Testing which axes of species differentiation underlie covariance of phylogeographic similarity among montane sedge species

Sample collection Samples were field-collected from 10 sites that represented the majority of the range for each species (see maps in manuscript: Fig. 2 and Fig. 3, and SI Table S1). In our experimental design, the small range species were sampled from 10 identical locations and the large range species were also sampled from 10 identical locations. Across all species, an average of 7.8 individuals were collected per sampling location. Individuals were sampled from the same sites for the two small-range species (C. bella and C. chalciolepis) across Colorado, New Mexico, and Utah (a range of approximately 250,000 km2) and for the two large-range species (C. epapillosa and C. pelocarpa) across Colorado, Idaho, Montana, Nevada, Oregon, Utah, and Wyoming (a range of approximately 1,000,000 km2). Due to the close proximity of different habitat types in montane regions, we were able to sample species with different habitat preferences from the same sampling sites. The average distance among sampled plants at a locality was 300m, and the minimum distance between samples was 35m to minimize the chance of sampling siblings or close relatives. For each sampled individual, leaf tissue was collected and placed directly into silica gel desiccant. Leaf tissue was kept in the dark and stored at room temperature until DNA extraction. DNA was extracted with DNeasy Plant Mini Kits (Qiagen, Hilden, Germany) using the manufacturer’s standard protocol. RAD-seq library construction and processing We used a restriction site associated DNA sequencing (RAD-seq) approach to generate thousands of anonymous loci using six single-end Illumina HiSeq 2500 sequencing lanes following the protocol of Peterson et al. 2012. We used ipyrad 0.7.20 (Eaton 2014) to demultiplex and filter reads from each library, allowing one mismatch per barcode and using the most stringent filtering to identify and exclude contamination from adapters and/or primers. As some individuals were represented in more than one library, every library was initially demultiplexed and filtered separately (ipyrad steps one and two). Reads corresponding to the same individual were then combined and all reads were trimmed to a length of 41 bp using the fast-x toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). We used a 90% clustering threshold for de novo assembly; all other parameter settings that affect steps three through seven were set as default (see SI Methods for additional details). Across all species, between two and 10 individuals were sequenced per sampling location, for a total of 89 C. bella individuals, 97 C. chalciolepis individuals, 71 C. epapillosa individuals, and 55 C. pelocarpa individuals (SI Table S1, S2).