FIGURE LEGENDS
FIGURE 1 Map of southeastern China, showing the sampling locations of C. chuniana populations used in this study. The mountain ranges involved in the study are shown in orange shadows. Populations sampled in the Nanling Mts. are indicated with black circles; populations in the east are indicated with triangles. Vicariance events detected in S-DIVA are shown with blue lines. The key at bottom right indicates elevational ranges. Additional accession information is given in Table 1.
FIGURE 2 Potentially suitable areas for C. chunianapredicted by ecological niche modeling (ENM) and corresponding variation in temperature. Four different periods were applied, i.e., the Last Interglacial (LIG), the Last Glacial Maximum (LGM), the Middle Holocene (MH), and current. Suitable and unsuitable habitats are indicated in red and grey, respectively, where red represents the habitat suitability (occurrence probability) higher than 44.93%. The temperature variations in each period are indicated in the gray boxes. (a) LIG; arrows point to relevant mountain ranges. The temperature during secondary contact (TSEC ) is indicated. (b) Potentially suitable areas projected in comparison with a layer of GIS-based vegetation map at LGM. Numbers 1‒8 represent different vegetation types at the LGM: 1, tropical thorn scrub and scrub woodland; 2, open boreal woodland; 3, semi-arid temperate woodland or scrub; 4, steppe-tundra; 5, polar and alpine desert; 6, temperate desert; 7, forest steppe; and 8, dry steppe. (c) The simulated distribution range at MH. (d) The current potential distribution range, with black dots representing localities based on herbarium specimen data and samples collected for the present study. The most influential factors are listed in Table S1.
FIGURE 3 Phylogenetic tree of C. chuniana populations based on maximum-likelihood (ML) method. Bootstrap percentages (> 50) in the ML tree are indicated above the branches.Cercis chingii was used as the outgroup.
FIGURE 4 Chronogram of the Bayesian tree for divergence time estimates, population structural clustering and ancestral area reconstruction. Branch lengths were transformed via Markov chain Monte Carlo (MCMC) simulations in the Bayesian time estimation. The light blue bars indicate 95% confidence intervals. Individuals assigned to different clusters in FastStructure are shown in corresponding colours with K = 2, 3 and 7, with 7 as the optimal value. In the S-DIVA analysis, colour legends indicate different geographical regions and ancestral areas. Vicariance events (V1–V6) with high probabilities (P ≥ 0.70), are shown for nodes. Populations in the Nanling Mts. are distinguished within the dotted line frame. Time scale bar is shown at the bottom. Various glacial and interglacial periods are indicated by the braces, the former of which is below the bar while the latter above the bar.
FIGURE 5 Schematic representation of the best fit demographic models investigated in our study. Model names and the parameters correspond to those in Tables S2 and S3, respectively. The left vertical line shows time (Ma). Time of divergence (TDIV ) and secondary contact (TSEC ) are indicated as dashed lines, with the corresponding glacial or interglacial periods shown on the right. The top grey box represents the ancestral populations with effective population size. The current population sizes are shown at the bottom after the split between the Nanling Mts. (NL) and the eastern mountains (ES). The migration directions and corresponding rates (MES ) in individual migrants per generation are shown between NL and ES.