Evaluation of Coarse Gold-Bearing Conglomerate Mineralisation at Beatons Creek, Pilbara, Western Australia: Sampling for Resource Development and Grade Control

Abstract

Many styles of gold mineralisation are challenging to sample because of the presence of coarse gold and a high spatial heterogeneity. The coarse gold-bearing conglomerates of the Beatons Creek deposit provide some challenges related to the presence of gold particles up to 8 mm and gold particle clusters (up to 500 mm³) at low in situ grades (<2 g/t Au). Novo has attempted to address these issues over the last six years of exploration, resource development and mining. The Beatons Creek open pit operation was the first Pilbara conglomerate-hosted gold deposit to go into production. Between January 2021 and September 2022, it produced 2.5 Mt at 1.2 g/t Au for 87,300 oz Au recovered. Gold is present within a matrix of multiple, narrow-stacked oxide and fresh (sulphide) conglomeritic reef horizons, which are interbedded with unmineralised conglomerate, sandstones and grits. It is strongly associated with detrital pyrite and authigenic nodules. Several sampling techniques have been applied across the project, including diamond core and RC drilling, trench channel sampling and bulk sampling. Assay methods applied include fire assay, screen fire assay, LeachWELL™ and more recently PhotonAssay™. The dominant sampling protocol applied for resource development and grade control utilised 0.5 m length RC samples; a 50% rig split (c. 8.5 kg) and laboratory crushing to 3 mm, followed by a 2.5 kg split and total assay via PhotonAssay™. For part of the 2022 RC programme, the detectORE™ technique was used to screen primary RC samples and reduce the feed to the laboratory. Novo operated a sampling and assay programme that aimed to reduce the impact of coarse gold on sample and assay preparation biases and to improve estimation.

1. Introduction

1.1. Background

The importance of high-quality sampling throughout the gold mine value chain has been stressed by many authors [1,2,3,4,5,6]. The sampling process, inclusive of sample collection, preparation and assaying or testing, is a critical component to all stages of a mining project, as it forms the basis for Mineral Resource and Ore (or Mineral) Reserve estimates. Sampling includes in situ material and broken (or crushed) rock for both geological and metallurgical purposes. All resource and reserve reporting frameworks, such as the 2012 JORC Code (JORC, 2012) [7], require the Competent Person(s) to consider the quality and implications of sampling and assaying programmes. In all cases, the aim is to derive a representative sample to accurately describe the material in question. Field sample collection is followed by sample reduction, in both mass and fragment size, to provide an assay charge or test sample. This process is challenging for gold mineralisation, and particularly when coarse gold is present. Samples should be collected and prepared within the framework of the Theory of Sampling [1,2,3,4,5,6].

Mineralisation containing substantive quantities of coarse gold (nominally >15% above 100 µm) is often typified by a high geological nugget effect, which represents variations in (1) the in situ size distribution of gold particles (including the effects of gold particle clustering) and (2) gold particle abundance [8,9,10]. Where the sampling process is not optimised, the sampling nugget effect is enhanced, increasing the total nugget effect [10,11,12].

A key consequence of evaluating coarse gold-bearing high-nugget effect mineralisation is the challenge of applying a cut-off grade to support a selective mining approach. Even where grades are estimated using a kriging-based interpolator, block grades tend toward the mean grade and estimates have a high conditional bias, with a considerable risk of ore/waste misclassification. As a result, targeted selective mining can be difficult. In such a case, optimised sampling protocols are vital and must be supported by strong geological knowledge.

1.2. Challenges of Gold Sampling

There are several challenges that relate to the sampling of gold:

• Primary gold particle distribution is often erratic (high geological/in situ nugget effect), exacerbated by localised clustering [8];
• Critical grades, such the mineralisation indicator (MIG) and breakeven mining cut-off (BCOG) grades are low (<1 g/t Au), therefore coarse gold particles are rare ‘events’ displaying a Poisson distribution [13];
• Poor disintegration of gold particles during pulverisation [12,13,14,15];
• Extreme contrast between the densities of gold and gangue minerals [16,17].

These challenges lead to some key requirements for accurate gold sampling and assaying:

• Correct sampling procedures with no gold loss or contamination;
• Preparation procedures that recognise the impact of gold particle sizing and reduce the impact of segregation;
• Effective splitting of sub-samples;
• Assay method and sub-samples that provide acceptable accuracy.

These problems can be reduced, but not eliminated, by using larger samples and assay charges, in combination with careful procedures to minimise all sampling and sub-sampling errors. Where mineralisation is dominated by gold particles greater than 1 mm in size, particularly at low mining and/or cut-off grades, specialised protocols will be required [13,14,15,18,19,20].

1.3. Theory of Sampling

1.3.1. Introduction

Sampling errors are defined in the Theory of Sampling (TOS) as promulgated by the works of Dr Pierre M. Gy [1,2,5,6]. Uncontrolled sampling errors lead to an elevated nugget effect [10,11,12]. The fundamental sampling (FSE) and grouping and segregation (GSE) errors (collectively, the correct sampling errors: CSE) are irreducible random errors related to the inherent heterogeneity and characteristics of the material being sampled. They lead to poor precision and can only be minimised through good sampling protocols. The other errors (e.g., extraction error [EE]; delimitation error [PE]; preparation error [PE]; weighting error [WE]; analytical error [AE]), collectively, the incorrect sampling errors (ISE), arise because of the physical interaction between the material being sampled and the technology employed to extract the sample. They result in biases which may be reduced by the correct application of sampling methods and procedures.

A sample may be representative when it results in acceptable levels of bias and precision [1,2,3,4,5,6]. Precision refers to the degree of reproducibility among sample grades (e.g., field or laboratory duplicates). Accuracy describes how close a sample grade is to the actual, true grade. Bias is the difference between the expected sample grade and the actual, true grade. An unbiased sample is likely to be more accurate than a biased sample. The sampling process aims to be unbiased and strives to achieve precise and accurate results. The fundamental tenant is that a sample is correct (or representative) if each fragment in the lot has an equiprobable chance of being selected [1,2,3,4,5,6].

Representative sampling will lead to a maximised recovery of value from a mining operation, which is key to “Responsible Consumption and Production”, the United Nations Sustainable Development Goal #12. In addition, effective recovery is also key to the International Council on Mining and Metals Principal #8, “Responsible Production”.

1.3.2. FSE-Based Calculations Used in This Study

The FSE reflects grade heterogeneity manifested as the Constitution Heterogeneity (CH) within a lot of broken rock (e.g., mine stockpile, bag of RC cuttings or assay pulp) [1,2,5]. The FSE does not cancel out and remains even after a sampling operation is perfect. For any process where the FSE escalates, there is an associated loss due to uncertainty, which will ultimately drive ore/waste misclassification. Controlling the FSE during sampling and assaying programmes is therefore important. Heterogeneity tests [5,21,22,23,24,25] or integrated approaches [9,14,15,18,19,20] to characterisation form the basis of calibrating FSE equation inputs.

While the use of the FSE equation represents an idealised expectation that may or may not be attained in practice, it provides a starting point from which protocols can be evaluated and optimised [3,14,15,18,21,26]. Gy [1,2] stated that the FSE equation is a tool for order-of-magnitude prediction and that it only computes the error variance for an ideal model where all particles are sampled with equal probability.

The FSE can be modelled before the material is sampled, provided certain characteristics are determined or assumed [1,2,5]. The FSE is applicable to samples once they are collected (e.g., RC cuttings at the rig). The FSE equation (François-Bongarçon-modified) is given as Equation (1) [26]:

FSE_{(rel. var.)} = f g c dℓ^b d_N^α (1/M_S − 1/M_L)

(1)

where f = shape factor (ranges between 0.2 for flaky particles to 0.5 for spherical particles); g = granulometric factor (generally taken as 0.25); d_N = nominal 95% material passing size (in cm); c = mineralogical factor (for gold = density × 10⁶/grade in g/t); dℓ = liberation diameter (in cm) is the screen size that retains 5% of gold, given a theoretical lot of liberated gold [3,5,9,26]; b = (3 − α), where α is determined experimentally [5,21,22,23,24,26], or a value based on experience can be used [26]; M_S = sample mass (in grammes); and M_L = lot mass (in grammes). Further details on the inputs into the FSE equation are provided in various key references [1,2,5,6,26].

The FSE is calculated via Equation (1) as the relative variance and is usually presented as a precision value based on the relative standard deviation. All FSE calculations herein are reported at one standard deviation, e.g., at 68% reliability. Using this formula, it is possible to (a) calculate the error for a given sample size split from the original or (b) calculate what sub-sample size should be used to obtain a specified variance at a given reliability.

Intrinsic heterogeneity (IH_L) is the component of CH driven by differences in particle size and variation in composition between one particle and another. The greater the difference between the amount of target analyte in the particles, the greater the IH_L value. Key influences on the IH_L are the grade, mineralogy, shape and size distribution of the fragments in the lot.

IH_L is the key output from heterogeneity tests and comprises the following components of the FSE equation at a given d_N value (Equation (2)).

IH_L = f g c dℓ^b d_N^α

(2)

The estimated constant factor of CH, EST IH_L (given in grammes) can be calculated for a given size fraction according to Equation (3).

$$\mathrm{E}\mathrm{S}\mathrm{T} {\mathrm{I}\mathrm{H}}_{\mathrm{L}}=\mathrm{g}{\sum }_{\mathrm{q}}\frac{{\left({\mathrm{a}}_{\mathrm{q}}-{\mathrm{a}}_{\mathrm{Q}}\right)}^2}{{\mathrm{a}}_{\mathrm{Q}}^2}\frac{{\mathrm{M}}_{\mathrm{q}}^2}{{\mathrm{M}}_{\mathrm{Q}}}$$

(3)

where M_q (in grammes) is the mass of fragments by group; a_q (in g/t Au) is the grade of each fragment group; M_Q (given in grammes) is the total mass of all groups; and a_Q is the mean grade of all groups. The g is the granulometric factor, whose value is 0.25 for unscreened material and 0.55 for screened material.

