Building Bayesian Belief Networks to investigate how fishery performance responds to management interventions

Abstract

This paper describes how Bayesian Belief Networks (BBNs) were used to investigate how a management intervention affects multiple aspects of a fishery’s performance. The ideas were developed in the context of a specific case study in which fishery performance was measured using the Marine Stewardship Council (MSC) certification scores, and the management intervention was whether the fishery is under catch share management (a form of rights based management) or not. A fishery’s performance is scored against more than 30 indicators to obtain MSC certification. These indicators are grouped into three Principles that measure different aspects of sustainability. The effect of catch share management must be investigated in the light of other fisheries characteristics such as gear type and target species, which can also affect MSC scores. Statistical models can measure the effect of these characteristics on the scores for each individual indicator, but are not able to assess their effect across all of the Principles together at the same time. A BBN summarised and synthesized the results from each indicator’s statistical model. It was possible using the BBN to (i) compare the probability of scoring highly on all three Principles, or subsets of indicators, for fisheries with different characteristics and catch share management strategies, (ii) identify whether a fishery that scores highly on all three Principles is more likely to be managed using catch shares and (iii) identify the characteristics and indicators that are most associated with obtaining high scores across all three Principles. The BBN was able to address a wide range of questions and provide a mechanism for integrating a suite of statistical models describing a complex dataset with multiple response variables of interest.