1. Using measurements of naturally occurring stable isotopes in animal tissues is useful for monitoring diets of wide-ranging species that would otherwise be logistically difficult to evaluate. However, differential metabolic routing of macromolecules within a consumer can be problematic when using stable isotope analysis of bulk tissues to trace dietary input.
2. We used stable isotope (δ13C, δ15N) analysis to examine polar bear Ursus maritimus diet, which includes both lipid-rich blubber and the proteinaceous tissues of their marine mammal prey. Because the proportion of proteins and lipids consumed may depend on prey type and size, it was necessary to consider metabolic routing of these macromolecules separately in isotope mixing models.
3. Bayesian mixing models (MixSIR, version 1.04) were used to separately estimate protein (δ13C, δ15N) and lipid (δ13C) dietary inputs. We used existing knowledge of the relative lipid and protein intake for polar bears and isotopic information from both macromolecules to estimate overall diet composition.
4. The results for both male and female polar bears indicated that smaller prey (e.g. ringed seal Pusa hispida) contributed the largest proportion to the protein-metabolic pathway. In contrast, the largest proportion of the lipid-metabolic pathway for both sexes tended to consist of larger prey (e.g. bearded seal Erignathus barbatus).
5. The diet composition of male polar bears consisted of more large than small prey. Diet estimates for females overlapped to some degree with males but tended to consist of less large prey.
6. Synthesis and applications. Monitoring polar bear diet may help determine the effects of climate-induced environmental changes in Arctic marine ecosystems including shifts in prey composition. Additionally, tracing origins of anthropogenic pollutants is currently a priority for wildlife managers concerned with the health of marine mammals. However, our results indicate using stable isotopes to infer dietary inputs when proportions of macromolecules fluctuate amongst food sources requires the sampling and analysis of multiple tissues representing distinct macromolecular metabolic pathways. In such cases, utilizing only proteinaceous tissues for analysis will result in erroneous dietary source estimates and inaccuracies when examining trophic-level transfer of contaminants, especially those that are lipophylic.