|
British Geological Survey, Kingsley Dunham Centre, Nicker Hill, Keyworth, Nottingham NG12 5GG, UK bgr{at}bgs.ac.uk
Data from a regional soil survey in eastern England have been used to determine whether samples over the same parent material can be discriminated on the basis of both individual and multielement geochemistry. Discrimination was based on estimates of measurement uncertainty, which were calculated from the analysis of a series of duplicates and subsamples. In the multivariate analysis we estimated a covariance matrix for the two sources of uncertainty and compared this to Mahalanobis distances calculated for pairs of samples within each parent material group. For 12 of the 19 individual elements, it was possible on average to discriminate between more than 80% of the samples within parent material groups and typically between 15 and 17 of the 19 elements discriminated individual samples. In the multi-element analysis, typically more than 99.8% of samples within the same parent material group were discriminated from one another. Hence, the geochemistry of a natural soil sample, when collected and analysed according to a strict protocol, and compared to a database that adopted the same methods, could be used to help establish provenance within bedrock-derived soil types. However, there are significant differences between the nature of soil samples and the way they are collected or derived in soil surveys and forensic investigations. These questions need to be addressed thoroughly before any practical application to forensic cases in which an investigator is attempting to link a suspect to a location based on soil geochemical signatures.
