Automated mineralogical profiling of soils as an indicator of local bedrock lithology: a tool for predictive forensic geolocation

Abstract

The use of soil evidence to identify an unknown location is a powerful tool to determine the provenance of an item in an investigation. We are particularly interested in the use of these indicators in nuclear forensic cases, whereby identification of locations associated with, for example, a smuggled nuclear material, may be used to indicate the provenance of a find. The use of soil evidence to identify an unknown location relies on understanding and predicting how soils vary in composition depending on their geological/geographical setting. In this study, compositional links between the mineralogy of 40 soils and the underlying bedrock geology, as documented in local-scale geological maps, were established. The soil samples were collected from locations with broadly similar climate and land use across a range of geological settings in a ‘test bed’ 3500 km2 area of SW England. In this region, the soils formed through chemical weathering of the bedrock, representing a worst case scenario for this type of forensic geolocation owing to the high degree of alteration of the parent rock during soil formation. The mineralogy was quantified using automated scanning electron microscopy–energy dispersive X-ray spectrometry analysis based on QEMSCAN technology. Soil mineralogy and texture as measured using this technique are consistent with the underlying geology as indicated by regional-scale geological mapping. Furthermore, differences between individual units of the same bedrock lithology, such as different granites, were identified by examining trace mineralogical signatures. From an investigative viewpoint, this demonstrated that rapid automated mineral profiling of soil samples could be used, in conjunction with readily available geological mapping or similar datasets, to provide indication of the areas from which a soil sample of unknown origin could, or could not, have been sourced.