On the determination of the IH_L via Equation (3), Equation (2) can be rearranged to solve for dℓ, giving Equation (4).

dℓ = (IH_L/f g c d_N^α)^1/b

(4)

1.4. Aim of This Paper

This paper presents the approach used to address key sampling characteristics at Beatons Creek, which include:

• Dominant (25%–50%) > 1 mm coarse gold particles and very coarse gold particles to 8 mm in size;
• In situ gold particle clusters up to 500 mm³ (c. 8 mm cube);
• Low critical grades, e.g., MIG of 0.5 g/t Au, BCOG of 0.8 g/t Au, and run-of-mine (ROM) grade of 1.5 g/t Au.

The sampling studies aimed to provide an optimised and practical approach to resource development and grade control drilling, sampling and assaying methods. The 2022 and 2023 Mineral Resource estimates (MRE) at Beatons Creek represent some of the first reported in the context of National Instrument 43-101 (NI 43-101) [27] using the PhotonAssay™ (PA) method [28,29]. The discussions have general implications and an impact on the sampling of other heterogeneous gold mineralisation styles.

2. Beatons Creek Gold Project

2.1. Introduction

The Beatons Creek gold project is located within the East Pilbara Shire, between the major regional centres of Newman and Port Hedland, Western Australia. The project is situated west of the town of Nullagine, which is 296 km southeast of Port Hedland and 170 km north of Newman. Modern evaluation did not commence until 1983, with various companies drilling up to 2007. Novo Resources Corp. (Novo) (Vancouver, BC, Canada) gained control of the project in 2015, continued exploration drilling through to 2018 and undertook a bulk sampling programme in 2018 [30,31]. An extensive grade control and resource development RC drilling programme was undertaken from October 2020 to December 2022. Sample preparation and assaying associated with the latter part of the 2022 RC drilling programme continued until May 2023. The open pit mine went into production in January 2021, terminating in September 2022, when the project went into care and maintenance. The operation produced 2.51 Mt at 1.17 g/t Au for 94,148 oz Au recovered [28,29].

An initial MRE was released by Novo in 2019 [31]. This was followed by the 2022 MRE, which comprised an Indicated Mineral Resource of 234,000 oz Au (3 Mt at 2.4 g/t Au) and an Inferred Mineral Resource of 42,000 oz Au (0.8 Mt at 1.6 g/t Au) reported within an optimised open pit shell and at a 0.5 g/t Au cut-off [28]. A final MRE (2023 MRE) was completed on 15 December 2023, although this has not been publicly reported [29]. On 20 December 2023, it was announced that the project had been sold as part of a larger disposal of tenements and infrastructure in the Eastern Pilbara.

2.2. Geology and Gold Mineralisation of Beatons Creek

The mineralisation consists of up to 2 m thick auriferous conglomerate reefs. Mapping and drilling have confirmed the Nullagine sub-basin subdivision of the Hardey Formation at Beatons Creek [32]. The Beatons Mineralised Unit and Beatons Middle Unit form a 200 m thick package comprising a monotonous sequence of pebble-to-boulder conglomerates, with occasional thin interbeds of sandstone [32]. Conglomerate clasts comprise sandstone, siltstone, quartz and dromedary boulder conglomerates. Dromedary boulders resemble the Dromedary Hills Mosquito Creek conglomerate unit towards the east [32]. Regular 0.5 m to 2 m thick horizons feature cobble-to-boulder conglomerates, with increased resistive clasts and increased pyrite, and represent fluvial channels (proximal to the depositional fan) or zones of marine reworking.

Gold-bearing conglomerates are restricted to channels (fluvial) and marine lags that are readily recognisable from the outcrop and drill core. They are constrained to the 40 m thick “Mineralised Unit” at the top of the sequence. Fluvial-type conglomerates and marine lags have a clearly defined top and base and represent a higher energy depositional environment conducive to concentrating gold, as well as detrital pyrite and resistive clasts (Figure 1 and Figure 2).

The fluvial and marine lag conglomerates are interstratified, indicating that the depositional facies in which they formed were laterally proximal. The depositional environment is interpreted to have been a river fan delta along a coastline. During periods of low stand, a braided river delta prograded seaward, depositing channelised fluvial-type conglomerates. As sea levels rose, wave action winnowed out fine, light sediment, leaving behind a transgressive armoured lag deposit of large siliceous boulders and heavy minerals, including gold. This process repeated several times to create the interbedded conglomerates exposed currently.

Channel mineralisation is located in close proximity to the Mosquito Creek Formation contact and is the dominant mineralisation at South Hill and the southern parts of Golden Crown (Figure 3). Marine lags are the only form of mineralisation distal from the contact, with up to seven lags identified at Grant’s Hill and Golden Crown. Towards Edwards Lease (Edwards), only two dominant marine lags continue (Figure 3). Here, the M1 and M2 reefs are continuous for over 2.5 km and are only closed off by topography and faults.

The Golden Crown block represents a different fan, with imbrication suggesting sedimentation from the east as opposed to the southeast (Figure 3). Three marine lags have been defined in this domain, with an additional sequence of channel mineralisation towards the southern margin. The sequence of channel mineralisation appears to transition towards marine lag mineralisation from south to north, generating a complex geological setting where channels and lags overlap.

The palaeoplacer deposition model applied at Beatons Creek is based on detrital gold sourced from the nearby Mosquito Creek Formation and deposited locally. Mineralisation has further been concentrated by marine reworking of an already endowed sequence of conglomerates by marine processes, as noted above. The presence of significant concentrations of pyrite is a key factor in reef identification (Figure 2 and Figure 4).

2.3. Mineralogy

Mineralogical studies at Beatons Creek principally relate to work undertaken by Gough [33] and Galindo et al. [34]. Sulphide minerals identified, other than pyrite, include gersdorffite, pyrrhotite, galena, chalcopyrite, sphalerite and cobaltite. Their occurrences are generally observed as fracture-filling textures within detrital grains of pyrite or around pyrite margins. Pyrite may locally form up to 40% of fresh mineralisation (Figure 2), with the other sulphides present in minor quantities (<1%).

Pyrite overgrowths point to post-sedimentary fluid circulation and the mobilisation of arsenic, nickel and cobalt [33]. An analysis showed that at least some gold, as well as chalcopyrite, sphalerite, galena and cobaltite, have been remobilised during metamorphism in greenschist facies conditions [33]. Quantitative mineralogy confirms that rounded grains of pyrite are concentrated with other detrital heavy minerals (e.g., rutile and chrome spinel) in the conglomerate horizons. Abraded and rounded edges on the outer surfaces are potential evidence of sediment erosion and transport from a source [33]. Two types of rounded first-generation pyrite were observed and can be characterised by texture and geochemistry. Their coexistence, but with different textures and geochemistry, suggests multiple sources from different zones within the Mosquito Creek Formation [33]. Euhedral and overgrowth pyrite of authigenic origin is likely the product of in situ metamorphic or hydrothermal mineralisation.

Gold is located dominantly in the conglomerate matrix and minor fracture infill associated with sulphide minerals. Its irregular shape and fracture-filling texture suggests that it is at least in part diagenetic or hydrothermal in origin. A mixed placer–hydrothermal model is envisaged, though this line of research is on-going [34].

Recent work on fresh samples collected from the Grant’s Hill M2 reef confirms that gold particles are located dominantly in a fine-grained conglomerate matrix (Figure 1, Figure 2, Figure 3 and Figure 4) [34]. Limited X-ray tomography reveals localised gold particle clustering, with an in situ volume to c. 500 mm³, with individual gold particles ranging 100–3500 µm in size [34].

The conglomerates are present as oxide and fresh (sulphide) mineralisation types (Figure 1 and Figure 2). Based on an analysis of grade and duplicate data, it is concluded that, from a sampling perspective, the two types can be treated as the same [28,29]. In all discussions in this contribution, oxide and fresh mineralisation are presented as one unless otherwise stated.

2.4. Gold Particle Sizing

2.4.1. Overview

Gold mineralisation within the Beatons Creek conglomerates occurs as fine grains, and larger flakes (≤1 mm) and rounded particles occasionally exceeding 8 mm in size. Coarse and fine gold are spatially related to higher concentrations of pyrite, and there seems to be a broad correlation with gold and pyrite clast size and abundance (Figure 2 and Figure 4). Gold particles are commonly visible and coarse (Figure 4 and Figure 5).

2.4.2. Field Observations of Gold

During mapping (2018 and 2021–2022), core logging (2018 and 2022) and optimisation testwork (2020 and 2021), the occurrence of gold particles and clusters was recorded in various material types (

Table 1. Occurrence of gold particles at Beatons Creek in outcrop, rock samples and drill core, panned RC cuttings, and laboratory crushed and pulp products. The cluster volume is the “in situ cube volume” containing the cluster group. Panning was undertaken post-primary assay to ensure a range of grades were selected (0.5–10 g/t Au). For each of RC cuttings, crushed material and pulp, 25 samples were panned.

