Résumé
Wood growth is driven by cambial activity, which is influenced by internal and external factors 1 . During fossilization, wood records the conditions (climate, disturbances) of the environment in which it grows. Fossil wood thus represents a precious archive of past environments. The Permian and Triassic periods were marked by major climatic changes, in particular the end of the Late Paleozoic Icehouse and the mass extinction at the Permian-Triassic boundary, which led to major shifts in vegetation 2 . This study focuses on fossil trunks from Autun basin in France (Early Permian) and two localities from Antarctica (Late Permian & Middle Triassic). In this study, a new automated chain of analysis was developed to record individual growth of fossil wood and measure lumen area of tracheids along cell files in the trunk from pith to bark. Based on Artificial Intelligence (AI), this analysis chain proved to have good results, close to human expert measurements. It was used to measure a large number of cells in several fossil woods with different preparations (thin-sections & peels) to analyze growth variation for each individual and propose paleoecological interpretations for tree growth in the Permian and the Triassic periods. The results reveal marked differences between Early Permian fossil wood from Autun (~299 Ma) and those from the Late Permian (~253 Ma) and Middle Triassic (~243 Ma) of Antarctica. In addition, an intra-individual analysis of growth rings based on the work on Taylor and Ryberg 3 identified local environmental signatures specific to glossopterids and corystosperms from Antarctica. The use of an automated analysis chain in this study opens up new prospects for furthering paleoecological interpretations of fossil wood samples by providing rapidly a large volume of quantitative data for large scale studies.