This talk describes the process of generating a 3D model of the Kevitsa (Finland) ore body through wavelet transform of geochemistry obtained from drill core. Tesselation is then used to determine an appropriate scale of study for the data and 3D modelling. Subtle signals are identified, while the effects of analytical noise are dampened through this process. A genetic model for ore body formation was also formulated due to the success of the data filtering process.
Disentangling the origin of chemical differences using GHOST
3D model of a Ni-Cu-PGE ore body - Margaux Le Vaillant and June Hill (CSIRO)
1. 3D Model of a Ni-Cu-PGE Ore Body
MINERAL RESOURCES
Margaux Le Vaillant, June Hill and Stephen J. Barnes
May 2017
The Kevitsa Cu-Ni-Au-PGE mine, northern Finland
2. Layered Intrusions
A simplified ore body model | Margaux Le Vaillant and June Hill2 |
• Repositories of some of the largest ore
bodies on earth of Cu, Ni PGE, Cr, V…
• Large intrusions presenting
compositional layering
This layering gives us
information on the genetic
magmatic processes at
play…
Need for a meaningful way
to visualise these variations
in 3D
5. Kevitsa 3D Model
A simplified ore body model | Margaux Le Vaillant and June Hill5 |
Variations within the intrusion poorly
modelled
6. Kevitsa 3D Model
A simplified ore body model | Margaux Le Vaillant and June Hill6 |
But access to a gigantic assay database! (>92,000 analyses)
Variations within the intrusion poorly
modelled
7. Ore Classification
A simplified ore body model | Margaux Le Vaillant and June Hill7 |
Ni < 2% - ‘False Ore’
S < 0.5% - ‘Not Ore’
Ni > 2% and Pd > 1,500 ppm
‘Normal Ore – High Pd’
Ni > 2% and Pd < 1,500 ppm
‘Normal Ore – Low Pd’
Ni > 10%
‘High Ni-PGE Ore’
8. 3D visualisation
A simplified ore body model | Margaux Le Vaillant and June Hill8 |
Assay points visualised and classified individually / no ‘lumping’ or domaining
9. A simplified ore body model | Margaux Le Vaillant and June Hill9 |
3D visualisation
Need for an objective & fast simplification of the model in order to distinguish
the large scale variations from the small scale one
Upscaling method using continuous wavelet transform (CWT) and
tessellation methods
WORKFLOW
11. Multiscale Boundary Detection
A simplified ore body model | June Hill11 |
Signal
Signalsmoothing
inflection point in signal:
best estimate of
boundary location
Increasingscale
many boundaries
few boundaries
12. A simplified ore body model | June Hill12 |
continuous wavelet transform
using
2nd derivative of Gaussian wavelet
(convolve wavelet with signal over range of scales)
(smooth & find inflection points)
Method
zero contours of
2nd derivative
13. A simplified ore body model | June Hill13 |
Multiscale Spatial Domaining
“Tessellation”
of
continuous wavelet transform
Method
involves
depth-correction of
zero contours
14. A simplified ore body model | June Hill14 |
signal Filter One scaleTessellation
15. A simplified ore body model | June Hill15 |
Combine
Domain
Boundaries
16. A simplified ore body model | June Hill16 |
Classify
Combined
Domains
17. Classification
by Domain
vs Interval
A simplified ore body model | June Hill17 |
Low S
High S, Low Ni
Mod Ni, Low PGE
Mod Ni, High PGE
High Ni
ORETYPES
IntervalDomain
18. Effect of
Filtering Level
A simplified ore body model | June Hill18 |
Low S
High S, Low Ni
Mod Ni, Low PGE
Mod Ni, High PGE
High Ni
ORETYPES
Weak filter Strong filter
50% 70%
19. Effect of
Filtering Level
A simplified ore body model | June Hill19 |
Low S
High S, Low Ni
Mod Ni, Low PGE
Mod Ni, High PGE
High Ni
ORETYPES
weak changes disappear
strong changes are preserved
Filter
27. A simplified ore body model | Margaux Le Vaillant and June Hill27 |
Interpretation
28. Genetic Model
A simplified ore body model | Margaux Le Vaillant and June Hill28 |
Interconnected sill sediment-complex choked with
country rock inclusions
29. Genetic Model
A simplified ore body model | Margaux Le Vaillant and June Hill29 |
Larger magmatic chamber
30. Genetic Model
A simplified ore body model | Margaux Le Vaillant and June Hill30 |
Freely convecting magma chamber
32. Conclusions
A simplified ore body model | Margaux Le Vaillant and June Hill32 |
• Consistent and objective reduction of the number of
units in each drill hole, and creation of a simplified
3D model of the orebody
Insights on the genetic processes at play
• Advantages of automated domaining process:
Consistent results over whole data base
Time saving
• Domaining using CWT and tessellation results in
domains whose size is a reflection of the location of
major changes in the variable values – not fixed
length like in conventional compositing of drill holes
33. A simplified ore body model | Margaux Le Vaillant and June Hill33 |
Thank you!
35. 35 |
S NiS PdIntervals Domains
Low S
High S, Low Ni
Mod Ni, Low PGE
Mod Ni, High PGE
High Ni
A simplified ore body model | Margaux Le Vaillant and June Hill
Making Sense of the Results
Notas del editor
mathematical method for performing spatial domaining on numerical drill hole data in order to extract valuable information from large assay datasets
The Kevitsa deposit is a large (237 Mt), low-grade disseminated Ni-Cu-(PGE) sulfide orebody in Arctic Finland formed within a layered ultramafic-mafic intrusion that lacks obviously recognisable internal stratigraphy
This assay database represents a wealth of information how do you interrogate it to extract fundamental ore genesis information?
Major challenge: distinguishing significant trends and patterns from background and short range variability…
new method of analysis of large geochemical datasets specifically intended to address the problem of distinguishing signal from noise
Shallow inward dipping cryptic layering defined by sulfide composition
Increasing tenors from bottom towards top of the intrusion
Metal enrichment of sulfides is recognized to be the result of interaction between sulfide droplets and silicate melt.
Metal enrichment of sulfides is recognized to be the result of interaction between sulfide droplets and silicate melt R Factor
Explaination of the observed variations: increasing mixing efficiency in an expanded magma chamber leading to higher effective R factors with time
wholesale assimilation of country rocks, combined with a limited amount of stirring, triggered the production of high S - low tenor sulfides, or ‘False ore’
continuous flux of magma being pumped into the system, allowing for more convection of the magma – interaction with larger volumes of magma – enrichment of the sulfide droplets
Progressive trapping of the sulfides within the cumulus pile
large regions in which there is very little change in the value of a variable will be lumped into a single domain, while narrow regions that contain values in strong contrast to their neighbourhoods will be preserved. This is particularly useful for identifying narrow regions of high grade or unusual composition.