HERNANDEZ-CASTILLO, GENARO RODRIGO. Department of Environmental and Plant Biology, Ohio University, Athens, OH, 45701. - Use of multivariate analysis in reconstructing fossil conifers.
The fossil record of conifers is characterized by fragmentary remains
of simple organs with cryptic characters. There is also a wide range
of variation from branch to branch on the same plant. Therefore,
meaningful species delimitation is extremely difficult for many fossil
confers, particularly Paleozoic conifers. Different authorities often
assign the same specimen to different species, and the same authority
may assign the same specimen to different species at different times.
A large number of Upper Pennsylvanian conifer specimens from the 7-11
mine in Columbiana, Co., Ohio are preserved by a combination of
coalified compression and pyrite permineralization. These specimens
display features of external morphology, cuticular characters and
internal anatomy, and can all be assigned to a single species with a
high degree of confidence. The confident recognition of this single
species allows us to record ranges of variation for a large number of
characters, and to recognize differences that result from variables
such as ontogeny, taphonomy, and/or different positions on the plant.
Several multivariate methods have been employed to interpret
similarities and differences resulting from each of these variables.
Principal Component Analysis (PCA), Discriminant Function Analysis
(DFA), and Cluster Analysis (CA) have been used to compare the
features of different specimens. Individual specimens are best
characterized by leaf length and width, angle of divergence, and
diameter of vegetative shoots. Differences in these characters are
correlated with shoot length and diameter. Angles of divergence in
leaves seem to be a consistent feature among all the specimens. Such
characters may explain differences among fossil conifer specimens due
to biological or taphonomic variables, and they may be valuable for
reliably distinguishing among species.
Key words: conifers, multivariate methods, Upper Pennsylvanian