Recent Bayesian stable-isotope mixing models are highly sensitive to variation in discrimination factors

Stable isotopes are now used widely in ecological studies, including diet reconstruction, where quantitative inferences about diet composition are derived from the use of mixing models. Recent Bayesian models (MixSIR, SIAR) allow users to incorporate variability in discrimination factors (Δ ¹³C or Δ¹⁵N), or the amount of change in either δ¹³C or δ¹⁵N between prey and consumer, but to date there has been no systematic assessment of the effect of variation in Δ¹³C or Δ¹⁵N on model outputs. We used whole blood from Common Terns (Sterna hirundo) and muscle from their common prey items (fish and euphausiids) to build a series of mixing models in SIAR (stable isotope analysis in R) using various discrimination factors from the published literature for marine birds. The estimated proportion of each diet component was affected significantly by Δ¹³C or Δ¹⁵N. We also use recently published stable-isotope data on the reliance of critically endangered Balearic Shearwaters (Puffinus mauretanicus) on fisheries discards to show that discrimination factor choice can have profound implications for conservation and management actions. It is therefore crucial for researchers wishing to use mixing models to have an accurate estimate of Δ¹³C and Δ¹⁵N, because quantitative diet estimates can help to direct future research or prioritize conservation and management actions.