Approximate Bayesian computation
Tournament comparisons of invasion scenarios indicated that each of the
four invasive populations in north America were more closely related to
populations from Europe than to each other, or to an unsampled
extra-European “ghost” population, suggesting four separate
introduction events from Europe for winter moth (Figure 4). Comparison
of the Nova Scotia population to the Western European, Central European,
and Eastern European genetic clusters, indicated that it was most likely
introduced from Central Europe (Supplemental Appendix S1 Figure S10a),
and this relationship received high support based on logistic regression
analysis (P = 0.98). The Oregon population most likely originated from
Western Europe (Supplemental Appendix S1 Figure S10b), and this
relationship received high support based on logistic regression analysis
(P = 0.96). The British Columbia population most likely represents an
additional independent introduction from Central Europe (Supplemental
Appendix S1 Figure S10c), and this relationship received high support
based on the logistic regression analysis (P = 1.00). The population in
the northeastern United States population was also reconstructed as a
third introduction from Central Europe (Figure 4), and this relationship
was also highly supported (P = 1.00). Based on examinations of the
posterior estimates, each of the invasive populations experienced a
strong genetic bottleneck when they diverged from European populations
~100 ya (Supplemental Appendix S1 Table S3).
Visualization of the PCAs for the supported scenario from each
tournament based on the Perform Model Checking analysis in
DiyABC, indicated that posterior values from simulated datasets
formed a distinct cloud encompassing the sample dataset (Supplemental
Appendix S1 Figures S11 through S14).