Abstract
Understanding past human and animal mobility is essential for reconstructing the social and cultural dynamics of ancient societies. Strontium isotope analysis (87Sr/86Sr) offers a powerful tool to investigate provenance and movement. The bioavailable 87Sr/86Sr follows the underlying lithology, and increasing efforts have been made to map its spatial distribution across the landscape and produce 87Sr/86Sr isoscapes at local and more global scales. Mongolia's steppe landscapes have long supported highly mobile pastoralist societies whose movements were central to the formation of major polities, including Late Bronze Age cultures, Xiongnu first nomadic state and Mongol Empire. Despite the growing use of 87Sr/86Sr isotopic analysis to investigate past mobility, the lack of a robust 87Sr/86Sr baseline has hindered large-scale interpretations. Here, we generated a regionally-calibrated bioavailable 87Sr/86Sr isoscape for Mongolia by analysing modern plants collected at 534 sites, mostly from Arkhangai, central Mongolia. We used two machine learning approaches: the random forest regression (RF) and the ensemble machine learning (EML). Both methods produced comparable isoscapes with RF slightly outperforming EML. Three major regions have been identified, reflecting the complex geology of Mongolia: a western intermediate-to-high 87Sr/86Sr region (0.710-0.722), a central low 87Sr/86Sr region (0.707-0.711), and an eastern intermediate 87Sr/86Sr region (0.709-0.712). Comparison of archaeological 87Sr/86Sr data from 25 sites across Mongolia from the Late Bronze Age to the Medieval period with local isoscape predictions suggest that human, animal and plant samples are mainly of local origins for most of the sites, regardless of the period. This isoscape provides a strong baseline in Mongolia and not only enhances archaeological interpretations of past mobility but also holds significant potential for ecological monitoring and the authentication of regional products, such as cashmere.