Type	Mineralisation	Material Size	Sample Support	Gold Particle Size Range	Gold Clustering
Outcrop	Oxide and fresh	Pit walls	n/a	0.25 mm to 5 mm	Not recorded
Rock specimens	Oxide and fresh	<1 m	n/a	0.25 mm to 5 mm	Observed to c. 500 mm3
Bulk sample gravity concentrate	Oxide	P80 750 µm	2 t (primary mass)	<100 µm to 2 mm	Destroyed by pilot process
Drill core	Fresh	PQ core	c. 20 kg (1 m comps)	0.25 mm to 7 mm	Observed to c. 500 mm3
RC cuttings	Oxide and fresh	P80 4 mm	c. 17 or 34 kg (0.5 or 1 m comps)	<100 µm to 4 mm	Mostly destroyed by drilling
RC crushed	Oxide and fresh	P80 2 mm	5 kg	<100 µm to 3 mm	Mostly destroyed by crushing
Pulp	Oxide and fresh	P90 75 µm	3 kg	<100 µm to 1.5 mm	Destroyed by pulverisation

).

The occurrence of gold particles in rock samples confirms the coarse nature of the gold at Beatons Creek. Importantly, the panning of RC cuttings, and the crushed and pulverised product, also indicates the presence of coarse gold. This is particularly notable in the crushed and pulverised products, where the presence of coarse gold is material to the sample splitting, handling and assay processes.

2.4.3. Metallurgical Testwork

Based on the 2019 (from 2018 diamond core) gravity recoverable gold (GRG) testwork [35], 53% (M1 reef—5.5 g/t Au head grade) and 37% (M2 reef—4.4 g/t Au head grade), respectively, of gold reported to the Stage 1 concentrate at P₈₀ 550 µm [28,29,31]. Size-by-assay analysis of the two Stage 1 concentrates indicates 31% (M1 reef) and 23% (M2 reef) of the gold being >600 µm in size [28,29,31]. The GRG test programme in 2022 for three sample composites shows 46% (1.7 g/t Au head grade), 50% (2.6 g/t Au head grade) and 56% (2.5 g/t Au head grade) of gold reporting to the Stage 1 concentrate (P₈₀ 850 µm) [28,29]. Size-by-assay analysis of the Stage 1 concentrates indicates 65%, 53% and 47% of the gold being >600 µm in size [28,29]. The testwork shows that across a range of grades, the proportion of gold >600 µm in size is substantial even after grinding.

During the 2021–2022 operating period, the plant feed was ground to P₈₀ 150 µm, with the cyclone underflow passing through the gravity circuit. Gravity gold recovery was 55% for oxide and 65% for fresh mineralisation, respectively [28,29]. Observation of gravity concentrates confirmed the presence of >1 mm gold particles even after grinding.

2.5. Heterogeneity Testwork

Given the highly heterogenous nature of mineralisation at Beatons Creek, traditional heterogeneity studies such the “duplicate series analysis” or “heterogeneity test” on screened fractions were not undertaken [21,22,23,24]. Various authors have warned of the unrepresentative nature and high variability of such calibrations, and that a more integrated approach is required in coarse gold mineralisation [21,22,23,24].

A “uncalibrated” (e.g., unscreened crushed material) heterogeneity study was undertaken, using the crushed drill core of fresh mineralisation assayed by PA. Figure 6 displays a series of “heterogeneity” curves across five grades.

For each sample comprising 40–52 individual PA jars, the IH_L and dℓ values were calculated (

Table 2. Summary of heterogeneity testing across a series of grades using individual PA jars. IH_L rounded to the nearest 100, and dℓ rounded to the nearest 50 µm. Calc. IH_L and Calc. dℓ based on analysis of PA jar grades. Obs. dℓ_max is the maximum gold particle size observed from outcrop and samples, and core and metallurgical testwork (inc. bulk samples). Est. dℓ based on 70% of the Obs. dℓ_max value.

Grade (g/t Au)	No. PA Jars	Mean PA Jar Weight (g)	Tot. Assayed (kg)	RSV PA Jar Grades	Range PA Jar Grades (g/t Au)	Calc. IHL (g)	Calc. dℓ (µm)	Obs. dℓmax (µm)	Est. dℓ (µm)
0.2 (LG)	42	514	21.6	76%	0.02–0.83	3000	350–550	<250–500	<175–350
0.5 (MIG)	50	531	26.5	91%	0.19–2.96	5400	800–1000	500–1200	350–840
1.1 (ROM[1])	40	504	20.2	75%	0.13–7.63	8600	1500–1700	500–2000	350–1400
1.6 ROM[2]	52	512	26.6	117%	0.26–10.81	8800	1800–2000	1000–2500	700–1750
3.1 HG	48	508	24.4	235%	0.32–36.61	31,000	4500–4900	1500–5000	1125–3500

dℓ calculation inputs: f = 0.3; g = 0.25; d_N = 0.35 cm; alpha = 0.6 or 1.0; gold density = 18.3 g/cm³.

). Equation (3) was used to calculate the IH_L, where a fragment group (M_q and a_q) was taken as a single PA jar. dℓ was calculated using Equation (4).

Each PA jar contained non-screened crushed (P₈₅ 3 mm) material. The IH_L values were calculated for each set of assays and ranged from 3000 g to 31,000 g. The two highest grades (1.6 g/t Au and 3.1 g/t Au) gave the highest values at 8800 g and 31,000 g respectively. With an increasing grade, the IH_L increases and indicates the strong likelihood of coarse gold. An IH_L value of >5000 g is typical of coarse gold-bearing mineralisation, particularly alluvial systems and some veins [2,14,36].

A range of dℓ values are given based on alpha (α) values of 0.6 and 1.0. No formal determination of α was undertaken, with the range applied based on experience of similar mineralisation.

The calculated dℓ values (Table 2: Calc. dℓ) are compared against dℓ_max values based on gold particle sizing from field observations and metallurgical testwork. The values compare well, at least at an order-of-magnitude scale, though the dℓ_max values are higher. Based on experience, the dℓ value is between 60% and 80% of dℓ_max [36]. The estimated dℓ value (e.g., 0.7× dℓ_max: Est. dℓ in Table 2) consistently understates the calculated dℓ (Table 2: Calc. dℓ). It should be noted that the comparison between dℓ and dℓ_max is not exact, given the different samples involved, ranging from c. 40 kg to 2 t (Table 1 and Table 2). The results of this study indicate the general inefficiency of heterogeneity-testing approaches in coarse gold-bearing high-nugget mineralisation.

This analysis does not resolve the presence of gold particle clustering, which, given the crushed nature of the assayed material, will have been destroyed. It is possible that some small remnant clusters could exist, particularly at high grades. These may be indicated by the highest calculated dℓ values (Table 2). Crushed material (P₈₅ 3 mm) will contain rock fragments up to 5–6 mm in size, which may bear gold particles related to clusters or parts thereof.

3. Resource Development and Grade Control Sampling

3.1. RC Sampling Methodology

3.1.1. RC Sampling Methodology Pre-2020

RC cuttings were collected at 1 m intervals via a cyclone and fixed splitter attached to the side of the rig or trailer-mounted. For the 2011–2012 programme, the rig split generated a nominal 10 kg sample for laboratory submission. For the 2013–2017 RC programmes, a riffle splitter was used to collect and split material from the cyclone into a 50/50 split, generating a 15–20 kg sample.

3.1.2. RC Sampling Methodology Post-2020

RC drilling was the only method considered appropriate for the post-2020 resource development and grade control programmes. Diamond core drilling would have been optimal, though logistically impractical and economically prohibitive. Blasthole sampling was reviewed as a grade control option but discounted, given its well-documented disadvantages related to sample contamination and loss, challenges of collecting a representative sample from the cuttings pile (e.g., segregation and zoning effects) and sub-drill effects [3,5,37].

Between 2020 and 2022, both resource development (20 m by 20 m spacing) and grade control (10 m by 10 m spacing) RC drilling was undertaken. Drill spacings were optimised via kriging neighbourhood analysis and reviewed on a continual basis as more data became available [29,38]. Drilling was completed to expand the resource and control mining activities, which commenced in 2021. RC holes were drilled using a 5.5 inch (c. 140 mm) diameter bit.

RC cuttings were collected at 0.5 m intervals via a cyclone and fixed cone splitter attached to the side of the rig or trailer-mounted (Figure 7). The splitter produced two equal splits of 8 kg to 10 kg each: A and B splits. Between the commencement of the programme and mid-August 2021, both splits were submitted to the laboratory. After August 2021, only one of the A or B splits was submitted to the laboratory, unless a field duplicate was scheduled (target rate of 1 in 30 and actual rate of 1 in 33), in which case, both the A and B rig splits were submitted to the laboratory [29].

3.1.3. Application of detectORE™ to RC Composites

detectORE™ (DO) was applied at Beatons Creek during the period May to December 2022. It is an application that utilises portable XRF technology for low-level gold detection in parts per billion [39,40]. The work at Beatons Creek aimed to screen RC samples to identify the mineralised zones, thus reducing the number of samples submitted to the laboratory. The process is summarised in Figure 8.

The DO process relies on a proprietary leach reagent and a ‘widget’ that contains a collector device. The reagent separates the gold from interferences, and the collector device concentrates the gold to detectable levels. Gold is reported as detectORE units (dU), which is a qualitative measure of the gold extracted by the process. A new additive was developed to negate the effect of significant sulphides (up to 40% pyrite) in fresh material. The Beatons Creek oxide and fresh mineralisation displays a good correlation between the DO and primary assay results (Figure 8).

The initial grade control sample intervals were based on the resource wireframes, and were not always spatially accurate, resulting in either the wrong sample intervals or longer sample intervals sent for assay. The DO process was applied on a broader downhole sample selection, targeting the anticipated mineralisation positions. Results from the DO were then used to update the mineralised wireframes, from which a more accurate sample selection was then applied for assay. The DO results are qualitative and therefore cannot be used for resource grade estimation.

