2.5.1 Population genomic structure
We used “poppr” v2.9.3 (Kamvar 2023) to produce a multidimensional
scaling (MDS) plot of all individuals based on a matrix of Prevosti’s
genetic distances. To focus in more detail on the King Island
population, we then repeated the MDS analysis using only the 15
individuals sampled from King Island. We then created a pairwise Weir
and Cockerham (1984) fixation index (FST) matrix in
“StAMPP” v1.6.3 (Pembleton 2013) with 999 bootstraps to assess
significance of the estimated differentiation between populations. To
assess isolation by distance at the individual level, we fitted a linear
model of all standardised pairwise FST values against
the geographic distance between samples (Rousset, 1997).
We used the alternating projected least squares algorithm implemented in
“tess3r” to assign individuals to ancestral population
genomic clusters, investigate patterns of admixture between populations
and assess hierarchical population structure. This method applies a
model of genetic structure featuring a number of ancestral populations
(k ), allowing assessment of values for k that have low
cross-entropy metrics (Frichot et al., 2014; Frichot & François, 2015).
It also incorporates the spatial location of sampling, to remove
potential bias associated with isolation-by-distance. We calculated
cross-entropy criteria for values of k between 1 and 15, and
visualised a cross-entropy scree-plot to identify a plateau or
substantial change in curvature in the plot. We then extracted the
matrices of individual admixture coefficients for the most relevant
values of k for inference and plotted these as stacked bar plots
to visualize hierarchical population structure. We then interpolated
several values of k (2–4) across the landscape within the range
of the scrubtit, based on the geographical location of samples. We then
repeated this analysis using only the King Island scrubtit data for
values of k from 1 to 5.