Lead isotope analysis for provenancing ancient materials: a comparison of approaches

Lead isotope analysis has been used to determine the provenance of metals such as lead, silver and bronze for many decades. Nevertheless, different approaches to interpret lead isotopic ratios have been proposed. In this study, three methods to couple the lead isotopic signature of archaeological artefacts to their possible mineral resources will be compared: the conventional assessment of biplots, a clustering method combined with calculating model ages (as applied by F. Albarede et al., J. Archaeol. Sci., 2020, 121, 105194), and relative probability calculations using kernel density estimates (as proposed by De Ceuster and Degryse, Archaeometry, 2020, 62(1), 107-116). The three different approaches will be applied to a dataset of lead isotopic analyses of 99 Roman Republican silver coins previously analyzed, pointing to a primary origin of the silver in the mining regions of Spain, NW-Europe and the Aegean, but showing signs of mixing and/or recycling. The interpretations made through the different approaches are compared, indicating the strengths and weaknesses for each one. This study argues that, although the conventional biplot method gives valid visual information, it is no longer feasible due to ever growing datasets. Calculating the relative probabilities via kernel density estimation provides a more transparent and statistically correct approach that generates an overview of plausible provenance candidates per artefact. The geological perspective introduced in the cluster and model age method by F. Albarede et al., J. Archaeol. Sci., 2020, 121, 105194 broadens the analytical spectrum with geologically informed parameters and improved visualization. However, the results when applying their method as a stand-alone approach are of low resolution and may lose archaeological relevance. Their approach regarding clustering should be revised.

Automated summary

This study compares three different methods for determining the provenance of metals using lead isotope analysis. The traditional biplot method is no longer feasible for large datasets, while calculating relative probabilities through kernel density estimation provides a more transparent and statistically correct approach. The cluster and model age method broadens the analysis with geologically informed parameters and improved visualization, but may have low resolution results and lose archaeological relevance when used as a stand-alone approach.