Thousands of years ago, prehistoric human societies traded the rare green mineral variscite. People throughout Western Europe between the sixth and second millennia BC valued the mineral's distinctive turquoise hue and used it in necklaces, rings and bracelets. Variscite has been found at various archaeological sites throughout the Iberian Peninsula, but tracing the origins of these finds has proven difficult.
Now, an interdisciplinary team of artificial intelligence experts and archaeologists from Portugal and Spain have collaborated to create an analysis pipeline that tracks variscite from a mine to an archaeological find, in a recent study published in the journal Journal of Archaeological Science.
“This is not just about green beads: it is about using artificial intelligence to tell human stories from prehistory,” said Daniel Sanchez-Gomez, lead author of the study and an archaeologist at the University of Lisbon, in press release.
Green stone samples at archaeological sites
The researchers tracked variscite by comparing geological samples excavated from modern mines with ancient variscite samples from archaeological sites. In their latest study, the team used portable X-ray fluorescence chemistry to determine the elemental composition of 1,778 variscite samples from three ancient mines in what is now Spain.
They then ran their data through a random forest AI algorithm, which uses a series of decision trees to improve classification accuracy.
“Our model learns to recognize the unique geochemical signature of each mine,” Sanchez-Gomez said.
After training the model to associate elemental signatures with specific mines, the team tested it on 571 variscites. beads extracted from archaeological sites. The analysis pipeline allowed the team to determine the source of each bead with 95 percent accuracy.
“It is able to determine the origin of a prehistoric bead even thousands of years after it was made,” Sanchez-Gomez said.
Read more: The Aztecs oversaw a vast network of trade in precious obsidian goods.
Amazing trade routes
The study's findings suggested that archaeologists may have to reconsider previously mapped historical trade routes. Previously, the mine near Encinazola in southwestern Spain was considered the main site of variscite production. The new analysis suggests that Encinasola played a smaller role in the production and spread of variscitis.
Instead, the main sources of the mineral are likely to be the mines at Gava and Alista in the Spanish provinces of Catalonia and Zamora, respectively. Variscite found in northern France has been geolocated to mines in the northern Iberian Peninsula, suggesting that it was traded along routes crossing the Pyrenees, rather than along sea routes as previously thought.
Using artificial intelligence methods
Importantly, the team made their data and the code underlying their analysis available through an open access repository called Zenodo. This will allow other researchers to independently study the data as part of efforts to improve open science.
The team prioritized artificial intelligence techniques that make their solutions understandable to users, rather than ones that operate in so-called “black boxes.” Carlos Odrizola, project leader and professor at the University of Seville, said the approach would increase the usefulness of their results for other researchers.
“We used explainable artificial intelligence techniques that allow AI models, especially the most complex ones, to clearly explain how they make their decisions. In the case of our study, this means that it not only accurately predicts, but also shows us which chemical elements were decisive in each classification, providing transparency and rigor to archaeological interpretation,” he said in a press release.
The researchers hope their approach can be applied to other materials from archaeological sites, such as amber. They also intend to follow the newly mapped route of variscite through Western Europe to try to determine what made it so valuable.
Read more: AI helps decipher mysterious prehistoric cave markings known as finger flutes
Sources of articles
Our authors in discovermagazine.com use peer-reviewed research and high-quality sources for our articles, and our editors review scientific accuracy and editorial standards. Review the sources used below for this article:
