Introduction

Gold-rich volcanogenic massive sulfide (VMS) deposits are important exploration targets as their gold content contribute significantly to their total value and their polymetallic nature makes them less vulnerable to metal price fluctuations. Some deposits are mined primarily for their gold content (e.g., Poulsen and Hannington 1996; Dubé et al. 2007a). However, there is little consensus in the literature about what constitutes a “gold-rich” VMS, in terms of gold grade or total contained gold. More rigorous criteria are needed to identify such deposits so that research can be focused on examples that best represent the geological attributes associated with gold enrichment. This could help better define key exploration criteria and target areas.

The main geological characteristics, mechanisms of gold concentration, and genetic models for gold-rich VMS deposits have been reviewed for both ancient and modern settings (e.g., Hannington et al. 1986; Huston and Large 1989; Large et al. 1989; Hannington and Scott 1989a, b; Large 1992a, b; Poulsen and Hannington 1996; Sillitoe et al. 1996; Hannington et al. 1999; Huston 2000; Dubé et al. 2007a). It is clear from these and other reviews (e.g., Franklin et al. 1981, 2005; Franklin 1993; Barrie and Hannington 1999; Galley et al. 2007a) that VMS deposits contain highly variable amounts of gold, both in terms of average grade and tonnage, with many VMS deposits being significant gold producers, especially in the last few decades. For example, approximately 13% of the gold mined in Canada up to 2003 came from VMS deposits (Lydon 2007) and gold represented about a fifth of the total value of VMS reserves and resources in Canada from 1994 to 2005 (Lydon 2007). Similarly, approximately 80% of the gold produced in Sweden from 1999 to 2006 came from VMS (statistics compiled from the Metals Economics Group, the Bureau de Recherches Géologiques et Minières, and the Geological Survey of Sweden), whereas 15% of the gold produced in Tasmania for the same period came from VMS (ABARE, Australian Minerals Statistics). Moreover, the relative value of gold in a median grade VMS deposit (i.e., ~0.9 g/t Au, 1.3% Cu and 3.3% Zn) in 1950 was about 10% of the total value of the deposit, whereas it was about 23% in 2008, illustrating the increasing economic importance of the gold content of a deposit. For example, the gold at Horne, Quebec (54 Mt grading 6.1 g/t Au and 2.2% Cu) was worth about 14% of the total value of the deposit based on 1950 gold and copper prices, whereas it would represent about 57% of the total value of the deposit at average prices for 2008.

Various classification schemes for VMS deposits based on their compositions have been used to interpret geodynamic settings and other aspects of ore genesis (e.g., Hutchinson 1973; Solomon 1976; Franklin et al. 1981). Similar attempts have been made to interpret the genetic aspects specific to “gold-rich” VMS based on Cu–Au versus Zn–Au associations (e.g., Huston and Large 1989; Huston 2000) and the style of mineralization or the ore mineralogy (e.g., Poulsen and Hannington 1996; Sillitoe et al. 1996; Hannington et al. 1999; Dubé et al. 2007a). However, it remains unclear how much gold constitutes a truly anomalous enrichment that must reflect special ore-forming conditions.

When comparing the metal endowment of deposits, it is necessary to consider both the grade and the size of the deposit, as well as the metal ratios. In previous studies, gold-rich VMS have been identified as having average gold grades expressed in ppm or g/t that exceed the combined content of base metals expressed in wt.% (Au g/t or ppm > Cu + Zn + Pb wt.%; Poulsen and Hannington 1996; Poulsen et al. 2000; Dubé et al. 2007a). However, this approach misclassifies some “gold-rich” polymetallic deposits that are accompanied by high grades of base metals. Conversely, some VMS deposits with a relatively low gold grade nevertheless produced significant amounts of gold due to their large size. Deposits in which gold can be mined economically on its own without any contribution from base metals could be, although not objectively, considered gold-rich. However, a more rigorous analysis of gold grades and tonnages is presented here. This analysis considers the frequency distribution of gold grades in 513 deposits worldwide, as well as the absolute gold contents of the deposits and the gold-to-base metals ratio. This analysis has identified a group of deposits with truly anomalous gold contents which highlight a number of geological attributes that should be targeted in exploration for gold-rich VMS.

Gold in VMS deposits

The data used in this study are from a global compilation of VMS deposits in Franklin et al. (2005). Some modifications were made to the dataset; in particular information on deposits with a very high average gold grade and deposits with a large total gold tonnage was updated. Deposits for which there are no known gold values and those that are characterized by a very high gold grade resulting from supergene enrichment were excluded, resulting in a dataset comprising 513 deposits.

Singer (1995) and Franklin et al. (2005) discuss a number of the limitations of such deposit grade and tonnage compilations. For example, the compiled numbers include production data as well as estimates of reserves and resources. Production numbers differ from reserves and resources, as metal recovery, particularly for gold, is never complete. A mining operation may be optimized for base metal production at the expense of precious metals recovery resulting in a bias towards lower reported average gold grades for deposits that have been or are presently producing base metals as their main commodities. This is a common situation in VMS deposits as the recovery rates for gold, copper, and zinc are 67%, 86%, and 83%, respectively (based on Canadian deposits mined up to 2005: Lydon 2007). The opposite situation also exists, with some VMS deposits mined and optimized for gold production. For example, gold, copper, and zinc recovery rates are approximately 76%, 93%, and 86% at the Flin Flon concentrator (HudBay Minerals Inc. Annual Report 2008), whereas they are approximately 91%, 86%, and 87% at the LaRonde concentrator (Agnico-Eagle Mines Ltd. Annual Report 2008). Thus, the head grades, which are the best measure of metal contents in the deposits, can differ significantly from reported grades based on past production. Comparing past or current producers with deposits for which only resources or reserves are known or estimated also may affect the grade distribution, without significantly affecting overall trends, as noted in previous studies of base–metal ratios in VMS deposits (Sangster 1977, 1980; Franklin et al. 1981). The mining method also can have a major impact on reported grades and tonnages, as an open pit operation will typically mine large tonnages at lower grades, illustrating the importance of considering the total gold content and gold-to-base metal ratios when comparing deposits. Moreover, gold may not be recovered in some deposits, despite large total contained amounts due to the refractory nature of the ore (e.g., Brunswick No. 12 in the Bathurst Mining Camp; McClenaghan et al. 2009). Many deposits that are not considered gold-rich are however characterized by relatively small but richer lenses or zones that do not affect significantly the average grade of the deposit (e.g., Rambler-Ming in Newfoundland, Canada; Santaguida and Hannington 1996; Hyde 2008; Iron Blow lens, Mount Lyell, Tasmania: Corbett 2001).

