Dynamic Anisotropy Difficult To Say Harder To Get Right

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

CIM Journal

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ucim20

Dynamic anisotropy: difficult to say, harder to get


right

J. Lunnon & R. Pressacco

To cite this article: J. Lunnon & R. Pressacco (2023): Dynamic anisotropy: difficult to say, harder
to get right, CIM Journal, DOI: 10.1080/19236026.2023.2192603

To link to this article: https://doi.org/10.1080/19236026.2023.2192603

Published online: 24 May 2023.

Submit your article to this journal

Article views: 73

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=ucim20
CIM JOURNAL
https://doi.org/10.1080/19236026.2023.2192603

GEOLOGY

Dynamic anisotropy: difficult to say, harder to get right


a b
J. Lunnon and R. Pressacco
a
La Mancha Resource Capital, London, UK; bOakville, Canada

ABSTRACT KEYWORDS
Dynamic anisotropy (DA) is a tool practitioners can employ to reflect localized variations in Dynamic anisotropy (DA),
mineralization orientation observed in nature in a block model. DA adjusts the orientation of the Euler angles, Locally varying
search ellipsoid on a block-by-block basis, rather than using a static search ellipsoid. These anisotropy (LVA), Major axis
transformation, Reference
block-by-block DA angles can be generated from several sources and surfaces that are tailored
surface, Search ellipsoid,
to suit the desired level of “smoothness.” The angles can then be coded into the model in Trend surface
several ways that balance the time availability and potential sources of human error such as
averaging dip angles across the 0°/360° change or applying a static plunge to a varying MOTS-CLÉS
azimuth/dip. Further corrections are required if the practitioner wishes to apply a default plunge angles d’Euler, anisotropie
to a constantly varying search ellipsoid. Although DA can be used to properly optimize a model dynamique (AD), anisotropie
and fine-tune the estimation, when used incorrectly, it will result in a model containing localement variable (LVA),
ellipsoïde de recherche,
significant fatal flaws. This paper outlines the most common approaches to DA, potential pitfalls,
surface de référence, surface
and corrections for applying a plunge to the search. It also highlights the most important step of de tendance, transformation
undertaking a thorough visual and statistical validation on both the input parameters and the de l’axe principal
final model.
RÉSUMÉ
L’anisotropie dynamique (AD) est un outil que les praticiens peuvent utiliser pour refléter dans un
modèle de bloc les variations localisées de l’orientation de la minéralisation observées dans la
nature. L’anisotropie dynamique ajuste l’orientation de l’ellipsoïde de recherche bloc par bloc,
plutôt que d’utiliser un ellipsoïde de recherche statique. Ces angles AD bloc par bloc peuvent être
générés à partir de plusieurs sources et surfaces adaptées au niveau de « douceur » souhaité. Les
angles peuvent ensuite être codés dans le modèle de plusieurs manières qui tiennent compte du
temps disponible et des sources potentielles d’erreur humaine, comme le calcul de la moyenne des
angles de pendage sur le changement 0°/360° ou l’application d’un plongement statique à un
azimut/ pendage variable. D’autres corrections sont nécessaires si le praticien souhaite appliquer
une plongée par défaut à un ellipsoïde de recherche qui varie constamment. Bien que l’AD puisse
être utilisée pour optimiser correctement un modèle et affiner l’estimation, lorsqu’elle est utilisée
de manière incorrecte, elle aboutira à un modèle contenant d’importantes failles fatales. Cet article
présente les approches les plus courantes de l’AD, les pièges potentiels et les corrections à apporter
à l’application d’une plongée à la recherche. Il met également l’accent sur l’étape la plus impor­
tante, qui consiste à entreprendre une validation visuelle et statistique approfondie des paramètres
d’entrée et du modèle final.

REVIEW OF THE CONCEPT Although DA is a powerful tool to help practitioners


The ability to alter the orientation of a search ellipsoid produce models that better reflect the natural, local-scale
locally during estimation is not a new concept. Dynamic variations observed in mineral deposits, it also intro­
anisotropy (DA), also known as locally varying aniso­ duces sources of potential errors that practitioners may
tropy, or variable orientation has been available in geo­ overlook as a result of time constraints that often accom­
modeling software packages for many years. DA is the pany resource updates. Common sources of errors are
process of altering the orientation of the search ellipsoid, selection of the wrong type of input surface or incor­
typically for grade estimation on a block-by-block basis. rectly flagging the angles to the model by averaging
The availability and use of the DA method on a routine angles that are orientated similarly (e.g., 355° and 10°)
basis have significantly increased over the last few years but differ greatly numerically. As with many areas of
due to improvements in computing power and wire­ resource modeling these days, the process of incorporat­
framing methods, which have made the method quicker ing DA into a modeling workflow is becoming more
and easier to implement. easily executed. Although it may provide a quick way

