Eocene vegetation and ecosystem fluctuations inferred from a high-resolution phytolith record

A thorough understanding of the ecosystem factors that can act as drivers of vegetation change, an important factor for projecting impacts of future climate change, requires adequately distinguishing between plant community changes that occur on the local scale or over ecological time scales, and long-term trends. Phytolith assemblages have the potential to provide this information, but little work has been done so far to test their ability to do so in deep time. We examined vegetation patterns inferred from phytoliths (plant silica microfossils) from a single section of the Renova Formation, Timberhills region, Montana, dated to 39.2+ or -3 Ma (late middle Eocene). The section is composed of fluvial sediments overprinted by paleosols including Alfisols, Entisols, Inceptisols, and composite paleosols. Phytolith assemblages from 27 paleosol samples were used to reconstruct a high-resolution vegetation history for the area and were compared with paleosol data. Phytoliths are predominantly from forest plants and include tropical elements such as palms (Arecaceae) and gingers (Zingiberales), suggesting a paratropical forest. The high-resolution sampling records heterogeneous vegetation including shifts among closed forest, moderately open forest (forest gaps, edges, and woodland), and grass-dominated habitats. Grasses are interpreted as early open-habitat plants tolerant of drier conditions, but they appear to be replaced by forest vegetation with time, resulting in poor correspondence between paleosol type and phytolith assemblage at some stratigraphic levels, where paleosol morphology indicates a forest ecosystem and the phytolith assemblages indicate a changing grass-forest mixture as a function of depth within the profiles. This is the first study that has used phytoliths to construct a high-resolution record of vegetation from deep time, and we demonstrate that dense sampling and multiproxy approaches to paleoenvironmental reconstruction are necessary to capture vegetation heterogeneity in time and across microhabitats.