Comparison of SNP and microsatellite genotyping panels for spatial assignment of individuals to natal range: A case study using the American black bear (Ursus americanus)

Analysis of multilocus genotypes is fundamental to conservation genetics, allowing inference ranging from population delineation to identification of dispersers; additionally, the natal range of a sample of unknown origin may be assigned. We investigated the accuracy and precision of two methods (spatial smoothing and principal components regression) for natal assignment with five datasets that varied in marker type (microsatellites or SNPs), number of loci, and number of training samples. Accuracy varied between datasets where the median difference between true and estimated geographic locations ranged from 192 to 902 km. A dataset using 1000 SNP loci and the spatial smoothing method was both the most accurate and precise. We observed that natal inference from SNPs was more accurate and precise than when estimated using microsatellites, and that large numbers of SNP loci could overcome having few samples in the training dataset. Our results suggest cautious interpretation of natal assignments, as 52% or fewer test samples were assigned to the correct management jurisdiction. In addition, samples from continuous habitat had less accurate assignments than samples from isolated areas whether due to landscape barriers or anthropogenic fragmentation. The use of natal inference as a tool for management agencies may work well at the regional level, given sufficient input data; however, we clearly observed limits on the spatial scale of inference and consequently on the effectiveness of genotypes as a sole source for natal inference.