Ecological Interchangeability
We evaluated a proxy for ecological interchangeability by measuring the overlap between ENMs for the sister lineages being compared (generally following the approach of Stockman and Bond 2007) as well as taking into account previous morphological and behavioral criteria for distinguishing A. microunicolor (Hendrixson & Bond, 2005a). Current climate data from 1970-2000 for 19 bioclimatic variables at 30 arc-second resolution were downloaded for tiles 12 and 13 from WorldClim v.2 (http://worldclim.org/version2; Fick and Hijmans 2017). Climate data from the tiles were then merged into layers, cropped to the area of interest, and converted to a raster stack using the packages ‘raster’ (Hijmans, 2015) and ‘rgdal’ (Bivand et al., 2019) in R. The software ENMTools v1.4.4 (Warren et al., 2010) was used to estimate the amount of correlation between the 19 bioclimatic variables, with significant correlation defined as r > .90 following (Jezkova et al., 2011). Six variables were removed and the 13 retained variables (BIO 2-9,12,13,15,16,18) were used for generating ENMs. Maxent v.3.4.1 (Phillips et al., 2006) was used to estimate ENMs with default settings for the lineages. The receiver-operating characteristic (ROC) plot’s area under the curve (AUC) was used as a measure of model prediction accuracy, with values > 0.9 indicating optimal model performance as opposed to values < 0.7 indicating poor model performance (Swets, 1988). Occurrence records were based on specimens collected for this study and specimens from prior publications (Hendrixson & Bond, 2005a, 2005b) also included in this study.
To statistically compare the ENMs of each lineage, we conducted analyses of niche overlap, niche identity, and niche similarity (i.e. background) tests in ENMTools (Warren et al., 2008, 2010). Niche overlap for each sister lineage comparison was quantified using Schoener’s D(Schoener, 1968), which ranges from 0 (no overlap between ENMs) to 1 (complete overlap of ENMs). To assess the significance of D , we employed both niche identity and niche similarity tests. For the niche identity test, 100 pseudoreplicates were used to construct a null distribution of niche overlap compared to the observed overlap value using a one-tailed test (Supplemental Figure S3). Warren, Glor, and Turelli (2008) highlighted that the niche identity test may be too strict and results in the null hypothesis often being rejected, so the more conservative niche similarity test was also conducted. The background regions for each lineage were generated based on minimum area polygons from occurrence points in ArcMap v10.7 (ESRI), and occurrence points of one lineage were tested against random points from the background region of the other lineage and vice versa. A hundred pseudoreplicates were used to construct a null distribution of niche similarity compared to the observed overlap value using a two-tailed test (Supplemental Figures S4-7). In order to reject ecological interchangeability with the niche similarity test, both of the background tests for a pair of lineages should have a D value that is more different (niche divergence) than expected by chance. Since some of our comparisons involved parapatric/sympatric lineages which results could potentially be affected by the defined background region, we conducted sensitivity tests by defining alternative background regions based on buffer polygons from occurrence points of differing distances (25 km, 50 km, 75 km, and 100 km) in ArcMap v10.7.