Categorising features of geological terranes with geodiversity metrics: Enhancing exploration of multiple geological models

Most geoscientific research programmes benefit from three-dimensional (3D) representations of geology. Various elements of a geological target can be visualised, analysed and quantified to better understand the spatial properties of prospective terranes. For example, current technologies are able to produce useful measures that describe proven or prospective ore deposits. Essential information such as host and source rock proximity relationships can be estimated and analysed simultaneously with the location and prevalence of particular geological features (such as faults, lithologies, folds, resource estimates and mineral distribution) to generate 3D prospectivity maps that help to guide exploration activity. The geological elements of a 3D model are defined by a suite of data including field observations, geophysical interpretation and the prevailing tectonic evolution hypothesis. Field data (consisting of orientation measurements and lithological observations) are often supported by interpreted geophysics in covered terranes. In addition, the tectonic evolution hypothesis describing the timing of important geological events also has a large influence on the stratigraphic column, fault networks and interactions between the modelled elements. All of these input data are prone to error and uncertainty and may produce a model that does not adequately represent actual geology. In particular, a heavy reliance on geophysical interpretation introduces a high risk of ambiguity as it is difficult to explicitly identify lithological and structural fabric orientations. Subsequently much effort is made to remove error and uncertainty from the inputs to produce a single, optimised model that represents the geology in a useful and reliable manner. Removing error from the input data is difficult and, in some cases, almost impossible to perform. There is a risk that a reduced set of measurements that produces a model best representing the geology can be removed in the process. Our philosophy is to examine model reliability by simulating the error in input data. The data is subjected to uncertainty simulation prior to model input and involves varying strike and dip observations that determine modelled geological geometries. The subsequent sets of varied strike and dip observations are used to calculate multiple geological 3D models. The result is a suite of models that represent the range of possibilities offered by the input data set. We present this technique using a part of the palaeoproterozoic Ashanti Greenstone Belt, southwestern Ghana and the Gippsland Basin, southeastern Australia in a comparative case study. Geological knowledge in these regions can benefit from this technique as it produces an interesting set of 'what-if' scenarios, expanding our understanding of the interaction between geological elements considered important for gold mineralisation (Ashanti Greenstone Belt) or oil and gas prospectivity (Gippsland Basin). We perform analysis on the model suites using Principal Component Analysis (PCA) to determine the important features and characteristics. The mostdifferent models or 'end-members' of a model suite can be identified given a particular geological attribute, be it depth (deep or shallow) or volumes (large and small) of a particular stratigraphic unit, fault relationships or magnitude of deformation. These attributes, or 'geodiversity' metrics, can provide invaluable information to the geoscientist. The geodiversity metrics are then integrated into a combined study to answer questions regarding geological possibilities in the region providing a comprehensive understanding of geology in the respective geological terranes.