1988). a project. However, they are less valuable for understanding the geology Reversal of plate motion and formation of a N dipping subduction and structural framework of a deposit. zone led to the closing of the Khomas sea and the collision of the Recent developments in geological modelling software (e.g., 3D Kalahari craton in the S with the Congo craton in the N around implicit modelling) allow for the precise and objective generation of 540 Ma to 520 Ma. Convergence culminated in the formation of the spatial representations of lithological contacts, mineralisation (grade Cambrian supercontinent Gondwana and caused thrusting and folding shells) and structural trends in data-rich environments. Instead of labo- within the Damara belt. In the Karibib area, the thrusting and folding rious manual cross-section digitisation, spatial interpolations of geolog- at 560–540 Ma is related to an early phase of crustal thickening owing ical data are employed to produce models with minimised user-induced to continent–continent collision (Longridge, 2012). modelling bias. In our case study of the Navachab gold deposit, initial 3D Navachab is located in deformed, primarily metasedimentary rocks implicit modelling was followed by structural fieldwork and data collec- (calcsilicates, marbles, schists and metapelites) of the several km thick tion in the Main Pit and Eastern Zone 1 of the mine. Field observations Neoproterozoic Damara Supergroup, which is floored by Mesoproterozoic and structural data were compared to the orientation, size and spatial (~2.0 to ~1.7 Ga) gneisses of the Abbabis Metamorphic Complex (Jacob distribution of implicitly modelled ore bodies and host rock stratigra- et al., 1978; Kisters, 2005). Syn- and post-tectonic granitic magmatism phy. Additionally, a 3D implicit structural trend model computed from is widespread and voluminous in the Damara belt and is reflected as regional bedding orientation measurements was used to calculate and post-tectonic thermal peaks in thermochronological data (Miller and visualise the spatial distribution of shear strain related to the ampli- Frimmel, 2009; Tack and Bowden, 1999). Slab breakoff caused astheno- fication of a regional scale domal structure. This structural trend spheric upwelling and heating of the lower crust and provided heat for model allowed us to evaluate the regional structural context of the metamorphism and melting of the crust around 540–535 Ma to pro- gold mineralisation, suggesting that auriferous ore shoots within duce anatectic red granite (Longridge, 2012). Several diorite and the Navachab deposit were controlled by the regional-scale, non- leucogranite intrusions are located within 5 km of the Navachab deposit, cylindrical folding responsible for the formation of the Karibib dome. but there is no direct contact between the mineralised system and any Our case study illustrates how 3D implicit modelling in conjunction major intrusive body (Steven et al., 2011). with selective fieldwork can aid in the extraction of additional structural Metamorphic conditions within the Southern Central Zone decrease and ore-genetic information from existing datasets. At the Navachab towards the NE (along strike) from granulite facies (P ~ 5 kbar, T ~ 700– gold deposit, it improved and extended understanding of the structural 800 °C, Nex et al., 2001; Masberg, 2000; Masberg et al., 1992) with par- framework that controlled syn- and post-deformational mineralisation, tial melting (Toé et al., 2013) in the SW coastal regions to lower pointing towards a strong relationship with regional scale folding. amphibolite-facies (P ~ 3 kbar, T ~ 560–650 °C, Puhan, 1983; Steven, 1993) in the Karibib district, as progressively shallower stratigraphic 2. Regional geology levels are exposed (Kisters et al., 2004). The temperature-dominated metamorphism in the Karibib district has been associated with periods The Navachab gold deposit is situated approximately 10 km SW of of voluminous intrusions (Jung and Mezger, 2003; Miller, 1983), which the town of Karibib (Namibia). The current open pit activities represent led to pervasive recrystallisation and obliteration of tectonic fabrics, espe- the only fully operating gold mine in Namibia. However, in recent years cially in marbles throughout the Southern Central Zone (Kisters, 2005). new discoveries of gold mineralisation have been reported approxi- The Southern Central Zone is bound by two roughly parallel, NW mately 20 km and 150 km NE of the present mine site (Helio trending, major lineaments which coincide with large-scale facies 270 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

Fig. 1. Geological map of Namibia showing tectonostratigraphic domains of the Damara belt after Miller (1983). The Navachab gold deposit is located within the Southern Central Zone of the Panafrican Damara belt, between the Omaruru Shear Zone in the N and the Okahandja Shear Zone in the S. Data source: Geological Survey of Namibia.

changes (Miller, 1983). The Okahandja lineament (Okahandja shear 3. Main characteristics of the Navachab gold deposit zone) was the locus of continental break up and represents the N bounding fault of the S graben during rifting as well as the leading The Navachab gold deposit has been owned and operated by edge of the overriding Congo craton during convergence (Miller et al., AngloGold Ashanti Ltd since 1998, but was acquired by QKR Corporation 2009, 2010). The Omaruru lineament (Omaruru shear zone) forms the Limited in 2014 (G. Bell 2014, pers. comm.). At the end of 2012, S boundary of the N graben and has been interpreted to have acted as Navachab had gold mineral resources of 4.4 Moz and reserves of 2.1 a growth fault throughout the spreading phase, associated with substan- Moz. The mine produced 46,000 oz of gold in the first nine months of tial sedimentary-exhalative activity (Miller et al., 2009). The Navachab 2013. The Navachab gold deposit can be divided into three main deposit lies between these two first-order, major crustal structures areas: Main Pit, North Pit and Eastern Zone 1 (Fig. 3). (Fig. 1). Elongated, kilometre-scale NE trending domal structures define the 3.1. Lithostratigraphy regional structural grain of the Southern Central Zone. One such dome is the Karibib dome, which hosts the Navachab gold deposit at its steep The Navachab gold deposit is hosted within NE striking, steeply NW NW limb, formed by SE-NW shortening (Fig. 2). Kisters et al. (2004) sug- dipping amphibolite facies marbles and schists of the Damara Super- gested that a blind thrust led to the formation of the Karibib Dome as a group, which form the NW limb of the kilometre-scale Karibib dome tip-line fold. The moderately steep dipping S limb of the Karibib dome is (Fig. 2). The Karibib dome is a shallowly doubly-plunging, NW verging overridden from the SE by the Mon Repos thrust zone, where top-to- asymmetric antiform with a length of about 12 km and a width of ap- the-NW movement pushed crystalline gneisses of the Abbabis Metamor- proximately 4 km. The oldest rocks in the area around the Navachab phic Complex over the younger Damara Sequence (Kisters et al., 2004). gold deposit are Palaeoproterozoic gneisses, schists, amphibolites and SHRIMP U–Pb dating of single zircons from syn-kinematically intruded, pegmatitic rocks of the Abbabis Methamorphic Complex (Jacob et al., foliated diorites and granodiorites (Mon Repos diorite-granodiorite, 1978; Tack and Bowden, 1999), which are exposed in the core of the 546 +− 6 Ma and 563 +− 4 Ma) and post-tectonic, undeformed Karibib dome (Kisters, 2005), but do not crop out at the Navachab monzogranites (Rote Kuppe granite, 543 +− 5 Ma) has provided con- mine. The Damaran metasedimentary rocks unconformably overlie straints on the timing of this regional thrusting event (Jacob et al., 2000; the Abbabis Metamorphic Complex starting with arkoses, quartzites, Kisters et al., 2004). conglomerates and minor metavolcanic rocks of the Etusis formation S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 271

Fig. 2. Simplified geological map of the Usakos and Karibib domes after Kisters (2005) and Kisters et al. (2004). Note that the Navachab gold deposit is located within rocks of the Damara Supergroup at the NW limb of the Karibib dome. The Karibib dome is overridden by the Mon Repos thrust zone to the S.

(Nosib Group), representing the Damaran rift succession (Miller, 1983). The Okawayo formation is intruded by meta-lamprophyre sheets The up to 1500 m thick Etusis formation can be observed as dominant and dykes. U–Pb SHRIMP analyses of titanite grains from a meta- ridges to the SE of the Navachab gold deposit (Kitt, 2008), but is only lamprophyre dyke yielded an age of 496 +− 12 Ma, which probably re- exposed within the oxidised zone of the Eastern Zone 1 at the mine cords a metamorphic overprint (Jacob et al., 2000). At the Navachab site. It consists of highly oxidised and weathered, interbedded layers gold deposit, the youngest igneous activity occurs as bedding-parallel of sandstone and siltstone. Stratigraphically higher is the Chuos forma- and mainly highly discordant, subvertical pegmatite dykes of unknown tion, which also crops out in Eastern Zone 1 and consists mainly of age. metaquartzites in the upper part and a diamictite in the lower part. These have been interpreted as glaciomarine in origin and related to the 3.2. Gold mineralisation late Neoproterozoic (746 +− 2 Ma) Sturtian glaciation (Hoffmann et al., 2004). Finely distributed, bedding-parallel sulphides (mainly pyr- Mineralisation at the Navachab gold deposit is associated with pyr- rhotite) have been observed in the Chuos formation. The Chuos formation rhotite and minor amounts of pyrite, chalcopyrite, arsenopyrite, sphal- is overlain by the biotite-schist and calcsilicate rock dominated Spes Bona erite, maldonite, bismuthinite, native bismuth and bismuth tellurides formation. The contact of the Chuos formation with the overlying Spes (Dziggel et al., 2009; Nörtemann et al., 2000; Wulff et al., 2010). Gold oc- Bona formation has been mapped in Eastern Zone 1. The Spes Bona for- curs mainly as native gold, but is also present in minerals such as mation has a true thickness of up to 220 m and is overlain by the up to maldonite (Au2Bi) and as a solid solution with bismuth (Nörtemann 160 m thick Okawayo formation, which mainly consists of marbles and et al., 2000). Stable isotope (O, H, C, S) analyses by Wulff et al. (2010) also hosts the auriferous semi-massive sulphide ore shoots at its footwall suggest that gold precipitated from a metamorphic, mid-crustal fluid contact. This basal part is dominated by a cm to dm banded sequence of derived from Damaran metapelites that underwent prograde metamor- calcsilicate rocks and marbles. The uppermost part of the Okawayo forma- phism under amphibolite (ca. 550 °C and 2 kbars) to granulite-facies tion comprises greyish massive and brecciated (dolomitic) marbles. The conditions. latter have been interpreted to be sedimentary breccias (Kitt, 2008). At the Navachab gold deposit, mineralisation occurs in two main Overlying the Okawayo formation are metasedimentary and minor styles: 1) bedding-parallel semi-massive sulphide bodies, referred to calcsilicate rocks of the Oberwasser formation, which is up to 180 m as ore shoots (or massive skarns, Miller, 1983), and 2) bedding- thick in the Main Pit. The Oberwasser formation is compositionally parallel and highly discordant sets of auriferous quartz-sulphide veins. very similar to the Spes Bona formation. The ore shoots are discrete, elongate zones within a planar contact sur- The marble-dominated Karibib formation lies on top of the face that contain higher metal contents than the adjacent parts of the Oberwasser formation and is followed by metapsammitic and metapelitic host rock (Peters, 1993a). The two known ore shoots in the Navachab rocks of the Kuiseb formation. The latter represents the uppermost forma- gold deposit are economically defined by relatively high gold grades of tion of the Damara Supergroup, but it does not crop out at the Navachab 3–7 ppm (Steven and Badenhorst, 2002). One major ore shoot (main mine. shoot) is located in the Main Pit, the second, smaller and less extensive 272 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

