©2019 Society of Economic Geologists, Inc.

Economic Geology, v. XXX, no. XX, pp. X–X

Automated Core Logging Technology for Geotechnical Assessment:

A Study on Core from the Cadia East Porphyry Deposit
Cassady L. Harraden,1,† Matthew J. Cracknell,1 James Lett,2 Ron F. Berry,1 Ronell Carey,3 and Anthony C. Harris2
1 Transforming the Mining Value Chain (TMVC), an Australia Research Council (ARC) Industrial Transformation Research Hub,
University of Tasmania, Hobart, Tasmania 7001, Australia
2 Newcrest Mining Limited, 600 St. Kilda Road, Melbourne, Victoria 3004, Australia
3 Corescan Proprietary Limited, 1/127 Grandstand Road, Ascot, Western Australia 6104, Australia

Abstract
The Cadia East porphyry deposit, located approximately 20 km south of Orange, New South Wales, Australia,
contains a significant resource of copper and gold. This resource is hosted within the Forest Reefs Volcanics
and is spatially and temporally associated with the Cadia Intrusive Complex. To extract ore, the underground
mine currently uses the block cave mining method. The Cadia East geotechnical model provides data inputs
into a range of numerical and empirical analysis methods that make up the foundation for mine design. These
data provide input into the construction of stress models, caveability models, ground support design, and frag-
mentation analysis. This geotechnical model encompasses two commonly used rock classification systems that
quantify ground conditions: (1) rock mass rating (RMR) and (2) rock tunneling quality index (Q index). The
RMR and Q index are calculated from estimates of rock quality designation (RQD), number of fracture sets,
fracture roughness, fracture alteration, and fracture spacing. Geologists and geotechnical engineers collect
information used to produce these estimates by manually logging sections of drill core, a time-consuming task
that can result in inconsistent data.
Modern automated core scanning technologies offer opportunities to rapidly collect data from larger samples
of drill core. These automated core logging systems generate large volumes of spatially and spectrally consistent
data, including a model of the drill core surface from a laser profiling system. Core surface models are used to
extract detailed measurements of fracture location, orientation, and roughness from oriented drill core. These
data are combined with other morphological and mineralogical outputs from automated hyperspectral core
logging systems to estimate RMR and the Q index systematically over contiguous drill core intervals. The goal
of this study was to develop a proof-of-concept methodology that extracts geotechnical index parameters from
hyperspectral and laser topographic data collected from oriented drill core.
Hyperspectral data from the Cadia East mine were used in this case study to assess the methods. The results
show that both morphological and mineralogical parameters that contribute to the RMR and Q index can be
extracted from the automated core logging data. This approach provides an opportunity to capture consistent
geologic, mineralogical, and geotechnical data at a scale that is too time-consuming to achieve via manual data
collection.
Introduction recognition and orientations, they require suspending drilling
Standard methods for geotechnical data acquisition and operations so that the drill hole imaging tools can be lowered
ground condition modeling in hard-rock environments rely down to collect borehole images. Additionally, these meth-
on the identification of fractures from manual logging of ods are expensive and often difficult or impossible to apply in
drill core. While this approach has been and is currently suc- areas with bad ground conditions.
cessfully applied to mining, it is often laborious and has the Other authors have proposed using drill core imagery to
potential to generate inconsistent results. Automation of geo- identify and orient fractures (e.g., Quiniou et al., 2007; Berry
technical data collection from fractures identified in digital and Nguyen, 2016). Berry and Nguyen (2016) developed a
drill core data would allow for consistent, rapid assessment of technique for calculating the orientation of features from
key parameters as they relate to ground conditions. high-resolution drill core photographs. This method showed
Numerous automated and semiautomated fracture detec- good agreement with manually collected fracture orienta-
tion and orientation techniques have been proposed over tions; however, it requires a user to manually identify the loca-
the years (e.g., Quiniou et al., 2007; Ureel et al., 2013; Berry tion of fractures from the photographs. Quiniou et al. (2007)
and Nguyen, 2016). Some approaches use imaging data col- proposed automated methods for identifying and orienting
lected using either an optical or acoustic imager such as the drill core features from core images using a Hough transform
Televiewer system, which produces 360° images from inside technique. This approach produced orientations similar to
a drill hole (Trofimczyk and Du Pisani, 2009). These images those measured by a geologist but also identified a number
are then unfolded, and the amplitude of the sine wave fitted of false features.
to each fracture is used in combination with the orientation Automated core scanning systems are becoming increas-
of the borehole to calculate the fracture orientation (Ureel ingly popular for geologic assessment and logging (e.g., Hun-
et al., 2013). While these methods produce robust fracture tington et al., 2006; Tappert et al., 2011; Fresia et al., 2017).
Several projects have used automated core logging technol-
† Corresponding author: e-mail, cassady.l.harraden@gmail.com ogy applied to geometallurgical studies for the assessment
ISSN 0361-0128; doi:10.5382/econgeo.4649; 17 p.
1 Published online: May 17, 2019

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 2 HARRADEN ET AL.
of mineralogical textures as they relate to processing (e.g.,
Arqué Armengol, 2015; Tiu, 2017). A number of commercial
systems incorporating a series of mineralogical, geochemical,
and core surface topography sensors are currently on the mar-
ket. These can analyze large volumes of drill core rapidly and
consistently.
This paper presents a methodology that seeks to enhance
and streamline the current, manual data collection tech-
niques by using automated core logging technology. The goal
of this study was to develop a methodology that could be
incorporated into an automated hyperspectral core logging
system workflow to extract geotechnical index parameters.
The protocols proposed here represent a proof-of-concept
study that utilizes manual steps to demonstrate the applica-
tion of automated core logging technology to geotechnical
assessment.
Geotechnical Indices
In order to safely and effectively extract ore deposits from
underground mine workings, detailed geotechnical models
are required. Together with geologic models, geotechnical
models are 3-D representations of the variability of structural,
mineralogical, and rock mass characteristics in space (Hoek
et al., 2000). Two primary rock classification systems are com-
monly used to assess geotechnical ground conditions: (1) rock
mass rating (RMR), and (2) rock tunneling quality (Q) index.
The RMR is used, largely, to estimate caveability of the rock
mass, while the Q index is used to predict ground support
requirements in development areas.
Rock quality designation (RQD) is a common index used
as a crude estimation of the rock conditions from drill core
(Deere et al., 1967). The RQD value is calculated by measur-
ing the percentage of core that is unbroken (>10 cm in length)
as shown by the following equation:
                 (Σ length core > 10 cm)
RQD(%) = —————————— × 100%. (1)
                 (total length core run)
The RQD is a rock quality description on its own but is also
a key component in the RMR and Q index calculations.
RMR is calculated by categorizing six parameters: (1) uni-
axial compressive strength (UCS), (2) RQD, (3) spacing of
joints, (4) condition of joints, (5) groundwater conditions, and
(6) orientation of joints. In the case of the UCS parameter,
laboratory tests are required to calculate this value; however,
on-site testing tools (such as a point load tester) can be used
to produce UCS proxy values (Rusnak and Mark, 1999). The
remaining parameters are either measured directly from the
drill core (such as RQD and spacing) or are observed and
assigned a value based on a series of criteria outlined by Bien-
iawski (1989). For each parameter, a range of measured or
observed values defines the number rating. These number
ratings are then summed to obtain the final RMR value for the
examined interval. Bieniawski (1989) defined the excavation
type and ground support requirements based on the range of
calculated RMR values.
The Q index is defined as a numerical value from 0.001 to
1,000 with the following geotechnical inputs: (1) RQD, (2)
number of joint sets (Jn), (3) joint roughness (Jr), (4) joint alter-
ation number (Ja), (5) joint water reduction factor (Jw), and
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
(6) stress reduction factor (SRF). The relationship between
these six parameters is given by the following formula:
                      RQD   Jr     Jw
Q index = ——— × — × ——, (2)
                       Jn    Ja   SRF
where RQD/Jn is an approximation of block size, Jr/Ja is an
estimate of the inner block shear strength, and Jw/SRF repre-
sents a total active stress indicator (Barton et al., 1974).
While the RQD, RMR, and Q index calculations refer spe-
cifically to joints, joints are simply one type of discontinuity
observed in drill core. The general term fracture will be used
here to reference all types of naturally occurring breaks in the
drill core, including joints, faults, cracks, fractures, and other
discontinuities of unknown geologic origin that could affect
the geotechnical behavior of the rock mass.
Cadia East Study Site
The Cadia East Au-Cu porphyry deposit is located approxi-
mately 20 km south of Orange, New South Wales, Austra-
lia, and is wholly owned and operated by Newcrest Mining
Limited. The current resource estimate of 3 billion tonnes
(Gt) at 0.37 g/t gold and 0.26% copper contains over 35 Moz
of gold (Newcrest Mining Limited, 2018). This resource
represents the largest of the four Late Ordovician to early
Silurian porphyry Au-Cu deposits constituting the Cadia
district, with the Cadia East deposit containing over 70% of
the total gold in the district (Harris et al., 2010; Fox, 2012).
The Cadia East deposit is located in a zone of steeply N-
and S-dipping sheeted quartz-calcite-feldspar veins contain-
ing bornite-­chalcopyrite  ± molybdenite. This mineralization
occurs within the Late Ordovician Forest Reefs Volcanics and
is spatially and temporally associated with the Late Ordovi-
cian to early Silurian Cadia Intrusive Complex (Wilson, 2003;
Kitto, 2005; Harris et al., 2010). Re-Os dates from molybde-
nite yield a mineralization age of 441.8 ± 1.4 Ma (Wilson et
al., 2007). The deposit is bounded to the north and south by
steeply dipping normal and reverse faults, respectively (Kitto,
2005). Two zones of mineralization are observed within the
deposit area: (1) an upper, copper-dominant zone hosted in
a 200- to 300-m-thick package of the Forest Reefs Volca-
nics and (2) a deep, central, gold-rich zone characterized by
sheeted veins associated with the Cadia Intrusive Complex
and hosted within the Forest Reefs Volcanics (Wilson, 2003;
Kitto, 2005). The Cadia East deposit is contained in a geo-
logically and structurally complex environment. The mine is
an underground block caving operation, and characterization
of the geologic and geotechnical characteristics of each litho-
structural domain is required.
The Cadia East deposit has been divided into nine distinct
geotechnical domains based on the key rock mass characteris-
tics that affect geotechnical behavior: structure, rock type, and
general rock mass quality (C. Rossimel and J. Lett, pers. com-
mun., 2012). Currently, the Cadia East geotechnical model
is built by calculating modified versions of the RMR and Q
index values for each domain from dill core measurements.
Modeled values are then cross-checked against underground
observations as drive development advances (C. Rossimel and
J. Lett, pers. commun., 2012). Manual geotechnical logging
is completed on site by geotechnical engineering and geology
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 3

