Abstract

Ecological systems can be treated as following mechanistic rules with inherent and unavoidable uncertainty. Thus, suitably chosen mechanistic models combined with robust statistical methods to quantify and propagate this uncertainty are useful for forecasting the present and future of ecological phenomena.

One ecological application of interest is coral reef health, which can be quantified by net calcification rates. Ocean acidification threatens growth of coral reefs by reducing their net calcification rates. Calcification occurs inside the coral animal, in a compartment that is difficult to measure. Using laboratory data for Great Barrier Reef coral species, we calibrated mechanistic and statistical models of calcification rate responses to seawater conditions, to identify how much information about coral animal biology can be gleaned from this data and make predictions of net calcification rates (including uncertainty) for thousands of reefs within the Great Barrier Reef. The mechanistic model consisted of a set of coupled ordinary differential equations combined with a standard algorithm used to calculate equilibrium kinetics of carbon chemistry (CO2SYS).

For both mechanistic and statistical models, calibration to data was undertaken using Bayesian inference, implemented via a robust Sequential Monte Carlo sampling algorithm. Both mechanistic and statistical models plausibly fit the laboratory data and made similar reef-scale predictions. Reef-scale predictions indicated that cumulative acidification and light stress mediate localised reductions of coral reef calcification rate across the Great Barrier Reef, although forecasted uncertainty in predictions was unsurprisingly high.