Geochemical data are frequently collected from mineral exploration drill-hole samples to more accurately define and characterise the geological units intersected by the drill hole. However, large multi-element data sets are slow and challenging to interpret without using some form of automated analysis, such as mathematical, statistical or machine learning techniques. Automated analysis techniques also have the advantage in that they are repeatable and can provide consistent results, even for very large data sets. In this paper, an automated litho-geochemical interpretation workflow is demonstrated, which includes data exploration and data preparation using appropriate compositional data-analysis techniques. Multiscale analysis using a modified wavelet tessellation has been applied to the data to provide coherent geological domains. Unsupervised machine learning (clustering) has been used to provide a first-pass classification. The results are compared with the detailed geologist's logs. The comparison shows how the integration of automated analysis of geochemical data can be used to enhance traditional geological logging and demonstrates the identification of new geological units from the automated litho-geochemical logging that were not apparent from visual logging but are geochemically distinct.

Abstract Core analysts principally study the storage, flow and saturation properties of porous rocks and sediments. Some of the derived parameters are specific to hydrocarbon production but many have commonality with other subsurface disciplines such as hydrology and soil science. Traditional core analysis involves direct physical experimentation on core plugs to derive a range of parameters used as calibration for conventional well logs, and to predict hydrocarbon reserves and recovery. The mechanisms and processes for obtaining such data have evolved significantly during the last century, from the manual instruments of the mid-twentieth century to the accredited digital data collection and recording of the 1990s onwards. X-ray micro- and nano-scale computed tomography (CT) imaging led to the development of the digital rock physics subdiscipline in the early 2000s. This has subsequently allowed direct visualization of fluid flow at the pore scale, imaging the wetting phase and multiphase fluid mobility. Multiscale imaging workflows are being developed to overcome issues around heterogeneous rock and the limited field of view associated with the highest resolution X-ray CT images. Hybrid workflows, which combine digital rock physics with traditional core analysis, are becoming increasingly common to meet the challenges associated with some of the most difficult to constrain properties, such as relative permeability. At a larger scale, the recent development of multisensor core logging (MSCL) tools has allowed the cost-effective acquisition of essentially continuous high-resolution 1D, 2D and 3D datasets from both slabbed and unslabbed whole core. Often aided by artificial intelligence to manage and interpret these large physical and chemical datasets, both new and legacy core can be rapidly screened to allow representative subsampling for detailed laboratory experimentation. The context and data provided by the MSCL then allows effective upscaling of these time- and cost-intensive point-source measurements. In the last decade, extended reality (XR) has resulted in a step change in the ability to visualize and integrate core and core-derived information with other subsurface datasets. A very wide range of scales can be managed effectively, from micrometre- to centimetre-scale petrographical and core analysis data, to metre-scale well logs and up to kilometre-scale 3D and 4D seismic. These tools allow stakeholders to work and meet from any location in a common workspace, and efficiently scale and interrogate data in a virtual 3D environment. The various advances in core analysis and associated technologies during the early twenty-first century mean that the study of porous media to help enable the energy transition looks assured. During the coming decades, applications as diverse as carbon capture, utilization and storage (CCUS), hydrogen storage, geothermal energy generation, mining for critical minerals, palaeoclimate studies, radioactive waste management, and site surveys for windfarms will all continue to benefit from the data and understanding derived from core analysis.

Modelling of 3D domain boundaries using information from drill holes is a standard procedure in mineral exploration and mining. Manual logging of drill holes can be difficult to exploit as the results may not be comparable between holes due to the subjective nature of geological logging. Exploration and mining companies commonly collect geochemical or mineralogical data from diamond drill core or drill chips; however, manual interpretation of multivariate data can be slow and challenging; therefore, automation of any of the steps in the interpretation process would be valuable. Hyperspectral analysis of drill chips provides a relatively inexpensive method of collecting very detailed information rapidly and consistently. However, the challenge of such data is the high dimensionality of the data’s variables in comparison to the number of samples. Hyperspectral data is usually processed to produce mineral abundances generally involving a range of assumptions. This paper presents the results of testing a new fast and objective methodology to identify the lithological boundaries from high dimensional hyperspectral data. This method applies a quadrant scan analysis to recurrence plots. The results, applied to nickel laterite deposits from New Caledonia, demonstrate that this method can identify transitions in the downhole data. These are interpreted as reflecting mineralogical changes that can be used as an aid in geological logging to improve boundary detection.

