Variability and uncertainty associated to methods for estimating diet composition: The case of Champsocephalus gunnari in the South Orkney Islands

Diet studies are essential to understand predator-prey dynamics within a marine ecosystem and how diet changes with growth and habitat. Although a large number of diet estimation methods are currently used in fish ecology, few studies have assessed results sensitivity and variability within and between methods. Such sensitivity and variability might be higher for fishes inhabiting the Southern Ocean given strong seasonality, limited sampling opportunities and methodological assumptions derived from temperate waters. Being interested in characterizing the diet of Champsocephalus gunnari in the South Orkney Islands, we compared results from stomach content analysis (SCA) and stable isotope analysis (SIA), while assessing the sensitivity of SIA estimates to (i) prey lipid correction, (ii) trophic enrichment factors (TEFs), (iii) putative prey, and (iv) diet composition priors. Thus, 284 stomach contents and 600 muscle isotopic compositions (δ13C/δ15N) were used to produce SCA and SIA estimates using Bayesian methods. SIA sensitivity to TEF values was assessed by comparing McCutchan et al. (2003)'s global averages and Canseco et al. (2022)'s temperature- and baseline-adjusted values. SIA sensitivity to putative prey was tested by comparing 4- and 6-prey models based on our own and previously published SCA results, respectively. Tested informative priors corresponded to SCA estimates. SCA showed krill Euphausia superba contributed >98% of C. gunnari diet. SIA estimates tended to produce smaller and more uncertain krill contribution estimates, which resulted highly sensitive to Δ15N assumptions. Thus, when Δ15N decreased from 5.2‰ (adjusted value) to 2.9‰ (global average) the estimated krill contribution decreased from 94% to 8%, departing largely from SCA estimates. As expected, informative priors enhanced consistency between SCA and SIA results. Overall, contribution means and credible intervals were highly sensitive to all tested methodological choices and, particularly, to TEF assumptions. While contrasting SCA and SIA seems a sensible way to identify methodological issues, our results highlight that adjusting TEF by temperature and prey isotopic baselines yield more accurate and precise results for cold-water fishes.