Fig. 3. Digital elevation model (DEM) of the Navachab gold mine site, which is divided into the Main Pit, North Pit and Eastern Zone 1. In this study, fieldwork focused on high-grade areas (hot colours) within the Main Pit and Eastern Zone 1. See Figs. 6 and 8 for cross section A.

ore shoot (2nd shoot) is located at a structurally higher level and is (Kisters, 2005) support a relatively late stage and alteration-related or- mined in the North Pit. Both ore shoots are defined by numerous, igin for the most of the calcsilicate banding within the MC unit. mineralised semi-massive sulphide lenses that can be up to 5 m high The second style of gold mineralisation is hosted in flat-lying, verti- and 1.5 m wide in cross-section (Kisters, 2005). The semi-massive sul- cally stacked, quartz-sulphide veins that crosscut the subvertical phide lenses are aligned to form a shallowly NNE plunging, high-grade bedding of the host rocks at high angles. In marble host rocks, quartz- gold trend with a down-plunge extent of at least 1800 m. The semi- sulphide veins can mostly comprise pyrrhotite with minor quartz, massive sulphide lenses are elongated to typically rhomb-shaped bodies whereas in biotite schist and banded calcsilicate host rocks quartz- that are parallel to the compositional layering and locally show evidence sulphide veins mainly consist of quartz with up to 30 vol.% sulphide for hydraulic brecciation (Wulff, 2008). The semi-massive sulphide lenses minerals such as pyrrhotite (Dziggel et al., 2009). Additionally, a rela- consist of pyrrhotite (up to 50 vol.%), garnet (10–30 vol.%), clinopyroxene tively subordinate quartz-sulphide vein set occurs parallel to bedding, (5–10 vol.%, mainly diopside), minor chalcopyrite, sphalerite, bismuth, particularly in well-bedded units such as the Oberwasser Formation bismuthinite, bismuth-tellurides, arsenopyrite, as well as quartz, biotite and Spes Bona Formation (Kisters, 2005). and K-feldspar (Dziggel et al., 2009; Nörtemann et al., 2000). Several genetic models for mineralisation at the Navachab gold de- The ore shoots are located in the marble and calcsilicate rock domi- posit have been suggested. The close spatial and temporal association nated, approximately 30 m thick MC unit at the base of the Okawayo of the deposit to Pan-African granitoids (Nörtemann et al., 2000; formation, near the contact with the underlying (partly silicified) Pirajno and Jacob, 1991), the presence of skarn-type alteration assem- biotite-schist rich Spes Bona formation. Studies of the calcsilicate min- blages, as well as an unusual metal association of Au–Bi–As–Cu–Ag erals (Dziggel et al., 2009; Wulff, 2008) as well as structural evidence (Dziggel et al., 2009) have been used to suggest an origin as an S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 273 intrusion-related gold deposit (Thompson et al., 1999). However, recent reproducible process in which geological interpretation is inherited structural studies by Kisters (2005), Kitt (2008),andCreus (2011) and from the outset; hence its application for purposes of structural inter- geochemical investigations by Wulff et al. (2010) point to formation pretation must be viewed with caution. as an orogenic gold deposit. Alternatively, implicit modelling is a technique where surfaces and volumes (=closed surfaces) are not explicitly defined, but mathemati- 3.3. Structural history cally described as isovalues of a volumetric scalar field (Calcagno et al., 2008; Carr et al., 1997; Caumon et al., 2013; Frank, 2006; Jessell et al., Three main phases of deformation have been described in the Cen- 2014; Maxelon et al., 2009; Mcinerney et al., 2005)(Fig. 4). Scalar fields tral Zone of the Damara belt (Barnes and Sawyer, 1980) affecting both are computed based on a global interpolation function (e.g., radial basis pre-Damaran basement (Abbabis Metamorphic Complex) and the function; RBF) that is fitted to data points. Early applications of the RBF metasedimentary rocks of the Damara Supergroup. However, only two interpolation scheme were employed by Hardy (1971) to calculate ele- main deformation phases have been recorded in Damaran rocks within vation contour lines based on irregular topographical data. However, the Navachab deposit. The oldest deformation is expressed by a the main development of this scalar field-based technique originated bedding-parallel or bedding-subparallel S1 foliation defined by the pre- from developments in computer graphics, where the aim was to recon- ferred orientation of mica in metasedimentary rocks, visible within the struct solids and surfaces from scattered point data for applications in Spes Bona and Oberwasser formations. Flattened clasts in diamictites of tomography, LiDAR, animation and visualisation of complex shapes the Chuos formation observed in Eastern Zone 1 are also parallel to S1. (Carr et al., 1997, 2001; Savchenko et al., 1995; Turk and O'Brien, Transposition of bedding (S0) into the S1 fabric impedes a full character- 2002). The first time scalar-fi eld based modelling was used in a geolog- isation of S1. However, as also observed by Kisters (2005), rare isoclinal ical context was by Lajaunie et al. (1997), who referred to it as implicit intrafolial folds mapped in banded marbles and calcsilicates of the form and used it to spatially interpolate structural data. Another geolog-

Okawayo formation within the Main Pit indicate that S1 is a composite ical application of this interpolation method was by Billings et al. S0,1 fabric. S1 is most probably related to an early D1 low angle thrusting (2002), who used so-called continuous global surfaces to interpolate event (Kisters, 2005). geophysical datasets. Cowan et al. (2002) first introduced rapid geolog-

The main deformational phase is D2, which is characterised by up- ical modelling using Leapfrog, a software package that employs RBFs to right, km-scale NE trending folds with doubly-plunging fold axes, spatially interpolate large drillhole datasets to build 3D geological such as the Karibib dome. The NE striking axial planar foliation (S2)of models of ore deposits. This RBF-based interpolation is computationally the Karibib dome dips steeply SE and indicates an asymmetric, NW expensive and was initially considered to be impractical (Sibson and verging fold geometry that formed during NW-SE shortening. Conse- Stone, 1991). However, mathematical advances resulted in the ability quently, the NW limb of the Karibib dome dips more steeply than the of RBFs to rapidly interpolate large datasets (Beatson et al., 1999). SE limb. Cowan et al. (2003) introduced the term implicit modelling to the min- ing industry and highlighted the advantages that arise from an implicit- 4. 3D geological modelling ly defined 3D model compared to traditional explicit 3D modelling methods that rely on manual digitisation. Other implicit modelling Selecting an appropriate modelling technique not only depends on the schemes have been referred to as implicit surface representation, poten- geological complexity and the available data (Guillen et al., 2008), but also tial field interpolation or potential field method (not to be confused on the intended function or purpose of the model (Stachowiak, 1973). with geophysical modelling of potential field data) in subsequent publi- The two main methodologies used for 3D geometrical modelling of ore cations (Calcagno et al., 2006; Chilès et al., 2004; Lane et al., 2007; Ledez, deposits, termed explicit and implicit modelling (Cowan et al., 2003; 2003; McInerney et al., 2007; Mcinerney et al., 2005). Jessell et al., 2014; Ledez, 2003) are reviewed brieflybelow. The proliferation of terminology for 3D geological modelling Before the advent of modern software and hardware, geological methods that share a common mathematical approach has led to an in- findings were hand drawn and stored on paper (2D) as sections and consistent understanding of the term implicit modelling. To clarify, in maps. Even though routinely transferred into digital formats, such sec- implicit modelling, an implicit function (generic format f(x, y, z)=0) tions and maps still form the foundation of traditional 3D explicit geo- is computed and fitted to a set of spatial data (points, lines, structural logical models (Cowan et al., 2003). Explicit models rely on manual data) (Fig. 4a). This implicit function describes a volumetric scalar definition of geological boundaries by digitisation of sections that are field, which assigns a scalar value to every point in space (Fig. 4b). later linked to generate 3D bodies and surfaces. The data are stored Isovalues of this field can be extracted and used to define open or closed digitally and can be visualised in 3D, which provides major advantages 3D isosurfaces, which can be envisaged as 3D analogues of 2D contour (Vollgger and Kassl, 2010). It does, however, still underutilise the lines (Fig. 4c). Since the term implicit modelling refers to the mathemat- inherent possibilities of 3D modelling. The basic concept of this method- ical description of a surface, it should only be applied in geological ology emerged out of computer aided design (CAD) applications. CAD models that comprise surfaces obtained from a scalar field. To visualise programmes were originally developed for industrial design purposes, these surfaces an independent step, that commonly includes the where the manual (explicit) drawing approach is appropriate because marching cubes algorithm (Lorensen and Cline, 1987), is necessary to the shape and dimension of objects are precisely defined and known be- generate triangulated meshes that can be rendered on a screen. Theo- forehand. However, this does not necessarily apply to 3D geological retically, a surface of any resolution could be extracted from an implicit modelling in data-rich and geometrically complex geological environ- function. However, practical and computational limitations will impact ments. The manual definition of each element makes explicit modelling the final resolution of the rendered surface. workflows time consuming and also introduces a strong bias inherited In general, implicit modelling can be used to spatially interpolate nu- from geological interpretation during digitisation and linkage of cross- merical (e.g., assay) and non-numerical (e.g., lithology) data points to sections, resulting in unique and non-reproducible models (Cowan define scalar fields. In the case of non-numerical data so-called signed et al., 2002, 2003, 2011; Jessell et al., 2010; Lindsay et al., 2012, 2013a, distances are calculated for each data point, which describe the distance b). Additionally, manual linking of cross-sections with complex geome- to a surface of interest (e.g., the contact between two geological forma- tries regularly forces the modeller to adapt and simplify the model to tions) (Carr et al., 2001; Cowan et al., 2003). This boundary is then rep- stay within a practical timeframe (Cowan et al., 2003), hence inhibiting resented by an isosurface that smoothly fits all data points with a signed the graphical representation of the full structural and geological com- distance value of 0. In addition to isosurfaces based on lithological data, plexity of the ore deposit. In summary, the manual approach used in ex- geochemical data (assay data) can be used to compute isosurfaces (re- plicit geological modelling is a subjective, time-intensive and a non- ferred to as grade shells) at various threshold grade values. Moreover, 274 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

