Determining the provenane of Mediterranean obsidianartifats

Providing a generally valid model to lassify and determine the provenane of obsidian arhaeologial artifats of the Mediterranean area, was the aim of this researh. This has been carried out to overcome the actual problem of the wide uncertainty arising when trying to distinguish samples from similar, although separate sources or to compare them to literature data. Reported chemical analyses of obsidian are characterized by being obtained with a variety of analytical methods, laboratories or sets of analyzed elements, introducing a ma jor difficulty when trying to use these data for provenance studies of ancient artifacts. The development of a comprehensive statistical method applied to take into consideration all the chemical parameters characteristic of each obsidian source has been carried out using different statistical methods, organized in a sequence that allows the discrimination, step by step, of all the sources. The method is based on the building of a large chemical database reporting both geological and archaeometrical literature obsidian data from the volcanic districts of the Mediterranean - Aeolian, Pontine, Sardinian, Pantelleria, Carpathian and Aegean. As a first step, comparison of Normal Probability Plots between different sources and interrelations between elemental abundances allowed us to individuate discriminating elements and tracers, like Sc, Rb, Cs, Sr, Ba, Ta, Hf, Zr, Nb, Y, Th, U and REE. Then multivariate data analysis allowed us to develop effective discriminating models, using Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Soft Independent Modeling of Class Analogy (SIMCA) and Partial Least Squares Discriminant Analysis (PLS-DA). The established procedure allows to select the discriminating elements to be compared with the geological fingerprint of different sources. The procedure, tested on both literature and new data, has the advantage of making the provenance study of obsidian artifacts independent from the analytical method used as well as from the need of large number of reference samples.