The outcome of species delimitation depends on many factors, including conceptual framework, study design, data availability, methodology employed, and subjective decision-making. Obtaining sufficient taxon sampling in endangered or rare taxa might be difficult, particularly when non-lethal tissue collection cannot be utilized. The need to avoid overexploitation of the natural populations may thus limit the methodological framework available for downstream data analyses and bias the results. We test species boundaries in rare North American trapdoor spider genus Cyclocosmia Ausserer (1871) inhabiting the Southern Coastal Plain biodiversity hotspot with the use genomic data and two multispecies coalescent model methods. We evaluate the performance of each methodology within a limited sampling framework. To mitigate the risk of species over-splitting, common in taxa with highly structured populations, we subsequently implement a species validation step via genealogical diversification index (gdi), which accounts for both genetic isolation and gene flow. We delimited eight geographically restricted lineages within North American Cyclocosmia, suggesting that major river drainages in the region are likely barriers to dispersal. Our results suggest that utilizing BPP in the species discovery step might be a good option for datasets comprising hundreds of loci, but fewer individuals, which may be a common scenario for rare taxa. However, we also show that such results should be validated via gdi, in order to avoid over-splitting.

Any type of text editor; Geneious Funding provided by: National Science FoundationCrossref Funder Registry ID: https://ror.org/021nxhr62Award Number: DEB 1937604 Funding provided by: Evert and Marion Schligner Foundation*Crossref Funder Registry ID: Award Number: Funding provided by: Charles UniversityCrossref Funder Registry ID: https://ror.org/024d6js02Award Number: UNCE 204069

Genomic data were obtained via Anchored Hybrid Enrichment (AHE) sequencing.