Delineating species boundaries in the framework of the multi-species coalescent (MSC) proves to be a reliable, objective, and reproducible method in an increasing number of studies. However, the underlying model assumes the lack of gene flow after speciation; an assumption which may be frequently violated in plant evolution. This study evaluates the robustness of currently available species delimitation methods implemented in beast (BFD, BFD*, and dissect) in the closely-knit ox-eye daisy group around Leucanthemum ageratifolium Pau. Comprising five taxa being allopatrically distributed between northern Spain and southern Italy this study group shows signs of hybridization with the widespread and codistributed species Leucanthemum vulgare (Vaill.) Lam. to various extent. As expected, our empirical analyses based on both AFLP fingerprinting and sequence data demonstrate that the robustness of species delimitation results is considerably influenced by the intensity of hybridization among species and the number of hybrid individuals included. Therefore, we set up a methodological pipeline with a first step of identification and subsequent removal of individuals showing admixed genetic patterns caused by actual interbreeding using AFLP-fingerprint and morphometric data, followed by application of different Bayesian MSC species delimitation methods based on the remnant individuals using both AFLP-fingerprint and sequence data (four nuclear markers, five concatenated intergenic spacer regions of the plastid genome). The results argue for acknowledgement of Leucanthemum laciniatum, L. legraeanum, and L. ligusticum as independent species, show the close relationship of L. ageratifolium, L. monspeliense, and L. vulgare, and give rise to the description of three nothospecies new to science.
Keywords: dissect; Bayes factor delimitation; hybridization; marginal likelihoods; multispecies coalescent; species delimitation.
© 2017 John Wiley & Sons Ltd.