Demographic inference
Eight different models were compared to identify the most likely scenario under which UCSD juncos diverged from its closest relative, thepinosus form, using fastSIMCOAL2. AIC revealed that the most likely scenario was the one consistent with a recent population split due to a founder event (Fig. 2A) under a model of isolation with migration, and clearly rejected alternative models (Table 3). Parameter estimates revealed a very recent colonization time of the urban environment (TDIV = 44 years, 95% CI = [30 - 47]) by a very small number of effective founders, from two to three individuals (Ne = 2 individuals; 95% CI = [2 - 3]). The estimate for the current effective population size of UCSD was Ne = 252 individuals, 95% CI = [215 - 274]). The best estimate for UCSD population growth rate since colonization was estimated at r = 0.159, 95% CI = [0.150 – 0.244]). The estimate for pinosus effective population size was high, at Ne = 5.86x106 individuals, 95% CI = [5.86x106 - 5.87x106]. The analysis also recovered a narrow, limited rate of gene migration frompinosus to UCSD (MIG pinUCSD = 6.02x10-2, 95% CI = [5.87x10-2 – 8.03x10-2]). The migration rate range from UCSD to pinosus varied by several orders of magnitude and was close to zero, revealing little impact of the parameter in the SFS (MIG UCSDpin = 2.01x10-6, 95% CI = [9.38x10-10 - 2.57x10-5]) (Table 3).
Despite the marked difference in AIC scores supporting a recent population split due to a founder event, it should be noted that none of the next best-fitting models included a recent population split, and that the model including a recent population split due to a founder event in strict isolation was not among those at the top (Table 3). This may be due to the effect of gene flow during the establishment and genetic differentiation of the UCSD population, resulting in highly different site frequency spectra when comparing scenarios of recent colonization with and without migration. However, it could also reflect a lack of resolution of the analysis in discriminating the historical factors shaping the current patterns of genetic variability in the UCSD population of juncos, so that these results should be interpreted with caution.
Genotype association analysis In the RDA, four explanatory environmental variables accounted for 7.65% (R2 = 0.0765) of the variability in the SNP dataset, and the model was highly significant (P = 0.001). Both mean temperature of the warmest quarter and temperature seasonality showed large contributions to the RDA2 axis, while variability in NDVI and elevation was more evenly captured by both RDA1 and RDA2 axes (Fig. 6). RDA association scores revealed a strong positive correlation of mean temperature of the warmest quarter with genetic differences of UCSD residents, and negative correlations with the remaining explanatory variables. Correlation patterns were less conspicuous in other junco forms, mostly differentiated along RDA1. Forms townsendi andpontilis showed negative correlations with vegetation greenness, with the former also loading heavily on elevation. The formpinosus presented general associations of the same sign as UCSD, yet much less pronounced. The remaining, more boreal forms presented moderate and positive correlations with vegetation greenness and temperature seasonality, while loading negatively on mean temperature of the warmest quarter, an inverse pattern with respect to UCSD juncos (Fig. 6).