Enhancing potential field inversion techniques using geological uncertainty: improving three-dimensional geological models Lindsay, Mark Douglas 10.4225/03/58b4f72604052 https://bridges.monash.edu/articles/thesis/Enhancing_potential_field_inversion_techniques_using_geological_uncertainty_improving_three-dimensional_geological_models/4701109 3D modelling aims to solve geological problems, but these are always underdetermined and require prediction to produce geologically reasonable results. There are a series of related issues associated with 3D modelling techniques that require analysis, including geophysical ambiguity, sparse data and the subjective nature of geological interpretation. The combination of these modelling issues results in geological models being subject to uncertainty. In this thesis, we have developed ‘stratigraphic variability’ to detect and quantify geological uncertainty within 3D models. Stratigraphic variability determines the likelihood of finding different geological units at any given location within a model. A model suite is produced during uncertainty analysis, which contains a collection of models calculated from a single input data set. The effect of geological uncertainty on 3D model architecture is analysed with a set of ‘geodiversity’ metrics, a collection of analytical techniques that characterise each model within the model suite geometrically and geophysically. Geometrical metrics include: depth and volume of a geological unit; curvature and surface area of a contact; and geological complexity. Geophysical metrics analyse the observed geophysical response (representing nature), the calculated geophysical response (representing the model) and the residual (the difference between the observed and calculated responses). Geophysical metrics include: the root-mean-square misfit of the residual; standard deviation of the calculated response; information entropy of the residual; the 2D correlation coefficient of the calculated and observed response and; the Hausdorff distance between the calculated and observed response. Model space is mapped by identifying models that exhibit common and diverse characteristics, with the most diverse models defining the boundaries of model space and most common models defining the centre. The combination of uncertainty detection and model space exploration reveals the range of geological possibility. This thesis demonstrates a set of techniques that better describe the geological problem and provide guidance to geophysical inversion procedures. We demonstrate stratigraphic variability, geodiversity and inversion techniques on two case studies: the Gippsland Basin, a mature oil and gas prospective region off the coast of southeastern Australia; and the gold prospective Proterozoic Ashanti Greenstone Belt located in southwestern Ghana. Significant results are that: (1) model architecture can vary significantly due to uncertainty; (2) considered addition of data significantly reduces model uncertainty; (3) characterisation of models using geodiversity reveals geologically relevant characteristics of a model; (4) not constructing multiple 3D models from a dataset misrepresents the geological terrane; (5) model uncertainty is complex and affects the architecture of each model differently; (6) geophysical inversion can be assisted using results from both stratigraphic variability and geodiversity and (7) all these techniques can be applied to 3D models that represent different geological terranes. Thesis submitted for the degree of Doctor of Philosophy (PhD) - cotutelle. 2017-02-28 04:05:55 Inversion 3D Modelling 1959.1/909387 thesis(doctorate) Uncertainty Geodiversity Ashanti Greenstone Belt Model space exploration monash:120378 ethesis-20130612-142115 Gippsland Basin 2013 Open access