IMPLICITY MODELING IN LEAPFROG

What is Implicit Modelling?

Implicit modelling is the fast and automated formation of surfaces such as
grade, faults and alteration directly from geological data. Explicit modelling is
what is considered the ‘traditional’ manual method of wire framing and
digitising.
Leapfrog pioneered the implicit modelling method of geological modelling,
which has been rapidly adopted around the world as it significantly speeds up
the modelling process, and allows models to be automatically updated as data
is adjusted. The Leapfrog engine takes care of the labour intensive and
repetitive work, to give you more time for interpretation and the consideration
of multiple hypotheses.

Are all implicit modelling engines created equal?

The quality of an implicit modelling engine lies in the algorithms that
determines how known data points are used to imply (or estimate) unknown
data points to create surfaces. Radial Basis Functions (RBF’s) are the basic
method to do this, however they have their limitations.
Many years of research by leading mathematicians culminated in the creation
of FastRBF™, the industry revolutionising algorithm that powers Leapfrog.
The main difference between traditional RBF’s and FastRBF™ is its ability to
deal with very large datasets of well over 1,000,000 points incredibly quickly,
and on ordinary computing hardware. FastRBF™ also has extraordinary
extrapolation capabilities, even when large gaps occur in a data set.
FastRBF™ and the algorithms behind it have been continually improved to
specialise in modelling geology since Leapfrog was launched in 2004. It is
widely regarded as the most superior implicit modelling engine on the market.

The evolution of earth modelling

The high-tech earth modelling we love today started with a little gold (a
filament to be exact) and the spirit of exploration.
In the early 1800s, continental explorers like Charles Darwin sketched maps of
land that included geological observations. These could be considered the
earliest examples of geological modelling – as well as mapping.
It’s suited that Darwin also contributed to the evolution of earth modelling with
his early geological observations. Image credit: John van Wyhe, ed. 2002-. The
Complete Work of Charles Darwin Online (http://darwin-online.org.uk/).
“Then, in 1833, the scientist Carl Gauss published a paper that
became famous because it talked about the Earth’s magnetic field,”
says Lorraine Godwin, Director, Segment Business Development at
Seequent.
“This would be the moment that most geophysicists would tie to the
advent of modern geophysics.”
Gauss noticed that a fine filament of gold hanging on a string seemed to
always point in the same direction – North. He’d discovered the Earth’s
magnetic field and invented a simple compass, the building block of today’s
magnetometers.
Lorraine herself has had an evolving career in geoscience technology,
including 23 years with Seequent companies. She worked in the field as a
geophysicist before moving into software. She’s watched the industry change
rapidly as new geophysical equipment, industries, and needs emerged.
With her colleague Darren Mortimer, Lorraine presented “The evolution of
earth modelling” at the virtual Symposium on the Application of Geophysics to
Engineering and Environmental Problems (SAGEEP) conference.
Here, she’ll share how – from compasses, to war submarines, to punchcards –
the world at large has shaped the geo-software that we use today.
1930s Scanning for metal with magnets
A century after Gauss, scientist Victor Vacquier discovered a new use for
compass technology: mineral exploration.
“Victor at the time was working on fluxgate magnetometers for ore
deposits,” explains Lorraine.
“When you have a magnetic ore deposit, then the Earth’s magnetic
field plus the signal strength of the deposit would add up to a higher
signal. You could measure that and see that in data.”
The early mining industry had relied mostly on luck and surface observations
to find metals. Now, magnetics gave them a better idea of what was hiding
below.
To solve a similar problem during World War II, the military found a new use
for Vacquier’s device at sea: detecting submarines.

Modern magnetic survey data, paired with software like Oasis

montaj and DAP, evolved from these early devices.
Vacquier also used this new marine application of his technology after the war
to map the magnetic patterns on the ocean floor.
“This led to the theory of seafloor spreading and was a major
impetus behind what we now know as plate tectonics: the idea that
the plates are moving and colliding, resulting in earthquakes and
mountains ranges forming, and even changing the underlying
geology,” says Lorraine.

1970s Rise of the machine

Fast forward a few decades and these early discoveries in magnetics have
transformed into calculators and basic computers.
“Around this time, a geologist had a hammer and a map, and a
geophysicist had a calculator and a computer with a stack of floppy
disks – and they had to be software developers to get the job done.”
Geo-professionals had to do more than collect and analyse the data – they
had to create the computer programs they needed to analyse and process it.
They had to become early programmers.
Geoslope (now part of Seequent) was founded by two siblings in
Saskatchewan, Canada in 1977. One was an engineer and the other was a
computer programmer. They saw an opportunity for computers to make
faster, more accurate predictions about slope stability by analysing more data
than a person could.
Punch cards used for data entry into the original Geoslope software.
“Geoslope has a long history of the engineering software business.
They’re arguably the first company worldwide to commercialize
software targeted at geotechnical engineers,” says Lorraine.
Their first two products were PC-SLOPE and PC-SEEP, which were made to run
on a personal computer (thus the “PC” names).
“And that was unique in the industry, as it represented a move away
from mainframe computing.”
Personal computers allowed more people to run programs and analyse data,
without waiting their turn on a single, shared mainframe.
“These were the days of digitisers and huge plotters, and you’d have
big, printed manuals that would help guide you through the
software.”

1980s Computers do the drawing

Early Geoslope and Geosoft software, and robust user manuals.
Meanwhile in Toronto, Canada, two geophysicists were tired of hand drawing
or “contouring” their geophysical data.
“They believed microcomputers would change the world of
geophysics and they were right,” says Lorraine.
Many early geo-measurement systems were analogue, which meant that they
just had gauges and paper recording charts.
“Geoscientists then had to take that data and manually enter it or
plot it by hand and hand contour it, which if anyone’s done it – and I
did this in my early geophysical field days – it’s really painful and
time-consuming,” explains Lorraine.
The Toronto duo wrote specialised programs on floppy disks to do their
geophysical analysis. They realised that software could be used to visualise
and interpret geophysical data and move away from hand contoured maps.
They founded Geosoft (now part of Seequent) and created Oasis
montaj geophysics software, supposedly named as an “oasis for your montage
of geophysics data.” Geophysicists rejoiced at no longer having to hand
contour data or hire artists to draw renditions of the subsurface.
Software provided a place to bring together a “montaj” of all their geophysical
data and tie it together with other geo-data such as drillholes, geochemical,
etc. for a better understanding of the subsurface.
1995 The Earth in digital 3D
Applied Research Associates in New Zealand (ARANZ, now Seequent) started
in 1995 in Christchurch, New Zealand. They originally created 3D medical
imaging software to model inside the human body using data.
It became clear this could be applied to see inside the Earth as well and solve
a big program in geological modelling.

Leapfrog software made data more visual and interactive, improving

communication.
“They were the developers of the fast RBF technology that we know
and love today in Leapfrog. The software made the modelling
process not only dramatically faster, but highly visual and dynamic.”
Geo-professionals wanted to bring all their data together, quickly, and make it
3D to clearly communicate insights with non-technical stakeholders. Much
software at the time was designed to be understood by experts only.
“At the same time, we’re seeing collaboration across multiple levels
of stakeholders: from very technical to non-technical people, like
investors, the environmental protection agency, and stakeholders
that are looking at the projects from a different perspective.”
Non-technical stakeholders didn’t need to know the nitty-gritty of the data, but
they needed to know the outcomes of it. Leapfrog’s 3D visualisation made it
much easier to communicate expert insights so everyone could make
informed decisions.
Leapfrog’s pioneering implicit modelling also helped predict the geology
between drillhole or point data, guided by the geologist’s knowledge. 3D and
implicit modelling tools were quickly adopted across industries to help better
map the subsurface.
“This was a paradigm shift in the depth of understanding of the earth
subsurface,” says Lorraine.
“It’s really allowed the integration of disparate data sets together.
Bringing our geology, our geotechnical, hydrogeology, geochemistry,
or geophysics together, and gave us deeper insight into the
subsurface.”

