Sei sulla pagina 1di 21

Bull V olcanol DOI 10.

1007/s00445-011-0515-7

RESEARCH ARTICLE

The influence of volcanological and sedimentological processes on diamond grade distribution in kimberlites: examples from the EKATI Diamond Mine, NWT, Canada
Lucy A. Porritt & R. A. F. Cas & L. Ailleres & P. Oshust

Received: 26 February 2010 / Accepted: 5 June 2011 # Springer-V erlag 2011

Abstract The distribution of diamonds within individual kimberlite pipes is poorly documented in the public domain due to the proprietary nature of the data. The study of the diamond distribution within two pipes, Fox and Koala, from the EKATI Diamond Mine, NWT, Canada, in conjunction with detailed facies models has shown several distinct relationships of deposit type and grade distribution. In both pipes, the lithological facies represent grade units which can be distinguished from each other in terms of relative size and abundance of diamonds. A positive relationship between olivine grain size and abundance with diamond size and abundance is observed, indicating that sorting of fragmental kimberlites influences diamond distribution. Though surface geological processes do not control the diamond potential of the erupting magma, they can be responsible for concentrating diamonds into eco-

nomically significant proportions. A good understanding of the eruption, transport and depositional processes responsible for the individual lithological units and the diamond distribution within them is important for successful resource estimation. This may lead to recognition of areas suitable for selective mining, making a marginal deposit economic. Keywords Kimberlite . Diamond . Grade distribution . V olcanology . Sedimentology

Introduction Diamonds form at depths of over 150 km in the subcontinental lithospheric mantle, in the diamond stability field where the pressures and temperatures are favourable for carbon to exist as diamond (Meyer 1987; Harris 1992). The main mantle lithologies in which diamonds form are peridotite (dunite, harzburgite, lherzolite), websterite and eclogite. Diamonds and other mantle minerals are transported to the surface by kimberlite and lamproite magmas, which randomly entrain xenolithic and xenocrystic mantle and crustal material en route. Where present, diamonds are found in very low abundances (<3 ppm, Gurney et al. 2005), which, depending on the value of the diamonds, may or may not be considered sufficient to be economically mined. Globally, the majority of known kimberlite bodies are in fact considered to be non-diamondiferous or contain sub-economic grades, with only as few as 1% eventually becoming diamond mines (Rombouts 1999). There are several factors which influence the diamond carrying potential of the magmas before they reach the near surface. If the kimberlite magma does not rise through diamondiferous mantle, then it cannot entrain diamonds.

Editorial responsibility: R.S.J. Sparks This paper constitutes part of a special issue: Cas RAF, Russell JK, Sparks RSJ (eds) Advances in Kimberlite V olcanology and Geology. L. A. Porritt (*) : R. A. F. Cas : L. Ailleres School of Geosciences, Monash University, Clayton 3800, VIC, Australia e-mail: lporritt@eos.ubc.ca L. A. Porritt Department of Earth and Ocean Sciences, University of British Columbia, 6339 Stores Road, V ancouver V6T 1Z4, BC, Canada P . Oshust AMEC Americas Ltd, 111 Dunsmuir St, V ancouver V6B 5W3, BC, Canada

Bull V olcanol

Even if the magma rises through a fertile zone, it will not necessarily entrain diamonds in sufficient quantities to form economic deposits. If diamonds are entrained by the magma, they must also survive the transport to the surface. The temperature and oxygen fugacity (O2) of the host magma influences diamond dissolution with a high-T, high-O2 magma being less favourable for diamond preservation (Fedortchouk et al. 2005). The degree of interaction with the kimberlite melt also controls dissolution, and diamonds residing within xenoliths are better protected than those liberated during transport to the surface. The nature, as well as the abundance, of the diamonds within any diamondiferous mantle may also affect the economic potential of the magma-batch, with stone value affected by size, clarity, shape and colour of the stones. The abundance and value of the diamonds that each individual magma or magma pulse carries to the surface can be inconsistent, and grade variations within individual intrusions have been documented e.g. W3 kimberlite at Wesselton and in the De Beers pipe (Clement et al. 1986). Diamond type (value) and abundance typically varies widely between pipes of the same kimberlite field, though it may show similarities within a temporally and spatially related cluster (Gurney et al. 2005). At the surface, volcanological processes operate with the result that the diamond abundance within the deposits may be modified compared to the initial proportion carried within the magma. Subsequent sedimentary processes may also occur, which can further redistribute the diamonds and influence their concentration in any given deposit. Diamond abundance commonly varies within individual pipes, due in part to their low overall abundance and to the varied relative abundance of other components of the deposits. Several authors have described grade variation between lithofacies with a distinct geological control on diamond grade, they attribute it to sorting and country rock dilution (e.g. at the Mwadui kimberlite, Tanzania, Stiefenhofer and Farrow 2004), and difference in grade of individual erupting magma pulses (e.g. several southern African occurrences, Clement 1982; and the Victor kimberlite, Canada, Webb et al. 2004). Grade variations within individual lithofacies have also been described where diamond grade is elevated in the coarse basal portion of a mega-graded bed at the 140/141 kimberlite, Canada (Berryman et al. 2004). Diamond has a high specific gravity (3.52) with respect to common silicate minerals such as quartz; thus, diamonds (and other mantle minerals such as olivine and pyrope) are easily sorted by particle size and density contrasts under tractional sedimentary conditions. Sorting due to size and density is apparent in sedimentary diamond deposits where diamonds have been eroded from the kimberlite source and deposited within alluvial and marine gravels (and in Brazil, Svisero 1995; e.g. in southern Africa, de Wit 1999). The physical processes affecting the distribution of diamonds

within kimberlite pipes are not well documented, due to the proprietary nature of such data, the poor understanding of intra-vent processes and products and the inherent difficulties involved in diamond sampling at a scale that can be related to depositional features. Kimberlite pipe-filling deposits are dominated by pyroclastic and resedimented volcaniclastic lithofacies. A variety of volcanic fragmentation and transport processes has been recognised from the deposits within kimberlite pipes, which show a range of lithofacies with different textural characteristics such as particle sorting and internal structure. For the most part, transport distances are short with the pipe being the source vent as well as the site of deposition; however, pyroclastic flow deposits from other source vents are now being recognised (Moss et al. 2008; Gernon et al. 2009a). Here, we assess the volcanic and sedimentary processes resulting in the deposits within two kimberlite pipes on the EKA TI property, Fox and Koala, and how those processes may have influenced the internal grade distribution. The Koala kimberlite is smaller in volume but carries a higher grade than the Fox kimberlite, with a total measured, indicated and inferred resource of 18.5 Mt at 0.95 ct/t in Koala and 28.7 Mt at 0.4 ct/t in Fox (based on a 2002 inventory, Dyck et al. 2004). The effects of grade dilution by country rock lithic fragments and disaggregated sediment, elutriation both within the eruption column and during subsequent transport and deposition and deposit grain size and sorting are discussed. Absolute diamond grades are not presented due to the proprietary nature of the data; however, normalised data are presented (normalised to volume), analysed and interpreted.

Methods and data Diamond distribution data, stones and carats per cubic meter, have been provided by BHP Billiton Canada Inc. for the Fox and Koala kimberlites, two of the pipes currently exploited as part of the EKA TI Diamond Mine, NWT, Canada. Large diameter reverse circulation drill holes were used to sample the lithologies within the pipes for diamond grade (using methods outlined by Dyck et al. 2004); 3D geological models of each pipe have been compiled by BHP Billiton Canada Inc., in Vulcan software, and have been used throughout the grade sampling and determination process providing a framework for resource evaluation. These models have been built on the basis of lithological data from >70 diamond drill holes per pipe, plus the logging of chips from the reverse circulation drill holes. Previous work included detailed logging of 23 diamond drill cores from Fox and 17 from Koala which enabled facies and emplacement models to be determined for each of the lithological units (Porritt 2008; Porritt and Cas 2009,

Bull V olcanol

2011). Factors such as grain size distribution and extent of dilution by country rock lithic fragments are considered as important in predicting grade variations between and within facies (Field et al. 2009; Scott Smith and Smith 2009). Olivine abundances were determined using visual percentage comparison charts for reference (based on Terry and Chilingar 1955). Average and maximum olivine grain size was also visually determined during core logging to give a continuous estimate of down-hole grain size variation. Both visual estimates and lithic fragment counts of grains >2 mm were used to track changes in abundance and proportion of country rock dilution within the Fox drill core. More detailed crystal size distribution methods (e.g. Holden et al. 2009; Jerram et al. 2009; Moss et al. 2010) were not utilized here due to their limited frequency of observations: detailed analysis of representative samples, rather than continuous logging. The Koala database consists of 301 samples and the Fox database consists of 947 samples; of which all but ten are large diameter reverse circulation (RC) drill hole samples, the remaining ten are sampled from underground drift rounds from the Koala pipe. The geological affiliation of each sample has, for the most part, been characterised by Ekati geologists based on the logging of the reverse circulation drill chips; however, in a few cases (<15 per pipe), the samples have straddled lithological boundaries and have been reassigned to the relevant units based on the 3D geological model. At Fox, one of the lithologies (F3) has been recognised by detailed core logging only and was not determined during grade sampling; the samples within this unit have been assigned based on their location within the 3D model alone. The average length of samples from