Fig. 4. 3D implicit models are constructed using isosurfaces (c) that are derived from a scalar field (b). This scalar field is defined by the spatial interpolation of points, lines, and structural measurement data (a). The latter influence the gradient of the scalar field and control the shape and orientation of the associated isosurface. when dealing with multi-element datasets, various grade shells for each introduced based on knowledge and experience to generate a model element can be rapidly computed and visualised. that complies with the geological understanding at the time of creation. Implicit modelling also permits the direct integration and processing To avoid any bias that may be inherited from models that have been mod- of planar structural measurements, producing structural isosurfaces by ified based on assumptions instead of measured data, the modelling adjusting the gradient of the scalar field, as described in detail by workflow described in this paper is limited to data-rich environments, Hillier et al. (2013). Computed surfaces are referred to as structural where high-data density allows direct computation of boundaries and trend surfaces and can be used to visualise the structural grain of an surfaces without manual, knowledge driven digitisation. For objects area in 3D. An important characteristic of implicitly defined (iso-) sur- with a narrow geometry such as dykes, this criterion is rarely met and un- faces is their ability to describe complex geometries, such as overturned aided 3D implicit modelling without manually introduced constraints do folds or domes (Cowan et al., 2003; Jessell et al., 2010, 2014; Wellmann not deliver satisfactory results. Another shortcoming is that RBF-based et al., 2010). spatial interpolation tends to generate balloon-like grade shells in areas The direct link between data and the 3D model is a major advantage where insufficient control data are available, such as the outer edges of of the implicit modelling method; updates and changes are straightfor- a model (e.g., Fig. 8, long section). Furthermore, the type of RBF (Gaussian, ward to make when additional or new data becomes available, which is multiquadric, or polyharmonic spline) used for the spatial interpolation an attribute that has partly driven its development (McInerney et al., will affect the resulting model (Jessell et al., 2014). However, in dense 2007). This direct link also eliminates the directional bias that is inherent data environments the impact of the chosen RBF appears to be minimal in traditional, manually defined explicit models, which are commonly (Hillier et al., 2014). Nevertheless, care must be taken when interpreting based on a series of parallel cross-sections that are often complemented such models and also when using the calculated grade-shell volumes for by additional long-sections, surface and level maps. Geological trends or an economic assessment of an ore deposit. structures oriented parallel or at low angles to these sections are com- Several 3D mining and geological modelling software packages such monly overlooked and become under-represented in such models. There- as GOCAD (Paradigm, 2014), earthVision (Dynamic Graphics Inc., 2014), fore, in geologically complex and poly-deformed environments where GeoModeller (Intrepid Geophysics, 2014), Vulcan (Maptek, 2014), Surpac mineralisation and structures are expected to have various orientations, (Geovia, 2014), Micromine (Micromine, 2014)andPetrel(Schlumberger, an implicit modelling approach should be preferred. 2014) have incorporated (semi-) implicit modelling engines. Implicit modelling removes the need for time-consuming rebuilding of wireframes, allowing the continual development of geological 4.1. 3D implicit modelling of the Navachab gold deposit models throughout the life of a project (Mortimer, 2010). In addition, it also supports the testing of multiple hypotheses (Jessell, 2001)ina In our study we aim to structurally interpret a 3D geological model of reasonable amount of time by modifying parameters (e.g. isotropic the Navachab gold deposit, based on available drillhole data and structur- versus anisotropic interpolations), applying structural trends (directly al data extracted from geological maps and collected in the field. It is calculated from structural measurements) or by adapting boundaries therefore important to generate objective 3D models that delineate geo- or trend. Concatenated files are automatically updated, keeping the logical boundaries, geochemical anomalies (ore bodies, mineralised model coherent and giving the geologist more time to focus on geological zones) and structural trends. Our 3D implicit modelling workflow is con- issues and to evaluate the plausibility of different interpretations. Fur- fined to areas of high drillhole data density which allows us to directly thermore, implicit modelling is attractive because it makes 3D geological compute 3D models without manual interaction (e.g., digitisation). This modelling a reproducible task, which is required for the quantification of is quite different from a traditional explicit modelling approach, where geological uncertainty in 3D models (Ailleres et al., 2010; Chilès et al., 3D models are based on cross-sections that have assumptions about the 2004; Lindsay et al., 2010, 2012, 2013a,b; Wellmann et al., 2010). geological and structural framework of the deposit already embedded. There are, however, some limitations in using an implicit modelling The aim of our workflow is to identify and analyse trends in large drillhole workflow. Computed surfaces will always try to achieve a smooth fit datasets and relate them to structures observed in the field. We also test due to the nature of the RBF-based spatial interpolation. In data-rich the hypothesis that if mineralisation is structurally controlled or influ- environments with large amounts of control data this is a less relevant enced, modelled ore-body geometries, orientations and spatial locations problem. However, with sparse datasets the resulting model will will outline and reveal associated first- or lower-order structures and con- become over-smoothed and will not reflect sudden changes or steps tribute to the understanding of the processes that formed them. such as fault offsets. The lack of control data is generally problematic For this study, the implicit modelling software package Leapfrog for all types of 3D modelling (explicit or implicit). In such under- Mining 2.6 (ARANZ Geo, 2014) was chosen because it has the capability constrained environments, additional data are commonly manually to process large amounts of data (several thousands of drillholes, S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 275 millions of samples) using an RBF-based algorithm on ordinary comput- 3D implicit model. For the 3D implicit structural trend model a regional ing hardware. Data for implicit modelling were extracted from a topographic model was generated based on SRTM 90 m digital elevation drillhole database provided by AngloGold Ashanti Ltd in 2012. It includ- data (source: http://srtm.csi.cgiar.org/). ed gold assay data, lithological data (drillhole logs), collar data and sur- Data verification and error checking are the first important steps of a vey data for boreholes drilled between May 1986 and January 2012. 3D implicit modelling workflow and are essential for the robustness of Maps of the regional geology and digital elevation models (DEM) of the resulting 3D geological model. Automated procedures during data the region and the mine were also provided and incorporated into the import and 3D visualisation helped to identify inconsistencies within

Fig. 5. 3D implicit modelling workflow applied to drillhole data (a & b lithology, c & d gold assays) and regional structural data (e & f). All datasets were processed using an isotropic in- terpolation (no trend applied) without any manual digitisation of contacts or cross-sections. In a subsequent step, the three resulting models were used for visual data analysis. These geo- metric models were combined with field observations to develop a conceptual model, which aims to explain the structural evolution and control of mineralisation within the Navachab gold deposit. 276 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 the extensive drillhole dataset from the Navachab gold deposit 70°–75° to the NW, but steepens to 85°–90° at depth and rotates slightly (i.e., duplicate samples, missing samples, missing survey information, to a more E orientation in the NE; and 2) the true thickness of the incorrect positioned drillholes). In summary, 20 out of 2,908 reverse marble-dominated Okawayo formation varies laterally and vertically circulation and diamond core drillholes were discarded, providing a from approximately 50 m to 160 m. total usable drillhole length of 388.5 km. Three implicit models were computed based on lithology codes, gold assay values and bedding 4.1.2. 3D implicit assay model measurements S0/1. A total of 190,255 gold assay samples from reverse circulation and diamond core drillholes, sampled from 2 m core segments (99%) and 4.1.1. 3D implicit lithological model sub-metre segments (1%), were processed into 4 m composites to The six main geological formations that make up the host rocks of reduce the number of data points, resulting in 88,580 composites the Navachab gold deposit (Section 3.1) were used to generate the 3D (Fig. 5c). These show a maximum Au grade of 336 ppm, median of lithological model (Fig. 5a and b). The known age relationships of the 0.1 ppm, mean of 0.5 ppm, 90 percentile value of 1.1 ppm and 95 per- formations were also incorporated into the model. The youngest, over- centile value of 2.1 ppm (ppm equivalent to g/t). The histogram of lying sediments (including alluvium, calcrete and calcretised alluvial 4 m gold composites revealed a positively skewed distribution, which conglomerates) were grouped into a unit called cover sediments. means that the frequency of low Au values is much higher than the Meta-lamprophyre sheets and pegmatite dykes were not included in frequency of high Au values, and that there are some samples with the 3D implicit model due to their narrow geometry, which requires man- extreme values. This is very common for gold assay distributions, but ual adjustments and must be modelled separately. However, drillhole in- is unfavourable when modelling grade values with an interpolant that tercepts were used to investigate and visualise the spatial relationship uses a weighted sum of the data, which is the case for implicit modelling. between pegmatite dykes and gold mineralisation. This is because it will place too much emphasis on what are essentially ex- The lithological model is exclusively based on drillhole data, without ceptional values. A nonlinear transformation (log transformation) was any manual digitisation or adjustments that could reflect a subjective applied to the gold assay data to reduce this effect. interpretation. An isotropic interpolation (i.e., no anisotropic trend ap- To minimise any bias in the 3D assay model, an isotropic interpola- plied to the interpolation) was chosen to minimise any bias. A radius tion (i.e., no imposed trend) was used to calculate ore grade shells for of 60 m (adjusted to the drillhole spacing) was applied to generate a vi- cut-off grades of 0.5 ppm, 1.1 ppm and 2.1 ppm, based on the descrip- sualisation buffer along every drillhole trace (Fig. 5b). The model is only tive statistical values of the 4 m composites described above (Fig. 5d; visualised inside this buffer, which means that the distance to a sample for clarity only the 1.1 ppm grade shell is shown). or data point is 60 m at most to guarantee that subsequent visual inter- Subsequent analysis of the 1.1 ppm gold grade shell in combination pretations were constrained to an area or volume of confidence. The 3D with the 3D implicit lithological model revealed that 1) most gold implicit lithological model revealed that 1) bedding strikes NE and dips mineralisation is located in the Okawayo and Spes Bona formations;

Fig. 6. Cross-section A of the 3D implicit lithological model and 1.1 ppm gold grade shell of the 3D implicit assay model showing distinctive mineralisation trends within the Main Pit. The shoot-like, semi-massive sulphide lenses (Main shoot and 2nd shoot) are parallel to bedding and located close to the contact of the marble-dominated Okawayo formation with the biotite schist-dominated Spes Bona formation. Additionally, bedding-parallel mineralisation can be observed in the Spes Bona formation. Discordant mineralisation crosscuts the steeply NW dip- ping formations at high angles. Dispersed mineralisation appears to be related to young, NW striking pegmatite dykes, which have not been modelled due to their narrow geometry. S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 277

2) the highest grade and most extensive high-grade gold zones occur Dispersed high-grade mineralisation within the Spes Bona, Okawayo close to lithological contacts; and 3) three geometrically and spatially and Oberwasser formations was also identified, which is impossible to distinct high-grade mineralisation trends can be defined within the recognise in cross sections (and traditional 3D models that are based 1.1 ppm grade shell (Fig. 6). One prominent high-grade mineralisation on them) due to its unfavourable orientation. However, a trend within trend that can be linked to the semi-massive sulphide lenses is parallel the dispersed high-grade mineralisation becomes perceptible during to bedding, located near the contact between the Spes Bona and 3D visual analysis of the implicit model and appears to be spatially relat- Okawayo formations and has a width of about 30 m and height of ed to subvertical, young pegmatite dykes (Fig. 8). Additionally, gaps 190 m in cross-section. In long-section, it has a 1900 m long shoot-like within the 1.1 ppm gold grade shell are spatially associated with appearance that plunges shallowly at higher levels, steepening slightly drillhole intercepts of these NW trending, crosscutting pegmatite dykes. at depth (Fig. 8). Additionally, the 1.1 ppm grade shell outlines more bedding-parallel but less extensive gold mineralisation within the 4.1.3. 3D implicit structural trend model stratigraphically lower Spes Bona Formation. This trend resembles the Planar structural data (bedding S0,1)withinanareaofabout orientation of bedding-parallel quartz-sulphide veins that have been 11 km × 17 km around the Navachab deposit, covering the Usakos observed in the field (for more details, see Section 5). dome and Karibib dome have been digitised from a georeferenced geo- The second high-grade gold trend dips shallowly ENE, crosscuts the logical map compiled by Kisters (2005) and height adjusted using a dig- subvertical Okawayo and Spes Bona formations at a high angle and is lo- ital elevation model (DEM), based on SRTM 90 m digital elevation data cated in both the hanging and footwall of the aforementioned shoot-like (Fig. 5e and f). Leapfrog Mining's interpolation of planar structural data mineralisation trend (Fig. 6). Laterally, this shallowly ENE dipping trend works best for relatively uniformly sampled data points in geological completely crosscuts the Okawayo formation and continues 50 m into environments that are gently, openly or closely folded (i.e., most the Oberwasser formation. The flat-lying trend terminates to the SE domal structures), but is limited when dealing with tight to isoclinal within the Spes Bona formation and has a maximum horizontal extent folds, unless the data density is very high. In our case study, we comput- of about 100 m. In long section, this mineralisation type appears to be ed structural trend surfaces solely based on regional S0,1 measurements less voluminous and extensive compared to the shoot-like style of to visualise the 3D structure of the Karibib dome and Usakos dome. mineralisation. In the field, this type of mineralisation is observed to Detailed, regional geological maps by Kisters et al. (2004) and Creus be hosted in packages of sheeted, shallowly ENE dipping quartz- (2011) show no evidence for large scale faults cutting through the sulphide veins that crosscut subvertical bedding at high angles (for Karibib or Usakos dome, which could have an important influence on more details, see Section 5). the topology. However, major thrust faults have been identified at the