staff. Information about core recovery, RQD, fracture min-
eralogy, microfracture intensity, intact rock strength, fracture
spacing, number of fracture sets, fracture roughness, and
fracture fill thickness over a 1- to 4-m interval is recorded.
The precision for geotechnical logging is dependent on each
logger’s experience, and, therefore, the relevant metrics are
not always captured accurately.
The logistical challenges involved with manually logging
drill core for geotechnical purposes prevent all holes drilled at
Cadia East from being logged from top to bottom for geotech-
nical parameters. Key drill holes that are expected to intersect
challenging rock types as well as areas requiring more data are
prioritized for geotechnical logging. This ensures that suffi-
cient data are available for input into the geotechnical model.
Unfortunately, inconsistencies in manual core logging tech-
niques (including variable experience and training and per-
sonal observation bias) mean that not all the data collected
from drill core can be used in the geotechnical model. Incon-
sistencies in the data collected manually from drill core mean
that the underground geotechnical observations weigh heavily
in determining robust RMR and Q index values. While this
dependence on underground measurements has been and
is currently applied to mining successfully, the geotechnical
characteristics would be more useful if accurate models were
available at an earlier stage in the project.
Automated Core Logging Technology
Recent technological developments in increased processing
capacity, high-precision robotics, and high-resolution spec-
trometers has resulted in a new generation of high-speed
hyperspectral core logging systems. The application of these
multisensory, automated platforms represents a key advance-
ment in comprehensive, drill core-based, mine-scale rock
characterization studies (Jackson et al., 2018). The scanning
capabilities and sensor array of automated hyperspectral
imaging systems allow for the rapid, nondestructive analysis
of drill core to produce high-resolution laser profiler images
representing a digital surface model of the drill core (Fig. 1).
Several core scanning systems incorporating a series of
mineralogical and core surface topography sensors are cur-
rently on the market, including the HyLogger, SisuROCK,
and Corescan (Table 1; Mason and Huntington, 2010; SGS,
2014; Specim, 2014). These systems incorporate a number
of sensors that collect data, including core photography and
hyperspectral-based mineralogical information, and can ana-
lyze large volumes of drill core rapidly and consistently. The
robotics utilized by these systems allow data collected from
the onboard sensors to be coregistered in space.
Minerals suitable for hyperspectral analysis exhibit distinc-
tive absorption patterns over different wavelength values of
the electromagnetic spectrum. These distinct patterns can be
used for identification of the minerals present (Thompson et
al., 1999). In general, hydroxide (OH), water (H2O), and car-
bonate (CO3) bonds are active in the visible near-infrared and
short-wave infrared (VNIR-SWIR) range (450–2,500 nm).
Minerals with very similar compositions and lattice structures
(e.g., illite and sericite) may display similar absorption features
but can often be distinguished by subtle shifts in the positions,
depths, or shapes of these features (Thompson et al., 1999).
Certain mineral groups that do not have distinguishable fea-
tures in the VNIR-SWIR range may have unique spectral fea-
tures in longer wavelengths (thermal and long-wave infrared)
(Green and Schodlok, 2016; Shodlok et al., 2016).
Methods
Recent advances in automated core logging technology pro-
vide a unique opportunity to extract key geotechnical index
parameters as they relate to the ground conditions in a mining
environment. To demonstrate the advantage of this opportu-
nity, a methodology that seeks to enhance and streamline the
current manual data collection techniques by using automated
core logging technology was developed and is summarized
from Harraden (2018). For this study, a 76-m interval of a
Cadia East drill hole was used to develop and test the perfor-
mance of the automated methodology. This sample contains
sections of the Forest Reefs Volcanics and Cadia Intrusive

Pixel Height
(cm)
7

montmorillonite
Fe oxide
chlorite
sericite
quartz
no spectral signature
0
A B C
Fig. 1. Example image showing automated core logging outputs for half of a meter of HQ core. The system simultaneously
collects high-resolution true-color photography (A), surface topography from a laser profiler (B), and hyperspectral imagery
data (C). The mineralogical products interpreted from the hyperspectral data can be draped over the surface topography to
observe the relationship between the fractures and mineralogy (modified from Harraden et al., 2017).

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 4

by Miami University user

Table 1. Summary of the Specifications, Advantages, and Challenges of the Hylogger, SisuROCK, and Corescan Automated Hyperspectral Core Logging Systems

Core
Core topography
Photography Spectral Spectral topography height
Developed pixel size resolution pixel size pixel size resolution Scan speed
System name by (µm) Spectral range (nm) (nm) (mm) (µm) (µm) (m/day) Advantages Challenges

HyLogger-3 CSIRO 100 400–1,000 (VNIR) 4 10 × 14 mm 62.5 250 700–1,0001 Low operating cost, Low spatial
field of view at longer wavelength resolution for
1,000–2,500 (SWIR) 4 4 mm sampling spectrometer photography
interval and spectral
6,000–14,500 (TIR) 25 nm at analysis
6,000 nm to
120 nm at
14,500 nm

SisuROCK Specim 16–1601 400–1,000 (VNIR) 2.8 0.09–0.641 n/a n/a 1,200 avg Two scanning speeds No core
(high resolution vs. topography
970–2,500 (SWIR) 10 0.2–21 high speed), longer system
wavelength
HARRADEN ET AL.

