This paper discusses the applicability of the Principal Component Analysis-Agglomerative Hierarchical Clustering (PCA-AHC) approach to provenance studies of non-ferrous metals using combined Pb isotope and chemistry data. Pb isotopic ratios were converted to the natural abundance of individual isotopes and then to weight units. Next, all relevant variables (Pb isotopes and trace elements) were processed with PCA and AHC to examine the relationships between observations. The method is first verified on three literature-based case studies (1, 2, and 3). It is argued that, as is the case in archaeological iron provenance studies, the PCA-AHC method is also viable for non-ferrous metals. This method can greatly facilitate research, compared to conventional biplots with ratios of Pb isotopes and trace elements. Additionally, PCA-AHC can become part of the initial deposit selection process, and it can help clarify less obvious classification cases. The main problem with a practical application of this approach is insufficient deposit datasets with complete Pb isotopic and chemistry data. In such cases, it is possible to use the PCA-AHC method separately on Pb isotopic and chemistry data and then to compare and contrast results. Alternatively, the proposed approach can be used solely with Pb isotopic data. This application is shown in two additional case studies (4 and 5), which demonstrate the method’s application for tracing artefacts to their parent ores using datasets with a few thousand observations.

La base de données Oxalid rassemble des analyses isotopiques effectuées tant sur des minerais que des produits (semi) finis datant majoritairement de l’âge du Bronze. Cette étude propose l’application d’un traitement statistique multivarié à cette base de données partagée. Un prétraitement est effectué dans le but de sélectionner les données de minerais. Cinq ratios isotopiques définissant la signature isotopique de ces minerais sont traités à l’aide des statistiques multivariées (CAH, AFD). Les regroupements effectués d’après les données analytiques sont confrontés à la localisation des échantillons analysés. Une fois la signature isotopique des gisements cuprifères définie, l’outil de détermination de provenance est testé. On utilise pour cela des lingots dont le gisement d’origine est connu. Le lien entre minerais et objets (semi-finis) est ainsi validé. Il devient alors possible de proposer une provenance aux artefacts archéologiques cuivreux pour lesquels la signature isotopique a été déterminée.

The question of sources of metal supply and the resulting trade circuits are at the heart of archaeological issues. Provenance studies using lead isotopy has been used since the 1980s to identify the ore deposits from which the metal composing an archaeological object is derived. Indeed, metallurgical processes do not affect the isotopic signature of lead, which remains unchanged between ore and finished product. However, such studies require mineral signature repositories. Archaeologists have therefore built up databases, rich in thousands of analyses. However, these databases only very rarely include gitological information. In addition, lead isotopes are also used in geology but in order to characterize the metallogenic phenomena that lead to the creation of ore deposits. This different type of study requires different sampling: unlike archaeological databases, which include many analyses by mine, geological repositories have very few measurements by deposit. However, although containing few analyses per mine, geological data allow reflection in terms of restricted mineralized subsets and observation of marked groupings in binary diagrams as well as in multidimensional projection. The integration of gitological contexts as qualitative data could thus avoid the problems of statistical discrimination that are common in studies of the provenance of archaeological artifacts. The databases created by geologists, which also count thousands of mineral analyses, can thus be perfectly integrated into production source tracing studies. The finesse of the statistical breakdowns provided by gitological data also opens up new possibilities for data processing through the use of multivariate statistics. Geographical uncertainties are then a function of the quality of the available gitological metadata. Lead isotopic analyses carried out as part of geological studies have much to contribute to the tracing of archaeological production sources.

No abstract