A comparison of diet estimates of captive beluga whales using fatty acid mixing models with their true diets

Abstract Beluga whales (Delphinapterus leucas) are sentinel species for Arctic environmental change; however, whether belugas are resilient to anticipated changes in prey dynamics remains uncertain, partly due to a lack of dietary information. Mixing models are promising tools to evaluate the diets of marine mammals, using dietary tracers such as fatty acid signatures. Additionally, captive feeding trials are useful to test the effectiveness of these tools for application to wild populations. In our study, we compared the ability of two models, quantitative fatty acid signature analysis in R (QFASAR) and Fatty Acid Source Tracking Algorithm in R (FASTAR) to reconstruct the diets of captive beluga whales that had been fed a long-term diet consisting mostly of Pacific herring (Clupea harengus pallasi), and in smaller proportions, capelin (Mallotus villosus) and squid (Doryteuthis opalescens). To account for fatty acid metabolism, we compared a suite of calibration coefficients derived from captive feeding trials of American mink (Mustela vison), harp seal (Pagophilus groenlandicus), harbour seal (Phoca vitulina richardsi), and grey seal (Halichoerus grypus) fed monotypic diets. For both methods, diets estimated using calibration coefficients derived from feeding trials of mink fed herring and/or seal oil were able to correctly identify herring as the dominant prey source of belugas. In addition, there was higher precision among diet estimates using these calibration coefficients in comparison to those generated using calibration coefficients derived from feeding trials of pinnipeds or mink fed poultry. QFASAR more accurately approximated the diet proportions of the whales than FASTAR, possibly because of the greater robustness to errors in the calibration coefficients. Based on QFASAR, diet estimates best reflect the whales' true dietary proportions consumed approximately two weeks to one month prior to sampling, similar to timeframes seen in previous studies of harbour seals. We recommend caution when using these models with prey that have high within-species variability in fatty acid signatures or small sample sizes, as well as prey with similar fatty acid signatures. The application of mixing models is promising to identify the prey of wild beluga whales and other marine mammals.