Reconstructing Body Size in Extinct Crown Cetacea (Neoceti) Using Allometry,Phylogenetic Methods and Tests from the Fossil Record

Living cetaceans exhibit interspecific size ranging across several orders of magnitude, and rank among the largest vertebrates ever. Details of how cetaceans evolved different body sizes, however, remain obscure, because they lack basic morphological proxies that have been traditionally used in other fossil vertebrates. Here, we reconstruct the body size of extinct crown group cetaceans (Neoceti) using different regression methods on extant skull and length data, in a phylogenetic context. Because most fossil cetaceans are fragmentary, we developed regression equations to predict total length based on cranial metrics that are preserved on most fossil crania. The resultant regression equations are based on a database of skull and length data from most extant lineages of cetaceans (n = 45 species; 272 specimens), sampling all living mysticete genera and all major clades of odontocetes. In generating predictive equations, we compared both conventional species data regression and independent contrast regression methods, as well as single trait predictors and a new approach that combines the advantages of a partial least squares (PLS) multivariate regression with independent contrasts. This last approach leverages the predictive power of using multiple correlated proxies. Lastly, we used the rare occurrences of fossil cetaceans with preserved total lengths to test the performance of our predictive equations for reconstructing body size from skull measurements alone. Our results demonstrate that incorporating information about phylogenetic relationships and multiple cranial measures in PLS scaling studies increases the accuracy of reconstructed body size, most notably by reducing prediction intervals by more than 70%. With this empirical foundation, we highlight the outline of major features in the evolution of body size for Neoceti and future opportunities to use these metrics for paleobiological questions.