Brown bears Ursus arctos consume a wide range of organisms, including ungulates and plants, but Pacific salmon Oncorhynchus spp. are especially important to their diet where their ranges overlap. Although some brown bears minimize antagonistic encounters with other brown bears or infanticide by avoiding streams where salmon spawn, studies generally assume that brown bears with ready access to salmon feed heavily on them. To test this assumption, and the hypothesis that male brown bears would feed more heavily on salmon than females (owing to their sexual size dimorphism), we collected hair samples from brown bears by using barbed wire placed on six small tributaries of Lake Aleknagik, Alaska, USA, where adult Sockeye Salmon Oncorhynchus nerka are readily accessible and frequently consumed by brown bears. Analysis of DNA distinguished among the different brown bears leaving the hair samples, some of which were sampled multiple times within and among years. We assessed the contribution of salmon to the diet of individual brown bears by using carbon and nitrogen stable isotope signatures. The 77 samples analyzed from 31 different bears over 4 y showed isotopic ratios consistent with reliance on salmon, but the wide range of isotopic signatures included values suggesting variable, and in one case considerable, use of terrestrial resources. Stable isotope signatures did not differ between male and female brown bears, nor did they differ between two sides of the lake, despite marked differences in Sockeye Salmon density. We collected the hair samples when salmon were present, so there was some uncertainty regarding whether they reflected feeding during the current or previous season. Notwithstanding this caveat, the results are consistent with the hypothesis that salmon were sufficiently available to provide food for the brown bears and that the considerable isotopic variation among brown bears with access to salmon reflected their age, status, and behavior.

Hair samples (n = 2026)  were collected during the months of July and August of 2012 through 2015, from six streams that flow into Lake Aleknagik (southcentral Alaska as part of the Wood River sytem): Happy, Hansen, and Eagle creeks on the northeastern side, and Yako, Whitefish, and Bear creeks on the southwestern side (Quinn et al. 2014; Wirsing et al. 2018). We designate these two trios of streams as separate foraging neighborhoods because bears often foraged on two or three of the streams on one side of the lake but almost never crossed the lake within or between years (Wirsing et al. 2018, and additional unpublished data). Annual estimates of sockeye salmon abundance in each stream, based on multiple counts each year (Quinn et al. 2017), indicated much higher densities on the northeastern side compared to the southwestern side (Table 1). To collect the hair samples, two unbaited barbed wires (average: 8 m long) with a barb every 12 cm were strung across each stream, attached to trees on either side using fencing staples. The wires, about 50 – 55 cm above the streambed at mid-channel in different reaches of each stream, were carefully checked for samples every second day to minimize sample degradation that results from prolonged exposure (Dumond et al. 2015). Hair samples were removed from the wire with tweezers, placed in dry coin envelopes in the field, and then stored with desiccant in the field. At the end of the season, samples were sent to the Laboratory for Evolutionary, Ecological, and Conservation genetics at the University of Idaho for DNA analysis to verify the species (i.e., brown rather than black bear), identify the individual, and determine the sex of each. 

Genetic material was extracted from hair samples using the DNeasy Blood and Tissue Kit (Qiagen, Inc.). Bear genotypes were generated for each sample using 10 nuclear DNA microsatellite loci (Paetkau and Strobeck 1994, Paetkau et al. 1998, DeBarba et al. 2010) and one sex identification marker (Ennis and Gallagher 1994). The observed and expected heterozygosities for these 10 loci are 0.71 and 0.69. Each sample was amplified two to four times to ensure accuracy. Consensus genotypes were derived following the rule that each allele must be observed twice at each locus and must contain data at eight or more loci to be included in the matching analysis. Two genotypes were considered a match in the program Genalex (Peakall and Smouse 2006) if they were identical or included a one allele mismatch at two or fewer loci that could be due to allelic dropout. Under this protocol, the probability of a match for unrelated individuals across 10 loci was 0.0000000011, and the probability of a match between siblings across 10 loci was 0.00025 (Waits et al. 2001). The probabilities of a match for unrelated individuals and siblings at eight loci ranged from 0.000000011 to 0.00000021 and 0.00072 to 0.0017, respectively. For all single captures the reliability of their genotype was estimated using the program Reliotype (Miller et al. 2002) and retained in the dataset if the reliability score was ≥ 90%. For the years used in this analysis, the amplification success rate was 63%.