CONTACT J. Lunnon jack.lunnon@lamancha.com


© 2023 Canadian Institute of Mining, Metallurgy and Petroleum
2 J. LUNNON AND R. PRESSACCO

to make it appear that the accuracy of the estimate at the adequate understanding of the variability of orientation
local scale has been improved, it can leave the practi­ of the mineralization grades within a deposit. There is
tioner not understanding how potential errors may no “sweet spot” for the correct quantity of input data:
occur and how to identify them. they will be deposit specific, and the practitioner will be
The goals of this paper are to: (1) provide required to make a decision based upon their judgment,
a description of the generally accepted workflows that experience, and knowledge of the deposit.
are in common use for preparing a DA grade estimate, In an example of a simple “kinked” vein that changes
(2) describe some of the common pitfalls that are significantly in orientation locally, the static search ellip­
encountered, and (3) share some solutions for com­ soid (Figure 1a) honors the majority of the deposit but
monly encountered errors. Additional information is does not reflect the orientation where this localized
provided in Glacken, Blackney, Gray, and Fogden deformation occurs. These localized deformations can
(2014). be associated with significant variations in grade. Thus,
ensuring that the search ellipsoid is correctly oriented
here is critical to improving the accuracy of the esti­
DESCRIPTION OF WORKFLOW
mated grade at this local scale. In the sub-domained
The traditional approach for completing a grade estima­ example (Figure 1b), there is a better reflection of this
tion with a search ellipse is to provide the software with varying orientation; however, the changes between the
three orientations to represent the strike, dip, and plunge sub-domains are often abrupt and may poorly honor the
that the practitioner wishes to use to model the continuity orientation in these transition zones, which are observed
of an anisotropic attribute, such as grade within a volume as estimation artifacts in the model. The DA example
or block model. Although a static ellipsoid orientation can (Figure 1c) illustrates the search ellipse orientation that
be used for a given estimation domain, the individual follows a trend surface and changes its orientation at
ellipse orientations can be varied for different estimation each sample location. In so doing, the estimated grades
domains where significant differences in the direction of can better reflect the local distribution of the grades
the mineralization exist (e.g., a set of steeply dipping veins within the vein at all locations.
and a mineralized colluvium body). Domains that change
their orientation significantly along the strike or dip can be
SOURCE OF SURFACES
further sub-domained for grade estimation purposes using
fixed search ellipses that reflect the local-scale geometries. To carry out DA, as a minimum, a dip and dip-direction
The data informing the estimate would not necessarily be (or strike) are sampled at multiple locations to provide
sub-domained to avoid the creation of estimation artifacts. input for the search ellipsoid during grade interpolation.
This sub-domaining approach can provide acceptable These sampled locations are often provided by extracting
results in mineralized bodies that have only occasional the dip and dip directions of individual triangles from
“bends” along the strike. In nature, however, the orien­ a wireframe surface that reflects spatial variations in
tations of many deposits such as shear or vein-hosted mineralization on a local scale. To provide a reasonable
deposits, can vary significantly, even within the same outcome, the informing wireframe surface must be of
general trend. In addition, there may be a high-grade sufficient resolution (i.e., have a sufficient number of
contact on the footwall/hanging wall or a high-grade triangles) to allow for these changes in direction to be
core, which should be preserved in the model by ensur­ described in a smoothly transitioning manner. The use of
ing these samples are highly correlated during the inter­ large wireframe triangles is generally not preferred
polation. The ability to alter the search ellipsoid because they may have significant and rapid changes of
orientation on a detailed block-by-block scale would orientation between them, which can produce unwanted
therefore help maintain the high-grade mineralization or unacceptable results or might not reflect shorter-scale
proximal to the contact, reduce lateral grade smearing, localized changes desired by the practitioner
and avoid samples being missed entirely during search Some examples of input surfaces that can be used to
estimation. Once a practitioner is familiar with using extract the triangle angles are described in the following
DA in their modeling package, the setup is often quicker sections.
than creating unique sub-domains and modifying inter­
polation scripts to alter the search orientation, especially
Hanging wall and footwall surfaces
when dealing with a large number of domains and/or
variables. Hanging wall and footwall surfaces are often selected
The practitioner should collect sufficient data and because they have been generated to constrain the block
conduct quality control of the data to provide an model during the estimation workflow. However, they
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 3

Figure 1. (a) Static search ellipsoid within a varying orientation vein, (b) sub-domained search ellipsoid, and (c) dynamic anisotropy-
controlled search ellipsoid in varying orientation vein

