Small-scale spatial resolution of the soil phytolith record in a rainforest and a dry forest in Costa Rica: applications to the deep-time fossil phytolith record

Phytolith analysis, a well-established tool in archaeology and Quaternary paleoecology, has become a source of data for deep-time paleoecological studies only in the last 15 years. Recent years have also witnessed the publication of numerous soil phytolith inventories from extant vegetation types, representing modern analogues. However, this work suffers from several methodological shortcomings limiting inter-study comparisons, and the development of a single, and repeatable protocol for soil sample collection. In this paper, we focus on two fundamental methodological questions that must be answered before modern analogue studies can be meaningfully applied to deep-time paleoecology. (1) Do phytoliths from the lower portion of the A-horizon (more commonly preserved in paleosols) reflect standing vegetation equally well as the upper A-horizon (which is more commonly used in modern analogue studies but often truncated by erosion in paleosols)? (2) Does a point soil sample (usually collected from paleosols) capture standing vegetation as accurately as a composite soil sample (commonly used in modern analogue studies), which is made from subsamples taken from a small area? To address these questions, we collected soil from 10 x 10 m quadrats in a rainforest (La Selva) and a dry forest (Palo Verde) in Costa Rica and compared phytolith assemblages from upper vs. lower soil A-horizons, and composites vs. single phytolith assemblages. At both sites, phytolith assemblages from the lower vs. upper A horizon, and from single vs. composite samples provide similar vegetation structure signal despite some compositional differences. These results suggest that the typical approach in deep-time paleoecology of taking point samples from the lower A-horizon of paleosols is likely justifiable (at least for paleosols reflecting rainforest and dry forest soils). We discuss possible mechanisms driving differences between sites, and future approaches to further expand and test our results.