Fig. 7. Field observations from the Main Pit at the Navachab gold deposit. a) Packages of shallowly dipping, auriferous qtz-sulphide veins with sub-metre spacing crosscutting bedding at high angles. b) Openly folded auriferous qtz-sulphide vein crosscutting subvertical bedding of the more competent (silicified) biotite schist within the Spes Bona Formation. c) Tightly disharmonically folded auriferous qtz-sulphide vein which crosscuts bedding within the marbles of the Okawayo formation. d) Qtz-sulphide vein (green) related to the modelled bed- ding-parallel mineralisation trend within the Spes Bona formation. e) Subvertical pegmatite dykes (red) crosscutting flat lying, auriferous qtz-sulphide veins within the Spes Bona forma- tion, hence postdating veining. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 278 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

SE margin of the Karibib dome, within the Mon Repos thrust zone 5. Structural fieldwork and links to 3D implicit models (Fig. 2), which is located at the SE edge of the 3D structural trend model. Movement within this zone is generally sub-parallel to bedding Subsequent structural fieldwork was carried out in order to ground-

S0,1 along the SE limb of the Karibib dome, hence no significant changes truth the 3D implicit models and to link them to observed geological in orientation of structural measurements that might influence the structures. Spatially interpolated assay data (3D implicit assay model) model are expected. Moreover, the 3D structural trend model aims to were overlaid onto the mine site digital elevation model (DEM) to identi- only resolve large-scale (kilometre-scale) structures, meaning that fy key areas of high-grade mineralisation for targeted fieldwork at the smaller-scale features such as minor thrust faults or parasitic folds are Navachab gold deposit (Fig. 3). Mapping was restricted due to limit- not represented. Nevertheless, the general 3D structure of the chosen ed exposure and accessibility of high-grade semi-massive sulphide model area is appropriately reflected by the computed surfaces. An iso- mineralisation as a consequence of ongoing mining activities in the tropic interpolation of the regional bedding measurements S0,1 was cho- Main Pit and North Pit. However, we were able to link implicitly modelled sen as to not force the interpolation to follow a manually-defined trend. 1.1 ppm grade shells within the Spes Bona and Okawayo formation to The computed 3D trend surfaces were used subsequently to investigate ~24° ENE dipping packages of sheeted quartz-sulphide veins with sub- the spatial location of the Navachab gold deposit within the regional metre spacing located in the Main Pit. Approximately 50% of all quartz- structural context. The 3D implicit structural trend model outlines the sulphide veins mapped within the Spes Bona Formation have a sub- shape and dimensions of the Usakos dome, Karibib dome and the inter- metre spacing (Fig. 7a), and about 75% of these veins have a thickness vening Navachab syncline (Fig. 5f). In NW-SE cross-section, the Karibib of less than 40 mm. Fire assay information from 40 veins sampled within dome has an amplitude of ~5 km and a wavelength of ~10 km. On the the Spes Bona and Okawayo formation (sample size approximately 5 kg) NW limb of the Karibib dome, structural trend surfaces dip steeply at revealed an average gold grade of 13.9 ppm with a maximum value of around 70° where they intersect topography, but are (sub-)vertical at 97 ppm. Statistical analysis of these vein samples indicate that there is deeper levels. The SE limb dips about 50° SE at the surface. The inferred no correlation between vein thickness and gold grade. fold axes of the Karibib dome are shallowly double plunging to the NE An hypothetical 1 m × 1 m × 1 m sized block within the Spes Bona and SW, and steepen slightly at depth. The mineralised semi-massive formation (biotite schist), which is crosscut by a single mineralised sulphide ore shoots of the Navachab gold deposit resemble the orienta- vein (40 mm thick, average grade of 13.9 ppm, as detailed above) tion of the NE trending fold axis of the Karibib dome. roughly represents the structural setting of the economic cut-off grade

Fig. 8. Comparison of structural field measurements and the 1.1 ppm grade shell of the 3D implicit assay model. Orientations from structural data of bedding S0,1,F2 Karibib dome axial planar foliation S2, qtz-sulphide veins and S0,1/S2 intersection lineations can be linked to high grade mineralisation trends. Crosscutting pegmatite dykes postdate and remobilise gold mineralisation and lead to holes within the grade shell and are related to dispersed mineralisation, which has been identified in the 3D implicit assay model. S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 279 along the SE face of the Main Pit. Based on previous constraints and an assumed average density of 2900 kg/m3, this 1 m3 block would have an average grade of 0.6 ppm. Consequently, to form high-grade mineralisation, as illustrated by the implicitly modelled 1.1 ppm grade shells, the right combination of vein spacing, gold grade and vein thick- ness is required. However, as noted above, the lack of correlation be- tween vein thickness and Au grade implies that vein spacing must be the governing quantity. If the frequency of veins exceeds a defined threshold, high-grade mineralisation becomes visible within the grade shells of the 3D implicit assay model, which in turn outlines areas of sig- nificant economic interest. Auriferous veins within the Spes Bona and Okawayo formation crosscut subvertical compositional bedding and intrafolial isoclinal folds (F1). This suggests that vein development postdates D1 and oc- curred when bedding was already rotated to subvertical attitudes, pointing towards emplacement during the final lock-up stage of region- al scale folding (D2). Additionally, sheeted mineralised quartz-sulphide veins show small-scale upright folds with sub-metre amplitudes and wavelengths and shallowly NNE plunging fold axes. Quartz-sulphide veins within the Spes Bona formation are openly folded (Fig. 7b). On the other hand, veins within the Okawayo Formation are affected by tight, disharmonic folding (Fig. 7c). Both vein sets were probably emplaced contemporaneously, but rheological differences between their country rocks (relatively soft marble versus competent schist) led to variations in the intensity of folding. Due to the fact that the amplitudes and wavelengths of these folded veins are much smaller than the sample and composite spacing (4 m) of the drillholes, these small-scale structures are not directly reflected in the 3D implicit assay model. However, the mean principal vein orienta- tion calculated from field measurements and the attitude of the enveloping surface resemble the orientation of implicitly modelled shallow-dipping, high-grade mineralisation within the Spes Bona for- mation (Fig. 8, cross-section A). Because of the disharmonic tight folding Fig. 9. Four main stages of buckle-fold development (Cobbold, 1976; Ramsay, 1974). of the veins and the limited exposure on the NW face of the Main Pit, we (1) Initial layer-parallel shortening, (2) nucleation of buckling instability, (3) growth or amplification of a buckle-fold, (4) decay or lock up stage. were not able to collect representative field measurements of folded veins within the marble-dominated Okawayo formation. The horizontal extent of the mineralised quartz-sulphide veins is re- stricted and most probably controlled by changes in host rock lithology. such splayed meta-lamprophyre sheets have been folded. The fold

An abrupt termination is not only visible in the 3D implicit assay model, vergence is converse to F2 parasitic folds, but consistent with asymmet- but also in the field within the Spes Bona formation in Eastern Zone 1, ric folds that have formed due to the oblique orientation of the meta- where flat-lying veins terminate at the contact between biotite schist lamprophyre sheets in relation to the maximum NW-SE shortening and interbedded calcsilicate rocks. direction. We also mapped a set of sub-vertical, bedding-parallel quartz- Numerous, mainly NW-trending subvertical pegmatite dykes (dm to sulphide veins in the Spes Bona Formation within the Main Pit several metre thick) intruded into the metasedimentary rocks at (Fig. 7d). Because these veins are oriented parallel to the pit face and Navachab and crosscut auriferous quartz-sulphide veins, semi-massive are widely spaced, their recognition is limited in the field, but they sulphide ore shoots and the meta-lamprophyres sheets and dykes, have been identified in drill cores by Deltenre (2012). The geometry therefore postdating gold mineralisation and the mafic intrusions and orientation of the computed high-grade ore shells within the Spes (Fig. 7e). Additionally, it is noticeable that the abundance of pegmatite Bona formation also suggests the existence of bedding-parallel gold dykes is greater in the S and N of the Main Pit as well as in the North mineralisation (Figs. 6 and 8). Pit. Occasionally, remnants of flat-lying quartz-sulphide veins are pre-

The S2 axial planar foliation related to the formation of the Karibib served within the crosscutting pegmatite dykes in the field, confirming dome is best preserved in mica-rich rocks such as biotite schists of the that the dykes postdate vein hosted gold mineralisation (Fig. 7e).

