Professional Documents
Culture Documents
MULTIVARIATE RESOURCES
ESTIMATION
Diego Vergara & Xavier Emery
Contents
• Introduction
• Statement of the problem
• Methodology
• Case study
• Results
• Conclusions
Introduction
• Production plans rely on a model that provide
estimates of mineral grades on a block-by-block basis
Material
sent to dump
Material
sent to mill
Arsenic grade
3878 0 34.94 1.00 2.16
(g/kg)
Case study
Case study
• Choice of search neighbourhood
Three moving neighbourhoods are considered
– Neighbourhood 1: 4 data of each variable, in order to minimize the
smoothing effect of cokriging
– Neighbourhood 2: 16 data of each variable
– Neighbourhood 3: 60 data of each variable, in order to get as
accurate results as possible
Case study
• Choice of block size
Simulation and estimation are performed on a grid containing
9072 blocks with size 10 × 10 × 10 m
Results
• Comparison of true and cokriged copper grades
subject to a cut-off on cokriged copper grade
Results
• Comparison of true and cokriged arsenic grades
subject to a cut-off on cokriged arsenic grade
Results
• Comparison of true and kriged grades subject to
cut-offs on kriged copper and arsenic grades
(As < 1500 g/t; neighbourhood with 60 data)
Results
• Comparison of true and cokriged grades subject to
cut-offs on cokriged copper and arsenic grades
(As < 1500 g/t; neighbourhood with 60 data)
Conclusions
• The use of neighbourhoods with few data leads to a
considerable conditional bias in the grade estimates,
irrespective of whether kriging or cokriging is
considered.