The future of earth modelling

Geological models can now update in near real time as teams around the world add and
refine their data.
In 2017, Seequent was founded to bring together geo-data of all types, to help
everyone gain a better understanding of the Earth.
“The real power is not just in the data. It’s in connecting that data
together with the knowledge and insights that we get by connecting
different, disparate data types together,” says Lorraine.
Teams began growing and diversifying, and there was a need to find a way to
make better decisions together – no matter their expertise or location.
“Today, geoscientific software is allowing teams to work
collaboratively from a single source of truth, to create more than just
3D models but what we call digital twins, a term originally from the
civil industry.”
Digital twins provide the most accurate reflection of the Earth by combining all
data and knowledge available.
Cloud-based Seequent Central came from this need for teams to collaborate
on geo-data and models in near real time, from anywhere. Which mirrored the
global shift to more remote work.
Data itself keeps growing and growing, along with new sensor types and ways
to collect it.
“So, how do we make sense of it all? That’s what our vision is for
Seequent Evo, which is an easy flow of information across all
solutions,” explains Lorraine.
“It shouldn’t matter which software solution you’re using. We want
the Evo data translation tool to be accessible to all. That’s why we’re
going to make it accessible via a secure API to any third party – even
if they’re a competitor or not.”
Seequent Evo is the connector between geo-data, software, and tools to reduce
time, risk, and project costs. It allows multidisciplinary teams to collaborate
seamlessly on models, and track their data and files.
When all data comes together, everyone gains a better understanding of the
Earth and can make better decisions.
“We’ve seen the evolution from hand-drawn maps to 3D models and
digital twins that are more than just geoscience, but incorporate
other types of data, like flow models and building information
management data,” says Lorraine.
The future of geo-modelling is connection between teams and data, still driven
by the human spirit to explore the Earth.

Why implicit modelling?

Geology is a science rather than an engineering discipline because it models

the real world instead of building structures. Models evolve with the
collection of new data and the scientist’s improving understanding.
Geologists require tools that enable them to create and compare possible
models quickly and efficiently, engineers need tools with a greater emphasis
on precision because they form the basis for precise construction. In mining,
the traditional engineering design approach of using sections, plans and
elevations to define a structure in 3D is also used to specify natural geological
structures, despite geological structures almost never being straight. The
contrast is evident in Figure 1.
Figure 1a: A comparison of engineering and geological structures.

Figure 1b

Explicit modelling
Explicit modelling is essentially akin to an engineering drawing process. The
modeller defines geological structures such as veins and faults by explicitly
drawing them on regularly spaced sections and joining them. The data may
constrain where they draw and tools may accelerate the process but
fundamentally it is a drawing process.
However, geology doesn’t come in boxes, triangles, straight lines or even
sophisticated Bezier curves, these are simply ways of representing the
geology on a computer.

Implicit modelling
Implicit modelling is generated by computer algorithms directly from a
combination of measured data and user interpretation. The modelling requires
a geologist’s insight, but this is made in the form of trends, stratigraphic
sequences and other geologically meaningful terms. This approach is faster,
more flexible and fundamentally better suited to modelling geology.
Models can also satisfy important geological constraints, e.g. lithological
units can fill the space under the ground with no gaps and spaces, cutting
through any section at any position will always be consistent with other
sections. Unlike explicit modelling where sections are created independently
and fitted together to try and create a 3D model. In figure 2 they appear
similar, however the explicit section (left) is an input used to create the model
formed by manually joining contact points, whereas the implicit section (right)
is an output created by slicing through a 3D model created directly from the
data.

Figure 2a
Figure 2b: The explicit section in the first image is created by manually joining
contact points, whereas the implicit section in the second image is created
directly from the data.

Figure 3: A complex vein system When first created, implicit models were
considered to be lacking detail and were mainly used in early exploration.
However, implicit modelling algorithms have evolved rapidly and can now
easily exceed the complexity and level of detail in hand-drawn models. This
process of improvement continues. Modellers were much more familiar with
wireframes, sections and block models than the concept of mathematical
functions. For certain practical problems, these are entirely appropriate. E.g.
modern computer graphics hardware is optimised to display wireframes, an
axially-aligned section may succinctly explain the geology and a block model
may map very closely to an open-pit mining process.
The critical advantage of an implicit model is that it allows the user to answer
a simple question: If I went to a particular point under the ground, what would I
expect to find based on the data? When we know this at a resolution
determined by the data, we can:
• Convert an implicit model into blocks by averaging all the points
within a block
• Create wireframes by connecting all the points with the same value
• Create sections by interrogating the model at a defined plane
The speed of implicit modelling means that models can incorporate the latest
data and be kept up-to-date. But another significant benefit is that more than
one geological interpretation can be considered and maintained. In future, the
principal benefit of implicit modelling will be that geological risk can finally be
quantified in a usable manner.
The famous mathematical modeller George Box said:
All models are wrong, but some are useful.
In essence, implicit modelling allows geologists to produce useful models
quickly and efficiently and then adapt them to the needs of the end-user, be
they engineers, managers or investors.

Dr. Richard Lane

Prevoiusly Research Director
Seequent

Reducing geological risk

Geologic risk is the subjective measure of the willingness of the modeller to

put all their eggs in one basket. Committing to a single model is all too often
the case in resource consulting the result of established polygon-based
modelling and estimation practices paired with time limitations.
Integrating implicit modelling into the workflow has not only expedited the
first iteration of a model, but has facilitated construction of multiple versions
and refinements to test and compare with known data. Done in the same time
a single model would once have been.
Section-view blinders
In my formative years, my primary tool for constructing geologic models was
the extruded polygon on section, well-established as providing complete
control. It was simply drawing polygons to define a body on section and
connecting those polygons with triangulated extrusions to form a solid mass
to represent geology. A simple model would take weeks, if not months, of
painstaking digitisation.

I would seldom introduce faulting or any additional shape-morphing factors

unless substantial time was available. In most cases, complexities were
ignored or simplified, the model would flex and undulate like a battered
accordion, only generally making geologic sense. It was my attempt to
simulate a complex system while working within the limitations of “section-
view blinders.”

The risk of the single model

Typically, new project data needs to be vetted, corrected, modelled, estimated
and reported within a 2-3 month window. When the raw data has been cleaned
up remaining time is a premium asset. Spending multiple weeks or months on
a polygonal model would leave insufficient time to review, refine or even
reasonably understand the deposit before a resource needs to be calculated
and reports written.

Additional hypotheses would be overlooked in favor of the first-pass model.

The polygon approach of rendering a single model to represent absolute
reality leads to a high degree of risk, as no alternative hypotheses were ever
investigated.