The DO process applied was as follows (Figure 8):

• A sample selection list is produced by the geological team based on the anticipated mineralisation positions;
• The A or B split not flagged for laboratory submission is spear-sampled and sieved to <1 mm. The coarse material in the sieve is used for geological logging;
• 500–600 g of screened material is then funnelled into a DO pouch. The pouch is assigned a unique bar code and a ‘widget’ is inserted;
• Samples are dosed with reagent;
• Pouches are loaded and mixed in a bottle roller for 6 h;
• The widget collector device is recovered, dried and analysed with a portable XRF;
• Data is uploaded to a cloud-based database and converted into dU.

Novo found that using DO generated a significant cost saving and improved the assay turnaround time by reducing the volume of samples submitted to the laboratory. In addition, DO was also helpful in identifying “missed” minor reefs located between the known principal reefs.

3.1.4. Diamond Core Sampling

Diamond drilling generated a PQ or HQ core. The core was oriented, marked up and validated against driller core blocks prior to measuring core recoveries. For the pre-2018 core, an automatic core cutter was used to cut the core in half. Samples were typically 1 m in length, although this varied based on geological contacts. A minimum sample length of 0.5 m ensured a sufficient sample for further analysis. The maximum sample length was set at 1.1 m.

For the 2018 and 2022 programmes, the whole PQ core was crushed, and a rotary sample divider (RSD) was used to collect sub-samples for PA. Due to the needs of metallurgical testwork, the assay samples were returned to each composite prior to the recovery testwork. This was facilitated by the PA method being non-destructive. Prior to crushing, the core was scanned using Minalyze CS technology, with resulting outputs of digital core photographs, bulk density measurements on 0.5 m core lengths and continuous down-hole XRF multi-element assays [41].

3.1.5. Trench Channel Sampling

Trench channel sampling was undertaken during 2014–2018. Where outcropping oxide conglomerate horizons were present, channel samples were collected from trenches at 20 m to 70 m spacings along the strike. The sample interval size did not exceed 1 m (vertical). If a conglomerate horizon was <1 m thick, a sample was collected from the top to the bottom of the layer. If the horizon thickness exceeded 1 m, two or more samples were collected. Samples were collected using a handheld compressed air pick to loosen material, and a tarpaulin was used to catch the material. A sample weighing between 40 kg and 65 kg was collected and submitted to the laboratory. The 2018 channel sampling was associated with the bulk sampling programme and was executed to minimise sampling errors such as DE and EE. This included a markup of the sampling zone and proportional collection of hard versus soft material within the target area [30,31].

3.1.6. Bulk Sampling Programme

The bulk sampling programme attempted to quantify the magnitude and distribution of gold grades within oxide mineralisation. Novo collected 45 primary and 13 duplicate bulk samples, where the samples were approximately 2.3 t (range 1–4 t), across approximate 1 m increments of conglomerate. All bulk samples were processed in their entirety through a gravity-based pilot plant located at an independent laboratory in Perth. Novo applied a considerable effort to minimise sampling errors during sample collection, which resulted in the highest-quality grade determinations at Beatons Creek [30].

4. Resource Development and Grade Control Sample Preparation and Assaying

4.1. Overview

The 2023 MRE was estimated from 35,422 samples [29]. Pre-2020 assays are determined by the LeachWELL™ (LW) technique (9% of total assays). Some samples were assayed by the fire assay (FA) or screen fire assay (SFA) methods (1% of total assays). Assays from 2020 onwards were determined by the PA technique (90% of total assays), using a 5 kg (23% of PA), then a 2.5 kg (77% of PA), assay charge. The rationale for changing from a 5 kg PA charge size to 2.5 kg is detailed in Section 4.8.4.

4.2. RC Sample Preparation and Assay (Pre-2011)

The exact protocols were not well documented. Post sorting, drying and weighing, the 3 kg sample was treated as follows:

• Crushed to P₈₀ 2 mm;
• Split (riffle?) to 1 kg;
• Pulverised to P₈₅ 75 µm;
• Split (riffle?) for FA (30 g) or SFA (500 g or 1 kg) or LW (1 kg).

4.3. RC Sample Preparation and Assay (2011–2012)

Samples were sorted, dried and weighed. Thereafter, the 10 kg sample was treated as follows:

▪

Crushed to P₈₀ 2 mm;

▪

RSD split to 3 kg;

▪

Pulverised (Mixermill) to P₈₅ 75 µm;

▪

Riffle split 1 kg for 6 h LW assay, followed by inductively coupled plasma mass spectrometry (ICP-MS) analysis;

▪

Selected LW results over 0.2 g/t Au triggered an FA on the residue to quantify any gold not dissolved during leaching.

4.4. RC and Trench Channel Sample Preparation and Assay (2013–2017)

RC and trench channel samples were prepared and analysed using the following protocols:

▪

RC (15 kg) and trench channel (40–65 kg) samples were sorted, dried and weighed;

▪

Crushed (the entire sample) to P₈₀ 2 mm with a jaw crusher, followed by a Boyd crusher;

▪

RSD split to 9 kg;

▪

Pulverised (LM5) the 9 kg to P₈₅ 75 µm—this was done in three 3 kg batches, due to the limited size of the pulveriser;

▪

The three pulverised splits were re-homogenised to 9 kg of pulp;

▪

Re-split, the 9 kg pulp was re-split into three 3 kg bags;

▪

One 3 kg pulp bag subjected to a 6 h LW assay and ICP-MS analysis;

▪

Selected LW results over 0.2 g/t Au triggered an FA on the LW tails.

For the 2018 trench channel sampling programme, the entire c. 50 kg sample was batch-pulverised and then split, to produce one 3 kg lot for LW assay.

4.5. Diamond Drill Core Sample Preparation and Assay (2013–2014)

Samples were prepared and analysed using the following protocol:

▪

Crushed to P₈₀ 2 mm with a Boyd crusher;

▪

Pulverised (LM5) all material to P₈₅ 75 µm;

▪

RSD split the pulp to generate two 1 kg bags;

▪

Subjected 1 kg pulp to a 24 h LW assay, followed by ICP-MS analysis. For any sample within the mineralised sequence, two 1 kg pulp samples were assayed;

▪

Selected LW results over 0.2 g/t Au triggered an FA on the LW tails.

4.6. Diamond Drill Core Sample Preparation and Assay (Metallurgical Samples 2018 and 2022)

For the 2018 and some of the 2022 PQ core, comminution testwork was undertaken on individual intersections. Post-comminution testwork, the samples were individually recombined, prepared and analysed using the following protocol [28,29]:

▪

Dry and weigh;

▪

Crush to P₈₀ 1.5 mm with a Boyd crusher;

▪

RSD split approximately 10 kg;

▪

The 10 kg is manually poured into 20× PA jars of 500 g each.

4.7. Diamond Drill Core Sample Preparation and Assay (2018 and 2022)

For the 2018 and 2022 PQ core, which were not subjected to metallurgical testwork:

▪

Dry and weigh;

▪

Crush to P₈₀ 3 mm in a Boyd crusher;

▪

A sub-sample of approximately 2.5 kg is split off automatically;

▪

The 2.5 kg is manually poured into five PA jars of 500 g each;

▪

For the 2018 samples, assay was via LW 1 kg.

4.8. RC Sample Preparation and Assay 2020 Onwards

4.8.1. Sample Preparation

Resource development and grade control RC drilling undertaken from October 2020 produced 0.5 m length samples. The rig splitter produced two equal splits (A and B) of approximately 8–9 kg each. Initial sample preparation was undertaken at commercial laboratories in Perth and Kalgoorlie. PA was undertaken in Perth and Kalgoorlie. This work commenced in October 2020, terminating in late August 2021. From late August 2021, samples were prepared at the Nullagine site laboratory (Figure 9). All PA analysis was undertaken in Perth.

4.8.2. PhotonAssay™

PA is a significant development in the gold analysis field, whose key drivers include reduced environmental impact and operating costs, increased assay speed and improved accuracy [42,43]. It is a novel X-ray method that provides accurate, fully automated and non-destructive measurements of large samples (350–500 g per assay jar: Figure 9). The method is agnostic to material composition and granulometry. No chemicals are used and no waste produced, other than the sample material, which can be stored or used again as required. It provides faster turnaround times and lower costs than most competing approaches. PA can analyse coarse crushed samples at a rate of 70 samples per hour. Once jarred, samples are computer-registered by the operator and automatically fed into the PA unit (Figure 10).

This methodology provides distinct advantages in terms of time and cost. These characteristics make it applicable to gold ores, particularly those bearing coarse gold, as only crushing is required (reduction in liberated gold) and multiple lots can be assayed [42,43]. The method has been ISO/NATA certified, and the results have been included in Exploration Results, Mineral Resource and Ore/Mineral Reserve estimates reported in accordance with the JORC Code 2012 and NI 43-101 (inclusive of the CIM Standard 2014) [7,27,44].

PA is an alternative to traditional assay methods such as FA, SFA and LW, with the advantages noted previously. However, as with all sampling protocols, using PA still requires proper optimisation, from rig collection to the laboratory assay [3,43].

4.8.3. Sample Preparation and Assay (October 2020 to August 2021)

Between commencement and mid-August 2021, both splits were submitted to the laboratory. The process was as follows:

• The A or B splits were sorted, dried and weighed;
• Crushed to P₈₅ 3 mm in a Boyd smart crusher;
• A sub-sample of approximately 5 kg was split off automatically;
• 5 kg manually poured into ten PA jars of 500 g each.

