Genomic clustering and admixture analyses
The quality filtered and trimmed reads were mapped to the M. longifolia reference genome (GCA_001642375.1_Mlong1.0; Vining et al., 2017) using bowtie2 with default settings. Mapped reads were quality filtered as outlined above and reads previously identified to map to the plastome were removed. The quality filtered mapping files were used to call single nucleotide polymorphisms (SNPs) in angsd v.0.931 (Korneliussen, Albrechtsen, & Nielsen, 2014). Only SNPs with less than 50% missing data across all individuals were retained and maximum depth to call a SNP within an individual was set to 50. SNPs were called using the SAMtools option and only genotypes with a posterior probability above 0.95 were used in downstream analyses. Genomic clusters were evaluated from the nuclear SNPs in PCangsd v.0.95 (Meisner & Albrechtsen, 2018) with default settings. The resulting covariance matrices were converted to eigen-values and visualized in R. Genomic admixture was evaluated using the nuclear SNPs in NGSadmix v3.2 (Skotte, Korneliussen, & Albrechtsen, 2013) with K increasing from 1 to 10. The best fit K was evaluated using the method of Evanno, Regnaut, and Goudet (2005) implemented in CLUMPAK (Kopelman et al., 2015) using ten independent runs of NGSasmix for each K .