Travertine, a Building Stone Extensively Employed in Umbria From Etruscan to Renaissance Age: Provenance Determination Using Artificial Intelligence Technique

This work is focused on determining the provenance of travertine stones employed in the construction on a number of monuments in Umbria Region (Italy) from the Etruscan to the Renaissance age. To this aim we propose a new methodological approach based on the combined use of petrographic observations and statistical analysis of geochemical data. Analyses are performed on samples from monuments and quarries whose activity is documented since ancient times. Statistical analysis is performed using the conventional Principal Component Analysis and two new methods based Artificial Intelligence (a Self-Organizing Map and a Fuzzy Logic System). Results show that the Principal Component Analysis is a very poor technique to discriminate travertine provenance because wide overlapping between samples from different quarries is observed. On the contrary, the two Artificial Intelligence techniques show an excellent discriminative power for quarry samples. Application of these two techniques to monument samples also produces very good and concordant results, although some uncertainties in the determination of travertine for some monuments are observed. These uncertainties can be solved, in most cases, by combining results of the statistical analysis with petrographic observations. It is evidenced that a local provenance of travertine employed in the construction of ancient buildings is a common feature at any age in the past. In addition, it is suggested that a non-local provenance of this building stone may furnish information on the historical background in which a monument was conceived and built. Results from this study indicate that the combined used of Artificial Intelligence techniques and petrographic observations is a powerful tool for provenance determination of travertine employed in the construction of ancient buildings.