Agent-based modeling for environmental management. Case study: virus dynamics affecting Norwegian fish farming in fjords

Abstract

Background: Norwegian fish-farming industry is an important industry, rapidly growing, and facing significant challenges such as the spread of pathogens1, trade-off between locations, fish production and health. There is a need for research, i.e. the development of theories (models), methods, techniques and tools for analysis, prediction and management, i.e. strategy development, policy design and decision making, to facilitate a sustainable industry. Loss due to the disease outbreaks in the aquaculture systems pose a large risk to a sustainable fish industry system, and pose a risk to the coastal and fjord ecosystem systems as a whole. Norwegian marine aquaculture systems are located in open areas (i.e. fjords) where they overlap and interact with other systems (e.g. transport, wild life, tourist, etc.). For instance, shedding viruses from aquaculture sites affect the wild fish in the whole fjord system. Fish disease spread and pathogen transmission in such complex systems, is process that it is difficult to predict, analyze, and control. There are several time-variant factors such as fish density, environmental conditions and other biological factors that affect the spread process. In this thesis, we developed methods to examine these factors on fish disease spread in fish populations and on pathogen spread in the time-space domain. Then we develop methods to control and manage the aquaculture system by finding optimal system settings in order to have a minimum infection risk and a high production capacity. Aim: The overall objective of the thesis is to develop agent-based models, methods and tools to facilitate the management of aquaculture production in Norwegian fjords by predicting the pathogen dynamics, distribution, and transmission in marine aquaculture systems. Specifically, the objectives are to assess agent-based modeling as an approach to understanding fish disease spread processes, to develop agent-based models that help us predict, analyze and understand disease dynamics in the context of various scenarios, and to develop a framework to optimize the location and the load of the aquaculture systems so as to minimize the infection risk in a growing fish industry. Methods: We use agent-based method to build models to simulate disease dynamics in fish populations and to simulate pathogen transmission between several aquaculture sites in a Norwegian fjord. Also, we use particle swarm optimization algorithm to identify agent-based models’ parameters so as to optimize the dynamics of the system model. In this context, we present a framework for using a particle swarm optimization algorithm to identify the parameter values of the agent-based model of aquaculture system that are expected to yield the optimal fish densities and farm locations that avoid the risk of spreading disease. The use of particle swarm optimization algorithm helps in identifying optimal agent-based models’ input parameters depending on the feedback from the agentbased models’ outputs. Results: As the thesis is built on three main studies, the results of the thesis work can be divided into three components. In the first study, we developed many agent-based models to simulate fish disease spread in stand-alone fish populations. We test the models in different scenarios by varying the agents (i.e. fish and pathogens) parameters, environment parameters (i.e. seawater temperature and currents), and interactions (interaction between agents-agents, and agents-environment) parameters. We use sensitivity analysis method to test different key input parameters such as fish density, fish swimming behavior, seawater temperature, and sea currents to show their effects on the disease spread process. Exploring the sensitivity of fish disease dynamics to these key parameters helps in combatting fish disease spread. In the second study, we build infection risk maps in a space-time domain, by developing agent-based models to identify the pathogen transmission patterns. The agent-based method helps us advance our understanding of pathogen transmission and builds risk maps to help us reduce the spread of infectious fish diseases. By using this method, we may study the spatial and dynamic aspects of the spread of infections and address the stochastic nature of the infection process. In the third study, we developed a framework for the optimization of the aquaculture systems. The framework uses particle swarm optimization algorithm to optimize agent-based models’ parameters so as to optimize the objective function. The framework was tested by developing a model to find optimal fish densities and farm locations in marine aquaculture system in a Norwegian fjord. Results show so that the rapid convergence of the presented particle swarm optimization algorithm to the optimal solution, - the algorithm requires a maximum of 18 iterations to find the best solution which can increase the fish density to three times while keeping the risk of infection at an accepted level. Conclusion: There are many contributions of this research work. First, we assessed the agent-based modeling as a method to simulate and analyze fish disease spread dynamics as a foundation for managing aquaculture systems. Results from this study demonstrate how effective the use of agentbased method is in the simulation of infectious diseases. By using this method, we are able to study spatial aspects of the spread of fish diseases and address the stochastic nature of infections process. Agent-based models are flexible, and they can include many external factors that affect fish disease dynamics such as interactions with wild fish and ship traffic. Agent-based models successfully help us to overcome the problem associated with lack of data in fish disease transmission and contribute to our understanding of different cause-effects relationships in the dynamics of fish diseases. Secondly, we developed methods to build infection risk maps in a space-time domain conditioned upon the identification of the pathogen transmission patterns in such a space-time domain, so as to help prevent and, if needed, combat infectious fish diseases by informing the management of the fish industry in Norway. Finally, we developed a method by which we may optimize the fish densities and farm locations of aquaculture systems so as to ensure a sustainable fish industry with a minimum risk of infection and a high production capacity. This PhD study offers new research-based approaches, models and tools for analysis, predictions and management that can be used to facilitate a sustainable development of the marine aquaculture industry with a maximal economic outcome and a minimal environmental impact.