BayesReef

Overview

BayesReef is a Bayesian inference framework for pyReef-Core, a stratigraphic forward model that simulates coral reef growth and sediment deposition. It quantifies uncertainty in reef predictions by estimating parameters as full probability distributions using Markov Chain Monte Carlo methods.

BayesReef is capable of inference from two perspectives: the depth structure (which coralgal assemblage occurs at a given depth interval) and the time structure (which sediment is deposited at each time interval over the simulation). Studies have shown accurate parameter estimation and mean model prediction for both two- and four-dimensional problems.

MCMC Methods

BayesReef uses the Metropolis-Hastings algorithm, which allows for asymmetrical proposals in probabilistic space. The random walk generates dependent sample chains designed to concentrate sampling near regions of best fit, approximating the distribution of parameter values.

The approach is grounded in Bayes' Rule — combining prior beliefs about parameters with the likelihood that chosen values explain observed data to determine the posterior distribution.

Parameterisation

The framework handles up to 27 free parameters including:

• Population dynamics — Malthusian parameter, assemblage interaction matrix values
• Hydrodynamic flow thresholds — exposure thresholds per assemblage
• Sediment input thresholds — exposure thresholds per assemblage

Constrained variants allow low-dimensional experiments with reduced parameter sets. Supporting scripts generate marginal and bivariate likelihood visualisations for both population dynamics and environmental threshold parameters.

Reference

Pall, J., Chandra, R., Azam, D., Salles, T., Webster, J.M. and Cripps, S. — BayesReef: A Bayesian inference framework for modelling reef growth in response to environmental change and biological dynamics.