Bayesian stable isotope mixing model methods are evolving rapidly, in large part due to their ability to account for error in stable isotope data. While the basic Bayesian mixing model formulation accounts for error in isotopic signatures of sources and mixtures (e.g. prey and predators), little work has been done to account for error resulting from variability in the underlying biological mixing process. For instance, variability in predator isotope signatures may result from different prey preferences across predators, differences in group-level diets among predators, or error resulting from predators eating prey items not included in the mixing model.  We demonstrate the analytic approaches to quantifying these different error structures in the Bayesian mixing model framework. We subsequently apply model selection methods in order to evaluate data support for these different error structures using data from grey wolves (Canis lupus) that occupy a spatially heterogeneous landscape in coastal British Columbia comprising a mainland area and an adjacent archipelago.

Results/Conclusions

Our findings suggest that variability in individual diets of wolves was substantial among subpopulations, social groups, and individual grey wolves. These model selection results support the theory that this grey wolf population has extensive terrestrial-marine isotopic niche variation across levels of population structure.  Moreover, they demonstrate the flexibility of Bayesian mixing models in partitioning variability in stable isotope signatures due to variability in the associated mixing process.