Koala is 14 m and from Fox 16 m, though shorter sample intervals down to 6 m and larger sample intervals up to 30 m in length are also included. RC samples were taken over a period of 7 to 8 years using different size equipment, boring a range of hole sizes (~2771 cm diameter) with the majority between 31 and 35 cm (Dyck et al. 2004). The volume of the in situ sample was determined from drill hole data using down-hole calliper measurements and sampling intervals. Despite the apparently wide variation in sample lengths and hole diameter over time, the mass of the samples remained relatively constant at approximately 7 t varying between 5 and 9 t; samples less than 4 t were excluded from the dataset. The diamonds obtained from the recovery circuit, giving an effective diamond cut-off size of 1 mm (Dyck et al. 2004), were counted and weighed and then normalised by sample volume to give comparable data as carats per cubic metre and stones per cubic metre (see Table 1 for some basic statistics). The 1-mm bottom screen slot size was consistently used during the processing of all samples used in this study. Average stone weight (ct/st) and size in millimetre of derived equivalent mean diameter of spheres with the same volume is also presented on Table 1. The location of each sample within the pipe was assigned an x-, y-, z-coordinate based on the mid-point of each sample interval to enable 3D visualisation and modelling. Based on these data, a comparison between the distribution of diamond size (given as weight measured in carats) and abundance (number of stones) in relation to the transport and depositional processes involved in the formation of each of the individual lithofacies within the pipes is presented. The differences in diamond distribution between pipes are also considered, and the influences of

Table 1 Basic statistics describing the diamond size distribution within each pipe and within the individual facies Average ct m3 Koala All P1 P3 P4 P5a P5b P6 P7 Fox All UFA F3 F1 Average st m3 Correlation coefficient (ct m3 : st m3) Average ct/st Spherical diameter equivalent (mm) Standard deviation (ct/st) Kurtosis (ct/st) No. samples

2.33 2.18 1.98 0.75 0.77 1.61 4.25 2.17 0.58 0.23 0.87 0.63

19.59 24.72 13.59 6.36 7.19 12.32 32.01 20.91 5.55 2.38 7.33 6.74

0.87 0.89 0.80 0.96 0.72 0.90 0.72 0.82 0.72 0.70 0.59 0.69

0.12 0.07 0.10 0.11 0.11 0.12 0.13 0.10 0.10 0.09 0.12 0.09

2.3 1.9 2.4 2.2 2.2 2.3 2.5 2.2 2.1 1.9 2.3 2.1

0.06 0.06 0.06 0.05 0.07 0.04 0.04 0.03 0.07 0.07 0.07 0.06

1.77 3.67 0.67 1.12 2.73 0.74 0.38 -0.38 1.53 1.39 1.61 1.81

301 67 42 15 74 13 79 11 947 305 339 303

Diamond data has been normalised by volume to give values as carats per m3 and stones per m3

Bull V olcanol

primary diamond content of the magma and eruption style are discussed. Limitations on geological grade prediction One of the intrinsic problems related to predicting diamond grade from diamond data is the low abundance of diamonds and the inherent high nugget effect. The nugget effect is an indication of the variability over short distances, less than the sampling interval (Davis 1986), and in kimberlite deposits, this is high mainly due to the stone size variance (Rombouts 2003). The low abundance of diamonds means that the size of the samples plays a role in influencing the observed grade. The absolute grade of a deposit is only known once it has been completely mined and all the recoverable diamonds removed. Feasible grade prediction relies on achieving a balance between an adequate amount of samples of sufficient sizes to produce a reliable estimate of the grade and the cost of sampling. Large sample sizes, commonly from large-diameter RC drill holes, are required to provide representative assay data, with samples at EKA TI typically ranging from 5 to 9 tonnes (Dyck et al. 2004). Verification of sample assays by analysing duplicate samples, as is common for metal deposits, is not possible with such large sample sizes. Due to the cost and logistical difficulties of increasing the diameter of RC drilling, sample lengths have to be relatively large (>6 m but more commonly ~1530 m at EKA TI) providing low resolution compared to observed geological variations. The RC sampling method therefore has only limited selectivity, especially where thin beds or rapid facies variations occur. Realistically, for mining, this sample resolution is sufficient due to the scale of the bulk mining methods employed at EKA TI; however, a finer resolution would provide more detail when assessing the geological controls on diamond distribution. Resource evaluation for diamond deposits is not straightforward as it is not only the total carats but also the number, size and value of the stones that influences the economic potential of the ore. There is the potential for the diamond size distributions to have been modified by breakage and losses during RC drilling and processing. The sample collection and diamond recovery processes employed at EKA TI were designed to provide reliable, representative data for use in resource modelling, as outlined by Dyck et al. (2004). Evaluation of diamond populations from RC drilling and mine production has shown minimal additional stone damage associated with drilling (Dyck et al. 2004). Losses during sampling are hard to quantify; however, the efficiency of the processing plant was closely monitored and diamond recovery maximised (Dyck et al. 2004).

The normalised data for this study were provided by BHP Billiton Canada Inc. as carats per cubic metre and stones per cubic metre, and no validation or statistical analysis of the original stone counts and sizes was possible by the authors. However, these data have been used in resource evaluation, and it can be assumed that they have been validated for this purpose and a satisfactory control on the data obtained. See Dyck et al. (2004) and Harrison et al. (2009) for more details on the resource modelling methods used at EKA TI.

Pipe geology and volcanology The Koala and Fox pipes are part of the Ekati Diamond Mine, which is part of the Lac de Gras Kimberlite Field, located within the Contwoyto Terrane of the Central Slave Province, Canada (Fig. 1). The dominant country rocks in the area are Archean granitoids which have intruded into the metasediments of the Yellowknife Supergroup. Both pipes are hosted by the granodioritic Koala Batholith, in the south-central portion of the Ekati Property. This composite batholith has an age of 2.6 Ga (Kjarsgaard et al. 2002). Fossil evidence from sedimentary xenoliths, mainly comprising silts, muds and quartz sands, indicates that the bedrock was covered by Cretaceous marine and Palaeocene lacustrine strata (Nassichuk and Dyck 1998) during the time of kimberlite volcanic activity. The fluidal forms of mudstone xenoliths within both pipes indicate that at least some of the cover sequence sediments were partially consolidated or unconsolidated at the time of kimberlite eruption. The Koala kimberlite has been dated using RbSr giving an age of ~53 Ma (Creaser et al. 2004), the Fox kimberlite has not been successfully dated, though the sedimentary xenoliths found within the Fox deposits would place it within a similar age range to other kimberlites on the property (7545 Ma, Creaser et al. 2004). Current estimates are that the surface at the time of eruption was perhaps only a few hundred metres above that of the present day (Nowicki et al. 2004). Fox pipe The Fox pipe has been divided into four facies and two sub-facies (F1 to F4 and sub-facies F1a and F4a, Porritt and Cas 2009). F2 is a granodiorite boulder breccia with only a minor kimberlite component, and the sub-facies are volumetrically unimportant, so for simplicity here, we focus on the three main kimberlite facies (F1, F3 and F4) which are easily distinguished from each other by differences in grain size, olivine content, presence or absence of accretionary lapilli and granodiorite country rock abundance (Fig. 2, Table 2). The lowermost facies (F1) is a

Bull V olcanol Fig. 1 Map showing the local geology and location of the Fox and Koala pipes on the EKA TI Property, NWT, Canada, modified from Nowicki et al. (2004). Other kimberlite pipes within the current mine plan are also shown. Insert showing location of EKA TI and the Lac de Gras kimberlite field
540 000
Slave Craton

Bathurst Inlet

N
7200000

7200000

66N

EKATI

Lac de Gras KimberliteField


64N

Yellowknife
100 km 114W 110W

Koala Fox

7160000

7160000

Post tectonic granitoids Pre / syn-tectonic granitoids Metasediments Kimberlite (Ekati Mine) Migmatite 0 Mafic Inclusion Undefined
480 000

