Format: ORAL

Richard Hodel1,2 Alicia Talavera1 Will Weaver3 Alex White2 Mike Trizna2 Rebecca Dikow4 Liz Zimmer1 Ze-Long Nie5 Jun Wen1

1 Department of Botany, National Museum of Natural History, MRC166, Smithsonian Institution, Washington, DC, 20013-7012, USA 2 Data Science Lab, Office of the Chief Information Officer, Smithsonian Institution, Washington, DC, 20560, USA 3 Department of Ecology and Evolutionary Biology, University of Michigan, 1105 N. University Ave., Ann Arbor, MI 48109, USA 4 Computational Methods and Data, Yale University Library, Yale University, 130 Wall Street, New Haven, CT, 06511, USA 5 Key Laboratory of Plant Resources Conservation and Utilization, College of Biology and Environmental Sciences, Jishou University, Jishou, Hunan 416000, China

The phylogenomic age has revealed that the number of plant lineages that can successfully transition between temperate and tropical biomes is surprisingly low. One explanation for this trend invokes the difficulty of tropical lineages acquiring the necessary traits to tolerate the cooler, drier, and more seasonal temperate zone. Nevertheless, in rare cases, some lineages have been able to successfully migrate from the tropics to temperate zones during their evolutionary history. Presumably, these transitions were facilitated by the acquisition of morphological characters enabling adaptation to novel habitats (i.e., temperate environments). Synthesis of biogeographic studies of many plant lineages improved our understanding of the frequency of tropical-temperate transitions, but our understanding of associated trait evolution lags behind. Here, we use novel approaches for quantifying phenotypemachine learning algorithms that can classify and extract leaf, flower, and fruit features from digitized herbarium specimen imagesto infer the morphological characters that were important for enabling transitions from tropical to temperate biomes. Specifically, we used the machine learning package LeafMachine2 to extract leaf, flower, and fruit traits from tens of thousands of specimens in targeted groups. We used the plum/cherry genus (Prunus, Rosaceae) and the grape genus (Vitis, Vitaceae) because they represent rare lineages that successfully transitioned from tropical to temperate biomes. We used morphological data from tens of thousands of digitized specimens representing hundreds of species to trace continuous character evolution on time-calibrated phylogenies. This approach was selected so that the effects of both evolutionary history and past environmental conditions on observed phenotype could be considered. We measured leaf and reproductive characters and assessed phenology to identify features associated with biome shifts. Additionally, for comparative purposes, we investigated morphological variation within extant widespread species. The workflow developed for connecting morphological data with phylogenetic and environmental data is available as a user-customizable pipeline.