Many VMS deposits are characterized by disseminated and/or semi-massive sulfide zones that can be more economically important than massive sulfide zones. This is the case for a number of deposits with minor massive sulfide zones but high gold grades in stockwork zones (e.g., Iron Dyke, Oregon: Bussey and LeAnderson 1994; Juhas et al. 1980). In other deposits, the mined gold does not occur in the massive sulfide lenses but around it as stockworks, disseminations, or in barite lenses and exhalites (e.g., Kali Kuning and Lerokis on Wetar Island, Indonesia: Sewell and Wheatley 1994b; Sillitoe et al. 1996; Hannington et al. 1999; Huston 2000; Scotney et al. 2005). In some cases, gold has been remobilized into veins or veinlets peripheral to or within the VMS deposit during deformation and metamorphism (Mobrun, Quebec: Larocque et al. 1993; Bousquet 2, Quebec: Tourigny et al. 1993; Eastern Australian VMS deposits: Huston et al. 1992; Boliden, Sweden: Wagner et al. 2007; Myra Falls, British Columbia: Sinclair et al. 2000a, b) or late-stage hydrothermal remobilization (e.g., Chisel Lake deposit, Manitoba: Galley et al. 1993; Ansil deposit, Quebec: Galley et al. 1995). This commonly results in significant coarsening of gold, upgrading of the deposits, and improvements in gold recovery.

In many cases, a deposit’s genesis and the origin of the gold enrichments remain controversial (e.g., Mount Morgan, Queensland: Cornelius 1969; Taube 1986, 1990; Arnold and Sillitoe 1989; Ulrich et al. 2002). For example, some deposits are thought to have formed in (shallow) submarine volcanic-dominated environments and have distinctive epithermal characteristics (e.g., Sillitoe et al. 1996; Hannington and Herzig 2000), suggesting a hybrid VMS-epithermal classification (e.g., Mount Lyell, Tasmania: Large 2000; Corbett 2001; Huston and Kamprad 2001; Large et al. 2001; Eskay Creek, British Columbia: Roth et al. 1999; Johnson River, Alaska: Steefel 1987; Iron Dyke, Oregon: Bussey and LeAnderson 1994), or even as intermediates between VMS and porphyry (e.g., Balta Tau and Baimak-type deposits of the South Urals; Prokin and Buslaev 1999).

The average gold grades of VMS deposits tend to cluster at low values, with ~78% of the data below 2 g/t (Fig. 1). As is typical of the grade distribution in many ore deposit types (e.g., Sangster 1977; Parker 1991; Singer 1995; Franklin et al. 2005), gold grades in VMS deposits worldwide show a highly skewed frequency distribution (Fig. 1) that is log-normal (i.e., with a large number of deposits having very low gold grades and a small number of deposits having very high gold grades; Fig. 2). The arithmetic mean gold grade for VMS deposits is 1.50 g/t (Fig. 2), whereas the median grade is much lower (0.86 g/t). The large difference between the mean and the median reflects the skewness of the data, and the arithmetic mean is more effective at characterizing normally distributed data. Other measures of central tendency, such as the geometric mean, are less sensitive to extreme values (e.g., Borradaile 2003) and may be more useful in characterizing the distribution of gold grades in VMS deposits. The geometric mean is very helpful when dealing with such skewed distributions. The geometric mean gold grade for VMS deposits is 0.76 g/t Au (Fig. 2), which better reflects the large number of low-grade deposits. The standard deviation on the geometric mean is +2.70 g/t. These calculated values are in agreement with the values estimated graphically on the cumulative probability frequency plot for average gold grades in VMS deposits shown in Fig. 2. However, the cumulative probability frequency diagram shows that the distribution is not homogeneous, which may be due to the presence of more than one population (see below).

Fig. 1
figure 1

a Total base metals content (Cu + Zn + Pb in wt.%) versus gold grade for VMS deposits worldwide. b Cu (wt.%) and c Zn (wt.%) versus gold grades for VMS deposits worldwide. These plots show that there are no direct correlations between the average gold grade of the VMS deposits and the average concentration in base metals. Data modified from Franklin et al. (2005)

Full size image
Fig. 2
figure 2

Frequency histogram of average gold grade in VMS deposits worldwide. The distribution is strongly skewed (log-normal after log transformation of the data; inset, left). A probability plot of the data (inset, right), indicates a non-homogeneous distribution that reflects multiple populations. Deposits with a reported average gold grade of 0 were omitted from the dataset to allow for calculation of the logarithm. Std. dev. standard deviation. Data modified from Franklin et al. (2005)

Full size image

Figure 1 shows that there is no relationship between the total base metals content or Zn or Cu grades of a deposit and its gold grade. However, many base metal-rich deposits have relatively low gold values, and vice versa, mainly reflecting mining economics (i.e., a gold-rich deposit may be economic despite low base metal grades). Similarly, the total gold content, which is also characterized by a highly skewed frequency distribution that is log-normal (Fig. 3), is not only a function of grade. The arithmetic mean gold tonnage for VMS deposits is 17.7 t of Au. However, the cumulative frequency distribution shown in Fig. 3 suggests that the median value (50% of cumulated data), which is a good estimate of the geometric mean (Lepeltier 1969), is ~4.7 t Au with a geometric standard deviation of +26.3 t.