8,000–12,000 (LWIR) 100–400 1.7 spectrometer

Corescan HCI-3 Corescan 50 448–1,000 (VNIR) 3.84 0.5 200 15 200–1,0001 High spatial and Lower acquisition
spectral resolution for rates for higher
1,000–2,500 (SWIR) 3.84 hyperspectral data, spatial resolution
core topography settings
imaging, high
resolution core height

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

Compiled from Mason and Huntington (2010), SGS (2014), and Specim (2014)
Abbreviations: CSIRO = Commonwealth Scientific and Industrial Research Organisation, LWIR = long-wave infrared, n/a = not applicable, SWIR = short-wave infrared, TIR = thermal infrared, VNIR
= visible near-infrared
1Variable, depends on operating speed and conditions
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 5
Complex. These two rock units represent the dominant host
rocks of the Cadia East deposit.
When selecting the automated core logging technology to
be used in this study, the resolution and imaging capacity of
the core logging system was considered. The Corescan system
utilizes the Hyperspectral Core Imager Mark-III (HCI-3)
logging technology to collect high-resolution, true-color pho-
tography, 2.5-D laser height profiles, and VNIR-SWIR spec-
tra (SGS, 2014). The photography is collected at a resolution
of 50 µm per pixel, while the laser profile data is collected
at a pixel size of 200 µm with a vertical resolution of 15 µm
(Table 1). VNIR-SWIR spectra is collected at a 3.84-nm spec-
tral resolution with a 500-µm pixel size (Corescan, 2018). In
total, 23 trays of HQ drill core were scanned using the HCI-3
automated core logging system. The spatial and spectral reso-
lution of this system provides high-resolution surface topogra-
phy and mineralogy data, making the HCI-3 system ideal for
testing geotechnical assessment protocols.
Photography, laser profiler data, and hyperspectral images
are coregistered so that the pixels of all three data sets cor-
respond to the same points in space. All three data outputs
are spatially referenced so that the x-, y-, and z-values rep-
resent true measurements in meters. In the HCI-3 system,
x measures the width of drill core referenced to the center
line, y measures the down hole depth, and z measures the
core height in the tray and the digital surface topography. For
geotechnical analysis, the laser profiling data are treated as a
surface topography, and a series of spatial analysis tools are
applied to locate and estimate geotechnical characteristics
(e.g., fracture location, orientation, and roughness) of frac-
tures in the core.
Minerals that are not active in the VNIR-SWIR range or are
low in reflectance do not produce distinctive and unique spec-
tra. Quartz is typically not active in the VNIR-SWIR range;
however, hydrous species (such as opal and chalcedony) can
be identified, because these minerals exhibit H2O features in
the SWIR region (Thompson et al., 1999). Minerals such as
feldspar and sulfides do not give VNIR-SWIR spectra with
identifiable and characteristic absorption features and, there-
fore, cannot be identified using VNIR-SWIR alone. Where
VNIR-SWIR spectra with no absorption features are col-
lected, the system assigns pixels as aspectral. No assumptions
about the mineralogy of aspectral pixels were made, and this
information was not used in the fracture mineralogy estimates.
To extract the parameters required to calculate the RMR
and the Q index from automated core logging data, the down-
hole location, orientation, and morphology of fractures need
to be recognized and measured. When estimating geotech-
nical parameters automatically, the mode of data collection
should be carefully considered. Normally, hyperspectral core
imaging is conducted on halved core, as it provides a flat sur-
face and constant core height to ensure that the focal length
of the sensors is consistent. However, for the geotechnical
method used here, full core must be scanned so that fracture
orientations can be determined, and the orientation mark on
the surface of the drill core must be visible so that the original
orientation of the fracture can be calculated. In addition, the
cutting and sampling of the drill core induces rock breaks not
related to natural fracture patterns, and these modifications of
the core make geotechnical analysis less robust.
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
The geotechnical parameters required for RMR and Q
index estimation that can be assessed from automated core
logging data are: (1) RQD, (2) fracture orientation, (3) num-
ber of fracture sets, (4) fracture roughness, (5) fracture spac-
ing, (6) fracture mineralogy, and possibly (7) fracture aperture
(i.e., the width of open fracture perpendicular to the fracture
surface). Other parameters such as UCS, groundwater condi-
tions, and SRF require independent testing and need to be
evaluated separately. The automated extraction of fracture-
related geotechnical parameters from these data is achieved
through a series of postprocessing protocols (Fig. 2). These
protocols and the criteria used to extract each geotechnical
parameter, as well as the automated core logging data used in
each step, are summarized in Table 2.
Fracture recognition
Since the fractures within drill core represent discontinuities
in what should be a relatively consistent cylindrical shape,
any deviation from the typical cylinder represents a potential
fracture. From the laser profiler data, slope values were cal-
culated across the surface of the drill core. In spatial analysis,
slope derivatives are obtained by calculating, on a pixel-by-
pixel basis, the maximum rate of change of height (z-value)
of one pixel compared to its neighbors (Zevenbergen and
Thorne, 1987). Regions of drill core where the calculated
slope is steeper than the curvature of the surface of the drill
core represent a discontinuity, such as a natural fracture or a
break resulting from mechanical damage. A 2-D plane was
fitted to the z-values from the profiler data within a 3 × 3
neighborhood of pixels. The maximum rate of change of this
plane is then calculated using the average maximum approach
described by Burrough et al. (1998). The resulting slope value
Recognize
fractures
Filter out
mechanical
breaks
Calculate α = 40
fracture β = 160
orientation
Calculate
fracture
roughness
Calculate
number Jn = 2
fracture sets
Measure
fracture 2.5 cm
spacing
Fig. 2. General processing steps required for the automated extraction of
fracture-related geotechnical index parameters (modified from Harraden et
al., 2016).
 6 HARRADEN ET AL.

Table 2. Summary of the Protocols and Criteria Used to Extract Each Geotechnical Parameter from Automated Core Logging Data

Geotechnical parameter Core logger data set(s) used Methods and criteria

Fracture recognition Core topography Steep slope + not facing right (90º) or left (270º)
Only recognized if
>20 pixels identified
all gaps between pixels <25 mm
pixels cover >50% width drill core
orientation line visible
Filter out mechanical breaks Core topography, core photography Flagged if fracture is
at beginning or end of core tray
<3 cm from driller’s mark
<3 cm from core block
Logger evaluation discriminates natural from mechanical fractures
Fracture orientation Core topography Fit 3-D plane to fracture points by LSR
Spatially reorient 3-D plane by 2-D linear transformations
Fracture roughness Core topography Model 3-D surface from fracture points
Calculate true and nominal surface areas
Calculate Rs roughness proxy by equation (3)
Fracture sets Core topography Plot calculated orientations on equal area stereographic diagram
Use density of orientations to determine number fracture sets
Fracture spacing and frequency Core topography Use y-value in meters as location of fracture
Calculate average distance between fractures per interval
Calculate average number of fractures per interval
Fracture mineralogy Core topography, hyperspectral mineralogy Query each fracture pixel for hyperspectral mineralogy
Determine proportion group H, S, and V minerals in each fracture
Measure apparent aperture and calculate true aperture by equation (4)

Specific data sets used for each parameter are also included
Abbreviations: H = hard, high-friction, nonswelling minerals, LSR = least squares regression, S = soft, moderate- to low-friction, nonswelling minerals, V =
very soft, low-friction, swelling minerals

(in degrees) is assigned to each pixel based on this relationship
(Fig. 3A, B).
To remove steep slope values associated with the curvature
of the cylindrical drill core, the downslope direction of the
best-fit plane for a 3 × 3 neighborhood of pixels (also known
as aspect or facing direction) was calculated based on 360° car-
dinal direction system (Fig. 3C). The edges of the core have
steep slopes relative to the center of the core and face either
right (90°) or left (270°) in the laser profiler image. Filtering
out pixels with aspect values near 90° and 270° excludes those
pixels with steep slope values that are associated with the cur-
vature of the drill core and not discontinuities. The per-pixel
combination of laser profiler image slope and aspect allows
for identification of discontinuities in the drill core (Fig. 3D).

