Quantifying parameter uncertainty in a coral reef model using Metropolis-Coupled Markov Chain Monte Carlo

摘要: Abstract Coral reefs are threatened ecosystems, so it is important to have predictive models of their dynamics. Most current coral fall into two categories. The first simple heuristic which provide an abstract understanding the possible behaviour in general, but do not describe real reefs. second complex simulations whose parameters obtained from a range sources such as literature estimates. We cannot estimate these single data set, and we little idea uncertainty predictions. developed compromise between extremes, enough reef data, that can for specific time series. In previous work, fitted this model long-term set Heron Island, Australia, using maximum likelihood methods. To evaluate predictions model, need estimates our parameters. Here, obtain Bayesian Metropolis-Coupled Markov Chain Monte Carlo. versions corals aggregated state variable (the three-state model), separated four variables six-state order determine appropriate level aggregation. also posterior distribution predicted trajectories each case. both cases, were close observed had doubts about biological plausibility some parameter suggest informative prior distributions incorporating expert knowledge may resolve problem. frequencies after 40 years contained divergent community types, one dominated by free space soft corals, acroporid, pocilloporid, massive corals. predicts only type. conclude hides too much heterogeneity, more if reliable model. It likely there will be similarly large, currently unevaluated, other models, many harder fit data.