La datation des aqueducs de Lyon : nouvelles données et contribution de l’analyse statistique bayésienne

Abstract

In this note, we present the results of the application of Bayesian analysis to the chronology of the Lyon aqueducts.Four aqueducts were built to supply water to the ancient city of Lyon/Lugdunum. For a long time, there was uncertainty over the dating of these aqueducts because of a lack of documentation. In recent years, preventive archaeology has provided substantial new data on the Yzeron, Brévenne and Gier aqueducts, with the result that it has been possible to narrow down their construction dates using ChronoModel chronological modelling software, based on Bayesian statistics. By applying Bayesian statistics to series of dates obtained by one or more chronometric methods, such as radiocarbon, archaeomagnetism, etc., we can construct chronological schemas that combine archaeometric dating with temporal knowledge such as historical dates or stratigraphic correspondences observed at the archaeological site.This also makes it possible to combine dates established by different methods to arrive at a more precise estimate of the date of a single event, such as the construction of a building. We used this model to process the dates of the aqueducts.For each aqueduct we had a number of absolute dates (obtained through carbon-14, dendrochonological and archaeomagnetic measurements) as well as chronotypological and epigraphical data. We ran all this temporal information through the (ChronoModel) software event date model developed by Lanos and Philippe (2017). We began by treating each aqueduct separately. We assumed that all available chronological data for the aqueduct were contemporaneous with its construction. We explained how we set out our model in ChronoModel, i.e. which dates were used and in what form form. We completed the model with a terminus post quem (TPQ) since none of the aqueducts could have been built before the foundation of Lugdunum. This statistical approach enabled us to propose a construction date, with a 95% confidence interval, for each aqueduct. We concluded our study by integrating the separate models established for the three aqueducts within ChronoModel, without adding any relation of anteriority or posteriority between them, as well as the TPQ of 50 BC. The resulting chronology, based solely on the use of the Bayesian model, is still awaiting discussion with archaeologists.