A B C D
Slope Aspect
0° 0°

5 cm
90° 360°
Fig. 3. Visualizations of the processing outputs of the fracture recognition protocols on 10 cm of whole HQ drill core. (A)
Red-green-blue image of the drill core shows the locations of the fractures. The orientation line (indicating the location of
the bottom of drill hole) is indicated by the orange arrow. (B) Fractures return high slope values, but the curved sides of the
core surface also return high slope values. (C) The aspect analysis identifies those pixels which face directly right (90°) and
directly left (270°). (D) By isolating the pixels with high slope values and removing those with aspects near 90° and 270°, the
pixels associated with fractures are identified.

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 7

Additional selection criteria were applied to ensure that
fracture selection was as robust as possible. A fracture was
considered recognized if (Table 2) (1) at least 20 pixels in total
are identified, (2) the largest gap between pixels across the top
of the drill core is less than 25 mm, (3) the pixels cover at least
50% of the distance across the width of the drill core (greater
than 30 mm in the x-direction), and (4) the orientation line is
clearly visible on the core surface. Figure 4 shows examples of
fractures that were rejected because they did not meet these
selection criteria.
Identifying mechanical and other nonfracture breaks is
important, since these features do not represent natural frac-
tures. The decision to discriminate these two types is based
largely on the experience of the geologist or geotechnical
engineer. Mechanical breaks are often rough and either per-
pendicular to the long core axis or at orientations uncommon
to structures in the deposit. To filter out mechanical breaks,
manual evaluation by a geologist or geotechnical engineer
is required. To aid in the distinction between natural and
mechanical fractures, a flagging protocol was developed based
on several criteria (Table 2). Potential mechanical breaks are
flagged if they occur (1) at the beginning or end of a core tray
row, (2) less than 3 cm away from a visible driller’s mark (in this
case, a red “x”), or (3) less than 3 cm away from a core block.
Fractures meeting these criteria were flagged and inspected
visually using the visible light core photography to determine
if they were natural fractures or mechanical breaks.
Fracture orientation
From the laser profiler data, the orientations of fractures are
determined by fitting a 3-D plane through recognized frac-
ture pixels. This is accomplished by analyzing the x-, y-, and
z-values of the selected pixels and applying a least squares
regression to calculate the orientation of the best-fit plane
(apparent orientation).
Once the orientation of the apparent fracture plane in the
scanned image is calculated, the original orientation of the
plane is spatially reoriented from the trend and plunge of the
drill hole and the position of the orientation mark (Fig. 5).
This is accomplished by a series of three 2-D linear rotations
(1) about the x-axis, (2) about the y-axis, and (3) about the
z-axis. The result is a true dip and dip direction for the identi-
fied fracture. The resulting true orientations are then used
in the RMR calculation and subsequent processing steps. A
summary of the fracture orientation protocols and criteria
used is outlined in Table 2.
Fracture roughness
In the Q index system, Barton (1976) defined six general
classifications of fracture roughness characteristics based on
geometry and general roughness. By modeling the 3-D frac-
ture surface using the recognized fracture pixels (Fig. 6A), a
proxy for roughness (Rs) is calculated by comparing the cal-
culated, 3-D surface area of the fracture to the nominal area
(or 2-D footprint area) of the fracture (Fig. 6B). The Rs value
is calculated by the following equation:
                  true 3–D surface area
Rs = ——————————. (3)
                     nominal area
This Rs value is modified from the 2-D real length profile
value, which is commonly used in fracture roughness calcula-
tions (Yu and Vayssade, 1991). The input value for roughness
when calculating the Q index uses a qualitative system where
the logger visually compares the fracture roughness to a refer-
ence roughness profile to assign a roughness value, so a proxy
value should be sufficient when modeling Q index parame-
ters. The fracture roughness protocols and criteria used are
outlined in Table 2.
Fracture sets
The number of fracture sets present in a given interval of drill
core is a key geotechnical parameter for the Q index. The
pole to plane for each calculated fracture orientation is plot-
ted on an equal area stereographic diagram. The distribution
of poles to planes on the stereographic diagram can then be
used to determine the number of fracture sets present in a
given group of fracture orientations (Table 2).

A B C

2 cm

Fig. 4. Examples of natural fractures that met two of the three fracture recognition criteria but were not considered recog-
nized, as they failed to meet a third criterion. Selected fracture pixels are shown in red. (A) Fracture that failed to have more
than 20 pixels identified. (B) Fracture with gaps greater than 25 mm between pixels. (C) Fracture with less than 50% coverage
across the surface of the drill core. Note: the blue line (B) and green lines (A and C) are orientation lines (recording the bot-
tom of the drill hole). The red line in C is the field orientation line that was later corrected in the core shed (green line in C).

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 8 HARRADEN ET AL.

41/90

55/93
64/101
59/112
56/273
55/227

34/105
71/226
54/226
27/88 76/245
39/112

83/101
80/105
A B C
Fig. 5. (A) Sequence used in calculating true orientation of fractures shown against a visible-light image draped over the drill
core surface topography. (B) Using the automated core logging data outputs, the pixels along the fracture are recognized, and
a least squares regression is applied to calculate the apparent orientation of the fractures in the core (green numbers). (C) A
series of 2-D rotations are used to account for the position of the orientation line (green line) and the trend and plunge of the
drill hole. The result is the true fracture orientation (blue numbers).

Fracture spacing and frequency Fracture mineralogy

The average fracture spacing is required to calculate RMR.
The spatial location coordinate (y-value, in meters) where
the fracture intersects the orientation line is measured, then
exported into a table for each fracture. Fracture spacing is the
average distance between fractures for a given, continuous
interval. The fracture frequency is equal to the average num-
ber of fractures per meter or to another user-defined interval
(Table 2).
A the hyperspectral system into three main groups (Table 3):
true surface
area
nominal area
B VNIR-SWIR hyperspectral match values are used, as these
Fig. 6. Example of a 3-D surface created from the recognized fracture points.
The gray points represent the fracture pixels used to create the 3-D surface.
By comparing the calculated, 3-D surface area (A) to the nominal area (or
2-D footprint area; B) of the fracture, a roughness proxy can be calculated.
The geotechnical behavior of a fracture is strongly influenced
by the properties of the minerals present within and proximal
to that fracture. The mineralogy within and near the fractures
is important because fractures represent planes of weakness.
Any degradation of the integrity of the rock near this weakness
due to the presence of soft, swelling, or low-friction minerals
greatly affects the overall geotechnical properties of the rock
mass (Pitts, 1985). To understand the geotechnical properties
of a rock mass, the hardness, swelling potential, and friction
potential of minerals present in and around fractures, and
their relative proportions, are critical. Based on their geotech-
nical properties, the authors divided the minerals detected by
(1) hard, high-friction, nonswelling minerals (group H); (2)
soft, moderate- to low-friction, nonswelling minerals (group
S); and (3) very soft, low-friction, swelling clays (group V).
In the RMR geotechnical system, one of the key param-
eters is fracture condition, which is calculated by assigning
values for three fracture parameters: (1) infilling, (2) weather-
ing, and (3) separation (aperture) (Bieniawski, 1989). The Q
index requires an input of the fracture alteration (Ja) value,
used to describe the mineralogy and weathering of a fracture
(Barton et al., 1974). Fracture mineralogy is a key parameter
in determining RMR infill and weathering and the Ja mineral-
ogy and weathering. For fracture mineralogy assessment, the
represent the relative proportion of each infrared-active min-
eral reported in a pixel. Fracture pixel locations are queried
for mineralogy from the coregistered mineralogical and laser
profiler outputs to determine the fracture and near-fracture