4.8.4. Sample Preparation and Assay (August 2021 to May 2023)

After August 2021, only one of the A or B split was submitted to the laboratory, unless a field duplicate was indicated, in which case, A and B splits were submitted. The process was as follows (Figure 9):

• The A or B split sample is sorted, dried and weighed;
• Crushed to P₈₅ 3 mm in a Orbis smart crusher;
• A sub-sample of approximately 2.5 kg is split off automatically;
• The 2.5 kg is manually poured into five PA jars of 500 g each.

On commencement of the grade control programme, the A and B splits were both submitted to the laboratory for analysis. Whilst optimal from a sampling protocol perspective, the bags required substantial manual handing and were a bottleneck at the preparation laboratory, particularly at the sample-crushing and splitting stage. Based on the evaluation of 2525 oxide and 1139 fresh A-B assay pairs (of 2.5 kg or five PhotonAssay pots each), the decision was made in mid-August 2021 to submit only one (A or B split) sample to the laboratory. This decision was based on the analysis of pair variances and the scenario-testing of various combinations of assays (PA jars) during the estimation of a trial area at Grant’s Hill (Figure 3) [38].

An analysis showed that, above 3 kg of samples (six PA jars), precision did not notably improve, and that estimates using six to ten PA jars were within ±5% globally. Critically, the change improved sample turnaround time and reduced costs. The change to a 2.5 kg assay charge was enacted based on continued well-controlled RC drilling, chip logging and in-pit geological mapping.

The downside of reducing to a 2.5 kg assay charge relates to geological modelling, where a 0.5 g/t Au cut-off is applied during the construction of the mineralised wireframes. Review of the duplicate field data (where either A or B shows a grade ≥0.5 g/t Au) shows that 66% of the pairs have A and B values ≥0.5 g/t Au. Thus, there is a 34% chance that, if the A or B assay is taken, it might not be ≥0.5 g/t Au. If the combined A and B grades are taken, then 91% of pairs are ≥0.5 g/t Au.

Taking only the A or B assay results in a higher probability of a given sample not being included in the wireframes. For the more continuous marine lags, this is not so problematic, given that realistic assumptions about their gross continuity can be made. This risk is higher for the more complex channel areas (e.g., Golden Crown: Figure 3), where discontinuity is likely.

4.9. Quality Assurance/Quality Control

The key period of Novo resource development and grade control drilling and associated laboratory activity was November 2020 to May 2023. During this period, 184,738 individual sample assays (as groups of five or ten PA jars) were returned. The QC programme included field and laboratory duplicates, and analytical repeats; CRMs and field blanks; umpire assays; barren flush assays; and crusher split quality monitoring [29]. Laboratory QC results were also monitored, and regular visits were made to the laboratories by company staff [29].

Crushers were cleaned between each sample, although this was restricted to brushing and air-blasting the easily accessible parts of the unit. At the beginning, middle and end of each shift, the crusher units were run through with blank rock material and vacuum-cleaned. Blank flush material was selected at random for assay to check for contamination; this was undertaken in additional to routine Novo field blanks.

At the beginning of each shift, barren material was run to check that the splitters were taking splits that were within ±5% (Figure 9). The splitters were generally in compliance and if not, were tagged out for further investigation.

5. Analysis of Errors

5.1. Fundamental Sampling Error Analysis

FSE was calculated using the François-Bongarçon modified FSE equation, as presented in Equation (1) (

Table 3. Calculated FSE values for RC, channel and core sampling protocols using the MIG (0.5 g/t Au) and ROM (1.5 g/t Au) grades. For each of the two grade scenarios, FSE was calculated using the low, median and high dℓ values. The three FSE values given in each line represent the low, median and high dℓ values. FSE values are rounded to the nearest 1%. Lot and sub-sample masses applied based on averages. Bold values are worst-case scenarios for the dominant sampling protocol informing the 2023 MRE. Mean oxide/fresh mineralisation mass applied at 34 kg/m for RC samples.

Protocol Period	Protocol	Total FSE for MIG (%)	Total FSE for ROM (%)
Pre-2011 RC	1 m composites (34 kg); rig split 3 kg; crush, then pulverise; split for assay for FA 30 g or SFA 1 kg or LW 1 kg	±83–166–332 [FA] ±52–104–208 [LW/SFA]	±48–96–192 [FA] ±30–60–120 [LW/SFA]
2011–2012 RC	1 m composites (34 kg); rig split c. 10 kg; crush, then split 3 kg and pulverise; split 1 kg for LW	±41–83–165	±24–48–95
2013–2017 RC	RC: 1 m composites (34 kg); rig split c. 15 kg or CH: 50 kg; crush, then split 9 kg and pulverise; split 3 kg for LW	±23–47–93 [RC] ±21–41–82 [CH]	±15–28–57 [RC] ±12–24–47 [CH]
2018 channel	c. 50 kg crush and pulverise in total; split 3 kg for LW	±6–12–25	±4–7–15
2018 and 2022 core (Met. protocol)	1 m composites (c. 20 kg); crush, then split 10 kg for PA	±12–23–46	±7–13–27
2020–2021 RC	0.5 m composites (17 kg); rig take A and B bags; crush, then split 5 kg for PA	±25–49–99	±14–29–57
2021–2023 RC	0.5 m composites (17 kg); rig split, take A or B bag (8.5 kg); crush, then split 2.5 kg for PA	±42–83–166	±24–48–96

FSE inputs: f = 0.3; g = 0.25; alpha = 1.0; gold density = 18.3 g/cm³; dℓ = 250 (low), 500 (median) and 1000 (high) µm.

). The target total FSE is ±32% or less [45]. It should be noted that the calculation of the FSE is just that, and no account is made for other TOS errors.

Except for the 2018 channel samples, all FSE values are high, reflecting high dℓ values at low grades. For the pre-2011, 2011–2012 and 2013–2017 RC protocols, the dominant component of the total FSE is the rig split, although in the case of the pre-2011 RC protocol, the FA split (30 g from 3 kg) is dominant (62%). The 2018 channel sample FSE values are low, given that the entire field sample is crushed and pulverised. However, the resulting 50 kg of pulp is difficult to homogenise and will be prone to enhanced GSE. This is confirmed from duplicate pulp analyses presented in Section 5.2 and from panned pulp (Table 1).

The 2018 and 2022 diamond core protocol display dominantly acceptable FSE values, which relate to the large 10 kg split taken after crushing to 1.5 mm. In addition, the entire 10 kg was assayed by PA. For the 2020–2021 and 2021–2023 RC protocols, the laboratory split becomes dominant. For the initial 2020–2021 protocol, there was effectively no rig split, as the two halves taken at the rig were recombined to take the final 5 kg split. So, the total FSE relates to splitting 5 kg from c. 17 kg. For the final 2021–2023 protocol, the split route was 17 kg to 5.5 kg to 2.5 kg.

The use of the FSE equation represents an idealised expectation that may or may not be attained in practice, but it provides a starting point from which protocols can be compared, evaluated and optimised.

5.2. Sampling Protocol Duplicate Pair Error Analysis

5.2.1. Introduction

Errors representing the repeatability of assay results can be empirically estimated by pairwise analysis of field, laboratory/coarse and pulp duplicates [46,47,48]. Sampling protocols include several stages of comminution and subsampling, where duplicates can be taken at every stage to allow an estimation of the sampling precision error and the relative contributions at the various stages of the sampling protocol (e.g., sampling, preparation and analysis error). The errors calculated from duplicate pairs contain all TOS errors (e.g., FSE, GSE, DE, EE, PE, WE and AE). Stanley and Lawrie [46] and Abzalov [47] have shown that the average coefficient of variation (e.g., average relative standard deviation) estimated from paired data produces a robust estimate of the total sampling error (from collection to assay).

Pairwise errors (reported here as the relative sampling variability—RSV) are calculated using the average coefficient of variation equation provided in Stanley and Lawrie Table 1 [46] and Abzalov equation #26 [47]. The RSV is calculated for each set of duplicate data (e.g., field, laboratory and assay duplicates) to give the RSV for that cumulative error, e.g., the field RSV is the total error, including the laboratory and assay RSVs. Precision values were then converted into relative variances to maintain additivity, and the subsequent value was subtracted from the original to give the stage value. The stage value relative variance was then converted back to relative standard deviation to give the RSV. For example, the field RSV—laboratory RSV gives the field stage RSV.

The total sampling error (as RSV) will have potential stage components across ±20%–90% for sampling, ±5%–40% for sample preparation and ±5%–15% for analytical/pulp errors, respectively [48]. The precision from duplicate pairs is calculated using the RSV at one standard deviation (68% reliability).

All data have been filtered at 10× the assay method detection limit to calculate precisions. No removal of potential outlier values has been undertaken. Given the coarse gold nature of the mineralisation, occasional variability is to be expected [10,48].

For this analysis, the combined oxide and fresh mineralisation results are presented. Separate analysis of each shows minimal differences in the results.

5.2.2. Sampling Protocol Error Analysis—Channel Samples

The channel sample protocol errors are summarised in

Table 4. Stagewise error estimate for 2018 channel sampling protocol (refer to Table 3).

RSV	Field RSV	Laboratory RSV	Pulp RSV
Sample split	c. 50 kg	50 kg to 3 kg	3 kg
Total RSV	±65%	±44%	±25%
Stage RSV	±48%	±37%	±25%
Relative proportion	54%	32%	14%
No. of pairs	110	85	85

.

The dominant error relates to field collection and then a reducing stagewise error across the laboratory coarse-to-pulp splits. The pulp split at ±25% is high, reflecting the presence of coarse gold and a high FSE and GSE during a split with liberated gold present. It should be noted that the pulp RSV also includes the AE and other errors such as the PE, DE and EE for the pulp-splitting stage. The potential presence of coarse gold in both the coarse and pulp products was confirmed in testwork (Table 3).

