Statistical models of marine multispecies ecosystems

• Main Author: Bjarki Þór Elvarsson 1981-
• Other Authors: Háskóli Íslands
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: 2015
• Subjects: Doktorsritgerðir; Iceland; North Atlantic
• Online Access: http://hdl.handle.net/1946/21560

The analyses presented in this thesis can broadly be split into two categories; 1) methodology for assessing goodness of fit, estimation of uncertainty and model selection, and 2) an application in the form of investigation into the link between minke whales and cod on the continental shelf around Iceland. These analyses are primarily reported in the five papers. The first paper presents the results from the first comprehensive analysis of the feeding ecology of minke whales on the Icelandic continental shelf area. The findings suggest that sandeels appear to be a major component in the diet of minke whales and gadoids are a greater part of the diet than previously considered. The results are presented with uncertainty estimated using a bootstrap approach treating an individual whale as a sampling unit. Paper V then details a potential use of the data obtained from the scientific survey by modelling the interaction between minke whales and cod. Paper II details the development of a novel bootstrap approach applied to highly disparate datasets is developed and demonstrated on cod in Icelandic waters. The method is contrasted with a more commonly used normal approximation where the covariance matrix is derived from the inverse Hessian matrix of the negative log likelihood function at the optimum. This method appears to perform favorably to the Hessian–based method for this class of models. Paper III presents the development of R-gadget, an R-package containing a set of useful utitilities when developing models in Gadget. Paper IV presents a power analysis of a proposed genetic tagging experiment contrasting two stock structure hypotheses based on information on genetically determined close relatives. The results indicate that in the case of small populations, such as fin whales in the North Atlantic fin whales a tagging experiment can be improved considerably by augmenting it with information on genetically determined close relatives. Hafrannsóknarstofnun