MULTIVARIATE ORDINATION OF AN OPPORTUNIST-DOMINATED FAUNA FROM THE FRASNIAN OF VIRGINIA: FINDING THE PALEOECOLOGIC GRADIENTS

The ordination of (paleo)ecologic samples faces a number of methodological challenges. For example, ordination techniques must overcome the classic arch effect. Tests of several ordination methods on samples from simulated coenoplanes (i.e., two-dimensional gradients) suggest that FSPA (flexible shortest path adjustment of ecologic distances) using a threshold of one, followed by nMDS (non-metric multidimensional scaling), can recover multi-dimensional gradients without introducing methodological artifacts. This method recovers a depth-related gradient in a data set from the Frasnian of Virginia, as well as a second ecologic gradient. However, the ordination is complicated by the presence of two opportunistic species that can occur at high abundance in more than one environment. Samples dominated by these species clump together, regardless of environmental provenance. Additional coenoplane simulations show that opportunistic species have this effect on a number of different ordination methods. Thus, the positions of opportunist-dominated samples in ordination space do not always represent their true placement with respect to paleoecologic gradients. Once the samples dominated by opportunists are removed, the depth gradient and secondary ecologic gradient are exposed much more clearly by FSPA.