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 9
Table 3. Geotechnical Properties of Minerals Detected by the Hyperspec-
tral VNIR-SWIR System.
Mohs Swelling Low-friction Geotechnical
Mineral name hardness potential potential mineral group
Amphibole 5–6 H
Apophyllite 4.5–5 H
Epidote 6–7 H
Prehnite 6–6.5 H
Quartz 7 H
Tourmaline 7–7.5 H
Carbonate 4 X S
Clinochlore 2–2.5 X S
Chlorite 2–2.5 X S
Iron carbonate 3.5–4.5 X S
Iron oxide 4.5–5.5 X S
Kaolinite 2–2.5 X S
Phlogopite 2–2.5 X S
Sericite 2.5 X S
Dickite 1.5–2 X S
Gypsum 2 X X V
Laumontite 3.5–4 X X V
Montmorillonite 1–2 X X V
Nontronite 1.5–2 X X V
Vermiculite 1.5–2 X X V
Mohs hardness from Mohs (1825); swelling potential from Sabtan (2005) and
Rauh and Thuro (2007); low-friction potential from Barton et al. (1974)
Abbreviations: H = hard, high-friction, nonswelling minerals, S = soft, mod-
erate- to low-friction, nonswelling minerals, SWIR = short-wave infrared,
V = very soft, low-friction, swelling minerals, VNIR = visible near-infrared
mineralogy. To assess weathering effects beyond the extent of
the fracture surface, the surrounding mineralogy is assessed.
In this study, a search radius of 5 mm was used to evaluate the
changes in mineralogy away from the fracture surface. This
search distance effectively maps fracture mineral assemblages
that do not differ significantly from the wall-rock mineralogy.
The determination of fracture aperture (which does not
include fracture mineralogy information but is required by
the RMR fracture condition parameter) is a two-step pro-
cess: (1) measure the apparent thickness of the fracture and
(2) convert apparent thickness to true aperture. The apparent
thickness of the fracture is calculated by multiplying the num-
ber of pixels across the fracture in the y-direction by the pixel
resolution (in this case 200 µm). In most cases, fractures are
at an oblique angle to the drill core; therefore, to calculate the
true aperture of a fracture its orientation is required. The true
aperture is calculated by the following equation:
t = L cos ρ, (4)
where t = true aperture, L = apparent thickness, and ρ = angle
between the pole to fracture plane and the drill hole (Charles-
worth and Kilby, 1981). A summary of the protocols and cri-
teria used to extract geotechnical parameters related to the
fracture mineralogy is outlined in Table 2.
Results
Twenty-three trays (76 m) of HQ drill core were scanned by
the HCI-3 system, and 199 fractures were manually identi-
fied within this interval. The results of the automated fracture
protocols were compared to expected values (i.e., based on
previous geotechnical studies of the deposit), photographs of
the fractures, or, in the case of orientations, measured values.
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
Fracture recognition
The majority (75%) of fractures identified by manual logging
were successfully recognized by the automated protocol. After
eliminating fractures that did not meet the selection criteria
and those that were flagged and deemed to be mechanical
breaks, 150 of the 199 fractures contained in the Cadia East
comparison data set were successfully identified. Some false
fractures were automatically detected by the automated frac-
ture detection protocols. Initially, 295 discontinuities were
automatically identified in the 76-m interval of drill core. After
filtering based on the fracture selection criteria (as described
in Table 2), 150 of those discontinuities were deemed to be
natural fractures, 113 were eliminated as mechanical breaks,
and 31 were classified as false positives.
Fracture orientation
The orientation of fractures in the drill core were measured on
site manually using the industry standard procedures for manual
core orientation measurements outlined in Holcombe (2013).
To assess the difference in orientation between the measured
and calculated fracture orientations, the dot product of two
unit vectors representing the orientation of the two planes was
used (Thomas, 1960; Groshong, 2006). Errors associated with
measuring fracture orientations manually are often the result
of an incorrectly placed orientation line, inaccurate selection
of the bottom of the fracture ellipse (especially as the fracture
approaches 90° to the core axis), and rounding to the nearest 5°
when using fracture measurement tools. Based on the findings
of Vearncombe (2013), the highest acceptable error between
the calculated and measured orientations should be 25°. Of the
150 fractures recognized from the automated core logging data,
111 (74%) have calculated orientations within 25° of the mea-
sured orientation value (Fig. 7).
Fracture roughness
A subset consisting of ~10% of the manually derived compari-
son data set was used to compare the 3-D roughness calculation
protocol results with a visual inspection of the fractures from
the high-resolution photography. The automatically calculated
Rs values provided reasonable roughness rankings when com-
pared to a visual roughness ranking assessment (Fig. 8).
Fracture sets
The calculated number of fracture sets is a function of the
calculated orientations of fractures within a given interval.
The measured fracture orientations were compared to the
calculated orientations for the Forest Reefs Volcanics and
the Cadia Intrusive Complex (Fig. 9). For both rock types,
stereographic projections for measured and calculated orien-
tations show similar patterns; however, the comparison data
set is limited. To adequately assess the number of fracture
sets present, more fracture orientations from a larger spatial
extent are required. Despite the limited comparison data set,
these results show a pattern similar to the measured fracture
orientations for the same data set.
Fracture spacing and frequency
For the 76-m drill core test interval, the fracture spacing
and frequency protocols (Table 2) returned an average of 2.4
fractures/m for the Forest Reefs Volcanics with an average
 10 HARRADEN ET AL.
Fracture Recognition Protocol
40
35
Maximum
error of 25°
30
25
Frequency
20
15
10
5
0
5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0
0- 5-1 0-1 5-2 0-2 5-3 0-3 5-4 0-4 5-5 0-5 5-6 0-6 5-7 0-7 5-8 0-8 5-9
1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8
Angle Between Measured and Calculated Orientation (°)
Fig. 7. Histogram of the angle between the measured and the calculated fracture orientations.
fracture spacing of 0.5 m. The Cadia Intrusive Complex
yielded 4.4 fractures/m with average fracture spacing of 0.2 m.
C. Rossimel and J. Lett (pers. commun., 2012) reported the
average range of fracture frequencies for the Forest Reefs
Volcanics in the Cadia East mine area to be between 2 and 4
fractures/m. The Cadia Intrusive Complex is expected to have
between 2.5 and 4 fractures/m (C. Rossimel and J. Lett, pers.
commun., 2012). While the fracture frequency estimated
from the automated fracture recognition protocols is higher
than expected for the Cadia Intrusive Complex rocks, it is
close to the expected range determined for both rock units
within the Cadia East mine area.
Fracture mineralogy
As fracture mineralogy information was not manually col-
lected on the comparison data set, a visual assessment from
drill core photographs was carried out. A subset of 17 fractures
representing ~10% of the comparison data set was selected.
This subset covers a range of fracture mineralogy and overall
fracture conditions.
The hyperspectral mineralogy groups for each fracture are
generally in agreement with photographs (Fig. 10). For exam-
ple, fractures containing primarily group H minerals are associ-
ated with fractures that appear to have a good condition in the
photographs, whereas fractures containing higher abundances
of group S and V minerals generally appear to have an overall
poor fracture condition in the photograph. The wall-rock min-
eralogy, in most cases, is very close to the fracture mineralogy,
which is consistent with most of the fractures lacking significant
weathering along their surface. The true fracture apertures
vary and, in some cases, do not appear to correlate with the
aperture visible in the photograph. True fracture apertures are
difficult to estimate from drill core by any method.
Discussion
The goal of this study was to demonstrate that key geotechni-
cal index parameters can be estimated from quantitative core
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
n = 150
logging techniques at a scale and consistency that is otherwise
challenging for effective manual data collection methods.
To assess the performance of each of the fracture protocols,
the calculated values were compared to either measured,
observed, or expected values.
The recognition of fractures in the laser profile data identi-
fied 75% of manually observed fractures. Many of the frac-
tures not recognized in the imagery were not detected either
because the aperture was too narrow given the image resolu-
tion (200 µm) or because the core was too blocky and broken
to distinguish individual fractures. The pixel resolution of the
laser profiler data does not allow for the recognition of frac-
tures with gaps of less than 200 µm. Higher-resolution laser
scanned topography would enhance fracture recognition and
should be considered as this technology advances.
Seventy-five percent of the calculated fracture orientations
are within 25° of the measured orientation values (the maxi-
mum acceptable error; Vearncombe, 2013). The errors asso-
ciated with measuring fracture orientations both manually
and automatically are difficult to quantify since they can be
the result of numerous factors, such as (1) incorrectly placed
orientation line, (2) incorrect downhole survey, (3) inaccurate
selection of the apical trace in manual measurement, espe-
cially as the fracture approaches 90° to the core axis, and (4)
rounding to the nearest 5° on both the protractor and wrap-
around beta ruler in manual measurements. Despite these
potential challenges, the results from the assessment of the
fracture orientation protocols clearly show that the orienta-
tion of fractures identified by the fracture recognition pro-
tocol is possible, while ongoing improvement on the quality
of structural data is required to assess its performance more
accurately.
One challenge common to both manual and automated
geotechnical logging is that fractures parallel or nearly paral-
lel to the long axis of the core are difficult to detect and orient.
In manual logging, the logger has the opportunity to pick the
core up and examine it, but parallel fractures are often missed,
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 11

Rs = 1.20 Rs = 1.30 Rs = 1.96

Rs = 2.05 Rs = 2.17 Rs = 2.47

5 cm

Rs = 7.43 Rs = 20.25
Fig. 8. Photographs of typical fractures in the test subset ordered by visual ranking of fracture roughness. Rs values below
the images are the calculated Rs roughness proxy values.

