SERVICE ET AL. WILEY 5 of 9 TABLE 1 Observed (O) versus expected (E) mc1r genotypes of black bears (Ursus americanus) by landmass in coastal British Columbia, Canada (2012-2017). Monte Carlo Exact Chi-square p-values reflect results from tests for departures from genotype frequencies expected under Hardy-Weinberg equilibrium. U-score p-values reflect tests for an excess or deficiency of heterozygosity at mc1r Landmass GG (white) AG (black) AA (black) о E ° E O E Chi² p U-score p Island Hawkesbury 0 0.00 0 0.00 24 24.00 Gribbell 1 1.13 7 6.75 10 10.12 1.00 0.74 Princess Royal 8 5.32 25 30.36 46 43.32 0.14 0.09 Roderick 0 0.01 1 0.98 21 21.01 $ 1.00 1.00 Pooley 0 0.00 0 0.00 6 6.00 - Yeo 0 0.00 0 0.00 5 5.00 Mainland West of Hawkesbury 0 0.03 East of Princess Royal 0 0.15 North of Roderick 0 0.02 Kynoch 0 0.00 Don Peninsula 0 0.01 24101 1.94 34 34.03 1.00 0.99 3.70 23 23.15 1.00 0.89 0.96 11 11.02 1.00 1.00 0.00 15 15.00 0.98 21 21.01 1.00 - 1.00 alleles. We fit a Matern variogram model (Cressie, 1990) using maxi- mum likelihood to account for spatial structure in G allele frequency values. The selected variogram model was used to fit a cost-based weighted kriging model using the krig.conv function in the geoRcb pack- age (López-Quílez & Muñoz, 2009) to create the interpolated G allele frequency surface. We used the predicted surface to identify hotspots of G allele frequencies and assess their spatial alignment with pro- tected areas (Section 2.5.3). 2.5.3 Assessing alignment of protected areas with G allele hotspots We assessed the alignment between our kriged G allele frequency surface and protected areas through two complementary approaches. First, we identified hotspots of allele frequency, defined as pixels with values in the 90th percentile (i.e. top 10% of the entire interpolated raster surface from Section 2.5.2) and assessed the percentage of the hotspot that corresponded with protected areas. To assess sensitivity of this arbitrary, but logical, cutoff value, we also report values for the 95th (top 5%) and 85th percentile (top 15%) hotspots. Secondly, we cal- culated the percentage of protected area in each landmass and tested if this value was predicted by G allele frequency through a linear regres- sion model and a Pearson's correlation. 3 RESULTS to 0.26 (Princess Royal Island), were mostly lower than previously reported estimates, and were significantly lower on Roderick (Odds ratio=0.091; p = 0.018) and Gribbell (Odds ratio = .261; p = 0.007) Islands (Table 2). Also, in contrast with previous research, we did not detect a heterozygote deficiency. Rather, we failed to detect a statis- tically significant departure from Hardy-Weinberg equilibrium in any landmass (Table 1; Table 2). We found that multiple regions of high G allele frequency occurred outside of protected areas. Specifically, approximately 50% of the 90th percentile hotspots corresponded with protected areas (Figure 2; Table 2). Similar results emerged for the 95th and 85th percentiles, at approximately 50% and approximately 45% protected area cover- age, respectively. Across landmasses, protected area coverage was not related spatially to the G allele frequency (R = -0.012; F = 0.001; p = 0.971). The two landmasses with the highest G allele frequency differed strongly in protected area coverage, with Princess Royal Island (G frequency = 0.26) having high coverage (52% of area) and Gribbell Island (G frequency = 0.25) having very low protection (0.05%; Table 2). 4 DISCUSSION Our results suggest that landscape-level frequency of the G allele is lower than previously estimated, and that populations previously reported to demonstrate a heterozygote deficiency are in fact in Hardy-Weinberg equilibrium. Additionally, despite the role of Spirit