A Bayesian inference approach to account for multiple sources of uncertainty in a macroalgae based integrated multi-trophic aquaculture model. (April 2016)

• Record Type: Journal Article
• Authors: Hadley, Scott; Jones, Emlyn; Johnson, Craig; Wild-Allen, Karen; Macleod, Catriona
• Is Part Of: Environmental modelling & software. Volume 78(2016:Apr.)
• Page Start: 120
• Page End: 133
• Publication Date: 2016-04
• Subjects: Bayesian inference -- Macroalgae -- IMTA -- Sequential Monte Carlo -- Particle filter -- Aquaculture; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2015.12.020
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: A Bayesian inference method was employed to quantify uncertainty in an Integrated Multi-Trophic Aquaculture (IMTA) model. A deterministic model was reformulated as a Bayesian Hierarchical Model (BHM) with uncertainty in the parameters accounted for using "prior" distributions and unresolved time varying processes modelled using auto-regressive processes. Observations of kelp grown in 3 seeding densities around salmon pens were assimilated using a Sequential Monte Carlo method implemented within the LibBi package. This resulted in a considerable reduction in the variability in model output for both the observed and unobserved state variables. A reduction in variance between the prior and posterior was observed for a subset of model parameters which varied with seeding density. Kullback–Liebler (KL) divergence method showed the reduction in variability of the state and parameters was approximately 90%. A low to medium seeding density results in the most efficient removal of excess nutrients in this simple system. Highlights: An IMTA model was recast in a Bayesian Hierarchical Modelling (BHM) framework. Particle MCMC methods were employed to estimate the uncertainty in the posterior distributions for the state variables and parameters. Our results suggest that low to medium density seeding (3–6 sporeling's per metre) of macroalgae gave greatest remediation. Several novel techniques were introduced to assist in visualising and interpreting high dimensional parameter distributions.