can often be a poor choice as input parameters because may reflect the practitioners goals if there are high
one or both of these surfaces can significantly change grades associated with one of the contacts, or one
direction: they may pinch and swell on a highly localized surface is a flat contact such as a hanging wall in an
scale as a result of small-scale changes in intersection erosional nickel deposit, and the practitioner wishes to
width, down-hole grade smearing, down-hole survey not have the continuity impacted by post-
issues, nuggety deposits, or wireframe construction mineralization changes.
issues (Figure 2a). Whereas localized issues with pinch
and swelling can to some extent be reduced by using
both hanging wall and footwall surfaces as inputs, other
Reference surface
issues can also occur through using multiple surfaces,
which are covered later in this paper. In addition, the A reference surface (referred to as a trend surface or
combination of different data sources such as historic viewing plane in some software packages) is sometimes
data, grade control data, and underground channel sam­ generated by modeling software to provide as a “guide
ple data can exacerbate this issue, particularly if the data trend” for the algorithm to join together selected inter­
are clustered. vals during automated wireframe construction. This
Potential solutions include increasing the resolution reference surface can then be manually edited by the
of the triangles or filtering some of these input data, practitioner to better reflect their understanding of the
although the impact of either solution should be con­ geology and any mapping.
sidered. If the estimation uses hanging wall or footwall If the reference surface is automatically generated by
surfaces, it may cause the search ellipsoid to be angled the software, it provides an easy source as an input to
incorrectly when this is a very localized issue and not DA; however, the practitioner must be aware that it
reflected by the general orientation of the mineraliza­ represents a generalized trend of mineralization used
tion. These changes are often pronounced where edge for wireframing. Therefore, where data are more closely
crinkling of wireframes occurs or if a hanging wall/ spaced and vary significantly (e.g., grade control data),
footwall intersects another surface. If these surfaces the reference surface may “wander” outside of the hang­
are being used as an input to DA, is often better to ing wall and footwall (Figure 2b). The practitioner must
extract the surfaces before any clipping/Boolean opera­ consider whether this is acceptable if they wish to have
tion (e.g., clipping of intrusions/topographic surfaces). a general orientation change or whether they want to
However, the use of a hanging wall or footwall surface capture more localized trends.
4 J. LUNNON AND R. PRESSACCO

Figure 2. Surfaces available for dynamic anisotropy: (a) trend following hanging wall surface; (b) reference surface following the
general mineralization trend; (c) center surface following interval midpoints; and (d) pole intersections surface following intersection
midpoints
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 5

Center surface amount of time to be of sufficient quality to be


used reliably.
An alternative to the reference surface is the use of
When manual wireframes are generated on sectional
a center surface (best-fit plane) that reflects the middle
interpretations that are joined using explicit wireframes,
of the mineralized domain. The benefit of a center sur­
the orientation change between neighboring wireframe
face is that it will always reflect the varying orientation
triangle angles can be very significant, which would
locally around a drill hole, although many software
cause abrupt orientation changes in the resulting model.
packages will not automatically generate these surfaces.
In these instances, it may be beneficial to generate interim
If a manual center surface is required, it can be created
sectional and/or plan interpretations in some areas to
by extracting full-width composite mid-points and gen­
reduce the orientation change between neighboring tri­
erating a surface from them.
angles, although care must be given if modeling in two
Although a center surface will better reflect the loca­
planes because interfering triangles can occur.
lized changes in orientation in some instances (e.g., broad
When undertaking DA with wireframes that have
orientation changes that have been interpreted between
been generated using implicit modeling methods (such
widely spaced drill holes), a center surface process could
as grade shell or indicator modeling), a manual trend
produce a wireframe that may still crosscut the hanging
surface may be required because the mineralization sur­
wall or footwall (Figure 2c). In all cases, the practitioner
faces may not be appropriate to use as a source due to
must review the input points and resulting surfaces to
the pinching and swelling of the domains, and they will
ensure that they reasonably represent the data.
be affected by many of the parameters used by the
geologist to create them (e.g., probability and range).
This pinching and swelling may cause undesirable vec­
Pole intersection tors that do not reflect the mineralization trend.
If the practitioner requires a center surface that exactly Use of a series of trend strings rather than a surface is
follows the mid-point between drill holes (e.g., if there is not recommended because they will only provide
a high-grade core that the practitioner wants to ensure a reflection of the orientation on the sections and not
remains in the center of the domain between the drill any orientation changes that may be present between the
holes), then a pole intersection surface can be created section lines.
(Figure 2d).
This surface is generated from a series of poles that Summary
traverse across the entire mineralization perpendicu­
larly, typically on a regular spacing. The intersection All of the surfaces discussed above can be used as inputs
of the polylines with the hanging wall and footwall can for a DA estimate. It is up to the practitioner to judge
then be converted into points, and the mid-point which surface will provide the optimal input and
generated as half the distance (in three-dimensional whether the additional time to improve a surface or
space) between the footwall and hanging wall contact manually generate a surface is warranted. It is highly
points. This pole midpoint is combined with the drill- recommended that practitioners review their choice of
hole midpoints, and a surface is generated with them. surface in both sectional/plan view and three dimen­
The spacing between the poles can be adjusted down­ sions before using its information for estimation pur­
wards if a higher resolution surface is required to poses. In addition, some software packages allow the
better honor the center of the modeled mineralization. user to visualize the trend orientation as a three-
This surface would not be suitable where one surface dimensional disc or triangle that can be visually com­
is a flat contact and the other is undulating, such as an pared with the input data. If the domain is faulted, the
erosional nickel deposit. practitioner should ensure that these surfaces are gener­
ated and flagged to the model separately for each faulted
domain because a single surface that crosses the fault
Manual trend surface would likely cause erroneous directions around the
contact.
A manual trend surface can be created by the practi­ An alternative approach to using a surface is to gen­
tioner, typically on sectional interpretations, which erate a structural control model using a series of points,
can be used as the reference surface. While manually which have come from structural measurements and/or
created trend surfaces have the benefit of exactly interpretations. This approach is not covered here
reflecting the practitioner’s geological knowledge because a structural trend will be subject to many para­
and desired input, they may require a significant meters and decisions on such items as trend decay. In
6 J. LUNNON AND R. PRESSACCO

