The spatial modelling of geometallurgical domains is an important step in the characterisation of an ore deposit allowing for the quantification of various rock types, their differing rock properties and the overall value of a project. The traditional approach to modelling such features is through the use of deterministic models, where a single estimate of the extents and structure of the ore deposit is mapped out by geologists based on available data and their interpretation of the geological processes. Any uncertainty in the layout of different domains is not accounted for and heterogeneities of chemical and physical properties of the orebody are likely underestimated. Spatial machine learning techniques can be used to derive geometallurgical categories, or classes, from multiscale, multiresolution and high dimensional measured rock properties. Subsequently, geostatistical simulations of these classes can be applied to obtain multiple equiprobable realisations that define the layout of the geometallurgical classes at unknown locations between boreholes, or domains. In-turn these realisations can be used quantify the uncertainties in the location of boundaries between different domains. The development of a workflow for defining geometallurgical domains and later geostatistically simulating them across the Orebody H; a complex stratabound Bedded Iron Ore deposit located in Western Australia’s Pilbara region is demonstrated. This could be used to identify zones of high uncertainty where collection of additional data might help mitigate or minimise risks and in turn improve forecast production performances.