The chemistry of hydrothermal monazite from the Carrapateena and Prominent Hill iron oxide-copper-gold (IOCG) deposits in the IOCG-rich Gawler Craton, South Australia, is used here to define geochemical criteria for IOCG exploration in the Gawler Craton as follows: Monazite associated with IOCG mineralisation: La + Ce > 63 wt% (where La > 22.5 wt% and Ce > 37 wt%), Y and/or Th < 1 wt% and Nd < 12.5 wt%; Intermediate composition monazite (between background and ore-related compositions): 45 wt% < La + Ce < 63 wt%, Y and/or Th < 1 wt%. Intermediate monazite compositions preserving Nd > 12.5 wt% are considered indicative of Carrapateena-style mineralisation; Background compositions: La + Ce < 45 wt% or Y or Th > 1 wt%. Mineralisation-related monazite compositions are recognised within monazite hosted within cover sequence materials that directly overly IOCG mineralisation at Carrapateena. Similar observations have been made at Prominent Hill. Recognition of these signatures within cover sequence materials demonstrates that the geochemical signatures can survive processes of weathering, erosion, transport and redeposition into younger cover sequence materials that overlie older, mineralised basement rocks. The monazite geochemical signatures therefore have the potential to be dispersed within the cover sequence, effectively increasing the geochemical footprint of mineralisation.

Manually interpreting multivariate drill hole data is very time-consuming, and different geologists will produce different results due to the subjective nature of geological interpretation. Automated or semi-automated interpretation of numerical drill hole data is required to reduce time and subjectivity of this process. However, results from machine learning algorithms applied to drill holes, without reference to spatial information, typically result in numerous small-scale units. These small-scale units result not only from the presence of very small rock units, which may be below the scale of interest, but also from misclassification. A novel method is proposed that uses the continuous wavelet transform to identify geological boundaries and uses wavelet coefficients to indicate boundary strength. The wavelet coefficient is a useful measure of boundary strength because it reflects both wavelength and amplitude of features in the signal. This means that boundary strength is an indicator of the apparent thickness of geological units and the amount of change occurring at each geological boundary. For multivariate data, boundaries from multiple variables are combined and multiscale domains are calculated using the combined boundary strengths. The method is demonstrated using multi-element geochemical data from mineral exploration drill holes. The method is fast, reduces misclassification, provides a choice of scales of interpretation and results in hierarchical classification for large scales where domains may contain more than one rock type.

Manual interpretation of data collected from drill holes for mineral or oil and gas exploration is time-consuming and subjective. Identification of geological boundaries and distinctive rock physical property domains is the first step of interpretation. We introduce a multivariate technique, that can identify geological boundaries from petrophysical or geochemical data. The method is based on time-series techniques that have been adapted to be applicable for detecting transitions in geological spatial data. This method allows for the use of multiple variables in detecting different lithological layers. Additionally, it reconstructs the phase space of a single drill-hole or well to be applicable for further investigations across other holes or wells. The computationally cheap method shows efficiency and accuracy in detecting boundaries between lithological layers, which we demonstrate using examples from mineral exploration boreholes and an offshore gas exploration well.

An important step in mineral resource estimation process is the grouping of drill hole samples into domains that reflect zones of homogeneous properties for accurate grade estimation and practical exploitation purposes. In practice, this challenging task is performed through a subjective, time-consuming manual interpretation of the mineral deposit. Therefore, various interpretations are possible. The definition of domains can be viewed as a clustering problem consisting of grouping samples into clusters, herein called domains, so that samples belonging to the same cluster are more similar than those in different clusters. Several methods exist for this purpose; however, groups of samples created through traditional clustering tend to show poor spatial contiguity. Alternatively, spatially contiguous clusters can be obtained through geostatistical clustering where the spatial dependency between samples is considered. This paper is devoted to the application of geostatistical clustering to support domaining of an iron ore deposit located in Western Australia.

