Radiolarite provenance using machine learning methods as a tool for understanding Palaeolithic human mobility

• LA-ICP-MS and pXRF elemental methods provide valuable information for the provenance of Central European radiolarites . • Machine learning algorithms, including LDA and RF, displayed promising results in classifying geological and archaeological samples. • West Slovakian radiolarites were used at the Nová Dědina I Aurignacian site (Moravia, Czech Republic). Elemental analysis of radiolarite samples from Central Europe was performed to estimate their provenance and to track the mobility of Upper Palaeolithic hunter-gatherers in Moravia (Czech Republic). The methods used were portable X-ray fluorescence spectrometry (pXRF) and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The primary aim of provenance studies is to distinguish between different possible sources, such as radiolarite outcrops. To address this problem, we created a training set for classification algorithms, including Linear Discriminant Analysis (LDA) and Random Forest (RF), based on geological samples from the outcrops. The classified geological samples were then used in an archaeological case study involving lithic artefacts from the Aurignacian site at Nová Dědina I (Czech Republic), mapping these artefacts to specific radiolarite outcrops. This approach allowed us to trace a particular hunter-gatherer mobility pattern during the early Upper Palaeolithic.