10

20 km
540 000

massive, poorly sorted, matrix supported, granodioritelithic-rich lapilli-tuff, containing 1030% strongly altered olivine grains. F1 is considered to be the result of the collapse of a large eruption column early in the eruption, before complete vent clearing had occurred, indicated by the high abundance of granodiorite lithic fragments
Fig. 2 Cross-sections of the Koala and Fox pipes showing the internal geology. Koala: P1 mega-graded bed; P3 multiple normally graded beds; P4 woodrich silty fine sandstone; P5a massive poorly sorted, mud-rich mass flow deposit; P5b granodiorite boulder breccia; P6 massive poorly sorted, olivinerich column collapse lapilli tuff; P7 macrocrystic coherent kimberlite. Fox: UF A weakly bedded, sandy mudstone; F3 massive to weakly bedded, poorly sorted accretionary lapilli bearing, column collapse lapilli tuff; F2 granodiorite cobbleboulder breccia; F1 massive, poorly sorted, granodiorite-rich, column collapse lapilli tuff

(>40%). Column collapse was most probably triggered by vent over-widening leading to overloading of the eruption column with cold dense country rock lithic fragments (Porritt et al. 2008a, b; Porritt and Cas 2009). The dynamics of kimberlite column collapse are not well constrained, and an alternative model involving waningFOX

KOALA

P1

UFA F2 F3 F2 F2

P3 P4 P5a P5b P6 F1

100 m

P7

Bull V olcanol Table 2 Textural and petrographic characteristics of the individual facies from both Fox and Koala kimberlites
Pipe and unit Bedding Sorting Texture Olivine macrocryst content 010% Avg grain size Medium ash Non-kimberlite material Minor granodiorite (up to 10% in F4a) >30% granodiorite of all sizes >40% granodiorite of all sizes Interpretation

Fox UFA Minor (F4 and F4a) Fox F3 Massive

Moderate

Mudstone

Resedimented phreatomagmatic crater-rim material Pyroclastic column collapse lapilli tuff with phreatomagmatic input Pyroclastic column collapse lapilli tuff Pyroclastic flow deposit modified by deposition into crater lake Pyroclastic flow deposits modified by deposition into crater lake Crater-lake sediment

Poor

Matrix supported Matrix supported

2025%

Lapilli

Fox F1 Koala P1

Massive

Poor

2025%

Lapilli Fine ash - Lapilli

Single normally Moderate graded bed ~100 m Normally graded Laminated Moderate

Clast to matrix 3540% at the supported base, to 0% at the top Clast to matrix 3035% (up to supported 80% in coarse bases) Mudstone <10% (though up to 20% in few discrete horizons) 1020%

Koala P3

< 5% lithic clasts in base to abundant 1015% quartz sand grains in top Coarse ash lapilli Minor granodiorite (fine ash tops to some beds) Abundant quartz sand, silt, wood fragments Abundant mudstone, wood and minor granodiorite Granodiorite boulders >40%

Koala P4

Well

Medium-ash/ fine-sand Coarse ash-lapilli

Koala P5a

Massive

Poor

Matrix supported Undetermined in drill core Matrix supported

En masse crater rimcollapse resedimented volcaniclastic deposit Pipe wall collapse boulder breccia

Koala P5b

Massive

Poor

Koala P6 Koala P7

Massive None

Poor

Block/boulder 1020% in matrix to boulders only 2565% (often Lapilli with narrow rims) 030% Fine-grained groundmass, coarse macrocrysts (up to ~10 mm)

<10% granodiorite Pyroclastic column and mudstone clasts collapse lapilli tuff Minor granodiorite along pipe walls Coherent kimberlite intrusion

Inequigranular Macrocrystic, minor flow alignment

stage in-vent deposition and large-scale fluidization of pyroclasts has been proposed (Sparks et al. 2006; Brown et al. 2008). Though we favour the column collapse model here, both processes could result in a massive poorly sorted pyroclastic deposit where limited internal grain-size sorting is invoked. F1 is overlain by a massive to weakly bedded, poorly sorted, matrix supported, accretionary lapilli-bearing (<1%) and granodiorite-lithic-rich lapilli tuff (F3). The two facies are very similar in appearance to each other, with the presence of accretionary lapilli being the distinguishing feature. The contact between the two facies varies between drill holes from gradational to sharp, with a 10 20-cm thick horizon of accretionary lapilli and/or a granodiorite boulder breccia horizon marking the sharp contact. F3 is considered to be the result of the modification of the column collapse material by phreatomagmatic interaction of magma or hot pyroclasts with influxing water (Porritt and Cas 2009). The uppermost facies (F4) comprises a package of massive to weakly bedded, moderately sorted, matrix supported, olivine-poor, accretionary lapilli-bearing (<5%), kimberlitic mud and fine sand. A granodiorite lithic-rich

sub-facies (F4a) occurs at the base of F4, and together, these lithofacies are termed the Upper Facies Association (UFA, Fig. 2, Porritt and Cas 2009). The UFA is considered to represent the products of a late phreatomagmatic stage of eruption at Fox, continuous from the F3-forming event, which were deposited onto the crater rim. This material was subsequently resedimented into the vent during a period of post-eruption stabilization and incremental wasting of the tephra rim (Porritt and Cas 2009). Koala pipe The Koala pipe is divided into seven distinct geological facies all distinguished by internal textural characteristics, olivine grain size and abundance and detrital material abundance (Fig. 2, Table 2). The pipe is dominantly filled by fragmental deposits, with a small volume coherent kimberlite body (P7) occupying the base of the pipe (Crawford et al. 2006; Porritt 2008; Porritt and Cas 2011). Based on the morphology, the contacts with the surrounding fragmental deposits and its coherent appearance P7 is considered to be a late-stage intrusion (Crawford et al. 2006; Porritt 2008; Porritt and Cas 2011).

Bull V olcanol

The fragmental deposits are arranged in a subhorizontally layered sequence, with each distinct facies representing different depositional events or conditions. The lowermost fragmental facies (P6) is a massive, poorly sorted, matrix supported, olivine-rich, pyroclastic deposit (Nowicki et al. 2004; Crawford et al. 2006; Porritt and Cas 2011). The origin of massive volcaniclastic kimberlites remains enigmatic; here, we interpret P6 as the intra-vent deposit from a collapsing waning-stage eruption column (Porritt 2008; Porritt et al. 2008a; Porritt and Cas 2011). The paucity of country rock lithic fragments (<10% of the deposit, Table 1) indicates that significant vent clearing had occurred prior to the deposition of this lithofacies, which differs from the Fox pipes F1 column-collapse deposit. Alternative mechanisms for the formation of this type of deposit include phreatomagmatic fragmentation and mixing (e.g. Lorenz et al. 1999) and fluidisation of waning-stage intra-vent deposits (e.g. Sparks et al. 2006; Gernon et al. 2009b). Regardless of which of these three models one adopts, it is the pyroclasic, massive, poorly sorted nature of this deposit which is important as, however it was deposited, limited sorting of the components has occurred. P6 is overlain by a massive, poorly sorted, matrix supported, mud-rich, volcaniclastic lithofacies (P5a, Table 2) with a distinct granodiorite-boulder breccia horizon at the base (P5b) containing up to 80% granodiorite boulders in a mixed matrix of P5a and P6 material. This deposit is interpreted to represent the post-eruption collapse of the pipe walls (P5b) and tephra cone (P5a) into the pipe (Nowicki et al. 2004; Crawford et al. 2006; Porritt and Cas 2011), through mass wasting and debris flow (Porritt 2008; Porritt and Cas 2011). Overlying P5a is 10 to 30 m of laminated, moderate to well-sorted, matrix supported, wood-rich, silty fine-grained sandstone (P4) which is dominated by quartz sand and silt. P4 contains only minor kimberlitic material, mainly concentrated in discrete, <1 m thick olivine-bearing horizons. This lithofacies is interpreted as crater lake sediments signifying the cessation of volcanic activity at Koala (Nowicki et al. 2004; Crawford et al. 2006; Porritt 2008; Porritt and Cas 2011) and most likely representing several thousands of years of sedimentation (Porritt 2008; Porritt and Cas 2011). P4 has a sharp erosional contact with the overlying facies which comprises a variably bedded sequence of normally graded, moderately well-sorted, olivine-rich beds (P3) which grade from granule-sized olivine grains at the base to fine-grained olivine sand or black mud at the top of individual graded beds. Erosional contacts between beds and incomplete or partial graded beds are common. In general, individual beds are not correlatable between drill holes and the beds appear to be coarser, thicker and fewer in number in the central portion of the pipe (Fig. 9b). This