A 3D geology model is a simplified version of the true geology, designed to give a visual summary of the geometry and distribution of major geological elements in a specified region. Drill holes provide detailed data of the subsurface that can be classified into geological units that are the fundamental elements of the 3D model. Due to software limitations, upscaling ('lumping') is usually required to reduce the number of geological units in the drill holes prior to model building. Upscaling is a subjective process, which means that different geologists will group in different ways and will typically not record the rationale behind their decision; this means the "experiment" is not reproducible. In our study we use a method of upscaling geological units, in this case based on assay data, using the continuous wavelet transform (CWT) and tessellation methods. This method reduces subjectivity and can easily be repeated (e.g. on an updated or new drill hole) by using the same parameters, ensuring that the upscaling process is consistent over all drill hole data. We apply this technique to a large assay database (>90,000 samples) from the Kevitsa Nickel-Copper-Platinum group element (PGE) deposit in Finland. The Kevitsa Ni-Cu-(PGE) disseminated sulfide orebody is hosted in a Proterozoic layered intrusion in northern Finland. Internal geological subdivision and correlation within the intrusion is very difficult to do consistently using lithological observations, owing to general homogeneity of rock types and an overprint of alteration, but distinct variability is evident in Ni and PGE sulfide tenors. In its raw form, the tenor variation dataset appears noisy and unsystematic. We have applied the tessellation method to classifying ore types based on tenor variations, consistently and objectively reducing the number of units in each drill hole to create a simplified 3D model of the orebody. Our results reveal shallow inward dipping cryptic layering defined by sulfide composition, which are interpreted as reflecting an increase in Ni and PGE tenor with time during emplacement of the sulfide-bearing cumulates. We interpret this as a progressive increase in silicate-sulfide mixing efficiency (R factor) as the intrusion developed from an interconnected sill sediment-complex choked with country rock inclusions into a freely convecting magma chamber. Based on this case study, we show that the tessellation method can add considerable value by distinguishing the wood from the trees in large 3D geochemical databases. The method may be widely applicable in other Ni-Cu-PGE deposits where tenor variations appear, at first sight, to be chaotic and uninterpretable. (C) 2017 Elsevier B.V. All rights reserved.

There is a need for rapid and reliable techniques for extracting geological information from geochemical data derived from exploration drill hole samples because geochemical data bases are becoming too large to interpret manually. Automated boundary detection techniques which use the continuous wavelet transform are a popular method for extracting multi-scale geological boundaries from drill hole signals. However, these boundary detection techniques do not distinguish between sharp and gradational boundaries, which may be an important factor when interpreting the geology of the subsurface or calculating resource estimates. This paper demonstrates how a scale-dependant measure of relative sharpness of the boundary can be extracted from the results of the continuous wavelet transform. The results of the automated method are compared against the interpretations of 15 geologists. The comparison demonstrates the need for alternative boundary selection methods when boundaries are not sharp. The results also demonstrate how the multi-scale nature of the sharpness measure allows for reliable gradient analysis in noisy signals.

New drilling methods, currently under development for minerals exploration, combined with rapid data collection by a range of sensors means that the end-user is confronted with increasingly large data sets. In order to reduce the stream of data into objects which represent meaningful geological features we need to incorporate spatial information into the analysis. Boundary detection incorporating multiscale considerations has previously been carried out using a scale–space plot from a wavelet transform. We present a method which applies a rectangular tessellation to the wavelet transform, this has the advantage of being easier to interpret, as it resembles a geological log. In addition, the tessellation records hierarchical information for different scale objects. The tessellation can be filtered in order to remove unwanted variation (including noise) from the results. When applied to geochemical data, the resulting tessellation provides a basis for classification of lithochemical units that is more reliable than classification by considering individual samples without spatial context.

In high-nugget gold ore bodies, samples taken from drill core for gold assay are typically too small to allow for the extreme spatial variability in grade and are a poor representation of the underlying distribution of mineralisation. The GQ North lode at the Sunrise Dam Gold Mine in Western Australia is a good example of an orebody which has a very strong nugget effect (coefficient of variation >20) and that has proved very problematic to model. Gold is hosted in vein stockworks and shear zones and although there is a clear spatial relationship between mineralisation and alteration, high vein density and well-developed foliations, the relationship is best defined statistically because the association between high gold grades and various combinations of these features is non-trivial. We present a method for automating the inclusion of geological data (proxies for gold mineralisation) into the prediction of mineralised rocks, using conditional probability. The method uses the gold assays and the logged geological data to calculate the probability that rocks with particular geological features will be mineralised. The ore body can then be modelled automatically using interpolation software with isosurfaces indicating the regions with highest probability of gold mineralisation. A good understanding of the geological features associated with mineralisation and consistent geological logging are important prerequisites for successful conditional probability modelling of drill hole data.