Taking a global view

Implicit modelling significantly reduces the degree of “geologic risk” in model
construction by allowing many models to be built in a relatively short time
while still viewing the data globally. Complex geometry becomes easily
understood if viewed globally, factors such as fault movement make much
more sense viewing the entire relationship. Complex geometry becomes much
more practical when the computer is doing most of the heavy lifting.
Questionable interpretations are quickly identified when not trapped in a
limited section-view (i.e. “section-view blinders”).
New modifications can be seen at once, quickly resolving inconsistencies and
errors. Additionally, geostatistical analysis can be performed to identify which
model is statistically the most valid and choose that with the least amount of
waste material in an ore boundary, the best variography results, the most
normal distribution of grade values, and so on. By reducing the geologic risk,
statistical risk also diminishes.

Process of refinement
I build revisions on a theme rather than drastically different and mutually-
exclusive models. An early iteration might be “usable” (comparable to a
polygonal approach) but with questions remaining. The existing model can be
copied to retain the original, then the copy refined. I may come up with half a
dozen or more iterations that are copies of copies, each one a modification of
the last.

For instance, I might have a completely un-faulted model, with a second

introducing the major faults, a third including minor faults, and fourth
incorporating dike intrusions. I can test how different hypotheses will affect
my model differently. Implicit modelling software will typically update models
automatically when new data is available, quickly proving some iterations
false once a new season of drilling or mapping data is imported, while other
modelling iterations will likely have a degree of validation. Remaining models
can be refined and expanded upon with additional models to test when new
data is available.

Channel for discussion

Creating multiple models is extremely pragmatic when communicating with a
client. With a polygonal model I would provide rough versions to which they
would advise on refining. Today, when I present a half-dozen model iterations,
it opens a channel for discussion about which is considered the best, given
their knowledge of the field aspects.

Often presenting a hypothesis not considered that may become a new target
for investigation. This ability to contribute to the conversation regarding
geologic interpretation, potentially aiding future endeavors beyond the model
estimation, is a value-added benefit I couldn’t provide without implicit
modelling.
The geology model frames all decisions in the resource extraction industry;
calculation of resources; mine planning; recovery and ultimately, profitability.
Every downstream decision is at least as wrong as the geology model,
starting with the best model is critical.
A geology model can be thought of as a single realisation of geologic
interpretation within the sample data set, permissive of many possible
interpretations. Furthermore, most geologic sample data are themselves
based on interpretation during logging. This demonstrates that more than one
model outcome is possible and more than one should be constructed and
evaluated. In fact, a final interpretation may be presented to downstream
users with error bars when more than one model is evaluated acknowledging
the risk involved.
This is ideal, where the evaluation of several models allows understanding of
project risk and likelihood of success. However, up to the present day the
connection between the economic evaluation and the underlying assumptions
are usually poorly understood. The model is usually taken as ground truth,
regardless of how well it was constructed or its risks understood. If it is
insufficiently characterised, the risk of the resulting business decision is an
unknown.

Two-dimensional process
The standard construction of a geology model typically relies on the sectional
construction of lines and polygons, either hand-drawn on paper sections or on
a computer screen. These 2D representations are then converted to 3D by
tying successive 2D slices together, like stacking slices of bread to make a
loaf.

Important geologic features may change from deposit or commodity type, but
the basic method of drawing lines, then erasing them to redraw them with the
addition of more data, has been the same. Usually, only one model will be
produced because it is simply too time- intensive. The shapes of geological
objects are frequently much more complicated than stacking slices of bread.

Enter implicit modelling

The heavy lifting is now completed in a fraction of the time and in 3D by fast
processors and efficient mathematical algorithms. Implicit modelling allows a
geologist to focus on the geology and examine possible outcomes of the data
using repeatable, objective algorithms rather than drawing only to complete
one possible subjective version.
The precipice of change
The industry is at the precipice of change, implicit techniques are gaining
traction. The attraction of fostering multiple hypotheses and having time to
fully examine them in 3D is quickly approaching critical mass.

Obstacle to acceptance
Yet geologists and managers sometimes worry that the speed of implicit
modelling provides less intimacy with the data and that it can feel a bit like a
black box where the geologist has lost control. This fundamental obstacle to
wider acceptance will fade with familiarity and training.

Speed is the great advantage of implicit modelling; one can arrive at the wrong
answer very quickly and the first model or even the second and third can be
quickly dismissed as important details and the geology it represents are
realised.

Driving improvement
Implicit modelling software is just another tool in the geologist’s toolbox to be
properly applied to the appropriate situation. Itis, however, fundamentally
different: poor logging techniques and sloppy standards are made obvious to
the simplicity and unbiased eye of a computer algorithm. Not only will implicit
modelling help geologists examine more geologic possibilities faster, but it
will also drive improvement in data management and logging practices.
Putting appropriate focus on the data that drives business decisions.
Modelling speed and a focus on the data will drive an important understanding
of geologic risk and uncertainty. Fundamentally we will be able to quantify risk
and, with the easy days of mining behind us, understand the geologic
uncertainty of resource extraction projects so critical to continued success.

Patrick ‘PJ’ Hollenbeck & Marc

Independent Consulting Geologists
 Role models

Adapted from the medical industry, the increasing speed and flexibility of
implicit modelling software have made it a powerful decision-making tool for
geologists.

Integra Gold’s Triangle Zone deposit is structurally complex, a common trait to

the many mines and gold deposits in Quebec’s Abitibi gold district. When the
company continued a drilling program on the site in early 2015 it found one
hidden structural trend, which changed the resource estimate for the better.
The site’s Indicated Resources increased by 21 per cent to 627,810 ounces of
gold, with an average grade of 7.37 grams per tonne, thanks to the discovery
of “steeply dipping ‘C’ structures.”
The company detected that structural trend with the help of Leapfrog, 3D
implicit modelling software made by ARANZ Geo. The software, and others
based on the same principles can create accurate, detailed 3D models that
highlight the particular challenges and potential of an orebody in a fraction of
the time and for a fraction of the cost of models done using explicit modelling,
the traditional approach.
Traditional explicit modelling methods require geologists to input the data and
connect the points by hand. Frank Bilki, technical product manager at software
company Micromine, compares this to making a computer model of a loaf of
bread by building a 2D model of each slice and then combining all the slices.
This kind of technique could take weeks or months, depending on the size and
complexity of the deposit, but it had been the standard for modelling – until
implicit modelling became available.
The mathematical functions behind implicit modelling have been
around for some time. It is the relatively recent advances in
computing power,
however, that have helped unlock their potential for commercial
application.
Figure 1: Staff from ARANZ Geo demonstrating the power of the Leapfrog
software.

