A NUMERICAL MODEL OF GRADIENT CHANGE AND ITS IMPLICATIONS FOR EMPIRICAL DATA

A prominent hypothesis in ecology and paleoecology is that taxa commonly maintain static, or similar, positions along environmental gradients over hundreds of thousands to millions of years. Testing this hypothesis, however, faces two hitherto unresolved obstacles. First, characterizing any two ecological gradients as "similar" is inherently a qualitative decision, and attempts to translate this judgment into quantitative terms are generally arbitrary and made outside the context of a forward model. Second, variations in sampling and faunal turnover between successive intervals may confound observations of gradient similarity. Here, we use a series of forward, numerical models that simulate the similarity of successive ecological gradients under varying conditions of sampling, faunal turnover, and taxon habitat preferences. These simulations indicate that sampling and faunal turnover do not confound assessments of gradient similarity, and any dissimilarity between two gradients reflects genuine, internally driven changes in the positions of taxa along the environmental gradient. Against this backdrop, we then interpret empirical data from the Middle Devonian upper Hamilton Group and Late Ordovician type-Cincinnatian Series. When both basins are analyzed in tandem with the model, we find significant differences both between and within basins in the level of observed gradient similarity. We affirm longstanding claims that the Hamilton records a period of pronounced gradient stasis, marked by enduring associations between each taxon and its associated environment, both when compared relative to the Cincinnatian and to the predictions of the numerical model. In contrast, we find that habitat preferences in the Late Ordovician are less well conserved, and some sequence-to-sequence comparisons are indistinguishable from random ecological change.