3.5 Inference of demographic history
Six vicariance events (V1‒V6) among the geographical regions were inferred from the S-DIVA analysis (Figure 1). V1 is between WYS1 and the rest of the populations. V2 is between WYS2 and LXS2, located in the western Wuyi Mts. and southern Luoxiao Mts., respectively. V3 is between WYS2/LXS2 and the rest of the populations, including LXS1 and the populations in the Nanling Mts. V4 is between LXS1 and the rest of the populations. V5 is between the eastern and western Nanling Mts., separating [NLE1 + NLE2] and [NLW1 through NLW5], whereas V6 is between NLW1 and NLW2 through NLW5. Across the six vicariance events, the eastern populations diverged from the rest of the species first, and the western populations later.
The best fit model for the demographic analysis with FSC2 is model SECEXP, indicating an isolation model followed by secondary contact (SEC) and demographic expansion (EXP; Table 3, Figure 5, S2, Table S2‒S4). The time scale of 548,000 generations (2.74 Ma) was confirmed based on the lowest AIC value. Based on the mutation rate, we converted the genome-wide estimates of nucleotide diversity into effective population sizes. The current effective population sizes of the Nanling Mts. (NL) and eastern regions (ES) are NeNL = 57,495 and NeES = 14,955, respectively. From the current effective population size, the ancestral effective population size was calculated as NeANC = 755,955 (Table 3, Figure 5, Table S3). Using the ancestral effective population size, we converted the divergence time between NL and ES into the number of generation times, TDIV = 319,472 generations ago, i.e., about 1.6 Ma. Secondary contact (SEC) was estimated at ca.T SEC = 0.10 Ma. This time is within the Lushan-Tali Interglacial period in China (Duan, Pu, & Wu, 1980; Zhu, Liu, & Jackson, 2004), when temperature increased and was ca. 5℃ higher than at present (Figure 2). The ancestral effective population size of NL was estimated to be much smaller (Ne-pre-exp = 866) than at present. In contrast, the ES population sizes remained more or less constant (Table 3, Figure 5, Table S3). The migration rateMNL-ES (2.14) was much higher thanMES-NL (0.33), with migration occurring after NL and ES divergence.