Gold distribution in vein-hosted hydrothermal ore deposits is commonly nuggety (i.e. occurs as very localised concentrations of gold). In these cases samples for gold assay from diamond drill core may be too small to model the underlying heterogeneity of gold distribution and result in poorly constrained ore body models and underestimated gold resources. Hence, it is common practice to use more spatially continuous proxies for mineralisation to help define the boundaries of mineralised regions. We present a method for automating the use of geochemical proxies for nuggety gold ore bodies.Sunrise Dam Gold Mine, in Western Australia, is a world-class gold deposit with a very high nugget effect. Multi-element geochemical data has been collected at this site in order to improve prediction of mineralised regions. Suitable proxy elements have been selected from this data set, in particular, those that are spatially related to gold mineralisation but do not display nuggety distribution, such as Sb, Rb and Cr.We applied a probabilistic approach to the problem of quantifying the relationship between gold assay values and geochemical elements. It is shown that a kernel density estimator and Bayes conditional probability can provide an effective method for calculating the probability of a sample having elevated gold content and that this measure will be more spatially continuous than gold assay values if the appropriate geochemical proxies are selected. Using conditional probability and suitable cut-off values, we reclassified approximately 27% of samples as mineralised which returned low Au assay results. When plotted on drill holes conditional probability values provided a much more spatially continuous guide to mineralised regions than Au assay values alone. (C) 2014 Elsevier B.V. All rights reserved.

In orebodies with very high-nugget gold it is not possible to reliably interpolate gold grades between drill holes. In such cases geologists may resort to using proxies for mineralisation to help define the orebody. We present a method for automating the use of proxies for mineralised zones. Proxies may be derived from categorical data logged by geologists or may be continuous numerical values measured using instruments, such as multi-element geochemistry or hyperspectral analysis. We use conditional probability of finding high grade gold to combine and evaluate the proxies. The conditional probability value can be interpolated and isosurfaced to generate a 3D model of the regions favourable for mineralisation.

A formal method for describing data collected from field studies is used to generate stochastic geological models of sedimentary successions using a method based on syntactic pattern recognition. Using this method an analogue model developed from field data can be encoded as a grammar. The grammar is composed of symbols which represent geological entities. Valid patterns formed by the symbols are described by a set of production rules. In order to demonstrate the potential of the syntactic method, 2D simulations of interpreted cross-sections from Brushy Canyon outcrops are presented here, as well as 2D simulations of seismic facies from the Bengal Fan.

This paper documents a new method for describing channel-related sedimentary deposits based on formal language theory. Using this method an analogue model of a sedimentary deposit can be encoded as a grammar. A program, called a parser, has been developed which can generate stochastic maps of these sedimentary deposits based on information in a specified grammar. The maps of sedimentary deposits generated by the parser have the same type, spatial arrangement, shape and size distribution as the analogue model. The successful generation of depositional maps represents a crucial step in the ongoing development of a new technique designed to generate 3D static geological models of sedimentary successions. The maps can be conditioned to match sparse hard data in the form of channel segments interpreted from seismic horizon maps.

Sections from a sedimentary succession can be simulated using a process which includes both probabilistic and deterministic information. The inclusion of both of these types of information allows the production of geologically realistic simulations which contain the required level of heterogeneity. The process uses syntactic pattern recognition techniques and is based on the formal description of a geological model using a grammar. The simulations can be conditioned on well data.

We show that mapping data from a bounded region onto a toroidal surface prior to running the DR (Dog–Rabbit) clustering algorithm greatly increases the rate of success of finding all the clusters in the region.

As lithospheric plates are subducted, rocks are metamorphosed under high-pressure and ultrahigh-pressure conditions to produce eclogites and eclogite facies metamorphic rocks. Because chemical equilibrium is rarely fully achieved, eclogites may preserve in their distinctive mineral assemblages and textures a record of the pressures, temperatures and deformation the rock was subjected to during subduction and subsequent exhumation. Radioactive parent–daughter isotopic variations within minerals reveal the timing of these events. Here we present in situ zircon U/Pb ion microprobe data that dates the timing of eclogite facies metamorphism in eastern Papua New Guinea at 4.3 ± 0.4 Myr ago, making this the youngest documented eclogite exposed at the Earth's surface. Eclogite exhumation from depths of ∼75 km was extremely rapid and occurred at plate tectonic rates (cm yr-1). The eclogite was exhumed within a portion of the obliquely convergent Australian–Pacific plate boundary zone, in an extending region located west of the Woodlark basin sea floor spreading centre. Such rapid exhumation (> 1 cm yr-1) of high-pressure and, we infer, ultrahigh-pressure rocks is facilitated by extension within transient plate boundary zones associated with rapid oblique plate convergence.