Data from: Testing classical species properties with contemporary data: how 'bad species' in the brassy ringlets (Erebia tyndarus complex, Lepidoptera) turned good

All species concepts are rooted in reproductive, and ultimately
genealogical, relations. Genetic data are thus the most important source
of information for species delimitation. Current ease of access to genomic
data and recent computational advances are blooming a plethora of
coalescent-based species delimitation methods. Despite their utility as
objective approaches to identify species boundaries, coalescent-based
methods i) rely on simplified demographic models that may fail to capture
some attributes of biological species, ii) do not make explicit use of the
geographic information contained in the data, and iii) are often
computationally intensive. In this paper we present a case of species
delimitation in the Erebia tyndarus species complex, a taxon regarded as a
classic example of problematic taxonomic resolution. Our approach to
species delimitation used genomic data to test predictions rooted in the
biological species concept and in the criterion of coexistence in
sympatry. We 1) obtained RAD sequencing data from a carefully designed
sample, 2) applied two genotype clustering algorithms to identify genetic
clusters and 3) performed within-clusters and between-clusters analyses of
isolation by distance (IBD) as a test for intrinsic reproductive barriers.
Comparison of our results with those from a Bayes Factor Delimitation
(BFD*) coalescent-based analysis, showed that coalescent-based approaches
may lead to overconfident splitting of allopatric populations, and
indicated that incorrect species delimitation is likely to be inferred
when an incomplete geographic sample is analysed. While we acknowledge the
theoretical justification and practical usefulness of coalescent-based
species delimitation methods, our results stress that, even in the
phylogenomic era, the toolkit for species delimitation should not dismiss
more traditional, biologically grounded, approaches coupling genomic data
with geographic information.