5.2.3. Sampling Protocol Error Analysis—2013–2017 RC

The pre-2020 protocol errors are summarised in

Table 5. Stagewise error estimate for resource development RC sampling protocol (2013–2017).

RSV	Field (Rig) RSV	Laboratory RSV	Pulp RSV
Sample split	c. 34 kg to 15 kg	15 kg to 9 kg	9 kg to 3 kg
Total RSV	±62%	±40%	±23%
Stage RSV	±47%	±33%	±23%
Relative proportion	58%	28%	14%
No. of pairs	430	175	1050

.

The dominant error relates to the 50/50 rig split and then a reducing stagewise error across the laboratory coarse-to-pulp splits. The pulp split at ±23% is high, reflecting the presence of coarse gold and a high FSE and GSE during a split with liberated gold present. It should be noted that the pulp RSV also includes the AE and other errors. The potential presence of coarse gold in both the coarse and pulp products was confirmed in testwork (Table 3).

5.2.4. Sampling Protocol Error Analysis—2020–2021 RC

The 2020–2021 protocol errors are summarised in

Table 6. Stagewise error estimate for the grade control and resource development sampling protocol (2020–2021).

RSV	Field (Rig) RSV	Laboratory RSV	Assay RSV
Sample split	Full field c. 17 kg	17 kg to 5 kg	5 kg
Total RSV	±0%	±38%	±6%
Stage RSV	±0%	±38%	±6%
Relative proportion	0%	98%	2%
No. of pairs	-	555	200

.

This protocol shows a zero field RSV as the entire 17 kg sample was collected, albeit as two 8.5 kg samples. The dominant error relates to the laboratory coarse split, where 5 kg is split from 17 kg. The potential presence of coarse gold in the coarse product was confirmed in testwork (Table 3). The analytical RSV is the repeat assay of the ten PA jars—hence, it is the true AE.

5.2.5. Sampling Protocol Error Analysis—2021–2023 RC

The 2021–2023 protocol errors are summarised in

Table 7. Stagewise error estimate for the grade control and resource development sampling protocol (2021–2023).

RSV	Field (Rig) RSV	Laboratory RSV	Assay RSV
Sample split	c. 17 kg to 8.5 kg	8.5 kg to 2.5 kg	2.5 kg
Total RSV	±52%	±42%	±8%
Stage RSV	±31%	±41%	±8%
Relative proportion	35%	63%	2%
No. of pairs	4455	5873	5658

.

The field RSV appears low, though it represents a 50% (8.5 kg) split of the 17 kg primary lot. The highest RSV is seen in the laboratory, where the stage precision is ±41%, representing 80% of the total error. The potential presence of coarse gold in the coarse product was confirmed in testwork (Table 1). Given that this step of reducing the field split from 8.5 kg to 2.5 kg shows the highest proportion of error, this is the step that could be optimised. Optimisation could include taking two 2.5 kg splits (e.g., 5 kg in total), as previously undertaken between commencement and August 2021 (Table 3 and Table 6).

6. Discussion

6.1. Overview

A number of different sample and assay types have been used at Beatons Creek (

Table 8. Summary of sampling and principal assay methods used at Beatons Creek.

Sample Type	Sample Support	Location/Mineralisation Type	#Samples in 2023 MRE	Principal Assay Route	Application
Bulk	c. 2.3 t	Surface/oxide	51	Pilot plant	Metallurgical testwork and MRE
Channel	40–65 kg	Surface/oxide	57	LW3000	MRE
RC	17–34 kg (0.5 m & 1 m comps)	Depth/oxide and fresh	34,807	FA30; SFA500–1000; LW1000 PA2500 or 5000	MRE and grade control
Diamond core (HQ and PQ)	HQ: 8–14 kg PQ: 9–16 kg (c. 1 m comps)	Depth/oxide and fresh	507	Half core LW1000 MET whole core PA10000	Metallurgical testwork and MRE

).

The most representative samples collected were the bulk samples, though they were limited in spatial coverage to surface exposure. Similarly, the channel samples were also limited to surface exposure. The RC and diamond drill samples extended spatially across the deposit and to the extent of its depth (c. 200 m).

6.2. Bulk Sampling

The bulk sampling programme provides the most rigorous option for grade determination at Beatons Creek. From a grade perspective, the programme confirmed the tenor of the oxide mineralisation, with an average grade at 2.2 g/t Au. Some 52% of the gold was recovered by gravity at a 750 µm grind size. It should be noted that the bulk samples were collected purposefully to minimise dilution, so the grades achieved did not reflect those that may be achieved during mining. Such an approach is routinely impractical and comes with a prohibitive cost and difficult logistics, although it is well-suited to coarse gold high-nugget-style mineralisation [10,14,18,19,20,30,49].

6.3. RC Drilling

6.3.1. Introduction

In coarse gold mineralisation, RC drilling provides both advantages and disadvantages compared to diamond core drilling [37,50]. Key positives include the fact that the mass of the primary sample is larger by generally one order of magnitude (e.g., NQ to PQ core, 5.5–15.5 kg/m versus 5–5.5″ RC, 34–41 kg/m) and that any gold particle clusters are destroyed, and the gold particles are redistributed in the sample. A key disadvantage with RC drilling is that “in situ” geology, particularly contacts, are destroyed and detail lost. Whilst the logging of RC cuttings can recover some geological knowledge, it is limited compared to drill core. Additionally, RC drilling can result in grade-smearing due to contamination between sample intervals.

RC drilling provides the dominant sample type at Beatons Creek, 97% by holes and 98% by samples used in the 2023 MRE. The method provides a large sample mass per m, at 32 kg for oxide and 36 kg for fresh mineralisation. The initial 1 m composite length applied was reduced to 0.5 m from 2020 to improve the geological resolution of the conglomerates.

6.3.2. Geological Dilution of Conglomerates by RC Drilling

The marine lag conglomerate bodies show good global continuity. The channelised conglomerate bodies are complex and more difficult to resolve from RC drilling. The use of 0.5 m RC sampling leads to potential dilution of the true mineralisation thickness, as the 0.5 g/t Au cut-off results in some samples spanning the true thickness boundaries. This dilution is unavoidable, given the nature of the RC drilling and sampling process. This results in a partially diluted geological model, whose effects are not uniform across the deposit. The presence of dromedary boulders in the marine lags also has the effect of locally diluting the mineralisation, as drillholes pass through all or part of them. Their scale is no more than a few meters. Where such low grades are returned, the resource/grade control model will be lower-grade. Logging of the RC cuttings attempts to resolve this problem by identifying the presence of dromedary boulders.

6.3.3. TOS Errors in RC Drilling

RC drilling is not a panacea that results in consistently low sampling errors. During the drilling process (bit to cyclone), DE and EE can be pervasive. DE are generated through variability in the sample length drilled (e.g., 1 m and 0.5 m at Beatons Creek). Variability may occur due to incorrect or obliterated driller’s depth marks, driller inattention to the job, bit diameter changes as the shift progresses and excessive hole-blowing between rods. EE are generated through outside return loss due to a lack of pressure, which may relate to worn O-rings, poor compressor capacity, a blown inner tube or insufficient attention paid to ensuring that the sample interval has time to clear from bit to bag. In addition, there can be a “plucking” effect, related to the preferential removal of soft material at the hole walls, and fraction size segregation, where heavy and large particles remain at the bottom of the hole, while fines tend to rise by pressurised air. Smearing relates to EE, where contamination between samples is caused by the previous sample material not being cleared from the hole bottom before the next length is drilled.

6.3.4. RC Sample Recovery

RC recovery was monitored through the weights of the A and B rig splits collected routinely to August 2021, and as part of the duplicate programme after August 2021. A 5.5 inch (140 mm) diameter drill bit was used across the rigs that were active during the 2021–2023 period. Bits were changed after their diameter reduced in size to 130 mm by wear. This leads to DE, where the expected mass will change as the hole/shift progresses. The 2017 RC programme was omitted from the 2023 MRE due to poor recovery through high dust loss (i.e., EE). Assuming the expected median bit size of 135 mm, the expected mass recoveries were 17.9 kg and 20 kg for oxide and fresh samples, respectively. The average oxide recovery was 89%, with 69% of all data showing between 85% and 100% recovery. The mean mass was 16 kg. The average fresh recovery was 90%, with 72% of all data showing between 85% and 100% recovery. The mean fresh mass was 18 kg. In both cases, the proportion of data indicating >85% recovery was less than the expectation, which is that 80% of the samples have better than 85% recovery.

The variable and sub-optimal recoveries can be explained by the bit diameter change and bulk density variability. The oxide mineralisation displayed bulk density variability related to the relative proportion of oxide minerals and voids. For the fresh mineralisation, bulk density variability was principally related to the relative proportion of pyrite present (Figure 2 and Figure 4). The 0.5 m sample length may also have caused additional DE, as it requires more attention to drill than 1 m lengths. Some fines (sub-100 µm material) loss from rig cyclones was also noted, which related to a lack of a dust emission filter.

6.3.5. RC Cyclone Splitting

The cyclone feed-to-bag process is also not devoid of sampling errors [5,37,50]. Recovery of the A and B samples at the rig was via a static cyclone/fixed cone splitter attached to the side of the rig or trailer-mounted (Figure 7). The splitter was set to recover a 50/50 split. Sample splitting at the rig was monitored through the weights of the A and B splits collected routinely to August 2021 and as part of the duplicate programme thereafter. For both the oxide and fresh mineralisation, the split precision was ±19% at 90% reliability. These figures are high and only borderline-acceptable. Static cone splitters often produce biased samples, because it is difficult to ensure that material is presented to the splitter consistently. The process is prone to DE (e.g., uneven feeding to the splitter and/or a non-level splitter unit), EE (sample lost as dust and material adherence to the inside of the unit) and PE (inter-sample contamination, driller intervention, sample loss during bagging and sample bag numbering mix-ups).