Spes Bona and Oberwasser formations. S2 strikes NE and dips steeply Several young, brittle, shallow SE-dipping faults with minor dis- SE, reflecting the asymmetry of the Karibib dome. The lack of an addi- placement (b1 m) have been mapped on the NW face of the Main Pit. tional (axial planar) foliation trending NW and oriented at high angles We interpret these to be related to the post-folding relaxation of the to S2 suggests that the Karibib dome (and Usakos dome) did not form Damara orogen or an even younger deformational event. They do not by two different deformation stages, but one non-cylindrical folding seem to play a significant role in the mineralisation or the main struc- event. Intersection lineations between S0,1 and S2 are oriented parallel tural framework at Navachab. the fold axis of the Karibib dome and plunge shallowly NNE, and also parallel to the orientation of the semi-massive sulphide ore shoots, as 6. Discussion and interpretation indicated by the 1.1 ppm grade shell trend within the MC unit of the Okawayo formation (Fig. 8). One of the most complex structural attributes of many ore deposits The least competent, marble-dominated Okawayo formation has is folding, which is a prevalent feature at macro and micro scale and been intruded by (meta-) lamprophyre sheets, which are generally par- often described to influence mineralisation (Aerden, 1993; Blenkinsop allel to bedding, but also show a number of bedding-oblique splays. On and Doyle, 2014; Cox et al., 1991; Leader et al., 2010; Morey et al., the uppermost, non-accessible pit walls in the NE part of the Main Pit, 2005, 2007; Richard and Tosdal, 2001; Wilson et al., 2013). However, 280 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 associated regional scale folding is frequently overlooked as a major the importance of irregularities in localising the site of fold initiation structural control on mineralisation. Additionally, folding is a rather (Abbassi and Mancktelow, 1990; Biot et al., 1961; Cobbold, 1975; Dubey under-appreciated control on ore emplacement compared to faults and Cobbold, 1977). Such heterogeneities are manually introduced prior and shear zones, even though the formation of these structures is to experimentation to induce folding during the experiment. These two often a consequence of strain localisation during regional scale folding. examples emphasise the importance of initial heterogeneities, which may also play a role in the development and preservation of certain 6.1. Relationship between folding and mineralisation types of mineral deposits. For example, in the case of VMS deposits, the typical mound shape of such deposits, accompanying faults or the rela- In compressional tectonic settings such as fold and thrust belts, ac- tively soft massive sulphides themselves have the potential to act as het- tive folding (buckling) initiates when multilayered rock packages of erogeneities and favour and/or control the nucleation of folds. This might different viscosities are shortened parallel to layering (Fossen, 2010). explain why VMS deposits like Kidd Creek, Canada (Richardson, 1998)or The development of a buckle fold can be divided into four main stages Que River, Tasmania (Large et al., 1988) are associated with fold closures, (Fig. 9), namely (1) initial layer-parallel shortening, (2) nucleation or although further work is required to evaluate this hypothesis. Alternative- birth of buckling instability, (3) growth or amplification of the buckle- ly, sulphide-rich mineralisation can also be remobilised during deforma- fold and (4) decay or locking-up (Cobbold, 1976; Ramsay, 1974; tion and metamorphism. Many sulphide ores are less competent than Schmalholz, 2006). Mineralisation, on the other hand, is commonly their host rocks (Duuring et al., 2007). When folded, one would expect classified as pre-, syn- or post-deformational based on structural rela- stretched ore bodies along the limbs or migration and accumulation of tionships, petrographical analysis and/or geochronological evidence. sulphide ore in fold hinges due to mass transport towards low pressure This temporal classification of mineralisation can be linked to the four sites. main stages of folding, which are thought to influence the spatial distri- During the subsequent fold amplification and kinematic growth stage bution, trends and relationship of mineralisation. Depending on the of folding, the highest shear strain is reached along the inflection lines of folding stage under which mineralisation occurs, geometrically distinct folds. Maximum values of flexural shear/flexural flow are confined to high grade ore zones can form at preferred sites with respect to the fold. these areas, which are also associated with zones of low pressure and Furthermore, due to ongoing tectonic processes these ore zones might maximum dilation. Such linear zones of enhanced permeability have overprint and overlap each other and can be refolded as well. Their the potential to efficiently transport large volumes of fluid through mid- crosscutting relationships can be used to establish the relative timing crustal rocks, which is important for the genesis of an economic grade of the different generations of mineralisation. mineralisation. Moreover, these linear structures might not only act as The nucleation of folds is mainly determined by initial heterogene- favourable conduits, but also trigger the precipitation of gold because of ities. In numerical models, initial perturbations are introduced to nucle- changed physical conditions (e.g., lowered pressure), therefore acting as ate folds (Fernandez and Kaus, 2014; Frehner, 2014; Grasemann and physical traps. As described by Peters (1993b), the location and shape of Schmalholz, 2012; Mancktelow, 1999; Schmalholz, 2008; Schmid ore shoots are usually controlled by dilatant zones caused by changes in et al., 2008). Moreover, scaled analogue experiments have highlighted attitude, splays, lithologic contacts and intersections.

Fig. 10. Topography based on SRTM90 data (grey) intersected by the regional structural trend surface for S0,1 (structural data from Kisters (2005)). Colours refer to calculated shear strain γ which is directly related to dip (γ = tan(dip)) of originally horizontal layers, which are deformed into upright folds (Fossen, 2010). Note the distinctive geometrical differences between the Usakos dome in the W compared to the Karibib dome in the E. The Navachab deposit is located where shear strain reaches peak values. Regional structural trend surface generated with Leapfrog Mining 2.6; shear strain calculated and visualised with Move 2013.1. S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 281

The final lock-up stage of folding occurs when folding can no longer to NNE plunging axis of the Karibib dome. There is no evidence that the accommodate layer-parallel shortening. This leads to the formation of aforementioned bedding-parallel flexural shear was active during veins and faulting, which is a major feature of orogenic gold deposits. folding of the highly discordant vein set, suggesting that they were Vein-hosted mineralisation that crosscuts steepened bedding and emplaced later during the lock-up stage of the Karibib dome when mineralisation controlled by pre-existing foliations and late thrust faults host rocks have already reached steep attitudes and doming related form at this stage. Famous examples are the gold deposits of the Bendigo flexural slip had mostly ceased (Kisters, 2005). Zone, where gold is hosted by quartz veins associated with steeply- Ambiguous and mutually-crosscutting relationships between the dipping reverse faults (Leader et al., 2012). subvertical, bedding-parallel and the discordant, flat-lying vein systems are consistent with contemporaneous emplacement during the late 6.2. Structural history and gold mineralisation at the Navachab gold deposit stages of regional fold (dome) lock-up. The highly discordant vein set is oriented sub-parallel to the horizontal principal compressive stress

The 3D implicit structural trend model for bedding S0/1 shows that direction, σ1, expected for the NW-SE regional shortening event and the Navachab gold deposit is located near the inflection line of the associated folding. In contrast, bedding-parallel veins are oriented al- steep NW limb of the regional scale Karibib dome (Fig. 10). During most perpendicular to σ1 and were most likely generated under high NW-SE shortening and associated buckling of the Karibib dome, this fluid overpressure conditions, which are required for emplacement would have been an area where shear strain reached peak values. In along planes of strong mechanical anisotropy such as bedding (Sibson, contrast, shear strain induced by the same process would have reached 1996). Their unfavourable orientation with respect to the main stress minimum values in the hinge regions. To quantify and visualise shear field might also explain their limited abundance and lower thickness. strain distribution along the Karibib dome, we calculated shear strain However, we cannot rule out that some bedding-parallel quartz- based on the layer dip. For originally horizontal layers that are folded sulphide veins could have formed as shear veins during amplification into upright folds, shear strain γ is directly related to layer dip (γ = of the Karibib dome, contemporaneously with emplacement of the tan(dip)) (Fossen, 2010). This formula was used to determine shear auriferous semi-massive sulphide ore shoots. Bedding-parallel veins strain based on local dip values of the triangulated mesh from the 3D logged in drillhole N766 located within the Oberwasser Formation are implicit structural trend model (Fig. 10). The software package Move boudinaged. This is consistent with the bulk strain regime that folded 2013.1 (Midland Valley Exploration Ltd, 2013) was chosen to perform the discordant vein set and lamprophyre sheets during the late lock- advanced analyses on (triangulated) 3D meshes. It also allows colour- up stage of folding. Boudinage could also explain the lenticular shape coding of desired properties such as dip, curvature or in our case shear of massive-sulphide bodies that form the ore shoots. strain. The results are consistent with the location of the Navachab The third type of gold mineralisation is associated with pegmatite gold deposit along the NW inflection line of the Karibib dome where dikes that intruded parallel to a NW trending, subvertical joint set, shear strain reached the highest values of 4 to 5 during fold amplifica- which is interpreted here to have formed in an extensional fracture ori- tion (Fig. 10). Maximum layer-parallel displacements due to flexural entation during the lock-up stage of the Karibib dome. These pegmatite shear or flexural flow would have been confined to these areas, poten- dykes are responsible for local remobilisation and redistribution of gold, tially forming low-pressures zones such as shear veins or dilational which has not been described previously at the Navachab gold deposit jogs, thereby generating suitable conduits and traps for ore bearing (Fig. 8). This suggests that earlier deposited gold has been locally re- fluids. The semi-massive sulphide lenses that delineate the auriferous placed and remobilised by pegmatite intrusions and re-deposited at ore shoots were probably controlled by this folding-related strain (mainly) higher levels in the form of dispersed mineralisation. partitioning mechanism. This is in agreement with structural fieldwork In summary, results from 3D implicit modelling in combination with by Kisters (2005), who relates the emplacement of auriferous semi- structural fieldwork suggest that structures formed during the amplifi- massive sulphide lenses to dilational jogs based on their rhomb-like ge- cation and lock-up stage of the Karibib dome acted as conduits and traps ometry. Additionally, the orientation of calculated S0,1/S2 intersection for gold mineralisation and also controlled deposit-scale remobilisation. lineations (=F2 fold axes) resembles the trend of the central, shallowly These deformation related structural features best classify the Navachab NE plunging 1.1 ppm grade shell that is related to semi=massive gold deposit as an orogenic gold deposit. sulphide ore shoots, suggesting that they are structurally related to the formation of the Karibib dome (Fig. 8). No semi-massive sulphide 7. Conclusions mineralisation is reported within the hinge regions of the Karibib dome, where flexural slip-related shear strain is expected to have We have demonstrated that ore body geometries obtained from 3D attained minimum values, and where the potential to form associated, implicit models can be accurately linked to local and regional structural dilatant structures is low. Therefore, an epigenetic, syn-deformational patterns. First-order controls are most important for economic origin of the semi-massive sulphide ore shoots is more likely. mineralisation and are represented in our 3D implicit models for the The potential to form such dilatant structures depends on the Navachab gold deposit. Small-scale structures that cannot be resolved amount of shear strain, but also rheological properties of affected lithol- in these models but have been observed in the field were used to com- ogies. This could explain the second order control on the localisation of plement and constrain our structural interpretation. Our workflow fo- semi-massive sulphide ore shoots. As seen in the 3D implicit models, the cused on minimising the modelling bias by using an isotropic spatial main shoot (Main Pit) and 2nd shoot (North Pit) are both located close interpolation in order to utilise the resulting 3D implicit models for to the contact of the Okawayo Formation with the Spes Bon Formation structural interpretation purposes. The results from 3D implicit model- (Fig. 6). Generally, layer parallel shearing will not be uniformly distrib- ling in conjunction with structural fieldwork suggest that this workflow uted throughout a sequence (Price and Cosgrove, 1990) and strain will enhances the identification and (re-)assessment of structural controls be preferentially partitioned close to lithological contacts due to relative on mineralisation. Our structural interpretation of the Navachab deposit differences in physical properties. In the case of the Navachab gold depos- builds upon work by Kisters (2005). We demonstrate that the spatial lo- it, the rheological contrast between relatively soft marbles and competent cation of ore shoots at the Navachab gold deposit is systematic and con- biotite schists (party silicified) may have controlled the ‘stratigraphic’ po- trolled by different stages in the development of a regional scale domal sition of the ore shoots. structure (Karibib dome). Additionally, we described the role of cross- At the Navachab gold deposit, high grade mineralisation is also cutting pegmatite dykes in locally replacing and remobilising previous hosted in bedding-parallel and highly discordant quartz sulphide vein-hosted and semi-massive sulphide-hosted mineralisation. veins. The highly discordant veins dip shallowly to the ENE and are af- The Navachab gold deposit case study highlights the importance of fected by sub-metre scale, upright folding with fold axis that are parallel folding as a first order structural control on emplacement of ore bodies. 282 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