lithofacies has been interpreted to represent the input of volcanic material from small collapses of the tephra rim (Nowicki et al. 2004; Crawford et al. 2006), or influx of material from eruptions of a nearby volcano (Porritt 2008; Porritt and Cas 2011). During the several thousands of years of P4 deposition, it is likely that some degree of induration of the remnant crater rim deposits would have occurred. Such induration would both inhibit the ability of the tephra rim material to collapse into sediment gravity flows and increase the probability of cohesive slumping incorporating clasts showing original crater rim layering into the subsequent deposits. Such clasts are not observed; therefore, an alternative volcanic source for this material has been proposed (Porritt 2008) and is favoured here. The uppermost facies at Koala is a ~100-m thick megagraded bed (P1). This bed grades almost continuously from a coarse olivine-rich lapillistone at the base to fine olivine sand and quartz siltstone at the top; it is matrix supported and moderately to poorly sorted throughout. Grading is defined by olivine size and abundance, with olivine disappearing in the upper ~30 m and fine quartz sand becoming the primary framework grain type. The contact with P3 is sharp and erosional. The origin of P1 is considered enigmatic and mass collapse of the Koala tephra cone (Nowicki et al. 2004) and input from a pyroclastic flow sourced from a nearby volcano (Crawford et al. 2006; Porritt 2008; Porritt and Cas 2011) have been proposed. Here, we favour input from a pyroclastic flow due to the absence of clasts of tephra rim material that would have likely indurated to some degree during the P4 period of deposition. The presence of such clasts would be expected from mass collapse of the tephra rim. No large blocks or xenoliths are observed, which may have been lagged out during transport, and it is possible that this represents a distal pyroclastic flow deposit. The key feature of this bed is the large-scale grading.

Grade distribution The grade of the ore is dependent on several measured parameters: number of recovered stones; total weight of recovered stones; volume of sample they were recovered from and the average density of the material they were recovered from. For simplicity, bulk density measurements are not included here; instead, diamond grade data are presented in carats per cubic metre and stones per cubic metre, and the average weight of the stones is calculated from these data as carats per stone. The average size of the stones is useful in comparing the dataset as a whole and assessing the influence of stone sizes on the population. The data have been analysed based on the host lithofacies and vertical position within each pipe, and then also in

Bull V olcanol

three dimensions using the GoCAD modelling software. The data and data analysis presented here are not intended as an example of resource estimation; good descriptions of the resource estimation methods employed at EKA TI are presented in Dyck et al. (2004) and Harrison et al. (2009). Data analysis Histogram plots of average stone size (ct/st) show that the data from both pipes have a positively skewed distribution with an abundance of smaller diamond grain sizes (Fig. 3). The stone size populations approximate a log-normal distribution, which is not unexpected (Rombouts 1995). The Fox dataset shows an apparent single sample population (mono-modal) which is also evident in the histograms of samples from the individual facies. However, F3 exhibits a mode value slightly higher than UFA and F1. The Koala dataset is apparently polymodal and shows a larger variance, which may relate to the presence of smaller subpopulations than at Fox. This may indicate that more than one population exists when considering small sample populations. The presence of different sample populations is supported by the histograms of samples from the
Fig. 3 Histograms showing the distribution of average size data (ct/st) for all samples from both Fox and Koala. The data from the individual lithofacies at Fox (F1, F3 and UFA) show similar distributions to each other and to the whole sample set plotted together, showing a single population (shown by curve). Histograms for the individual lithofacies at Koala are also shown, with P1, P3, P5a and P6 showing different distribution to each other. When combined, the data show a more complicated size distribution than at Fox, with three possible populations (highlighted by curves)
200

individual Koala facies, which show the following distributions: P1 shows a mono-modal log-normal distribution; P3 has a main log-normal distribution with one to two other modes perhaps indicating the presence of other populations; P5a shows a mono-modal normal distribution and P6 appears to show a distribution similar to P5a mixed in with a log-normal population (Fig. 3). Within both pipes, carats per cubic metre and average size vary with depth (Fig. 4), showing associations with the individual geological units, stones per cubic metre also follows similar trends. The relationship between carats and stones per cubic metre in both pipes is not strictly linear, with a large spread of data showing that there are variations in individual stone sizes which contribute to the total grade in any individual sample (Fig. 5). The correlation coefficient between carats and stones per cubic metre varies between pipes as well as between the facies within each pipe with a value of 0.87 for all samples at Koala and 0.72 at Fox (Table 1). Higher correlation coefficients indicate either a better sorted diamond population in which the stones contribute evenly to the diamond grade or a poorly sorted population where abundance and contribution to total grade of the different size fractions increases proportionally.
60

Fox all

UFA
n = 305

n = 947 st dev = 0.065


80

Frequency

F3

0.5

n = 339

100
80 0.5

F1

n = 303

0.5

0 0 0.1 0.2 Ct/st 0.3 0.4 0.5

50

Koala all

20

P1

14

P5a
n = 74

n = 67 n = 301 st dev = 0.063

Frequency

0 9

0 0.5

0.5

25

P3

18

P6
n = 79

n = 42

0 0.5

0.5

0 0 0.1 0.2 0.3 Ct/st 0.4 0.5

Bull V olcanol Fig. 4 Scatter plots showing how both carats per cubic metre and average size (ct/st) vary with depth in both Koala and Fox. Horizontal axes are on the same scale for each pipe but are unlabelled due to the proprietary nature of the data. Different symbols are shown for the different lithologies within each pipe. Distinct trends of carats per cubic metre with lithology can be seen in both pipes and the higher-grade nature of Koala compared to Fox is illustrated. When comparing the average size of the diamonds, lithological control appears less important, with both pipes showing similar diamond size distribution with Fox perhaps showing a slightly larger population of diamonds. In both the cpm3 and ct/st plots, the samples from Koala P1 show a welldeveloped grading with depth which reflects the grain size grading of the deposit; the smaller average size in the upper portions of the deposit are not replicated in any of the other units within the pipe indicating that extreme sorting, fragmentation during transport and/or a different original diamond grain size distribution may be represented here
500

Koala ct m-3

500

Fox ct m-3

400 Depth metres above sea level

400

300

300

200

200

100

100

-100

-100

6 ct m-3

12

-200 0 6 ct m-3 12

500

Koala ct/st

500

Fox ct/st

400 Depth metres above sea level

400

300

300

200

200

100

100

-100

-100

0.2 0.4 Average size (ct/st)

0.6

-200 0 0.2 0.4 0.6 Average size (ct/st)


P1 P3 P4 P5A P5B P6 P7

F1

F3

UFA

3D spatial distribution patterns The diamond data have been modelled in 3D using GoCAD to assess the spatial relationship of data points to each other and to determine whether internal grade patterns can be seen within the different lithologies, which may be related

to smaller scale deposit characteristics observed in drill core. Original data points were spatially located and colourcoded from black/blue (low relative values) through to red/ white (high relative values) for the stone abundance data for both pipes (Fig. 6). This display of the raw data shows that the different lithofacies have different stone abundance

Bull V olcanol Fig. 5 Plots showing the covariance between carats per cubic metre and stones per cubic metre in both Koala (a, c) and Fox (b, d), with the whole datasets represented in (a) and (b) and focussing on the lowgrade end of the spectrum in (c) and (d) (area shown by boxes on (a) and (b)). A large spread of data points shows that there is no simple linear relationship between carats per cubic metre and stones per cubic metre, indicating a range of diamond sizes contributing to the population. The difference in grade between the pipes is obvious, and similarities in the shape of the low-grade graphs can be seen. Covariance trend lines are shown for single populations comprising 1, 0.1 and 0.01 ct stones
12
All stones = 1 ct All stones = 0.1 ct All stones = 0.1 ct

12

ct m-3

All stones = 0.01 ct

ct m-3

All stones = 0.01 ct

st m-3
2
All stones = 1 ct

120

st m-3
2
All stones = 1 ct

120

c
All stones = 0.1 ct

d
All stones = 0.1 ct

ct m-3

All stones = 0.01 ct

ct m-3

All stones = 0.01 ct

20

20

st m-3
P1 P3 P4 P5A P5B P6 P7 F1

st m-3
F3 UFA

characteristics, with Koala P4 and P5, and Fox UFA representing low grade mining units. Also of note is the vertical change in abundance of stones seen within the Koala P1 mega-graded bed, where the high concentration of stones observed towards the base decreases upwards, as does the overall grain size of the bed. Similar variations are seen for carats per cubic metre (Fig. 4) and average stone size (Figs. 4, 7b and 8b). Modelling of the average size data (ct/st) has been carried out firstly on the data set as a whole, to illustrate a non-geologically controlled model, and then by individual lithofacies using sharp boundaries between lithofacies (Figs. 7 and 8). Large-scale variations in the data can be seen after crude 3D gridding using discrete smooth interpolation (DSI, Mallet 1992; Figures 7c and 8c). DSI is a computer-based interpolation method used as a general tool to get an approximation of values over a grid (Mallet 1989). Within the corresponding 3D grids (Figs. 7c and 8c), one can identify low grade areas which coincide with lithological units such as the Fox UFA and Koala P5.