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 12 HARRADEN ET AL.

Measured Calculated

Forest Reefs Volcanics

A n = 125 B n = 125
Cadia Intrusive Complex

C n = 25 D n = 25

Fig. 9. Heat map showing the distribution of poles to fracture surfaces on a lower hemisphere equal area stereographic pro-
jection. Measured (A) and calculated (B) fracture orientations for detected fractures in the Forest Reefs Volcanics. Measured
(C) and calculated (D) fracture orientations for fractures in the Cadia Intrusive Complex. In general, the trends observed in
the calculated and measured stereographic projections are similar.

as they may not have a surface expression that is visible to the
logger. In addition, these long-axis parallel fractures are chal-
lenging to orient as they can intersect long sections of drill
core, making it difficult to physically measure alpha and beta
angles. With automated logging, only the surface directed at
the sensors can be measured; therefore, fractures that do not
have a surface expression visible to the automated system can-
not be detected or oriented.
The calculated Rs values used as a proxy for fracture rough-
ness agree with the relative roughness of fractures that are
observed visually. Many of the extreme calculated Rs values
are instances where there is a small gap across the fracture,
and the fracture recognition protocols detect fracture points
on either side of the fracture surface. In these instances, frac-
ture surface area is exaggerated relative to the nominal area,
and calculated Rs values are artificially high. One of the chal-
lenges with estimating roughness from the automated core
logging system data is that the laser profiler images only a few
millimeters into most fractures. Particularly where the frac-
ture is tight, very little of the fracture surface is imaged. While
some of the subset roughness proxy values seem unreasonable
when compared visually to the core photos, it is possible to
set Rs threshold values so that unreasonable roughness proxy
values are rejected. In order to address this limitation, a larger
sample of calculated Rs values should be evaluated to deter-
mine practical threshold values. This work represents the next
phase of research in the application and performance of this
methodology.
The trends observed in the stereographic projections pro-
duced using measured values and calculated values are visu-
ally similar, indicating that the fracture set protocols in the
automated system are successfully tracking the number of
fracture sets. The protocols estimating the number of fracture
sets rely entirely on correctly recognizing fractures and cal-
culating their orientations. The fracture orientation protocols
were generally successful, so it is expected that the calculated
number of fracture sets would be comparable to the mea-
sured values. Overall, the fracture recognition protocols were
successful, so it is expected that the fracture spacing calcula-
tions would also be successful.
In most cases, the wall-rock mineralogy as determined from
VNIR-SWIR data is similar to the fracture mineralogy, which
is consistent with most of the fractures lacking visible wall-
rock weathering. Both the RMR and Q index geotechnical
indices utilize mineral properties to determine fracture condi-
tion. Commonly, geotechnically unstable (group V) minerals,

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 13

True Aperture = 1.7 mm True Aperture = 0.1 mm True Aperture = 0.0 mm

F: F: F:
W: W: W:

True Aperture = 0.0 mm True Aperture = 0.3 mm True Aperture = 0.0 mm

F: F: F:
W: W: W:

True Aperture = 6.5 mm True Aperture = 1.4 mm

2 cm

Group H
Minerals
Group S
Minerals
Group V
Minerals

F: F:
W: W:

Fig. 10. Photographs of eight fractures in the test subset with the corresponding calculated fracture mineralogy (upper bar,
F) and immediately adjacent (within 5 mm) wall-rock mineralogy to the fracture (lower bar, W) from the hyperspectral core
logging data. The blue arrows point to the fracture that was measured. Groups H, S, and V as in Table 3.

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

by Miami University user
 14 HARRADEN ET AL.
such as montmorillonite, are difficult to distinguish visually
from more stable (group S) minerals, such as kaolinite. The
hyperspectral VNIR-SWIR system readily distinguishes these
minerals from one another; this approach potentially provides
more consistent mineral identification than visual inspection,
particularly where the minerals are mixed or difficult to distin-
guish visually. True fracture apertures calculated by the auto-
mated protocols based on mineralogy are variable. Applying
reasonable filtering threshold values for the true aperture will
ensure that erroneous aperture values are rejected prior to
RMR and Q index calculations. Assigning these values will
require analyzing a larger set of data than was used here and
will be investigated as the research moves forward.
Implications for RMR and Q index
Overall, the fracture protocols performed well when com-
pared to expected values, photographs of the fractures, or,
in the case of orientations, measured values. A comparative
study by Morelli (2015) evaluated the relative influence that
each input parameter used in the RMR and Q index calcula-
tions has on the final calculated value. This study determined
that the RMR is most heavily influenced by the RQD and
UCS values. The fracture spacing and roughness each con-
tribute less than 10% to the overall variance of the RMR (Fig.
11A). With respect to the Q index, the RQD and fracture
roughness contribute the most variance to these calculations,
with the number of fracture sets contributing less than 10% to
the overall variance (Fig. 11B). The geotechnical parameters
RQD, fracture roughness, spacing, and number of fractures
sets are measurable using an automated core logging system,
but UCS will need to be assessed outside of the automated
system. The fracture roughness contributes the second larg-
est variance to the geotechnical index calculations, and the
Fracture Spacing
Fracture Roughness
RQD
UCS
0 10
A
No. Fracture Sets
Fracture Roughness
RQD
B 0
Fig. 11. Summary of the contribution of each geotechnical index parameter to the overall variance of the RMR (A) and Q
index (B) (modified from Morelli, 2015).
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
roughness protocols resemble reasonably well measured val-
ues overall. Based on our test data set, it is therefore expected
that RMR and Q index calculations derived from these pro-
tocols will be suitable to develop robust stress models, cave-
ability models, ground support designs, and fragmentation
analysis studies at Cadia East.
Advantages of Automated Hyperspectral Imaging
for Geotechnical Analysis
High-resolution surface topography and mineralogy of drill
core provided by automated core logging systems can rapidly
capture a large volume of geotechnical data. Mine sites com-
monly prioritize manual logging of select drill core intervals,
while automated core logging methods allow for data collection
continuously and consistently down the length of the drill hole.
One advantage of the automated protocols developed in this
study is the ability to collect a higher density of data than is
feasible by manual methods. A comparison of all geotechnical
parameters collected on the test data set using the developed
methods and the select intervals of geotechnical data logged
by the Cadia East site team over the same interval is shown
in Figure 12. The data density using automated core logging
methods is greatly increased and can be collected in a much
shorter amount of time. Manual geotechnical drill core log-
ging can take roughly 5 min to an hour per meter, depending
on the condition of the drill core and level of detail required.
Automated core logging systems can capture data at a rate
of around 3 min/m, with minimal additional processing time,
regardless of the level of detail or condition of the drill core.
The scanning rate of the automated systems combined with
the increased data density ensures that geotechnical data is
collected rapidly and consistently, providing major advantages
over manual geotechnical logging methods.
RMR Comparisons
20 30 40 50 60 70
Contribution to the Variance (%)
Q-Index Comparisons
10 20 30 40 50 60 70
Contribution to the Variance (%)


by Miami University user

Automated Core Logging-derived Data Cadia East Site Data
Fracture Fracture Fracture Fracture Fracture Fracture 5 mm 10 mm Calculated Calculated Calculated Fracture Fracture Fracture Fracture
Depth (m) Orientation Roughness Sets Spacing RQD% Aperture Mineralogy Mineralogy Mineralogy RMRinfill RMRweath Ja Roughness Sets Spacing RQD% Aperture RMRinfill
0.0