Figure 3. Change in an observed azimuth as a result of applying a plunge within a dipped plane

the authors’ experience, these have been infrequently angles as a result of applying a plunge must therefore be
used to date, but they are seeing more use alongside corrected at all informing surface nodes (or at each
implicit modeling approaches. block within the model) before using the empty
model in a grade estimation. The two main methods
the authors have used to account for the plunge are to
PLUNGE ESTIMATE transform the major axis and adjust the Euler angle
rotation axes, as outlined below.
Up to this point in the paper, only changes in the strike At this junction, the practitioner must remember that
and dip of the mineralization have been considered. the plunging value should be measured in the plane of
However, many deposits display a measurable plunge, the dipping surface (i.e., in a longitudinal section that is
which should be reflected in the orientation of the search angled so that it is perpendicular to both the strike and
ellipsoid. If the plunge of the mineralization is observed the dip of the mineralization), otherwise an apparent
to be either horizontal (0°) or vertical (90°) or if there is plunge would be measured. When used as the input,
no observed plunge, that is, there is equal anisotropy in the apparent plunge would result in a sub-optimal
both the major and semi-major direction (a pancake- search.
shaped search), then it is safe to select a default plunge
value because the search ellipsoid is likely to have the
same radius in both directions. Depending on the soft­
Transforming major axis
ware, these default plunges either need to be coded
directly to the block model or they can be used as default One way to add a static plunge to a DA angle is to
values as input parameters during the grade estimation recalculate the azimuth, plunge, and rotation from the
process. dip and dip direction using the following workflow:
The issue with using a plunge arises from a single
plunge value being measured despite the fact that the Rotation = IF ([Dip direction] < 180) THEN [Plunge]
azimuth and dip angles are constantly changing with ELSE [Plunge] × –1
the DA surface. Since plunge values are applied to the Plunge = [Dip] –90
search ellipsoid within the dipped surface, this will Azimuth = [Dip direction]
result in the azimuth being modified. Figure 3 illus­
trates an example, where an arrow oriented north The major direction then becomes axis 2, the semi-
(labeled 1) is on a 45° dipped plane. When a plunge is major direction is axis 3, and the minor direction is
applied to the arrow, it is applied within the dipped axis 1. For example, if the search ellipsoid was 100 m ×
plane. Therefore the resulting observed azimuth rotates 50 m × 10 m in axis 1/2/3, it would now become 10 m ×
away from the north (labeled 2). This change in the DA 100 m × 50 m (Table 1).

Table 1. Major axis transformation of the input orientation to correct for the azimuth change with a dip applied; see Figure 4 for
ellipsoid descriptions
Axis (m)

Ellipsoid 1 2 3 Azimuth Angle (°) Dip/Rotation Angle (°) Plunge Angle (°)
Orange (pre-transformation) 100 50 10 270 82 50
Blue (post-transformation) 10 100 50 180 −50 −8
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 7

Figure 4. (a) Two-dimensional slice and (b) sectional projections of a wireframe surface dipping 82° south illustrating transformed
(blue) and untransformed (orange) ellipsoids in-plane described by the strike and dip looking north

Figure 4 illustrates an example where two search set. The conversion of angles using Euler rotations fol­
ellipsoids have been generated from the same input lows the following steps:
surface (dip 82° and dip direction 180°). The orange
ellipsoid uses the static dip and dip direction with (1) Convert the expected plunge using: New plunge
a 50° plunge, whereas the blue ellipsoid has been = 90 – Plunge.
modified as above using the major axis transformation (2) Convert the dip, dip direction, and new plunge
to produce new inputs. The result is that the blue angles into Euler radians.
ellipsoid replicates the static ellipsoid anisotropy (3) Transform the Euler radians to a rotation matrix
while reflecting the local trend represented by around axes ZXZ (static or intrinsic rotation).
a wireframed surface. The blue ellipsoid correctly (4) Back-transform the rotation matrix to Euler radians
applies the 50° plunge through all undulations or around axes ZXY (static or intrinsic rotation).
bends of the surface. The result is 9.4° net difference (5) Back-convert the dip, dip direction, and new
in the dip direction, which if not corrected, would plunge radians to degrees.
counter the effort put in to generate the most appro­ (6) Use the newly converted degrees as input for
priate surface type discussed earlier and may misalign flagging.
samples during estimation and thus affect the grade
continuity that may be crucial for the deposit type. The key to the above process is converting the rotation
matrix from being around a ZXZ rotation (a static
proper Euler angle) to a ZXY rotation (a static Tait–
Adjusting Euler angle rotation axes
Bryan angle). Essentially, this converts the plunge from
Another option to account for plunge is to use rotation being a rotation around the Z-axis to being an inde­
matrices to change the plunge from being applied within pendent rotation around the Y-axis after the first two
the Z plane to being applied around its own plane (Y) angles (Z and X) have been applied (referred to as an
once the azimuth (Z) and dip of the plane (X) have been intrinsic rotation).
8 J. LUNNON AND R. PRESSACCO