Geostatistical analysis of the data shows limited spatial relationships between data points. Continuity analysis of the grade data is made more complicated by the relatively small numbers of widely spaced data points. Short-range continuity is especially impacted by the widely spaced data. The low abundance of samples from Koala, especially in some of the less volumetrically important units such as P4 and P7, made semi-variogram modelling challenging. The data were subjected to kriging using a variety of search ellipsoids to explore for spatial anisotropy. The influence of sample spacing, widely spaced drillholes with more evenly spaced samples down the hole, greatly influences the observed anisotropy. In general, kimberlite grade samples tend to be weakly correlated and can be modelled using spherical variograms, high nugget effects and a range of a few tens of metres (Rombouts 2003), and here, good models were produced using both 50 and 100 m isotropic search ellipsoids (Figs. 7 and 8). Kriging prevents interpolation beyond the dataset, whereas DSI allows interpolation to the limits of the grid used, however large or small that is.

Bull V olcanol Fig. 6 3D models of both pipes showing the internal geology and location of sample points. Sample points are coloured from black (low), through blue and red, to white (high) based on their measured stones per cubic metre; values are not shown due to the proprietary nature of the data, and high-grade Koala is not equal in grade to high-grade Fox. The horizontal layering of the geology is reflected by the stone abundance, with obvious low abundances present in the Koala P5a and Fox UFA

KOALA

FOX

P1 P3 P4 P5a P5b P6

UFA

F3

F1

P7

Low
Regions for kriging were defined by the geological model allowing clipping to the pipe walls (Figs. 7 and 8). More detailed interpretation of the data would be performed during resource estimation (see Dyck et al. 2004; Harrison et al. 2009); however, for the purposes of this contribution which is to enable visualisation of data trends, this is a sufficient representation. Presentation of the data in three dimensions allows visualisation of vertical and lateral trends, which, when combined with detailed logging can be used to better resolve the geological influences on diamond distribution. Stratigraphic logging of four drill holes which intersect the multiply bedded, upward-fining Koala P3 (Fig. 9), reveals a coarser-grained, thicker-bedded intersection in the central pipe (KDC-41 and 71) whereas the intersections towards the eastern and western sides of the pipe (KDC-65 and 66) show thinner, finer-grained beds with more abundant mud-rich tops. The DSI model of Koala reveals a band across the pipe (trending roughly N S) located at the level of P3, where average diamond size is coarser than in the east and western regions (Fig. 9c, d), which roughly coincides with the observed coarser part of the unit. However, when using the kriged model, which incorporates sharp lithological boundaries (Fig. 9e), the reverse is observed whereby the central NS zone is a zone of reduced average diamond size compared to the eastern and western regions. By treating the dataset globally (here with DSI), different results are obtained than by treating the data lithology by lithology (here using kriging).

High Low

High

Mantle mineralsa good proxy for diamond grade? Kimberlite magmas transport xenocrysts and xenolithic fragments from the mantle to the surface, sampling the different levels of the mantle as they pass through. Mantle minerals such as garnet and chrome diopside are commonly used in diamond exploration as kimberlite indicator minerals, with their presence indicating that mantle material has been sampled and, if the chemistry of the minerals is favourable, that diamonds are likely to be present (e.g. Gurney 1989; Gurney et al. 1993; Nowicki et al. 2007). Olivine is also entrained from the mantle and is ubiquitous in kimberlites whether they are diamondiferous or not. The weight and abundance of indicator minerals at Koala correlates broadly with diamond grade; however, variations in mineral geochemistry do not appear to relate to diamond grade variations (Nowicki and Gurney 1998). The densities of many of these mantle minerals are sufficiently similar to those of diamonds so that volcanological and sedimentary processes leading to concentration of diamonds are also likely to affect the distribution of olivine, chrome diopside and garnet in a similar manner. Components with lower densities such as country rock material or finer grain sizes such as kimberlite ash will also be sorted with hydraulic equivalence being more important than density or size alone. Indicator minerals, though of a higher abundance than diamonds, represent only a small proportion of the pipe infill (often <1%) and may not be visually abundant in drill core, polished slabs or thin sections. The process of

Bull V olcanol Fig. 7 3D models showing the average size distribution within Koala, with black representing low values and red representing high values. a Plan view of the Koala pipe showing the internal layers and the position of the cross section used to illustrate the data. b Distribution of samples showing original data values prior to modelling. c Simple 3D DSI of global data, extrapolates data across a bounding box and edge effects can be seen. d 3D blocks formed from data bounding box and geological units. e Kriged data using a 50-m isotropic search ellipsoid, samples restricted to the geological blocks shown in (d). f Kriged data using a 100-m isotropic search ellipsoid, samples restricted to the geological blocks shown in (d)

collecting indicator mineral data is similar in expense and effort to collecting diamond data itself, requiring both rock crushing and heavy mineral separation. Macrocrystic olivine, however, often comprises the dominant mineral phase in kimberlites and is generally easily distinguished from other components in drill core and has been proposed as a tool for predicting diamond grade (Field et al. 2009; Scott Smith and Smith 2009). It is apparent from the distribution of diamond data between the different lithological units discussed above that there is a general

relationship between olivine abundance and diamond grade with low grade units such as Fox UFA and Koala P5 also containing relatively low olivine abundances when compared to the other units in the pipes (Table 2, Figs. 10 and 11). By comparing olivine grain size and abundance from logged diamond drill cores, with carats per cubic metre and average size of diamonds from nearby RC holes in both Koala and Fox (Figs. 10 and 11) a similar correspondence can be seen where diamond grade increases with increasing

Bull V olcanol

D
Fig. 8 3D models showing the average size distribution within Fox, with black representing low values and red representing high values. a Plan view of the Koala pipe showing the internal layers and the position of the cross section used to illustrate the data. b Distribution of samples showing original data values prior to modelling. c Simple 3D DSI of global data, extrapolates data across a bounding box and

edge effects can be seen. d 3D blocks formed from data bounding box and geological units. e Kriged data using a 50-m isotropic search ellipsoid, samples restricted to the geological blocks shown in (d). f Kriged data using a 100-m isotropic search ellipsoid, samples restricted to the geological blocks shown in (d)

olivine abundance and size. Diamond abundance (stone per cubic metre) also follows similar trends; however, the correspondence is less pronounced than for carats per cubic metre. In particular, the upward fining of the Koala P1 mega-graded bed is reflected by both the grain size and abundance of olivine, and the grade and average size of diamonds (Fig. 10). Unfortunately, the resolution of the diamond data is not as detailed as the olivine data (obtained from continuous visual estimations during detailed core logging); several boundaries are crossed by the grade samples, and distinct olivine rich horizons are incorporated in the samples where the background olivine is low in abundance. Also, when assessing the trends seen in the Koala example (Fig. 10), it

should be noted that the spacing between the logged drill core and the RC hole is 13 m, and lateral variations in the diamond/olivine abundance should also be considered. Ideally, the pilot drill holes for the RC sampling should be studied, as in the Fox example (Fig. 11); however, these were not available for Koala. This study shows that there is a geological influence on diamond distribution at Fox and Koala and that there is a positive relationship between olivine and diamond abundance. However, it also highlights the limitations of comparing grade with lithology due to the differing scales of observation and the diamond sampling techniques, which require a large amount of material to obtain a representative grade. The drill hole sample interval length

Bull V olcanol Fig. 9 Plots illustrating the average size distribution in Koala P3. a Plan view of the pipe showing location of diamond drill holes; b portions of stratigraphic logs with P3 highlighted in green, grain size scale is fine ash <0.0625 mm, ash 0.06252 mm; lapilli 264 mm, vertical scale marks represent 20 m; c horizontal slice through the DSI model from the middle of P3 showing a NS oriented band across the centre of the pipe with high average diamond size; d horizontal slice through the DSI model from the top of P3 showing the vertical continuity of the coarse diamond band. Average size ranges from low (black to blue) to high (red and white); e kriged data showing location of the drill holes and slices of the model in 3D

KDC-66 KDC-71

N
KDC-41 KDC-65 50 m

KDC-41

a)
KDC-66 KDC-71 P1 KDC-65

P3

b)
fine ash ash lapilli

P4

c)

d)
Low High

e)

may exceed the thickness of an individual lithology, and a diamond assay sample may thus sample more than one lithological unit. The relationship of olivine size and abundance to diamond content within the coherent kimberlite at Koala

has not been investigated due to the limited amount of grade data available. However, it remains unclear whether the macrocrystic olivine distribution within coherent rocks is influenced by large-scale sorting processes. The observed relationship between abundant coarse olivine grains and