6.3.6. RC Error Analysis

Both the FSE and duplicate pair analysis highlight the split error at the rig, particularly where a 3 kg cut was taken (pre-2014 programme: Table 3). The 2020–2023 programme reduced this effect by initially taking the entire 0.5 m composite, and then half (Table 3). With the revised protocols, the dominant error relates to the laboratory coarse split post-crush. This is driven by the presence of coarse gold at a −3 mm crush size.

The laboratory split is a key consideration when using PA, given that pulverisation is not required. Riffle- or RSD-type splitters are appropriate as either benchtop or automated/smart units. Any kind of scooping or grabbing is likely to lead to high bias, unless the entire sample is assayed, which was the dominant case at Beatons Creek. Crush quality and split precision must be monitored as part of the QC process.

6.4. Diamond Core Drilling

Core drilling represents less than 1% of all drilling used in the 2023 MRE. It was principally used to support geotechnical logging and testing, followed by bulk density determination and metallurgical testwork. Core recovery was excellent (>95%), enabling high-quality samples to be collected (Figure 2). From a geological perspective, diamond drilling provides the best scope for high-quality data. From a cost perspective, routine use of diamond drilling was prohibitive. For the 2018 and 2022 diamond drill programmes, the application of non-destructive PA facilitated grade determination and metallurgical recovery testwork.

6.5. Trench Channel Sampling

6.5.1. TOS Errors in Channel Samples

After the 2019 MRE, it became apparent that the channel samples were positively biased. Consequently, most of the channel samples were not used for the 2022 and 2023 MREs. The channel samples were prone to high DE and EE during sample collection. The soft oxide matrix material contains the dominant part of the gold inventory, and, by virtue of its relative softness, it is easily over-sampled (e.g., high EE). More silicified, less oxidised and dromedary boulder areas are hard and are prone to relative under-sampling without the use of a diamond saw. This issue of high DE and EE leading to grade bias is typical of channel-style sampling, unless collected under close supervision [51].

6.5.2. Channel Sampling Bias

Channel samples collected during 2018 and informing bulk sample locations were compared to the bulk sample grades and found to be biased by +21%. Given the effort made to collect and process the bulk samples, they were assumed to be the most representative sample type collected at Beatons Creek [30,31]. The 2018 channel samples were collected to minimise DE and EE. The bias level of the pre-2018 channel samples was likely to be higher, potentially +100%, based on a small number of channel samples located close (not spatially coincident) to the bulk samples. Consequently, most of the channel samples are excluded from the 2022 and 2023 MREs. It should be noted that any comparison between the 45–65 kg channel samples and c. 2 t bulk samples is not direct, given the different supports they represent.

6.6. Sample Preparation and Assay Methods

Earlier programmes (pre-2021) took a more traditional approach to assaying, using 30 g FA and 1 kg SFA, then 1 kg or 3 kg LW assays. Numerous authors have warned about the biased assay results when using FA to determine a grade on coarse gold mineralisation [13,15,52,53,54,55].

Minimal assay comparisons were undertaken before 2021. However, a small dataset of 52 pulps assayed by 30 g FA were re-split and assayed via 1 kg SFA. The SFA grades display a +28% bias (19% uncertainty) compared to the original FA. This is an expected result, where the larger SFA charge has a greater probability of encountering coarse gold particles [13,36,52,53].

The 2011–2012 and 2013–2017 protocols yielded large, pulverised lots (3 kg and 9 kg, respectively), with sub-samples of 1 kg and 3 kg taken respectively for assay. Coarse gold particles displayed poor disintegration during pulverising, which was confirmed at Beatons Creek by the panning of a small number of pulps and the poor pulp precision on 3 kg LW duplicates (Table 1). Together with the high density of gold, it makes the correct splitting of pulps difficult, unless RSD or riffle spitters are used. If any kind of homogenisation (e.g., mat rolling) is attempted, then the GSE is likely to dominate above the FSE [3,16,17]. Any scooping from pulp for sub-sampling will also be problematic [3,6,17].

For the pre-2011, 2011–2012 and 2013–2017 protocols, the splitting of the crusher product was undertaken by RSD. The splitter type for the pulps was not recorded for pre-2011, and a riffle spitter was used for the latter protocols. For the 2013–2017 samples, the crusher split was 9 kg. Pulverisation was undertaken as three 3 kg lots in an LM5 mill. After pulverisation, the three 3 kg lots were “homogenised”, prior to re-splitting to 3 kg for the LW assay. The staged pulverisation led to a greater opportunity for gold loss and contamination in the pulveriser bowl (e.g., PE). The homogenisation process was unlikely to be efficient, although if the split were undertaken via an RSD or riffle spitter, then any GSE would have been reduced.

The application of SFA and LW are valid, as they permit the use of a large assay charge size of >1 kg. The PA method is applicable to coarse gold assaying, as it allows for a large assay charge size on the scale of kg and avoids pulverisation leading to excessive gold liberation [43].

The DO process provided a capability for screening all samples quickly and cheaply. During the extended trial period, it successfully reduced the number of samples submitted to the laboratory, particularly those in the sub-economic interburden mineralisation, thus reducing costs and improving turn-around times for critical samples.

6.7. Nugget Effect

The total nugget effect derived from the assay data reflects both geological and sampling variability within mineralisation [10,11,12]. A major emphasis of sampling optimisation is to reduce the sampling component of the total nugget effect.

Table 9. Total nugget effect values across various iterations of MRE for Grant’s Hill (M1 and M2 reefs) and Golden Crown (all reefs). Nugget effect values based on downhole 1 m composites.

MRE Date	Dominant Drill Spacing	Prop. of RC Holes	M1 Reef Nugget	M2 Reef Nugget	Golden Crown (All Reefs)
2015	>20 by 20 m	72%	75%		-
2019	>20 by 20 m	64%	61%	60%	56%–60%
2022	≤20 by 20 m	96%	41%	53%	35%–63%
2023	≤20 by 20 m	97%	54%	65%	37%–75%

provides a summary of nugget effect values from the different MREs between 2015 and 2023. The nugget effect (variance) was determined from downhole variograms and is reported as a function of the total sill (total sample variance) and the proportion of RC holes in the total dataset.

The M1 and M2 reefs are reported as the dominant marine lags present, making up 33% of the 2023 MRE tonnes and 54% of the contained gold. The range of nugget values is provided for Golden Crown as an example of fluvial channels.

By MRE, there was a change in nugget effect values, reflecting various issues and not simply the sampling protocol. The differences include the mix of samples used for the variography, drill spacing, the number of RC holes available and revisions to the geological model (wireframes). For the 2015 and 2019 MREs, the data spacing was sparser and channel samples contributed to the high-nugget effect. For the 2022 and 2023 MREs, fewer channel samples were used, with the dominant sample type being RC. The 2020–2023 PA-based protocols were used for the 2023 MRE (Table 3). The higher-nugget values in 2023, relative to 2022, are not readily explained, beyond notable changes in the geological model and additional infill drill holes, which better depict the short-scale variability.

Sensitivity analysis was undertaken on a well-drilled (10 m by 10 m) area of Grant’s Hill, which used the initial Novo protocol (2020–2021 with 5 kg PA—Table 3). This enabled variography and estimation to be run on the same area but using different combinations of PA results [38].

Table 10. Nugget effect values for a test area on Grant’s Hill drilled at a 10 m by 10 m spacing based on different sample masses. Nugget effect values based on 0.5 m and 1 m composites. Scenario [1] is the original assay used for all samples in the test area.

Scenario/Source	#PA Jars	PA Mass	M1 Reef Nugget		M2 Reef Nugget
			0.5 m	1 m	0.5 m	1 m
1. A and B bags	10	5 kg	71%	67%	53%	50%
2. A bag	5	2.5 kg	70%	65%	53%	43%
3. B bag	5	2.5 kg	65%	60%	53%	58%
4. Random 2× PA jars from A and B each	4	2 kg	74%	69%	58%	53%
5. Random 1× PA jar from A and B each	2	1 kg	80%	60%	62%	47%

summarises the results of the analysis.

All samples were collected at 0.5 m lengths (Table 3; 2020–2021 RC) and then composited to 1 m for estimation. The 0.5 m composites generally display a higher-nugget effect than the 1 m composites, according with expectation and the well-documented volume–variance relationship. The change in nugget effect with sampling protocol, in Scenarios (1) to (5) in Table 10, is broadly within expectation, with some aberrations. For the 0.5 m composites, the nugget effect generally increases from Scenario (1) to (5), with some exceptions. For the 1 m composites, the relative variability is higher, but still lower than for the 0.5 m composites. It should be noted that, in Scenarios (4) and (5), only one set of four and two PA jars was selected and modelled. An exhaustive study across all combinations of jars was not undertaken. Overall, the nugget effect does not seem to be sensitive to the sampling protocol applied, which may reflect a high geological/in situ component to the total nugget effect [10,11,12].

6.8. Representative Mass on a Primary Sample, Composite and Estimation Block Basis

Metallurgical testwork, trial mining and bulk sampling programme results permit a gold particle size–grade relationship to be inferred, where the range of observations were used to apply Poisson statistics to define an optimal field sample mass to achieve a precision of ±20% at 68% reliability [36,56]. Dominy, van Roij and Graham [30] propose a representative sample mass of 2 t to achieve the target precision over a range of grades (0.5–5 g/t Au) and dℓ (500–5000 µm). Across the scenarios, the optimal mass ranged from <0.5 t to 14 t, with 2 t being a median practical value, covering grades between 0.5 g/t Au (MIG) and 1.5 g/t Au (ROM). This mass was used during the 2018 bulk-sampling programme, where limited field duplicates achieved a precision of ±22% at 68% reliability [23].