Fig. 3
figure 3

Frequency histogram of total gold tonnage in VMS deposits worldwide. The distribution is strongly skewed (log-normal after log transformation of the data; inset, left). The data are also plotted on a probability diagram (inset, right), showing a relatively homogeneous distribution, except for the very low values. Std. dev. standard deviation. Data modified from Franklin et al. (2005)

Full size image

Eighteen VMS deposits in the dataset were among the top 10% of gold deposits worldwide in terms of contained gold in 1995 (Fig. 4a). Each of these deposits contained more than 100 t of gold and is among the world-class gold deposits according to the definition of Singer (1995) (i.e., >100 t Au). Five of these deposits, with more than 250 t of Au (see Appendix), are among the top 5% of gold deposits worldwide (Fig. 4a). Although there are many supergiant VMS deposits (Franklin et al. 2005; Galley et al. 2007a), none contains enough gold to be considered a giant gold deposit according to the definitions of Sillitoe (2000) and Singer (1995) (i.e., >600 t Au for a giant gold deposit and ≥1,000 t Au for a “supergiant” deposit). The cumulative frequency plot of the gold content of VMS deposits (Fig. 4b) shows that the top 10% of deposits contain more than 45 t of gold each and account for about 64% of the total amount of gold contained in VMS deposits worldwide. The top 5% of the deposits each contained more than 83 t of gold, accounting for about 47% of all the gold hosted in VMS deposits (Fig. 4b).

Fig. 4
figure 4

a Cumulative frequency plots of contained gold in gold-bearing deposits of all types, worldwide (modified from Singer 1995). Eighteen gold-bearing VMS deposits are among the top 10% of gold deposits worldwide in terms of total contained gold, and five are among the top 5%. b Plot of cumulative frequency of deposit size and total contained gold among VMS deposits worldwide for which gold grades are known. Data modified from Franklin et al. (2005). The top 5% and 10% of the deposits in terms of size contain 47% and 64% of the total gold in VMS deposits, respectively. Deposits with 31 t of Au or more represent less than 13% of the total number of deposits but contain about 70% of the VMS gold. Thus, a large proportion of the VMS-hosted gold worldwide is contained in just a few deposits. These deposits form a unique group that best illustrate the geological parameters that control the formation of gold-rich VMS

Full size image

When individual districts or mining camps are considered, the distribution of gold grades is generally log-normal, but one or more deposits almost always have anomalous grades that are distinct from the rest of the deposits in the camp (see also Hannington et al. 1999). In some cases (e.g., Abitibi Greenstone Belt and Flin Flon–Snow Lake Camps in Canada) the gold grade distribution is bimodal. This appears to indicate that a particular process or set of geological conditions responsible for gold enrichment in some deposits did not operate in others. Examples of gold grade distribution in a number of VMS-bearing belts and districts are shown in Fig. 5. Most belts or districts contain at least one discordant or distinctly gold-rich deposit and some districts contain a number of such deposits (e.g., Noranda and Doyon-Bousquet-LaRonde Mining Camps in the Blake River Group of the Abitibi Greenstone Belt, Quebec: Dubé et al. 2007a; Mercier-Langevin et al. 2007b; Mount Read Volcanics, Tasmania: Huston 2000). There are no clear relationships between the size and the grade of the deposits as shown on Fig. 6, except perhaps for the Abitibi Greenstone Belt and the Skellefte District where the richest deposits are among the largest VMS deposits of the district.

Fig. 5
figure 5

Frequency histograms of gold grades in VMS deposits for selected major camps and districts of various ages. a Archean Abitibi Greenstone Belt in Canada. b Paleoproterozoic Skellefte District in Sweden c Paleoproterozoic Flin Flon and Snow Lake Camps in Canada. d Cambrian Mount Read Volcanics in Tasmania, Australia. e Ordovician Bathurst Mining Camp in Canada. f Devonian South Urals in Kazakhstan and Russia. g Carboniferous Iberian Pyrite Belt in Spain and Portugal. h Miocene Kuroko Belt in Japan. The geometric mean plus one standard deviation from the geometric mean obtained from the global database is shown on each plot (arrow), as well as the value of the geometric mean plus one standard deviation from the geometric mean calculated for the individual camps (dashed line). The distribution of gold grades in individual belts, camps, or districts is generally log-normal, but one or more deposits almost always have anomalous grades that are distinct from the rest of the deposits in the belt, camp, or district. In some cases, the gold grade distribution is bimodal. The global geometric mean plus one standard deviation from the geometric mean is a useful discriminator of gold-rich deposits in most cases. However, the geometric mean plus one standard deviation in each belt, district, or camp varies. Std. dev. standard deviation

Full size image
Fig. 6
figure 6

Bubble plots showing the relative contribution of VMS deposits to the total gold tonnage and total ore tonnage of their respective camps or districts for selected major belts, camps, and districts of various ages. Bubble size is proportional to the average gold grade of the deposit: a Archean Abitibi Greestone Belt in Canada. b Paleoproterozoic Skellefte District in Sweden c Paleoproterozoic Flin Flon and Snow Lake Camps in Canada. d Cambrian Mount Read Volcanics in Tasmania, Australia. e Ordovician Bathurst Mining Camp in Canada. f Devonian South Urals in Kazakhstan and Russia. g Carboniferous Iberian Pyrite Belt in Spain and Portugal. h Miocene Kuroko Belt in Japan. The plots show that there are no clear relationships between the size and the grade of the deposits, except perhaps for the Abitibi Greenstone Belt and the Skellefte District where the richest deposits are among the largest VMS of the district