3D in real-time
The algorithms Leapfrog and Micromine use for implicit modelling are based
on radial basis functions (RBFs). These functions help interpolate the data
collected during drilling programs. They do this by identifying trends in the
data set, based on parameters set by the geologist, and by creating surfaces
that honour those trends. Implicit modelling provides the flexibility to model
many types of deposits, including complex vein systems, stratiform and
porphyry deposits.
As new information and data become available the model is dynamically
adjusted, all in real time. Geologists can tweak the model to examine multiple
hypotheses; for example, making it more or less conservative to aid decision
making. “In an exploration program when you’re spending as much money as
you’re spending – and we’ve got between five and 10 drill rigs running – you
need to be able to make changes on the fly to your drill patterns,” said George
Salamis, a geologist and chairman of Integra Gold. “The quicker you have the
data … the more efficient you’re going to be drilling and the less money you’re
going to spend. It’s pretty simple.” Better yet, implicit modelling software can
take models and present them with 3D contours, allowing geologists to
quickly spot trends or structures. “You really need to see things in 3D to fully
understand them,” said Salamis. “It’s especially key in these structurally
controlled deposits that you see in Quebec and Ontario.”
Before miners got their hands on the software, ARANZ first applied implicit
modelling to medicine and movies. The company’s techniques led to a hand-
held portable laser scanner, which has been used to make prosthetics as well
as some of the computer animations in the Lord of the Rings trilogy and the
Star Wars prequels.
Leapfrog’s program has matured over the past decade as a result of working
closely with the industry, according to its director of product and innovation
Tim Schurr. “We pioneered this method of modelling in the mining industry in
2003, and it definitely caught people’s attention. After more than a decade of
refinement the Leapfrog engine you see today, and the models you see today,
are light years ahead,” he said.
The software uses radial basis functions to generate models such
as grade shells using drill hole data.
Computers have been evolving as well. “The implicit modelling algorithm is
actually quite old, but computers were simply unable to create an implicit
model until they became powerful enough to handle the calculation,” said
Micromine’s Bilki. “Only in the last decade or so have we seen computers fast
enough to do so.”

Uptake expanding
Though ARANZ has established itself as the leading implicit modeller in the
geology market over the past decade, the company is now facing new
competition; Micromine first included implicit modelling in their 2013 software
release.
“We actually love the way that others [like Micromine] are embracing implicit
modelling. It’s really highlighting how game changing, effective and important
it is,” Schurr said. Integra has used implicit modelling software on-site for
several years, Salamis said, ever since the company hired new employees who
had Leapfrog experience from their previous work with major mining
companies
“Before, when we weren’t using Leapfrog as much as we are now, getting the
data into a quick 3D visualization module took time. Now, with our more
intensive use of Leapfrog, the guys and gals who know how to use it pretty
well can visualize things on the fly,” he said.
However, implicit modelling has faced challenges in the past and some people
remain unconvinced. Users need to be familiar with the software in order to
get good results, said Peter Gleeson, a consultant with SRK. “It really does
take about a year to become proficient,” he noted. “And you’ve got to
understand your geology, first and foremost. It’s not a black box. You’ve got to
have a good geologist behind the computer.”
Bilki echoed that users must understand what went into an implicit model if
the results are to be trusted.
[Implicit modelling] is simply another tool in the geologists’ toolkit, to be used
under the appropriate conditions,” he said. “Some ore bodies are ideally suited
to an implicit modelling approach, others are best modelled using traditional
interpretation-and-wireframing or stratigraphic methods.”

Feedback essential
Some regulatory steps have also been taken to recognize implicit modelling’s
potential. In 2014, the technique was included in the AusIMM Monograph 30,
which gives a guide to best practices for mineral resource and ore reserve
estimation. Although Schurr noted that some companies have begun to use
implicit modelling techniques for reporting standards-compliant resource
estimations and classifications, the technique has not yet entered the
mainstream.
Gleeson is optimistic. “NI 43-101 does not stipulate the method you use. So
long as you can justify what you’ve done, and show that it’s geologically
realistic, there is no reason not to use it,” he said.
Geologists may soon be able to bring their implicit models out of the office
and into the field. In September, ARANZ Geo announced a beta version of an
augmented reality tool was in the field with testers. The tool should allow
geologists to superimpose their models on the actual topography of the site. A
release date, at this point, has not been announced.
Feedback will be key, the company stated in the release, and Schurr noted that
this is the way the software has always developed. “This implicit modelling
engine that we’ve built has been developed carefully, through a lot of feedback
from our customers.”
Salamis said that he would love to see conditional simulations, providing
what-if scenarios that could be used to determine if an ore body is
economically viable. “If I was to drill 10 holes, 100 metres away from the ore
body and hit this grade or this width, what does that do to the economics of
this deposit? Right now, we don’t have the capability to do that. We basically
have to go back from scratch and redo resource estimates to get that answer,”
he explained.
But that does not mean that he is not happy with the software’s current
capabilities. In fact, the company has a lot of faith in the software – so much
so that they have selected Leapfrog as the 3D modelling partner for their Gold
Rush Challenge. Entrants are given a Leapfrog licence for the duration of the
competition to help them find new, potentially profitable trends at their Sigma-
Lamaque project. This competition allows the Integra geologists to focus
more on the Triangle Zone.
As implicit modelling itself becomes more established, companies like
Leapfrog and Micromine continue to develop new, cutting-edge applications.
Both Leapfrog and Micromine have new versions of their software on the
market or nearing release; Micromine’s should be launched in the second
quarter of 2016 and a new version of Leapfrog Geo was made available in
November with new tools for data analysis and geologists working on
stratified or layered geologies.

Figure 2: Stratigraphic model designed with Leapfrog Geo 3.0 software.

Kate Sherldan
Journalist
CIM Magazine

Can implicit methods be used to model complex geology?

How Votorantim used Leapfrog to model a complex nickel deposit

Company
Votorantim Metais (Brazilian multinational working with base metals)

Project
A lateritic, mafic-ultramafic nickel deposit located in Niquelandia, 300Km
north of Brasilia, Brasil. The open pit mine produces nickel carbonate using a
hybrid pyro-hydrometallurgical process. The deposit is 26 km (N-S) by 6 km
(W-E) drilled at a 25×25 m space.

Data
Approximately 372,000 metres of drilled cores.

The problem
Until 2011, Voltorantim Metais used explicit modelling to construct and update
geological models, taking approximately six months, depending on the
modeller’s skill. Due to the excessive time spent generating and updating
domains, not enough time was available for developing subsequent stages of
mineral resource evaluation.

Geological challenges
• Strong structural variation along ore bodies
• Dips ranging from sub-horizontal to sub-vertical
• Thickness variations from 1-120m
• Very large dataset (372,000 of drilled core)

Business benefits
Despite learning a new geological modelling tool, it took only four months to
complete work, compared to at least 5 months previously.
Benefits are even more remarkable considering time saved updating the
model with approximately 15,000 metres of new drillholes, from three months
to only three weeks with Leapfrog.

Further Advantages
• Able to process very large datasets including drillhole, channel
samples, geologic contact points, structural dataset and
topography.
• Flexibility to add points and control lines to better reproduce
geology.
• Ability to reproduce the model to avoid subjective bias. Assisting
with due diligence and auditing. Using the same database and
parameters, the same isosurfaces and geological domains are
created.
• Rapidly built and updated, able to evaluate more than one scenario.
The RBF function can be easily fitted by the user to make the model closely
represent reality. Many different parametric settings were tested to fit the 3D
interpolation for each and every lithotype (such as isotropic/anisotropic,
capping values, nugget effect, resolution, directional bias, structural trends).
Assumptions
Lithotypes were grouped into six geological domains, by modelling sequence:
• Bedrock
• Dunite Saprolite (Basic Ore)
• Chalcedony
• Clay (waste)
• Oxide Ore
• Silicate Ore
Geological expertise is essential in defining the relationship between
geological contacts and stratigraphic sequence.

