Effective sample size and assignment
Breeding populations ranged in number of samples from 27 (Maritime Provinces and Basin Rockies) to 47 (Southern Temperate) and ranged in effective sample size from 12.3 – 24.6. Mean accuracy of leave-one-out assignment with these individuals was 89.3% (151 out of 169 individuals), and accuracy by breeding population ranged from 63.0% (Maritime Provinces) to 100% (Southern Temperate and Basin Rockies). All 18 individuals that were inaccurately assigned were assigned to a breeding population with higher effective sample size than the known breeding population (Figure 2). All posterior probabilities of assignment (accurate and inaccurate) were greater than 0.8 (Table S3).
The SSBPs all had 27 samples but ranged in effective sample size from 12.3 – 16.1. The mean accuracy of assignment was 97.0% (164 out of 169 individuals) for the leave-one-out assignment with the 135 individuals in the SSBPs and standard assignment for the remaining 34 individuals. Three individuals were incorrectly assigned in the leave-one-out assignment test and two individuals were incorrectly assigned in the standard assignment test. All five incorrectly assigned individuals were assigned to a breeding population with higher effective sample size than the known breeding population (Figure 2).
The ESSBPs ranged in number of samples from 21 (Basin Rockies) to 27 (Maritime Provinces) but had minimal variation in their effective sample sizes (range: 12.0 – 12.5). Mean assignment accuracy was 99.4% (168 out of 169 individuals) for the leave-one-out-assignment with the 122 individuals in the ESSBPs and standard assignment for the remaining 47 individuals. Only one individual was incorrectly assigned from the Northern Temperate population to the Southern Temperate population, and this did not correspond to a breeding population with higher effective sample size. Interestingly, this same individual (sampled in Minnesota from the Northern Temperate population) also stands out in the PCA results as clustering more closely with individuals from the Southern Temperate population (Figure S1). Given the higher accuracy of assignment with the ESSBPs, compared to the other sets of breeding individuals, we continued subsequent assignment testing using only the ESSBPs data set as the source populations.