Bull V olcanol
400
0 ct m-3 10 ct m-3

350

P1
300

250

P3 P4
elevation (masl)

200

P5a

150

100

P5b P6

50
fine ash ash lapilli

0 ct/st 0% 0 mm 80% 4 mm

0.25 ct/st

Abundance olivine (%)

Avg size olivine (mm)

Total diamond weight (ct m-3)

Avg diamond weight (ct/st)

Fig. 10 Plot illustrating the relationship of olivine to diamond seen at Koala. Stratigraphic log and olivine data collected from KDC-41 and diamond data from a large diameter RC-hole collared 13 m away from KDC-41 (K-17). Olivine abundance and average size are visual estimates based on detailed core logging. The rough correlation of

olivine size and abundance with carats per cubic metre can be seen, average size of diamonds does not correlate as well but a similar pattern is observed. Xenoliths are shown on the stratigraphic log, granodiorite (red), mudstone (black), sandstone (yellow) and siltstone (grey)

diamonds is only considered to be relevant in fragmental deposits. However, in cases where multiple intrusions occur, the olivine content may be useful in distinguishing between magma batches carrying different diamond loads (Clement et al. 1986; Field et al. 2009).

Discussion Lithological units are mirrored by grade units in both the Fox and Koala pipes, and the grain size, olivine content and sorting, and abundance of crustal xenolithic material of the individual units can be used as an indication of the relative grade differences between genetically related lithologies (such as Koala P5 and P6). The data presented above support the notion that volcanic and sedimentary processes occurring during kimberlite pipe infilling influence the distribution of diamonds within the subsequent deposits. The absolute diamond content of the erupting magma is impossible to determine from the fragmental deposits found

within kimberlite pipes. Diamond grade may decrease with depth within a single coherent intrusion (e.g. the Kimberley and De Beers pipes, Clement 1982; Clement et al. 1986) and also vary between early to late kimberlite intrusions as observed at the Helam Mine, South Africa (Swartruggens dyke swarm, Mc Kenna et al. 2004). An increase in grade between early and late cross-cutting fragmental phases is described at the Victor pipe, Canada (Webb et al. 2004; van Straaten et al. 2008). Individual pulses of magma contributing to an eruption may contain varying diamond grades, and though spatially and temporally related pipes may constitute economic clusters (Creaser et al. 2004), grade variations between pipes are common. Therefore, whether a kimberlite is diamondiferous or not and to what extent is not determined by the eruption or deposition processes but by the route of the magma from the mantle to the surface. The same eruptive and depositional processes can occur in any kimberlite eruption and will concentrate the heavy minerals in the same way, regardless of the presence or absence of diamonds. Surface processes, however, have the potential to concentrate diamonds

Bull V olcanol
400
0 ct m-3 2.5 ct m-3

UFA
300

F3
200

elevation (masl)

100

F1

0 ct/st

0.3 ct/st

Abundance olivine (%) Avg size olivine (mm)


-100
fine ash ash lapilli

Total diamond weight (ct m-3) Avg diamond weight (ct/st)


0% 0 mm 70% 2.5 mm

-200

Fig. 11 Plot illustrating the relationship of olivine to diamond seen at Fox. Stratigraphic log and olivine data collected from FDC-22 drilled as a pilot hole to a large diameter RC hole from which the diamond data is taken (F-32). Olivine abundance and average size are visual estimates based on detailed core logging. The rough correlation of olivine size and abundance with carats per cubic metre can be seen, average size of diamonds does not correlate as well but a similar pattern of low size in the UFA, elevated size in F3 and then slightly

reduced again in F1 is observed. The distinction between the UFA and F1 and F3 is obvious in all plots. An increased abundance of granodiorite (red), dolerite (blue) and mantle xenoliths (green) with depth is shown on the stratigraphic log and this increased dilution may account for the apparent decrease in diamond size and grade in the F1 with depth, though grade sampling did not cover the full depth of FDC-22

from sub-economic magmas into economic abundances within the pipes. Therefore, simply identifying a pyroclastic flow deposit, for instance, will not enable prediction of the grade of that unit without first having determined the presence of diamonds. Sorting of diamonds during sedimentary transport and deposition is well known from detrital diamond placer deposits, where diamonds generally decrease in size away from the source and are often found concentrated in bedrock traps (Gurney 1989). Sedimentary transport also affects the quality of the diamonds due to preferential breakage of low-quality diamonds (boart) and diamonds with flaws or inclusions due to attrition (Gurney 1989;

Censier and Tourenq 1995). However, the influences of volcanic processes on diamond size distribution in primary and resedimented kimberlitic volcaniclastic deposits have not often been documented. The liberation of diamonds from the host magma/ xenoliths will influence the physical sorting process. If they are not liberated, then the size/density of the enclosing host clast will be the important factor in sorting. Understanding the location of diamonds either as free crystals or inclusions in larger clasts is therefore important. Unfortunately, whether they are contained within larger clasts or not is difficult to assess due to the scarcity of diamonds seen in drill core.

Bull V olcanol

The degree of physical sorting during eruption, transport and deposition can lead to an increase in relative diamond grade by elutriation/removal of fine material and concentration of the coarser denser components. In particular, the free crystal distribution of a pyroclastic deposit, which is likely to have more uniform initial size and density properties when compared to lithic fragments and juvenile clasts, can be used to interpret the transport system and emplacement process (Taddeucci and Palladino 2002). In kimberlites, the main crystal phase is olivine; however, olivine is dominantly xenocrystic and the initial crystal size distribution in the erupting kimberlite magma is likely to be more variable (Field et al. 2009; Jerram et al. 2009). Enhancement of the grade Nothing is known about the initial diamond (or olivine) distribution in the erupting magma at either Fox or Koala, though they were likely to have been different. At Koala, the late-stage intrusion (P7) suggests early pulses of magma may have carried different grades (e.g. Mc Kenna et al. 2004). Assuming the P6 erupting magma carried similar grades to P7 (~2 ct m3), then significant enhancement of the grade (see Table 2) has occurred during the formation of P6 (~4 ct m3). Magma fragmentation during explosive eruptions generates fine ash. Removal of ash by elutriation within an eruption column or during fluidisation would act to enhance the diamond grade in the subsequent deposit relative to that of the erupting magma. For Koala, P6 removal of 50% of the ash (original fluid component of the magma) would be required to produce the observed doubling of the diamond grade. Geochemical studies of pyroclastic kimberlites at Ekati, including the Koala P6, by Nowicki et al. (2008) suggest that up to 75% of the groundmass component may actually be removed by elutriation of ash. Grain size sorting also influences the diamond distribution, though not necessarily the bulk grade of the deposit. The Koala P1 lithofacies shows a distinct grain size sorting with coarse lapilli sized material, dominated by olivine, at the base grading upwards to medium to fine ash at the top of the 100-m thick bed (Fig. 11). The diamond distribution also follows a similar trend with more abundant and larger diamonds at the base grading upwards to low-grade material at the top (Figs. 4, 7 and 11). Comparable olivine grain-size grading is observed in the P3 lithofacies, though the graded beds are much smaller and of a scale that diamond distribution cannot be determined. A similar distribution of diamonds within a mega-graded bed has also been described at the 140/141 kimberlite in the Fort la Corn area (Berryman et al. 2004). Processes such as large-scale sorting and removal of fine material from the system (e.g. elutriation in an eruption