Modelling application
To model bedrock as a continuous surface, the contact points were extracted
directly and interpolated, creating a contact surface. Additional points were
included at the end of holes that did not intercept the bedrock and, in areas
with poor geological information. Both sets of points helped define the
bedrock surface.
The dunite saprolite ore bodies have a flat and smooth shape, usually lying on
the bedrock surface, formed by the weathering process of dunites. To
generate this surface, hanging wall contact points from dunite saprolite were
used. In areas with no dunite saprolite mineralisation, a minus 2 m offset from
the bedrock surface was used to interpolate (figure 1a), resulting in a
continuous surface along the entire deposit. There are only dunite saprolite
orebodies where their surface is above the bedrock surface, to guarantee this
chronology, Boolean calculations were carried out (figure 1b).
For other rock types exhibiting complex shapes (strong variations in thickness,
length, dip and along the strike), the contact points between the litho types
were extracted and converted to a volume function (f (x,y,z)) within Leapfrog.
Points inside (>0), points outside (<0) and contact point on surface (=0) were
defined and arranged according to the position (x,y,z) related to the contact
point. This equation described the infinite numbers of coordinates that lay on
the surface. Positive and negative values increased linearly as they moved
away from their point of reference (figure 2). Figure 3a shows the control
points that were used for chalcedony modelling. Figure 3b shows the
geological domain constructed from the control points.
Figure 1a) Dunite saprolite and bedrock surfaces. Dark grey represents the
dunite saprolite.
b) Dunite saprolite on bedrock surface after the Boolean process.

Figure 2: Positive and negative values increase linearly as they move away
from their point of reference (contact point).
a) Control points used for chalcedony modelling.
b) Chalcedony geological domain constructed from control points.

Laerico Guedes Bertossi & Gustavo Usero

Resource Geologists

Rapid visualisation and modelling of geological data

The rapid development of a 3D model has been a goal in MMG’s exploration programs
to speed up decision making. The ease and speed at which a geological model can be
built has far-reaching implications, as greater workload can be directed at formulating
and testing hypotheses, generating meaningful targeting criteria for success.
Increasing pressure throughout mining to produce results using fewer resources translates to less
time generating accurate and meaningful models, easily leading to a lack of detail in a
depositional model. Where consultants are used for explicit modelling, a culture of external
expertise is fostered, where very few in-house geologists study the resulting model in any detail.
Using an intuitive implicit geological modelling package, such as Leapfrog, has allowed MMG
to expand modelling expertise, with many exploration geologists taking on their own projects
and adding value with relatively little time investment.
Using Leapfrog has resulted in an increase in the speed of modelling by an order of magnitude.
The generation of simple objects has never been easier or more accessible. Directly using raw
data, previously proven tedious, is now streamlined to the point of being transformational for
exploration projects.
3D modelling is still in its infancy at MMG Exploration. The resource department uses explicit
methods to maintain a standard, while Exploration has started modelling some solid volumes
implicitly, the real value has been in the use of Leapfrog Geo and Mining earlier in the process.
Ideas are easily drawn from the modelling before it is complete.
We are developing workflows that promote positive feedback
loops in implicit
modelling and geological understanding, which is moving us
towards our vision of ‘Every geologist using 3D every day’.

The value of fast, intuitive data visualisation and validation

Simple 3D data visualisation makes complex trends instantly recognisable (Figure 1). Many
explicit modelling environments require significant training and data manipulation. Leapfrog
Geo’s interfaces allow large and complex datasets to be inspected and interpreted using intuitive,
fast, and easy-to-learn tools. More geologists gain the confidence to use the 3D environment and
share meaningful interpretation.
Many 3D environments require raw data to be validated manually or with only partial
automation within a database or spreadsheet, applying Boolean rules. This cannot guarantee
error-free data sets and can be days of work. Common issues include drill hole collars floating in
mid-air (Figure 2), subtle survey errors, and irregularity in assay values that can be missed in a
normal rule-based validation as thresholds are often not representative of all project situations.
Leapfrog validates within the 3D environment, allowing issues to be easily detected and fixed
visually. Geologists have a greater sense of data ownership, meaning better collection practices,
more confidence, improved buy-in and involvement.

Figure 1: NEIC
Earthquake Database displayed in Leapfrog Geo after a spherical transformation. A fast,
easy interface is critical to allowing an increase in the number of geologists able to
conduct their own studies.

Figure 2: Drillhole collars before and after a visual comparison in 3D with an accurate
DEM. Database errors such as these are not picked up in automated validation.

Interpretation directly within the modelling environment

Leapfrog Geo’s speed has allowed geologists at MMG to conduct sectional interpretation with
greater accuracy. While ‘pencil and paper’ still provide a certain freedom, the ability to flick
between variables, downhole logging fields and geochemical calculations directly from ioGAS
has been an incentive to conduct2D and 3D interpretation within the 3D view. Implicit modelling
comes to the fore with the ability to instantly propagate onto the area of interest, and while useful
for regularly spaced and oriented interpretation, this becomes an essential tool if using
perpendicular, oblique, or inclined sections. The task has become iterative, resulting in a far
more robust interpretation than otherwise.

Speed of modelling solid volumes

In many situations, the implicit approach has proved far more efficient. A good case study is that
of a stratiform deposit required to maintain a minimum unit thickness while honouring all
available intersection points. Initial explicit modelling took four weeks manually adjusting
surfaces to align with requirements (unsuccessfully). Leapfrog Geo’s stratigraphic sequence
object type was released at this time, and the model was redone in less than a day (Figure 3) and
further refined due to errant assays and incorrect drill hole logging that could not be quantified or
readily detected explicitly.

Figure 3: The stratigraphic model object provides a critical combination of continuous

strata while incorporating pierce points.

Radial basis functions: mapping the quick wins

RBF grade interpolants have become a staple of MMG’s exploration processes. Models can now
be used at the initial investigation phase, rather than the traditional approach of using them
towards the end after most interpretation had been completed.

Generating new ideas

For a new advanced project, an initial geometrical study of the distribution of grade is
undertaken with no bias (Figure 4), allowing the RBF interpolant to simply model the strongest
natural trends. These results offer completely new ideas and hypotheses on ore grade control can
be formulated immediately. Successive refinement using observed trends as controls, form an
iterative process where controlling structures are retained and data biases recognised and
removed (Figure 5).
 Figure 4: Initial
RBF interpolant without bias using a linear variogram. Primary trends of the deposit are
clearly visible, an initial estimate of the magnitude of the deposit is possible.

Figure 5
Figure 5: Surfacing with a volume points function (dynamically converting a subset of text-based
logging into a cloud of distance-based values) is an extremely fast method of understanding the
distribution and geometry of logged geology. This method saves an exorbitant amount of time
compared to purely explicit methods. RBF interpolants provide an unbiased view of geometry
providing further insight absent in mapping or logging.

Communication
Communication implications at this stage are also major, with sometimes only a few minutes
work resulting in a product that can be shared amongst the group, with the majority of the
overarching geometry clearly visible.

Driving decisions
We use Leapfrog to effectively understand primary properties, possibly driving a go/no- go
decision, or as a catalyst to drive further targeted work. Many deposits require review in a short
timeframe so gaining this information and being able to compare our own first pass interpretation
has proven invaluable.
When more than one commodity or element is required to understand or delineate the deposit,
using the same set of controls gives an opportunity to pick multiple generations of mineralisation
or alteration.
MMG is bringing structural geology back from being a specialism to making it mainstream.
Where geometrical relationships are built using overprinting rules and anisotropic trends (Figure
5), validating them requires further fieldwork. The 3D geological model becomes part of an
iterative process that undergoes rigorous testing. With the ability to easily add new data, implicit
models become dynamic, generated to answer scientific questions (Figure 6).
Implicit interpolation of structural surfaces is an essential part of MMG’s workflow (Figure 7),
where structural measurement data are propagated throughout a defined space, and form surfaces
are generated at defined intervals. This enables a quick interpretation of regional geometry where
there is little or no sub-surface data, where such critical elements as depth of prospective
horizons can be extracted. Formline mapping in 2D is an underutilised (and under acquired)
dataset that provides the most significant information about geometry, the ability to instantly
generate the equivalent surfaces in 3D has encouraged more efficient inspection. For stratiform
deposits, a structural form surface is the ideal input for a heterogeneous bias.

