Model selection in statistical historical biogeography of Neotropical insects—theExophthalmusgenus complex (Curculionidae: Entiminae)

ABSTRACTStatistical historical biogeographical methods rely on the use of models that assume various biogeographic processes. Until recently model selection remains an explored topic and the impacts of using different models on inferring biogeographic history are poorly understood. Focusing on the Neotropical weevils in theExophthalmusgenus complex (Insecta: Curculionidae: Entiminae), we compare three commonly used biogeographic models – DIVA (Dispersal-Vicariance Analysis), DEC (Dispersal-Extinction-Cladogenesis) and BayArea (Bayesian Analysis of Biogeography), and examine the impact of modeling founder-event jump dispersal on biogeographic history estimation. We also investigate the biogeographic events that have shaped patterns of distributions, diversification, and endemism in this group of weevils. We sample representatives of 65 species of theExophthalmusgenus complex and 26 outgroup terminals from the Neotropics including Caribbean islands and mainland. We reconstruct a molecular phylogeny based on six genes and performed molecular dating using a relaxed clock with three fossil calibration points. We conduct biogeographic history estimations and compare alternative biogeographic models with the R package BioGeoBEARS. Model selection strongly favors biogeographic models that include founder-event jump dispersal. Without modeling jump dispersal, estimations based on the three biogeographic models are dramatically different, especially at early diverging nodes. When jump dispersal is modeled, however, the three biogeographic models perform similarly. Accordingly, we show that the Neotropical mainland was colonized by Caribbean species in the early Miocene, and thatin situdiversification accounts for a majority (~75%) of the biogeographic events in theExophthalmusgenus complex. Our study highlights the need for testing for wide-ranging historical biogeographic processes in the study of Caribbean biogeography and the importance of comparing and selecting the best-fitting model in statistical biogeographic inferences. We demonstrate that modeling founder-event jump dispersal significantly improves the fit of the biogeographic history estimation of Caribbean and Neotropical mainland weevils. We establish thatin situdiversification acts as a dominant biogeographic force in the evolution of theExophthalmusgenus complex. The colonization of the Neotropical mainland from Caribbean islands reinforces the notion that islands can be an important source of continental diversity.