Multivariate geostatistical simulation using intrinsically correlated factors

Geostatistical simulation of multiple variables is a frequently encountered situation in mining. A polymetallic mineral deposit is a typical example of this. When the number of variables is more than one, there is a need for geostatistical simulation methods that will produce cross and auto spatial relationships. Cosimulation is developed for this purpose and it is basically based on cokriging. When the number of variables is four or more, using the cokriging method may not be practically possible. Especially modeling and reproducing cross variograms, also solving a number of large size cokriging systems limits the practicality of this method. For this reason, more practical and faster methods are needed for multivariate geostatistical simulation. One approach would be to transform spatially cross correlated variables into autokrigeable factors so that it is easy and fast to produce multivariate simulation by using the autokrigeable factors. First of all, the autokrigeable factors are independently simulated and then the simulations are back transformed into original space. There are two conditions causing autokrigeability. (1) cross variograms are zero at all distances (spatial orthogonality), (2) the ratio between cross and auto variograms is constant at all distances (intrinsic relationship). In order to facilitate multivariate simulations, the orthogonality method has been taken into consideration in the literature up to now. This thesis has developed a new multivariable geostatistical simulation method based on intrinsically correlated factor production. For this purpose, a method has been suggested which enables the production of intrinsically correlated factors and multivariable geostatistical simulations have been performed by independent simulations of each factor produced by this method. The method for geostatistical simulation of intrinsically correlated factors is coded in MATLAB. The method has been applied on 2 and 3 dimensional datasets. Multivariable geostatistical simulation methods are used in a wide range of applications from the calculation of grade-tonnage curves, the assessment of the risk of mining projects to the classification of mineral resources / reserves. When such methods are developed, the process intensity in polymetallic deposits will be considerably reduced compared to standard methods.