We describe a highly isolated population of hairstreak butterfly from Waterton Lakes National Park, Alberta, Canada, previously recognized as the Half-moon Hairstreak (Satyrium semiluna), as a new species, Satyrium curiosolus sp. nov. We propose “Curiously Isolated Hairstreak” as the common name due to its disjunct and unusual distribution. Previous whole-genome analyses revealed S. curiosolus has extremely low genomic diversity and is highly divergent from the nearest S. semiluna populations in British Columbia and Montana, more than 400 km distant. Further analysis suggests prolonged inbreeding and isolation for up to ~40,000 years BP. Ecological niche modeling indicated that S. curiosolus occupies environmental conditions that are distinct from S. semiluna, suggesting niche divergence driven by long-term geographical and ecological separation. While host plant and ant associations have not been definitively resolved, they likely differ between S. curiosolus and S. semiluna. As part of this description, we provide whole-genome consensus sequences for each individual of the type series and identify 21,985 single nucleotide polymorphisms (SNPs) that are divergently fixed between S. curiosolus and S. semiluna, including 117 unlinked SNPs distributed across the genome as putative diagnostic markers. Previously listed as endangered in Canada as the Waterton population of S. semiluna, S. curiosolus should retain this conservation status due to its extreme isolation, small population size, and flatlined genomic diversity. We propose species recognition as a testable hypothesis under the General Lineage Concept, and recommend further research to explore the taxonomy, ecological relationships, and conservation of the greater species complex, including S. semiluna and S. fuliginosa.

Nuclear whole-genome consensus sequences (fastq format) were generated for each individual using individual-level BAM files (produced in genotype calling) and the mpileup command (-C 50, -Q 30, and -q 30) in samtools (Danecek et al. 2021). This was piped into the vcf2fq command from vcfutils.pl using our genome assembly as the reference. Filtering included sites with inferred consensus quality <20 and a read depth less than 8× or greater than two times each individual sample’s mean coverage, calculated from BAM files using samtools “depth”.