The 3D structural trend model demonstrates a geometrical dependence the 1st International Conference on Computer Graphics and Interactive Techniques in Australasia and South East Asia http://dx.doi.org/10.1145/604471.604495. of shear strain that potentially controls the localisation of associated Carr, J.C., Beatson, R.K., Evans, T.R., 2001. Reconstruction and representation of 3D objects mineralised structures (ore shoots). We propose that inflection lines with radial basis functions. SIGGRAPH, Computer Graphics, pp. 67–76. (where shear strain reaches peak values during fold amplification) of Caumon, G., Gray, G., Antoine, C., Titeux, M.-O., 2013. Three-dimensional implicit strati- graphic model building from remote sensing data on tetrahedral meshes: theory close, tight or even overturned folds are preferred sites of mineralisation and application to a regional model of La Popa Basin, NE Mexico. IEEE Trans. Geosci. and represent pipe-like traps and conduits for ascending fluids of vari- Remote Sens. 51, 1613–1621. http://dx.doi.org/10.1109/TGRS.2012.2207727. ous origins. Our results have implications for exploration activities in Chilès, J.P., Aug, C., Guillen, A., Lees, T., 2004. Modelling the geometry of geological units fi similar tectonic environments and should be considered in the planning and its uncertainty in 3D from structural data: the potential- eld method. Orebody Modelling and Strategic Mine Planning. The Australasian Institute of Mining and Met- stage of drillhole targets. Additionally, this work indicates that state-of- allurgy, Perth, WA, pp. 355–362. the-art modelling workflows can lead to a better structural insight in al- Cobbold, P., 1975. Fold propagation in single embedded layers. Tectonophysics 27, – ready developed and explored mine settings. 333 351. Cobbold, P.R., 1976. Fold shapes as functions of progressive strain. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 283, 129–138. http://dx.doi.org/10.1098/rsta.1976.0073. Acknowledgements Cowan, E.J., Beatson, R.K., Fright, W.R., Mclennan, T.J., Mitchell, T.J., 2002. Rapid geological modelling. Applied Structural Geology for Mineral Exploration and Mining, Interna- tional Symposium, 23–25 September 2002. Kalgoorlie, pp. 23–25. We thank AngloGold Ashanti Ltd for their generous support as well Cowan, E.J., Beatson, R.K., Ross, H.J., Fright, W.R., Mclennan, T.J., Evans, T.R., Carr, J.C., Lane, as AusIMM for funding S. A. V. (Bicentennial Gold Endowment). We R.G., Bright, D.V., Gillman, A.J., Oshust, P.A., T., M., 2003. Practical implicit geological — modelling. In: Dominy, S. (Ed.), Fifth International Mining Geology Conference Pro- are grateful to Jane Allen (AngloGold Ashanti Ltd Exploration Manager ceedings. AusIMM Publication Series No 8/2003, pp. 89–99. Brownfields, Continental Africa Region), Frik Badenhorst and especially Cowan, E.J., Spragg, K.J., Everitt, M.R., 2011. Wireframe-free geological modelling — an ox- to Navachab's chief geologist Graham Bell and his team for their abun- ymoron or a value proposition? Eighth International Mining Geology Conference. The Australasian Institute of Mining and Metallurgy, Queenstown, New Zealand dant and helpful support. S.A.V. and A.R.C. also acknowledge Monash Cox, S., Etheridge, M., 1987. The role of fluids in syntectonic mass transport, and the local- University for scholarship and research support. Stereonets were gener- ization of metamorphic vein-type ore deposits. Ore Geol. Rev. 2, 65–86. ated using Rod Holcombe's programme GEOrient. Academic licenses for Cox, S.F., Ruming, K., 2004. The St Ives mesothermal gold system, Western Australia—a – the Leapfrog Mining software package were kindly provided by ARANZ case of golden aftershocks? J. Struct. Geol. 26, 1109 1125. http://dx.doi.org/10. 1016/j.jsg.2003.11.025. Geo Pty. Ltd. In addition, we acknowledge Midland Valley Exploration Cox, S., Wall, V., Etheridge, M., 1991. Deformational and metamorphic processes in the Ltd. for providing an academic license for the Move 2013.1 geological formation of mesothermal vein-hosted gold deposits — examples from the Lachlan – modelling software package. We are grateful to two anonymous re- Fold Belt in central Victoria. Ore Geol. Rev. 6, 391 423. Cox, S.F., Sun, S.-S., Etheridge, M.A., Wall, V.J., Potter, T.F., 1995. Structural and geochemi- viewers for their comments and suggestions, which improved the qual- cal controls on the development of turbidite-hosted gold quartz vein deposits, Wattle ity of this paper. Gully Mine, Central Victoria, Australia. Econ. Geol. 90, 1722–1746. Creus, P.K., 2011. Geology and Structural Controls of Lode-gold Mineralization Around the Navachab Gold Mine in the Pan-African Damara Belt of Stellenbosch. University of References Stellenbosch. Deltenre, A., 2012. Understanding the Different Auriferous Quartz Vein Orientations in the Abbassi, M.R., Mancktelow, N.S., 1990. The effect of initial perturbation shape and symme- Footwall of Navachab Gold Mine, Namibia. Camborne School of Mines. try on fold development. J. Struct. Geol. 12, 273–282. http://dx.doi.org/10.1016/0191- Dubey, A., Cobbold, P., 1977. Noncylindrical flexural slip folds in nature and experiment. 8141(90)90011-M. Tectonophysics 38, 223–239. Aerden, D.G.A.M., 1993. Formation of massive sulfide lenses by replacement of folds; the Duuring, P., Bleeker, W., Beresford, S.W., 2007. Structural modification of the komatiite- Hercules Pb–Zn mine, Tasmania. Econ. Geol. 88, 377–396. http://dx.doi.org/10.2113/ associated Harmony nickel sulfide deposit, Leinster, Western Australia. Econ. Geol. gsecongeo.88.2.377. 102, 277–297. http://dx.doi.org/10.2113/gsecongeo.102.2.277. Ailleres, L., Lindsay, M., Jessell, M., DeKemp, E., 2010. The role of geological uncertainty in Dynamic Graphics Inc., 2014. earthVision. [WWW Document] URL, http://www.dgi.com/ developing combined geological and potential field inversions. ASEG 2010 — 21st In- earthvision/evmain.html (accessed 11.7.14). ternational Geophysical Conference and Exhibition. CSIRO publishing, Sydney, New Dziggel, A., Wulff, K., Kolb, J., Meyer, F.M., 2009. Processes of high-T fluid–rock interaction South Wales, Australia. during gold mineralization in carbonate-bearing metasediments: the Navachab gold ARANZ Geo, 2014. Leapfrog Mining. [WWW Document] URL, http://www.leapfrog3d. deposit, Namibia. Miner. Depos. 44, 665–687. http://dx.doi.org/10.1007/s00126-009- com/products/leapfrog-mining (accessed 10.31.14). 0231-9. Barnes, S., Sawyer, E., 1980. An alternative model for the Damara mobile belt: ocean crust Fernandez, N., Kaus, B.J.P., 2014. Influence of pre-existing salt diapirs on 3D folding pat- subduction and continental convergence. Precambrian Res. 13. terns. Tectonophysics http://dx.doi.org/10.1016/j.tecto.2014.10.021. Bauer, T.E., Skyttä, P., Hermansson, T., Allen, R.L., Weihed, P., 2014. Correlation between Fossen, H., 2010. Structural Geology. Cambridge University Press. distribution and shape of VMS deposits and regional deformation patterns, Skellefte Frank, T., 2006. Advanced Visualization and Modeling of Tetrahedral Meshes. Th. Doct. district, northern Sweden. Miner. Depos. 49, 555–573. http://dx.doi.org/10.1007/ Inst. Natl. Polytech. Lorraine. Technische Universitaet Bergakademie Freiberg. s00126-013-0503-2. Frehner, M., 2014. 3D fold growth rates. Terra Nova 26, 417–424. http://dx.doi.org/10. Beatson, R.K., Cherrie, J.B., Mouat, C.T., 1999. Fast fitting of radial basis functions: methods 1111/ter.12116. based on predonditioned GMRES iteration. Adv. Comput. Math. 11, 253–270. Geovia, 2014. Surpac. [WWW Document] URL, http://www.geovia.com/products/Surpac Billings, S.D., Beatson, R.K., Newsam, G.N., 2002. Interpolation of geophysical data using (accessed 10.31.14). continuous global surfaces. Geophysics 67, 1810–1822. http://dx.doi.org/10.1190/1. Grasemann, B., Schmalholz, S.M., 2012. Lateral fold growth and fold linkage. Geology 40, 1527081. 1039–1042. http://dx.doi.org/10.1130/G33613.1. Biot, M.A., Ode, H., Roever, W.L., 1961. Theory of folding of stratified viscoelastic media Groves, D.I., Ridley, J.R., Bloem, E.M.J., Gebre-Mariam, M., Hagemann, S.G., Hronsky, and its implications in tectonics and orogenesis. Geol. Soc. Am. Bull. 72, 1595–1632. J.M.A.,Knight,J.T.,McNaughton,N.J.,Ojala,J.,Vielreicher,R.M.,McCuag,T.C., Blenkinsop, T.G., 2004. Orebody geometry in lode gold deposits from Zimbabwe: implica- Holyland, P.W., 1995. Lode-gold deposits of the Yilgarn block: products of Late tions for fluid flow, deformation and mineralization. J. Struct. Geol. 26, 1293–1301. Archaean crustal-scale overpressured hydrothermal systems. In: Coward, M.P., http://dx.doi.org/10.1016/j.jsg.2003.11.010. Ries, A.C. (Eds.), Early Precambrian Processes. Geological Society Special Publica- Blenkinsop, T.G., Doyle, M.G., 2014. Structural controls on gold mineralization on the mar- tion 95, pp. 155–172. gin of the Yilgarn craton, Albany–Fraser orogen: the Tropicana deposit, Western Groves, D.I., Goldfarb, R.J., Knox-Robinson, C.M., Ojala, J., Gardoll, S., Yun, G.Y., Holyland, P., Australia. J. Struct. Geol. 1–16. http://dx.doi.org/10.1016/j.jsg.2014.01.013. 2000. Late-kinematic timing of orogenic gold deposits and significance for computer- Blenkinsop, T.G., Kadzviti, S., 2006. Fluid flow in shear zones: insights from the geometry based exploration techniques with emphasis on the Yilgarn Block, Western Australia. and evolution of ore bodies at Renco gold mine, Zimbabwe. Geofluids 6, 334–345. Ore Geol. Rev. 17, 1–38. http://dx.doi.org/10.1111/j.1468-8123.2006.00154.x. Guillen, A., Calcagno, P., Courrioux, G., Joly, A., Ledru, P., 2008. Geological modelling from Breitkopf, J.H., Maiden, K.J., 1988. Tectonic setting of the Matchless Belt pyritic copper de- field data and geological knowledge: Part II. Modelling validation using gravity and posits, Namibia. Econ. Geol. 83, 710–723. http://dx.doi.org/10.2113/gsecongeo.83.4. magnetic data inversion. Phys. Earth Planet. Inter. 171, 158–169. 710. Hardy, R., 1971. Multiquadric equations of topography and other irregular surfaces. Calcagno, P., Courrioux, G., Guillen, A., Fitzgerald, D., Mcinerney, P., 2006. How 3D implicit J. Geophys. Res. 76, 1905–1915. geometric modelling helps to understand geology: the 3D GeoModeller methodolo- Helio Rescources, 2012. Damara Gold Project (DGP). [WWW Document] URL, gy. Society for Mathematical Geology XIth International Congress. http://www.helioresource.com/s/Namibia.asp?ReportID=95922 (accessed Calcagno, P., Chilès, J.P., Courrioux, G., Guillen, A., 2008. Geological modelling from field 9.25.13). data and geological knowledge: part I. Modelling method coupling 3D potential- Hill, E.J., Oliver, N.H.S., Cleverley, J.S., Nugus, M.J., Carswell, J., Clark, F., 2013. Charac- field interpolation and geological rules. Phys. Earth Planet. Inter. 171, 147–157. terisation and 3D modelling of a nuggety, vein-hosted gold ore body, Sunrise Carr, J.C., Beatson, R.K., McCallum, B.C., Fright, W.R., McLennan, T.J., Mitchell, T.J., 1997. Dam, Western Australia. J. Struct. Geol. 67, 222–234. http://dx.doi.org/10.1016/ Smooth surface reconstruction from noisy range data. GRAPHITE '03 Proceedings of j.jsg.2013.10.013. S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284 283

