Realistic representations of multivariate natural phenomena such as many mineral deposits or petroleum reservoirs need to consider and reproduce the relationship between the variables. However, there is currently no method that can practically simulate multivariate real-size deposits. In this thesis, a method for the conditional simulation of a non-Gaussian vector random field on different support is presented. The method, derived from the direct block simulation algorithm, is shown to efficiently joint simulate large multivariate deposits directly on block support. Such a method permits the simulation of multivariate real-size deposit and is computationally very efficient.
The method, termed DBMAFSIM, is a multistage process. First, a vector random function is orthogonalised with minimum/maximum autocorrelation factors. Blocks are then simulated by performing a LU simulation on their discretised points, which are later
back-rotated and averaged to yield the block value. The internal points are then discarded. Only the block value is stored in memory and is used for further conditioning, resulting in reduction of memory requirements and file storage. This method is successfully applied at Yandi Central 1, an iron ore deposit located in Western Australia. Five attributes are jointly simulated; iron, phosphorus, silica, alumina, and the loss on ignition.
The capacity to obtain multivariable stochastic representations of mineral deposits allows for the use and the development of more complete transfer functions, such as mine planning or production scheduling that requires all the significant attributes of a deposit. This would yield more reliable figures for highlighting risk associated with the exploitation of the resource.