Dietary complexity and hidden costs of prey switching in a generalist top predator

Variation in predator diet is a critical aspect of food web stability, health, and population dynamics of predator/ prey communities. Quantifying diet, particularly among cryptic species, is extremely challenging, however, and differentiation between demographic subsets of populations is often overlooked.We used prey remains and data taken postmortem from otter Lutra lutra to determine the extent to which dietary variation in a top predator was associated with biotic, spatial, and temporal factors.Biotic data (e.g., sex, weight, and length) and stomach contents were taken from 610 otters found dead across England and Wales between 1994 and 2010. Prey remains were identified to species where possible, using published keys and reference materials. Multi-model inference followed by model prediction was applied to test for and visualize the nature of associations.Evidence for widespread decline in the consumption of eels (Anguilla anguilla) reflected known eel population declines. An association between eel consumption and otter body condition suggested negative consequences for otter nutrition. Consumption of Cottus gobio and stickleback spp. increased, but was unlikely to compensate (there was no association with body condition). More otters with empty stomachs were found over time. Otter sex, body length, and age-class were important biotic predictors of the prey species found, and season, region, and distance from the coast were important abiotic predictors.Our study is unique in its multivariate nature, broad spatial scale, and long-term dataset. Inclusion of biotic data allowed us to reveal important differences in costs and benefits of different prey types, and differences between demographic subsets of the population, overlaid on spatial and temporal variation. Such complexities in otter diet are likely to be paralleled in other predators, and detailed characterization of diet should not be overlooked in efforts to conserve wild populations.