10.1 n = 24

80.1

89.9
148.7 n = 43

n = 31
168.8
228.5
No data collected
No data collected
No data collected
No data collected
No data collected
No data collected

n = 31

248.2
317.7

n = 10
327.4
396.7

406.7 n = 12

Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf

0 20 0 2.5 0 100 0 20 0 100 0 100 0 100 0 6 0 6 4.0 0.75 0 20 0 4 0 2.5 0 100 0 20 0 6
JRC m % mm % % % JRC m % mm
Group H Group S Group V
AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT

A B

Fig. 12. Comparison of geotechnical parameters derived from automated core logging data using the Corescan system (A) with the data collected in the Cadia East
routine site logging campaign (B) within the same interval of drill core. Company site measurements are usually restricted to a number of short intervals similar to that
shown in (B) here. Note: for calibration purposes the entire 76 m of core was manually measured independent of the Cadia East site program. Fracture orientations
from this calibration data were shown in Figures 7 and 9 but are not included in this figure. Abbreviations: JRC = joint roughness coefficient, RMRinfill = RMR infilling,
RMRweath = RMR weathering.
15
 16 HARRADEN ET AL.
Whole, oriented drill core is required to extract fracture ori-
entations and the number of fracture sets. It may be possible,
however, to collect fracture roughness, fracture spacing, frac-
ture condition, and fracture alteration on whole, unoriented
drill core. The work by Morelli (2015) shows that the number
of fracture sets contributes less than 10% to the variability of
the RMR and Q index calculations, so it may be possible to
calculate partial RMR or Q index values (without calculating
the number of fractures sets) using whole, unoriented drill
core. This potential should be investigated, as it may provide
meaningful geotechnical information without the cost and
logistics of obtaining oriented drill core.
Conclusions
Assessing the morphological and mineralogical properties of
fractures as they relate to geotechnical indices is time-con-
suming and challenging where fractures are abundant, where
minerals are mixed or difficult to identify visually, and where
core is very broken or blocky. These factors commonly lead to
inconsistencies when assigning RMR and Q index values and
modeling the geotechnical characteristics of a deposit. Since
the geologic properties of a rock mass dictate rock behavior,
understanding these properties is vital for the production of
robust geotechnical models. Recent advances in automated
core logging technology provide an opportunity to rapidly
and consistently collect coincident surface topography (laser
height data) and mineralogical information (hyperspectral
data) from whole (uncut) exploration and production drill
core. While determining the fracture orientation and num-
ber of fracture sets requires oriented drill core, parameters
such as fracture roughness, frequency, and mineralogy can be
assessed using any whole drill core, even if it is not oriented.
By combining surface topography and mineralogical data
obtained from automated core logging systems, key proper-
ties that affect the geotechnical response of a rock mass can
be rapidly and consistently estimated. Integration of laser
height data with hyperspectral derived mineralogical data
generates large volumes of consistent data for geotechnical
assessments, increasing the underlying statistical support for
rock mass characterization.
Acknowledgments
This research project was undertaken as part of the University
of Tasmania’s ARC Research Hub for Transforming the Min-
ing Value Chain (project IH130200004). The authors would
like to thank the geology and geotechnical teams at the Cadia
mine, especially Maya Secheny, Chris Chester, Jo Morrison,
Mel Morgan, Sarah Stark, and Stephen Guy. Special thanks
should be extended to Laura Jackson, Ekaterina Savinova,
and Neil Goodey for their insights and knowledge that have
contributed to this project. We would also like to thank David
Cooke, Mary Harris, Brian McNulty, and Chip and Rebecca
Harraden for reviewing this work and providing meaningful
edits in previous versions. Karin Olson Hoal and Pim van Gef-
fen are graciously thanked for their thorough and constructive
reviews, which greatly improved this manuscript. Finally, we
would like to thank Newcrest Mining Ltd., Corescan Pty Ltd.,
and the University of Tasmania for their continued support
throughout this project.
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
REFERENCES
Arqué Armengol, A., 2015, Towards automated logging of ore: Positive iden-
tification of sulphides in the ores of Agnico Eagle Kittilä and New Boli-
den mines: Unpublished M.Sc. thesis, Luleå, Sweden, Luleå University of
Technology, 96 p.
Barton, N., 1976, Unsupported underground openings: Rock Mechanics
Meeting, Stifrelsen Bergteknisk Forskning, Stockholm, Sweden, 1976, Pro-
ceedings, p. 62–94.
Barton, N., Lien, R., and Lunde, J., 1974, Engineering classification of rock
masses for the design of tunnel support: Rock Mechanics, v. 6, p. 189–236.
Berry, R., and Nguyen, K., 2016, Structure from photographs of oriented
core: STORC: Economic Geology, v. 111, p. 1525–1527.
Bieniawski, Z.T., 1989, Engineering rock mass classifications: A complete
manual for engineers and geologists in mining, civil, and petroleum engi-
neering: New York, Wiley, 257 p.
Burrough, P.A., McDonnell, R.A., and Lloyd, C.A., 1998, Principles of geo-
graphical information systems: Oxford, Oxford University Press, 329 p.
Charlesworth, H., and Kilby, W., 1981, Calculating thickness from out-
crop and drill-hole data: Bulletin of Canadian Petroleum Geology, v. 29,
p. 277–292.
Corescan, 2018, Hyperspectral core scanning, processing and interpretation:
www.corescan.com.au.
Deere, D.U., Hendron, A., Patton, F., and Cording, E., 1967, Design of
surface and near-surface construction in rock: U.S. Symposium on Rock
Mechanics (USRMS), 8th, New York, 1967, Proceedings, p. 237–302.
Fox, N., 2012, Controls on alteration and mineralisation at the Cadia East
alkalic porphyry Au-Cu deposit: Unpublished Ph.D. thesis, Hobart, Austra-
lia, University of Tasmania, 222 p.
Fresia, B., Ross, P.-S., Gloaguen, E., and Bourke, A., 2017, Lithological dis-
crimination based on statistical analysis of multi-sensor drill core logging
data in the Matagami VMS district, Quebec, Canada: Ore Geology Reviews,
v. 80, p. 552–563.
Green, D., and Schodlok, M., 2016, Characterisation of carbonate miner-
als from hyperspectral TIR scanning using features at 14 000 nm and
11 300 nm: Australia Journal of Earth Sciences, v. 63, p. 951–957.
Groshong, R.H., 2006, 3-D structural geology: A practical guide to quantita-
tive surface and subsurface map interpretation: New York, Springer, 400 p.
Harraden, C.L., 2018, Geotechnical and grain size assessment using Corescan
automated core logging and complementary microanalytical techniques:
Unpublished Ph.D. thesis, Hobart, Australia, University of Tasmania, 396 p.
Harraden, C.L., Berry, R., and Lett, J., 2016, Proposed methodology for
utilising automated core logging technology to extract geotechnical index
parameters: Australasian Institute of Mining and Metallurgy (AusIMM)
International Geometallurgy Conference, 3rd, Perth, Australia, 2016, Pro-
ceedings, p. 119–123.
Harraden, C.L., Cracknell, M.J., Lett, J., and Berry, R., 2017, The use of
automated core logging technology to improve estimation of fracture min-