Figure 6: Mineral
modes, vein abundances, granular or grouped stratigraphic packages can be quickly
interpolated with an RBF to understand gross geometries, structural complexity and
important relationships.

Figure 7:
Structural form surface interpolants are essential in tracing stratiform deposits where
data is sparse.

Spatial geochemistry
At MMG the ioGAS link has made Geochemistry far more accessible to other geologists
encouraging them to add value to their study. Leapfrog users find that the ability to answer
geochemical questions in real-time a good reason to begin using ioGAS in earnest, while
Geochemists and ioGAS users have effectively used Leapfrog to add more detailed spatiality to
their work (Figure 8), allowing a comparison with geophysical datasets and other modelled
objects.
Figure 8: Alteration domains modelled from ratios calculated in ioGAS, observed and
then modelled using an RBF interpolant in Leapfrog Geo.

Rapid implicit processes lead to better decisions

At MMG, working with an implicit modelling package that combines an easy-to-use and easy-to-
learn interface, a rapid modelling engine, partnerships with leading software and intuitive
modelling methods have not only increased efficiency but have caused an expansion of skillsets.

Structural geology, geochemistry, geophysics, and mineralisation characteristics are increasingly

more accessible. The ability to experiment, combine, and interpret quickly allows far more time
for us to be geologists, mapping the distribution of complex relationships, testing hypotheses,
focussing effort on increasing critical ore-body knowledge elements, and ultimately drilling the
best targets.

Anthony Reed
3d Structural & Information Geologist
MMG
Related
 Validating an implicit model

Implicit modelling was first made well known by Leapfrog, now more have picked up
the process including MICROMINE, Mintec’s Minesight and Maptek’s EurekaTM,
others are sure to follow.
Implicit modelling is fast and easy to run multiple times to test hypotheses and rapidly assess
new datasets. The ease with which the geologist can create and recreate allows refining to
something approaching reality. To create meaningful and robust models, proper forethought,
planning and understanding of the area’s geology is required.
Implicit modelling allows you to build the model you have in mind and apply various changes to
test possibilities suggested by analysis of the data.
Both wireframed and implicit geological models require an understanding of the geology and the
processes around creating them. For implicit modelling, you need a basic understanding of the
interpolation process.
The following help ensure models are robust and the process properly understood.

1. Has the modeller used historic knowledge?

Often those that came before were better trained in mapping, thinking and fieldwork. In this day
of 24/7 connectivity and constant demands, we rarely have the time to spend thinking.
Geological understanding may have changed, but the fundamental basics and measurements
remain. I regularly come across fully qualified geologists that mix strike and dip direction, a
mistake that makes the data worthless. Sandstone mapped 100 years ago is still sandstone,
granite 100 years ago is still granite today, even if the understanding of its origins has changed.
Combining the sampling, mapping and past thought, even from the last year, with the current
understanding always results in a more robust model.
Figure 1: Historical understanding is that grade was contained in short scale veins. The implicit
modelling grade shells indicate there may be a series of cross-cutting faults disrupting the grade.
Pit mapping shows repetitive faults cut the ore body with throws of 8m vertical and up to 20m
horizontally, disrupting the continuity of the grade more than expected.

2. Has the modeller utilised the information encapsulated inside

available data?
Fractal patterns occur in geology over and again. The geological model must make sense on both
the deposit and the larger regional scale. This also occurs across the datasets, where the grade
distribution reflects underlying geology. A folded host may be replicated in folded grade, a
faulted ore body will be visible in the grade, perhaps before it is seen in the geology (Figure 1).

A simple flexure in the geology might be shown to be a fault with sharp offset in the grade. This
works both ways and at different scales. The use of grade control data in a pit or underground
environment can show significant discrepancies not seen in the geology. If there is a fault
modelled but the grade shows no offset is a fault really there and is it material?
 Figure 2
Figure 2: Implicit modelling used in targeting gold hosted in a ductile shear along an ultramafic
contact. Simple implicit grade modelling combined with in-depth assessment showed the gold
tended towards supergene enrichment in and around a folded volcanic horizon hosted within the
sediments. A significant step in the basement co-incident with the grade shows the location of
what was later modelled as a cross-cutting fault and which subsequently became the feeder for
the mineralisation, completely turning the exploration model on its head. (Legend; Ultramafic =
purple, volcanic = green, sediments = blue and cover = brown).

3. Can the modeller describe and explain the model?

The implicit modelling algorithm may be indicating a dip on an intrusion normally thought to be
vertical. We may write off the result because we “know” the intrusion is vertical. But the
interpolant will always replicate the geology. We need to be able to look at the results and read
what the algorithm is telling us (Figure 2). Is the algorithm wrong or is our understanding of the
geology flawed? A modeller who has thought about the model and crafted a result using all
aspects of geology and having a full understanding of the mineralisation process should be able
to intelligently explain the resulting model. Equally important is being able to tell which aspects
of geology are most important for each deposit modelled and how they affect the result.
Figure 3a: A series of porphyries were modelled explicitly on the “knowledge” that in these
environments they always intrude vertically. Whilst the model was complex and never really
hung together with respect to the geology in the drilling, it was never fully challenged.

Close inspection of grade shows that it appears to start and stop abruptly within and outside of
various porphyries. It was postulated that the grade was in fact thrusted apart by a series of east-
dipping faults (green) and that the whole sequence may be tilted 20 degrees to the east as it
ascends a thrust ramp.

Figure 3b: Mapping and regional studies showed that there were indeed several smaller thrusts
cutting the deposit and a large regional thrust just to the west (left) of the image, the geology was
in fact tilted to the east. With this knowledge, the deposit was remodelled implicitly and the
multitude of small low-continuity dykes filled out to become several large consistent but
disaggregated porphyry bodies.
4. Does the output look realistic given what we know?
Sometimes what we know about a deposit is simply wrong (Figure 3). So question the model,
test and dissect it and the knowledge of the geologist who built it. The model, implicit or explicit,
might be poorly constructed, poorly thought out and poorly presented. Don’t dismiss the model
outright. It may be that something important has been missed and the data is trying to spit out
(Figure 4).

Figure 4: The same deposit as Figure 3 but showing a very simple isotropic undirected
interpolant on the grade. The data clearly shows that “implicitly” the grade is dipping to the west
and dissected by a series of faults (there are more thrusts in this system than shown). Despite
seeing this, it took time to convince the audience as it clashed with current understanding.

