Addressing the effects of sampling on ecometric-based paleoenvironmental reconstructions
J Tyler Faith, Andrew Du, John Rowan
Ecometric analysis involves the examination of quantifiable functional traits across the taxa in a biotic community. Well-documented relationships between certain functional traits and environmental gradients in the present provide the empirical framework for a large body of research that uses ecometrics to reconstruct environments in the fossil record. In current applications of the technique, the taxa present in a fossil assemblage are summarized using the mean value of an environmentally informative trait. This study explores some of the quantitative pitfalls inherent to this approach. Through analysis of dental crown height—a trait that is widely used to infer paleo-precipitation—of Late Pleistocene ungulate assemblages from the Lake Victoria Basin in western Kenya, we illustrate how ecometric means vary as a function of sample size. Sampling artifacts have the potential to bias ecometric means, and the environmental inferences derived from them, whenever there is a non-random distribution of traits across the species abundance distribution (e.g., if abundant taxa have different traits than rare taxa). This sampling issue also means that the degree of analytical precision implied by quantitative paleoenvironmental reconstructions (e.g., annual precipitation at time X was 500 mm/yr) derived from ecometrics may be unwarranted. We recommend that analytical approaches be modified to circumvent these problems and explore three potential solutions: (1) specimen-based rarefaction, (2) coverage-based rarefaction, and (3) weighting ecometric means by taxonomic abundances. Of these, only the latter is robust to variation in sampling effort. Because abundance data are not always available and are potentially unreliable, we outline alternative approaches that could be implemented to contend with sample bias.