eralogy and weathering for geotechnical index calculations: Drilling for
Geology II, Australian Institute of Geoscientists (AIG), Brisbane, Australia,
2017, Proceedings, p. 73–80.
Harris, A., Cooke, D.R., Fox, N., Cuison, A., Tosdal, R., Groome, M., Percival,
I., Dunham, P.D., Collett, D., Holliday, J.R., and Allen, C., 2010, Architec-
tural controls on Paleozoic porphyry Au-Cu mineralisation in the Cadia Val-
ley, NSW: Mines and Wines Symposium 2010, Mudgee, New South Wales,
Australia, Conference Presentation, http://smedg.org.au/M&W%202010/
Harris%20Cadia%20Au-Cu%20Mineralization.pdf.
Hoek, E., Kaiser, P.K., and Bawden, W.F., 2000, Support of underground
excavations in hard rock: Netherlands, CRC Press, 225 p.
Holcombe, R., 2013, Orientated drillcore: Measurement, conversion, and
QA/QC procedures for structural and exploration geologists: Holcombe
Coughlin Oliver and Associates, p. 1–36, https://docs.wixstatic.com/ugd/baa
4fe_9d93bff830d740e8a6f724e134839054.pdf.
Huntington, J.F., Mauger, A.J., Skirrow, R.G., Bastrakov, E.N., Connor, P.,
Mason, P., Keeling, J.L., Coward, D.A., Berman, M., and Phillips, R., 2006,
Automated mineralogical core logging at the Emmie Bluff iron oxide-­
copper-gold prospect: MESA Journal, v. 41, p. 38–44.
Jackson, L.M., Parbhaker-Fox, A., Fox, N., Cooke, D.R., Harris, A.H., Meffre,
S., Danyushevsky, L., Goemann, K., Rodemann, T., Gloy, G., and Savinova,
E., 2018, Assessing geo-environmental risk using intact materials for early
life-of-mine planning—a review of established techniques and emerging
tools: From start to finish: A life-of-mine perspective: Australasian Institute
of Mining and Metallurgy (AusIMM), Spectrum Series, no. 24, p. 9–26.
 AUTOMATED CORE LOGGING, CADIA EAST PORPHYRY DEPOSIT 17
Kitto, J., 2005, Lithostratigraphy, alteration and geochemistry at the Cadia
East gold-copper porphyry deposit, NSW: Unpublished Honours thesis,
Hobart, Australia, University of Tasmania, 136 p.
Mason, P., and Huntington, J., 2010, HyLogger 2 components and prepro-
cessing: An overview: Northern Territory Geological Survey Technical
Note, v. 1, p. 7.
Mohs, F., 1825, Treatise on mineralogy (translated by W. Haidinger): Edin-
burgh, Caledonian Mercury Press, 458 p.
Morelli, G.L., 2015, Variability of the GSI index estimated from different
quantitative methods: Geotechnical and Geological Engineering, v. 33,
p. 1–13.
Newcrest Mining Limited, 2018, Annual mineral resources and ore reserves
statement—31 December 2017: www.newcrest.com.au/media/resource_
reserves/2017/December_2017_Mineral_Resources_and_Ore_Reserves_
Statement.pdf.
Pitts, J., 1985, A manual of geology for civil engineers: Singapore, World Sci-
entific, p. 17–62.
Quiniou, T., Selmaoui, N., Laporte-Magoni, C., and Allenbach, M., 2007,
Calculation of bedding angles inclination from drill core digital images:
Conference on Machine Vision Applications, Tokyo, Japan, 2007, Proceed-
ings, p. 252–255.
Rauh, F., and Thuro, K., 2007, Investigations on the swelling behavior of pure
anhydrites: Canada-U.S. Rock Mechanics Symposium, 1st, Vancouver, Brit-
ish Columbia, Canada, 2007, Proceedings, 7 p.
Rusnak, J., and Mark, C., 1999, Using the point load test to determine the
uniaxial compressive strength of coal measure rock: International Confer-
ence on Ground Control in Mining, 19th, Morgantown, West Virginia,
1999, p. 362–371.
Sabtan, A.A., 2005, Geotechnical properties of expansive clay shale in Tabuk,
Saudi Arabia: Journal of Asian Earth Sciences, v. 25, p. 747–757.
Schodlok, M.C., Whitbourn, L., Huntington, J., Mason, P., Green, A., Ber-
man, A., Coward, D., Connor, P., Wright, W., Jolivet, M., and Martinez, R.,
2016, HyLogger-3, a visible to shortwave and thermal infrared reflectance
spectrometer system for drill core logging: Functional description: Austra-
lia Journal of Earth Sciences, v. 63, p. 929–940.
SGS, 2014, Automated core logging: SGS Minerals Services Technical
Report—T3 SGS 1344, www.sgs.com/-/media/global/documents/flyers-
and-leaflets/sgs-1344-automated-core-logging.pdf.
Specim, 2014, SisuROCK hyperspectral core logger: Product brochure, www.
specim.fi/downloads/SisuROCK_Brochure-1_17_web.pdf.
Tappert, M., Rivard, B., Giles, D., Tappert, R., and Mauger, A., 2011,
Automated drill core logging using visible and near-infrared reflectance
spectroscopy: A case study from the Olympic Dam IOCG deposit, South
Australia: Economic Geology, v. 106, p. 289–296.
Thomas, G.B., 1960, Calculus and analytic geometry: Boston, Addison-Wes-
ley, 818 p.
Thompson, A.J., Hauff, P.L., and Robitaille, A.J., 1999, Alteration mapping in
exploration: Application of short-wave infrared (SWIR) spectroscopy: SEG
Newsletter, no. 39, p. 16–27.
Tiu, G., 2017, Classification of drill core textures for process simulation in
geometallurgy: Aitik mine, Sweden: Unpublished M.Sc. thesis, Luleå, Swe-
den, Luleå University of Technology, 71 p.
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user
Trofimczyk, K., and Du Pisani, P., 2009, Integration of borehole radar and
acoustic televiewer data in geotechnical boreholes—a case study of the use
of downhole geophysical data in mitigating risk to a new mining shaft devel-
opment: South African Geophysical Association (SAGA) Biennial Technical
Meeting and Exhibition, 11th, Swaziland, 2009, Proceedings, p. 149–154.
Ureel, S., Momayez, M., and Oberling, Z., 2013, Rock core orientation for
mapping discontinuities and slope stability analysis: International Journal
of Research in Engineering and Technology, v. 2, p. 1–8.
Vearncombe, J.R., 2013, Structural geology from oriented drill core: SJS
Resource Management (blog), August 20, 2013, www.sjsresource.com.au/
blog-layout/structural-geology-from-oriented-drill-core.
Wilson, A., 2003, The geology, genesis and exploration context of porphyry
copper-gold deposits, New South Wales, Australia: Unpublished Ph.D. the-
sis, Hobart, Australia, University of Tasmania, 316 p.
Wilson, A.J., Cooke, D.R., Stein, H.J., Fanning, C.M., Holliday, J.R., and
Tedder, I.J., 2007, U-Pb and Re-Os geochronologic evidence for two alkalic
porphyry ore-forming events in the Cadia district, New South Wales, Aus-
tralia: Economic Geology, v. 102, p. 3–26.
Yu, X., and Vayssade, B., 1991, Joint profiles and their roughness parameters:
International Journal of Rock Mechanics and Mining Sciences and Geome-
chanics Abstracts: v. 28, no. 4, p. 333–336.
Zevenbergen, L.W., and Thorne, C.R., 1987, Quantitative analysis of land sur-
face topography: Earth Surface Processes and Landforms, v. 12, p. 47–56.
Cassady Harraden currently works as a spectral
geologist and geometallurgist with Corescan. Her
work is focused on the interpretation of hyperspec-
tral data and developing new geometallurgical and
geotechnical applications of data collected using
automated core logging technology. She recently
finished her Ph.D. degree with the CODES-
TMVC group at the University of Tasmania in Australia. Her Ph.D. research
focused on extracting geotechnical parameters from Corescan-derived data
and assessing the ability of numerous microanalytical techniques to measure
grain size at a rate similar to Corescan. Prior to her Ph.D. studies, she worked
as an exploration geologist on projects in Alaska, Colorado, Nevada, Utah,
and Arizona.
Downloaded from https://pubs.geoscienceworld.org/segweb/economicgeology/article-pdf/doi/10.5382/econgeo.4649/4695998/4649_harraden_et_al.pdf
by Miami University user