This conversion is typically well handled using Inverse distance


python scripting, along with the NumPy (2022) and
To provide a level of “smoothing” between the data
Transforms3D (2016) libraries. The setup of this is far
points, inverse distance weighting to a relatively low
beyond the scope of this paper. Additional information
power can be used. The search ellipsoid orientation is
is provided by Diebel (2006) and Slabaugh (1999).
The benefit of using this conversion is that the updated generally set to the average strike/dip/plunge for the
azimuth, dip, and plunges can be directly used as inputs domain. Care and additional visual validation should
without adjusting the major, semi-major, and minor axis as be used where there are strong variations in domain
required by major axis transformation. However, it typi­ orientation. When properly performed, the inverse dis­
cally requires a knowledge of advanced scripting and/or tance method will provide a smooth variation in the
python to be a workable option. Therefore, the major axis estimated values, particularly where there are few
transformation is often preferred because it can be used informing DA data points. However, when mistakes
directly within many geological software packages. occur, such as averaging similar azimuths of 359° and
1° resulting in an incorrect value of 180°, the resultant
search orientation can be wildly incorrect. Validation of
CODING THE DA ANGLES TO THE BLOCK MODEL the results after each estimation run is critical to com­
pleting a successful DA estimation run.
If a surface has been generated for use as inputs to a DA
Typically, these issues arise from averaging dips or
estimation run, the orientations and XYZ coordinates of
dip directions that are computationally considered as
the individual triangles can be extracted to a table by the
discrete values rather than as angles that have limits on
geological software so they can be read during grade esti­
their minimum and maximum values. Figure 5 illus­
mation, and any plunge corrections required can be
trates an example where data from a NNE trending
performed.
surface are extracted into a series of data points having
Depending on the requirements of the practitioner’s
azimuths generally between 0° and 10°. However, due to
geological software, the dip-directions may need to be
a local variation, one point has an azimuth of 359°. If one
converted into azimuths. The exact input types, rotation
averages the azimuths at the estimated location (ignor­
convention, and if a dip is expected to be negative/
ing the distance component of inversed distance for
positive can depend on the software. Therefore, it is
simplicity), then the large value overpowers the remain­
highly recommended that the practitioner undertake
der, and an average angle of 76.6° is returned (calcula­
background reading to understand their software’s rota­
tion B), which is oblique to the trend surface and is
tion convention, particularly if they will be swapping
incorrect. It is necessary to correct for this large azimuth
between different software packages.
value to ensure that the average azimuth comes out at
Most geological software codes read the DA angles
the correct value of 4.6° (calculation A). One correction
directly from the block model on a block-by-block basis.
method is to apply a logic statement to convert dip
Thus, the extracted DA angles need to be flagged to the
directions/azimuths so that the angles are continuous.
block model prior to executing a DA grade estimate.
The example in Figure 5 converts the 359° azimuth
Nearest neighbor or inverse distance interpolation are typi­
reading to −1° for the purpose of grade interpolation.
cally used to assign point angle information to the model.
A variation of this issue commonly occurs when
a deposit consists of steeply dipping mineralized zones
that might overturn their dips past the vertical, which
Nearest neighbor
effectively flips the dip direction. This can be corrected by
By far the simplest out of these two methods is the nearest converting the azimuths into one direction and increasing
neighbor method, which simply writes the DA angle infor­ the dip angle through −90° to compensate (e.g., a −85° dip
mation to each block based on the sample that is located toward NE045° becomes −95° dip toward SW225).
closest to the block centroid in question. The advantage of Any corrections undertaken should be checked in the
using a nearest neighbor estimation is that there are fewer resultant model and azimuth angles above 360° or below
opportunities for human error to occur because there are 0° and dips greater than −90° corrected in the final
fewer input values required. However, a disadvantage is model to be within the expected range.
that if the input information is widely spaced (e.g., large Another source of potential error is when an input
triangles or widely spaced trend surfaces) or multiple sur­ wireframe is clipped against another wireframe, causing
faces are used (e.g., both hanging wall and footwall), then an abrupt angle change that is an artifact rather than
there can be abrupt changes in orientation between neigh­ a change in the mineralization orientation. If the practi­
boring blocks. tioner is sure that the orientation is not outside a certain
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 9

