70-12 Using Grass Phytolith 3-D Morphology to Reconstruct the Evolution of Grasses

Session: Paleontology, Biogeography/Biostratigraphy & Phylogenetic/Morphological Patterns (Posters)

Poster Booth No.: 182

Presenting Author:

Jewel Wass de Czege

Authors:

Wass de Czege, Jewel K.¹, Hermans, Rosalie M², Gallaher, Timothy J³, Horn, Ash⁴, Lee, Hannah⁵, Scofield, Molly⁶, Smythies, Chiara⁷, Li, Mao⁸, Brightly, William H⁹, Stromberg, Caroline A. E.¹⁰

(1) Earth and Planetary Sciences, University of Washington, Seattle, WA, USA, (2) Archaeology, Environmental Changes and Geo-Chemistry Research Group, Vrije Universiteit Brussel, Brussels, Belgium, (3) Bishop Museum, Honolulu, Hawai'i, USA, (4) Biology, University of Washington, Seattle, WA, USA, (5) Biology, University of Washington, Seattle, WA, USA, (6) Biology, University of Washington, Seattle, WA, USA, (7) Biology, University of Washington, Seattle, WA, USA, (8) Donald Danforth Plant Science Center, Saint Louis, MO, USA, (9) University of Sheffield, Sheffield, United Kingdom, (10) Biology, University of Washington, Seattle, WA, USA; Burke Museum of Natural History and Culture, University of Washington, Seattle, WA, USA,

Abstract:

Grasses (family Poaceae) evolved in the Cretaceous and rose to ecological prominence, forming grasslands, during the mid-late Cenozoic, thereby dramatically reshaping terrestrial ecosystems and spurring animal evolution. Fossil phytoliths—microscopic silica bodies that form in/around cells in living plants—is the ideal tool to study grass and grassland evolution, given their diagnostic morphological variation within Poaceae. However, traditional ways of identifying phytoliths rely largely on 2-D shape and subjective, (semi-)qualitative morphotype definitions and outdated Poaceae taxonomy; thus, it remains uncertain how precisely fossil grass phytoliths can be classified.

Here, we present our efforts to develop robust, quantitative ways to use grass silica short cell phytoliths (GSSCPs) to determine taxonomic affinities of ancient grasses. We extracted GSSCPs from leaves of extant grass taxa, stained them with fluorescent dye, and used confocal microscopy to generate 3‑D surface models. The dataset currently encompasses ~5,800 3-D models from ~200 species representing all 12 Poaceae subfamilies. We analyze these data through three distinct approaches that provide complementary information on morphological variation: a) landmark‑based 3‑D geometric morphometrics, b) contour geometry from orthogonal 2‑D projections of 3‑D models (persistent homology), and c) 3‑D surface geometry represented by sampling points on a surrounding sphere. Using supervised classification algorithms, we develop models and evaluate their efficacy in placing modern GSSCP into Poaceae subfamilies, tribes, and subtribes. We also documented the distribution (costal vs. intercostal) and relative abundance of GSSCPs in situ in cleared leaf tissue from the same taxa. This information was used to correctly align (establish homology among) GSSCP morphotypes as well as to reconstruct the evolution of dominant GSSCP morphotypes within Poaceae.

Our preliminary results suggest consistent differences in phytolith composition across subfamilies, but overlap at lower taxonomic levels and class imbalance at every taxonomic level influence classification accuracy. Our goal is to apply our Poaceae-wide classification models to some of the oldest known fossil grass phytoliths, from the Late Cretaceous of central India. These fossils were previously assigned to several different grass subfamilies, including the Oryzoideae, but, given the subjective nature of traditional phytolith analysis these taxonomic placements have been viewed with skepticism. Application of our models will provide a robust test of earlier identifications and elucidate the timing of Cretaceous diversification and evolution within the grass family.

Geological Society of America Abstracts with Program. Vol. 57, No. 6, 2025

doi: 10.1130/abs/2025AM-10540

© Copyright 2025 The Geological Society of America (GSA), all rights reserved.