Full size image

The VMS deposits of the Abitibi Greenstone Belt (n = 56; Fig. 5), which are clustered in a number of districts of various sizes, have a mean gold grade of 1.43 g/t, with a geometric mean of 0.64 g/t. Of these deposits, 43 (~77%) have gold grades that are lower than the arithmetic mean indicating the major influence of a few deposits. In contrast, for the Flin Flon and Snow Lake Camps, 15% of the deposits have gold grades higher than 3 g/t (Figs. 5 and 6). The two most gold-rich deposits in the Bathurst Mining Camp, New Brunswick (Goodfellow and McCutcheon 2003; McClenaghan et al. 2003, 2004) also stand out from the other deposits in the camp (Figs. 5 and 6), although they would not be considered auriferous in comparison to deposits worldwide (e.g., Fig. 1). Nevertheless, these atypical deposits in terms of gold grade indicate an underlying process or condition of ore formation that is unique in the camp. Two of the largest (Horne and LaRonde Penna) and four of the richest (Westwood-Warrenmac, Bousquet 2-Dumagami, Quemont, Bousquet 1) gold-rich VMS deposits worldwide are located in the Blake River Group of the Southern Abitibi Greenstone Belt. Moreover, the VMS deposits of the Blake River Group represent approximately 48% of the total VMS tonnage of the Abitibi Greenstone Belt, but contain about 92% of the total VMS gold of the belt. Similarly, the VMS deposits of the Macuchi Belt in Ecuador are particularly rich in gold (Chiaradia and Fontboté 2001) and the Mount Read Volcanics in Tasmania are “lacking” the typical dominance of low-grade deposits (Figs. 5 and 6) with a median grade of 2.38 g/t Au. This illustrates the provinciality of some gold-rich VMS deposits (Hannington et al. 1999; Huston 2000; Dubé et al. 2007a) and strongly suggests a large-scale geological control on the enrichment of gold (see below).

Discriminating anomalous, auriferous, and gold-rich VMS deposits

Poulsen and Hannington (1996) defined gold-rich VMS as having an average gold content (expressed in ppm or in g/t) that exceeds the total base metals content (expressed in weight percent; Fig. 7). However, some ambiguity arises in this classification for deposits that have both high gold grades and high base metal grades (e.g., Lemoine, Quebec; Guha 1984; Guha et al. 1988; La Plata, Ecuador; Chiaradia et al. 2008; Iron King, Arizona; Creasey 1952) or very low gold and base metals contents (e.g., Mount Lyell, Tasmania; Corbett 2001; Seymour et al. 2007). The latter may be related to the style of mineralization and/or the mining method (e.g., Mount Lyell, which is dominated by disseminated sulfides and pyrite-rich ore, was mined by open pit: Corbett 2001). The problem of classifying very low grade deposits has been discussed by Poulsen et al. (2000) among others, who proposed that the classification be restricted to only those deposits that were or could be mined economically.

Fig. 7
figure 7

Bivariate plots of total base metals grade (Cu + Zn + Pb in wt.%) versus gold grade of VMS deposits worldwide. Deposits are considered anomalous in terms of gold content if they meet at least one of three criteria: (1) the gold grade is higher than 3.46 g/t, which corresponds to the geometric mean plus one standard deviation from the geometric mean, (2) the total gold tonnage is equal to or greater than 31 t Au (geometric mean plus one standard deviation from the geometric mean; see Fig. 8), (3) the gold grade in g/t equals or exceeds the total concentration of base metals in wt.% (Au/total b.m. ≥1). Johnson R. Johnson River. Geo. Mean = geometric mean, Geo. Std. Dev. standard deviation from the geometric mean

Full size image

Discordant or anomalous sample populations are characteristic of trace element abundance data (e.g., Limpert et al. 2001). In many cases, the discordance serves to identify an important feature of practical geological importance, especially if it is caused by natural variability (Beckman and Cook 1983). The standard deviation from the mean is the most common metric to identify discordance in a dataset (e.g., de Wijs 1951; Grubbs 1969) and has been used in this analysis. Because gold grades and gold tonnage of VMS deposits worldwide define positively skewed distributions (Figs. 2 and 3), the standard deviation has been calculated on the geometric mean (geometric standard deviation), which appears to be in closest agreement with the observed distribution of gold grades in most VMS camps or districts. On a global scale, discordant or anomalous deposits are those having gold grades higher than 3.46 g/t (geometric mean value plus one geometric standard deviation calculated from the database of 513 deposits). Forty-six VMS deposits have gold grades exceeding this value (Fig. 8 and Appendix). Examination of the gold grade distribution in different districts (Fig. 5) shows that the global geometric mean and geometric standard deviation are a useful discriminator of auriferous deposits in most cases. Deviations from a strictly log-normal distribution in the cumulative frequency plot (Fig. 2) are related to the superposition of different populations (e.g., log-normal grade distribution in some districts, bimodal in others).

Fig. 8
figure 8

Bivariate plot of gold grade versus tonnage for VMS deposits. Data are shown for 513 deposits worldwide (dataset modified from Franklin et al. 2005). The shaded regions show the geometric mean plus one standard deviation from the geometric mean of the gold grade and the total contained gold for all 513 deposits. Auriferous deposits are considered to have a gold grade greater than 3.46 g/t and/or gold grades in g/t equal or exceeding the total concentration of base metals in wt.%. VMS deposits with more than 31 t Au (n = 50) are considered anomalous. VMS deposits with more than 3.46 g/t Au and 31 t of Au (n = 11) are considered gold-rich. One hundred thirteen deposits meet at least one of these criteria, 29 deposits meet at least two criteria, and only nine deposits meet all three criteria. Canoe-L. Canoe Landing Lake, Iron D. Iron Dyke, Iron K. Iron King, Johnson R. Johnson River, Lomero Lomero-Poyatos, Petiknas N. Petiknas North, Geo. Mean geometric mean, Std. Dev. standard deviation