Ronald Reid
Group Resource Geologist
Harmony Gold SE Asia
 The future of geological modelling

One of the biggest gaps in mining-industry geoscience is between our understanding

of ore deposit formation and application in resource estimation. Significant advances
including imaging and analytical methods, has given better understanding of
interactions between hydrothermal fluids and host rocks.
Advanced geostatistical methods and improved hardware and software capability allows rapid
processing and interpolation of huge amounts of drillhole data. However, there is sometimes a
significant lack of communication between geoscience and geostatistics resulting in reduced
confidence in resource classification and, sometimes, poor business decisions. This is the
opportunity and future of geological modelling.
The mineral system approach has been successfully used in the mining industry for decades and
describes the process whereby concentrations of chemical elements resulting from natural
processes are transported, focused and delivered from a source area to the deposit. The approach
originally came from the oil and gas industry, where an understanding of the entire source-
pathway-reservoir/trap system is critical for expensive exploration efforts.
Advances in software provide an opportunity to include mineral system components in the
deposit modelling workflow. For example, a better understanding of fluid pathways (usually
structures) and fluid-rock interactions, highlight the shape and continuity of high- grade domains
as well as the boundaries between high- grade, low-grade and waste domains.

Consistent application of geological rules

Advanced implicit modelling software has the ability to include geological rules and timing
relationships as constraints, This includes cross-cutting relationships between faults and other
structural features, and the nature and position of contacts between geological units, e.g.
depositional, conformable, erosional or intrusion-related. Critical to building a geologically valid
and consistent model, particularly if complex relationships need to be expressed, such as with
larger and/or high-grade ore deposits. Hopefully in the near future software that has intuitive and
simple tools for applying rules and relations will become the industry standard ensuring that
models honour our understanding of complex geology in 3D (Figure 1).
The ability to rapidly build and continuously test a range of geological interpretations is another
significant advantage. This provides the capability to build multiple geologically valid scenarios
from the same dataset, from pessimistic to optimistic interpretations. In mining, this will
improve understanding of uncertainty and reduce reliance on a single resource and reserve model
for planning and strategic decisions. This will also make the modelling process more scientific.
Figure 1: Example of complex geological model built with implicit modelling workflow that honours
geological rules and timing relationships.

Implicit modelling used for resource estimation

The main advantage of implicit modelling is that mathematical interpolants build the surfaces
between known points in drillholes, rather than ‘explicit’ or manual construction. In some
modelling software, the interpolants are mathematically equivalent to kriging, one of the best
known and consistently used linear geostatistical interpolation methods. There is now the
opportunity to use the interpolants that generate the geological domains to also do the grade
estimation, vastly simplifying and integrating geological and geostatistical modelling workflows.
This potentially combines geoscience and geostatistics into a seamless workflow, providing good
fit-for- purpose results and allowing for more geological controls in the estimation workflow and
final block model.
Plan view of open pit with grade control contour lines following structural trends.

Bringing more geology into grade control models

The grade control stage of mining has the most geological information and usually the least
amount of time to make decisions about boundaries. Poor decisions can be very costly and
produce poor reconciliation results. At a recent mining geology conference, a world-renowned
geostatistician stated:
Increasingly, reconciliation issues are now found to be related to a
lack of orebody knowledge in terms of location of the geometric
limits of ore, and inadequate understanding of the irregularities of
contacts and the ability of mining to follow them.
This exactly relates to the knowledge gap between geology and geostatistics. Implicit modelling
software can bridge the gap rapidly building and updating with geological rules intact.
Traditional grade control dig lines are generally straight lines with sharp corners (Figure 2),
commonly constructed using data from a single bench. Structural trends in implicit modelling
(which mimic structural pathways for ore fluids) means we can rapidly build more realistic grade
control models (Figure 3) using all available data above and below that particular bench. The
resulting dig lines look more like what you expect to see as the result of ore fluids flowing
through host rocks. Also, with more GPS-guided and automated mining equipment, we can
follow more irregular and curved contours.

Figure 2: Plan view of open pit with grade control blocks, which result in angular and unrealistic dig lines.
Figure 3: Plan view of open pit with grade control contour lines following structural trends.

Implicit modelling for geoscience training: from classroom to

resource
The most critical component is that geoscientists have the appropriate knowledge and training to
build complex and multi-discipline resource models that unlock value and benefits for their
company, the industry and society.

New specialist skills required will include:

• Downhole geophysical tools and sensors
• Real-time sample analysis and systems
• (e.g. laser-induced breakdown spectroscopy, portable x-ray fluorescence and x-ray
diffraction)
• Hyperspectral systems for scanning drill samples and mine faces
• Monitor-while-drilling and ‘smart bit’ systems
• Automated drill rigs and associated equipment
• Data integration and analytics
Geological understanding will always be the most important skill in building geologically valid
models. Implicit modelling will provide maximum benefit from the flood of real-time geoscience
data.
It will be important that geoscientists develop multi-discipline skills to better understand the
requirements of all technical disciplines and downstream customers of resource models. This is
an opportunity for industry, service providers and universities to work together to develop
training programs to attract and retain the best and showcase mining as a sophisticated, high-tech
industry.
Models and scenarios that are easy to build and update are fundamental in helping communicate
multi-discipline understanding. This is the future of geological modelling: implicit modelling
workflows that honour geological rules will result in a better understanding of uncertainty,
opportunities and risks and will enable ore bodies to be mined more efficiently, more
environmentally responsible and more safely.

Paul Hodkiewicz
Senior Manager Technology Development

Is implicit modelling reliable enough for use in resource

domaining?

The mining industry has resisted the acceptance of resource domain

wireframes derived using implicit methods. There’s also been bad press
because it provides a conduit for some rapid and poorly thought through
‘blobby’ models. However, this is changing as more become familiar with the
techniques and see the advantages, which, we argue, outweigh the limitations
of traditional sectional wireframing. This brief review touches on key
questions and is a starting point for discussion.
The development of a robust resource domain model requires familiarity with
the commodity and geology of the deposit, with the geological and statistical
justification of the domain and with the strengths and weaknesses of the
methods and software used.
In the past, some models built using implicit tools were sub- optimal,
reflecting the form of a mathematical function rather than a geological shape.
This can be put down to the practitioner’s poor knowledge of the software’s
algorithms and an unfortunate acceptance of such output. Geologists are now
more adept in implicit techniques and the software itself has similarly evolved.
This was equally true of sectional methods when the GMPs replaced older 2D
techniques.
Careful and relevant geological domaining in resource estimation is crucial to
developing a robust and reliable resource estimate. Mathematical models
alone are not sufficient to accurately define a resource.

Reluctance to accept the implicit approach for resource domaining

Conservatism
Resource practitioners are probably the most conservative people in the
mining industry (as Competent Persons they carry the responsibility for public
signing off on Resource statements). They feel most comfortable with
methods used early in their careers.
Distrust of the unknown
Especially when implicit models produce unexpected results (no matter how
fundamentally correct and unbiased)
Skill
Lack of training makes it difficult to overcome early problems in controlling
implicit models, so practitioners revert back to known techniques.
Resistance to Change
Feelings that new technology may take away or diminish the role of the
professional modeller.
Organic look
Unease at the organic look of implicit models compared to traditional
structured boxy shapes of polygonal models (an irrational but common
observation).
Black box
The mathematics behind the implicit models is unknown and tends not to be
trusted.
JORC compliance?
No software or method is JORC compliant as this requires the input of a
Competent Person and appropriate methodology at all levels in the estimate.
Bad initial experience
Commonly due to lack of necessary skills at an early stage to control the
model outcome effectively.
Perceived lack of control
A perception of lack of control on development and building compared to
manual techniques.
Numerous outcomes
Which is the correct model?
Snapping
Does the implicit model snap to points on drillholes?
Experience
Becoming competent in new software methods requires considerable
investment in time to train and gain experience.

