196
– Only a single mineralised horizon was defined for each drillhole, even if two or more horizons were evident
– Where mineralisation transgressed stratigraphy, the 2D modelling approach aligned mineralised horizons located in different stratigraphic positions for estimation
– Estimation using ID2 did not incorporate geological controls known from syn-sedimentary growth faults.
Later improvements to the 2D model were made possible through the addition of significant drilling to close the drill grid to 100 m or 200 m, resulting in an improved understanding
of the project geology and controls on mineralisation. Improvements included the introduction of non-vertical faults, the use of an innovative smoothing routine in Datamine utilising the ‘centre-of-gravity’ points for individual wireframe triangles to reduce the dimpling
effects, and a time-consuming process drawing control strings along fault block edges. Grade
estimation gradually incorporated greater geological controls through the use of anisotropic
searches aligned with trends evident in the variography and thickness trends of stratigraphic
units.
Dilution skins were added for improved mining study results. The initial steps towards 3D
modelling was incorporated into the Kakula model, where five stacked gridded models introduced limited vertical grade variability at 1%, 2% and 3% copper cut-offs.
To enable the construction of a full 3D model constrained within stratigraphic and mineralised horizon wireframes, certain key challenges to modelling had to be overcome:
– Stacked stratigraphic units representing changes in basin depositional environments are
extensive over kilometres, but some only a few metres thick
– Gentle folding, extensional and compressional faulting, and erosion on elevated domes
had to be accounted for, as did localised onlapping of stratigraphic units
– A large data set (nearly 1500 drillholes) made manual coding exercises time consuming
and ruled out surface creation through sectional interpretation and joining of interpreted
strings.
3.2 Overcoming surface modelling constraints
The development of uneven surfaces, pinch outs and abnormal thinning or thickening of
units was overcome through internal testwork (Nielsen 2014) through combining key functionality of Datamine and Leapfrog modelling software.
Modelling a surface in Leapfrog using just point data is susceptible to the same dimpling
effect and pinch-outs encountered using the trend modelling approach in Datamine. The use
of structural data in Leapfrog, where dip and dip direction are defined at specific coordinates, effectively defines a trend surface at each individual drillhole. Surfaces generated from
structural data are smoother and honour the expected trends far more effectively. The challenge is to determine the appropriate dip and dip direction for each drillhole position.
The contacts from each stratigraphic or mineralised unit are identified in Datamine and
exported to Leapfrog. A reliable marker horizon is identified and modelled from the point
data to create a reference surface for individual fault blocks. This surface is dimpled, but is a
good general representation of the marker horizon. The wireframe is exported to Datamine,
where the ANISOANG function is used to determine the dip and dip direction of each triangle of the wireframe. These values are estimated into a block model, with the estimated dip
and dip direction tagged back to the point data and reimported into Leapfrog as structural
data (Fig. 4). A key step, to prevent pinch outs or the creation of abnormal thinning or thickening between layers, is the assignment per drillhole of the identical dip and dip direction for
each unit modelled.
Surfaces modelled with the structural data are smooth (Fig. 5), realistic and successful
when tested by drilling, and require very little control along fault surfaces. The contact points
and the dip and dip direction used are generated in Datamine through a process that can be
carefully monitored and adjusted; the user retains full control of the modelling. The process
is robust, and largely automated through scripted modelling routines, creating a high quality
– Only a single mineralised horizon was defined for each drillhole, even if two or more horizons were evident
– Where mineralisation transgressed stratigraphy, the 2D modelling approach aligned mineralised horizons located in different stratigraphic positions for estimation
– Estimation using ID2 did not incorporate geological controls known from syn-sedimentary growth faults.
Later improvements to the 2D model were made possible through the addition of significant drilling to close the drill grid to 100 m or 200 m, resulting in an improved understanding
of the project geology and controls on mineralisation. Improvements included the introduction of non-vertical faults, the use of an innovative smoothing routine in Datamine utilising the ‘centre-of-gravity’ points for individual wireframe triangles to reduce the dimpling
effects, and a time-consuming process drawing control strings along fault block edges. Grade
estimation gradually incorporated greater geological controls through the use of anisotropic
searches aligned with trends evident in the variography and thickness trends of stratigraphic
units.
Dilution skins were added for improved mining study results. The initial steps towards 3D
modelling was incorporated into the Kakula model, where five stacked gridded models introduced limited vertical grade variability at 1%, 2% and 3% copper cut-offs.
To enable the construction of a full 3D model constrained within stratigraphic and mineralised horizon wireframes, certain key challenges to modelling had to be overcome:
– Stacked stratigraphic units representing changes in basin depositional environments are
extensive over kilometres, but some only a few metres thick
– Gentle folding, extensional and compressional faulting, and erosion on elevated domes
had to be accounted for, as did localised onlapping of stratigraphic units
– A large data set (nearly 1500 drillholes) made manual coding exercises time consuming
and ruled out surface creation through sectional interpretation and joining of interpreted
strings.
3.2 Overcoming surface modelling constraints
The development of uneven surfaces, pinch outs and abnormal thinning or thickening of
units was overcome through internal testwork (Nielsen 2014) through combining key functionality of Datamine and Leapfrog modelling software.
Modelling a surface in Leapfrog using just point data is susceptible to the same dimpling
effect and pinch-outs encountered using the trend modelling approach in Datamine. The use
of structural data in Leapfrog, where dip and dip direction are defined at specific coordinates, effectively defines a trend surface at each individual drillhole. Surfaces generated from
structural data are smoother and honour the expected trends far more effectively. The challenge is to determine the appropriate dip and dip direction for each drillhole position.
The contacts from each stratigraphic or mineralised unit are identified in Datamine and
exported to Leapfrog. A reliable marker horizon is identified and modelled from the point
data to create a reference surface for individual fault blocks. This surface is dimpled, but is a
good general representation of the marker horizon. The wireframe is exported to Datamine,
where the ANISOANG function is used to determine the dip and dip direction of each triangle of the wireframe. These values are estimated into a block model, with the estimated dip
and dip direction tagged back to the point data and reimported into Leapfrog as structural
data (Fig. 4). A key step, to prevent pinch outs or the creation of abnormal thinning or thickening between layers, is the assignment per drillhole of the identical dip and dip direction for
each unit modelled.
Surfaces modelled with the structural data are smooth (Fig. 5), realistic and successful
when tested by drilling, and require very little control along fault surfaces. The contact points
and the dip and dip direction used are generated in Datamine through a process that can be
carefully monitored and adjusted; the user retains full control of the modelling. The process
is robust, and largely automated through scripted modelling routines, creating a high quality