Full size image

The geometric mean plus one geometric standard deviation of the total contained gold (i.e., 31 t Au) also identifies deposits with anomalous gold endowment, irrespective of their gold grade (Fig. 8). Deposits with total contained gold of 31 t or more (n = 63: Appendix) are termed “anomalous VMS” and account for 69.8% of the total gold contained in VMS deposits worldwide (Fig. 4b). This indicates that approximately 70% of the VMS gold is hosted in about 13% of the VMS deposits. Eleven of these deposits have gold grades higher than 3.46 g/t (Fig. 8 and Table 1). These deposits, termed “gold-rich VMS,” account for 17.8% of the total gold contained in VMS deposits worldwide, even though they only represent 2.9% of the total VMS tonnage. A majority of these 11 deposits also have a gold-to-base metals ratio that is greater than 1 (Fig. 7, Table 1 and Appendix). The 46 deposits that contain more than 3.46 g/t represent about 9% of the deposits worldwide for which gold grades are known (database of 513 deposits) and account for about 20% of the VMS gold but only 3.5% of the global VMS tonnage. These statistics highlight the fact that a large proportion of the total gold hosted in VMS worldwide is found in a relatively small number of deposits. Overall, 113 deposits meet at least one of these criteria (see Appendix and Fig. 9). Twenty-nine deposits meet at least two criteria. Only nine deposits meet all three criteria (≥31 t of contained Au, an average gold grade higher than 3.46 g/t, and a gold-to-base metals ratio ≥1: Fig. 8, Table 1 and Appendix); six of these are in the Blake River Group of the Abitibi Greenstone Belt.

Table 1 Selected characteristics for some examples of gold-rich VMS and auriferous deposits of various ages
Full size table
Fig. 9
figure 9

Geographical distribution of ancient anomalous, auriferous, and gold-rich VMS deposits highlighted in this study and listed in Appendix. Generalized geology of the world from Chorlton (2007)

Full size image

Characteristics of anomalous, auriferous and gold-rich VMS deposits

Table 1 summarizes the main characteristics (grades, age, interpreted geodynamic setting, lithostratigraphic association, host rocks, mineralization, and alteration styles) of 31 deposits highlighted in this analysis, and Appendix lists all gold-rich, auriferous, and anomalous VMS deposits identified in our analysis. The distribution of these deposits is shown on Fig. 9. The data presented in Table 1 emphasize specific geological attributes that distinguish them from other less gold-rich deposits.

The sizes and grades vary significantly, but at the camp scale, the richest deposits are commonly either the largest or among the largest deposits of the district or the belt (e.g., Horne, LaRonde Penna, Quemont and Bousquet 2-Dumagami in the Abitibi Greenstone Belt; Boliden in the Skellefte District; Flin Flon in the Flin Flon and Snow Lake Camps; Caribou in the Bathurst Mining Camp; La Zarza in the Iberian Pyrite Belt; Greens Creek in the Juneau-Admiralty District; Fig. 6 and Table 1). It is noteworthy that six of the 15 largest VMS deposits in the Abitibi Greenstone Belt, accounting for 42% of the total tonnage of VMS in the Abitibi, are also the most gold-rich, accounting for 88% of the VMS gold. Moreover, Horne (H-G and Zone 5) and LaRonde Penna, two world-class gold-rich VMS deposits, account for 35% of the total VMS tonnage of the entire Abitibi and for 60% of the total VMS gold of the belt. In other districts, the most gold-rich deposits are among the smallest (e.g., Delbridge and Deldona in Noranda: Boldy 1968; Gibson and Galley 2007; Petiknas North and Holmtjarn in Skellefte; Lomero-Poyatos in the Iberian Pyrite Belt; Photo Lake in Flin Flon/Snow Lake; Balta Tau in the South Urals; Fig. 6 and Table 1). The gold-to-base metals ratios are highly variable among the gold-rich and auriferous VMS deposits (Table 1), including at the camp scale (e.g., Skellefte District in Sweden and Macuchi Camp in Ecuador), with some deposits having an elevated (>2) gold-to-base metals ratio (Boliden in Skellefte and Mercedes in Macuchi) located close to deposits with relatively low ratios (Petiknas North and Holmtjarn in Skellefte and La Plata in Macuchi). The richest deposits are not necessarily those having the highest gold-to-base metals ratio (e.g., South Urals: Balta Tau at 4.5 g/t Au with a ratio of 0.56 and Yubileinoe at 2.5 g/t Au with a ratio of 0.76; Table 1). Despite relatively high gold grades, the deposits of the Mount Read Volcanics are characterized by very low gold-to-base metals ratios due to very high base metal contents (Table 1).

As noted by Hannington et al. (1999), Huston (2000), and Dubé et al. (2007a), gold-rich VMS deposits can be found in belts and districts of all ages (e.g., Figs. 5 and 6, Table 1 and Appendix). However, in some districts, the gold-rich VMS appear to have formed at specific times. Although there are few camps where detailed and very precise geochronology are available to establish the position of the VMS deposits, stratigraphic reconstructions in some cases provide insight on the relative timing of hydrothermal events and magmatic evolution of the host volcanic complex. For example, the Horne and Quemont deposits are older than the Cu–Zn VMS deposits of the Noranda Central Camp (McNicoll et al. 2008) and are among the oldest deposits of the Blake River Group, whereas the gold-rich VMS deposits of the Doyon-Bousquet-LaRonde Camp 50 km east of Noranda (LaRonde Penna, Bousquet 2-Dumagami, Bousquet 1, and Westwood-Warrenmac) are among the youngest deposits of the Blake River Group (Lafrance et al. 2005; Mercier-Langevin et al. 2007c). The Photo Lake deposit in Snow Lake is possibly the youngest deposit of the camp, although tectonic disturbance obscures the stratigraphic relationships of this area (Galley et al. 2007b). The Caribou and Canoe Landing Lake deposits of the Bathurst Mining Camp are hosted by the Middle Ordovician California Lake Group, and are among the earliest formed VMS deposits in the region (Walker and McDonald 1995; Goodfellow 2003). The California Lake Group tends to host deposits that are, on average, slightly richer in gold than deposits of the coeval Tetagouche Group (McClenaghan et al. 2003, 2004) and may result from their formation on differing crustal blocks as reflected in base metals and isotopic signature (Goodfellow and McCutcheon 2003). The specific timing of formation of gold-rich deposits in certain VMS districts may be directly related to the geodynamic evolution of the arc–back-arc systems and the nature of the corresponding magmatism (e.g., Macuchi Arc in Ecuador; Chiaradia et al. 2008; Baimak-type VMS deposits of the South Urals; Prokin and Buslaev 1999; Herrington et al. 2005b).

