Bridging Dimensions in Cereal Phytolith Identification: A Comparative Study of 2D and 3D Morphometric Data

This research focuses on silica plant microfossils, called phytoliths, and particularly, a specific phytolith morphotype that is frequently encountered in grass inflorescences, ELONGATE DENDRITIC. In archaeological contexts, this morphotype is often ascribed to the presence of cereals. However, its formation in both cereals (wheat, rye, oats, and barley) and wild grasses forms a challenge in making definitive taxonomic distinctions, casting uncertainty on their reliability as a cereal marker. The primary objective is to explore the capability of ELONGATE DENDRITIC to differentiate between cereals and wild grasses, and to determine the extent to which they can be used to accurately classify cereals at genus and species level.

Last year, we prepared a substantial phytolith reference dataset of cereals and wild grasses that we used to develop a 3-D phytolith model dataset (n = 1.850 surface meshes) This dataset was created using confocal microscopy and Imaris image processing software. This year, we introduce a 2D phytolith image dataset (n = 9.100 phytolith outlines), generated using conventional light microscopy and ImageJ processing software. This dataset aims to provide a more accessible and cost-effective alternative for the identification of cereal phytoliths.

To quantify the morphology of the phytoliths, morphometric features were extracted from the 3D surface meshes and the 2D phytolith outlines. This included both traditional shape and size features and features deriving from Persistent Homology, which is a technique from algebraic topology used to quantify complex shapes.

By using a set of machine learning models, the study evaluates the performance of both datasets to identify phytoliths at species and genus level. The study also reflects on the usage of Persistent Homology as a quantification technique in phytolith research. Our goal is to ascertain the optimal balance between taxonomical precision and practical applicability in the identification of cereal phytoliths in archaeobotanical studies.