Hillier, M., de Kemp, E., Schetselaar, E., 2013. 3D form line construction by structural field Maptek, 2014. Vulcan. [WWW Document] URL, http://www.maptek.com/products/ interpolation (SFI) of geologic strike and dip observations. J. Struct. Geol. 51, vulcan/ (accessed 10.31.14). 167–179. http://dx.doi.org/10.1016/j.jsg.2013.01.012. Masberg, H.P., 2000. Garnet growth in medium pressure granulite-facies metapelites Hillier, M.J., Schetselaar, E.M., de Kemp, E.a., Perron, G., 2014. Three-dimensional model- from the central Damara orogen: igneous versus metamorphic history. Commun. ling of geological surfaces using generalized interpolation with radial basis functions. Geol. Surv. Namib. 12, 115–124. Math. Geosci. http://dx.doi.org/10.1007/s11004-014-9540-3. Masberg, H.P., Holler, E., Hoernes, S., 1992. Microfabrics indicating granulite-facies meta- Hodkiewicz, P.F., Weinberg, R.F., Gardoll, S.J., Groves, D.I., 2005. Complexity gradients in morphism in the low-pressure central Damara orogen, Namibia. Precambrian Res. 55, the Yilgarn Craton: fundamental controls on crustal-scale fluid flow and the forma- 243–257. tion of world-class orogenic-gold deposits. Aust. J. Earth Sci. 52, 831–841. http://dx. Maxelon, M., Renard, P., Courrioux, G., Brändli, M., Mancktelow, N., 2009. A workflow to doi.org/10.1080/08120090500304257. facilitate three-dimensional geometrical modelling of complex poly-deformed geo- Hoffmann, K.-H., Condon, D.J., Bowring, S.a., Crowley, J.L., 2004. U–Pb zircon date from the logical units. Comput. Geosci. 35, 644–658. http://dx.doi.org/10.1016/j.cageo.2008. Neoproterozoic Ghaub Formation, Namibia: constraints on Marinoan glaciation. Ge- 06.005. ology 32, 817. http://dx.doi.org/10.1130/G20519.1. Mcinerney, P., Guillen, A., Courrioux, G., Calcagno, P., Lees, T., 2005. Building 3D geological Intrepid Geophysics, 2014. GeoModeller. [WWW Document] URL, http://www.intrepid- models directly from the data? A new approach applied to Broken Hill, Australia. Dig- geophysics.com/ig/index.php?page=geomodeller (accessed 10.31.14). ital Mapping Techniques. Jacob, R.E., Kröner, A., Burger, A.J., 1978. Areal extent and first U–Pb age of the pre- McInerney, P., Goldberg, A., Calcagno, P., Courrioux, G., Guillen, R., Seikel, R., 2007. Im- Damaran Abbabis complex in the central Damara belt of South West Africa proved 3D geology modelling using an implicit function interpolator and forward (Namibia). Geol. Rundsch. 67, 706–718. modelling of potential field data. Proceedings of Exploration '07, pp. 1–4. Jacob, R.E., Moore, J.M., Armstrong, R.A., 2000. Zircon and titanite age determinations from Micklethwaite, S., Cox, S., 2006. Progressive fault triggering and fluid flow in aftershock igneous rocks in the Karibib District, Namibia: implications for Navachab vein-style domains: examples from mineralized Archaean fault systems. Earth Planet. Sci. Lett. gold mineralization. Commun. Geol. Surv. Namib. 12, 157–166. 250, 318–330. http://dx.doi.org/10.1016/j.epsl.2006.07.050. Jessell, M., 2001. Three-dimensional geological modelling of potential-field data. Comput. Micromine, 2014. Micromine. [WWW Document] URL, http://www.micromine.com/ Geosci. 27, 455–465. products-downloads/micromine (accessed 10.31.14). Jessell, M.W., Ailleres, L., de Kemp, E.a., 2010. Towards an integrated inversion of Midland Valley Exploration Ltd, 2013. Move. [WWW Document] URL, http://www.mve. geoscientific data: what price of geology? Tectonophysics 490, 294–306. com/software/move (accessed 10.25.13). Jessell, M., Aillères, L., Kemp, E. De, Lindsay, M., Wellmann, F., Hillier, M., Laurent, G., Miller, R.M., 1983. The Pan-African Damara orogen of South West Africa/Namibia. Evolu- Carmichael, T., Martin, R., 2014. Next generation three-dimensional geologic model- tion of the Damara Orogen of South West Africa/Namibia. The Geological Society of ing and inversion. SEG Special Publication 18, pp. 261–272. South Africa, pp. 431–515. Jung, S., Mezger, K., 2003. Petrology of basement-dominated terranes I. Regional meta- Miller, R.M., Frimmel, H.E., 2009. Syn- to post-orogenic magmatism. Neoproterozoic to Early morphic T–t path from U–Pb monazite and Sm–Nd garnet geochronology (Central Palaeozoic evolution of Southwestern Africa. In: Gaucher, C., Sial, A.N., Halverson, G.P., Damara orogen, Namibia). Chem. Geol. 198, 223–247. http://dx.doi.org/10.1016/ Frimmel, H.E. (Eds.), Syn- to Post-orogenic Magmatism. Neoproterozoic to Early S0009-2541(03)00037-8. Palaeozoic Evolution of Southwestern Africa. Elsevier, pp. 219–226 http://dx.doi.org/ Kisters, A.F.M., 2005. Controls of gold-quartz vein formation during regional folding in 10.1016/S0166-2635(09)01615-6. amphibolite-facies, marble-dominated metasediments of the Navachab Gold Mine Miller, R.M., Frimmel, E.F., Will, T.M., 2009. Geodynamic synthesis of the Damara orogen in the Pan-African Damara Belt, Namibia. S. Afr. J. Geol. 108, 365–380. http://dx.doi. Sensu Lato. Dev. Precambrian Geol. 16, 231–235. http://dx.doi.org/10.1016/S0166- org/10.2113/108.3.365. 2635(09)01617-X. Kisters, A.F., Jordaan, L.S., Neumaier, K., 2004. Thrust-related dome structures in the Miller, R.M., Frimmel, H.E., Halverson, G.P., 2010. Passive continental margin evolution. Karibib district and the origin of orthogonal fabric domains in the south Central Neoproterozoic to Early Palaezoic evolution of Southwestern Africa. In: Gaucher, C., Zone of the Pan-African Damara belt, Namibia. Precambrian Res. 133, 283–303. Sial, A.N., Halverson, G.P., Frimmel, H.E. (Eds.), Neoproterozoic-Cambrian Tectonics, http://dx.doi.org/10.1016/j.precamres.2004.05.001. Global Change and Evolution: A Focus on Southwestern Gondwana. Developments Kitt, S., 2008. Structural Controls of Auriferous Quartz Veins in the Karibib Area, Southern in Precambrian Geology. Elsevier, pp. 161–181 http://dx.doi.org/10.1016/S0166- Central Zone of the Pan-African Damara Belt, Namibia. Stellenbosch University. 2635(09)01612-0. Kröner, A., Stern, R., 2004. Africa: Pan-African orogeny. Encycl. Geol. 1, 1–12. Morey, A.A., Weinberg, R.F., Bierlein, F.P., 2005. Deformation history and multiple gold Lajaunie, C., Courrioux, G., Manuel, L., 1997. Foliation fields and 3D cartography in geolo- mineralisation events within the Bardoc Tectonic Zone, Eastern Goldfields, Western gy: principles of a method based on potential interpolation. Math. Geol. 29, 571–584. Australia. Miner. Depos. Res. 1–4. http://dx.doi.org/10.1007/BF02775087. Morey, A.A., Weinberg, R.F., Bierlein, F.P., 2007. The structural controls of gold Lane, R., Beckitt, G., Duffett, M., 2007. 3D geological mapping and potential field modelling mineralisation within the Bardoc Tectonic Zone, Eastern Goldfields Province, Western of West Arnhem Land, Northern Territory. ASEG Ext. Abstr. 1. http://dx.doi.org/10. Australia: implications for gold endowment in shear systems. Miner. Depos. 42, 1071/ASEG2007ab072 583–600. http://dx.doi.org/10.1007/s00126-007-0125-7. Large, R.R., McGoldrick, P.J., Berry, R.F., Young, C.H., 1988. A tightly folded, gold-rich, mas- Mortimer, S., 2010. Implicit modeling — a new era in geological evaluation? In: Castro, R., sive sulfide deposit; Que River Mine, Tasmania. Econ. Geol. 83, 681–693. http://dx. Emery, X., Kuyvenhoven, R. (Eds.), Proceedings of the 4th International Conference doi.org/10.2113/gsecongeo.83.4.681. on Mining Innovation. Santiago, Chile Leader, L.D., Robinson, J.a., Wilson, C.J.L., 2010. Role of faults and folding in controlling Mueller, A.G., 2014. Structure, alteration, and geochemistry of the Charlotte quartz vein gold mineralisation at Fosterville, Victoria. Aust. J. Earth Sci. 57, 259–277. http://dx. stockwork, Mt Charlotte gold mine, Kalgoorlie, Australia: time constraints, down- doi.org/10.1080/08120090903552726. plunge zonation, and fluid source. Miner. Depos. http://dx.doi.org/10.1007/s00126- Leader, L.D., Wilson, C.J.L., Robinson, J.A., 2011. Structural constraints and numerical sim- 014-0527-2. ulation of strain localization in the Bendigo goldfield, Victoria, Australia. Econ. Geol. Nex, P.A.M., Kinnaird, J.A., Oliver, J.H.O., 2001. Petrology, geochemistry and uranium 108, 279–307. mineralisation of post-collisional magmatism around Goanikontes, southern Central Leader, L.D., Robinson, J.a., Wilson, C.J.L., 2012. Numerical modelling of fluid infiltration Zone, Damaran orogen, Namibia. J. Afr. Earth Sci. 33, 481–502. constrained by fault and bedding relationships in the Fosterville goldfield, Victoria, Nörtemann, M.F.-J., Mücke, A., Weber, K., Meinert, L.D., 2000. Mineralogy of the Navachab Australia. Ore Geol. Rev. http://dx.doi.org/10.1016/j.oregeorev.2012.05.005. skarn deposit, Namibia: an unusual Au-bearing skarn in high-grade metamorphic Ledez, D., 2003. Modélisation d'objets naturels par formulation implicites. Institut Nation- rocks. Commun. Geol. Surv. Namib. 12, 149–156. al Polytechnique de Lorraine. Paradigm, 2014. Gocad. [WWW Document] URL, http://www.pdgm.com/products/ Lindsay, M.D., Jessell, M., De Kemp, E., Betts, P., Ailleres, L., 2010. Integrating geological un- gocad/ (accessed 10.31.14). certainty into combined geological and potential field inversions. European Peters, S.G., 1993a. Nomenclature, concepts and classification of oreshoots in vein de- Geomodeller User Meeting. BRGM, Orleans (October). posits. Ore Geol. Rev. 8, 3–22. http://dx.doi.org/10.1016/0169-1368(93)90025-T. Lindsay, M.D., Aillères, L., Jessell, M.W., de Kemp, E.a., Betts, P.G., 2012. Locating and quan- Peters, S.G., 1993b. Formation of oreshoots in mesothermal gold-quartz vein deposits: ex- tifying geological uncertainty in three-dimensional models: analysis of the Gippsland amples from Queensland, Australia. Ore Geol. Rev. 8, 277–301. http://dx.doi.org/10. Basin, southeastern Australia. Tectonophysics 546–547, 10–27. http://dx.doi.org/10. 1016/0169-1368(93)90020-Y. 1016/j.tecto.2012.04.007. Phillips, G.N., Powell, R., 2010. Formation of gold deposits: a metamorphic devolatilization Lindsay, M.D., Jessell, M.W., Ailleres, L., Perrouty, S., de Kemp, E., Betts, P.G., 2013a. model. J. Metamorph. Geol. 28, 689–718. http://dx.doi.org/10.1111/j.1525-1314. Geodiversity: exploration of 3D geological model space. Tectonophysics 594, 27–37. 2010.00887.x. http://dx.doi.org/10.1016/j.tecto.2013.03.013. Pirajno, F., 2008. Hydrothermal Precesses and Mineral Systems. Springer Science. Lindsay, M.D., Perrouty, S., Jessell, M.W., Ailleres, L., 2013b. Making the link between Pirajno, F., Jacob, R.E., 1991. Gold mineralisation in the intracontinental branch of the geological and geophysical uncertainty: geodiversity in the Ashanti Greenstone Damara orogen, Namibia: a preliminary survey. J. Afr. Earth Sci. 13, 305–311. Belt. Geophys. J. Int. 195, 903–922. http://dx.doi.org/10.1093/gji/ggt311. Price, N.J., Cosgrove, J.W., 1990. Analysis of Geological Structures. Lombard, A.P.J., Fletcher, B.A., MacKinnon, H.F., 2013. Otjikoto gold deposit: discovery and Puhan, D., 1983. Temperature and pressure of metamorphism in the central Damara exploration of Namibia's second major gold mine. Conference Proceedings to orogen. In: Miller, R.M. (Ed.), Evolution of the Damara orogen of South West Africa/ NewGenGold 2013, Nov 26–27, 2013. Paydirt Media Pty Ltd., Perth. Namibia. Special Publication of the Geological Society of South Africa, pp. 219–224. Longridge, L., 2012. Tectonothermal Evolution of the Southwestern Central Zone, Damara Ramsay, J., 1974. Development of chevron folds. Geol. Soc. Am. Bull. http://dx.doi.org/10. Belt, Namibia. University of the Witwatersrand, Johannesburg. 1130/0016-7606(1974)85b1741. Lorensen, W., Cline, H., 1987. Marching cubes: a high resolution 3D surface construction Richard, J.P., Tosdal, R.M., 2001. Structural controls on ore genesis. In: Richards, J.P., Tosdal, algorithm. ACM Siggraph Comput. Graph. 21, pp. 163–169 R.M. (Eds.), Reviews in Economic Geology. Society of Economic Geologists, p. 181. Mancktelow, N., 1999. Finite-element modelling of single-layer folding in elasto- Richardson, D.M., 1998. An alteration study of the Archean Kidd creek volcanogenic mas- viscous materials: the effect of initial perturbation geometry. J. Struct. Geol. 21, sive sulphide deposit, Abitibi greenstone belt, Canada. Implications For Genetic 161–177. Models And Exploration Criteria. Laurentian University, Sudbury, Ontario, Canada. 284 S.A. Vollgger et al. / Ore Geology Reviews 69 (2015) 268–284