Barrie and Hannington (1999), Franklin et al. (2005), and Galley et al. (2007a) noted that the average gold grades of VMS deposits are highest in mafic and bimodal-felsic volcanic sequences. The gold-rich and auriferous VMS deposits listed in Table 1 have a variety of lithostratigraphic associations; however, most are thought to have formed in arc-related rifts in their early stages, in oceanic or continental arc or back-arc settings, and at times of evolving arc magmatism. Transitional to calc-alkaline, intermediate to felsic rocks, including andesites, dacites, rhyodacites, and rhyolites appear to be common hosts (e.g., Table 1), similar to many sub-aerial epithermal deposits (Sillitoe and Hedenquist 2003; Simmons et al. 2005). Gold-rich VMS are associated with particularly voluminous felsic volcanism in many districts (e.g., Table 1). The Horne deposit is hosted in a thick succession of rhyolitic domes and volcaniclastics with only minor intermediate to mafic rocks (Barrett et al. 1991; Kerr and Gibson 1993; Monecke et al. 2008): this contrasts sharply with the dominantly andesitic sequences that host the other Cu–Zn VMS lenses of the Noranda Central Camp (Spence and de Rosen-Spence 1975; Kerr and Gibson 1993). The Horne rhyolites are tholeiitic to transitional and characterized by a slight depletion in rare earth elements that distinguishes them from rhyolites in the Noranda Central Camp hosting Cu–Zn deposits. This suggests a different magmatic evolution for the Horne sequence (Kerr and Gibson 1993). The gold-rich VMS deposits of the Doyon-Bousquet-LaRonde Camp are hosted by a sequence consisting of transitional to calc-alkaline dacite, rhyodacite, and rhyolite overlying Archean tholeiitic basalts (Mercier-Langevin et al. 2007c). The La Zarza and Lomero-Poyatos deposits are located in the northern part of the Iberian Pyrite Belt, which has a much higher proportion of calc-alkaline felsic volcanic rocks (dacites and lesser amounts of rhyolite) and less sedimentary rocks than the southern part of the belt (Leistel et al. 1998a, b; Thiéblemont et al. 1998; Tornos 2006). The Paleoproterozoic Skellefte District (Allen et al. 1996), the Cambrian Mount Read Volcanics (Corbett 1992; Crawford et al. 1992; Large et al. 2001), the Ordovician Bathurst Mining Camp (Lentz 1996; Yang and Scott 2003), the Devonian South Urals (Herrington et al. 2005b), and the Miocene Hokuroku District (Dudas et al. 1983), which contain numerous auriferous and/or anomalous VMS, are similarly characterized by voluminous transitional to calc-alkaline felsic volcanic rocks (Table 1).

The transitional to calc-alkaline felsic volcanic rocks of Precambrian age have been considered to be of a very limited prospectivity for VMS deposits in the literature (e.g., Lesher et al. 1986; Hart et al. 2004) based on the preferential association between VMS deposits and tholeiitic sequences in the Superior Province of Canada. However, the association between significant Archean gold-rich VMS deposits and HREE-depleted, transitional to calc-alkaline felsic volcanic rocks formed in high-pressure settings (e.g., incipient arc rifting or back-arc rifting over a relatively thick arc–back-arc lithosphere) demonstrates the potential of such sequences (Mercier-Langevin et al. 2007c). It also suggests that this particular geodynamic setting and the associated petrogenetic processes are related to the elevated gold content of the associated VMS deposits of Archean age and perhaps of younger ages as well. Even though not all gold-rich VMS deposits highlighted in this study are directly associated to transitional to calc-alkaline rocks, most appear to be associated with an abrupt change in magmatic affinity in response to changes from arc volcanism to arc rifting to back-arc volcanism (e.g., Eskay Creek, British Columbia; Hokuroku District, Japan). While a link to the formation of specific gold-rich deposits remains uncertain, these step changes in magmatic evolution and corresponding changes in host rock composition appear to have been favorable for the formation of the most auriferous and gold-rich deposits. A better assessment of the lithogeochemistry of the volcanic sequence hosting gold-rich VMS deposits could help highlight specific magmatic processes (and geodynamic settings) associated with the mineralizing systems.

