Knowledge-guided data-driven evidential belief modeling of mineral prospectivity in Cabo de Gata, SE Spain

Spurious evidence and spurious spatial associations between target mineral deposits and certain classes of spatial data undermine GIS-based data-driven modeling of mineral prospectivity. In a case study application of data-driven evidential belief functions, such problems were recognized and then, based on sound geological judgment, were addressed accordingly. By invoking knowledge of genetic associations between mineral deposits of the type sought and spatial geological attributes (lithology, fault/fracture density, hydrothermal alteration intensity), spurious spatial associations depicted in ‘original’ evidence maps were addressed by treatment of input spatial data via applications of certain basic GIS functionalities in order to derive ‘treated’ evidence maps. By invoking knowledge of geological processes involved in the formation of mineral deposits of the type sought and knowledge of how operations to combine evidence maps function, the integration of evidence maps was guided such that the inter-play of geological processes involved in the formation of mineral deposits of the type sought is represented in the modeling procedure and such that spurious evidence is filtered and not transmitted into the output map representing likelihood of mineral deposit occurrence in every location within a study area. The results show that: (a) using ‘treated’ evidence maps, instead of ‘original’ evidence maps, results in better mineral prospectivity maps and, thus, (b) knowledge-guided data-driven modeling of mineral prospectivity is better than a ‘purely’ data-driven modeling of mineral prospectivity.