On a 0.5 m primary RC sample-by-sample basis, the FSE by protocol is presented in Table 3. Primary samples are used for the construction of geological wireframes based on an MIG of 0.5 g/t Au and geological logging. Geostatistical estimation using ordinary kriging of a given resource block is based on a number of 1 m (2 m × 0.5 m primary samples) composites. Figure 11 shows the reduction in FSE based on compositing to 1 m and the number of composites by block applied in the 2023 MRE at Grant’s Hill to achieve Indicated and Inferred Mineral Resources (e.g., the JORC Code 2012 and the CIM Standard 2014) [7,44]. The FSE values given in Figure 11 were calculated via the FSE equation (Equation (1)) for each number of composites.

Indicated Mineral Resource blocks (as defined by ≤20 m by 20 m drilling) are based on a minimum of 8 (272 kg) and maximum of 22 (748 kg) composites, with a mean of 16 (544 kg) and mode of 22 composites. The composites informing a given block are captured within the neighbourhood search based upon variogram ranges [29]. Similarly, Inferred Mineral Resource blocks (defined by >20 m by 20 m drilling) are based on a minimum of 4 (136 kg) and maximum of 22 composites, with a mean of 13 (442 kg) and mode of 15 (510 kg) composites. The total composite masses are based on a single composite weighing 34 kg (e.g., 2 × 17 kg 0.5 m samples). It can be seen that the FSE value becomes acceptable at >8 composites for grades above the BCOG of 0.5 g/t Au. At this level, the total composite mass is above 272 kg, rising to 748 kg for 22 samples. This approach provides a relative measure of the performance of the sampling protocol through the FSE, although it should be noted that the other sampling errors (e.g., the GSE and CSEs) have not been accounted for. In all cases, the calculated representative sample mass of 2 t has not been achieved, which is a common situation in most coarse gold deposits. As part of the 2023 MRE, a conditional simulation study using 1 m composites was undertaken on the Grant’s Hill M1 reef [29]. This yielded grade precisions (absolute) at the 80% confidence limits of ±11% and ±31%, respectively, for indicated and inferred block grades. This indicates reasonable precisions on grade and includes all sampling errors.

The above analysis addressed grade only, though geological confidence is an also a critical factor in resource estimation. The conditional simulation study also modelled reef thickness and bulk density [29]. This yielded tonnage precisions (absolute) at the 80% confidence limits of ±9% and ±26%, respectively, for Indicated and Inferred Mineral Resource blocks. This indicates reasonable precision in the geological model and that a drill spacing for Indicated Mineral Resources at <20 m by 20 m is appropriate.

7. Conclusions

In the context of Beatons Creeks and other coarse gold-bearing mineralisation types, the following conclusions and recommendations can be made:

• Coarse gold-bearing mineralisation, such as that encountered at Beatons Creek, is challenging to sample effectively. The large gold particle size (>1 mm) and relatively low grades (MIG: 0.5 g/t Au; BCOG: 0.8 g/t Au and ROM: 1.5–2 g/t Au) combine to drive sampling challenges. This is exacerbated by a high geological/in situ nugget effect. Even where the estimation is via geostatistics, with its theoretical and actual advantages, block grades will tend toward the deposit or reef mean grade and estimates will have a high conditional bias [10,11]. As a result, mining to a cut-off grade can be difficult and may result in ore/waste misclassification. In such a case, optimised drilling and sampling protocols are vital but may be impractical. Strong geological control (mapping or dig control) during exploitation is likely to improve the effectiveness of selective mining [10].
• A “best of breed” drilling, sampling and assaying programme to optimise grade and geological definition would be a costly exercise. The resource development programme would include RC drilling at <20 m by 20 m, likely 15 m by 15 m and whole 0.5 m composite assaying via either PA (c. 34× jars) or laboratory-scale gravity processing and tails assay. For grade control, the RC drill spacing would be c. 7.5 m by 7.5 m, with full PA assay or laboratory-scale processing. Close-spaced drilling benefits both grade and geological continuity resolution. RC cuttings should be carefully logged for all programmes. The DO method would be used to screen the samples prior to assay and reduce the load to the laboratory. Diamond core drilling would be integrated into the resource development stage, at a spacing of c. 45 m by 45 m, to provide geological information and material for bulk density determination. The drill spacings will only achieve Indicated Mineral Resources, with the resulting grade control model more local and better suited to selective mining. The inherent geological and grade variability in high-nugget coarse gold deposits precludes the definition of Measured Mineral Resources, even at a close drill spacing [10,57]. On-going geological mapping and control during mining would be critical. The best of breed approach is impractical from cost, time and operational management perspectives and highlights the challenges of evaluating coarse gold-bearing mineralisation. This flags the importance of the Competent Person(s)/Qualified Person(s) in managing project expectations and risk at the resource development and operational stages [10].
• With appropriate data, a theoretical representative primary sample mass can be calculated using a Poisson statistic-based methodology. This incorporates dℓ values associated with a series of critical grades (e.g., MIG, BCOG and ROM), with an output precision of ±20% at 68% reliability (or as deemed appropriate by the practitioner). At Beatons Creek, a value of 2 t was supported by limited field duplicates collected during the bulk-sampling programme.
• Based on duplicate pair analysis, sample precision values range from ±65% (channel, 3 kg LW), ±62% (RC, 3 kg LW), ±52% (RC, 2.5 kg PA), ±38% (RC, 5 kg PA) and ±22% (bulk, pilot plant). The best precision obtained was via bulk samples processed through a pilot plant. Channel sampling was prone to poor precision and high bias due to the over-collection of high-grade conglomerate matrix. The 2020–2021 protocol used the whole 0.5 m composite to form a 5 kg PA sub-sample; however, this was reduced to a 50% rig split and 2.5 kg PA sub-sample to provide cost and time savings. Sensitivity work in a 10 m by 10 m drilled area indicates that this change had a small impact on the block model (within 5% globally).
• A high-quality RC sample provides >30 kg/m. Large splits (>3–5 kg) at the rig are required, potentially taking the entire 17 kg primary sample. The rig split error will generally dominate the total sampling error, though in the case of the 2021–2023 programme, the laboratory split error was dominant (Table 7). The practitioner must consider the relative pros and cons and design an appropriate “rig to assay” protocol, considering the key CSE and ISE. Even when the sample composite is 0.5 m, there is an inherent risk of dilution within a narrow mineralised zone. All efforts must be made to log chippings and undertake mapping to support geological interpretation.
• Whole diamond core sampling, followed by a full or large-sample assay via PA (or SFA or LW or Pulverise-and-Leach), is a valid approach. It effectively yields zero to very low FSE. Arguments against whole core sampling revolve around having no core remaining for reference, although, with modern digital photography, geochemical and spectral sensors and detailed logging this should not be an issue (e.g., the application of Minalyze CS). The PA assay technique is non-destructive; therefore, all (crushed) material can be retained, with the possibility for post-assay metallurgical or geoenvironmental testwork.
• Large mass assays (e.g., >1 kg, LW and PA) are important for coarse gold grade determination. An alternative is to process samples in their entirety via a laboratory-scale unit, though this approach is time-consuming and costly, particularly when large numbers of samples are involved. Such an approach proves more rigorous in the presence of coarse gold, particularly that dominated by >1 mm particles, reducing sampling errors and potentially providing metallurgical data [19,20].
• Heterogeneity tests are well-documented as problematic in the presence of coarse gold, resulting in potentially unreliable results. Multiple PA assays provide a possible source of calibration of unscreened or screened material. This provides at least order-of-magnitude values for the HI_L. No α value is calculated, unless a DSA approach is applied [21]. Given certain assumptions, the dℓ values can be back-estimated.
• There are instances where impractical or economically prohibitive protocols will be required to achieve correct sampling. In this case, the sampling strategy should consider the cost–benefit of protocols and work backwards to achieve an outcome that is both reasonable and cost-effective. It is inherent in coarse gold deposits that highly confident levels of resource classification (e.g., Measured and/or Indicated Mineral Resource categories) may not be achievable, even with close-spaced drilling (<10 m) and large-sample processing [57].
• PhotonAssay™ is a significant development in the field of gold analysis [42,43]. It provides fast, automated and non-destructive measurements on large samples. The method is agnostic to material composition and granulometry. No chemicals are used and no waste produced. The sample material does not require pulverising and can be assayed in a crushed form. This provides distinct advantages in terms of time and cost and allows for multiple jars to be determined to achieve a large assay lot.
• Practitioners should not accept so-called “standard” or “best practice” protocols and methodologies for the sampling, preparation and assay of coarse gold mineralisation. The optimisation of a sampling protocol comes from understanding the mineralisation and desired programme outputs. It is not a mathematical process, but a process taking advantage of orebody knowledge and an application of TOS.
• Conduct systematic QC programmes to measure the reliability of each of the sampling, preparation and assaying steps and then optimise the process. QC cannot be divorced from the TOS and is a mandatory step in representative fit-for-purpose sampling. Proper documentation, regular staff training and peer review are required.
• Identification and resolution of individual relative errors across the complete sampling, preparation and analysis stages (total measurement error) can be gained from duplicate sample pairs. The application of the modified FSE equation is a useful tool to investigate the effectiveness of a sampling protocol. It may not represent exact reality, but it provides a focus for the practitioner to review protocol stages and make changes accordingly. Monitoring of the nugget effect provides an ultimate measure of sampling performance, albeit inclusive of geological/in situ effects.