Figure 5. Estimating an average azimuth at an unknown location with no correction (A) and an azimuth correction (B)

range (e.g., the deposits dip is always between −85° and recommended that the DA angles are still flagged to the
−60°), then a filter can be applied to ignore all values model to facilitate internal validation efforts and internal/
outside this range. external peer reviews or audits.
Several correction steps may be required when selecting
inverse distance as the estimation method of coding the DA
angle information from the informing data points to the AXIS ORIENTATION
block model. Due to these challenges, it is often more
simple and time-efficient to use a nearest neighbor estimate Unfortunately, rotation conventions differ among geo­
to code the DA information into the block model, filter the logical software packages, and detailing them is outside
input data to an expected range of values, and reduce the of the scope of this paper. It is important to review the
wireframe triangle size to provide a better “smoothing” of software manuals and the accompanying Help sections
orientations. Whereas increasing the number of points to ensure that the rotation conventions are being imple­
used in the estimate could help reduce the impact of an mented correctly. This is especially true when some of
incorrect input datum, this approach should be taken care­ the analysis (such as variogram modeling) is being
fully because often there is a cluster of incorrect data that undertaken in a software package other than the one
significantly influences a larger area if several input values being used for DA/grade estimation. Further informa­
are used. tion on axis orientations is provided by Deutsch (2015)
and Reid (2017).

Summary
VARIOGRAPHY
The use of nearest neighbor or inverse distance methods to
flag the DA angles to the block model typically requires that Directional variograms model the similarity between
the practitioner ensures they are correctly setting up their paired data points in a particular direction. Therefore,
search window for interpolation of these angles, including when using DA, this relationship should be considered
selecting a reasonable ellipsoid/spheroid size and potential when the orientation of the mineralization is changing
orientation. When there are multiple domains, each with its along with the deposit. A misaligned variogram will result
own informing surfaces, these must be coded in such a way in higher variance and shorter ranges, which will give
that each domain is only being informed by the correct DA a higher weighting to the local mean rather than the
angles for that domain. individual data points. This runs counter to the intention
Some software packages have the functionality to read of localizing the model using DA. The orientation of the
the DA orientation information directly from a surface variography should be carefully considered in three dimen­
during the grade estimation procedure, rather than reading sions, and the angle tolerance and bandwidth should be
the information from a block model. It is strongly manipulated to provide optimum results.
10 J. LUNNON AND R. PRESSACCO

If there are marked changes in the variogram model block model is retained in a copy of the original model
along the length of the deposit, variograms will often be for peer review/auditing purposes if the final output
modeled in several locations and the corresponding model file is required to be as small a file size as possible
models applied locally to reflect the mineralization in with minimal fields. As with grade estimation, the vali­
that location. dation methods are both statistical and visual.
If the practitioner chooses to manipulate the data
before undertaking the variography, it is worth consid­
Statistical validation methods
ering the following:
Several basic checks can be taken on the DA angles
(1) The data are likely being estimated in nontrans­ directly on the orientation point file and the resulting
formed space, so longer ranges may not be justi­ block model. These include:
fied because, although a transformed variogram
may indicate a range of 250 m (for example), if ● checks on azimuths/dip directions below 0° or
DA is being used rather than unfolding, the above 360°;
change in domain orientation may mean that ● checks on positive or negative dips (software-
samples are not selected by the search ellipsoid dependent);
at this distance. ● loading values onto stereonet or histograms to
(2) Transforming the data to a single plane to create check for clusters of anomalous orientations (e.g.,
the variogram could be an additional source of many low angle dips present in a steeply dipping
error if it is undertaken incorrectly, and the body, or an azimuth that is oblique to the main
resulting variograms should be compared to orientation, Figure 6);
ones generated in the nontransformed space to ● calculating the variance between the DA angles on
ensure they make sense. neighboring blocks to highlight where a significant
(3) The most influential parts of the variogram are the change occurs; and
nugget, range, and sill of the first part of the vario­ ● calculating the standard deviation, variance, skewness
gram (likely corresponding to the first structure) and kurtosis between the model interpolated using
because the majority of the sample weight is applied DA and a standard search (Ruskov et al., 2020).
here. Additional effort to extend the range (particu­
larly when above 90% of the sill) may not be war­
Visual validation methods
ranted because the weighting on those samples may
be minimal, and the search ellipsoid may not A visual review of the data is a very important step in the
require those data to fill the block. validation process. One technique is to cut sections and
plans of the block centroids loaded as a point file that is
A further step could be employed to locally adjust the displayed as an arrow orientated using the azimuth
variography orientation on a block-by-block basis to flagged into the block. The trend surfaces can then be
fully optimize the weighting of the samples in each loaded in the same view and compared for compatibility
individually interpolated block. Significant care is and any abrupt changes. This should also be repeated for
required by the practitioner because this could be an dip angles in sectional view.
additional source of error, and the net result of making A second way of visually reviewing the orientation in
this change could be negligible, depending on a number sections and plan is to color the block model by the dip
of factors such as: direction/azimuth and then look for any colors in unex­
● the degree of orientation variation; pected orientations. The practitioner should consider the
● the nugget and range of the variogram; choice of color scheme and of gradational or discrete color
● the relative position of the selected input samples changes between value bins as well as the impact of a dip
for the estimation; and passing through −90° and effectively changing the azimuth.
● additional restrictions, such as number of samples All of these may either hide issues or overemphasize very
per drill hole. small changes.
Figure 7 illustrates an example of these visual valida­
tions. Both images present a plan section of a steeply
VALIDATION OF DA ANGLES
dipping body with a generally southerly azimuth repre­
The most crucial step in the DA process is validating the sented by the green arrows. However, two DA informing
final angles coded into the block model. It is therefore points have dipped past vertical, so the corresponding
very important that the DA information within the azimuth is now northerly at these locations. Because the
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 11

