Introduction
The influence of genetic relatedness on social interactions of
conspecifics has been an ongoing question in conservation of wild
populations for a long time (Wilson, 1975). On the one hand, pedigree
reconstructions have provided insights into mating patterns, individual
fitness, and social and genetic structure of populations (Harcourt,
Kingston, Cameron, Waas, & Hindell, 2006; Lucena-Perez et al., 2018;
vonHoldt et al., 2008). On the other hand, analyses of social
connectivity and stability of individuals have produced valuable
information on the viability of wild populations, especially at current
selective environments due to human-induced environmental changes
(Snijders, Blumstein, Stanley, & Franks, 2017). Yet social network
analyses and genetically-derived pedigree reconstruction have been used
as two separate methodological frameworks to assist conservation of wild
populations. The combination of these methods may highlight the
interconnectedness between individuals, differences in reproductive
success, and ultimately inform on the demographic structure of a
population.
Reconstructing a reasonably complete and accurate familial network from
pedigree data is especially relevant for endangered species, providing
information on mating patterns and reproductive success (Lucena-Perez et
al., 2018; Manlik et al., 2016). However, collecting reliable parentage
information for cryptic and elusive species is difficult or directly
unfeasible; pedigree information obtained through direct field
observations are often limited to females, and may consistently overlook
cryptic mating (Coltman et al., 1999; Gottelli, Wang, Bashir, & Durant,
2007). Molecular markers, such as microsatellites, have been used to
infer parentage and familial relationships in wild populations
(Pemberton, 2008) and assess individual heterogeneity in survival and
reproduction (Bolnick et al., 2011; Hamel, Gaillard, Festa-Bianchet, &
Côté, 2009; Kendall, Fox, Fujiwara, & Nogeire, 2011). Such
heterogeneity can be the result of a number of common processes, such as
persistent social rank (e.g. Holst et al., 2002; Stockley &
Bro-Jørgensen, 2011), unequal allocation during parental care
(e.g. Manser & Avey, 2000; Johnstone, 2004), fine-scale spatial habitat
heterogeneity (Bollinger & Gavin, 2004; Franklin, Anderson, Gutiérrez,
& Burnham, 2000; Manolis, Andersen, & Cuthbert, 2002), and genetics
(Meyers & Bull, 2002; Nussey, 2005).
Social networks have been used to investigate complex webs of
interconnections between individuals, providing an array of measurements
of individual sociality and the extent to which an individual is
connected to other individuals (Borgatti, Mehra, Brass, & Labianca,
2009; Wasserman & Faust, 1994). Several network-based measures are
commonly used in social network analysis to quantify indirect
connections between individuals in a network (Table 1). Although some
network-based centrality measures may overlap, each measure captures a
distinct aspect of the social network of a population; individuals with
high scores for one measure may not necessarily have a high score in
other measures (Brent, 2015; Sueur, Jacobs, Amblard, Petit, & King,
2011). Highly directly connected individuals may not necessarily be
highly indirectly connected; individuals with the same degree (same
number of social ties) may have different betweenness, depending if that
individual’s partners are from the same subgroup (low betweenness) or
from different subgroups (high betweenness). In sperm whale
(Physeter macrocephalus ) association networks, centrality varied
between and within individuals, with one sperm whale having the highest
scores for strength and eigenvector centrality, but the lowest
clustering coefficient (Lusseau, Whitehead, & Gero, 2008). In a captive
chimpanzee (Pan troglodytes ) population, the individuals with the
largest degree (number of grooming partners) were not the individuals
with the greatest betweenness or clustering coefficient (Kanngiesser,
Sueur, Riedl, Grossmann, & Call, 2011). The extent to which an
individual is directly and indirectly connected to the network has
considerable quantifiable differences. Strongly directly connected
individuals with weak indirect connections may have great influence over
their immediate social partners, but have minimal influence over the
rest of the population, whereas individuals that are weakly directly
connected but with strong indirect connections may be the single tie
that links otherwise unconnected network sections together, exerting
considerable influence over the population (Brent, 2015).
Here, we infer population demographic structure by assessing different
node-based metrics of centrality obtained from a familial pedigree
network. First, we use microsatellite data to identify parent-offspring
relationships and construct a spatial familial network from all
relationships (familial pedigree) of boreal caribou in Saskatchewan,
Canada. Then we create a spatial familial network to identify local area
networks with varying distributions of centrality metrics, determining
whether high centrality metrics and edge-to-node ratios at the fine
scale correspond to high centrality in the full network. We also assess
the community structure and cohesiveness within the full network using
edge removal to identify boundaries that run between subgroups, with a
particular focus on parts of the range presenting different levels of
anthropogenic disturbance. Our findings allow us to discuss how
different metrics of network centrality can be used to spatially
identify areas of highest fitness levels and social cohesion across the
landscape in support of population monitoring and recovery efforts.