Ninth International Geostatistics Congress, Oslo, Norway
June 11 – 15, 2012


Mining 5

Abstract No.:



High order simulation of a fault-controlled gold deposit using spatial cumulants.


D. F. Machuca-Mory, McGill University, Department of Mining and Materials Engineering (CA)
R. Dimitrakopoulos, McGill University, Department of Mining and Materials Engineering (CA)


High-order simulation using spatial cumulants are able to reproduce the low and high-order spatial statistics informed by a data-set and a training image. Spatial cumulants allow capturing the non-linear spatial features present in the data that traditional spatial statistics, such as the variogram and the covariance, cannot. This methodology is aimed to the modelling of non-Gaussian random fields showing non-linear spatial structures. However, its application to real datasets poses various challenges, such as the construction of a representative continuous training image, data clustering, alignment of samples along drill-holes, and underrepresentation of non-linear structures in the data. A real drilling dataset containing gold grades is used to illustrate these issues and to present some solutions to them. Methods for building a representative training image, such as sequential Gaussian simulation using local variograms and combined with single normal equation simulation are presented and discussed. Data clustering and structures underrepresentation is assessed by including declustering weights in the calculation of the low and high order cumulants. The preferential sampling along drill-hole traces is treated using a search template divided in octants. The results of implementing these measures are presented and discussed.




    Copyright 2012 International Geostatistics Congress

Produced by X-CD Technologies Inc.