Figure 6. Illustration of a stereonet plot of a single domain with an unexpected cluster of azimuths indicating a potential error

model has been estimated using inverse distance weight­ Visual validation of the estimated grades should be
ing with no correction, these validations highlight that at undertaken after grade estimation is complete. The prac­
the transition between the points (indicated by 1), the titioner should review the continuity of the grade within
blocks are colored yellow, and a review of the centroid the block model and compare it to the informing compo­
arrows show that the azimuths are pointing east. site and assay samples and the practitioners understand­
Therefore, a search ellipsoid in this location would be ing of the deposit. The practitioner should question
oblique to the major mineralization orientation, result­ whether the grade distribution reflects what they expect
ing in incorrect grade estimations. of the model and whether there is any smearing/mixing of
grade as a result of misaligned search ellipsoids that
should be rectified through improving the DA or through
GRADE ESTIMATION AND VALIDATION better domaining/sub-domaining.

Once DA angles have been flagged to the block model,


the model can have grades estimated into it. At this
ALTERNATIVES TO DA
point, it is important to ensure that the parameters
selected for DA (e.g., azimuth over dip-direction) The two main alternative techniques to DA for reflecting
match the inputs and orientation convention required variable grade orientations in the input data are sub-
by the software. domaining and unfolding.
12 J. LUNNON AND R. PRESSACCO

Figure 7. Block model (a) colored by azimuth showing incorrect block azimuths and (b) centroid illustrated by icons orientated along
block azimuth. DA: dynamic anisotropy
Sub-domaining estimation. Typically, this is undertaken by transform­
ing the coordinates of the composite samples and block
As outlined above, a deposit can be sub-domained during
model using a surface (often a hanging wall and foot­
estimation. Along with careful application of soft boundaries
wall) so that the folds are “straightened out” (Garrido,
between sub-domains, separate search ellipses orientations
Navarro, Ortiz, & Moreira, 2016; Heriawan, Indarto,
are applied to these sub-domains. If the deposit is not
a continuous body, and the change in orientations is well Widayat, & Perdana, 2014). Data analysis and estima­
defined and not continuous (e.g., there is a series of offset tion are then performed in the new unfolded coordi­
faults or jogs with the mineralization continuing in nate system before returning the data to the original
a uniform manner between them), then this procedure coordinates for further analysis.
may be appropriate. The benefits are that it is easier to This unfolding technique tends to be more appropriate
implement, and there are fewer areas where procedural for sedimentary deposits that have undergone tectonic
errors may occur. The disadvantage is that the performance deformation because it can be assumed that the continuity
of this technique is compromised where there is either and characteristics of the deposit were fairly symmetrical
a gradual change in the orientation (e.g., a ring structure before deformation. This technique could also be consid­
or contact mineralization around intrusive bodies) or where ered where the model has been interpreted to be very
there is higher deformation (such as at the axis of a fold) and tightly folded and the drilling is typically wider spaced;
the mineralization shallows or steepens the down dip. thus using DA will not be possible to line one drill hole
from a limb to another. However, this type of deposit
comes with unique substantial geological/modeling risks,
and any resultant model would likely have low confidence.
Unfolding/unfaulting
A positive feature of this approach is that data analy­
Several geological software packages have the ability to sis, such as variography, will likely have longer ranges
unfold the deposit into a single plane prior to that represent the true nature of the deposit, rather than
DYNAMIC ANISOTROPY: DIFFICULT TO SAY, HARDER TO GET RIGHT 13

being limited by changes in orientation and variogram most significant errors can be caught. It is important to
controls, such as angle tolerance and bandwidth. ensure that such visual reviews are undertaken in cross-
Disadvantages to using this technique where there are section, long-section, and plan views.
high levels of deformation or local changes are that it can
be time consuming and highly susceptible to the input
ACKNOWLEDGMENTS
surfaces, which on a sedimentary bed are likely fairly
good. However, a pinching and swelling or nuggety gold The authors wish to acknowledge the contributions of Sean
deposit can make the hanging wall and/or footwall sur­ Horan (SLR Consulting—Canada) for his assistance on Euler
faces unreliable for this purpose. In addition, unfolding rotations and Donald Cameron (Cameron Resource
Consulting LLC) for his input on major axis transformation
is not appropriate where the deformation is the reason
and pole intersection surfaces.
the mineralization is concentrated in that zone. The authors also wish to thank the two CIM Journal peer
reviewers, whose comments and suggestions have greatly
improved this paper.

