RESUMEN
We aimed to understand the diversification history of jumping spiders in the Habronattus tarsalis species complex, with particular emphasis on how history in this system might illuminate biogeographic patterns and processes in deserts of the western United States. Desert populations of H. tarsalis are now confined to highly discontinuous oasis-like habitats, but these habitats would have been periodically more connected during multiple pluvial periods of the Pleistocene. We estimated divergence times using relaxed molecular clock analyses of published transcriptome datasets. Geographic patterns of diversification history were assessed using phylogenetic and cluster analyses of original sequence capture, RADSeq and morphological data. Clock analyses of multiple replicate transcriptome datasets suggest mid- to late-Pleistocene divergence dates within the H. tarsalis group complex. Coalescent and concatenated phylogenetic analyses indicate four early-diverging lineages (H. mustaciata, H. ophrys, and H. tarsalis from the Lahontan and Owens drainage basins), with remaining samples separated into larger clades from the Mojave desert, and western populations from the California Floristic Province of California and northern Baja California. Focusing on desert populations, there is a strong correspondence between RAD lineages and modern and/or paleodrainages, mirrored more finely in STRUCTURE and machine learning results. Non-metric multidimensional scaling analysis reveals strong congruence between morphological clusters and genetic lineages, whether the latter represent previously described species or H. tarsalis RAD lineages. Here we have uncovered a system that adds to our regional biogeographic knowledge in unique ways, using multiple types of evidence in a broadly-distributed terrestrial taxon. At the same time, we have discovered rapid evolution of both novel morphological forms and diverging genetic lineages. The hierarchical nature of variation in the H. tarsalis complex, the minute range sizes of many forms, the high likelihood that geographic distributions have shrunk and expanded through time, and signs of introgression all align with an ephemeral speciation model.