Investigating variability in the diet, distribution and demographics of Barents Sea cod, 1991-2017: A combined statistical modelling approach

Rising global ocean temperatures are changing marine ecosystems globally. Over the past decade, temperatures in the Barents Sea have been at a historic high due to the increasing temperature of Atlantic water and a decrease in sea ice volume. Studies investigating the effects of temperature variability on the Barents Sea ecosystem have highlighted the complexity and difficulty of predicting how future change may affect the species and communities here. Environmental variability has ramifications for every aspect of the ecosystem, but one particularly complex relationship is between the environment and trophic dynamics. As environmental changes may affect species within a community in different ways and to different extents, predicting how the whole system may be affected can be challenging. This thesis seeks to quantify some aspects of the relationship between environmental variability and trophic dynamics in the Barents Sea by applying a novel statistical modelling approach to a spatio-temporal diet database for Barents Sea cod, a top predator. We examine spatio-temporal patterns in diet, distribution and demographics for cod and several key prey species. We utilise a species distribution model, Maxent, and a nonlinear system identification model, NARMAX, to synthesise existing data and identify the factors driving spatio-temporal change in these aspects of the ecosystem. We found a trend of distribution spreading into the northern Barents Sea following increasing water temperatures. We found further evidence for the borealization of the arctic environment in the Barents Sea, with Boreogadus saida, a landmark arctic species, increasing in significance as a prey item with consumption rising as Mallotus villosus and Pandalus borealis consumption falls. We found that higher order biotic effects were more informative in our models than direct measurements of climate, and some of the measures we attempted to predict could not be adequately explained by either the biotic or abiotic data that were used.