GSA Annual Meeting in Denver, Colorado, USA - 2016

Paper No. 76-9
Presentation Time: 9:00 AM-5:30 PM


ARENS, Nan Crystal, FISCHER, Mikhail Y., ANDES, Richard E. and CARR, Caroline E., Department of Geoscience, Hobart & William Smith Colleges, Geneva, NY 14456,

Traditionally, researchers group fossil leaves into morphotypes/morphospecies by hand based on expertise and intuition. Such groupings may be difficult to reproduce, with different experts circumscribing broader or more restrictive ranges of variation. If the goal is to recognize morphotypes that approximate the variation seen in biological species, a more objective approach would be desirable. We test an approach that combines hierarchical cluster analysis (which forms groups) with logistic regression (which tests for group homogeneity across variable ranges of variation) to identify species in a known set of modern woody eudicot leaves. We used the well-tested leaf description protocol of the Manual of Leaf Architecture (Ellis et al., 2009, Cornell University Press) to define 236 binary (to minimize magnitude biases) characters that are scored for each leaf in the collection. Size effects were minimized during character definition; proportional/shape relationships were retained. The collection includes leaves from the same stem, including inner and outer canopy leaves, different stems of the same species, different species in the same genus, different genera in the same family and a variety of families across the woody eudicots. The collection includes members of the Betulaceae, Cornaceae, Fabaceae, Fagaceae, Jugladaceae, Moraceae, Platanaceae, Rosaceae, Salicaceae, Sapindaceae, Tiliaceae, and Ulmaceae collected from developed areas on and around the Hobart & William Smith campus during June and July 2017. The region was experiencing a drought during the collection period. Leaves from the same stem, irrespective of canopy position, produced coherent groups, as did most defined species, irrespective of cultivar. Principle components analysis was used to identify the characters most influential in defining groups. After sensitivity analysis, the next phase of the research will apply the grouping protocol to Cretaceous-age leaf floras for which morphotypes were initially defined in the traditional way.