column or during fluidisation) causes grade enhancement (Fig. 12); smaller-scale sorting (e.g. in density currents producing graded beds such as P3 and P1) also enhances grade in the coarser portions of the deposit and also effectively reduces the grade in the finer portions of the deposit. Depending on the scale of the graded beds related to mining method (typical open pit bench heights at Ekati are 10 m) and level of detailed geology, this sorting of diamonds within the deposit may be able to be exploited during mining. Resedimentation of material from the crater rim back into the vent may lead to further grain-size sorting into discrete coarse and fine horizons if sediment gravity flows rather than slumping occurs. This may not influence the overall grade of a mineable unit as the fine grain sizes will not be removed from the material as it is deposited within the vent but simply segregated from the coarser grain-sizes, as the vent acts like a closed system. Rapid, large-scale debris flows, such as the Koala P5, are not likely to improve the overall sorting, as the matrix sediment has some degree of cohesion, preventing settling of larger particles (Boggs 2006). Lowering the grade A similar deposition mechanism is inferred for F1 as for P6; however, the diamond grades are very different. The original grades of the erupting magmas are likely to have been different to each other; however, the influence of dilution by xenolithic crustal material on diamond grade is apparent. P6 contains <10% country rock lithic fragments whereas F1 contains >40% (Table 1), dilution of P6 by 40% would result in lowering of the grade to 2.55 ct m3. Likewise, if we calculate the grade of the F1 deposit without the 40% dilution, a grade of 1.05 ct m3 is obtained. Dilution clearly adversely influences the diamond grade of any deposit. Transient events leading to ventwidening during eruption can increase the abundance of lithic fragments (Cas and Wright 1987). Deposits resulting from early-stage column collapse (e.g. Fox F1, Porritt et al. 2008a; Porritt and Cas 2009), occurring before vent clearing was complete are likely to contain higher abundances of country rock lithics than column collapse deposits from the waning phase of an eruption (e.g. Koala P6, Porritt 2008; Porritt and Cas 2011). The affects of dilution can also be seen when comparing the olivine content and diamond grade of the lowermost fragmental deposits in the Koala pipe, P6 and P5a (Tables 1, 2). These deposits are genetically related in that they are the pyroclastic and volcaniclastic (resedimented from the tuff rim) deposits of the Koala eruption. Theoretically, they should have comparable olivine and diamond contents if they are the products of the same erupting magma. However, both the

Bull V olcanol

Fig. 12 A simplified flow diagram illustrating the sequence of volcanological and sedimentary processes which can influence the grain-size sorting of kimberlitic fragmental deposits. Grain size sorting of deposits can both upgrade diamond content, in the coarse grain sizes, or decrease diamond content where sorting concentrates the finer, lighter material e.g. Koala P1. The affects of dilution may be

obvious, in the form of abundant host rock lithics as seen in the Fox F1 or subtle where fine mud or kimberlitic ash may be incorporated into the unit. Each lithofacies should be assessed in terms of transport and deposition processes and how those processes may have led to concentration of diamonds in particular horizons or areas within the unit

olivine content and grade of P5a (1020% olivine, 0.77 ct m3) are significantly lower than in P6 (2565% olivine, 4.25 ct m3). Resedimentation of the crater rim material may result in better internal sorting of the deposits, as mentioned above, but may also result in the incorporation of non-kimberlitic material from the crater

walls into the deposits, as seen in the Koala P5a unit, thus diluting the grade. The instability of vent walls during, and immediately prior to, the cessation of the eruption may also lead to incorporation of country rock material into the pipe, creating boulder breccias dominated by non-kimberlitic material (e.g. Koala

Bull V olcanol

P5b, Nowicki et al. 2004; Crawford et al. 2006; Porritt and Cas 2011), and the distribution of large boulders within such a unit influences the distribution of diamond grade. P5b is slightly anomalous in that although it is a boulder breccia, dominated by large granodiorite boulders; it returns a higher grade than the overlying P5a lithofacies. This is most likely due to the presence of some P6 material in the breccia matrix, the small sample population, and also the difficulties in defining the boundaries of this irregular zone during RC drilling and grade sampling. The dilution factor for any kimberlite may be hard to quantify due to the physical nature of the rocks. A visual estimation of the abundance of country rock lithic fragments has been used in this study, supplemented by point counting of clasts >2 mm. In some instances, the non-kimberlitic material is not readily apparent, either due to fine grain sizes, alteration overprints or incorporation of disaggregated material into the matrix. In these cases, whole rock geochemical analysis may be used to quantify the degree of dilution (Nowicki et al. 2008; Porritt and Cas 2009). The varieties of eruption and depositional processes that can have an influence on sorting/diluting the kimberlite material are shown in Fig. 12. Detailed geological assessment of the lithofacies within each pipe and their genetic relationships to each other lead to a better understanding of the grade distribution and can help to refine grade models.

Conclusions At both the Fox and Koala kimberlite pipes, the individual lithological units are mirrored by the grade units, indicating that geological processes are important in distributing diamonds both between and within the individual units. A positive relationship between diamond grade and macrocrystic olivine size and abundance is seen, showing that sorting processes are involved in the distribution of these minerals throughout fragmental deposits. There is also a negative relationship between diamond grade and dilution (commonly in the form of granodiorite clast abundance) which is especially obvious when comparing overall grade between the P6 and F1 pyroclastic deposits. The overall grade of F1 is significantly lower than that in P6; however, when recalculated to remove the effects of significant dilution by crustal xenolithic debris, the grades become more comparable. When predicting the grade potential of different lithofacies, those with abundant coarse olivine grains, few lithic fragments and low proportions of fine-grained material are expected to yield the highest diamond grades. The influence of eruption, transport, deposition and resedimentation processes on the distribution of diamonds within kimberlite pipes has been poorly documented, in

part due to the limited understanding of the geological processes, which have resulted in the deposits, and also in part due to the proprietary nature of diamond data. Without knowing the original diamond content of the erupting magma, it is impossible to quantify the relative increase or decrease of diamond grade within any unit due to the geological processes occurring at the surface. However, based on an understanding of the grain size sorting, dilution and olivine abundance, we can predict which lithofacies are likely to carry the significant diamond grade and possibly where within these lithofacies higher grade zones might be located. Processes such as elutriation in the eruption column, through large-scale fluidisation of deposits and within pyroclastic flows all help to upgrade the diamond content by removing fine ash from the final deposit. Further sorting of material could occur if the vent was filled with water at the time of deposition; however, the fines are not removed from the system altogether. Facies developed in this manner may have concentrated layers of coarse diamond bearing material, interbedded with, or overlain by, the finer material segregated by the water column. In these cases, concentration of diamonds may have occurred on a scale too fine to be exploited by conventional bulk mining methods, but which might be recovered by more selective methods such as free digging. Recognising depositional units from a different source vent, which most likely contain different diamond types as well as abundances, is crucial in diamond grade modelling where the value of the diamonds, not just the abundance, becomes important. A good understanding of the eruption and deposition processes may lead to recognition of areas for selective mining perhaps enhancing a sub-economic deposit.
Acknowledgements This research comprises part of a PhD thesis by LAP on the volcanology of the Ekati kimberlites. The authors would like to thank BHP Billiton Canada Inc. for their financial support for this research project. In particular, we would like to thank Jon Carlson and the geologists and other staff members for their continued support and valuable knowledge. We would like to thank Tom Nowicki, Pat Hayman and Stephen Moss for their invaluable insight and intellectual discussions. LAP would also like to acknowledge the current financial support of CBIE. This manuscript has benefitted greatly from reviews by P . Kokelaar, one anonymous reviewer, and the editor RSJ Sparks.

References
Berryman AK, Scott Smith BH, Jellicoe BC (2004) Geology and diamond distribution of the 140/141 kimberlite, Fort a la Corne, central Saskatchewan, Canada. Lithos 76(14):99114 Boggs S (2006) Principles of sedimentology and stratigraphy. Pearson Prentice Hall, Upper Saddle River, p 662 Brown RJ, Field M, Gernon T, Gilbertson MA, Sparks RSJ (2008) Problems with an in-vent column collapse model for the emplacement of massive volcaniclastic kimberlite. J V olcanol Geotherm Res 178(4):847850

