Modelling and prediction of the evolution of fish farm networks under aquatic disease spread.

Abstract

The movement of live fish, from ova to market sized stock, between aquaculture sites and recreational fisheries plays a vital role in food security and leisure industries. These transfers form a network linking entities such as farms, hatcheries, and suppliers. When consignments carrying pathogens enter this densely connected system, disease can spread rapidly, disrupting trade flows and creating significant logistical problems. If suppliers cannot receive or send stock of the required health status, or when official movement restrictions are imposed, they seek alternative sources of supply. This alters established trade routes, relationships, and flow patterns across the network. This study developed a dynamic network simulation framework that represents, simulates, and predicts the structure and evolution of live fish trade. It began with a systematic review of the applications of complex network analysis (CNA) in aquaculture and capture fisheries, classifying existing research into a thematic taxonomy covering disease control, ecological assessment, genetic analysis, and production flows. The review revealed three key gaps: (i) the lack of dynamic models that capture changes in the network, (ii) limited use of predictive algorithms accompanied by rigorous evaluation methods, and (iii) a lack of rewiring frameworks to model disruptions. To address these gaps, the research developed the Edge-Weighted Katz Index (EWKI), a spatial link prediction method that embeds proximity and network topology to accurately predict future trade links. It then conducted a rigorous evaluation of the EWKI and other link prediction methods under conditions of class imbalance and sparse data, resulting in the development of a targeted edge removal technique to improve predictive performance and ensure reliable results. Building on these predictions, a rewiring decision-support framework was developed to simulate network restructuring during supply disruption while preserving spatial constraints and operational feasibility. Although the research was applied to aquaculture, the mathematical methodologies and simulation framework have value in other spatially distributed, dynamically rewired systems such as healthcare logistics, food supply chains, transport networks, and energy grids. By integrating spatially informed link prediction with adaptive network rewiring, this thesis provides a novel approach for improving resilience, operational efficiency, and evidence-based decision-making under uncertainty.