Classification of samples to species and hybrid categories
The probability of assignment (Z ) of each sample as a pure Bruce
spanworm, a pure winter moth, or a hybrid (either F1, F2, Bruce
spanworm-backcross, or winter moth-backcross) was estimated using the
Bayesian-assignment program NewHybrids v 1.1 b3 (Anderson,
2008; Anderson & Thompson, 2002). We used uniform priors, random
starting seeds, burn-in periods of 100,000 generations, and a
post-burn-in runtime of 1,000,000 generations. A separate dataset was
run for each transect and year combination to reduce assignment errors,
given that individuals from one year could be the offspring of
individuals from the previous year. Datasets were then filtered so that
only individuals with ≥ 10 successfully scored loci were included. Four
independent runs were performed for each dataset, and the assignment
scores were then averaged across runs. We interpreted samples withZ ≥ 0.75 to any one category as obtaining “strong support”, and
samples with 0.5 ≤ Z < 0.75 as obtaining “moderate
support”. If a sample was not assigned to any category with Z ≥
0.5, it was classified as having “weak support” to the category with
the highest Z score.
To determine whether there was temporal or spatial variation in
hybridization rates across each transect, the mean proportions of
hybrids were calculated by dividing the number of genotyped individuals
from each trap classified to one of the four hybrid categories by
NewHybrids (as described above). Differences between years and
between traps were compared using an analysis of variance (ANOVA) as
implemented in R v 4.0.2 (R Core Team, 2020).