Results
A total of 2,198 samples were collected (2,099 fecal and 99 blood blot).
1,970 were successfully scored (average success rate of 91.4%), and 933
unique individuals were identified (Table 2). Overall, the average
dropout rate was 0.0028% and the average false allele rate was 0.011%.
Pedigree reconstruction inferred 310 females and 319 males, for a total
familial network of 1,562 individuals. 355 inferred and sampled males
and 360 inferred and sampled females were identified as parents. 1,487
(95.2%) individuals were linked in one network, with the remaining 75
individuals linked in five smaller clusters (Figure S2.1). We used the
1,487 individuals identified in the primary network for calculating
node-based metrics of centrality.
Spatial network analysis
Local area networks
We identified 18 local area networks in order to determine the
cohesiveness and centrality of individuals. The local areas with the
lowest edge-to-node ratios were all located in the northern Boreal
Shield, with the high edge-to-node ratio areas found further south in
the western Boreal Plains and southern Boreal Shield (Figure 1). We
found significant differences between the distribution of centrality
metrics between high and low edge-to-node ratio local areas (Table S2.2,
Figure 2). The largest edge-to-node ratio was Canoe Lake in the western
Boreal Plains (ratio of 15; Table S2.1, Figure S2.3). We identified
three other local areas with similarly high edge-to-node ratios (Figure
S2.4, Figure S2.5, Figure S2.6, Table S2.1). The smallest edge-to-node
ratio (Central SK Shield) had zero parent-offspring relationships (Table
S2.1; Figure S2.7). We identified two other local areas with similarly
low edge-to-node ratios, with very few parent-offspring relationships
occurring within these local areas (Figure S2.8-Figure S2.9, Table
S2.1), indicating that Boreal Shield individuals are not presenting the
same proximity to related individuals as observed in the Boreal Plains.
Overall, edge-to-node ratios correlated positively to closeness (Figure
S2.2A), alpha (Figure S2.2C), betweenness (Figure S2.2D), and degree
centrality (Figure S2.2E). However, edge-to-node ratios decreased with
eccentricity centrality (Figure S2.2B).
When bringing in the first neighbours of all individuals within a local
area, the high edge-to-node ratio areas formed a tighter cluster of
individuals than in the low edge-to-node ratio areas. Including first
neighbours in the area with the highest edge-to-node ratio (Canoe Lake)
increased the ratio to 1.14 and connected 73.6% of individuals into one
cluster (Figure S2.3). A large proportion of each high edge-to-node
ratio local area became connected into one or two large clusters with
the inclusion of first neighbours (Figure S2.4, Figure S2.5, Figure
S2.6). In comparison, including first neighbours in the lowest
edge-to-node ratio local area (Central SK Shield) increased the ratio to
0.86, but did not connect many individuals into one cluster (only 12.8%
of individuals; Figure S2.7). The large variation in local area
edge-to-node ratios show that individuals within the high edge-to-node
ratio areas in the Boreal Plains and southern Boreal Shield are forming
tighter family groups (high number of parent-offspring relationships
within each local area).
Full network
Individuals from high edge-to-node ratio local areas were located more
centrally within the full family network and clustered with other
individuals from the same local area. Individuals from low edge-to-node
ratio local areas were dispersed throughout the network and primarily
found on the outer edges of the network (Figure 3). Although all local
areas were of similar geographic size (Figure 1), individuals from low
edge-to-node ratio local areas were not closely connected to each other
in the network. Individuals from these local areas were not found within
a few edges of other individuals from the same local area, indicating
that individuals encountered in each low edge-to-node ratio local area
are from different familial lines, or are dispersers that were sampled
in that local area (Figure 3); as the edges in the familial network
represent parent-offspring relationships, these individuals are likely
not highly related to one another, and do not form a cohesive social
group. In contrast, individuals from high edge-to-node local areas were
highly connected to one another within the full network, indicating they
are closely related, with a high density of familial ties
(parent-offspring relationships).
Removal of edges with high betweenness did not alter the overall network
structure (Figure S2.10). Most edges within the network had low
betweenness centrality (score of 1 – 81.5% of edges; Table 3). Only
2.97% of edges were removed after sequentially removing edges with the
highest edge betweenness score until only edges with an edge betweenness
> 4 remained (Table 3), with no clear delineation of groups
within the main network (Figure S2.10), but the removal of these edges
led to individuals becoming disconnected from the main network. The high
edge-to-node ratio local areas of Canoe Lake, Peter Pond Lake, and
SK2West remained central and clustered within the edge removal network
(Figure S2.10). Individuals from Trade Lake maintained a high level of
clustering, but became separated from the main network, forming a
separate subgroup (Figure S2.10). Removal of high betweenness edges did
not result on individuals from low edge-to-node ratio areas becoming
separate subgroups; individuals remained dispersed throughout the
network (Figure S2.10).