A field-based analysis of the accuracy of niche models applied to the fossil record

Abstract

The use of ecological niche modeling (ENM) to estimate the geographic ranges of species is widely employed with modern fauna and is becoming more widespread in paleontology. Herein, field validation is utilized to assess the predictive accuracy of ENM methods for Paleozoic brachiopod species. This study represents the first field validation analysis of ENM methods in the fossil record. Previously published species distributions models for 8 Late Ordovician brachiopod species from the Cincinnati, Ohio region (United States) developed using GARP (Genetic Algorithm using Rule-set Prediction) were assessed for accuracy by comparing species occurrence data from a newly available set of 18 localities with the original species distribution models. Based on this data, the statistical significance of the original model set was assessed; 18 of the 22 original models were demonstrated to be statistically significant, based on field validation. Of the 140 individual species occurrences assessed, 60.8% were accurately predicted, 9.2% exhibited over prediction, and 30% exhibited under prediction. Accurate results were more common for species modeled from the greatest number of species occurrence points. The least accurate species models developed were for eurytopic species or those for which taxonomic affinities are unclear. Results indicate that with ample outcrop, well-defined stratigraphy, and sufficient fossil occurrence data, ENM methods could be successfully applied to many intervals in Earth history.