TESTING THE VALIDITY OF ECOLOGICAL NICHE MODELS FOR PALEOBIOGEOGRAPHIC RECONSTRUCTION: A CASE STUDY FROM THE LATE ORDOVICIAN OF THE CINCINNATI ARCH

Ecological niche modeling (ENM) is widely employed to estimate the geographic ranges of modern species. ENM predicts the geographic range of a species by first estimating the environmental parameters of a species’ fundamental niche and then projecting the distribution of the species to occur wherever that set of environmental conditions occurs in the geographic space of the study region. This method can be modified for use in the fossil record since the sedimentary units that preserve a fossil also record clues to the environmental conditions under which that organism lived. The ENM program, GARP (Genetic Algorithm for Rule-set Prediction), has previously been applied to model geographic ranges of Ordovician brachiopods, Devonian brachiopods and bivalves, and Miocene horses. In these studies, internal model testing indicated high model accuracy, but this was not tested through secondary field validation.

In this study, model results from GARP analysis of the Late Ordovician Corryville and Mount Auburn Formations of the Cincinnati Arch are field validated by secondary collection of species occurrence data at 18 localities. Locations of predicted species presence and absence from the original GARP models of eight brachiopod species are compared to the new field data. Of the 140 individual species occurrences assessed, 60.8% are accurately predicted, 9.2% exhibit under prediction, and 30% exhibit over prediction. The statistical significance of the original GARP models was further assessed using the binomial distribution; 19 of the original 22 GARP models are statistically significant.

The high accuracy of these results validates the utility of the GARP modeling system for paleobiogeographic studies. Results indicate that ENM methods have the potential to successfully elucidate biogeographic patterns during intervals in Earth history with ample outcrop, detailed stratigraphic control, and a dense set of fossil occurrence data.