In a number of cases, the auriferous deposits are located in distinctly different volcanic and/or structural settings from other deposits in the district. The Horne and Quemont gold-rich deposits that are separated in time and space from the Noranda Cu–Zn VMS are among the best examples; both deposits are located in the southern part of the Noranda Central Camp in fault-bounded structural blocks separated from the slightly younger Cu–Zn deposits. The gold-rich VMS deposits of the Doyon-Bousquet-LaRonde Camp are part of the Bousquet Formation, which is approximately coeval with the volcanic rocks that host the Cu–Zn VMS of the Noranda Central Camp (Lafrance et al. 2005; Mercier-Langevin et al. 2007d; McNicoll et al. 2008). The Bousquet Formation and its deposits are thought to have been formed in a volcanic complex at the periphery of the Blake River Group (Lafrance et al. 2003), possibly in an area characterized by thicker crust basement and closer to an inferred arc (immature or early arc-rift stage: Mercier-Langevin et al. 2007c). Recent dating in the Noranda and Bousquet camps indicates that Horne and Quemont similarly formed during an episode of early rifting and felsic volcanism at Noranda at about 2,702–2,701 Ma (McNicoll et al. 2008). Rifting gradually migrated eastward to the Doyon-Bousquet-LaRonde Camp by 2,698–2,697 Ma, at which time Horne and Quemont had already formed and the relatively gold-poor Cu–Zn deposits of the Noranda Central Camp were being deposited in the more mature mafic-dominated extensional setting. The inferred geological setting of the gold-rich VMS deposits of the Blake River Group is not unique. The gold-rich Nurukawa deposit in Japan is located northeast of the main Hokuroku Basin that hosts many of the Kuroko VMS deposits of the Green Tuff Belt or Kuroko Belt (e.g., Ohmoto 1983; Ohmoto and Takahashi 1983; Tanimura et al. 1983). The Nurukawa deposit is coeval with the other VMS deposits of the district but was possibly formed in a shallower setting (Yamada et al. 1988), outside the main Hokuroku Basin where magmatic fluids or volatiles of a volcanic origin are thought to have directly contributed to the gold enrichment of the deposit (Sasaki et al. 1995; Ishiyama et al. 2001). The Nurukawa deposit, especially the number 5 orebody, with its gold-bearing stockwork and bedded siliceous ores, bears a strong resemblance to parts of the LaRonde-Bousquet 2 complex (Dubé et al. 2007b; Mercier-Langevin et al. 2007a). Other gold-rich and auriferous deposits are not associated with major felsic volcanic complexes (e.g., Greens Creek in Alaska, Abyz in Kazakhstan, La Plata and Mercedes in Ecuador, Iron King in Arizona; Table 1). The Late Triassic Greens Creek deposit in Alaska is the only significant deposit of the Admiralty District (Newberry et al. 1997; Freitag 2000), but it is located in the Alexander Terrane that hosts the giant Windy Craggy Cu–Co–Au deposit (Peter and Scott 1999; Appendix) and numerous base and precious metals occurrences. The Greens Creek deposit has been interpreted to be a VMS-SEDEX hybrid (Taylor et al. 1999, 2008) formed on a rift basin margin in a transitional setting between near arc (shallow water) and intra-arc rift (deep water) environments (Nelson and Colpron 2007; Taylor et al. 2008). The Iron Dyke deposit in Oregon is hosted in a succession of arc-related, low-K tholeiitic felsic volcanic and volcaniclastic rocks that are overlain by epiclastic sedimentary rocks and is thought to have been formed in a basin or rift setting adjacent to a sub-aerial magmatic arc (Bussey and LeAnderson 1994). Similarly, the La Plata and Mercedes deposits located in the Western Cordillera in Ecuador were formed in association with oceanic arc-related tholeiitic volcanism in an intra-arc extensional environment (Chiaradia et al. 2008). In the South Urals, the polymetallic Baimak-type deposits (e.g., Balta Tau; Table 1) are clearly enriched in gold compared to the Cyprus and Urals-type deposits and are related with the calc-alkaline andesite-dacite volcanism characterizing the more mature island-arc stage of the South Urals evolution (Prokin and Buslaev 1999; Herrington et al. 2005b). These examples illustrate the common association between gold-rich VMS forming systems and (early) arc rifting. These examples also suggest that shallow water settings are favorable for the generation of gold-enriched VMS deposits. A shallow-water setting has been demonstrated for a number of deposits and inferred for some older deposits. This common association between gold enrichments in VMS and shallow-water settings suggests that gold-rich VMS deposits are, in many cases, VMS-epithermal hybrids as previously proposed (e.g., Sillitoe et al. 1996; Hannington et al. 1999; Galley et al. 2007b). Shallow-water settings are in agreement with the inferred geodynamic setting proposed for most auriferous and gold-rich VMS deposits (i.e., early arc–back-arc rifting). As for sub-aerial epithermal deposits, VMS deposits formed in shallow water settings would probably be more prone to be eroded than their counterparts formed in deep water settings, perhaps explaining the relatively limited number of auriferous and gold-rich VMS deposits in the geologic record.

Another distinguishing feature of many gold-rich VMS deposits is the nature of their associated hydrothermal alteration. A number of gold-rich and auriferous VMS deposits worldwide are associated with zones of advanced argillic alteration and intense silicification and their metamorphosed equivalents (Sillitoe et al. 1996). Although advanced argillic or aluminous alteration assemblages are found in some gold-poor or ordinary VMS deposits (e.g., Mattabi deposit in Sturgeon Lake, Ontario, Canada; Franklin et al. 1975; Undu deposit in Fiji; Colley and Rice 1975; Sillitoe et al. 1996; some Kuroko deposits; Marumo 1989), a high proportion of the most gold-rich VMS exhibit this style of alteration (Table 1). Moreover, the aluminous alteration or intense silicification is also common around subordinate gold-rich zones in some otherwise “gold-poor” VMS deposits (e.g., Einarsson zones at Kristineberg in the Skellefte District; Areback et al. 2005; Barrett et al. 2005). Other important auriferous VMS deposits or gold-rich zones are associated with potassium feldspar alteration (e.g., Eskay Creek in British Columbia; Barrett and Sherlock 1996, and Que River in Tasmania; McGoldrick and Large 1992; Offler and Whitford 1992). The alteration assemblages at Horne are not exceptional, but the quartz-sericite-chlorite alteration affects a much larger volume of rock than the pipe-like alteration zones at other Cu–Zn deposits of the Noranda Central Camp with an overall extent that is much larger than the deposit itself (e.g., Knuckey et al. 1982). The Horne system is also characterized by relatively large zones of silica alteration that are locally anomalous in gold (Cattalani et al. 1993).

Stockwork, vein style, and disseminations of auriferous sulfides commonly represent significant parts of the ore in gold-rich VMS deposits, suggesting that subsea-floor replacement processes are important in controlling the enrichment of gold. Gibson et al. (1999) and Hannington et al. (1999) noted that deposits formed in volcaniclastic-dominated environments are commonly slightly richer in gold than their counterparts formed in volcanic-dominated environments. Examples include the LaRonde Penna, Bousquet 2-Dumagami, Bousquet 1, Westwood-Warrenmac, and Horne and Quemont deposits of the Doyon-Bousquet-LaRonde and Noranda camps in the Abitibi Greenstone Belt, the deposits of the Mount Read Volcanics in Tasmania, the Boliden deposit in Sweden, the Nurukawa Zone 5 in Japan, the Mount Morgan deposit in Australia, the Separate Gold Zones at Lalor in Canada, and the Balta Tau deposit in Russia (Table 1) as well as the Johnson River and Iron Dyke deposits located in Alaska and Oregon, respectively. Other deposits are characterized by barite-rich tops that can contain significant amount of gold. Examples include Kali Kuning and Lerokis in Indonesia, Balta Tau in the South Urals, Lomero-Poyatos in the Iberian Pyrite Belt and Iron Dyke in Oregon. Deposits dominantly formed by seafloor exhalative activity are often characterized by large amounts of gold but have a relatively low gold grade (e.g., Caribou in the Bathurst Mining Camp, La Zarza in the Iberian Pyrite Belt, Flin Flon in Manitoba; Table 1).

