Bias associated with diet samples in Audouin's gulls

We analyzed five different types of food samples from Audouin's Gull (Larus audouinii), collected during the breeding seasons of 1994 and 1995 at its two main breeding colonies, the Ebro Delta and the Chafarinas Islands. These food samples included spontaneous regurgitates, dry boli containing partially digested food, food remains, pellets, and prey identified during direct observations of chick provisioning. We compared estimates of biomass, levels of taxonomic determination allowed by each kind of food sample, and the associated potential biases to assess which sampling method provides the best estimate of diet in gulls. Regurgitates allowed identification of most prey to species level and reliable biomass estimates, but their collection was time-consuming and invasive. Dry boli provided almost the same information as regurgitates at order level and were easy to collect. However, both underestimated soft-bodied prey and prey with large, hard parts. Food remains provided an estimate of diet composition that was highly biased towards prey with large distinctive hard parts. However, food remains were a good complement to dry boli, enhancing biomass estimates for food items that had a good relationship of weight and linear measurements of prey-hard parts; Direct observation allowed identification of prey only to upper taxonomic categories, and is useful when only a broad categorization of prey types is required. Pellets showed important biases towards fish with robust otoliths, and inaccurate conversion to biomass, but they can be useful for monitoring variations in the consumption of certain prey items. Several factors such as time spent collecting, sample availability, disturbance to animals, and the status of the species studied need to be considered when deciding on a method of diet assessment sampling.