Abstract:
We used BEASTmasteR (a collection of R scripts to write Beast2 analyses) to conduct a Bayesian total evidence analysis (tip-dating) to produce a dated phylogeny of 222 extant and extinct Canidae (64 in subfamily Borophaginae, 29 in subfamily Hesperocyoninae, and 129 in subfamily Caninae). To maximise taxon coverage, fossil taxa with no character data were included a priori based on expert opinion and analysed together with extant and fossil taxa. Morphological characters were extracted from previous analyses, while molecular data was gathered from GenBank. We assigned the approximate positions of the additional fossil taxa based on expert opinion a priori because this allows the inclusion of their temporal information into the analysis; in tip-dating the dates of fossil taxa are used as the primary source of dating information, rather than traditional node-based dating. We used last occurrence dates for our tip-dating that were gathered from fossil databases (primarily New and Old Worlds [NOW] fossil mammal database, but also Paleobiology Database [PBDB]) and the literature. Our results are moderately consistent with previous studies by other authors but the inclusion of over 50 fossil taxa previously excluded resulted in differences in the clade topology and timings. We also included extant Caninae from Asia, Africa, and South America that were not included in the previous large-scale analyses. We then conducted a historical biogeographical analysis only for subfamily Caninae because Borophaginae and Hesperocyoninae are almost entirely exclusive to North America. We assembled an occurrence database from NOW, PBDB, and the literature for geographic ranges, and compared the fit of many different models of biogeography using the R package BioGeoBEARS. Models that include jump dispersal gained over 99% of the AIC model weight. A newly developed detection model in BioGeoBEARS allowed us to estimate whether an absence of a fossil taxon from a certain region is a 'true absence' or just 'absence of evidence' based on the completeness of the fossil record (estimated from large mammal occurrences in the NOW database through time), resulting in more robust biogeographical models.