Bull V olcanol Cas RAF, Wright JV (1987) V olcanic Successions. Chapman and Hall, New York, p 528 Censier C, Tourenq J (1995) Crystal forms and surface textures of alluvial diamonds from the western region of the CentralAfrican-Republic. Mineral Deposita 30(34):314322 Clement CR (1982) A comparative geological study of some major kimberlite pipes in the Northern Cape and Orange Free State. University of Cape Town, Cape Town, p 2 vols Clement CR, Harris JW, Robinson DN, Hawthorne JB (1986) The De Beers kimberlite pipe - a historic South African diamond mine. In: Anhaeusser CR, Maske S (eds) Mineral Deposits of Southern Africa. Geological Society of South Africa, Johannesburg, pp 21932214 Crawford B, Porritt LA, Nowicki TE et al (2006) Key geological characteristics of the Koala kimberlite, Ekati Diamond Mine, Canada. In: 2006 Kimberlite Emplacement Workshop. Long Abstracts, Saskatoon, Canada Creaser RA, Grutter H, Carlson J, Crawford B (2004) Macrocrystal phlogopite Rb-Sr dates for the Ekati property kimberlites, Slave Province, Canada: evidence for multiple intrusive episodes in the Paleocene and Eocene. Lithos 76(14):399414 Davis JC (1986) Statistics and data analysis in geology, 2nd edn. Wiley, Chichester, p 646 de Wit MCJ (1999) Post-Gondwana drainage and the development of diamond placers in western South Africa. Econ Geol 94 (5):721740 Dyck DR, Oshust PA, Carlson JA, Nowicki TE, Mullins MP (2004) Effective resource estimates for primary diamond deposits from the EKA TI Diamond Mine(TM), Canada. Lithos 76(14):317335 Fedortchouk Y , Canil D, Carlson JA (2005) Dissolution forms in Lac de Gras diamonds and their relationship to the temperature and redox state of kimberlite magma. Contrib Mineral Petrol 150 (1):5469 Field M, Gernon T, Mock A, Walters A, Sparks RSJ, Jerram DA (2009) V ariations of olivine abundance and grain size in the snap lake kimberlite intrusion, Northwest Territories, Canada: a possible proxy for diamonds. Lithos 112(1):2335, 9th International Kimberlite Conference Proceedings Gernon T, Fontana G, Field M, Sparks RSJ, Brown RJ, Mac Niocaill C (2009a) Pyroclastic flow deposits from a kimberlite eruption: the Orapa South Crater, Botswana. Lithos 112(Supplement 1):566578 Gernon T, Gilbertson MA, Sparks RSJ, Field M (2009b) The role of gas-fluidisation in the formation of massive volcaniclastic kimberlite. Lithos 112(Supplement 1):439451 Gurney JJ (1989) Diamonds. In: Ross J, Jaques AL, Ferguson J, Green DH, Reilly SY O, Danchin RV , Janse AJA (eds) Kimberlites and Related Rocks, Proceedings of the Fourth International Kimberlite Conference, Perth 1986, Geological Society of Australia Special Publication No. 14. Blackwell, Oxford, pp 935965 Gurney JJ, Helmstaedt HH, Moore RO (1993) A review of the use and application of mantle mineral geochemistry in diamond exploration. Pure and Applied Chemistry 65:24232442 Gurney JJ, Helmstaedt HH, Le Roex AP et al (2005) Diamonds: Crustal Distribution and Formation Processes in Time and Space and an Integrated Deposit Model. Econ Geol 100th Anniversary V olume:143177 Harris JW (1992) Diamond geology. In: Field JE (ed) The Properties of Natural and Synthetic Diamonds. Academic, London, pp 345393 Harrison S, Leuangthong O, Crawford BB, Oshust PA (2009) Uncertainty-based grade modelling of kimberlite: a case study of the Jay kimberlite pipe, EKA TI Diamond Mine, Canada. Lithos 112(Supplement 1):7382 Holden E, Moss S, Russell JK, Dentith M (2009) An image analysis method to determine crystal size distributions of olivine in kimberlite. Comput Geosci 13:255268 Jerram DA, Mock A, Davis GR, Field M, Brown RJ (2009) 3D Crystal size distributions: a case study on quantifying olivine populations in kimberlites. Lithos 112(Supplement 1):223235 Kjarsgaard BA, Wilkinson L, Armstrong JP (2002) Geology, Lac de Gras kimberlite field, Central Slave Province, Northwest TerritoriesNunavut. In: Open File 3228. Geological Survey of Canada, Lorenz V , Zimanowski B, Buettner R (1999) Discussion on the formation of kimberlite pipes: the phreatomagmatic model. CEV Newsletter, IA VCEI:1117 Mallet JL (1989) Discrete smooth interpolation. ACM Trans Graph 8 (2):121144 Mallet JL (1992) Discrete smooth interpolation in geometric modelling. J Comp Aid Des 24(4):177191 Mc Kenna N, Gurney JJ, Klump J, Davidson JM (2004) Aspects of diamond mineralisation and distribution at the Helam Mine, South Africa. Lithos 77(14):193208 Meyer HOA (1987) Inclusions in diamond. In: Nixon PH (ed) Mantle Xenoliths. Wiley, Chichester, pp 501523 Moss S, Russell JK, Andrews GDM (2008) Progressive infilling of a kimberlite pipe at Diavik, Northwest Territories, Canada: insights from volcanic facies architecture, textures, and granulometry. J V olcanol Geotherm Res 174(13):103116 Moss S, Russell JK, Scott Smith BH, Brett RC (2010) Olivine crystal size distribution in kimberlite. Am Mineral 95:527536 Nassichuk WW, Dyck DR (1998) Fossils recovered from Kimberlite Pipes in the Lac de Gras Field, Slave Province, Northwest Canada; Geological Implications. In: Extended Abstracts 7th International Kimberlite Conference. Cape Town, pp 612614 Nowicki TE, Gurney JJ (1998) The relationship between indicator minerals and diamond grade in samples from the Panda and Koala kimberlites: a test study. In: Mineral Services (Pty) Ltd, unpublished company report, p 10 Nowicki T, Crawford B, Dyck D, Carlson J, McElroy R, Oshust P , Helmstaedt H (2004) The geology of kimberlite pipes of the Ekati property, Northwest Territories, Canada. Lithos 76(14):127 Nowicki TE, Moore RO, Gurney JJ et al (2007) Chapter 46 Diamonds and Associated Heavy Minerals in Kimberlite: A Review of Key Concepts and Applications. Developments in Sedimentology. Elsevier, Amsterdam, p 1235 Nowicki T, Porritt L, Crawford B, Kjarsgaard B (2008) Geochemical trends in kimberlites of the Ekati property, Northwest Territories, Canada: insights on volcanic and resedimentation processes. J V olcanol Geotherm Res 174(13):117127 Porritt LA (2008) The volcanology and sedimentology of the Ekati kimberlites, NWT, Canada, with consideration of the implications for diamond grade. In: School of Geosciences. Monash University, Clayton, Australia Porritt LA, Cas RAF (2009) Reconstruction of a kimberlite eruption, using an integrated volcanological, geochemical and numerical approach: a case study of the Fox Kimberlite, NWT, Canada. J V olcanol Geotherm Res 179(34):241264 Porritt LA, Cas RAF (2011) The influence of complex intra- and extra-vent processes on the deposit characteristics of the Koala Kimberlite, NWT, Canada: volcanology, sedimentology and root zone processes. Bull V olcanol. doi:10.1007/s00445-011-0452-5 Porritt LA, Cas RAF, Crawford BB (2008a) In-vent column collapse as an alternative model for massive volcaniclastic kimberlite emplacement: an example from the Fox Kimberlite, Ekati Diamond Mine, NWT, Canada. J V olcanol Geotherm Res 174 (13):90102 Porritt LA, Cas RAF, Crawford BB (2008b) Reply to: Discussion by Brown et al. on In-vent column collapse as an alternative model for massive volcaniclastic kimberlite emplacement: An example

Bull V olcanol from the Fox kimberlite, Ekati Diamond Mine, NWT, Canada. J V olcanol Geotherm Res 178(4):851854 Rombouts L (1995) Sampling and statistical evaluation of diamond deposits. J Geochem Explor 53(13):351367 Rombouts L (1999) Extreme value analysis of diamond sizes and values. In: Gurney JJ, Gurney JL, Pascoe MD, Richardson SH (eds) Proceedings of the VIIth International Kimberlite Conference. Red Roof Design, Cape Town, pp 721727 Rombouts L (2003) Assessing the diamond potential of kimberlites from discovery to evaluation bulk sampling. Mineral Deposita 38 (4):496504 Scott Smith BH, Smith SCS (2009) The economic implications of kimberlite emplacement. Lithos 112(Supplement 1 (9th International Kimberlite Conference Proceedings)):1022 Sparks RSJ, Baker L, Brown RJ, Field M, Schumacher J, Stripp G, Walters A (2006) Dynamical constraints on kimberlite volcanism. JV olcanol Geotherm Res 155(12):18 Stiefenhofer J, Farrow DJ (2004) Geology of the Mwadui kimberlite, Shinyanga district, Tanzania. Lithos 76(14):139160 Svisero DP (1995) Distribution and origin of diamonds in Brazil: an overview. J Geodyn 20(4):493514 Taddeucci J, Palladino DM (2002) Particle size-density relationships in pyroclastic deposits: inferences for emplacement processes. Bull V olcanol 64(34):273284 Terry RD, Chilingar GV (1955) Summary of Concerning some additional aids in studying sedimentary formations by M.S. Shretsor. J Sediment Petrol 25:229234 van Straaten BI, Kopylova MG, Russell JK, Webb KJ, Scott Smith BH (2008) Discrimination of diamond resource and non-resource domains in the Victor North pyroclastic kimberlite, Canada. J V olcanol Geotherm Res 174(13):128138 Webb KJ, Scott Smith BH, Paul JL, Hetman CM (2004) Geology of the Victor Kimberlite, Attawapiskat, Northern Ontario, Canada: cross-cutting and nested craters. Lithos 76(14):2950

Potrebbero piacerti anche