Methodology for analysing long-tern variation in harbour seal diet and relationships with fish abundance

No abstracts are to be cited without prior reference to the author.

Long time-series of dietary data can potentially provide information on how predator diets respond to changing prey abundance (e.g. as a result of overfishing and/or climate change). Here we analyse data on the summer diet of harbour seals (Phoca vitulina) in Orkney (northeast Scotland), based on faecal sampling at haulout sites on the island of Eynhallow during 1986–2006. The main component of the diet in summer is sandeels (Ammodytidae, mainly Ammodytes spp.). Aside from a patchy distribution of samples and involvement of various different researchers over the 20-year series, the data present several statistical challenges. Thus, individual otoliths in a faecal sample may not represent independent samples and the statistical distributions of prey numbers include an excess of zero values. Linear mixed effects modelling was applied to model temporal patterns in the size of fish eaten (using otolith length as a proxy of fish size). In addition, zero-inflated negative binomial generalized linear mixed models were applied to model trends in the numerical importance of individual prey species and to relate them to trends in prey abundance. We discuss the statistical robustness and biological significance of the trends identified in this dataset.