Ridley, J., 1993. The relations between mean rock stress and fluid flow in the crust: with deformation and remobilization. Econ. Geol. 100, 1441–1455. http://dx.doi.org/10. reference to vein-and lode-style gold deposits. Ore Geol. Rev. 8, 23–37. 2113/gsecongeo.100.7.1441. Ridley, J., Mengler, F., 2000. Lithological and structural controls on the form and setting of Tack, L., Bowden, P., 1999. Post-collisional granite magmatism in the central Damaran vein stockwork orebodies at the Mount Charlotte gold deposit, Kalgoorlie. Econ. Geol. (Pan-African) Orogenic Belt, western Namibia. J. Afr. Earth Sci. 28, 653–674. http:// 95, 85–98. dx.doi.org/10.1016/S0899-5362(99)00037-8. Robb, L., 2004. Introduction to Ore-forming Processes. Blackwell Publishing. Thompson, J.F.H., Sillitoe, R.H., Baker, T., Lang, J.R., Mortensen, J.K., 1999. Intrusion-related Savchenko, V.V., Pasko, A.A., Okunev, O.G., Kunii, T.L., 1995. Function representation of gold deposits associated with tungsten-tin provinces. Miner. Depos. 34, 323–334. solids reconstructed from scattered surface points and contours. Comput. Graphics Toé, W., Vanderhaeghe, O., André-mayer, A., Feybesse, J., Milési, J., 2013. From migmatites Forum 14, 181–188. http://dx.doi.org/10.1111/1467-8659.1440181. to granites in the Pan-African Damara orogenic belt, Namibia. J. Afr. Earth Sci. 85, Schlumberger, 2014. Petrel E&P software platform. [WWW Document] URL, www.slb. .62–74 http://dx.doi.org/10.1016/j.jafrearsci.2013.04.009. com/petrel (accessed 10.31.14). Tomkins, A.G., 2010. Windows of metamorphic sulfur liberation in the crust: implications Schmalholz, S.M., 2006. Scaled amplification equation: a key to the folding history of for gold deposit genesis. Geochim. Cosmochim. Acta 74, 3246–3259. http://dx.doi. buckled viscous single-layers. Tectonophysics 419, 41–53. http://dx.doi.org/10. org/10.1016/j.gca.2010.03.003. 1016/j.tecto.2006.03.008. Turk, G., O'Brien, J.F., 2002. Modelling with implicit surfaces that interpolate. ACM Trans. Schmalholz, S., 2008. 3D numerical modeling of forward folding and reverse unfolding of Graph. 21. a viscous single-layer: implications for the formation of folds and fold patterns. Vollgger, S., Kassl, K.H., 2010. Optimale Nutzung von Explorationsdaten für die Tectonophysics 446, 31–41. http://dx.doi.org/10.1016/j.tecto.2007.09.005. Abbauplanung. Berg-Huettenmaenn. Monatsh. 155, 357–364. http://dx.doi.org/10. Schmid, D.W., Dabrowski, M., Krotkiewski, M., 2008. Evolution of large amplitude 3D fold 1007/s00501-010-0586-3. patterns: a FEM study. Phys. Earth Planet. Inter. 171, 400–408. http://dx.doi.org/10. Wellmann, J.F., Horowitz, F.G., Schill, E., Regenauer-Lieb, K., 2010. Towards incorporating 1016/j.pepi.2008.08.007. uncertainty of structural data in 3D geological inversion. Tectonophysics 490, Sibson, R., 1996. Structural permeability of fluid-driven fault-fracture meshes. J. Struct. 141–151. http://dx.doi.org/10.1016/j.tecto.2010.04.022. Geol. 18. Willman, C.E., 2006. Regional structural controls of gold mineralisation, Bendigo and Sibson, R., Stone, G., 1991. Computation of thin-plate splines. SIAM J. Sci. Stat. Comput. 12, Castlemaine goldfields, Central Victoria, Australia. Miner. Depos. 42, 449–463. 1304–1313. http://dx.doi.org/10.1137/0912070. http://dx.doi.org/10.1007/s00126-006-0072-8. Stachowiak, H., 1973. Allgemeine Modelltheorie (General Model Theory). Springer. Willman, C.E., Korsch, R.J., Moore, D.H., Cayley, R.A., Lisitsin, V.A., Rawling, T.J., Morand, Steven, N.M., 1993. A study of epigenetic mineralization in the Central Zone of the V.J., O'Shea, P.J., 2010. Crustal-scale fluid pathways and source rocks in the Damara orogen, Namibia, with special reference to gold, tungsten, tin and rare Victorian gold province, Australia: insights from deep seismic reflection profiles. earth elements. Mem. Geol. Surv. Namib. 16, 166. Econ. Geol. 895–915. Steven, N.M., Badenhorst, F., 2002. Mesothermal gold deposits of the Damara orogen. Ex- Wilson, C.J.L., Schaubs, P.M., Leader, L.D., 2013. Mineral precipitation in the Quartz Reefs of cursion Guidebook for the 11th Quadrennial IAGOD Symposium and Geocongress. the Bendigo gold deposit, Victoria, Australia. Econ. Geol. 108, 259–278. Geological Survey of Namibia, Windhoek, p. 225. Wulff, K., 2008. Petrography, Geochemistry and Stable Isotope Characteristics of the Steven, N., Wulff, K., Bell, G., Hill, J., Sales, M., 2011. The increasing importance of Navachab Gold Deposit, Namibia. Rheinisch-Westfälische Technische Hochschule, Neoproterozoic (Pan-African) gold deposits in Africa and South America. 11th Bienni- Aachen. al SGA Conference — Let's Talk Ore Deposits. Ediciones Univ Catolica Norte, Antofa- Wulff, K., Dziggel, A., Kolb, J., Vennemann, T., Öttcher, M.E., Meyer, F.M., 2010. Origin of gasta, Chile. mineralizing fluids of the sediment-hosted Navachab gold mine, Namibia: constraints Stone, W.E., Beresford, S.W., Archibald, N.J., 2005. Structural setting and shape analysis of from stable (O, H, C, S) isotopes. Econ. Geol. 105, 285–302. nickel sulfide shoots at the Kambalda dome, Western Australia: implications for