Orogenic gold deposits as nonlinear systems: Nonlinear analysis of data

Despite many studies of orogenic gold systems, the underlying processes involved in their formation and in defining their location and endowment remain enigmatic. This arises because such processes are multiscale and nonlinear so that patterns of alteration and mineralisation are apparently irregular and unpredictable. The goal of a nonlinear dynamical analysis is to extract the dynamics of the underlying nonlinear and multiscale physical and chemical processes that produced these data. We review nonlinear analysis methodology and explore hyperspectral and gold assay data for a drill-hole in an orogenic gold system. The analysis is non-parametric and purely data driven. We use recurrence, cross- and joint-recurrence plots to extract quantitative measures of the system and construct the attractor for the mineral distributions. The resulting dynamical model is tested using nonlinear prediction algorithms. Cross recurrence analysis shows strong spatial correlations of gold with carbonates and weaker correlations with phengitic micas and chlorite. Joint recurrence analysis reveals that all components of the system belong to the same dynamical attractor and hence are parts of the same physical–chemical system. This means that data from one part of the system can be used to predict other parts of the system. The probability distributions for mineral abundance are members of the Generalised Extreme Value family, reflecting the details of the ways in which the systems grow. We speculate on the coupled processes responsible for the mineralising system, propose that autocatalytic reactions associated with quartz and carbonate deposition contribute to pH variations responsible for gold deposition and present a new view of mineralising systems where the probability distributions reflect the endowment of the system. This approach permits a good prediction of the grade distribution into substantial volumes of rock within the mineralized system, outside the zone where the data were obtained. We suggest that such an approach has potential to become an important component of the definition of ore reserves, as the current methods, based on Gaussian or log-normal probability distributions, commonly produce poor outcomes as ore deposits are developed.