Testing glacial refugia and post glacial recolonization
The TMRCA estimation of 96 pairs of populations were used to conduct ANCOVA for models of K=3 to K=6. In all tested models, the interaction terms are not significant and thus the type II sum of squares was used in ANCOVA tests. The K=6 drainage refugia model includes 30 pairs of populations from the same drainage system and 66 pairs of populations from different drainage systems. This model has the lowest AIC and BIC, and explains the data significantly better than either K=3 or K=4 population structure models based on Vuong’s test (Table 1 ). However, we cannot distinguish the K=5 and K=6 population structure model, as they possess the same set of paired populations due to the lack of multiple population samples form the Merced drainage.
In the best fit K=6 model, the result of ANCOVA shows that TMRCA estimates are significantly older (> 20,000 years ago) for pairs of populations from different drainages than from the same drainage (~ 10,000 years ago; p-value < 0.001), regardless of whether geographical distance is included as a covariant or not (Figure 7A ). However, there are a few of population pairs from the same drainage that do have older TMRCA, notably sites in the Kings or Kern drainages. Similarly, there are a few outliers with recent TMRCA when pairs of different drainages are compared, including sites from the San Joaquin and Kings drainages. In contrast, there is no significant difference between the two types of population pairs when measuring Weir and Cockerham’sF ST (p-value > 0.05). Geographical distance has a high R2 value (0.35 and 0.55) when included as a covariate, implying genetic variation is consistent with isolation by distance (Figure 7 B).