Gold-rich and auriferous VMS also commonly possess a complex ore mineral paragenesis, with the common occurrence of arsenopyrite, arsenian pyrite, tennantite and tetrahedrite, bornite and complex sulfosalts and tellurides, and trace element signatures that include enrichments in the epithermal suite (Ag, As, Sb, Hg) or interpreted magmatic associations (e.g., Bi, Te) (Hannington et al. 1999). The gold occurs in two principal metallogenic associations: Au–Cu and Au–Zn (Huston 2000). Both are represented in Table 1 and Appendix and can occur together in the same district (e.g., Au–Cu at Boliden and Au–Zn–Pb–Ag at Petiknas North and Holmtjarn in the Skellefte District in Sweden). The two metallogenic associations can even coexist in a single deposit (e.g., Au–Cu in the 20 North lens and Au–Zn–Pb–Ag in the 20 South lens at LaRonde Penna: Dubé et al. 2004, 2007a). Importantly, there is no correlation between the metallogenic association and the gold grade or total gold content of the deposits.

Gold in anomalous, auriferous and gold-rich VMS is present in four main mineralogical forms: (1) native gold, (2) electrum, (3) tellurides, and (4) “invisible” or refractory gold. Native gold is common in many deposits (e.g., Horne, Bousquet 1, Bousquet 2-Dumagami, Boliden, Hellyer, Johnson River, and Kali Kuning) as very fine grains in silicates, sulfides, and sulfosalts, filling fractures or cleavages or at grain boundaries. Native gold is also common as minute inclusions in sulfides and sulfosalts. Electrum is also very common and represents the main residence site for gold in a number of deposits (Huston 2000). Its composition varies significantly from a deposit to another and within single deposits or lenses. It can be gold-rich (e.g., Bousquet 2-Dumagami, Balta Tau and La Plata), silver-rich (e.g., LaRonde Penna, Eskay Creek, Greens Creek and Nurukawa), and even mercury-rich (electrum-amalgam; e.g., Boliden, La Zarza, Eskay Creek). Auriferous tellurides (mainly calaverite, petzite, sylvanite, and krenerite) are also quite common, especially in older deposits or high-grade zones (e.g., Bousquet 1, Bousquet 2-Dumagami, Horne, Mount Morgan, and some deposits of the Southern Urals). Gold is “invisible” or refractory in many VMS deposits, especially in anomalous deposits (e.g., deposits of the Bathurst Mining Camp; McClenaghan et al. 2009). Most of the invisible gold is located in primary arsenian pyrite, arsenopyrite, and pyrite. Varying amounts of primary invisible gold can be liberated to form native gold, electrum, and tellurides through zone refining and during regional metamorphic and tectonic events.

Some of the characteristics that appear to distinguish gold-rich and auriferous VMS from other VMS deposits can be mapped in the field and used as vectoring tools. At the district scale, the gold-rich deposits occupy a stratigraphic position and volcanic setting that commonly differs from other deposits of the district (e.g., outside a basin or main structure that hosts the majority of the deposits; in the stratigraphically oldest or youngest positions; associated with locally thickened packages of felsic rocks which may be related to a step change in the geodynamic and magmatic evolution of local volcanic complexes). At the deposit scale, uncommon alteration assemblages and trace element signatures may be recognized (e.g., advanced argillic, aluminous, strongly siliceous, or potassium feldspar alteration).

Conclusions

Volcanogenic massive sulfide deposits are a significant repository of gold, with several being world-class gold deposits. A number of VMS deposits are mined mainly for their gold content, with average gold grade in g/t higher than the combined content of base metals in weight percent (e.g., Fig. 7). However, the majority of VMS deposits are characterized by a relatively low gold grade (averaging less than 2 g/t Au). A significant number have grades between 2 and 3.46 g/t (geometric mean plus one standard deviation from the geometric mean). In some deposits, both the gold grade and total base metals grade are high. We suggest that the geometric mean plus one geometric standard deviation can be used to discriminate subclasses of VMS deposits worldwide and within individual districts based on their gold content. In our analysis, VMS deposits that contain more than 3.46 g/t Au are auriferous, regardless of their base metals content. Many of these deposits have gold grades (in g/t) higher than the combined content of base metals in weight percent, as previously noted by Poulsen and Hannington (1996) (Figs. 7 and 8) and are considered auriferous as well. However, some deposits with lower gold grades nevertheless contain significant amounts of gold because of their large size. Our analysis shows that deposits with 31 t of gold or more (~1 Moz) exceed the geometric mean plus one geometric standard deviation and are clearly anomalous (Fig. 8). Deposits with a grade of more than 3.46 g/t Au and 31 t Au or more are considered gold-rich. Consideration of the geometric mean for smaller populations in individual districts or mining camps can also help to identify those deposits with statistically significant gold enrichments (e.g., Fig. 5), even if they have gold grades that are far below the global mean value.

The identification of truly anomalous, auriferous, and gold-rich deposits helps to identify the geological parameters that may be responsible for atypical gold enrichments in VMS. There is a distinct provinciality controlling the distribution of some gold-rich deposits that strongly suggests a regional geological control on the enrichment in gold in some VMS districts. As previously noted by Hannington et al. (1999), the most gold-rich deposits in a district almost always have one or more attributes, apart from their high gold contents, that distinguish them from other “gold-poor” or ordinary VMS in the same district. Most of the gold found in the VMS environment is contained in a small number of deposits that are geologically distinct from other deposits in the same district and these differences may be valuable guides for exploration of other gold-rich VMS.