Cross-validation of diet determination methods for seabird conservation

This SRA dataset is part of DOI:Seabirds are recognised as one of the most vulnerable groups of birds with around a third of species identified as globally threatened. The conservation of seabirds is often linked with their feeding and diet, due to undesirable interactions with human-related fishing activities and fisheries depletion/climate change impacting food resources. Therefore, understanding the diet of seabirds is often a critical first step towards identifying effective conservation measures. DNA metabarcoding and hard parts analyses provides a foundation for assessing the diet of a predatory seabird species, giving insight to predator-prey relationships and ecosystem-wide food webs. Congruency between these two methods would increase confidence, providing validation that either method provides a reliable representation of the diet. This study on the diet of the endangered New Zealand king shag (Leucocarbo carunculatus) compared the frequency of occurrence of fishes detected from the same regurgitated pellets (n = 191) using both hard parts and DNA metabarcoding methodologies. The number of pellets with overlapping fish families between methods showed a significant positive correlation between methods (r = 0.96; p < 0.001), with 50 out of 191 pellets showing complete alignment and only two pellets without any alignment. Both methods confirmed the predominance of Bothidae (DNA: 71% of pellets, hard parts: 77%, total: 80%) and Rhombosoleidae (DNA: 45%, hard parts: 51%, total: 59%) taxa in the diet of king shag, while also revealing the consumption of a diversity of other fish species. Overall, this study demonstrates that the two methods provide a complementary approach for revealing the dominant fish prey species in the diet, as well providing an overview of the diversity of prey species (DNA: 14 unique species, hard parts: 8, total: 28). However, the inexact alignment between the two methods for detecting every taxon in every pellet suggests differences in detection, especially for less common taxa and for DNA metabarcoding where species level resolution is dependent on adequate DNA database sequence entries of taxa in the local area. In addition, filtering thresholds for DNA metabarcoding further influenced alignment. Overall, the results indicate that both methods provide consistent detection of major prey items, however, reliably capturing the overall full diversity of prey species with either method is reliant on a sufficient sample size.