Meta-analytical approaches to the study of fish population dynamics

The aim of this thesis is to reduce uncertainties in our understanding of fish population dynamics by: (a) developing statistical methods that allow information to be shared or combined; and (b) applying these methods to real fisheries problems. Meta-analysis is a statistical framework for making inferences by combining results across studies or experiments. In this thesis I demonstrated that meta-analysis is particularly useful for studies of fisheries population dynamics because: (1) there is often considerable observation error in fisheries data that creates difficulties in making useful inferences; and (2) there are potentially relevant data for many fish populations or years. I showed, using data for many populations, that catch-per-unit-effort abundance indices generally underestimate the declines in the population being fished—a finding critical to fisheries stock assessment. I developed a generalized framework to test hypotheses relating to the regulation of fish abundance during the larval and juvenile stages—thus allowing for future meta-analyses of this problem. I extended the traditional meta-analytical methods to combine estimates of the relationship between the size of a fish and its catchability to trawl surveys—this allowed for predictions to be made for species where no data were available. Using a method from the field of econometrics, I developed an approach to integrate random effects into a complex nonlinear model—this allowed for improved modelling and interpretation of million dollar acoustic surveys of a New Zealand fish population. I demonstrated how inferences from meta-analysis can be used in the assessment of stocks with little information (e.g., developing fisheries) and to obtain reliable estimates required for management decisions. To conclude, I provided a discussion of how the further extension of methods and fisheries databases would be critical to the future of meta-analysis, and highlighted important directions for future research.