Predictive spatial maps of a mineralising system's geological features are highly desirable and useful from the industrial standpoint of resource exploration, mining, civil and geotechnical engineering. Additionally, such maps have academic value as theoretical models of complex natural phenomena. However, the production of these maps is a challenging endeavour using models based on geochemical data that are sparely populated with uneven statistical distributions. Using a Conformal Geometric Algebra based formulation that predicts geological features from geochemical datasets in a recent work by the authors (referred to as ‘the hyperfield formulation’), this work extends the methods and techniques to the construction of predictive spatial maps of these geological features. This contribution further introduces a novel “intersection space minimisation procedure” and the “raytracing interpolation procedure” and they are used to create comprehensive and accurate spatialised maps from a limited number of predictions in conjunction with neural networks and the hyperfield formulation. Case studies are presented where the sample's distance to mineralisation and the whole rock Au concentrations are predicted and mapped for the Mount Porter deposit, Northern Territory, Australia.