Reasons for using implicit modelling for resource domaining

Comparing volume, grade and geometry of well thought-through implicit
models with traditional models is normally favourable, with the implicit model
often exceeding reliability.
Implicit models are easily repeatable, making the incorporation of new data
relatively straightforward. They are also more auditable than manual
wireframes.
Implicit methods make it much easier to deal with larger grade control data
sets, assisting with 3D interpretation of trends and improving reliability of
local grade estimates.
Implicit models provide a better 3D extrapolation when modelling from
irregularly spaced data and are not sectionally biased as can sometimes occur
with manual models.
With complex, high nugget deposits, implicit methods can better link
mineralised intervals between holes.
Grade and volume dilution can be incorporated with much greater control,
often eliminating unnecessary dilution that occurs in manual models.
Implicit modelling makes it feasible to produce multiple realisations of a
deposit according to a range of assumptions, highlighting the level of risk in a
single model approach.
Surfaces can be generated more rapidly from the data, saving modellers time
and companies money.
The new geological workflow built into several implicit modelling methods
enables geologists to bring more geology into the resource model and
produce more geologically refined models.
The speed of the implicit approach allows more time for gathering
fundamental information to inform.
The implicit approach allows management to overcome the One Model
Syndrome, where a projects technical and financial development hinges on a
single model. Multiple realisations of a resource model can be used alongside
the manual model to determine how conservative, liberal or otherwise, i.e. the
level of risk inherent in the model. The implicit approach allows management
to determine the weaknesses of a geological or resource domain model.
It is something of a myth that implicit technologies only work well in data rich
environments (e.g. grade control). Implicit modelling can work as well, if not
better, in data-poor environments. If data is genuinely sparse, both manual and
implicit methods often fail due to the realisation that constructing a model
that has no demonstrated geological control or geological continuity is going
to be very poorly constrained.
These views are commonplace in the metalliferous mining industry and are
partially justified by early experiences with implicit modelling software.
However, the software has developed significantly, addressing early concerns
and incorporating more geologically relevant workflows. The ability of the
geologists has also improved. Many mining companies, operations and
consultancies are realising the technical and financial advantages. As labour
costs rise and mines seek efficiencies, teams dedicated to time-consuming
wireframing is not cost effective. No model can yet be generated entirely
automatically, a hybrid approach with geologist input to steer implicit models
is expected to be required for some time. The pendulum has swung in favour
of implicit methods for many good reasons.

Focus on snapping
Vein modelling tools mean that exact intersection points can be chosen if
required, providing the means for snapping, but reasons should be examined
carefully.
Is snapping to points essential for a good model? One question frequently
asked is do and should implicit techniques honour contacts? Often, this refers
to the process of snapping in 3D to either grade or geological contacts points
on drillholes. The answer is not simply yes or no. If the contact is a hard
geological boundary such as a vein footwall or hanging wall snapping can be
necessary. However, if it is a soft boundary such as a nuggety gradational
grade change, then the answer is not so obvious. In this situation, it is likely
there is no definite contact discernable in the core or field and the “contact”
becomes a grade contour that serves the purpose of defining ore types for
mining.
 Figure 1
Figure 1 shows this problem of grade contouring with a typical gold deposit
intercept in a drillhole. The intercept was given a broad geological description,
minimum mining width and cut off grade, and several geologists were asked
to select the correct point to snap to to achieve a required cut off grade
domain boundary. All selected slightly different intervals. The implicitly
modelled answer was somewhere between them all. Each contact selected
was precise but not necessarily any more correct than the implicit solution.
Snapping to an absolute point is very subjective and the application of ‘rules’
means that snapping is not as important as other considerations, such as
complexity and minability of resultant shapes.
The aim of snapping or selecting a grade boundary is to ensure that below cut
off grade samples are excluded from the estimate. From a review of several
manually constructed wireframes, the number of samples below cut off
significantly exceeded that selected by implicit methods, despite the geologist
snapping to exact points.

Focus on geological modelling

Another common criticism of implicit techniques is the perceived lack of
geological and structural control on the grade or ore body outline. In early
versions of the software this was indeed the case, subsequent refinements
and improvements mean it is no longer. Complex structural trends can be
incorporated, analogous to key features of traditional wireframing of a skeletal
framework of the controlling structures underlying the grade distribution.
However, implicit modelling is far simpler and less time consuming.
It is now becoming a common approach amongst all implicit software that the
structural trends themselves can be controlled directly from primary
geological inputs of dip and strike (from structural measurements down hole,
surface and underground mapping. The ability to rapidly build hard geological
contacts such as footwall and hanging wall structures encourages the use of
rigorous geological domaining where once it may have been given less
importance due to time constraints. Ore bodies with multiple, complex strike
directions that were once difficult to build can now be tackled with greater
geological rigor. Recent improvements in vein modelling workflows and
grade/contact selections mean straight forward vein systems can also be
tackled.
The principles of stratigraphy are now also being incorporated ensuring easier,
logical geological building of models rather than using abstract Boolean
operations.
Modern techniques now also incorporate local manual control to overcome
this once frequently commented on issue.

Focus on the mathematical algorithms

Probably the most prevalent criticism of implicit approaches is that they are
“Black Box”. The algorithms used in GOCAD’s SKUA are based on the well
known and documented Discrete Surface Interpolator (DSI) algorithm. The
Leapfrog Radial Basis Function (RBF) is based on Dual Kriging, a rapid
optimised derivation of Ordinary Kriging.

All are openly available for review. As in any modelling, no outputs should be
trusted without rigorous comparison with the informing data and local
geological knowledge. Similarly, prudent validation by section and plan with
the informing data sets should always be performed.

Summary
The use of implicit technologies for resource domaining requires a paradigm
shift in thought and approach away from the sectional CAD-based techniques
that have dominated for over 30 years. Thomas Kuhn (Professor MIT, Harvard,
Berkeley 1922-1996) thought of science as a kind of mob psychology and not
a purely rational search for truth. “Only once the groundswell of opinion is
sufficient then the new science or technology become accepted despite how
fundamentally correct the theory or technology was in the first place.”
Failure to recognise the importance of new technology is summed up in this
quote by Sir William Preece, Chief Engineer of the British Post Office, circa
1876: “The Americans have need of the telephone, but we do not. We have
plenty of messenger boys.”
Evidence from numerous comparisons with pre- existing resource domain
models over the last few years has demonstrated that, when used correctly by
experienced geologists, implicit models can equal or exceed the quality and
reliability of manually- derived sectional wireframe methods. However, it is
critical that fundamental geological skills be employed when developing any
model, regardless of software or method employed.

Peter Gleeson
Corporate Consultant Mining Geology
SRK Consulting, Cardiff, UK

References:
Malet JL. Discrete smooth interpolation in geometric modelling. Computer
aided design 1992. Vol 24, p 178
Chiles J-P and Delfiner P, “Geostatistics Modelling Spatial Uncertainty”, Wiley
Interscience, 1999, p 186
Dowd P A, “A review of geostatistical techniques for contouring”, NATA ASI
Series, Vol F17, Fundamental Algorithms for Computer Graphics. Edited by
R.A. Earnshaw Springer Verlag..
Galli A. and Murillo E. “Dual Kriging – Its properties and its uses in direct
contouring” Geostatistics for Natural Resources Characterization, NATA ASI
Series C: Mathematical and Physical vol 122 part 2, Dohrecht, Holland, pp 621-
634
Kugn T. 1962. The Structure of Scientific Revolutions