Adaptation to local environments is common in widespread species and the basis of ecological speciation. The song sparrow (Melospiza melodia) is a widespread, polytypic passerine that occurs in shrubland habitats throughout North America. We examined the population structure of two parapatric subspecies that inhabit different environments: the Atlantic song sparrow (M. m. atlantica), a coastal specialist; and the eastern song sparrow (M. m. melodia), a shrubland generalist. These populations lacked clear mitochondrial population structure, yet coastal birds formed a distinct nuclear genetic cluster. We found weak overall genomic differentiation between these subspecies, suggesting either recent divergence, extensive gene flow, or a combination thereof. There was a steep genetic cline at the transition to coastal habitats, consistent with isolation by environment (IBE), not isolation by distance (IBD). A phenotype under divergent selection, bill size, varied with the amount of coastal ancestry in transitional areas, but larger bill size was maintained in coastal habitats regardless of ancestry, further supporting a role for selection in the maintenance of these subspecies. Demographic modeling suggested a divergence history of limited gene flow followed by secondary contact, which has emerged as a common theme in adaptive divergence across taxa.

Data were generated from genomic DNA of song sparrows using restriction-site associated DNA sequencing (RADseq). Libraries were prepared following the 3RAD protocol with restriction enzymes NheI, EcoRI, and XBalI. Sequencing was done on an Illumina HiSeq 4000 with paired-end 150bp reads at the Duke Center for Genomic and Computational Biology in 2019. Data were processed in STACKS using the de novo map approach with a catalog set of 45 individuals and parameters M=4, n=4, and m=3.

Included are a compressed FASTA of the catalog loci and a compressed VCF of genome-wide SNPs after running filters (-R 0.8, --max-obs-het 0.7, --write-single-snp) using the STACKS populations module (before the application of a MAF filter) for all 163 individuals of the final population structure dataset.

A FASTA file of ND2 sequences, which are available on GenBank, is also included.