Estimation of ancestral ranges
We used a Microsoft Access database to compile locality data from 4050 Preponini specimens deposited at five butterfly collections: the McGuire Center for Lepidoptera and Biodiversity, Florida Museum of Natural History, University of Florida (Gainesville, FL, USA), National Museum of Natural History, Smithsonian Institution (Washington D.C., USA), American Museum of Natural History (New York, NY, USA), Instituto Alexander von Humboldt (Villa de Leyva, Colombia), and the Instituto de Ciencias Naturales - Universidad Nacional de Colombia Sede Bogotá (Bogotá, Colombia). We georeferenced localities using Google Earth, literature and published/unpublished gazetteers, and cleaned the database to remove erroneous and imprecise localities. Our final database contained 1121 locality records for the 31 taxa in consideration.
We used the R package BioGeoBEARS 1.1.1 (Matzke, 2018) to estimate the ancestral range of Preponini under the Dispersal-Extinction Cladogenesis (DEC) model (Ree and Smith, 2008) and a maximum likelihood implementation of the Dispersal-Vicariance analysis (DIVALIKE) model (Ronquist, 1997). We did not include models with the parameter J since they seem less relevant to continental settings and have been shown to be difficult to compare statistically to other non-nested models (Ree & Sanmartín, 2018). We tested eight different hypotheses to evaluate the influence of different biogeographic events and distance among areas in the evolutionary history of Preponini (see Table 2 for details).
Briefly, the hypotheses account for the influence of the closing of the Panama isthmus, the formation of the Andes and distance among eight geographic areas in the Neotropics. We coded the areas as Central America, Caribbean, Chocó and Caribbean lowlands, Western Andes, Eastern Andes, Amazon (including Chaco and Cerrado), Guianas, and Atlantic Forest, are based on NatureServe’s classification of the Neotropics into ’Ecological Systems’ (Josse, 2003). Since one of our objectives was to test for the influence of major biogeographical events, we designated four time slices representing different paleo-geological stages. The four time slices used in our analyses were: (1) 32-23; (2) 23-10 Ma , (3) 10-7 Ma, and (4) 7 Ma-present which represent the gradual formation of the connection between South and Central America and Andean Uplift (Bacon et al., 2015; Condamine et al., 2012; Montes et al., 2015). We allowed the probability of movement across areas to change in time, accounting for geographic position, distance and for barriers to dispersal, and penalized accordingly (Table S4 – S11). We evaluated the effect of distance by penalizing dispersal probabilities with a factor proportional to the distance among the centroids of the areas (Table S5, S8, S9, S11).
Both DEC and DIVA are biased towards estimating widespread ranges in deep nodes (Buerki et al., 2011; Clark et al., 2008; Matzke, 2014; Ree and Smith, 2008). In an attempt to avoid this potential bias, the maximum number of areas any ancestor may occupy was set to six, since this is the maximum number of areas observed to be currently occupied by any single Preponini species (Ronquist and Sanmartin, 2011; e.g.Archaeoprepona demophoon ). We reduced the number of potential ranges in the reconstruction by including only geographic ranges with adjacent areas. This resulted in 113 potential ranges from the 248 possible combinations. To identify the hypotheses with strongest support, we compared AIC among the eight hypotheses and two possible models.
Results