CONCLUSIONS
DISCLOSURE STATEMENT
DA is a very powerful tool at the practitioner’s disposal
that, when used correctly, can significantly improve the No potential conflict of interest was reported by the author(s).
local accuracy of a mineral resource model.
The most significant downsides of DA compared to
ORCID
the traditional approach of static search ellipsoids is the
increased time required for setup and the opportunity J. Lunnon http://orcid.org/0000-0003-4427-9691
for errors to occur. At best, these errors may provide R. Pressacco http://orcid.org/0000-0001-9286-5498
non-optimum orientations of the search ellipsoid. At
worst, they may produce an unreliable mineral resource
REVIEW STATEMENT
model. However, if the practitioner performs the appro­
priate due diligence on their models, this risk can be Paper reviewed and approved for publication by the
largely mitigated, and the time required may be shorter Geological Society of the Canadian Institute of Mining,
Metallurgy and Petroleum.
than that required to individually set up orientated
search ellipsoids on multiple sub-domains. At mini­
mum, this due diligence should consider: NOTES ON CONTRIBUTORS

(1) whether the deposit style is appropriate for DA; Jack Lunnon (MGeol, CGeol) is a VP—Geology with La
Mancha Resource Capital in London and has over 12 years’
(2) the quality of the input mineralization wireframes experience in the industry, primarily for RPA/SLR Consulting.
and grade data, including whether the geological He focuses on resource estimation and due diligence studies
modeling is of sufficient resolution to infer con­ and has a Citation in Geostatistics from the University of
nectivity based upon drill hole spacing; Alberta. jack.lunnon@lamancha.com
(3) the most appropriate type of input surface for DA; Reno Pressacco (MSc(A), PGeo) is a Consulting Geologist
(4) whether the flagging/interpolation of DA angles with more than 35 years of experience in industry, govern­
into the block model scripts have been constructed ment, and consulting. He is a contributing author to the CIM
correctly; Mineral Exploration Best Practices Guidelines and the CIM
Estimation of Mineral Resources and Mineral Reserves Best
(5) whether a plunge is evident and if it needs to be Practices Guidelines. In 2020 he received the Canadian
corrected in the DA angles; Institute of Mining, Metallurgy and Petroleum Robert Elver
(6) outcomes of the visual and statistical review of Award for mineral economics.
the resulting DA angles within the block model;
(7) whether the grade interpolation setup has been
configured to correctly read the DA angles;
REFERENCES
(8) outcomes of the visual review of the grade esti­ Deutsch, M. (2015). The angle specification for GSLIB
mation in comparison to the input data; and software. www.geostatisticslessons.com/lessons/
(9) outcome of stringent peer review by an experi­ anglespecification
Diebel, J. (2006). Representing attitude: Euler angles, unit
enced practitioner on any final models.
quaternions, and rotation vectors. www.researchgate.
net/publication/215458871_Representing_Attitude_
The authors highlight that the process of visual and Euler_Angles_Unit_Quaternions_and_Rotation_
statistical reviews is critical and is the point at which Vectors
14 J. LUNNON AND R. PRESSACCO

Garrido, M., Navarro, F., Ortiz, J., & Moreira, J. (2016). NumPy. (2022). NumPy reference (release 1.24). www.numpy.
Resource assessment with unfolding methodology: A case org/doc/stable/reference
study. Minin2016 Conference Paper. www.researchgate.net/ Reid, R. (2017). What is it with rotation conventions and
publication/314034839_Resource_assessment_with_unfold mining software providers? https://wantokgeoscience.
ing_methodology_a_case_study blogspot.com/2017/11/rotation-madness-and-estimation-
Glacken, I., Blackney, P., Gray, D., & Fogden, N. (2014). screw.html
Resource estimation in folded deposits – A review of prac­ Ruskov, K., Marinov, I., & Popov, K. (2020). Application of
tice and case studies. In Mineral resource and ore reserve dynamic anisotropy for the resource estimation improve­
estimation. The AusIMM guide to good practice (2nd ed., pp. ment in Pisani Skali ore occurrence, Bulgaria. Review of the
351–361). Monograph 30. Bulgarian Geological Society, 81(3), 172–174.
Heriawan, M. N., Indarto, A., Widayat, A. H., & Perdana, A. Slabaugh, G. (1999). Computing Euler angles from a rotation
(2014). Optimizing the resources model of “horse shoe” Cu- matrix. http://eecs.qmul.ac.uk/~gslabaugh/publications/euler.
Au porphyritic deposit using unfolding geostatistical meth­ pdf
ods. International Symposium on Earth Science and Transforms3D. (2016). Defining rotation. www.matthew-brett.
Technology. github.io/transforms3d/reference/transforms3d.euler.html

You might also like