Correlation analyses of isolation patterns
Cost distance was calculated
while the route was plotted (Figure 3, Supporting information). A
penalty of f = 1 predicted that A. hueti and P.
pectoralis could take “shortcuts” through the center of the mountain
range (Figure 3, Supporting information). A better fit between cost
distance and genetic distance for the species with IBD patterns was
expected compared to spheric distance. IBE and IBD patterns with
(Scenario 2) and without cost-distances were tested to determine
patterns of isolation among populations within species. Mantel tests and
linear correlations based on the 6 bioclimatic variables and elevational
differences showed no significant relationship between genetic distance
and environment difference (Supporting information). Mantel testing and
linear regression of geographic distances and N-distances indicate
different correlation patterns among the species (Table 3 and Figure 4,
Supporting information), given penalties for dispersal routes outside
each species’ preferred elevational range. Only one of the five species,A. hueti, is significantly isolated by distance. The linear
correlation between genetic distance and cost distance is significant
under penaltized scenarios for P. pectoralis . While the Mantel
test of this species is quasi significant
(0.05<p<0.1), with low linear correlation
coefficient and low Mantel’s statistic r (Figure 4B, Supporting
information). For these 2 species, the extent of obstruction caused by
the montane landscape is variable. In A. hueti , a moderate
dispersal penalty of f = 0.1 leads to the highest correlation
according to significance and AIC values (Supporting information). InP. pectoralis, the least cost route with no penalty outside
preferred elevational range correlates with genetic distance, as
supported by AIC values. The Mantel test results in the other 3 species,I. mcclellandii, L. lutea and S. torqueola , indicated no
significant correlation between geographic and genetic distance.
The location and length of potential barriers to dispersal among the
sampled populations inferred by BARRIER are consistent with Mantel
tests. Overall, none of the inferred barriers concur with the
latitudinal layout of the mountain range. The main barriers withinA. hueti (Figure 5B), which was significantly isolated by
distance, separate populations with high diversity, suggesting the
existence of substantial dispersal obstacles during historical expansion
in high-diversity distribution centers. For example, the A. huetipopulations NL and JXS occur in the center of the Nanling Mountains and
are separated by three barriers (Figure 5B). In the other species, the
layout and direction of the barriers were either longitudinal or around
peripheral populations (Figure 5A, C, D and E). For A. hueti,when no penalty was applied, the least cost route distance was
correlated with genetic distance and corresponded to direct distance
(Supporting information). When the correlation was highest withf =0.1, the route taken was constrained to more mountainous areas,
compared to scenarios with no penalties (Supporting information).
To help explain genetic differences among populations, we performed ENM
for each species using LGM, BAW and current environmental variables
(Supporting information). Given likely environmental conditions for
those time periods, A. hueti and P. pectoralis should have
maintained a relatively stable distribution in the Nanling mountains and
the mountainous region east of Nanling during all three time periods.
The rest of the species should have experienced a major shift in
distribution from the LGM and BAW to the current environmental
situation. At all times, parts of the Nanling Mountains provided
appropriate habitat for the five species and were disjunct or patchily
distributed.