Approximate Bayesian computation (ABC) is a powerful and widely used approach in inference of population history. However, the computational effort required to discriminate among alternative historical scenarios often limits the set that is compared to those considered more likely a priori. While often justifiable, this approach will fail to consider unexpected but well-supported population histories. We used a hierarchical tournament approach, in which subsets of scenarios are compared in a first round of ABC analyses and the winners are compared in a second analysis, to reconstruct the population history of an oak gall wasp, Synergus umbraculus (Hymenoptera, Cynipidae) across the Western Palaearctic. We used 4,233 bp of sequence data across seven loci to explore the relationships between four putative Pleistocene refuge populations in Iberia, Italy, the Balkans and Western Asia. We compared support for 148 alternative scenarios in eight pools, each pool comprising all possible rearrangements of four populations over a given topology of relationships, with or without founding of one population by admixture and with or without an unsampled "ghost" population.We found very little support for the directional "out of the east" scenario previously inferred for other gall wasp community members. Instead, the best-supported models identified Iberia as the first-regional population to diverge from the others in the late Pleistocene, followed by divergence between the Balkans and Western Asia, and founding of the Italian population through late Pleistocene admixture from Iberia and the Balkans. We compare these results with what is known for other members of the oak gall community, and consider the strengths and weaknesses of using a tournament approach to explore phylogeographic model space.
K E Y W O R D Sapproximate Bayesian Computation, Cynipidae, Hymenoptera, oak, phylogeography, western
Palaearctic ----------------------------------------------------------------------------------------------------------------------------------------------------------------------This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. (Beaumont, 2010;Bertorelle, Benazzo, & Mona, 2010;Csill ery, Blum, Gaggiotti, & Franc ßois, 2010;Pelletier & Carstens, 2014). ABC allows comparison of support for alternative models (combinations of population relationships and prior distributions for population and genetic data parameters; Cornuet et al., 2008;Lopes, Balding, & Beaumont, 2009;Pudlo et al., 2016;Wegmann, Leuenberger, Neuenschwander, & Excoffier, 2010) and assessment of confidence in model choice (Bertorelle et al., 2010;Cornuet, Ravign e, & Estoup, 2010;Hickerson et al., 2013).Where the number of population scenarios to be compared is small, support for all of them can be compared directly (e.g., Hearn, Stone, Nicholls, . However, it remains challenging to compare the large numbers o...