Genetic analyses
Mme7 is a sex-linked marker and it deviated from Hardy Weinberg
Equilibrium (HWE), therefore, we coded it as a missing allele for
females and juveniles in further analyses (Rasner et al., 2004;
Rodríguez-Bardía et al., 2022). We genotyped 76 individuals, 28 males
from ‘period 1’, 25 males from ‘period 2’, 14 females from ‘period 1’,
and 9 females from ‘period 2’. We calculated expected heterozygosity
(\(H_{e}\)), observed heterozygosity (\(H_{o}\)), allelic richness
(\(A_{r}\)), and inbreeding coefficients (\(F_{\text{IS}}\)) for each
population and time period (e.g., MTV period 1 and MTV period 2). We
estimated confidence intervals for FIS values
using 9999 bootstraps as implemented in the “hierfstat” package in R
(R Core Team, 2020). To compare differences in heterozygosity for the
same populations between time periods we used z-scores from a Wilcoxon
signed rank test as implemented in the “coin” package in R (R Core
Team, 2020). To estimate changes in genetic structure among populations
over a decade, we calculated Nei’s coefficient (\(G_{\text{st}}\)) among
populations within time periods, using 9999 bootstrap values to estimate
confidence intervals. We assumed that nonoverlapping confidence
intervals indicated significant differences in genetic structure between
periods. We also tested for differences in genetic diversity between
time periods using an analysis of molecular variance (AMOVA) with 9999
permutations using the “poppr” package in R (R Core Team, 2020). We
clustered individuals in each time period into the most likely
configuration of clusters based on allele frequency similarity using the
Bayesian clustering algorithms implemented in the program STRUCTURE
V.2.3.4 (Pritchard et al., 2000). To determine the most probable number
K, we utilized a method implemented in a study that used data from some
of the same populations (Rodríguez-Bardía et al., 2022). Thus, we used
correlated allele frequencies and the admixture model with 300 000
Markov chains, and a burning of 30 000 chains. We grouped individuals in
1 to 6 clusters with 20 repetitions for each cluster (Rodríguez-Bardía
et al., 2022). To determine the most likely number of K clusters we used
Structure Harvester 0.6.94 (Evanno et al., 2005). After determining the
most likely number of K cluster, we conducted another structure analysis
with 1 000 000 Markov chains and a burn-in of 100 000 chains to assure
proper chain mixture (Rodríguez-Bardía et al., 2022).